The Safer Affordable Fuel-Efficient (SAFE) Vehicles Rule for Model Years 2021-2026 Passenger Cars and Light Trucks

 
CONTENT
Federal Register, Volume 85 Issue 84 (Thursday, April 30, 2020)
[Federal Register Volume 85, Number 84 (Thursday, April 30, 2020)]
[Rules and Regulations]
[Pages 24174-25278]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 2020-06967]
[[Page 24173]]
Vol. 85
Thursday,
No. 84
April 30, 2020
Part IV
Book 2 of 3 Books
Pages 24173-25278
Environmental Protection Agency
-----------------------------------------------------------------------
40 CFR Parts 86 and 600
Department of Transportation
-----------------------------------------------------------------------
National Highway Traffic Safety Administration
49 CFR Parts 523, 531, 533, et al.
-----------------------------------------------------------------------
The Safer Affordable Fuel-Efficient (SAFE) Vehicles Rule for Model
Years 2021-2026 Passenger Cars and Light Trucks; Final Rule
Federal Register / Vol. 85 , No. 84 / Thursday, April 30, 2020 /
Rules and Regulations
[[Page 24174]]
-----------------------------------------------------------------------
ENVIRONMENTAL PROTECTION AGENCY
40 CFR Parts 86 and 600
DEPARTMENT OF TRANSPORTATION
National Highway Traffic Safety Administration
49 CFR Parts 523, 531, 533, 536, and 537
[NHTSA-2018-0067; EPA-HQ-OAR-2018-0283; FRL 10000-45-OAR]
RIN 2127-AL76; RIN 2060-AU09
The Safer Affordable Fuel-Efficient (SAFE) Vehicles Rule for
Model Years 2021-2026 Passenger Cars and Light Trucks
AGENCY: Environmental Protection Agency and National Highway Traffic
Safety Administration.
ACTION: Final rule.
-----------------------------------------------------------------------
SUMMARY: EPA and NHTSA, on behalf of the Department of Transportation,
are issuing final rules to amend and establish carbon dioxide and fuel
economy standards. Specifically, EPA is amending carbon dioxide
standards for model years 2021 and later, and NHTSA is amending fuel
economy standards for model year 2021 and setting new fuel economy
standards for model years 2022-2026. The standards set by this action
apply to passenger cars and light trucks, and will continue our
nation's progress toward energy independence and carbon dioxide
reduction, while recognizing the realities of the marketplace and
consumers' interest in purchasing vehicles that meet all of their
diverse needs. These final rules represent the second part of the
Administration's action related to the August 24, 2018 proposed Safer
Affordable Fuel-Efficient (SAFE) Vehicles Rule. These final rules
follow the agencies' actions, taken September 19, 2019, to ensure One
National Program for automobile fuel economy and carbon dioxide
emissions standards, by finalizing regulatory text related to
preemption under the Energy Policy and Conservation Act and withdrawing
a waiver previously provided to California under the Clean Air Act.
DATES: This final rule is effective on June 29, 2020.
 Judicial Review: NHTSA and EPA undertake this joint action under
their respective authorities pursuant to the Energy Policy and
Conservation Act and the Clean Air Act. Pursuant to CAA section 307(b),
42 U.S.C. 7607(b), any petitions for judicial review of this action
must be filed in the United States Court of Appeals for the D.C.
Circuit. Given the inherent relationship between the agencies' action,
any challenges to NHTSA's regulation under 49 U.S.C. 32909 should also
be filed in the United States Court of Appeals for the D.C. Circuit.
ADDRESSES: EPA and NHTSA have established dockets for this action under
Docket ID Nos. EPA-HQ-OAR-2018-0283 and NHTSA-2018-0067, respectively.
All documents in the docket are listed in the http://www.regulations.gov index. Although listed in the index, some
information is not publicly available, e.g., confidential business
information (CBI) or other information whose disclosure is restricted
by statute. Certain other material, such as copyrighted material, will
be publicly available in hard copy in EPA's docket, and electronically
in NHTSA's online docket. Publicly available docket materials can be
found either electronically in www.regulations.gov by searching for the
dockets using the Docket ID numbers above, or in hard copy at the
following locations:
 EPA: EPA Docket Center, EPA/DC, EPA West, Room 3334, 1301
Constitution Ave. NW, Washington, DC. The Public Reading Room is open
from 8:30 a.m. to 4:30 p.m., Monday through Friday, excluding legal
holidays. The telephone number for the Public Reading Room is (202)
566-1744.
 NHTSA: Docket Management Facility, M-30, U.S. Department of
Transportation (DOT), West Building, Ground Floor, Rm. W12-140, 1200
New Jersey Ave. SE, Washington, DC 20590. The DOT Docket Management
Facility is open between 9 a.m. and 5 p.m. Eastern Time, Monday through
Friday, except Federal holidays.
FOR FURTHER INFORMATION CONTACT: EPA: Christopher Lieske, Office of
Transportation and Air Quality, Assessment and Standards Division,
Environmental Protection Agency, 2000 Traverwood Drive, Ann Arbor, MI
48105; telephone number: (734) 214-4584; fax number: (734) 214-4816;
email address: [email protected], or contact the Assessment
and Standards Division, email address: [email protected]. NHTSA: James Tamm,
Office of Rulemaking, Fuel Economy Division, National Highway Traffic
Safety Administration, 1200 New Jersey Avenue SE, Washington, DC 20590;
telephone number: (202) 493-0515.
SUPPLEMENTARY INFORMATION:
Does this action apply to me?
 This action affects companies that manufacture or sell new light-
duty vehicles, light-duty trucks, and medium-duty passenger vehicles,
as defined under EPA's CAA regulations,\1\ and passenger automobiles
(passenger cars) and non-passenger automobiles (light trucks) as
defined under NHTSA's CAFE regulations.\2\ Regulated categories and
entities include:
---------------------------------------------------------------------------
 \1\ ``Light-duty vehicle,'' ``light-duty truck,'' and ``medium-
duty passenger vehicle'' are defined in 40 CFR 86.1803-01. Generally
speaking, a ``light-duty vehicle'' is a passenger car, a ``light-
duty truck'' is a pick-up truck, sport-utility vehicle, or minivan
up to 8,500 lbs. gross vehicle weight rating, and a ``medium-duty
passenger vehicle'' is a sport-utility vehicle or passenger van from
8,500 to 10,000 lbs. gross vehicle weight rating.
 \2\ ``Passenger car'' and ``light truck'' are defined in 49 CFR
part 523.
---------------------------------------------------------------------------
[[Page 24175]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.000
 This list is not intended to be exhaustive, but rather provides a
guide regarding entities likely to be regulated by this action. To
determine whether particular activities may be regulated by this
action, you should carefully examine the regulations. You may direct
questions regarding the applicability of this action to the person
listed in FOR FURTHER INFORMATION CONTACT.
I. Executive Summary
II. Overview of Final Rule
III. Purpose of the Rule
IV. Purpose of Analytical Approach Considered as Part of Decision-
Making
V. Regulatory Alternatives Considered
VI. Analytical Approach as Applied to Regulatory Alternatives
VII. What does the analysis show, and what does it mean?
VIII. How do the final standards fulfill the agencies' statutory
obligations?
IX. Compliance and Enforcement
X. Regulatory Notices and Analyses
I. Executive Summary
 NHTSA (on behalf of the Department of Transportation) and EPA are
issuing final rules to adopt and modify standards regulating corporate
average fuel economy and tailpipe carbon dioxide (CO2)
emissions and use/leakage of other air conditioning refrigerants for
passenger cars and light trucks for MYs 2021-2026.\3\ These final rules
follow the proposal issued in August 2018 and respond to each agency's
legal obligation to set standards based on the factors Congress
directed them to consider, as well as the direction of the United
States Supreme Court in Massachusetts v. EPA, which stated that ``there
is no reason to think the two agencies cannot both administer their
obligations and yet avoid inconsistency.'' \4\ These standards are the
product of significant and ongoing work by both agencies to craft
regulatory requirements for the same group of vehicles and vehicle
manufacturers. This work aims to facilitate, to the extent possible
within the statutory directives issued to each agency, the ability of
automobile manufacturers to meet all requirements under both programs
with a single national fleet under one national program of fuel economy
and tailpipe CO2 emission regulation.
---------------------------------------------------------------------------
 \3\ Throughout this document and the accompanying FRIA, the
agencies will often use the term ``CO2'' or ``tailpipe
CO2'' to refer broadly to EPA's suite of light duty
vehicle GHG standards.
 \4\ 549 U.S. 497, 532 (2007).
---------------------------------------------------------------------------
 The CAFE and CO2 emissions standards established by
these final rules will increase in stringency at 1.5 percent per year
from MY 2020 levels over MYs 2021-2026. The ``1.5 percent'' regulatory
alternative is new for the final rule and was not expressly analyzed in
the NPRM, but it is a logical outgrowth of the NPRM analysis, being
well within the range of alternatives then considered and consistent
with discussions by both the agencies and commenters that there are
benefits to having standards that increase at the same rate for all
fleets. These standards apply to light-duty vehicles, which NHTSA
divides for purposes of regulation into passenger cars and light
trucks, and EPA divides into passenger cars, light-duty trucks, and
medium-duty passenger vehicles (i.e., sport utility vehicles, cross-
over utility vehicles, and light trucks). Both the CAFE and
CO2 standards are vehicle-footprint-based, as are the
standards currently in effect. These standards will become more
stringent for each model year from 2021 to 2026, relative to the MY
2020 standards. Generally, the larger the vehicle footprint, the less
numerically stringent the corresponding vehicle CO2 and
miles-per-gallon (mpg) targets. As a result of the footprint-based
standards, the burden of compliance is distributed across all vehicle
footprints and across all manufacturers. Each manufacturer is subject
to individualized standards for passenger cars and light trucks, in
each model year, based on the vehicles it produces. When standards are
carefully crafted, both in terms of the footprint curves and the rate
of increase in stringency of those curves, manufacturers are not
[[Page 24176]]
compelled to build vehicles of any particular size or type.
 Knowing that many readers are accustomed to considering CAFE and
CO2 emissions standards in terms of the mpg and grams-per-
mile (g/mi) values that the standards are projected to eventually
require, the agencies include those projections here. EPA's standards
are projected to require, on an average industry fleet-wide basis, 201
grams per mile (g/mi) of CO2 in model year 2030, while
NHTSA's standards are projected to require, on an average industry
fleet-wide basis, 40.5 miles per gallon (mpg) in model year 2030. The
agencies note that real-world CO2 is typically 25 percent
higher and real-world fuel economy is typically 20 percent lower than
the CO2 and CAFE compliance values discussed here, and also
note that a portion of EPA's expected ``CO2'' improvements
will in fact be made through improvements in minimizing air
conditioning leakage and through use of alternative refrigerants, which
will not contribute to fuel economy but will contribute toward
reductions of climate-related emissions.
 In these final rules, NHTSA and EPA are reaching similar
conclusions on similar grounds: even though each agency has its own
distinct statutory authority and factors, the relevant considerations
overlap in many ways. Both agencies recognize that they are balancing
the relevant considerations in somewhat different ways from how they
may have been balanced previously, as in the 2012 final rule and in
EPA's Initial Determination, but the current balancing is called for in
light of the facts before the agencies. The balancing in these final
rules is also somewhat different from how the agencies balanced their
respective considerations in the proposal, in part because of updates
to analytical inputs and methodologies, previewed in the NPRM and made
in response to public comments, that collectively resulted in changes
to the analytical outputs. For example, between the notice and final
rule, the agencies updated fuel price projections to somewhat greater
values, updated the analysis fleet to MY 2017, updated estimates of the
efficacy and cost of fuel-saving technologies, revised procedures for
calculating impacts on vehicle sales and scrappage, updated models for
estimating highway safety impacts, updated estimates of highway
congestion costs, and updated estimates of annual mileage accumulation,
holding VMT (before applying the rebound effect) constant between
regulatory alternative. Moreover, the cost-benefit analysis conducted
for these final rules has even been overtaken by events in many ways
over recent weeks. Based upon current events, and for additional
reasons discussed in Section VI.D.1 the benefits of saving additional
fuel through more stringent standards are potentially even smaller than
estimated in this rulemaking analysis.
 The standards finalized today fit the pattern of gradual, tough,
but feasible stringency increases that take into account real world
performance, shifts in fuel prices, and changes in consumer behavior
toward crossovers and SUVs and away from more efficient sedans. This
approach ensures that manufacturers are provided with sufficient lead
time to achieve standards, considering the cost of compliance. The
costs to both industry and automotive consumers would have been too
high under the standards set forth in 2012, and by lowering the auto
industry's costs to comply with the program, with a commensurate
reduction in per-vehicle costs to consumers, the standards enhance the
ability of the fleet to turn over to newer, cleaner and safer vehicles.
 More stringent standards also have the potential for overly
aggressive penetration rates for advanced technologies relative to the
penetration rates seen in the final standards, especially in the face
of an unknown degree of consumer acceptance of both the increased costs
and of the technologies themselves--particularly given current
projections of relatively low fuel prices during that timeframe. As a
kind of insurance policy against future fuel price volatility,
standards that increase at 1.5 percent per year for cars and trucks
will help to keep fleet fuel economy higher than they would be
otherwise when fuel prices are low, which is not improbable over the
next several years.\5\ At the same time, the standards help to address
these issues by maintaining incentives to promote broader deployment of
advanced technologies, and so provides a means of encouraging their
further penetration while leaving manufacturers alternative technology
choices. Steady, gradual increases in stringency ensure that the
benefits of reduced GHG emissions and fuel consumption are achieved
without the potential for disruption to automakers or consumers.
---------------------------------------------------------------------------
 \5\ For example, EIA currently expects U.S. retail gasoline
prices to average $2.14/gallon in 2020, compared to $2.69/gallon in
2019 (see https://www.eia.gov/outlooks/steo/archives/mar20.pdf), and
$3.68/gallon in 2012 (see https://www.eia.gov/dnav/pet/hist/LeafHandler.ashx?n=PET&s=EMM_EPM0_PTE_NUS_DPG&f=A). While gasoline
prices may foreseeably rise over the rulemaking time frame, it is
also very foreseeable that they will not rise to the $4-5/gallon
that many Americans saw over the 2008-2009 time frame, that caused
the largest shift seen toward smaller and higher-fuel-economy
vehicles. See, e.g., Figure VIII-2 below.
---------------------------------------------------------------------------
 Standards that increase at 1.5 percent per year represent a
reasonable balance of additional technology and required per-vehicle
costs, consumer demand for fuel economy, fuel savings and emissions
avoided given the foreseeable state of the global oil market and the
minimal effect on climate between finalizing 1.5 percent standards
versus more stringent standards. The final standards will also result
in year-over-year improvements in fleetwide fuel economy, resulting in
energy conservation that helps address environmental concerns,
including criteria pollutant, air toxic pollutant, and carbon
emissions.
 The agencies project that under these final standards, required
technology costs would be reduced by $86 to $126 billion over the
lifetimes of vehicles through MY 2029. Equally important, purchase
prices costs to U.S. consumers for new vehicles would be $977 to $1,083
lower, on average, than they would have been if the agencies had
retained the standards set forth in the 2012 final rule and originally
upheld by EPA in January 2017. While these final standards are
estimated to result in 1.9 to 2.0 additional billion barrels of fuel
consumed and from 867 to 923 additional million metric tons of
CO2 as compared to current estimates of what the standards
set forth in 2012 would require, the agencies explain at length below
why the overall benefits of the final standards outweigh these
additional costs.\6\
---------------------------------------------------------------------------
 \6\ 1.9 to 2.0 barrels of fuel is approximately 78 to 84 gallons
of fuel.
---------------------------------------------------------------------------
 For the CAFE program, overall (fleetwide) net benefits vary from
$16.1 billion at a 7 percent discount rate to -$13.1 billion at a 3
percent discount rate. For the CO2 program, overall
(fleetwide) societal net benefits vary from $6.4 billion at a 7 percent
discount rate to -$22.0 billion at a 3 percent discount rate. The net
benefits straddle zero, and are very small relative to the scale of
reduced required technology costs, which range from $86.3 billion to
$126.0 billion for the CAFE and CO2 programs across 7
percent and 3 percent discount rates. Likewise, net benefits are very
small relative to the scale of reduced retail fuel savings over the
full life of all vehicles manufactured during the 2021 through 2029
model years, which range from $108.6 billion to $185.1 billion for the
CAFE and CO2 programs across 7 percent and 3 percent
discount rates. Similarly, all of the alternatives have small net
benefits, ranging from $18.4 billion to -$31.1
[[Page 24177]]
billion for the CAFE and CO2 programs across 7 percent and 3
percent discount rates.\7\
---------------------------------------------------------------------------
 \7\ See Table II-12 to Table II-15 for costs, benefits and net
benefits.
---------------------------------------------------------------------------
 NHTSA and EPA believe their analysis of the final rule represents
the best available science, evidence, and methodologies for assessing
the impacts of changes in CAFE and CO2 emission standards.
In fact, the agencies note that today's analysis represents a marked
improvement over prior rulemakings. Previously, the agencies were
unable to model the impact of the standards on new vehicle sales or the
retirement of older vehicles in the fleet, and, instead, were forced to
assume, contrary to economic theory and empirical evidence, that the
number of new vehicles sold and older vehicles scrapped remained static
across regulatory alternatives. Today's analysis--as commenters to
previous rulemakings and EPA's Science Advisory Board have argued is
necessary \8\--quantifies the sales and scrappage impacts of the
standards, including the associated safety benefits, and represents a
significant step forward in agencies' ability to comprehensively
analyze the impacts of CAFE and CO2 emission standards.
---------------------------------------------------------------------------
 \8\ Science Advisory Board, U.S. EPA. Review of EPA's Proposed
SAFE rule at 4 (Feb. 27, 2020), available at https://
yosemite.epa.gov/sab/sabproduct.nsf/LookupWebProjectsCurrentBOARD/
1FACEE5C03725F268525851F006319BB/$File/EPA-SAB-20-003+.pdf
[hereinafter ``SAB Report''].
---------------------------------------------------------------------------
 However, the agencies also believe it is important to be
transparent about analytical limitations. For example, EPA's Science
Advisory Board stressed that the agencies account for ``evolving
consumer preferences for performance and other vehicle attributes,''
\9\ yet due to limitations on the agencies' current ability to model
buyers' choices among combinations of various attributes and their
costs, the primary analysis does not account for the consumer benefits
of other vehicle features that may be sacrificed for costly
technologies that improve fuel economy. The agencies' analysis assumes
that under these final standards, attributes of new cars and light
trucks other than fuel economy would remain identical to those under
the baseline standards, so that changes in sales prices and fuel
economy would be the only sources of benefits or costs to new car and
light truck buyers. In other words, the agencies' primary analysis does
not consider that producers will likely respond to buyers' demands by
reallocating some their savings in production costs due to lower
technology costs to add or improve other attributes that consumers
value more highly than the increases in fuel economy the augural
standards would have required. The agencies have long debated whether
and how best to model the consumer benefits of other vehicle
attributes, and note that they have made considerable progress.\10\
However, despite these potential analytical shortcomings, the agencies
reaffirm that today's analysis represents the most complete and
rigorous examination of CAFE and CO2 emission standards to
date, and provide decision-makers a powerful analytical tool--
especially since the limitations are known, do not bias the central
analysis' results, and are afforded due consideration.
---------------------------------------------------------------------------
 \9\ SAB at 10.
 \10\ In their evaluations of previous CAFE and CO2
rules, the agencies attempted to account for this possibility by
conducting sensitivity analyses that reduced the fuel savings and
other benefits to vehicle buyers by a significant fraction. For
example, NHTSA's analysis supporting the Final Rule establishing
CAFE standards for model year 2012-16 cars and light trucks tested
the sensitivity of their central estimates of social costs and
benefits to the assumptions that 25 percent and 50 percent of
benefits to buyers were offset by opportunity costs of foregone
improvements in attributes other than fuel economy; see NHTSA, Final
Regulatory Impact Analysis: Corporate Average Fuel Economy for Model
year 2012-16 Passenger Cars and Light Trucks, March 2010, at 563-565
and Table X-9, at 566-56; see also, NHTSA, Final Regulatory Impact
Analysis: Corporate Average Fuel Economy for Model year 2017-25
Passenger Cars and Light Trucks, August 2012, at 1087 and Tables X-
18a, X-18b, and X-18c, at 1099-1104. The agencies acknowledged that
this was not a completely satisfactory way to represent the
sacrifices in vehicles' other attributes that car and light truck
manufacturers might find it necessary to make in order to comply
with the increasingly stringent standards those previous rules
established. At the time, however, the agencies were unable to
identify specific attributes that manufacturers were most likely to
sacrifice, measure the tradeoffs between increased fuel economy and
improvements in those attributes, or assess the potential losses in
utility to car and light truck buyers. In an effort to improve on
their previous treatment of this issue, the agencies' evaluation of
this final rule includes a sensitivity case that assumes
manufacturers redirect their technology cost savings from complying
with less stringent standards to instead improve a combination of
cars' and light trucks' other attributes that offers benefits to
their buyers significantly exceeding those costs. The magnitude of
these (net) benefits is interpreted as the opportunity cost of the
improvements in vehicles' other attributes that would have been
sacrificed if the augural standards had been enacted. The method the
agencies use to approximate these benefits, together with its effect
on the rule's overall benefits and costs, is discussed in detail in
Section VI.D.1.b)(8). Briefly, the results of this sensitivity
analysis suggest the Final Rule would generate net benefits for the
CAFE and CO2 programs ranging from $34.9 to $55.4 billion
at 3% and 7% discount rates.
---------------------------------------------------------------------------
 In terms of the agencies' respective statutory authorities, EPA is
setting national tailpipe CO2 emissions standards for
passenger cars and light trucks under section 202(a) of the Clean Air
Act (CAA),\11\ and taking other actions under its authority to
establish metrics and measure passenger car and light truck fleet fuel
economy pursuant to the Energy Policy and Conservation Act (EPCA),\12\
while NHTSA is setting national corporate average fuel economy (CAFE)
standards under EPCA, as amended by the Energy Independence and
Security Act (EISA) of 2007.\13\ As summarized above and as discussed
in much greater detail below, the agencies believe that these represent
appropriate levels of CO2 emissions standards and maximum
feasible CAFE standards for MYs 2021-2026, pursuant to their respective
statutory authorities. Sections III and VIII below contain detailed
discussions of both agencies' statutory obligations and authorities.
---------------------------------------------------------------------------
 \11\ 42 U.S.C. 7521(a).
 \12\ 49 U.S.C. 32904(c).
 \13\ 49 U.S.C. 32902.
---------------------------------------------------------------------------
 Section 202(a) of the CAA requires EPA to establish standards for
emissions of pollutants from new motor vehicles that cause or
contribute to air pollution that may reasonably be anticipated to
endanger public health or welfare. Standards under section 202(a) thus
take effect only ``after providing such period as the Administrator
finds necessary to permit the development and application of the
requisite technology, giving appropriate consideration to the cost of
compliance within such period.'' \14\ In establishing such standards,
EPA must consider issues of technical feasibility, cost, and available
lead time, among other things.
---------------------------------------------------------------------------
 \14\ CAA Sec. 202(a); 42 U.S.C. 7512(a)(2).
---------------------------------------------------------------------------
 EPCA, as amended by EISA, contains a number of provisions governing
how NHTSA must set CAFE standards. EPCA requires that the Department of
Transportation establish separate passenger car and light truck
standards \15\ at ``the maximum feasible average fuel economy level
that the Secretary decides the manufacturers can achieve in that model
year,'' \16\ based on the agency's consideration of four statutory
factors: technological feasibility, economic practicability, the effect
of other standards of the Government on fuel economy, and the need of
the United States to conserve energy.\17\ EPCA does not define these
terms or specify what weight to give each concern in balancing them--
such considerations are left within the discretion of the Secretary of
Transportation (delegated to NHTSA) based upon current information.
Accordingly, NHTSA interprets these factors and determines the
appropriate weighting that leads to the maximum
[[Page 24178]]
feasible standards given the circumstances present at the time of
promulgating each CAFE standard rulemaking. While EISA, for MYs 2011-
2020, additionally required that standards increase ``ratably'' and be
set at levels to ensure that the CAFE of the industry-wide combined
fleet of new passenger cars and light trucks reach at least 35 mpg by
MY 2020,\18\ EISA requires that standards for MYs 2021-2030 simply be
set at the maximum feasible level as determined by the Secretary (and
by delegation, NHTSA).\19\
---------------------------------------------------------------------------
 \15\ 49 U.S.C. 32902(b)(1).
 \16\ 49 U.S.C. 32902(a).
 \17\ 49 U.S.C. 32902(f).
 \18\ 49 U.S.C. 32902(b)(2)(A) and (C).
 \19\ 49 U.S.C. 32902(b)(2)(B).
---------------------------------------------------------------------------
 In the NPRM, the agencies sought comment on a variety of possible
changes to existing compliance flexibilities that have been created
over the past several years. The vast majority of the existing
compliance flexibilities are not being changed, but a small number of
flexibilities related to real-world fuel efficiency improvements are
being finalized. In addition, EPA will continue to allow manufacturers
to make improvements relating to air conditioning refrigerants and
leakage and will credit those improvements toward CO2
compliance, and EPA is making no changes in the amounts of credits
available. EPA is also not making any changes to the existing
CH4 and N2O standards. EPA is also extending the
``0 g/mi upstream'' incentive for electric vehicles beyond its current
sunset of MY 2021, through MY 2026. EPA is also establishing a credit
multiplier for natural gas vehicles through the 2026 model year.
Otherwise, compliance flexibilities in the two programs do not change
significantly for the final rule. These changes should help to
streamline manufacturer use of those flexibilities in certain respects.
While manufacturers and suppliers sought a number of other additional
compliance flexibilities, the agencies have concluded that the
aforementioned existing flexibilities are reasonable and appropriate,
and that additional flexibilities are not justified.
 Table I-1 and Table I-2 present the total costs, benefits, and net
benefits for the 2021-2026 preferred alternative CAFE and
CO2 levels, relative to the MY 2022-2025 existing/augural
standards (with the MY 2025 standards repeated for MY 2026) and current
MY 2021 standard. The preferred alternative exhibits a stringency rate
increase of 1.5 percent per year for both passenger cars and light
trucks. The values in Table I-1 and Table I-2 display (in total and
annualized forms) costs for all MYs 1978-2029 vehicles, and the
benefits and net benefits represent the impacts of the standards over
the full lifetimes of the vehicles sold or projected to be sold during
model years 1978-2029.
 For this analysis, negative signs are used for changes in costs or
benefits that decrease from those that would have resulted from the
existing/augural standards. Any changes that would increase either
costs or benefits are shown as positive changes. Thus, an alternative
that decreases both costs and benefits, will show declines (i.e., a
negative sign) in both categories. From Table I-1 and Table I-2, the
preferred alternative (Alternative 3) is estimated to decrease costs
relative to the baseline by $182 to $280 billion over the lifetime of
MYs 1978-2029 passenger vehicles (range determined by discount rate
across both CAFE and CO2 programs). It will also decrease
benefits from $175 to $294 billion over the life of these MY fleets.
The net impact will be a decrease from $22 billion to an increase of
$16 billion in total net benefits to society over this roughly 52-year
timeframe. Annualized, this amounts to roughly -$0.8 to 1.2 billion in
net benefits per year.
BILLING CODE 4910-59-P
[GRAPHIC] [TIFF OMITTED] TR30AP20.001
[GRAPHIC] [TIFF OMITTED] TR30AP20.002
[[Page 24179]]
 Table I-3 and Table I-4 lists costs, benefits, and net benefits for
all seven alternatives that were examined.
[GRAPHIC] [TIFF OMITTED] TR30AP20.003
[GRAPHIC] [TIFF OMITTED] TR30AP20.004
 Table I-5 and Table I-6 show a summary of various impacts of the
preferred alternative for CAFE and CO2 standards. Impacts
are presented in monetized and non-monetized values, as well as from
the perspective of society and the consumer.
[[Page 24180]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.005
[[Page 24181]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.006
BILLING CODE 4910-59-C
 The agencies note that the NPRM drew more public comments (and,
particularly, more pages of substantive comments) than any rulemaking
in the history of the CAFE or CO2 tailpipe emissions
programs--exceeding 750,000 comments. The agencies recognized in the
NPRM that the proposal was significantly different from the final rules
set forth in 2012, and explained at length the reasons for those
differences--namely, that new information and considerations, along
with an expanded and updated analysis, had led to different tentative
conclusions. Today's final rules represent a further evolution of the
work that supported the proposal, based on improved quantitative
methodology and in careful consideration of the hundreds of thousands
of public comments and deep reflection on the serious issues before the
agencies. Simply put, the agencies have heard the comments, and today's
analysis and decision reflect the agencies' grappling with the issues
commenters raised, as well as all of the other information before the
agencies. These programs and issues are weighty, and the agencies
believe that a reasonable balance has been struck in these final rules
between the many competing national needs that these regulatory
programs collectively address.
II. Overview of Final Rule
A. Summary of Proposal
 In the NPRM, the National Highway Traffic Safety Administration
(NHTSA) and the Environmental Protection Agency (EPA) (collectively,
``the
[[Page 24182]]
agencies'') proposed the ``Safer Affordable Fuel-Efficient (SAFE)
Vehicles Rule for Model Years 2021-2026 Passenger Cars and Light
Trucks'' (SAFE Vehicles Rule). The proposed SAFE Vehicles Rule would
set Corporate Average Fuel Economy (CAFE) and carbon dioxide
(CO2) emissions standards, respectively, for passenger cars
and light trucks manufactured for sale in the United States in model
years (MYs) 2021 through 2026.\20\
---------------------------------------------------------------------------
 \20\ NHTSA sets CAFE standards under the Energy Policy and
Conservation Act of 1975 (EPCA), as amended by the Energy
Independence and Security Act of 2007 (EISA). EPA sets
CO2 standards under the Clean Air Act (CAA).
---------------------------------------------------------------------------
 The agencies explained that they must act to propose and finalize
these standards and do not have discretion to decline to regulate.
Congress requires NHTSA to set CAFE standards for each model year.\21\
Congress also requires EPA to set emissions standards for light-duty
vehicles if EPA has made an ``endangerment finding'' that the pollutant
in question--in this case, CO2--``cause[s] or contribute[s]
to air pollution which may reasonably be anticipated to endanger public
health or welfare.'' \22\ NHTSA and EPA proposed the standards
concurrently because tailpipe CO2 emissions standards are
directly and inherently related to fuel economy standards,\23\ and, if
finalized, the rules would apply concurrently to the same fleet of
vehicles. By working together to develop the proposals, the agencies
aimed to reduce regulatory burden on industry and improve
administrative efficiency.
---------------------------------------------------------------------------
 \21\ 49 U.S.C. 32902.
 \22\ 42 U.S.C. 7521; see also 74 FR 66495 (Dec. 15, 2009)
(``Endangerment and Cause or Contribute Findings for Greenhouse
Gases under Section 202(a) of the Clean Air Act'').
 \23\ See, e.g., 75 FR 25324, at 25327 (May 7, 2010) (``The
National Program is both needed and possible because the
relationship between improving fuel economy and reducing tailpipe
CO2 emissions is a very direct and close one. The amount
of those CO2 emissions is essentially constant per gallon
combusted of a given type of fuel. Thus, the more fuel efficient a
vehicle is, the less fuel it burns to travel a given distance. The
less fuel it burns, the less CO2 it emits in traveling
that distance. [citation omitted] While there are emission control
technologies that reduce the pollutants (e.g., carbon monoxide)
produced by imperfect combustion of fuel by capturing or converting
them to other compounds, there is no such technology for
CO2. Further, while some of those pollutants can also be
reduced by achieving a more complete combustion of fuel, doing so
only increases the tailpipe emissions of CO2. Thus, there
is a single pool of technologies for addressing these twin problems,
i.e., those that reduce fuel consumption and thereby reduce
CO2 emissions as well.'').
---------------------------------------------------------------------------
 The agencies discussed some of the history leading to the proposal,
including the 2012 final rule, the expectations regarding a mid-term
evaluation as required by EPA regulation, and the rapid process over
2016 and early 2017 by which EPA issued its first Final Determination
that the CO2 standards set in 2012 for MYs 2022-2025
remained appropriate based on the information then before the EPA
Administrator.\24\ The agencies also discussed President Trump's
direction in March 2017 to restore the original mid-term evaluation
timeline, and EPA's subsequent information-gathering process and
announcement that it would reconsider the January 2017
Determination.\25\ EPA ultimately concluded that the standards set in
2012 for MYs 2022-2025 were no longer appropriate.\26\ For NHTSA, in
turn, the ``augural'' CAFE standards for MYs 2022-2025 were never
final, and as explained in the 2012 final rule, NHTSA was obligated
from the beginning to undertake a new rulemaking to set CAFE standards
for MYs 2022-2025.
---------------------------------------------------------------------------
 \24\ See 83 FR at 42987 (Aug.24, 2018).
 \25\ Id.
 \26\ 83 FR 16077 (Apr. 2, 2018).
---------------------------------------------------------------------------
 The NPRM thus began the rulemaking process for both agencies to
establish new standards for MYs 2022-2025 passenger cars and light
trucks. Standards were concurrently proposed for MY 2026 in order to
provide regulatory stability for as many years as is legally
permissible for both agencies together. The NPRM also included revised
standards for MY 2021 passenger cars and light trucks, because the
agencies tentatively concluded, based on the information and analysis
then before them, that the CAFE standards previously set for MY 2021
were no longer maximum feasible, and the CO2 standards
previously set for MY 2021 were no longer appropriate. Agencies always
have authority under the Administrative Procedure Act to revisit
previous decisions in light of new facts, as long as they provide
notice and an opportunity for comment, and the agencies stated that it
is plainly the best practice to do so when changed circumstances so
warrant.\27\
---------------------------------------------------------------------------
 \27\ See FCC v. Fox Television, 556 U.S. 502 (2009).
---------------------------------------------------------------------------
 The NPRM proposed to maintain the CAFE and CO2 standards
applicable in MY 2020 for MYs 2021-2026, and took comment on a wide
range of alternatives, including different stringencies and retaining
existing CO2 standards and the augural CAFE standards.\28\
Table II-1, Table II-2, and Table II-3 show the estimates, under the
NPRM analysis, of what the MY 2020 CAFE and CO2 curves would
translate to, in terms of miles per gallon (mpg) and grams per mile (g/
mi), in MYs 2021-2026, as well as the regulatory alternatives
considered in the NPRM. In addition to retaining the MY 2020
CO2 standards through MY 2026, EPA proposed and sought
comment on excluding air conditioning refrigerants and leakage, and
nitrous oxide and methane emissions for compliance with CO2
standards after model year 2020, in order to improve harmonization with
the CAFE program. EPA also sought comment on whether to change existing
methane and nitrous oxide standards that were finalized in the 2012
rule. The proposal was accompanied by a 1,600 page Preliminary
Regulatory Impact Analysis (PRIA) and, for NHTSA, a 500 page Draft
Environmental Impact Statement (DEIS), with more than 800 pages of
appendices and the entire CAFE model, including the software source
code and documentation, all of which were also subject to comment in
their entirety and all of which received significant comments.
---------------------------------------------------------------------------
 \28\ The agencies noted that this did not mean that the miles
per gallon and grams per mile levels that were estimated for the MY
2020 fleet in 2012 would be the ``standards'' going forward into MYs
2021-2026. Both NHTSA and EPA set CAFE and CO2 standards,
respectively, as mathematical functions based on vehicle footprint.
These mathematical functions that are the actual standards are
defined as ``curves'' that are separate for passenger cars and light
trucks, under which each vehicle manufacturer's compliance
obligation varies depending on the footprints of the cars and trucks
that it ultimately produces for sale in a given model year. It was
the MY 2020 CAFE and CO2 curves that the agencies
proposed would continue to apply to the passenger car and light
truck fleets for MYs 2021-2026. The mpg and g/mi values which those
curves would eventually require of the fleets in those model years
would be known for certain only at the ends of each of those model
years. While it is convenient to discuss CAFE and CO2
standards as a set ``mpg,'' ``g/mi,'' or ``mpg-e'' number,
attempting to define those values based on the information then
before the agency would necessarily end up being inaccurate.
---------------------------------------------------------------------------
BILLING CODE 4910-59-P
[[Page 24183]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.007
[[Page 24184]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.008

---------------------------------------------------------------------------
 \29\ The carbon dioxide equivalents of air conditioning
refrigerant leakage, nitrous oxide emissions, and methane emissions
were included for compliance with the EPA standards for all MYs
under the baseline/no action alternative in the NPRM. Carbon dioxide
equivalent is calculated using the Global Warming Potential (GWP) of
each of the emissions.
 \30\ Beginning in MY 2021, the proposal provided that the GWP
equivalents of air conditioning refrigerant leakage, nitrous oxide
emissions, and methane emissions would no longer be able to be
included with the tailpipe CO2 for compliance with
tailpipe CO2 standards.
---------------------------------------------------------------------------
[[Page 24185]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.009
BILLING CODE 4910-59-C
 The agencies explained in the NPRM that new information had been
gathered and new analysis performed since publication of the 2012 final
rule establishing CAFE and CO2 standards for MYs 2017 and
beyond and since issuance of the 2016 Draft TAR and EPA's 2016 and
early 2017 ``mid-term evaluation'' process. This new information and
analysis helped lead the agencies to the tentative conclusion that
holding standards constant at MY 2020 levels through MY 2026 was
maximum feasible, for CAFE purposes, and appropriate, for
CO2 purposes.
 The agencies further explained that technologies had played out
differently in the fleet from what the agencies previously assumed:
That while there remain a wide variety of technologies available to
improve fuel economy and reduce CO2 emissions, it had become
clear that there were reasons to temper previous optimism about the
costs, effectiveness, and consumer acceptance of a number of
technologies. In addition, over the years between the previous analyses
and the NPRM, automakers had added considerable amounts of technologies
to their new vehicle fleets, meaning that the agencies were no longer
free to make certain assumptions about how some of those technologies
could be used going forward. For example, some technologies that could
be used to improve fuel economy and reduce emissions had not been used
entirely for that purpose, and some of the benefit of these
technologies had gone instead toward improving other vehicle
attributes. Other technologies had been tried, and had been met with
significant customer acceptance issues. The agencies underscored the
importance of reflecting the fleet as it stands today, with the
technology it has and as that technology has been used, and considering
what technology remains on the table at this point, whether and when it
can realistically be available for widespread use in production, and
how much it would cost to implement.
 The agencies also acknowledged the math of diminishing returns: As
CAFE and CO2 emissions standards increase in stringency, the
benefit of continuing to increase in stringency decreases. In mpg
terms, a vehicle owner who drives a light vehicle 15,000 miles per year
(a typical assumption for analytical purposes) \31\ and trades in a
vehicle with fuel economy of 15 mpg for one with fuel economy of 20
mpg, will reduce their annual fuel consumption from 1,000 gallons to
750 gallons--saving 250 gallons annually. If, however, that owner were
to trade in a vehicle with fuel economy of 30 mpg for one with fuel
economy of 40 mpg, the owner's annual gasoline consumption would drop
from 500 gallons/year to 375 gallons/year--only 125 gallons even though
the mpg improvement is twice as large. Going from 40 to 50 mpg would
save only 75 gallons/year. Yet each additional fuel economy improvement
becomes much more expensive as the easiest to achieve low-cost
technological improvement options are chosen. In CO2 terms,
if a vehicle emits 300 g/mi CO2,
[[Page 24186]]
a 20 percent improvement is 60 g/mi, so the vehicle would emit 240 g/
mi; but if the vehicle emits 180 g/mi, a 20 percent improvement is only
36 g/mi, so the vehicle would get 144 g/mi. In order to continue
achieving similarly large (on an absolute basis) emissions reductions,
the percentage reduction must also continue to increase.
---------------------------------------------------------------------------
 \31\ A different vehicle-miles-traveled (VMT) assumption would
change the absolute numbers in the example, but would not change the
mathematical principles.
---------------------------------------------------------------------------
 Related, average real-world fuel economy is lower than average fuel
economy required under CAFE and CO2 standards. The 2012
Federal Register notice announcing augural CAFE and CO2
standards extending through MY 2025 indicated that, if met entirely
through the application of fuel-saving technology, the MY 2025
CO2 standards would result in an average requirement
equivalent to 54.5 mpg. However, because the CO2 standards
provide credit for reducing leakage of AC refrigerants and/or switching
to lower-GWP refrigerants, and these actions do not affect fuel
economy, the notice explained that the corresponding fuel economy
requirement (under the CAFE program) would be 49.7 mpg. These estimates
were based on a market forecast grounded in the MY 2008 fleet. The
notice also presented analysis using a market forecast grounded in the
MY 2010 fleet, showing a 48.7 mpg average CAFE requirement.
 In the real world, fuel economy is, on average, about 20% lower
than as measured under regulatory test procedures. In the real world,
then, these new standards were estimated to require 39.0-39.8 mpg.
 Today's analysis indicates that the requirements under the
baseline/augural CAFE standards would average 46.6 mpg in MY 2029. The
lower value results from changes in the fleet forecast which reflects
consumer preference for larger vehicles than was forecast for the 2012
rulemaking. In the real world, the requirements average about 37.1 mpg.
Under the final standards issued today, the regulatory test procedure
requirements average 40.5 mpg, corresponding to 33.2 mpg in the real
world. Buyers of new vehicles experience real-world fuel economy, with
levels varying among drivers (due to a wide range of factors). Vehicle
fuel economy labels provide average real-world fuel economy information
to buyers.
[GRAPHIC] [TIFF OMITTED] TR30AP20.010
 Vehicle owners also face fuel prices at the pump. The agencies
noted in the NPRM that when fuel prices are high, the value of fuel
saved may be enough to offset the cost of further fuel economy/
emissions reduction improvements, but the agencies recognized that
then-current projections of fuel prices by the Energy Information
Administration did not indicate particularly high fuel prices in the
foreseeable future. The agencies explained that fundamental structural
shifts had occurred in global oil markets since the 2012 final rule,
largely due to the rise of U.S. production and export of shale oil. The
consequence over time of diminishing returns from more stringent fuel
economy/emissions reduction standards, especially when combined with
relatively low fuel prices, is greater difficulty for automakers to
find a market of consumers willing to buy vehicles that meet the
increasingly stringent standards. American consumers have long
demonstrated that in times of relatively low fuel prices, fuel economy
is not a top priority for the majority of them, even when highly fuel
efficient vehicle models are available.
 The NPRM analysis sought to improve how the agencies captured the
effects of higher new vehicle prices on fleet composition as a whole by
including an improved model for vehicle scrappage rates. As new vehicle
prices increase, consumers tend to continue using older vehicles for
longer, slowing fleet turnover and thus slowing improvements in fleet-
wide fuel economy, reductions in CO2 emissions, reductions
in criteria pollutant emissions, and advances in safety. That aspect of
the analysis was also driven by the agencies' updated estimates of
average per-vehicle cost increases due to
[[Page 24187]]
higher standards, which were several hundred dollars higher than
previously estimated. The agencies cited growing concerns about
affordability and negative equity for many consumers under these
circumstances, as loan amounts grow and loan terms extend.
 For all of the above reasons, the agencies proposed to maintain the
MY 2020 fuel economy and CO2 emissions standards for MYs
2021-2026. The agencies explained that they estimated, relative to the
standards for MYs 2021-2026 put forth in 2012, that an additional 0.5
million barrels of oil would be consumed per day (about 2 to 3 percent
of projected U.S. consumption) if that proposal were finalized, but
that they also expected the additional fuel costs to be outweighed by
the cost savings from new vehicle purchases; that more than 12,700 on-
road fatalities and significantly more injuries would be prevented over
the lifetimes of vehicles through MY 2029 as compared to the standards
set forth in the 2012 final rule over the lifetimes of vehicles as more
new and safer vehicles are purchased than the current (and augural)
standards; and that environmental impacts, on net, would be relatively
minor, with criteria and toxic air pollutants not changing noticeably,
and with estimated atmospheric CO2 concentrations increasing
by 0.65 ppm (a 0.08 percent increase), which the agencies estimated
would translate to 0.003 degrees Celsius of additional temperature
increase relative to the standards finalized in 2012.
 Under the NPRM analysis, the agencies tentatively concluded that
maintaining the MY 2020 curves for MYs 2021-2026 would save American
auto consumers, the auto industry, and the public a considerable amount
of money as compared to EPA retaining the previously-set CO2
standards and NHTSA finalizing the augural standards. The agencies
explained that this had been identified as the preferred alternative,
in part, because it appeared to maximize net benefits compared to the
other alternatives analyzed, and recognizing the statutory
considerations for both agencies. Relative to the standards issued in
2012, under CAFE standards, the NPRM analysis estimated that costs
would decrease by $502 billion overall at a three-percent discount rate
($335 billion at a seven-percent discount rate) and benefits were
estimated to decrease by $326 billion at a three-percent discount rate
($204 billion at a seven-percent discount rate). Thus, net benefits
were estimated to increase by $176 billion at a three-percent discount
rate and $132 billion at a seven-percent discount rate. The estimated
impacts under CO2 standards were estimated to be similar,
with net benefits estimated to increase by $201 billion at a three-
percent discount rate and $141 billion at a seven-percent discount
rate.
 The NPRM also sought comment on a variety of potential changes to
NHTSA's and EPA's compliance programs for CAFE and CO2 as
well as related programs, including questions about automaker requests
for additional flexibilities and agency interest in reducing market-
distorting incentives and improving transparency; and on a proposal to
withdraw California's CAA preemption waiver for its ``Advanced Clean
Car'' regulations, with an accompanying discussion of preemption of
State standards under EPCA.\32\ The agencies sought comment broadly on
all aspects of the proposal.
---------------------------------------------------------------------------
 \32\ Agency actions relating to California's CAA waiver and EPCA
preemption have since been finalized, see 84 FR 51310 (Sept. 27,
2019), and will not be discussed in great detail as part of this
final rule.
---------------------------------------------------------------------------
B. Public Participation Opportunities and Summary of Comments
 The NPRM was published on NHTSA's and EPA's websites on August 2,
2018, and published in the Federal Register on August 24, 2018,
beginning a 60-day comment period. The agencies subsequently extended
the official comment period for an additional three days, and left the
dockets open for more than a year after the start of the comment
period, considering late comments to the extent practicable. A separate
Federal Register notice also published on August 24, 2018, which
announced the locations, dates, and times of three public hearings to
be held on the proposal: One in Fresno, California, on September 24,
2018; one in Dearborn, Michigan, on September 25, 2018; and one in
Pittsburgh, Pennsylvania, on September 26, 2018. Each hearing started
at 10 a.m. local time; the Fresno hearing ended at 5:10 p.m. and
resulted in a 235 page transcript; the Dearborn hearing ran until 5:26
p.m. and resulted in a 330 page transcript; and the Pittsburgh hearing
ran until 5:06 p.m. and also resulted in a 330 page transcript. Each
hearing also collected several hundred pages of comments from
participants, in addition to the hearing transcripts.
 Besides the comments submitted as part of the public hearings,
NHTSA's docket received a total of 173,359 public comments in response
to the proposal as of September 18, 2019, and EPA's docket a total of
618,647 public comments, for an overall total of 792,006. NHTSA also
received several hundred comments on its DEIS to the separate DEIS
docket. While the majority of individual comments were form letters,
the agencies received over 6,000 pages of substantive comments on the
proposal.
 Many commenters generally supported the proposal and many
commenters opposed it. Commenters supporting the proposal tended to
cite concerns about the cost of new vehicles, while commenters opposing
the proposal tended to cite concerns about additional fuel expenditures
and the impact on climate change. Many comments addressed the modeling
used for the analysis, and specifically the inclusion, operation, and
results of the sales and scrappage modules that were part of the NPRM's
analysis, while many addressed the NPRM's safety findings and the role
that those findings played in the proposal's justification. Many other
comments addressed California's standards and role in Federal decision-
making; as discussed above, those comments are further summarized and
responded to in the separate Federal Register notice published in
September 2019. Nearly every aspect of the NPRM's analysis and
discussion received some level of comment by at least one commenter.
The comments received, as a whole, were both broad and deep, and the
agencies appreciate the level of engagement of commenters in the public
comment process and the information and opinions provided.
C. Changes in Light of Public Comments and New Information
 The agencies made a number of changes to the analysis between the
NPRM and the final rule in response to public comments and new
information that was received in those comments or otherwise became
available to the agencies. While these changes, their rationales, and
their effects are discussed in detail in the sections below, the
following represents a high-level list of some of the most significant
changes:
 Some regulatory alternatives were dropped from
consideration, and one was added;
 updated analysis fleet, and changes to technologies on
``baseline'' vehicles within the fleet to reflect better their current
properties and improve modeling precision;
 no civil penalties assumed to be paid after MY 2020 under
CAFE program;
 updates and expansions in accounting for certain over-
compliance
[[Page 24188]]
credits, including early credits earned in EPA's program;
 updates and expansions to CAFE Model's technology paths;
 updates to inputs defining the range of manufacturer-,
technology-, and product-specific constraints;
 updates to allow the model to adopt a more advanced
technology if it is more cost-effective than an earlier technology on
the path;
 precision improvements to the modeling of A/C efficiency
and off-cycle credits;
 updates to model's ``effective cost'' metric;
 extended explicit simulation of technology application
through MY 2050;
 expanded presentation of the results to include ``calendar
year'' analysis;
 quantifying different types of health impacts from changes
in air pollution, rather than only accounting for such impacts in
aggregate estimates of the social costs of air pollution;
 updated costs to 2018 dollars;
 updated fuel costs based on the AEO 2019 version of NEMS;
 a variety of technology updates in response to comments
and new information;
 updated accounting of rebound VMT between regulatory
alternatives;
 updated estimates of the macroeconomic cost of petroleum
dependence;
 updated response of total new vehicle sales to increases
in fuel efficiency and price; and
 updated response of vehicle retirement rates to changes in
new vehicle fuel efficiency and transaction price.
 Sections IV and VI below discuss these updates in significant
detail.
D. Final Standards--Stringency
 As explained above, the agencies have chosen to set CAFE and
CO2 standards that increase in stringency by 1.5 percent
year over year for MYs 2021-2026. Separately, EPA has decided to retain
the A/C refrigerant and leakage and CH4 and N2O
standards set forth in 2012 for MYs 2021 and beyond, and the stringency
of the CO2 standards in this final rule reflect the
``offset'' also established in 2012 based on assumptions made at that
time about anticipated HFC emissions reductions.
 When the agencies state that stringency will increase at 1.5
percent per year, that means that the footprint curves which actually
define the standards for CAFE and CO2 emissions will become
more stringent at 1.5 percent per year. Consistent with Congress's
direction in EISA to set CAFE standards based on a mathematical
formula, which EPA harmonized with for the CO2 emissions
standards, the standard curves are equations, which are slightly
different for CAFE and CO2, and within each program,
slightly different for passenger cars and light trucks. Each program
has a basic equation for a fleet standard, and then values that change
to cause the stringency changes are the coefficients within the
equations. For passenger cars, consistent with prior rulemakings, NHTSA
is defining fuel economy targets as follows:
[GRAPHIC] [TIFF OMITTED] TR30AP20.011
where:
TARGETFE is the fuel economy target (in mpg) applicable to a
specific vehicle model type with a unique footprint combination,
a is a minimum fuel economy target (in mpg),
b is a maximum fuel economy target (in mpg),
c is the slope (in gallons per mile per square foot, or gpm, per
square foot) of a line relating fuel consumption (the inverse of
fuel economy) to footprint, and
d is an intercept (in gpm) of the same line.
 Here, MIN and MAX are functions that take the minimum and maximum
values, respectively, of the set of included values. For example,
MIN[40,35] = 35 and MAX(40, 25) = 40, such that MIN[MAX(40, 25), 35] =
35.
 For light trucks, also consistent with prior rulemakings, NHTSA is
defining fuel economy targets as follows:
[GRAPHIC] [TIFF OMITTED] TR30AP20.012
where:
TARGETFE is the fuel economy target (in mpg) applicable to a
specific vehicle model type with a unique footprint combination,
a, b, c, and d are as for passenger cars, but taking values specific
to light trucks,
e is a second minimum fuel economy target (in mpg),
f is a second maximum fuel economy target (in mpg),
g is the slope (in gpm per square foot) of a second line relating
fuel consumption (the inverse of fuel economy) to footprint, and
h is an intercept (in gpm) of the same second line.
 The final CAFE standards (described in terms of their footprint-
based curves) are as follows, with the values for the coefficients
changing over time:
[[Page 24189]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.013
 These equations are presented graphically below, where the x-axis
represents vehicle footprint and the y-axis represents fuel economy,
showing that in the CAFE context, targets are higher (fuel economy) for
smaller footprint vehicles and lower for larger footprint vehicles:
BILLING CODE 4910-59-C
[[Page 24190]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.014
BILLING CODE 4910-59-P
[[Page 24191]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.015
BILLING CODE 4910-59-C
 EPCA, as amended by EISA, requires that any manufacturer's
domestically-manufactured passenger car fleet must meet the greater of
either 27.5 mpg on average, or 92 percent of the average fuel economy
projected by the Secretary for the combined domestic and non-domestic
passenger automobile fleets manufactured for sale in the U.S. by all
manufacturers in the model year, which projection shall be published in
the Federal Register when the standard for that model year is
promulgated in accordance with 49 U.S.C. 32902(b).\33\ Any time NHTSA
establishes or changes a passenger car standard for a model year, the
MDPCS for that model year must also be evaluated or re-evaluated and
established accordingly. Thus, this final rule establishes the
applicable MDPCS for MYs 2021-2026. Table II-8 lists the minimum
domestic passenger car standards.
---------------------------------------------------------------------------
 \33\ 49 U.S.C. 32902(b)(4).
 [GRAPHIC] [TIFF OMITTED] TR30AP20.016

 EPA CO2 standards are as follows. Rather than expressing
these standards as linear functions with accompanying minima and
maxima, similar to the approach NHTSA has followed since 2005 in
specifying attribute-based standards, the following tables specify flat
standards that apply below and above specified footprints, and a linear
function that applies between those footprints. The two approaches are
mathematically identical. For passenger cars with a footprint of less
than or equal to 41 square feet, the gram/mile CO2 target
value is selected for the appropriate model year from Table II-9:
[[Page 24192]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.017
 For passenger cars with a footprint of greater than 56 square feet,
the gram/mile CO2 target value is selected for the
appropriate model year from Table II-10:
[GRAPHIC] [TIFF OMITTED] TR30AP20.018
 For passenger cars with a footprint that is greater than 41 square
feet and less than or equal to 56 square feet, the gram/mile
CO2 target value is calculated using the following equation
and rounded to the nearest 0.1 grams/mile.
[[Page 24193]]
Target CO2 = [a x f] + b
Where f is the vehicle footprint and a and b are selected from Table
II-11 for the appropriate model year:
[GRAPHIC] [TIFF OMITTED] TR30AP20.019
 For light trucks with a footprint of less than or equal to 41
square feet, the gram/mile CO2 target value is selected for
the appropriate model year from Table II-12:
[[Page 24194]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.020
 For light trucks with a footprint greater than the minimum value
specified in the table below for each model year, the gram/mile
CO2 target value is selected for the appropriate model year
from Table II-13:
[GRAPHIC] [TIFF OMITTED] TR30AP20.021
[[Page 24195]]
 For light trucks with a footprint that is greater than 41 square
feet and less than or equal to the maximum footprint value specified in
Table II-14 below for each model year, the gram/mile CO2
target value is calculated using the following equation and rounded to
the nearest 0.1 grams/mile.
Target CO2 = (a x f) + b
Where f is the footprint and a and b are selected from Table II-14
below for the appropriate model year:
[GRAPHIC] [TIFF OMITTED] TR30AP20.022
 These equations are presented graphically below, where the x-axis
represents vehicle footprint and the y-axis represents the
CO2 target. The targets are lower for smaller footprint
vehicles and higher for larger footprint vehicles:
BILLING CODE 4910-59-P
[[Page 24196]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.023
[[Page 24197]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.024
BILLING CODE 4910-59-C
 Except that EPA elected to apply a slightly different slope when
defining passenger car targets, CO2 targets may be expressed
as direct conversion of fuel economy targets, as follows:
[GRAPHIC] [TIFF OMITTED] TR30AP20.025
where 8887 g/gal relates grams of CO2 emitted to gallons
of fuel consumed, and OFFSET reflects the fact that that HFC
emissions from lower-GWP A/C refrigerants and less leak-prone A/C
systems are counted toward average CO2 emissions, but
EPCA provides no basis to count reduced HFC emissions toward CAFE
levels.
 For the reader's benefit, Table II-15, Table II-16, and Table II-17
show the estimates, under the final rule analysis, of what the MYs
2021-2026 CAFE and CO2 curves would translate to, in terms
of miles per gallon (mpg) and grams per mile (g/mi).
BILLING CODE 4910-59-P
[[Page 24198]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.026
BILLING CODE 4910-59-C
[[Page 24199]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.027
 As the following tables demonstrate, averages of manufacturers'
estimated requirements are more stringent (i.e., for CAFE, higher, and
for CO2, lower) under the final standards than under the
proposed standards:
[GRAPHIC] [TIFF OMITTED] TR30AP20.028
[[Page 24200]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.029
E. Final Standards--Impacts
 This section summarizes the estimated costs and benefits of the MYs
2021-2026 CAFE and CO2 emissions standards for passenger
cars and light trucks, as compared to the regulatory alternatives
considered. These estimates helped inform the agencies' choices among
the regulatory alternatives considered and provide further confirmation
that the final standards are maximum feasible, for NHTSA, and
appropriate, for EPA. The costs and benefits estimated to result from
the CAFE standards are presented first, followed by those estimated to
result from the CO2 standards. For several reasons, the
estimates for costs and benefits presented for the different programs,
while consistent, are not identical. NHTSA's and EPA's standards are
projected to result in slightly different fuel efficiency improvements.
EPA's CO2 standard is nominally more stringent in part due
to its assumptions about manufacturers' use of air conditioning
leakage/refrigerant replacement credits, which are expected to result
in reduced emissions of HFCs. NHTSA's final standards are based solely
on assumptions about fuel economy improvements, and do not account for
emissions reductions that do not relate to fuel economy. In addition,
the CAFE and CO2 programs offer somewhat different program
flexibilities and provisions, primarily because NHTSA is statutorily
prohibited from considering some flexibilities when establishing CAFE
standards, while EPA is not.\34\ The analysis underlying this final
rule reflects many of those additional EPA flexibilities, which
contributes to differences in how the agencies estimate manufacturers
could comply with the respective sets of standards, which in turn
contributes to differences in estimated impacts of the standards. These
differences in compliance flexibilities are discussed in more detail in
Section IX below.
---------------------------------------------------------------------------
 \34\ See 49 U.S.C. 32902(h); CAA Sec. 202(a).
---------------------------------------------------------------------------
 Table II-20 to Table II-23 present all subcategories of costs and
benefits of this final rule for all seven alternatives proposed. Costs
include application of fuel economy technology to new vehicles,
consumer surplus, crash costs due to changes in VMT, as well as, noise
and congestion. Benefits include fuel savings, consumer surplus,
refueling time, and clean air.
BILLING CODE 4910-59-P
[[Page 24201]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.030
[[Page 24202]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.031
[[Page 24203]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.032
[[Page 24204]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.033
[[Page 24205]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.034
[[Page 24206]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.035
[[Page 24207]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.036
[[Page 24208]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.037
[[Page 24209]]
BILLING CODE 4910-59-C
F. Other Programmatic Elements
1. Compliance and Flexibilities
 Automakers seeking to comply with the CAFE and CO2
standards are generally expected to add fuel economy-improving
technologies to their new vehicles to boost their overall fleet fuel
economy levels. Readers will remember that improving fuel economy
directly reduces CO2 emissions, because CO2 is a
natural and inevitable byproduct of fossil fuel combustion to power
vehicles. The CAFE and CO2 programs contain a variety of
compliance provisions and flexibilities to accommodate better
automakers' production cycles, to reward real-world fuel economy
improvements that cannot be reflected in the 1975-developed test
procedures, and to incentivize the production of certain types of
vehicles. While the agencies sought comment on a broad variety of
changes and potential expansions of the programs' compliance
flexibilities in the NPRM, the agencies determined, after considering
the comments, to make a few changes to the flexibilities proposed in
the NPRM in this final rule. The most noteworthy change is the
retention, in the CO2 program, of the flexibilities that
allow automakers to continue to use HFC reductions toward their
CO2 compliance, and that extend the ``0 grams/mile''
assumption for electric vehicles through MY 2026 (i.e., recognizing
only the tailpipe emissions of full battery-electric vehicles and not
recognizing the upstream emissions caused by the electricity usage of
those vehicles). In the NPRM, EPA had proposed to remove and sought
comment on removing those flexibilities from the CO2
program, but determined not to remove them in this final rule. EPA and
NHTSA are also removing from the programs, starting in MY 2022, the
credit/FCIV for full-size pickup trucks that are either hybrids or
over-performing by a certain amount relative to their targets, and
allowing technology suppliers to begin the petition process for off-
cycle credits/adjustments.
 Table II-24, Table II-25, Table II-26, and Table II-27 provide a
summary of the various compliance provisions in the two programs; their
authorities; and any changes included as part of this final rule:
BILLING CODE 4910-59-P
[GRAPHIC] [TIFF OMITTED] TR30AP20.038
[[Page 24210]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.039
BILLING CODE 4910-59-P
[[Page 24211]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.040

---------------------------------------------------------------------------
 \35\ The CAFE program uses an energy efficiency metric and
standards that are expressed in miles per gallon. For PHEVs and
BEVs, to determine gasoline the equivalent fuel economy for
operation on electricity, a Petroleum Equivalency Factor (PEF) is
applied to the measured electrical consumption. The PEF for
electricity was established by the Department of Energy, as required
by statute, and includes an accounting for upstream energy
associated with the production and distribution for electricity
relative to gasoline. Therefore, the CAFE program includes upstream
accounting based on the metric that is consistent with the fuel
economy metric. The PEF for electricity also includes an incentive
that effectively counts only 15 percent of the electrical energy
consumed.
---------------------------------------------------------------------------
 Providing a technology neutral basis by which manufacturers meet
fuel economy and CO2 emissions standards encourages an
efficient and level playing field. The agencies continue to have a
desire to minimize incentives that disproportionately favor one
technology over another. Some of this may involve regulations
established by other Federal agencies. In the near future, NHTSA and
EPA intend to work with other relevant Federal agencies to pursue
regulatory means by which we can further ensure technology neutrality
in this field.
2. Preemption/Waiver
 As discussed above, the issues of Clean Air Act waivers of
preemption under Section 209 and EPCA/EISA preemption under 49 U.S.C.
32919 are not addressed in today's final rule, as
[[Page 24212]]
they were the subject of a separate final rulemaking action by the
agencies in September 2019. While many comments were received in
response to the NPRM discussion of those issues, those comments have
been addressed and responded to as part of that separate rulemaking
action.
III. Purpose of the Rule
 The Administrative Procedure Act (APA) requires agencies to
incorporate in their final rules a ``concise general statement of their
basis and purpose.'' \36\ While the entire preamble document represents
the agencies' overall explanation of the basis and purpose for this
regulatory action, this section within the preamble is intended as a
direct response to that APA (and related CAA) requirements. Executive
Order 12866 further states that ``Federal agencies should promulgate
only such regulations as are required by law, are necessary to
interpret the law, or are made necessary by compelling public need,
such as material failures of private markets to protect or improve the
health and safety of the public, the environment, or the well-being of
the American people.'' \37\ Section III.C of the FRIA accompanying this
rulemaking discusses at greater length the question of whether a market
failure exists that these final rules may address.
---------------------------------------------------------------------------
 \36\ 5 U.S.C. 553(c); see also Clean Air Act section
307(d)(6)(A), 42 U.S.C. 7607(d)(6)(A).
 \37\ E.O. 12866, Section 1(a).
---------------------------------------------------------------------------
 NHTSA and EPA are legally obligated to set CAFE and GHG standards,
respectively, and do not have the authority to decline to regulate.\38\
The agencies are issuing these final rules to fulfill their respective
statutory obligations to provide maximum feasible fuel economy
standards and limit emissions of pollutants from new motor vehicles
which have been found to endanger public health and welfare (in this
case, specifically carbon dioxide (CO2); EPA has already set
standards for methane (CH4), nitrous oxide (N2O),
and hydrofluorocarbons (HFCs) and is not revising them in this rule).
Continued progress in meeting these statutory obligations is both
legally necessary and good for America--greater energy security and
reduced emissions protect the American public. The final standards
continue that progress, albeit at a slower rate than the standards
finalized in 2012.
---------------------------------------------------------------------------
 \38\ For CAFE, see 49 U.S.C. 32902; for CO2, see 42
U.S.C. 7521(a).
---------------------------------------------------------------------------
 National annual gasoline consumption and CO2 emissions
currently total about 140 billion gallons and 5,300 million metric
tons, respectively. The majority of this gasoline (about 130 billion
gallons) is used to fuel passenger cars and light trucks, such as will
be covered by the CAFE and CO2 standards issued today.
Accounting for both tailpipe emissions and emissions from ``upstream''
processes (e.g., domestic refining) involved in producing and
delivering fuel, passenger cars and light trucks account for about
1,500 million metric tons (mmt) of current annual CO2
emissions. The agencies estimate that under the standards issued in
2012, passenger car and light truck annual gasoline consumption would
steadily decline, reaching about 80 billion gallons by 2050. The
agencies further estimate that, because of this decrease in gasoline
consumption under the standards issued in 2012, passenger car and light
truck annual CO2 emissions would also steadily decline,
reaching about 1,000 mmt by 2050. Under the standards issued today, the
agencies estimate that, instead of declining from about 140 billion
gallons annually today to about 80 billion gallons annually in 2050,
passenger car and light truck gasoline consumption would decline to
about 95 billion gallons. The agencies correspondingly estimate that
instead of declining from about 1,500 mmt annually today to about 1,000
mmt annually in 2050, passenger car and light truck CO2
emissions would decline to about 1,100 mmt. In short, the agencies
estimate that under the standards issued today, annual passenger car
and light truck gasoline consumption and CO2 emissions will
continue to steadily decline over the next three decades, even if not
quite as rapidly as under the previously-issued standards.
 The agencies also estimate that these impacts on passenger car and
light truck gasoline consumption and CO2 emissions will be
accompanied by a range of other energy- and climate-related impacts,
such as reduced electricity consumption (because today's standards
reduce the estimated rate at which the market might shift toward
electric vehicles) and increased CH4 and N2O
emissions. These estimated impacts, discussed below and in the FEIS
accompanying today's notice, are dwarfed by estimated impacts on
gasoline consumption and CO2 emissions.
 As explained above, these final rules set or amend fuel economy and
carbon dioxide standards for model years 2021-2026. Many commenters
argued that it was not appropriate to amend previously-established
CO2 and CAFE standards, generally because those commenters
believed that the administrative record established for the 2012 final
rule and EPA's January 2017 Final Determination was superior to the
record that informed the NPRM, and that that prior record led
necessarily to the policy conclusion that the previously-established
standards should remain in place.\39\ Some commenters similarly argued
that EPA's Revised Final Determination--which, for EPA, preceded this
regulatory action--was invalid because, they allege, it did not follow
the procedures established for the mid-term evaluation that EPA
codified into regulation,\40\ and also because the Revised Final
Determination was not based on the prior record.\41\
---------------------------------------------------------------------------
 \39\ Comments arguing that the prior record was superior to the
current record, and thus a better basis for decision-making, will be
addressed throughout the balance of this preamble.
 \40\ 40 CFR 86.1818-12(h).
 \41\ See, e.g., comments from the States and Cities, Attachment
1, Docket No. NHTSA-2018-0067-11735, at 40-42; CARB, Detailed
Comments, Docket No. NHTSA-2018-0067-11873, at 71-72; CBD et. al,
Appendix A, Docket No. NHTSA-2018-0067-12000, at 214-228.
---------------------------------------------------------------------------
 The agencies considered a range of alternatives in the proposal,
including the baseline/no action alternative of retaining the existing
EPA carbon dioxide standards. As the agencies explained in the
proposal, the proposal was entirely de novo, based on an entirely new
analysis reflecting the best and most up-to-date information available
to the agencies.\42\ This rulemaking action is separate and distinct
from EPA's Revised Final Determination, which itself was neither a
proposed nor a final decision that the standards ``must'' be revised.
EPA retained full discretion in this rulemaking to revise the standards
or not revise them. In any event, the case law is clear that agencies
are free to reconsider their prior decisions.\43\ With that legal
principle in mind, the agencies agree with commenters that the amended
(and new) CO2 and CAFE standards must be consistent with the
[[Page 24213]]
CAA and EPCA/EISA, respectively, and this preamble and the accompanying
FRIA explain in detail why the agencies believe they are consistent.
The section below discusses briefly the authority given to the agencies
by their respective governing statutes, and the factors that Congress
directed the agencies to consider as they exercise that authority in
pursuit of fulfilling their statutory obligations.
---------------------------------------------------------------------------
 \42\ 83 FR 42968, 42987 (Aug. 24, 2018).
 \43\ See, e.g., Encino Motorcars, LLC v. Navarro, 136 S. Ct.
2117, 2125 (2016) (``Agencies are free to change their existing
policies as long as they provide a reasoned explanation for the
change.''); FCC v. Fox Television Stations, Inc., 556 U.S. 502, 515
(2009) (When an agency changes its existing position, it ``need not
always provide a more detailed justification than what would suffice
for a new policy created on a blank slate. Sometimes it must--when,
for example, its new policy rests on factual findings that
contradict those which underlay its prior policy; or when its prior
policy has engendered serious reliance interests that must be taken
into account . . . . In such cases it is not that further
justification is demanded by the mere fact of policy change, but
that a reasoned explanation is needed for disregarding facts and
circumstances that underlay or were engendered by the prior
policy.'')
---------------------------------------------------------------------------
A. EPA's Statutory Requirements
 EPA is setting national CO2 standards for passenger cars
and light trucks under Section 202(a) of the Clean Air Act (CAA).\44\
Section 202(a) of the CAA requires EPA to establish standards for
emissions of pollutants from new motor vehicles which cause or
contribute to air pollution which may reasonably be anticipated to
endanger public health or welfare.\45\ In establishing such standards,
EPA considers issues of technical feasibility, cost, available lead
time, and other factors. Standards under section 202(a) thus take
effect only ``after providing such period as the Administrator finds
necessary to permit the development and application of the requisite
technology, giving appropriate consideration to the cost of compliance
within such period.'' \46\ EPA's statutory requirements are further
discussed in Section VIII.A.
---------------------------------------------------------------------------
 \44\ 42 U.S.C. 7521(a).
 \45\ See Coalition for Responsible Regulation v. EPA, 684 F.3d
102, 114-115 (D.C. Cir. 2012) (`` `If EPA makes a finding of
endangerment, the Clean Air Act requires the [a]gency to regulate
emissions of the deleterious pollutant from new motor vehicles . . .
. Given the non-discretionary duty in Section 202(a)(1) and the
limited flexibility available under Section 202(a)(2), which this
court has held related only to the motor vehicle industry, . . . EPA
had no statutory basis on which it could ground [any] reasons for
further inaction' '') (quoting Massachusetts v. EPA, 549 U.S. 497,
533-35 (2007).
 \46\ 42 U.S.C. 7521(a)(2).
---------------------------------------------------------------------------
B. NHTSA's Statutory Requirements
 NHTSA is setting national Corporate Average Fuel Economy (CAFE)
standards for passenger cars and light trucks for each model year as
required under EPCA, as amended by EISA.\47\ EPCA mandates a motor
vehicle fuel economy regulatory program that balances statutory factors
in setting minimum fuel economy standards to facilitate energy
conservation. EPCA allocates the responsibility for implementing the
program between NHTSA and EPA as follows: NHTSA sets CAFE standards for
passenger cars and light trucks; EPA establishes the procedures for
testing, tests vehicles, collects and analyzes manufacturers' data, and
calculates the individual and average fuel economy of each
manufacturer's passenger cars and light trucks; and NHTSA enforces the
standards based on EPA's calculations.
---------------------------------------------------------------------------
 \47\ EPCA and EISA direct the Secretary of Transportation to
develop, implement, and enforce fuel economy standards (see 49
U.S.C. 32901 et. seq.), which authority the Secretary has delegated
to NHTSA at 49 CFR 1.94(c).
---------------------------------------------------------------------------
 The following sections enumerate specific statutory requirements
for NHTSA in setting CAFE standards and NHTSA's interpretations of
them, where applicable. Many comments were received on these
requirements and interpretations. Because this is intended as an
overview section, those comments will be addressed below in Section
VIII rather than here, and the agencies refer readers to that part of
the document for more information.
 For each future model year, EPCA (as amended by EISA) requires that
DOT (by delegation, NHTSA) establish separate passenger car and light
truck standards at ``the maximum feasible average fuel economy level
that the Secretary decides the manufacturers can achieve in that model
year,'' \48\ based on the agency's consideration of four statutory
factors: ``technological feasibility, economic practicability, the
effect of other motor vehicle standards of the Government on fuel
economy, and the need of the United States to conserve energy.'' \49\
The law also allows NHTSA to amend standards that are already in place,
as long as doing so meets these requirements.\50\ EPCA does not define
these terms or specify what weight to give each concern in balancing
them; thus, NHTSA defines them and determines the appropriate weighting
that leads to the maximum feasible standards given the circumstances in
each CAFE standard rulemaking.\51\
---------------------------------------------------------------------------
 \48\ 49 U.S.C. 32902(a) and (b).
 \49\ 49 U.S.C. 32902(f).
 \50\ 49 U.S.C. 32902(g).
 \51\ See Center for Biological Diversity v. NHTSA, 538 F.3d
1172, 1195 (9th Cir. 2008) (hereafter ``CBD v. NHTSA'') (``The EPCA
clearly requires the agency to consider these four factors, but it
gives NHTSA discretion to decide how to balance the statutory
factors--as long as NHTSA's balancing does not undermine the
fundamental purpose of the EPCA: Energy conservation.'')
---------------------------------------------------------------------------
 EISA added several other requirements to the setting of separate
passenger car and light truck standards. Standards must be ``based on 1
or more vehicle attributes related to fuel economy and express[ed] . .
. in the form of a mathematical function.'' \52\ New standards must
also be set at least 18 months before the model year in question, as
would amendments to increase standards previously set.\53\ NHTSA must
regulations prescribing average fuel economy standards for at least 1,
but not more than 5, model years at a time.\54\ A number of comments
addressed these requirements; for the reader's reference, those
comments will be summarized and responded to in Section VIII. EISA also
added the requirement that NHTSA set a minimum standard for
domestically-manufactured passenger cars,\55\ which will also be
discussed further in Section VIII below.
---------------------------------------------------------------------------
 \52\ 49 U.S.C. 32902(b)(3)(A).
 \53\ 49 U.S.C. 32902(a), (g)(2).
 \54\ 49 U.S.C. 39202(b)(3)(B).
 \55\ 49 U.S.C. 32902(b)(4).
---------------------------------------------------------------------------
 For MYs 2011-2020, EISA further required that the separate
standards for passenger cars and for light trucks be set at levels high
enough to ensure that the achieved average fuel economy for the entire
industry-wide combined fleet of new passenger cars and light trucks
reach at least 35 mpg not later than MY 2020, and standards for those
years were also required to ``increase ratably.'' \56\ For model years
after 2020, standards must be set at the maximum feasible level.\57\
---------------------------------------------------------------------------
 \56\ 49 U.S.C. 32902(b)(2)(A) and (C). NHTSA has CAFE standards
in place that are projected to result in industry-achieved fuel
economy levels over 35 mpg in MY 2020. EPA typically provides
verified final CAFE data from manufacturers to NHTSA several months
or longer after the close of the MY in question, so the actual MY
2020 fuel economy will not be known until well after MY 2020 has
ended. The standards for all MYs up to and including 2020 are known
and not at issue in this regulatory action, so these provisions are
noted for completeness rather than immediate relevance to this final
rule. Because neither of these requirements apply after MY 2020,
they are not relevant to this rulemaking and will not be discussed
further.
 \57\ 49 U.S.C. 32902(b)(2)(B).
---------------------------------------------------------------------------
1. Factors That Must Be Considered in Deciding What Levels of CAFE
Standards are ``Maximum Feasible''
(a) Technological Feasibility
 ``Technological feasibility'' refers to whether a particular method
of improving fuel economy can be available for commercial application
in the model year for which a standard is being established. Thus, in
determining the level of new standards, the agency is not limited to
technology that is already being commercially applied at the time of
the rulemaking. For this rulemaking, NHTSA has evaluated and considered
all types of technologies that improve real-world fuel economy,
although not every possible technology was expressly included in the
analysis, as discussed in Section VI and also in Section VIII.
(b) Economic Practicability
 ``Economic practicability'' refers to whether a standard is one
``within the
[[Page 24214]]
financial capability of the industry, but not so stringent as to'' lead
to ``adverse economic consequences, such as a significant loss of jobs
or the unreasonable elimination of consumer choice.'' \58\ The agency
has explained in the past that this factor can be especially important
during rulemakings in which the automobile industry is facing
significantly adverse economic conditions (with corresponding risks to
jobs). Economic practicability is a broad factor that includes
considerations of the uncertainty surrounding future market conditions
and consumer demand for fuel economy in addition to other vehicle
attributes.\59\ In an attempt to evaluate the economic practicability
of different future levels of CAFE standards (i.e., the regulatory
alternatives considered in this rulemaking), NHTSA considers a variety
of factors, including the annual rate at which manufacturers can
increase the percentage of their fleet(s) that employ a particular type
of fuel-saving technology, the specific fleet mixes of different
manufacturers, assumptions about the cost of the standards to
consumers, and consumers' valuation of fuel economy, among other
things, including, in part, safety.
---------------------------------------------------------------------------
 \58\ 67 FR 77015, 77021 (Dec. 16, 2002).
 \59\ See, e.g., Center for Auto Safety v. NHTSA (``CAS''), 793
F.2d 1322 (D.C. Cir. 1986) (Administrator's consideration of market
demand as component of economic practicability found to be
reasonable); Public Citizen v. NHTSA, 848 F.2d 256 (D.C. Cir. 1988)
(Congress established broad guidelines in the fuel economy statute;
agency's decision to set lower standard was a reasonable
accommodation of conflicting policies).
---------------------------------------------------------------------------
 It is important to note, however, that the law does not preclude a
CAFE standard that poses considerable challenges to any individual
manufacturer. The Conference Report for EPCA, as enacted in 1975, makes
clear, and the case law affirms, ``a determination of maximum feasible
average fuel economy should not be keyed to the single manufacturer
which might have the most difficulty achieving a given level of average
fuel economy.'' \60\ Instead, NHTSA is compelled ``to weigh the
benefits to the nation of a higher fuel economy standard against the
difficulties of individual automobile manufacturers.'' \61\
Accordingly, while the law permits NHTSA to set CAFE standards that
exceed the projected capability of a particular manufacturer as long as
the standard is economically practicable for the industry as a whole,
the agency cannot simply disregard that impact on individual
manufacturers.\62\ That said, in setting fuel economy standards, NHTSA
does not seek to maintain competitive positions among the industry
players, and notes that while a particular CAFE standard may pose
difficulties for one manufacturer as being too high or too low, it may
also present opportunities for another. NHTSA has long held that the
CAFE program is not necessarily intended to maintain the competitive
positioning of each particular company. Rather, it is intended to
enhance the fuel economy of the vehicle fleet on American roads, while
protecting motor vehicle safety and paying close attention to the
economic risks.
---------------------------------------------------------------------------
 \60\ Center for Auto Safety v. NHTSA (``CAS''), 793 F.2d 1322,
1352 (D.C. Cir. 1986).
 \61\ Id.
 \62\ Id. (``. . . the Secretary must weigh the benefits to the
nation of a higher average fuel economy standard against the
difficulties of individual automobile manufacturers.'')
---------------------------------------------------------------------------
(c) The Effect of Other Motor Vehicle Standards of the Government on
Fuel Economy
 ``The effect of other motor vehicle standards of the Government on
fuel economy'' involves an analysis of the effects of compliance with
emission, safety, noise, or damageability standards on fuel economy
capability and thus on average fuel economy. In many past CAFE
rulemakings, NHTSA has said that it considers the adverse effects of
other motor vehicle standards on fuel economy. It said so because, from
the CAFE program's earliest years,\63\ the effects of such compliance
on fuel economy capability over the history of the program have been
negative ones. For example, safety standards that have the effect of
increasing vehicle weight lower vehicle fuel economy capability and
thus decrease the level of average fuel economy that the agency can
determine to be feasible. NHTSA has considered the additional weight
that it estimates would be added in response to new safety standards
during the rulemaking timeframe. NHTSA has also accounted for EPA's
``Tier 3'' standards for criteria pollutants in its estimates of
technology effectiveness.\64\
---------------------------------------------------------------------------
 \63\ 42 FR 63184, 63188 (Dec. 15, 1977). See also 42 FR 33534,
33537 (Jun. 30, 1977).
 \64\ See Section VI, below.
---------------------------------------------------------------------------
 The NPRM also discussed how EPA's CO2 standards for
light-duty vehicles and California's Advanced Clean Cars program fit
into NHTSA's consideration of ``the effect of other motor vehicle
standards of the Government on fuel economy.'' The agencies note that
on September 19, 2019, to ensure One National Program for automobile
fuel economy and carbon dioxide emissions standards, the agencies
finalized regulatory text related to preemption of State tailpipe
CO2 standards and Zero Emission Vehicle (ZEV) mandates under
EPCA and partial withdrawal of a waiver previously provided to
California under the Clean Air Act.\65\ This final rule's impact on
State programs--including California's--will therefore be somewhat
different from the NPRM's consideration. In the interest of brevity,
this preamble will hold further discussion of that point, along with
responses to comments received, until Section VIII.
---------------------------------------------------------------------------
 \65\ 84 FR 51310 (Sept. 27, 2019).
---------------------------------------------------------------------------
(d) The Need of the United States To Conserve Energy
 ``The need of the United States to conserve energy'' means ``the
consumer cost, national balance of payments, environmental, and foreign
policy implications of our need for large quantities of petroleum,
especially imported petroleum.'' \66\ Environmental implications
principally include changes in emissions of carbon dioxide and criteria
pollutants and air toxics. Prime examples of foreign policy
implications are energy independence and security concerns.
---------------------------------------------------------------------------
 \66\ 42 FR 63184, 63188 (1977).
---------------------------------------------------------------------------
(1) Consumer Costs and Fuel Prices
 Fuel for vehicles costs money for vehicle owners and operators. All
else equal (and this is an important qualification), consumers benefit
from vehicles that need less fuel to perform the same amount of work.
Future fuel prices are a critical input into the economic analysis of
potential CAFE standards because they determine the value of fuel
savings both to new vehicle buyers and to society, the amount of fuel
economy that the new vehicle market is likely to demand in the absence
of new standards, and they inform NHTSA about the consumer cost of the
nation's need for large quantities of petroleum. In this final rule,
NHTSA's analysis relies on fuel price projections estimated using the
version of NEMS used for the U.S. Energy Information Administration's
(EIA) Annual Energy Outlook for 2019.\67\ Federal government agencies
generally use EIA's price projections in their assessment of future
energy-related policies.
---------------------------------------------------------------------------
 \67\ The analysis for the proposal relied on fuel price
projections from AEO 2017; the difference in the projections is
discussed in Section VI.
---------------------------------------------------------------------------
(2) National Balance of Payments
 Historically, the need of the United States to conserve energy has
included consideration of the ``national balance of payments'' because
of concerns that importing large amounts of oil created a
[[Page 24215]]
significant wealth transfer to oil-exporting countries and left the
U.S. economically vulnerable.\68\ As recently as 2009, nearly half of
the U.S. trade deficit was driven by petroleum,\69\ yet this concern
has largely lain fallow in more recent CAFE actions, in part because
other factors besides petroleum consumption have since played a bigger
role in the U.S. trade deficit.\70\ Given significant recent increases
in U.S. oil production and corresponding decreases in oil imports, this
concern seems likely to remain fallow for the foreseeable future.\71\
Increasingly, changes in the price of fuel have come to represent
transfers between domestic consumers of fuel and domestic producers of
petroleum rather than gains or losses to foreign entities.
---------------------------------------------------------------------------
 \68\ See, e.g., 42 FR 63184, 63192 (Dec. 15, 1977) (``A major
reason for this need [to reduce petroleum consumption] is that the
importation of large quantities of petroleum creates serious balance
of payments and foreign policy problems. The United States currently
spends approximately $45 billion annually for imported petroleum.
But for this large expenditure, the current large U.S. trade deficit
would be a surplus.'')
 \69\ See ``Today in Energy: Recent improvements in petroleum
trade balance mitigate U.S. trade deficit,'' U.S. Energy Information
Administration (Jul. 21, 2014), available at https://www.eia.gov/todayinenergy/detail.php?id=17191.
 \70\ See, e.g., Nida [Ccedil]akir Melek and Jun Nie, ``What
Could Resurging U.S. Energy Production Mean for the U.S. Trade
Deficit,'' Mar. 7, 2018, Federal Reserve Bank of Kansas City.
Available at https://www.kansascityfed.org/publications/research/mb/articles/2018/what-could-resurging-energy-production-mean. The
authors state that ``The decline in U.S. net energy imports has
prevented the total U.S. trade deficit from widening further. . . .
In 2006, petroleum accounted for about 16 percent of U.S. goods
imports and about 3 percent of U.S. goods exports. By the end of
2017, the share of petroleum in total goods imports declined to 8
percent, while the share in total goods exports almost tripled,
shrinking the U.S. petroleum trade deficit. Had the petroleum trade
deficit not improved, all else unchanged, the total U.S. trade
deficit would likely have been more than 35 percent wider by the end
of 2017.''
 \71\ For an illustration of recent increases in U.S. production,
see, e.g., `U.S. crude oil and liquid fuels production,'' Short-Term
Energy Outlook, U.S. Energy Information Administration (Aug. 2019),
available at http://www.eia.gov/outlooks/steo/images/Fig16.png. EIA
noted in April 2019 that ``Annual U.S. crude oil production reached
a record level of 10.96 million barrels per day (b/d) in 2018, 1.6
million b/d (17%) higher than 2017 levels. In December 2018, monthly
U.S. crude oil production reached 11.96 million b/d, the highest
monthly level of crude oil production in U.S. history. U.S crude oil
production has increased significantly over the past 10 years,
driven mainly by production from tight rock formations using
horizontal drilling and hydraulic fracturing. EIA projects that U.S.
crude oil production will continue to grow in 2019 and 2020,
averaging 12.3 million b/d and 13.0 million b/d, respectively.''
``Today in Energy: U.S. crude oil production grew 17% in 2018,
surpassing the previous record in 1970,'' EIA, Apr. 9, 2019.
Available at http://www.eia.gov/todayinenergy/detail.php?id=38992.
---------------------------------------------------------------------------
 As flagged in the NPRM, some commenters raised concerns about
potential economic consequences for automaker and supplier operations
in the U.S. due to disparities between CAFE standards at home and their
counterpart fuel economy/efficiency and CO2 standards
abroad. NHTSA finds these concerns more relevant to technological
feasibility and economic practicability considerations than to the
national balance of payments. The discussion in Section VIII below
addresses this topic in more detail.
(3) Environmental Implications
 Higher fleet fuel economy can reduce U.S. emissions of various
pollutants by reducing the amount of oil that is produced and refined
for the U.S. vehicle fleet, but can also increase emissions by reducing
the cost of driving, which can result in more vehicle miles traveled
(i.e., the rebound effect). Thus, the net effect of more stringent CAFE
standards on emissions of each pollutant depends on the relative
magnitude of both its reduced emissions in fuel refining and
distribution and increases in its emissions from vehicle use. Fuel
savings from CAFE standards also necessarily results in lower emissions
of CO2, the main greenhouse gas emitted as a result of
refining, distributing, and using transportation fuels. Reducing fuel
consumption directly reduces CO2 emissions because the
primary source of transportation-related CO2 emissions is
fuel combustion in internal combustion engines.
 NHTSA has considered environmental issues, both within the context
of EPCA and the context of the National Environmental Policy Act
(NEPA), in making decisions about the setting of standards since the
earliest days of the CAFE program. As courts of appeal have noted in
three decisions stretching over the last 20 years,\72\ NHTSA defined
``the need of the United States to conserve energy'' in the late 1970s
as including, among other things, environmental implications. In 1988,
NHTSA included climate change concepts in its CAFE notices and prepared
its first environmental assessment addressing that subject.\73\ It
cited concerns about climate change as one of its reasons for limiting
the extent of its reduction of the CAFE standard for MY 1989 passenger
cars.\74\ Since then, NHTSA has considered the effects of reducing
tailpipe emissions of CO2 in its fuel economy rulemakings
pursuant to the need of the United States to conserve energy by
reducing petroleum consumption.
---------------------------------------------------------------------------
 \72\ CAS, 793 F.2d 1322, 1325 n. 12 (D.C. Cir. 1986); Public
Citizen, 848 F.2d 256, 262-63 n. 27 (D.C. Cir 1988) (noting that
``NHTSA itself has interpreted the factors it must consider in
setting CAFE standards as including environmental effects''); CBD,
538 F.3d 1172 (9th Cir. 2007).
 \73\ 53 FR 33080, 33096 (Aug. 29, 1988).
 \74\ 53 FR 39275, 39302 (Oct. 6, 1988).
---------------------------------------------------------------------------
(4) Foreign Policy Implications
 U.S. consumption and imports of petroleum products can impose
additional costs (i.e., externalities) on the domestic economy that are
not reflected in the market price for crude petroleum or in the prices
paid by consumers for petroleum products such as gasoline. NHTSA has
said previously that these costs can include (1) higher prices for
petroleum products resulting from the effect of U.S. oil demand on
world oil prices, (2) the risk of disruptions to the U.S. economy
caused by sudden increases in the global price of oil and its resulting
impact on fuel prices faced by U.S. consumers, and (3) expenses for
maintaining the strategic petroleum reserve (SPR) to provide a response
option should a disruption in commercial oil supplies threaten the U.S.
economy, to allow the U.S. to meet part of its International Energy
Agency obligation to maintain emergency oil stocks, and to provide a
national defense fuel reserve.\75\ Higher U.S. consumption of crude oil
or refined petroleum products increases the magnitude of these external
economic costs, thus increasing the true economic cost of supplying
transportation fuels above the resource costs of producing them.
Conversely, reducing U.S. consumption of crude oil or refined petroleum
products (by reducing motor fuel use) can reduce these external costs.
---------------------------------------------------------------------------
 \75\ While the U.S. maintains a military presence in certain
parts of the world to help secure global access to petroleum
supplies, that is neither the primary nor the sole mission of U.S.
forces overseas. Additionally, the scale of oil consumption
reductions associated with CAFE standards would be insufficient to
alter any existing military missions focused on ensuring the safe
and expedient production and transportation of oil around the globe.
See the FRIA's discussion on energy security for more information on
this topic.
---------------------------------------------------------------------------
 While these costs are considerations, the United States has
significantly increased oil production capabilities in recent years, to
the extent that the U.S. is currently producing enough oil to satisfy
nearly all of its energy needs and is projected to continue to do so
(or even become a net energy exporter in the near future).\76\ This has
added stable new supply to the global oil market, which ameliorates the
U.S.' need to
[[Page 24216]]
conserve energy from a security perspective even given that oil is a
global commodity. The agencies discuss this issue in more detail in
Section VIII below.
---------------------------------------------------------------------------
 \76\ See AEO 2019, at 14 (``In the Reference case, the United
States becomes a net exporter of petroleum liquids after 2020 as
U.S. crude oil production increases and domestic consumption of
petroleum products decreases.''). Available at https://www.eia.gov/outlooks/aeo/pdf/aeo2019.pdf.
---------------------------------------------------------------------------
(2) Factors That NHTSA Is Prohibited From Considering
 EPCA states that in determining the level at which it should set
CAFE standards for a particular model year, NHTSA may not consider the
ability of manufacturers to take advantage of several EPCA provisions
that facilitate compliance with CAFE standards and thereby can reduce
their costs of compliance.\77\ As discussed further below, NHTSA cannot
consider compliance credits that manufacturers earn by exceeding the
CAFE standards and then use to achieve compliance in years in which
their measured average fuel economy falls below the standards. NHTSA
also cannot consider the use of alternative fuels by dual-fueled
vehicles (such as plug-in hybrid electric vehicles) nor the
availability of dedicated alternative fuel vehicles (such as battery
electric or hydrogen fuel cell vehicles) in any model year. EPCA
encourages the production of alternative fuel vehicles by specifying
that their fuel economy is to be determined using a special calculation
procedure that results in those vehicles being assigned a higher fuel
economy level than they actually achieve. For non-statutory incentives
that NHTSA developed by regulation, NHTSA does not consider these
incentives subject to the EPCA prohibition on considering
flexibilities. These topics will be addressed further in Section VIII
below.
---------------------------------------------------------------------------
 \77\ 49 U.S.C. 32902(h).
---------------------------------------------------------------------------
(3) Other Considerations in Determining Maximum Feasible CAFE Standards
 NHTSA historically has interpreted EPCA's statutory factors as
including consideration for potential adverse safety consequences in
setting CAFE standards. Courts have consistently recognized that this
interpretation is reasonable. As courts have recognized, ``NHTSA has
always examined the safety consequences of the CAFE standards in its
overall consideration of relevant factors since its earliest rulemaking
under the CAFE program.'' \78\ The courts have consistently upheld
NHTSA's implementation of EPCA in this manner.\79\ Thus, in evaluating
what levels of stringency would result in maximum feasible standards,
NHTSA assesses the potential safety impacts and considers them in
balancing the statutory considerations and to determine the maximum
feasible level of the standards.\80\ Many commenters addressed the
NPRM's analysis of safety impacts; those comments will be summarized
and responded to in Section VI.D.2 and also in each agency's discussion
in Section VIII.
---------------------------------------------------------------------------
 \78\ Competitive Enterprise Institute v. NHTSA, 901 F.2d 107,
120 n. 11 (D.C. Cir. 1990) (``CEI-I'') (citing 42 FR 33534, 33551
(Jun. 30, 1977).
 \79\ See, e.g., Competitive Enterprise Institute v. NHTSA, 956
F.2d 321, 322 (D.C. Cir. 1992) (``CEI-II'') (in determining the
maximum feasible fuel economy standard, ``NHTSA has always taken
passenger safety into account,'' citing CEI-I, 901 F.2d at 120 n.
11); Competitive Enterprise Institute v. NHTSA, 49 F.3d 481, 483-83
(D.C. Cir. 1995) (same); Center for Biological Diversity v. NHTSA,
538 F.3d 1172, 1203-04 (9th Cir. 2008) (upholding NHTSA's analysis
of vehicle safety issues with weight in connection with the MYs
2008-2011 light truck CAFE rulemaking).
 \80\ NHTSA stated in the NPRM that ``While we discuss safety as
a separate consideration, NHTSA also considers safety as closely
related to, and in some circumstances a subcomponent of, economic
practicability. On a broad level, manufacturers have finite
resources to invest in research and development. Investment into the
development and implementation of fuel saving technology necessarily
comes at the expense of investing in other areas such as safety
technology. On a more direct level, when making decisions on how to
equip vehicles, manufacturers must balance cost considerations to
avoid pricing further consumers out of the market. As manufacturers
add technology to increase fuel efficiency, they may decide against
installing new safety equipment to reduce cost increases. And as the
price of vehicles increase beyond the reach of more consumers, such
consumers continue to drive or purchase older, less safe vehicles.
In assessing practicability, NHTSA also considers the harm to the
nation's economy caused by highway fatalities and injuries.'' 83 FR
at 43209 (Aug. 24, 2018). Many comments were received on this issue,
which will be discussed further in Section VIII below.
---------------------------------------------------------------------------
 The above sections explain what Congress thought was important
enough to codify when it directed each agency to regulate, and begin to
explain how the agencies have interpreted those directions over time
and in this final rule. The next section looks more closely at the
interplay between Congress's direction to the agencies and the aspects
of the market that these regulations affect, as follows.
IV. Purpose of Analytical Approach Considered as Part of Decision-
Making
A. Relationship of Analytical Approach to Governing Law
 Like the NPRM, today's final rule is supported by extensive
analysis of potential impacts of the regulatory alternatives under
consideration. Below, Section VI reviews the analytical approach,
Section VII summarizes the results of the analysis, and Section VIII
explains how the final standards--informed by this analysis--fulfill
the agencies' statutory obligations. Accompanying today's notice, a
final Regulatory Impact Analysis (FRIA) and, for NHTSA's consideration,
a final Environmental Impact Analysis (FEIS), together provide a more
extensive and detailed enumeration of related methods, estimates,
assumptions, and results. The agencies' analysis has been constructed
specifically to reflect various aspects of governing law applicable to
CAFE and CO2 standards, and has been expanded and improved
in response to comments received to the NPRM and based on additional
work by the agencies. The analysis aided the agencies in implementing
their statutory obligations, including the weighing of competing
considerations, by reasonably informing the agencies about the
estimated effects of choosing different regulatory alternatives.
 The agencies' analysis makes use of a range of data (i.e.,
observations of things that have occurred), estimates (i.e., things
that may occur in the future), and models (i.e., methods for making
estimates). Two examples of data include (1) records of actual odometer
readings used to estimate annual mileage accumulation at different
vehicle ages and (2) CAFE compliance data used as the foundation for
the ``analysis fleet'' containing, among other things, production
volumes and fuel economy levels of specific configurations of specific
vehicle models produced for sale in the U.S. Two examples of estimates
include (1) forecasts of future GDP growth used, with other estimates,
to forecast future vehicle sales volumes and (2) the ``retail price
equivalent'' (RPE) factor used to estimate the ultimate cost to
consumers of a given fuel-saving technology, given accompanying
estimates of the technology's ``direct cost,'' as adjusted to account
for estimated ``cost learning effects'' (i.e., the tendency that it
will cost a manufacturer less to apply a technology as the manufacturer
gains more experience doing so).
 The agencies' analysis makes use of several models, some of which
are actually integrated systems of multiple models. As discussed in the
NPRM, the agencies' analysis of CAFE and CO2 standards
involves two basic elements: First, estimating ways each manufacturer
could potentially respond to a given set of standards in a manner that
considers potential consumer response; and second, estimating various
impacts of those responses. Estimating manufacturers' potential
responses involves simulating manufacturers' decision-making processes
regarding the year-by-year application of fuel-saving technologies to
specific vehicles. Estimating impacts involves calculating resultant
changes in new vehicle costs, estimating a
[[Page 24217]]
variety of costs (e.g., for fuel) and effects (e.g., CO2
emissions from fuel combustion) occurring as vehicles are driven over
their lifetimes before eventually being scrapped, and estimating the
monetary value of these effects. Estimating impacts also involves
consideration of the response of consumers--e.g., whether consumers
will purchase the vehicles and in what quantities. Both of these basic
analytical elements involve the application of many analytical inputs.
 The agencies' analysis uses the CAFE Model to estimate
manufacturers' potential responses to new CAFE and CO2
standards and to estimate various impacts of those responses. The model
may be characterized as an integrated system of models. For example,
one model estimates manufacturers' responses, another estimates
resultant changes in total vehicle sales, and still another estimates
resultant changes in fleet turnover (i.e., scrappage). The CAFE model
makes use of many inputs, values of which are developed outside of the
model and not by the model. For example, the model applies fuel prices;
it does not estimate fuel prices. The model does not determine the form
or stringency of the standards; instead, the model applies inputs
specifying the form and stringency of standards to be analyzed and
produces outputs showing effects of manufacturers working to meet those
standards, which become the basis for comparing between different
potential stringencies.
 The agencies also use EPA's MOVES model to estimate ``tailpipe''
(a.k.a. ``vehicle'' or ``downstream'') emission factors for criteria
pollutants,\81\ and use four DOE and DOE-sponsored models to develop
inputs to the CAFE model, including three developed and maintained by
DOE's Argonne National Laboratory. The agencies use the DOE Energy
Information Administration's (EIA's) National Energy Modeling System
(NEMS) to estimate fuel prices,\82\ and use Argonne's Greenhouse gases,
Regulated Emissions, and Energy use in Transportation (GREET) model to
estimate emissions rates from fuel production and distribution
processes.\83\ DOT also sponsored DOE/Argonne to use Argonne's
Autonomie full-vehicle modeling and simulation system to estimate the
fuel economy impacts for roughly a million combinations of technologies
and vehicle types.84 85 Section VI.B.3, below, and the
accompanying final RIA document details of the agencies' use of these
models. In addition, as discussed in the final EIS accompanying today's
notice, DOT relied on a range of climate and photochemical models to
estimate impacts on climate, air quality, and public health. The EIS
discusses and documents the use of these models.
---------------------------------------------------------------------------
 \81\ See https://www.epa.gov/moves. Today's final rule used
version MOVES2014b, available at https://www.epa.gov/moves/latest-version-motor-vehicle-emission-simulator-moves.
 \82\ See https://www.eia.gov/outlooks/aeo/info_nems_archive.php.
Today's final rule uses fuel prices estimated using the Annual
Energy Outlook (AEO) 2019 version of NEMS (see https://www.eia.gov/outlooks/aeo/data/browser/#/?id=3-AEO2019&cases=ref2019&sourcekey=0).
 \83\ Information regarding GREET is available at https://greet.es.anl.gov/index.php. Today's notice uses the 2018 version of
GREET.
 \84\ As part of the Argonne simulation effort, individual
technology combinations simulated in Autonomie were paired with
Argonne's BatPAC model to estimate the battery cost associated with
each technology combination based on characteristics of the
simulated vehicle and its level of electrification. Information
regarding Argonne's BatPAC model is available at http://www.cse.anl.gov/batpac/.
 \85\ In addition, the impact of engine technologies on fuel
consumption, torque, and other metrics was characterized using GT
POWER simulation modeling in combination with other engine modeling
that was conducted by IAV Automotive Engineering, Inc. (IAV). The
engine characterization ``maps'' resulting from this analysis were
used as inputs for the Autonomie full-vehicle simulation modeling.
Information regarding GT Power is available at https://www.gtisoft.com/gt-suite-applications/propulsion-systems/gt-power-engine-simulation-software.
---------------------------------------------------------------------------
 As further explained in the NPRM,\86\ to prepare for analysis
supporting the proposal, DOT expanded the CAFE model to address EPA
statutory and regulatory requirements through a year-by-year simulation
of how manufacturers could comply with EPA's CO2 standards,
including:
---------------------------------------------------------------------------
 \86\ 83 FR 42986, 43003 (Aug. 24, 2018).
---------------------------------------------------------------------------
 Calculation of vehicle models' CO2 emission
rates before and after application of fuel-saving (and, therefore,
CO2-reducing) technologies;
 Calculation of manufacturers' fleet average CO2
emission rates;
 Calculation of manufacturers' fleet average CO2
emission rates under attribute-based CO2 standards;
 Accounting for adjustments to average CO2
emission rates reflecting reduction of air conditioner refrigerant
leakage;
 Accounting for the treatment of alternative fuel vehicles
for CO2 compliance;
 Accounting for production ``multipliers'' for PHEVs, BEVs,
compressed natural gas (CNG) vehicles, and fuel cell vehicles (FCVs);
 Accounting for transfer of CO2 credits between
regulated fleets; and
 Accounting for carried-forward (a.k.a. ``banked'')
CO2 credits, including credits from model years earlier than
modeled explicitly.
 As further discussed in the NPRM, although EPA had previously
developed a vehicle simulation tool (``ALPHA'') and a fleet compliance
model (``OMEGA''), and had applied these in prior actions, having
considered the facts before the Agency in 2018, EPA determined that,
``it is reasonable and appropriate to use DOE/Argonne's model for full-
vehicle simulation, and to use DOT's CAFE model for analysis of
regulatory alternatives.'' \87\
---------------------------------------------------------------------------
 \87\ 83 FR 42986, 43000 (Aug. 24, 2018).
---------------------------------------------------------------------------
 As discussed below and in Section VI.B.3, some commenters--some
citing deliberative EPA staff communications during NPRM development,
and one submitting comments by a former EPA staff member closely
involved in the origination of the above-mentioned OMEGA model--took
strong exception to EPA's decision to rely on DOE/Argonne and DOT-
originated models as the basis for analysis informing EPA's decisions
regarding CO2 standards. Some commenters argued that the EPA
Administrator must consider exclusively models and analysis originating
with EPA staff, and that to do otherwise would be arbitrary and
capricious. As explained below (and as explained in the NPRM), it is
reasonable for the Administrator to consider analysis and information
produced from many sources, including, in this instance, the DOE/
Argonne and DOT models. The Administrator has the discretion to
determine what information reasonably and appropriately informs
decisions regarding emissions standards. Some commenters conflated
models with decisions, suggesting that the former mechanically
determine the latter. The CAA authorizes the EPA Administrator, not a
model, to make decisions about emissions standards, just as EPCA
provides similar authority to the Secretary. Models produce analysis,
the results of which help to inform decisions. However, in making such
decisions, the Administrator may and should consider other relevant
information beyond the outputs of any models--including public
comment--and, in all cases, must exercise judgment in establishing
appropriate standards.
 Some commenters conflated models with inputs and/or with results of
the modeling. All of the models mentioned above rely on inputs,
including not only data (i.e., facts), but also estimates (inputs about
the future are estimates, not data). Given these inputs, the models
produce estimates--ultimately, the agencies' reported estimates of the
potential impacts of standards under
[[Page 24218]]
consideration. In other words, inputs do not define models; models use
inputs. Therefore, disagreements about inputs do not logically extend
to disagreements about models. Similarly, while models determine
resulting outputs, they do so based on inputs. Therefore, disagreements
about results do not necessarily imply disagreements about models; they
may merely reflect disagreements about inputs. With respect to the
Administrator's decisions regarding models underlying today's analysis,
comments regarding inputs, therefore, are more appropriately addressed
separately, which is done so below in Section VI.
 The EPA Administrator's decision to continue relying on the DOE/
Argonne Autonomie tool and DOT CAFE model rather than on the
corresponding tools developed by EPA staff is informed by consideration
of comments on results and on technical aspects of the models
themselves. As discussed below, some commenters questioned specific
aspects of the CAFE model's simulation of manufacturer's potential
responses to CO2 standards. Considering these comments, the
CAFE model applied in the final rule's analysis includes some revisions
and updates. For example, the ``effective cost'' metric used to select
among available opportunities to apply fuel-saving technologies now
uses a ``cost per credit'' metric rather than the metric used for the
NPRM. Also, the model's representation of sales ``multipliers'' EPA has
included for CNG vehicles, PHEVs, BEVs, and FCVs reflects current EPA
regulations or, as an input-selectable option, an alternative approach
under consideration. On the other hand, some commenters questioning the
CAFE model's approach to some CO2 program features appear to
ignore the fact that prior analysis by EPA (using EPA's OMEGA) model
likewise did not account for the same program features. For example,
some stakeholders took issue with the CAFE model's approach to
accounting for banked CO2 credits and, in particular,
credits banked prior to the model years accounted for explicitly in the
analysis. In the course of updating the basis for analysis fleet from
model year 2016 to model year 2017, the agencies have since updated
corresponding inputs. However, even though the ability to carry forward
credits impacts outcomes, EPA's OMEGA model used in previous
rulemakings never attempted to account for credit banking and, indeed,
lacking a year-by-year structure, cannot account for credit banking.
Therefore, at least with respect to this important CO2
program flexibility, the CAFE model provides a more complete and
realistic basis for estimating actual impacts of new CO2
standards.
 For its part, NHTSA remains confident that the combination of the
Autonomie and CAFE models remains the best available for CAFE
rulemaking analysis, and notes, as discussed below, that even the
environmental group coalition stated that the CAFE model is aligned
with EPCA requirements.\88\ In late 2001, after Congress discontinued
an extended series of budget ``riders'' prohibiting work on CAFE
standards, NHTSA and the DOT Volpe Center began development of a
modeling system appropriate for CAFE rulemaking analysis, because other
available models were not designed with this purpose in mind, and
lacked capabilities important for CAFE rulemakings. For example,
although NEMS had procedures to account for CAFE standards, those
procedures did not provide the ability to account for specific
manufacturers, as is especially relevant to the statutory requirement
that NHTSA consider the economic practicability of any new CAFE
standards. Also, as early as the first rulemaking making use of this
early CAFE model, commenters stressed the importance of product
redesign schedules, leading developers to introduce procedures to
account for product cadence. In the 2003 notice regarding light truck
standards for MYs 2005-2007, NHTSA stated that ``we also changed the
methodology to recognize that capital costs require employment of
technologies for several years, rather than a single year. . . . In our
view, this makes the Volpe analysis more consistent with the [manually
implemented] Stage analysis and better reflects actual conditions in
the automotive industry.'' \89\ Since that time, NHTSA and the Volpe
Center have significantly refined the CAFE model with each of
rulemaking. For example, for the 2006 rulemaking regarding standards
for MYs 2008-2011 light trucks, NHTSA introduced the ability to account
for attribute-based standards, account for the social cost of
CO2 emissions, estimate stringencies at which net benefits
would be maximized, and perform probabilistic uncertainty analysis
(i.e., Monte Carlo simulation).\90\ For the 2009 rulemaking regarding
standards for MY 2011 passenger cars and light trucks, we introduced
the ability to account for attribute-based passenger car standards, and
the ability to apply ``synergy factors'' to estimate how some
technology pairings impact fuel consumption,\91\ For the 2010
rulemaking regarding standards for MYs 2012-2016, we introduced
procedures to account for FFV credits, and to account for product
planning as a multiyear consideration.\92\ For the 2012 rulemaking
regarding standards for MYs 2017-2025, we introduced several new
procedures, such as (1) accounting for electricity used to charge
electric vehicles (EVs) and plug-in hybrid electric vehicles (PHEVs),
(2) accounting for use of ethanol blends in flexible-fuel vehicles
(FFVs), (3) accounting for costs (i.e., ``stranded capital'') related
to early replacement of technologies, (4) accounting for previously-
applied technology when determining the extent to which a manufacturer
could expand use of the technology, (5) applying technology-specific
estimates of changes in consumer value, (6) simulating the extent to
which manufacturers might utilize EPCA's provisions regarding
generation and use of CAFE credits, (7) applying estimates of fuel
economy adjustments (and accompanying costs) reflecting increases in
air conditioner efficiency, (8) reporting privately-valued benefits,
(9) simulating the extent to which manufacturers might voluntarily
apply technology beyond levels needed for compliance with CAFE
standards, and (10) estimating changes in highway fatalities
attributable to any applied reductions in vehicle mass.\93\ Also for
the 2012 rulemaking, we began making use of Autonomie to estimate fuel
consumption impacts of different combinations of technologies, using
these estimates to specify inputs to the CAFE model.\94\ In 2016,
providing analyses for both the draft TAR regarding light-duty CAFE
standards and the final rule regarding fuel consumption standards for
heavy-duty pickup trucks and vans, we greatly expanded the agency's use
of Autonomie-based full vehicle simulations and introduced the ability
to simulate compliance with attribute-based standards for heavy-duty
pickups and vans.\95\ And, as discussed at length in the NPRM and
below, for this rulemaking, we have, among other things, refined
procedures to account for impacts on highway travel and safety,
[[Page 24219]]
added procedures to simulate compliance with CO2 standards,
refined procedures to account for compliance credits, and added
procedures to account for impacts on sales, scrappage, and employment.
We have also significantly revised the model's graphical user interface
(GUI) in order to make the model easier to operate and understand. Like
any model, both Autonomie and the CAFE model benefit from ongoing
refinement. However, NHTSA is confident that this combination of models
produces a more realistic characterization of the potential impacts of
new standards than would another combination of available models. Some
stakeholders, while commenting on specific aspects of the inputs,
models, and/or results, commended the agencies' exclusive reliance on
the DOE/Argonne Autonomie tool and DOT CAFE model. With respect to
CO2 standards, these stakeholders noted not only technical
reasons to use these models rather than the EPA models, but also other
reasons such as efficiency, transparency, and ease with which outside
parties can exercise models and replicate the agencies' analysis. These
comments are discussed below and in Section VI.
---------------------------------------------------------------------------
 \88\ Environmental group coalition, NHTSA-2018-0067-12000,
Appendix A, at 24-25.
 \89\ 68 FR at 16885 (Apr. 7, 2003).
 \90\ 71 FR at 17566 et seq. (Apr. 6, 2006).
 \91\ 74 FR at 14196 et seq. (Mar. 30, 3009).
 \92\ 75 FR at 25599 et seq. (May 7, 2010).
 \93\ 77 FR 63009 et seq. (Oct. 15, 2012).
 \94\ 77 FR at 62712 et seq. (Oct. 15, 2012).
 \95\ 81 FR at 73743 et seq. (Oct. 25, 2016); Draft TAR,
available at Docket No. NHTSA-2016-0068-0001, Chapter 13.
---------------------------------------------------------------------------
 Nevertheless, some comments regarding the model's handling of CAFE
and/or CO2 standards, and some comments regarding the
model's estimation of resultant impacts, led the agencies to make
changes to specific aspects of the model. Comments on and changes to
the inputs and model are discussed below and in Section VI; results are
discussed in Section VII and in the accompanying RIA; and the meaning
of results in the context of the applicable statutory requirements is
discussed in Section VIII.
 As explained, the analysis is designed to reflect a number of
statutory and regulatory requirements applicable to CAFE and tailpipe
CO2 standard setting. EPCA contains a number of requirements
governing the scope and nature of CAFE standard setting. Among these,
some have been in place since EPCA was first signed into law in 1975,
and some were added in 2007, when Congress passed EISA and amended
EPCA. The CAA, as discussed elsewhere, provides EPA with very broad
authority under Section 202(a), and does not contain EPCA/EISA's
prescriptions. In the interest of harmonization, however, EPA has
adopted some of the EPCA/EISA requirements into its tailpipe
CO2 regulations, and NHTSA, in turn, has created some
additional flexibilities by regulation not expressly envisioned by
EPCA/EISA in order to harmonize better with some of EPA's programmatic
decisions. EPCA/EISA requirements regarding the technical
characteristics of CAFE standards and the analysis thereof include, but
are not limited to, the following, and the analysis reflects these
requirements as summarized:
 Corporate Average Standards: 49 U.S.C. 32902 requires standards
that apply to the average fuel economy levels achieved by each
corporation's fleets of vehicles produced for sale in the U.S.\96\ CAA
Section 202(a) does not preclude the EPA Administrator from expressing
CO2 standards as de facto fleet average requirements, and
EPA has adopted a similar approach in the interest of harmonization.
The CAFE Model, used by the agencies to conduct the bulk of today's
analysis, calculates the CAFE and CO2 levels of each
manufacturer's fleets based on estimated production volumes and
characteristics, including fuel economy levels, of distinct vehicle
models that could be produced for sale in the U.S.
---------------------------------------------------------------------------
 \96\ This differs from safety standards and traditional
emissions standards, which apply separately to each vehicle. For
example, every vehicle produced for sale in the U.S. must, on its
own, meet all applicable federal motor vehicle safety standards
(FMVSS), but no vehicle produced for sale must, on its own, federal
fuel economy standards. Rather, each manufacturer is required to
produce a mix of vehicles that, taken together, achieve an average
fuel economy level no less than the applicable minimum level.
---------------------------------------------------------------------------
 Separate Standards for Passenger Cars and Light Trucks: 49 U.S.C.
32902 requires the Secretary of Transportation to set CAFE standards
separately for passenger cars and light trucks. CAA Section 202(a) does
not preclude the EPA Administrator from specifying CO2
standards separately for passenger cars and light trucks, and EPA has
adopted a similar approach. The CAFE Model accounts separately for
passenger cars and light trucks, including differentiated standards and
compliance.
 Attribute-Based Standards: 49 U.S.C. 32902 requires the Secretary
of Transportation to define CAFE standards as mathematical functions
expressed in terms of one or more vehicle attributes related to fuel
economy. This means that for a given manufacturer's fleet of vehicles
produced for sale in the U.S. in a given regulatory class and model
year, the applicable minimum CAFE requirement (i.e., the numerical
value of the requirement) is computed based on the applicable
mathematical function, and the mix and attributes of vehicles in the
manufacturer's fleet. In the 2012 final rule that first established
CO2 standards, EPA also adopted an attribute-based standard
under its broad CAA Section 202(a) authority. The CAFE Model accounts
for such functions and vehicle attributes explicitly.
 Separately Defined Standards for Each Model Year: 49 U.S.C. 32902
requires the Secretary to set CAFE standards (separately for passenger
cars and light trucks) at the maximum feasible levels in each model
year. CAA Section 202(a) allows EPA to establish CO2
standards separately for each model year, and EPA has chosen to do so
for this final rule, similar to the approach taken in the previous
light-duty vehicle CO2 standard-setting rules. The CAFE
Model represents each model year explicitly, and accounts for the
production relationships between model years.\97\
---------------------------------------------------------------------------
 \97\ For example, a new engine first applied to given vehicle
model/configuration in model year 2020 will most likely be ``carried
forward'' to model year 2021 of that same vehicle model/
configuration, in order to reflect the fact that manufacturers do
not apply brand-new engines to a given vehicle model every single
year.
---------------------------------------------------------------------------
 Separate Compliance for Domestic and Imported Passenger Car Fleets:
49 U.S.C. 32904 requires the EPA Administrator to determine CAFE
compliance separately for each manufacturers' fleets of domestic
passenger cars and imported passenger cars, which manufacturers must
consider as they decide how to improve the fuel economy of their
passenger car fleets. CAA 202(a) does not preclude the EPA
Administrator from determining compliance with CO2 standards
separately for a manufacturer's domestic and imported car fleets, but
EPA did not include such a distinction in either the 2010 or 2012 final
rules, and EPA did not propose or ask for comment on taking such an
approach in the proposal. The CAFE Model is able to account explicitly
for this requirement when simulating manufacturers' potential responses
to CAFE standards, but combines any given manufacturer's domestic and
imported cars into a single fleet when simulating that manufacturer's
potential response to CO2 standards.
 Minimum CAFE Standards for Domestic Passenger Car Fleets: 49 U.S.C.
32902 requires that domestic passenger car fleets achieve CAFE levels
no less than 92 percent of the industry-wide average level required
under the applicable attribute-based CAFE standard, as projected by the
Secretary at the time the standard is promulgated. CAA 202(a) does not
preclude the EPA Administrator from correspondingly requiring that
domestic passenger car fleets achieve CO2 levels no greater
than 108.7 percent (1/0.92 = 1.087) of the projected industry-wide
average CO2
[[Page 24220]]
requirement under the attribute-based standard, but the GHG program
that EPA designed in the 2010 and 2012 final rules did not include such
a distinction, and EPA did not propose or seek comment on such an
approach in the proposal. The CAFE Model is able to account explicitly
for this requirement for CAFE standards, and sets this requirement
aside for CO2 standards.
 Civil Penalties for Noncompliance: 49 U.S.C. 32912 prescribes a
rate (in dollars per tenth of a mpg) at which the Secretary is to levy
civil penalties if a manufacturer fails to comply with a CAFE standard
for a given fleet in a given model year, after considering available
credits. Some manufacturers have historically demonstrated a
willingness to treat CAFE noncompliance as an ``economic'' choice,
electing to pay civil penalties rather than achieving full numerical
compliance across all fleets. The CAFE Model calculates civil penalties
for CAFE shortfalls and provides means to estimate that a manufacturer
might stop adding fuel-saving technologies once continuing to do so
would be effectively more ``expensive'' (after accounting for fuel
prices and buyers' willingness to pay for fuel economy) than paying
civil penalties. In contrast, the CAA does not authorize the EPA
Administrator to allow manufacturers to sell noncompliant fleets and
instead only pay civil penalties; manufacturers who choose to pay civil
penalties for CAFE compliance tend to employ EPA's more-extensive
programmatic flexibilities to meet tailpipe CO2 emissions
standards. Thus, the CAFE Model does not allow civil penalty payment as
an option for CO2 standards.
 Dual-Fueled and Dedicated Alternative Fuel Vehicles: For purposes
of calculating CAFE levels used to determine compliance, 49 U.S.C.
32905 and 32906 specify methods for calculating the fuel economy levels
of vehicles operating on alternative fuels to gasoline or diesel
through MY 2020. After MY 2020, methods for calculating alternative
fuel vehicle (AFV) fuel economy are governed by regulation. The CAFE
Model is able to account for these requirements explicitly for each
vehicle model. However, 49 U.S.C. 32902 requires that maximum feasible
CAFE standards be set in a manner that does not presume manufacturers
can respond by producing new dedicated alternative fuel vehicle (AFV)
models. The CAFE model can be run in a manner that excludes the
additional application of dedicated AFV technologies in model years for
which maximum feasible standards are under consideration. As allowed
under NEPA for analysis appearing in EISs informing decisions regarding
CAFE standards, the CAFE Model can also be run without this analytical
constraint. CAA 202(a) does not preclude the EPA Administrator adopting
analogous provisions, but EPA has instead opted through regulation to
``count'' dual- and alternative fuel vehicles on a CO2 basis
(and through MY 2026, to set aside emissions from electricity
generation). The CAFE model accounts for this treatment of dual- and
alternative fuel vehicles when simulating manufacturers' potential
responses to CO2 standards. For natural gas vehicles, both
dedicated and dual-fueled, EPA is establishing a multiplier of 2.0 for
model years 2022-2026.
 Creation and Use of Compliance Credits: 49 U.S.C. 32903 provides
that manufacturers may earn CAFE ``credits'' by achieving a CAFE level
beyond that required of a given fleet in a given model year, and
specifies how these credits may be used to offset the amount by which a
different fleet falls short of its corresponding requirement. These
provisions allow credits to be ``carried forward'' and ``carried back''
between model years, transferred between regulated classes (domestic
passenger cars, imported passenger cars, and light trucks), and traded
between manufacturers. However, these provisions also impose some
specific statutory limits. For example, CAFE compliance credits can be
carried forward a maximum of five model years and carried back a
maximum of three model years. Also, EPCA/EISA caps the amount of credit
that can be transferred between passenger car and light truck fleets,
and prohibits manufacturers from applying traded or transferred credits
to offset a failure to achieve the applicable minimum standard for
domestic passenger cars. The CAFE Model explicitly simulates
manufacturers' potential use of credits carried forward from prior
model years or transferred from other fleets.\98\ 49 U.S.C. 32902
prohibits consideration of manufacturers' potential application of CAFE
compliance credits when setting maximum feasible CAFE standards. The
CAFE Model can be operated in a manner that excludes the application of
CAFE credits after a given model year. CAA 202(a) does not preclude the
EPA Administrator adopting analogous provisions. EPA has opted to limit
the ``life'' of compliance credits from most model years to 5 years,
and to limit borrowing to 3 years, but has not adopted any limits on
transfers (between fleets) or trades (between manufacturers) of
compliance credits. The CAFE Model is able to account for the absence
of limits on transfers of CO2 standards. Insofar as the CAFE
model can be exercised in a manner that simulates trading of
CO2 compliance credits, such simulations treat trading as
unlimited.\99\ EPA has considered manufacturers' ability to use credits
as part of its decisions on these final standards, and the CAFE model
is now able to account for that.
---------------------------------------------------------------------------
 \98\ As explained in Section VI, the CAFE Model does not
explicitly simulate the potential that manufacturers would carry
CAFE or CO2 credits back (i.e., borrow) from future model
years, or acquire and use CAFE compliance credits from other
manufacturers. At the same time, because EPA has elected to not
limit credit trading, the CAFE Model can be exercised in a manner
that simulates unlimited (a.k.a. ``perfect'') CO2
compliance credit trading throughout the industry (or, potentially,
within discrete trading ``blocs''). The agencies believe there is
significant uncertainty in how manufacturers may choose to employ
these particular flexibilities in the future: for example, while it
is reasonably foreseeable that a manufacturer who over-complies in
one year may ``coast'' through several subsequent years relying on
those credits rather than continuing to make technology
improvements, it is harder to assume with confidence that
manufacturers will rely on future technology investments (that may
not pan out as expected, as if market demand for ``target-beater''
vehicles is lower than expected) to offset prior-year shortfalls, or
whether/how manufacturers will trade credits with market competitors
rather than making their own technology investments. Historically,
carry-back and trading have been much less utilized than carry-
forward, for a variety of reasons including higher risk and
preference not to ``pay competitors to make fuel economy
improvements we should be making'' (to paraphrase one manufacturer),
although the agencies recognize that carry-back and trading are used
more frequently when standards require more technology application
than manufacturers believe their markets will bear. Given the
uncertainty just discussed, and given also the fact that the
agencies have yet to resolve some of analytical challenges
associated with simulating use of these flexibilities, the agencies
consider borrowing and trading to involve sufficient risk that it is
prudent to support today's decisions with analysis that sets aside
the potential that manufacturers could come to depend widely on
borrowing and trading. While compliance costs in real life may be
somewhat different from what is modeled today as a result of this
analytical decision, that is broadly true no matter what, and the
agencies do not believe that the difference would be so great that
it would change the policy outcome.
 \99\ To avoid making judgments (that would invariably turn out
to be at least somewhat incorrect) about possible future trading
activity, the model simulates trading by combining all manufacturers
into a single entity, so that the most cost-effective choices are
made for the fleet as a whole.
---------------------------------------------------------------------------
 Statutory Basis for Stringency: 49 U.S.C. 32902 requires the
Secretary to set CAFE standards at the maximum feasible levels,
considering technological feasibility, economic practicability, the
need of the Nation to conserve energy, and the impact of other
government standards. EPCA/EISA authorizes the Secretary to interpret
[[Page 24221]]
these factors, and as the Department's interpretation has evolved,
NHTSA has continued to expand and refine its qualitative and
quantitative analysis. For example, as discussed below in Section
VI.B.3, the Autonomie simulations reflect the agencies' judgment that
it would not be economically practicable for a manufacturer to
``split'' an engine shared among many vehicle model/configurations into
a myriad of versions each optimized to a single vehicle model/
configuration. Also responding to evolving interpretation of these
EPCA/EISA factors, the CAFE Model has been expanded to address
additional impacts in an integrated manner. For example, the CAFE Model
version used for the NPRM analysis included the ability to estimate
impacts on labor utilization internally, rather than as an external
``off model'' or ``post processing'' analysis. In addition, NEPA
requires the Secretary to issue an EIS that documents the estimated
impacts of regulatory alternatives under consideration. The EIS
accompanying today's notice documents changes in emission inventories
as estimated using the CAFE model, but also documents corresponding
estimates--based on the application of other models documented in the
EIS, of impacts on the global climate, on tropospheric air quality, and
on human health. Regarding CO2 standards, CAA 202(a)
provides general authority for the establishment of motor vehicle
emissions standards, and the final rule's analysis, like that
accompanying the agencies' proposal, addresses impacts relevant to the
EPA Administrator's decision making, such as technological feasibility,
air quality impacts, costs to industry and consumers, and lead time
necessary for compliance.
 Other Factors: Beyond these statutory requirements applicable to
DOT and/or EPA are a number of specific technical characteristics of
CAFE and/or CO2 regulations that are also relevant to the
construction of today's analysis. These are discussed at greater length
in Section II.F. For example, EPA has defined procedures for
calculating average CO2 levels, and has revised procedures
for calculating CAFE levels, to reflect manufacturers' application of
``off-cycle'' technologies that increase fuel economy (and reduce
CO2 emissions) in ways not reflected by the long-standing
test procedures used to measure fuel economy. Although too little
information is available to account for these provisions explicitly in
the same way that the agencies have accounted for other technologies,
the CAFE Model does include and makes use of inputs reflecting the
agencies' expectations regarding the extent to which manufacturers may
earn such credits, along with estimates of corresponding costs.
Similarly, the CAFE Model includes and makes use of inputs regarding
credits EPA has elected to allow manufacturers to earn toward
CO2 levels (not CAFE) based on the use of air conditioner
refrigerants with lower global warming potential (GWP), or on the
application of technologies to reduce refrigerant leakage. In addition,
EPA has elected to provide that through model year 2021, manufacturers
may apply ``multipliers'' to plug-in hybrid electric vehicles,
dedicated electric vehicles, fuel cell vehicles, and hydrogen vehicles,
such that when calculating a fleet's average CO2 levels (not
CAFE), the manufacturer may, for example, ``count'' each electric
vehicle twice. The CAFE Model accounts for these multipliers, based on
either current regulatory provisions or on alternative approaches.
Although these are examples of regulatory provisions that arise from
the exercise of discretion rather than specific statutory mandate, they
can materially impact outcomes. Section VI.B explains in greater detail
how today's analysis addresses them.
Benefits of Analytical Approach
 The agencies' analysis of CAFE and CO2 standards
involves two basic elements: First, estimating ways each manufacturer
could potentially respond to a given set of standards in a manner that
considers potential consumer response; and second, estimating various
impacts of those responses. Estimating manufacturers' potential
responses involves simulating manufacturers' decision-making processes
regarding the year-by-year application of fuel-saving technologies to
specific vehicles. Estimating impacts involves calculating resultant
changes in new vehicle costs, estimating a variety of costs (e.g., for
fuel) and effects (e.g., CO2 emissions from fuel combustion)
occurring as vehicles are driven over their lifetimes before eventually
being scrapped, and estimating the monetary value of these effects.
Estimating impacts also involves consideration of the response of
consumers--e.g., whether consumers will purchase the vehicles and in
what quantities. Both of these basic analytical elements involve the
application of many analytical inputs.
 As mentioned above, the agencies' analysis uses the CAFE model to
estimate manufacturers' potential responses to new CAFE and
CO2 standards and to estimate various impacts of those
responses. DOT's Volpe National Transportation Systems Center (often
simply referred to as the ``Volpe Center'') develops, maintains, and
applies the model for NHTSA. NHTSA has used the CAFE model to perform
analyses supporting every CAFE rulemaking since 2001, and the 2016
rulemaking regarding heavy-duty pickup and van fuel consumption and
CO2 emissions also used the CAFE model for analysis.\100\
---------------------------------------------------------------------------
 \100\ While both agencies used the CAFE Model to simulate
manufacturers' potential responses to standards, some model inputs
differed EPA's and DOT's analyses, and EPA also used the EPA MOVES
model to calculate resultant changes in emissions inventories. See
81 FR 73478, 73743 (Oct. 25, 2016).
---------------------------------------------------------------------------
 NHTSA recently arranged for a formal peer review of the model. In
general, reviewers' comments strongly supported the model's conceptual
basis and implementation, and commenters provided several specific
recommendations. The agency agreed with many of these recommendations
and has worked to implement them wherever practicable. Implementing
some of the recommendations would require considerable further
research, development, and testing, and will be considered going
forward. For a handful of other recommendations, the agency disagreed,
often finding the recommendations involved considerations (e.g., other
policies, such as those involving fuel taxation) beyond the model
itself or were based on concerns with inputs rather than how the model
itself functioned. A report available in the docket for this rulemaking
presents peer reviewers' detailed comments and recommendations, and
provides DOT's detailed responses.\101\
---------------------------------------------------------------------------
 \101\ Docket No. NHTSA-2018-0067-0055.
---------------------------------------------------------------------------
 As also mentioned above, the agencies use EPA's MOVES model to
estimate tailpipe emission factors, use DOE/EIA's NEMS to estimate fuel
prices,\102\ and use Argonne's GREET model to estimate downstream
emissions rates.\103\ DOT also sponsored DOE/Argonne to use the
Autonomie full-vehicle modeling and simulation tool to estimate the
fuel economy impacts for roughly a million
[[Page 24222]]
combinations of technologies and vehicle types.104 105
---------------------------------------------------------------------------
 \102\ See https://www.eia.gov/outlooks/aeo/info_nems_archive.php. Today's notice uses fuel prices estimated
using the Annual Energy Outlook (AEO) 2019 version of NEMS (see
https://www.eia.gov/outlooks/archive/aeo19/ and https://www.eia.gov/outlooks/aeo/data/browser/#/?id=3-AEO2019&cases=ref2019&sourcekey=0).
 \103\ Information regarding GREET is available at https://greet.es.anl.gov/index.php. Availability of NEMS is discussed at
https://www.eia.gov/outlooks/aeo/info_nems_archive.php. Today's
notice uses fuel prices estimated using the AEO 2019 version of
NEMS.
 \104\ As part of the Argonne simulation effort, individual
technology combinations simulated in Autonomie were paired with
Argonne's BatPAC model to estimate the battery cost associated with
each technology combination based on characteristics of the
simulated vehicle and its level of electrification. Information
regarding Argonne's BatPAC model is available at http://www.cse.anl.gov/batpac/.
 \105\ Furthermore, the impact of engine technologies on fuel
consumption, torque, and other metrics was characterized using GT
POWER simulation modeling in combination with other engine modeling
that was conducted by IAV Automotive Engineering, Inc. (IAV). The
engine characterization ``maps'' resulting from this analysis were
used as inputs for the Autonomie full-vehicle simulation modeling.
Information regarding GT Power is available at https://www.gtisoft.com/gt-suite-applications/propulsion-systems/gt-power-engine-simulation-software.
---------------------------------------------------------------------------
 EPA developed two models after 2009, referred to as the ``ALPHA''
and ``OMEGA'' models, which provide some of the same capabilities as
the Autonomie and CAFE models. EPA applied the OMEGA model to conduct
analysis of tailpipe CO2 emissions standards promulgated in
2010 and 2012, and the ALPHA and OMEGA models to conduct analysis
discussed in the above-mentioned 2016 Draft TAR and Proposed and 2017
Initial Final Determinations regarding standards beyond 2021. In an
August 2017 notice, the agencies requested comments on, among other
things, whether EPA should use alternative methodologies and modeling,
including DOE/Argonne's Autonomie full-vehicle modeling and simulation
tool and DOT's CAFE model.\106\
---------------------------------------------------------------------------
 \106\ 82 FR 39551, 39553 (Aug. 21, 2017).
---------------------------------------------------------------------------
 Having reviewed comments on the subject and having considered the
matter fully, the agencies have determined it is reasonable and
appropriate to use DOE/Argonne's model for full-vehicle simulation, and
to use DOT's CAFE model for analysis of regulatory alternatives. EPA
interprets Section 202(a) of the CAA as giving the agency broad
discretion in how it develops and sets CO2 emissions
standards for light-duty vehicles. Nothing in Section 202(a) mandates
that EPA use any specific model or set of models for analysis of
potential CO2 standards for light-duty vehicles. EPA weighs
many factors when determining appropriate levels for CO2
standards, including the cost of compliance (see Section 202(a)(2)),
lead time necessary for compliance (id.), safety (see NRDC v. EPA, 655
F.2d 318, 336 n. 31 (D.C. Cir. 1981)) and other impacts on
consumers,\107\ and energy impacts associated with use of the
technology.\108\ Using the CAFE model allows consideration of a number
of factors. The CAFE model explicitly evaluates the cost of compliance
for each manufacturer, each fleet, and each model year; it accounts for
lead time necessary for compliance by directly incorporating estimated
manufacturer production cycles for every vehicle in the fleet, ensuring
that the analysis does not assume vehicles can be redesigned to
incorporate more technology without regard to lead time considerations;
it provides information on safety effects associated with different
levels of standards and information about many other impacts on
consumers, and it calculates energy impacts (i.e., fuel saved or
consumed) as a primary function, besides being capable of providing
information about many other factors within EPA's broad CAA discretion
to consider.
---------------------------------------------------------------------------
 \107\ Since its earliest Title II regulations, EPA has
considered the safety of pollution control technologies. See 45 FR
14496, 14503 (1980).
 \108\ See George E. Warren Corp. v. EPA, 159 F.3d 616, 623-624
(D.C. Cir. 1998) (ordinarily permissible for EPA to consider factors
not specifically enumerated in the Act).
---------------------------------------------------------------------------
 Because the CAFE model simulates a wide range of actual constraints
and practices related to automotive engineering, planning, and
production, such as common vehicle platforms, sharing of engines among
different vehicle models, and timing of major vehicle redesigns, the
analysis produced by the CAFE model provides a transparent and
realistic basis to show pathways manufacturers could follow over time
in applying new technologies, which helps better assess impacts of
potential future standards. Furthermore, because the CAFE model also
accounts fully for regulatory compliance provisions (now including
CO2 compliance provisions), such as adjustments for reduced
refrigerant leakage, production ``multipliers'' for some specific types
of vehicles (e.g., PHEVs), and carried-forward (i.e., banked) credits,
the CAFE model provides a transparent and realistic basis to estimate
how such technologies might be applied over time in response to CAFE or
CO2 standards.
 There are sound reasons for the agencies to use the CAFE model
going forward in this rulemaking. First, the CAFE and CO2
fact analyses are inextricably linked. Furthermore, the analysis
produced by the CAFE model and DOE/Argonne's Autonomie addresses the
agencies' analytical needs. The CAFE model provides an explicit year-
by-year simulation of manufacturers' application of technology to their
products in response to a year-by-year progression of CAFE standards
and accounts for sharing of technologies and the implications for
timing, scope, and limits on the potential to optimize powertrains for
fuel economy. In the real world, standards actually are specified on a
year-by-year basis, not simply some single year well into the future,
and manufacturers' year-by-year plans involve some vehicles ``carrying
forward'' technology from prior model years and some other vehicles
possibly applying ``extra'' technology in anticipation of standards in
ensuing model years, and manufacturers' planning also involves applying
credits carried forward between model years. Furthermore, manufacturers
cannot optimize the powertrain for fuel economy on every vehicle model
configuration--for example, a given engine shared among multiple
vehicle models cannot practicably be split into different versions for
each configuration of each model, each with a slightly different
displacement. The CAFE model is designed to account for these real-
world factors.
 Considering the technological heterogeneity of manufacturers'
current product offerings, and the wide range of ways in which the many
fuel economy-improving/CO2 emissions-reducing technologies
included in the analysis can be combined, the CAFE model has been
designed to use inputs that provide an estimate of the fuel economy
achieved for many tens of thousands of different potential combinations
of fuel-saving technologies. Across the range of technology classes
encompassed by the analysis fleet, today's analysis involves more than
a million such estimates. While the CAFE model requires no specific
approach to developing these inputs, the National Academy of Sciences
(NAS) has recommended, and stakeholders have commented, that full-
vehicle simulation provides the best balance between realism and
practicality. DOE/Argonne has spent several years developing, applying,
and expanding means to use distributed computing to exercise its
Autonomie full-vehicle modeling and simulation tool over the scale
necessary for realistic analysis of CAFE or average tailpipe
CO2 emissions standards. This scalability and related
flexibility (in terms of expanding the set of technologies to be
simulated) makes Autonomie well-suited for developing inputs to the
CAFE model.
 In addition, DOE/Argonne's Autonomie also has a long history of
development and widespread application by a much wider range of users
in government, academia, and industry. Many of these users apply
[[Page 24223]]
Autonomie to inform funding and design decisions. These real-world
exercises have contributed significantly to aspects of Autonomie
important to producing realistic estimates of fuel economy levels and
CO2 emission rates, such as estimation and consideration of
performance, utility, and driveability metrics (e.g., towing
capability, shift business, frequency of engine on/off transitions).
This steadily increasing realism has, in turn, steadily increased
confidence in the appropriateness of using Autonomie to make
significant investment decisions. Notably, DOE uses Autonomie for
analysis supporting budget priorities and plans for programs managed by
its Vehicle Technologies Office (VTO). Considering the advantages of
DOE/Argonne's Autonomie model, it is reasonable and appropriate to use
Autonomie to estimate fuel economy levels and CO2 emission
rates for different combinations of technologies as applied to
different types of vehicles.
 Commenters have also suggested that the CAFE model's graphical user
interface (GUI) facilitates others' ability to use the model quickly--
and without specialized knowledge or training--and to comment
accordingly.\109\ For the NPRM, NHTSA significantly expanded and
refined this GUI, providing the ability to observe the model's real-
time progress much more closely as it simulates year-by-year compliance
with either CAFE or CO2 standards.\110\ Although the model's
ability to produce realistic results is independent of the model's GUI,
the CAFE model's GUI appears to have facilitated stakeholders'
meaningful review and comment during the comment period.
---------------------------------------------------------------------------
 \109\ From Docket Number EPA-HQ-OAR-2015-0827, see Comment by
Global Automakers, Docket ID EPA-HQ-OAR-2015-0827-9728, at 34.
 \110\ The updated GUI provides a range of graphs updated in real
time as the model operates. These graphs can be used to monitor fuel
economy or CO2 ratings of vehicles in manufacturers'
fleets and to monitor year-by-year CAFE (or average CO2
ratings), costs, avoided fuel outlays, and avoided CO2-
related damages for specific manufacturers and/or specific fleets
(e.g., domestic passenger car, light truck). Because these graphs
update as the model progresses, they should greatly increase users'
understanding of the model's approach to considerations such as
multiyear planning, payment of civil penalties, and credit use.
---------------------------------------------------------------------------
 The question of whether EPA's actions should consider and be
informed by analysis using non-EPA-staff-developed modeling tools has
generated considerable debate over time. Even prior to the NPRM,
certain commenters had argued that EPA could not consider, in setting
tailpipe CO2 emissions standards, any information derived
from non-EPA-staff-developed modeling. Many of the pre-NPRM concerns
focused on inputs used by the CAFE model for prior rulemaking
analyses.111 112 113 Because inputs are exogenous to any
model, they do not determine whether it would be reasonable and
appropriate for EPA to use NHTSA's model for analysis. Other concerns
focused on certain characteristics of the CAFE model that were
developed to align the model better with EPCA and EISA. The model has
been revised to accommodate both EPCA/EISA and CAA analysis, as
explained further below. Some commenters also argued that use of any
models other than ALPHA and OMEGA for CAA analysis would constitute an
arbitrary and capricious delegation of EPA's decision-making authority
to NHTSA, if NHTSA models are used for analysis instead.\114\ As
discussed above, the CAFE Model--as with any model--is used to provide
analysis, and does not result in decisions. Decisions are made by EPA
in a manner that is informed by modeling outputs, sensitivity cases,
public comments, any many other pieces of information.
---------------------------------------------------------------------------
 \111\ For example, EDF previously stated that ``the data that
NHTSA needs to input into its model is sensitive confidential
business information that is not transparent and cannot be
independently verified, . . .'' and it claimed ``the OMEGA model's
focus on direct technological inputs and costs--as opposed to
industry self-reported data--ensures the model more accurately
characterizes the true feasibility and cost effectiveness of
deploying greenhouse gas reducing technologies.'' EDF, EPA-HQ-OAR-
2015-0827-9203, at 12. These statements are not correct, as nothing
about either the CAFE or OMEGA model either obviates or necessitates
the use of CBI to develop inputs.
 \112\ As another example, CARB previously stated that ``another
promising technology entering the market was not even included in
the NHTSA compliance modeling'' and that EPA assumes a five-year
redesign cycle, whereas NHTSA assumes a six to seven-year cycle.''
CARB, EPA-HQ-OAR-2015-0827-9197, at 28. Though presented as
criticisms of the models, these comments--at least with respect to
the CAFE model--actually concern model inputs. NHTSA did not agree
with CARB about the commercialization potential of the engine
technology in question (``Atkinson 2'') and applied model inputs
accordingly. Also, rather than applying a one-size-fits-all
assumption regarding redesign cadence, NHTSA developed estimates
specific to each vehicle model and applied these as model inputs.
 \113\ As another example, NRDC has argued that EPA should not
use the CAFE model because it ``allows manufacturers to pay civil
penalties in lieu of meeting the standards, an alternative
compliance pathway currently allowed under EISA and EPCA.'' NRDC,
EPA-HQ-OAR-2015-0827-9826, at 37. While the CAFE model can simulate
civil penalty payment, NRDC's comment appears to overlook the fact
that this result depends on model inputs; the inputs can easily be
specified such that the CAFE model will set aside civil penalty
payment as an alternative to compliance.
 \114\ See, e.g., CBD et al., NHTSA-2018-0067-12057, at 9.
---------------------------------------------------------------------------
 Comments responding to the NPRM's use of the CAFE model and
Autonomie rather than also (for CO2 standards) ALPHA and
OMEGA were mixed. For example, the environmental group coalition stated
that the CAFE model is aligned with EPCA requirements,\115\ but also
argued (1) that EPA is legally prohibited from ``delegat[ing] technical
decision-making to NHTSA;'' \116\ (2) that ``EPA must exercise its
technical and scientific expertise'' to develop CO2
standards and ``Anything less is an unlawful abdication of EPA's
statutory responsibilities;'' \117\ (3) that EPA staff is much more
qualified than DOT staff to conduct analysis relating to standards and
has done a great deal of work to inform development of standards; \118\
(4) that ``The Draft TAR and 2017 Final Determination relied
extensively on use of sophisticated EPA analytic tools and
methodologies,'' i.e., the ``peer reviewed simulation model ALPHA,''
``the agency's vehicle teardown studies,'' and the ``peer-reviewed
OMEGA model to make reasonable estimates of how manufacturers could add
technologies to vehicles in order to meet a fleet-wide [CO2
emissions] standard;'' \119\ (5) that NHTSA had said in the MYs 2012-
2016 rulemaking that the Volpe [CAFE] model was developed to support
CAFE rulemaking and incorporates features ``that are not appropriate
for use by EPA in setting [tailpipe CO2] standards;'' \120\
(6) allegations that some EPA staff had disagreed with aspects of the
NPRM analysis and had requested that ``EPA's name and logo should be
removed from the DOT-NHTSA Preliminary Regulatory Impact Analysis
document'' and stated that ``EPA is relying upon the technical analysis
performed by DOT-NHTSA for the [NPRM];'' \121\ (7) that EPA had
developed ``a range of relevant new analysis'' that the proposal
``failed to consider,'' including ``over a dozen 2017 and 2018 EPA peer
reviewed SAE articles;'' \122\ (8) that EPA's OMEGA modeling undertaken
during NPRM development ``found costs half that of NHTSA's findings,''
``Yet NHTSA did not correct the errors in its modeling and analysis,
and the published proposal drastically overestimates the cost of
complying . . . .;'' \123\ (9) that some EPA staff had requested that
the technology ``HCR2'' be included in the NPRM analysis, ``Yet NHTSA
overruled
[[Page 24224]]
EPA and omitted the technology;'' \124\ (10) that certain EPA staff had
initially ``rejected use of the CAFE model for development of the
proposed [tailpipe CO2] standards;'' \125\ (11) that there
are ``many specific weaknesses of the modeling results derived in this
proposal through use of the Volpe and Autonomie models'' and that the
CAFE model is ``not designed in accordance with'' Section 202(a) of the
CAA because (A) EPA ``is not required to demonstrate that standards are
set at the maximum feasible level year-by-year,'' (B) because EPCA
``preclude[s NHTSA] from considering vehicles powered by fuels other
than gas or diesel'' and EPA is not similarly bound, and (C) because
the CAFE model assumed that the value of an overcompliance credit
equaled $5.50, the value of a CAFE penalty.\126\ Because of all of
these things, the environmental group coalition stated that the
proposal was ``unlawful'' and that ``Before proceeding with this
rulemaking, EPA must consider all relevant materials including these
excluded insights, perform its own analysis, and issue a reproposal to
allow for public comment.'' \127\
---------------------------------------------------------------------------
 \115\ Environmental group coalition, NHTSA-2018-0067-12000,
Appendix A, at 24-25.
 \116\ Id. at 12.
 \117\ Id. at 14.
 \118\ Id. at 15-17.
 \119\ Id. at 17.
 \120\ Id. at 18.
 \121\ Id. at 19.
 \122\ Id. at 20.
 \123\ Id. at 21.
 \124\ Id. at 21-22.
 \125\ Id. at 23.
 \126\ Id. at 24-25.
 \127\ Id. at 27.
---------------------------------------------------------------------------
 Some environmental organizations and States contracted for external
technical analyses augmenting general comments such as those summarized
above. EDF engaged a consultant, Richard Rykowski, for a detailed
review of the agencies' analysis.\128\ Among Mr. Rykowski's comments, a
few specifically involve differences between these two models. Mr.
Rykowski recommended NHTSA's CAFE model replace its existing
``effective cost'' metric (used to compare available options to add
specific technologies to specific vehicles) with a ``ranking factor''
used for the same purpose. As discussed below in Section VI.A, the
model for today's final rule adopts this recommendation. He also states
that (1) ``EPA has developed a better way to isolate and reject cost
ineffective combinations of technologies . . . [and] includes only
these 50 or so technology combinations in their OMEGA model runs;'' (2)
``NHTSA's arbitrary and rigid designation of leader-follower vehicles
for engine, transmission and platform level technologies
unrealistically slows the rollout of technology into the new vehicle
fleet;'' (3) ``the Volpe Model is not capable of reasonably simulating
manufacturers' ability to utilize CO2 credits to smooth the
introduction of technology throughout their vehicle line-up;'' and (4)
``the Volpe Model is not designed to reflect the use of these [A/C
leakage] technologies and refrigerants.'' \129\
---------------------------------------------------------------------------
 \128\ EDF, NHTSA-2018-0067-12108, Appendix B. See also EPA, Peer
Review of the Optimization Model for Reducing Emissions of
Greenhouse Gases from Automobiles (OMEGA) and EPA's Response to
Comments, EPA-420-R-09-016, September 2009.
 \129\ EDF, op. cit., at 73-75.
---------------------------------------------------------------------------
 Mr. Rogers's analysis focuses primarily on the agencies' published
analysis, but mentions that some engine ``maps'' (estimates--used as
inputs to full vehicle simulation--of engine fuel consumption under a
wide range of engine operating conditions) applied in Autonomie show
greater fuel consumption benefits of turbocharging than those applied
previously by EPA to EPA's ALPHA model, and these benefits could have
caused NHTSA's CAFE model to estimate an unrealistically great tendency
toward turbocharged engines (rather than high compression ratio
engines).\130\ Mr. Rogers also presents alternative examples of year-
by-year technology application to specific vehicle models, contrasting
these with year-by-year results from the agencies' NPRM analysis,
concluding that ``that the use of logical, unrestricted technology
pathways, with incremental benefits supported by industry-accepted
vehicle simulation and dynamic system optimization and calibration,
together with publicly-defensible costs, allows cost-effective
solutions to achieve target fuel economy levels which meet MY 2025
existing standards.'' \131\
---------------------------------------------------------------------------
 \130\ Roush Industries, NHTSA-2018-0067-11984, at 17-21.
 \131\ Roush Industries, NHTSA-2018-0067-11984, at 17-30.
---------------------------------------------------------------------------
 Mr. Duleep's analysis also focuses primarily on the agencies'
published analysis, but does mention that (1) ``the Autonomie modeling
assumes no engine change when drag and rolling resistance reductions
are implemented, as well as no changes to the transmission gear ratios
and axle ratios, . . . [but] the EPA ALPHA model adjusts for this
effect;'' (2) ``baseline differences in fuel economy [between two
manufacturers' different products using similar technologies] are
carried for all future years and this exaggerates the differences in
technology adoption requirements and costs between manufacturers; (3)
``assumptions [that most technology changes are best applied as part of
a vehicle redesign or freshening] result in unnecessary distortion in
technology paths and may bias results of costs for different
manufacturers;'' and (4) that for the sample results shown for the
Chevrolet Equinox ``the publicly available EPA lumped parameter model
(which was used to support the 2016 rulemaking) and 2016 TAR cost data
. . . results in an estimate of attaining 52.2 mpg for a cost of $2110,
which is less than half the cost estimated in the PRIA.'' \132\
---------------------------------------------------------------------------
 \132\ H-D Systems, op. cit., at 48, et seq.
---------------------------------------------------------------------------
 Beyond these comments regarding differences between EPA's models
and the Argonne and DOT models applied for the NPRM, these and other
technical reviewers had many specific comments about the agencies'
analysis for the NPRM, and these comments are discussed in detail below
in Section VI.B.
 Manufacturers, on the other hand, supported the agencies' use of
Autonomie and the CAFE model rather than, in EPA's case, the ALPHA and
OMEGA models. Expressly identifying the distinction between models and
model inputs, Global Automakers stated that:
 The agencies provided a new, updated analysis based on the most
up-to-date data, using a proven and long-developed modeling tool,
known as the Volpe model, and offering numerous options to best
determine the right regulatory and policy path for ongoing fuel
efficiency improvements in our nation. Now, all stakeholders have an
opportunity to come to the table as part of the public process to
provide input, data, and information to help shape the final
rule.\133\
---------------------------------------------------------------------------
 \133\ Global Automakers, NHTSA-2018-0067-12032, at 2.
---------------------------------------------------------------------------
 This NPRM's use of a single model to evaluate alternative
scenarios for both programs provides consistency in the technical
analysis, and Global Automakers supports the Volpe model's use as it
has proven to be a transparent and user-friendly option in this
current analysis. The use of the Volpe model has allowed for a broad
range of stakeholders, with varying degrees of technical expertise,
to review the data inputs to provide feedback on this proposed rule.
The Volpe model's accompanying documentation has historically
provided a clear explanation of all sources of input and constraints
critical to a transparent modeling process. Other inputs have come
from modeling that is used widely by other sources, specifically the
Autonomie model, allowing for a robust validation, review and
reassessment.\134\
---------------------------------------------------------------------------
 \134\ Global Automakers, NHTSA-2018-0067-12032, Attachment A, at
A-12.
 The Alliance commented, similarly, that ``at least at this time,
NHTSA's modeling systems are superior to EPA's'' and ``as such, we
support the Agencies' decision to use NHTSA's modeling tools for this
rulemaking and recommend that both Agencies continue on this path. We
encourage Agencies to work together to provide input to the single
common set of tools.'' \135\
---------------------------------------------------------------------------
 \135\ Alliance, NHTSA-2018-0067-12073, at 134.
---------------------------------------------------------------------------
[[Page 24225]]
 Regarding the agencies' use of Argonne's Autonomie model rather
than EPA's ALPHA model, the Alliance commented that (1) ``the benefits
of virtually all technologies and their synergistic effects are now
determined with full vehicle simulations;'' (2) ``vehicle categories
have been increased to 10 to better recognize the range of 0-60
performance characteristics within each of the 5 previous categories,
in recognition of the fact that many vehicles in the baseline fleet
significantly exceeded the previously assumed 0-60 performance metrics.
This provides better resolution of the baseline fleet and more accurate
estimates of the benefits of technology. . . .;'' (3) ``new
technologies (like advanced cylinder deactivation) are included, while
unproven combinations (like Atkinson engines with 14:1 compression,
cooled EGR, and cylinder deactivation in combination) have been
removed;'' (4) ``Consistent with the recommendation of the National
Academy of Sciences and manufacturers, gradeability has been included
as a performance metric used in engine sizing. This helps prevent the
inclusion of small displacement engines that are not commercially
viable and that would artificially inflate fuel savings;'' (5) ``the
Alliance believes NHTSA's tools (Autonomie/Volpe) are superior to EPA's
(APLHA[sic]/LPM/OMEGA). This is not surprising since NHTSA's tools have
had a significant head start in development. . . .'' (6) ``the
Autonomie model was developed at Argonne National Lab with funding from
the Department of Energy going back to the PNGV (Partnership for Next
Generation Vehicles) program in the 1990s. Autonomie was developed from
the start to address the complex task of combining 2 power sources in a
hybrid powertrain. It is a physics-based, forward looking, vehicle
simulator, fully documented with available training,'' and (7) ``EPA's
ALPHA model is also a physics-based, forward looking, vehicle
simulator. However, it has not been validated or used to simulate
hybrid powertrains. The model has not been documented with any
instructions making it difficult for users outside of EPA to run and
interpret the model.'' \136\
---------------------------------------------------------------------------
 \136\ Id. at 135.
---------------------------------------------------------------------------
 Regarding the use of NHTSA's CAFE model rather than EPA's OMEGA
model, the Alliance stated that (1) NHTSA's model appropriately
differentiate between domestic and imported automobiles; (2) in NHTSA's
model, ``dynamic estimates of vehicle sales and scrappage in response
to price changes replace unrealistic static sales and scrappage
numbers;'' (3) NHTSA's model ``has new capability to perform
[CO2 emissions] analysis with [tailpipe CO2]
program flexibilities;'' (4) ``the baseline fleet [used in NHTSA's
model] has been appropriately updated based on both public and
manufacturer data to reflect the technologies already applied,
particularly tire rolling resistance;'' and (5) ``some technologies
have been appropriately restricted. For example, low rolling resistance
tires are no longer allowed on performance vehicles, and aero
improvements are limited to maximum levels of 15% for trucks and 10%
for minivans.'' \137\ The Alliance continued, noting that ``NHTSA's
Volpe model also predates EPA's OMEGA model. More importantly, the new
Volpe model considers several factors that make its results more
realistic.'' \138\ As factors leading the Volpe model to produce
results that are more realistic than those produced by OMEGA, the
Alliance commented that (1) ``The Volpe model includes estimates of the
redesign and refresh schedules of vehicles based on historical trends,
whereas the OMEGA model uses a fixed, and too short, time interval
during which all vehicles are assumed to be fully redesigned. . . .;''
(2) ``The Volpe model allows users to phase-in technology based on year
of availability, platform technology sharing, phase-in caps, and to
follow logical technology paths per vehicle. . . .;'' (3) ``The Volpe
model produces a year-by year analysis from the baseline model year
through many years in the future, whereas the OMEGA model only analyzes
a fixed time interval. . . .;'' (4) ``The Volpe model recognizes that
vehicles share platforms, engines, and transmissions, and that
improvements to any one of them will likely extend to other vehicles
that use them'' whereas ``The OMEGA model treats each vehicle as an
independent entity. . . .;'' (5) ``The Volpe model now includes sales
and scrappage effects;'' and (6) ``The Volpe model is now capable of
analyzing for CAFE and [tailpipe CO2] compliance, each with
unique program restrictions and flexibilities.'' \139\ The Alliance
also incorporated by reference concerns it raised regarding EPA's
OMEGA-based analysis supporting EPA's proposed and prior final
determinations.\140\
---------------------------------------------------------------------------
 \137\ Id. at 134.
 \138\ Id. at 135.
 \139\ Id. at 135-136.
 \140\ Id. at 136.
---------------------------------------------------------------------------
 The Alliance further stated that ``For all of the above reasons and
to avoid duplicate efforts, the Alliance recommends that the Agencies
continue to use DOT's Volpe and Autonomie modeling system, rather than
continuing to develop two separate systems. EPA has demonstrated
through supporting Volpe model code revisions and by supplying engine
maps for use in the Autonomie model that their expertise can be
properly represented in the rulemaking process without having to
develop separate or new tools.'' \141\
---------------------------------------------------------------------------
 \141\ Id. at 136.
---------------------------------------------------------------------------
 Some individual manufacturers provided comments supporting and
elaborating on the above comments by Global Automakers and the
Alliance. For example, FCA commented that ``the modeling performed by
the agencies should illuminate the differences between the CAFE and
[tailpipe CO2 emissions] programs. This cannot be
accomplished when each agency is using different tools and assumptions.
Since we believe NHTSA possesses the better set of tools, we support
both agencies using Autonomie for vehicle modeling and Volpe (CAFE) for
fleet modeling.'' \142\
---------------------------------------------------------------------------
 \142\ FCA, NHTSA-2018-0067-11943, at 82.
---------------------------------------------------------------------------
 Honda stated that ``The current version of the CAFE model is
reasonably accurate in terms of technology efficiency, cost, and
overall compliance considerations, and reflects a notable improvement
over previous agency modeling efforts conducted over the past few
years. We found the CAFE model's characterization of Honda's
``baseline'' fleet--critical modeling minutiae that provide a technical
foundation of the agencies' analysis--to be highly accurate. We commend
NHTSA and Volpe Center staff on these updates, as well as on the
overall transparency of the model. The model's graphical user interface
(GUI) makes it easier to run, model functionality is thoroughly
documented, and the use of logical, traceable input and output files
accommodates easy tracking of results.'' \143\ Similarly, in an earlier
presentation to the agencies, Honda included the following slide
comparing EPA's OMEGA model to DOT's CAFE (Volpe) model, and making
recommendations regarding future improvements to the latter: \144\
---------------------------------------------------------------------------
 \143\ Honda, EPA-HQ-OAR-2018-0283, at 21-22.
 \144\ Honda, NHTSA-2018-0067-12019, at 12.
---------------------------------------------------------------------------
[[Page 24226]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.041
 Toyota, in addition to arguing that the agencies' application of
model inputs (e.g., an analysis fleet based on MY 2016 compliance data)
produced more realistic results than in the draft TAR and in EPA's
former proposed and final determinations, also stressed the importance
of the CAFE model's year-by-year accounting for product redesigns,
stating that this produces more realistic results than the OMEGA-based
results shown previously by EPA:
 The modeling now better accounts for factors that limit the rate
at which new technologies enter and then diffuse through a
manufacturer's fleet. Bringing new or improved vehicles and
technologies to market is a several-year, capital-intensive
undertaking. Once new designs are introduced, a period of stability
is required so investments can be amortized. Vehicle and technology
introductions are staggered over time to manage limited resources.
Agency modeling now better recognizes the inherent constraints
imposed by realities that dictate product cadence. We agree with the
agencies' understanding that ``the simulation of compliance actions
that manufacturers might take is constrained by the pace at which
new technologies can be applied in the new vehicle market,'' and we
are encouraged to learn that ``agency modeling can now account for
the fact that individual vehicle models undergo significant
redesigns relatively infrequently.'' The preamble correctly notes
that manufacturers try to keep costs down by applying most major
changes mainly during vehicle redesigns and more modest changes
during product refresh, and that redesign cycles for vehicle models
can range from six to ten years, and eight to ten-years for
powertrains. This appreciation for standard business practice
enables the modeling to more accurately capture the way vehicles
share engines, transmissions, and platforms. There are now more
realistic limits placed on the number of engines and transmissions
in a powertrain portfolio which better recognizes manufacturers must
manage limited engineering resources and control supplier,
production, and service costs. Technology sharing and inheritance
between vehicle models tends to limit the rate of improvement in a
manufacturer's fleet.\145\
---------------------------------------------------------------------------
 \145\ Toyota, NHTSA-2018-0067-12098, Attachment 1, at 3 et seq.
 These comments urging EPA to use NHTSA's CAFE model echo comments
provided in response to a 2018 peer review of the model. While
identifying various opportunities for improvement, peer reviewers
expressed strong overall support for the CAFE model's technical
approach and execution. For example, one reviewer, after offering many
---------------------------------------------------------------------------
specific technical recommendations, concluded as follows:
 The model is impressive in its detail, and in the completeness
of the input data that it uses. Although the model is complex, the
reader is given a clear account of how variables are variously
divided and combined to yield appropriate granularity and efficiency
within the model. The model tracks well a simplified version of the
real-world and manufacturing/design decisions. The progression of
technology choices and cost benefit choices is clear and logical. In
a few cases, the model simply explains a constraint, or a value
assigned to a variable, without defending the choice of the value or
commenting on real-world variability, but these are not substantive
omissions. The model will lend itself well to future adaptation or
addition of variables, technologies and pathways.\146\
---------------------------------------------------------------------------
 \146\ NHTSA, CAFE Model Peer Review, DOT HS 812 590, Available
at https://www.nhtsa.gov/document/cafe-model-peer-review, at 250.
 Although the peer review charge focused solely on the CAFE model,
another peer reviewer separately recommended that EPA ``consider
opportunities for EPA to use the output from the Volpe Model in place
of their OMEGA Model output'' \147\
---------------------------------------------------------------------------
 \147\ Id. at 287-288 and 304.
---------------------------------------------------------------------------
 More recently, in response to the NPRM, Dr. Julian Morris, an
economist at George Washington University, commented extensively on the
superiority of the agencies' NPRM analysis to previous analyses,
offering the following overall assessment:
 I have assessed the plausibility of the analyses undertaken by
NHTSA and EPA in relation to the proposed SAFE rule. I found that
the agencies have undertaken a thorough--one might even say
exemplary--analysis, improving considerably on earlier analyses
undertaken by the agencies of previous rules relating to CAFE
standards and [tailpipe CO2] emission standards. Of
particular note, the agencies included more realistic estimates of
the rebound effect, developed a sophisticated model of the
[[Page 24227]]
scrappage effect, and better accounted for various factors affecting
vehicle fatality rates.\148\
---------------------------------------------------------------------------
 \148\ Morris, J., OAR-2018-0283-4028, at 6-11.
 The agencies carefully considered these and other comments
regarding which models to apply when estimating potential impacts of
each of the contemplated regulatory alternatives. For purposes of
estimating the impacts of CAFE standards, even the coalition of
environmental advocates observed that the CAFE model reflects EPCA's
requirements. As discussed below in Section VI.A, EPCA imposes specific
requirements not only on how CAFE standards are to be structured (e.g.,
including a minimum standard for domestic passenger cars), but also on
how CAFE standards are to be evaluated (e.g., requiring that the
potential to produce additional AFVs be set aside for the model years
under consideration), and the CAFE model reflects these requirements,
and the agencies consider the CAFE model to be the best available tool
for CAFE rulemaking analysis. Regarding the use of Autonomie to
construct fuel consumption (i.e., efficiency) inputs to the CAFE model,
the agencies recognize that other vehicle simulation tools are
available, including EPA's recently-developed ALPHA model. However, as
also discussed in Section VI.B.3, Autonomie has a much longer history
of development and refinement, and has been much more widely applied
and validated. Moreover, Argonne experts have worked carefully for
several years to develop methods for running large arrays of
simulations expressly structured and calibrated for use in DOT's CAFE
model. Therefore, the agencies consider Autonomie to be the best
available tool for constructing such inputs to the CAFE model. While
the agencies have also carefully considered potential specific model
refinements, as well as the merits of potential changes to model inputs
and assumptions, none of these potential refinements and input have led
either agency to reconsider using the CAFE model and Autonomie for CAFE
rulemaking analysis.
 With respect to estimating the impacts of CO2 standards,
even though Argonne and the agencies have adapted Autonomie and the
CAFE model to support the analysis of CO2 standards,
environmental groups, California, and other States would prefer that
EPA use the models it developed during 2009-2018 for that purpose.\149\
Arguments that EPA revert to its ALPHA and OMEGA models fall within
three general categories: (1) Arguments that EPA's models would have
selected what commenters consider better (i.e., generally more
stringent) standards, (2) arguments that EPA's models are technically
superior, and (3) arguments that the law requires EPA use its own
models.
---------------------------------------------------------------------------
 \149\ The last-finalized versions of EPA's OMEGA model and ALPHA
tools were published in 2016 and 2017, respectively.
---------------------------------------------------------------------------
 The first of these arguments--that EPA's models would have selected
better standards--conflates the analytical tool used to inform
decision-making with the action of making the decision. As explained
elsewhere in this document and as made repeatedly clear over the past
several rulemakings, the CAFE model (or, for that matter, any model)
neither sets standards nor dictates where and how to set standards; it
simply informs as to the potential effects of setting different levels
of standards. In this rulemaking, EPA has made its own decisions
regarding what CO2 standards would be appropriate under the
CAA.
 The third of these arguments--that EPA is legally required to use
only models developed by its own staff--is also without merit. The CAA
does not require the agency to create or use a specific model of its
own creation in setting tailpipe CO2 standards. The fact
that EPA's decision may be informed by non-EPA-created models does not,
in any way, constitute a delegation of its statutory power to set
standards or decision-making authority.\150\ Arguing to the contrary
would suggest, for example, that EPA's decision would be invalid
because it relied on EIA's Annual Energy Outlook for fuel prices for
all of its regulatory actions rather than developing its own model for
estimating future trends in fuel prices. Yet, all Federal agencies that
have occasion to use forecasts of future fuel prices regularly (and
appropriately) defer to EIA's expertise in this area and rely on EIA's
NEMS-based analysis in the AEO, even when those same agencies are using
EIA's forecasts to inform their own decision-making. Similarly, this
argument would mean that the agencies could not rely on work done by
contractors or other outside consultants, which is contrary to regular
agency practice across the entirety of the Federal Government.
---------------------------------------------------------------------------
 \150\ ``[A] federal agency may turn to an outside entity for
advice and policy recommendations, provided the agency makes the
final decisions itself.'' U.S. Telecom. Ass'n v. FCC, 359 F.3d 554,
565-66 (D.C. Cir. 2004). To the extent commenters meant to suggest
outside parties have a reliance interest in EPA using ALPHA and
OMEGA to set standards, EPA and NHTSA do not agree a reliance
interest is properly placed on an analytical methodology, which
consistently evolves from rule to rule. Even if it were, all parties
that closely examined ALPHA and OMEGA-based analyses in the past
either also simultaneously closely examined CAFE and Autonomie-based
analyses in the past, or were fully capable of doing so, and thus,
should face no additional difficulty now they have only one set of
models and inputs/outputs to examine.
---------------------------------------------------------------------------
 The specific claim here that use of the CAFE model instead of ALPHA
and OMEGA is somehow illegitimate is similarly unpersuasive. The CAFE
and CO2 rules have, since Massachusetts v. EPA, all been
issued as joint rulemakings, and, thus are the result of a
collaboration between the two agencies. This was true when the
rulemakings used separate models for the different programs and
continues to be true in today's final rule, where the agencies take the
next step in their collaborative approach by now using simply one model
to simulate both programs. In 2007, immediately following this Supreme
Court decision, the agencies worked together toward standards for model
years 2011-2015, and EPA made use of the CAFE model for its work toward
possible future CO2 standards. That the agencies would need
to continue the unnecessary and inefficient process of using two
separate combinations of models as the joint National Program continues
to mature, therefore, runs against the idea that the agencies, over
time, would best combine resources to create an efficient and robust
regulatory program. For the reasons discussed throughout today's final
rule, the agencies have jointly determined that Autonomie and the CAFE
model have significant technical advantages, including important
additional features, and are therefore the more appropriate models to
use to support both analyses.
 Further, the fact that Autonomie and CAFE models were initially
developed by DOE/Argonne and NHTSA does not mean that EPA has no role
in either these models or their inputs. That is, the development
process for CAFE and CO2 standards inherently requires
technical and policy examinations and deliberations between staff
experts and decision-makers in both agencies. Such engagements are a
healthy and important part of any rulemaking activity--and particularly
so with joint rulemakings. The Supreme Court stated in Massachusetts v.
EPA that, ``The two obligations [to set CAFE standards under EPCA and
to set tailpipe CO2 emissions standards under the CAA] may
overlap, but there is no reason to think the two agencies cannot both
administer their obligations and yet
[[Page 24228]]
avoid inconsistency.'' \151\ When agency experts consider analytical
issues and agency decision-makers decide on policy, which is informed
(albeit not dictated) by the outcome of that work, they are working
together as the Court appears to have intended in 2007, even if
legislators' intentions have varied in the decades since EPCA and the
CAA have been in place.\152\ Regulatory overlap necessarily involves
deliberation, which can lead to a more balanced, reasonable, and
improved analyses, and better regulatory outcomes. It did here. The
existence of deliberation is not per se evidence of unreasonableness,
even if some commenters believe a different or preferred policy outcome
would or should have resulted.\153\
---------------------------------------------------------------------------
 \151\ Massachusetts v. EPA, 549 U.S. 497, 532 (2007).
 \152\ For example, when wide-ranging amendments to the CAA were
being debated, S. 1630 contained provisions that, if enacted, would
have authorized automotive CO2 emissions standards and
prescribed specific average levels to be achieved by 1996 and 2000.
In a letter to Senators, then-Administrator William K. Reilly noted
that the Bill ``requires for the first time control of emissions of
carbon dioxide; this is essentially a requirement to improve fuel
efficiency'' and outlined four reasons the H.W. Bush Administration
opposed the requirement, including that ``it is inappropriate to add
this very complex issue to the Clean Air Act which is already full
of complicated and controversial issues.'' Reilly, W., Letter to
U.S. Senators (January 26, 1990). The CAA amendments ultimately
signed into law did not contain these or any other provisions
regarding regulation of CO2 emissions.
 \153\ See, e.g., U.S. House of Representatives, Committee on
Oversight and Government Reform, Staff Report, 112th Congress, ``A
Dismissal of Safety, Choice, and Cost: The Obama Administration's
New Auto Regulations,'' August 10, 2012, at 19-21 and 33-34.
---------------------------------------------------------------------------
 Over the 44 years since EPCA established the requirement for CAFE
standards, NHTSA, EPA and DOE career staff have discussed, collaborated
on, and debated engineering, economic, and other aspects of CAFE
regulation, through focused meetings and projects, informal exchanges,
publications, conferences and workshops, and rulemakings.
 Part of this expanded exchange has involved full vehicle
simulation. While tools such as PSAT (the DOE-sponsored simulation tool
that predated Autonomie) were in use prior to 2007, including for
discrete engineering studies supporting inputs to CAFE rulemaking
analyses, these tools' information and computing requirements were such
that NHTSA had determined (and DOE and EPA had concurred) that it was
impractical to more fully integrate full vehicle simulation into
rulemaking analyses. Since that time, computing capabilities have
advanced dramatically, and the agencies now agree that such integration
of full vehicle simulation--such as the large-scale exercise of
Autonomie to produce inputs to the CAFE Model--can make for more robust
CAFE and CO2 rulemaking analysis. This is not to say,
though, that experts always agree on all methods and inputs involved
with full vehicle simulation. Differences in approach and inputs lead
to differences in results. For example, compared to other publicly
available tools that can be practicably exercised at the scale relevant
to fleetwide analysis needed for CAFE and CO2 rulemaking
analysis, DOE/Argonne's Autonomie model is more advanced, spans a wider
range of fuel-saving technologies, and represents them in more specific
detail, leaving fewer ``gaps'' to be filled with other models (risking
inconsistencies and accompanying errors). These differences discussed
in greater detail below in Section VI.B.3. Perhaps most importantly,
the CAFE model considers fuel prices in determining both which
technologies are applied and the total amount of technology applied, in
the case where market forces demand fuel economy levels in excess of
the standards. While OMEGA can apply technology in consideration of
fuel prices, OMEGA will apply technology to reach the same level of
fuel economy (or CO2 emissions) if fuel prices are 3, 5, or
20 dollars, which violates the SAB's requirement that the analysis
``account for [. . .] future fuel prices .'' \154\ Furthermore, it
produces a counterintuitive result. If fuel prices become exorbitantly
high, we would expect consumers to place an emphasis on additional fuel
efficiency as the potential for extra fuel savings is tremendous.
---------------------------------------------------------------------------
 \154\ See SAB Report 10 (``Constructing each of the scenarios is
challenging and involve extensive scientific, engineering, and
economic uncertainties. Projecting the baseline requires the
agencies to account for a wide range of variables including: The
number of new vehicles sold, future fuel prices,. . . .'').
---------------------------------------------------------------------------
 Moreover, DOE has for many years used Autonomie (and its precursor
model, PSAT) to produce analysis supporting fuel economy-related
research and development programs involving billions of dollars of
public investment, and NHTSA's CAFE model with inputs from DOE/
Argonne's Autonomie model has produced analysis supporting rulemaking
under the CAA. In 2015, EPA proposed new tailpipe CO2
standards for MY 2021-2027 heavy-duty pickups and vans, finalizing
those standards in 2016. Supporting the NPRM and final rule, EPA relied
on analysis implemented by NHTSA using NHTSA's CAFE model, and NHTSA
used inputs developed by DOE/Argonne using DOE/Argonne's Autonomie
model. CBD questioned this history, asserting that, ``EPA conducted a
separate analysis using a different iteration of the CAFE model rather
than rely on the version which NHTSA used, again resulting and parallel
but corroborative modeling results.'' \155\ CBD's comment
mischaracterizes EPA's actual use of the CAFE Model. As explained in
the final rule, EPA's ``Method B'' analysis was developed as follows:
---------------------------------------------------------------------------
 \155\ CBD, et al., 2018-0067-12000, Appendix A, at 27.
 In Method B, the CAFE model from the NPRM was used to project a
pathway the industry could use to comply with each regulatory
alternative, along with resultant impacts on per-vehicle costs.
However, the MOVES model was used to calculate corresponding changes
in total fuel consumption and annual emissions for pickups and vans
in Method B. Additional calculations were performed to determine
corresponding monetized program costs and benefits.\156\
---------------------------------------------------------------------------
 \156\ 81 FR 73478, 73506-07 (October 25, 2016).
 In other words, a version of NHTSA's CAFE Model was used to perform
the challenging part of the analysis--that is, the part that involves
accounting for manufacturers' fleets, accounting for available fuel-
saving technologies, accounting for standards under consideration, and
estimating manufacturers' potential responses to new standards--EPA's
MOVES model was used to perform ``downstream'' calculations of fuel
consumption and tailpipe emissions, and used spreadsheets to calculate
even more straightforward calculations of program costs and benefits.
While some stakeholders perceive these differences as evidencing a
meaningfully independent approach, in fact, the EPA staff's analysis
was, at its core, wholly dependent on NHTSA's CAFE Model, and on that
model's use of Autonomie simulations.
 Given the above, the only remaining argument for EPA to revert to
its previously-developed models rather than relying on Autonomie and
the CAFE model would be that the former are so technically superior to
the latter that even model refinements and input changes cannot lead
Autonomie and the CAFE model to produce appropriate and reasonable
results for CO2 rulemaking analysis. As discussed below,
having considered a wide range of technical differences, the agencies
find that the Autonomie and CAFE models currently provide the best
analytical combination for CAFE and tailpipe CO2 emissions
rulemaking analysis. As discussed
[[Page 24229]]
below in Section VI.B.3, Autonomie not only has a longer and wider
history of development and application, but also DOE/Argonne's
interaction with automakers, supplier and academies on continuous bases
had made individual sub-models and assumptions more robust. Argonne has
also been using research from DOE's Vehicle Technology Office (VTO) at
the same time to make continuous improvements in Autonomie.\157\ Also,
while Autonomie uses engine maps as inputs, and EPA developed engine
maps that could have been used for today's analysis, EPA declined to do
so, and those engine maps were only used in a limited capacity for
reasons discussed below in Section VI.C.1.
---------------------------------------------------------------------------
 \157\ U.S. DOE Benefits & Scenario Analysis publications is
available at https://www.autonomie.net/publications/fuel_economy_report.html. Last accessed November 14, 2019.
---------------------------------------------------------------------------
 As also discussed below in Section VI.A.4, the CAFE model accounts
for some important CO2 provisions that EPA's OMEGA model
cannot account for. For example, the CAFE model estimates the potential
that any given manufacturer might apply CO2 compliance
credits it has carried forward from some prior model year. While one
commenter, Mr. Rykowski, takes issue with how the CAFE model handles
credit banking, he does not acknowledge that EPA's OMEGA model, lacking
a year-by-year representation of compliance, is altogether incapable of
accounting for the earning and use of banked compliance credits. Also,
although Mr. Rykowski's comments regarding A/C leakage and refrigerants
are partially correct insofar as the CAFE model does not account for
leakage-reducing technologies explicitly, the comment is as applicable
to OMEGA as it is to the CAFE model and, in any event, data regarding
which vehicles have which leakage-reducing technologies was not
available for the MY 2016 fleet. Nevertheless, as discussed in Section
VI.A.4, NHTSA has refined the CAFE model's accounting for the cost of
leakage reduction technologies.
 The agencies have also considered Mr. Rykowski's comments
suggesting that using OMEGA would be preferable because, rather than
selecting from hundreds of thousands of potential combinations of
technologies, OMEGA includes only the ``50 or so'' combinations that
EPA has already determined to be cost-effective. The ``better way'' of
making this determination is also effectively a model, but the
separation of this model from OMEGA is, as evidenced by manufacturers'
comments, obfuscatory, especially in terms of revealing how specific
vehicle model/configurations initial engineering properties are aligned
with specific initial technology combinations. By using a full set of
technology combinations, the CAFE model makes very clear how each
vehicle model/configuration is assigned to a specific initial
combination and, hence, how subsequently fuel consumption and cost
changes are accounted for. Moreover, EPA's separation of ``thinning''
process from OMEGA's main compliance simulation makes sensitivity
analysis difficult to implement, much less follow. The agencies find,
therefore, that the CAFE model's approach of retaining a full set of
vehicle simulation results throughout the compliance simulation to be
more realistic (e.g., more capable of reflecting manufacturer- and
vehicle-specific factors), more responsive to changes in model inputs
(e.g., changes to fuel prices, which could impact the relative
attractiveness of different technologies), more transparent, and more
amenable to independent corroboration the agencies' analysis.
 Regarding comments by Messrs. Duleep, Rogers, and Rykowski
suggesting that the CAFE model, by tying most technology application to
planned vehicle redesigns and freshening, is too restrictive, the
agencies disagree. As illustrated by manufacturers' comments cited
above, as reinforced by both extensive product planning information
provided to the agencies, and as further reinforced by extensive
publicly available information, manufacturers tend to not make major
changes to a specific vehicle model/configuration in one model year,
and then make further major changes to the same vehicle model/
configuration the next model year. There is ample evidence that
manufacturers strive to avoid such discontinuity, complexity, and
waste, and in the agencies' view, while it is impossible to represent
every manufacturer's decision-making process precisely and with
certainty, the CAFE model's approach of using estimated product design
schedules provides a realistic basis for estimating what manufacturers
could practicably do. Also, the relevant inputs are simply inputs to
the CAFE model, and if it is actually more realistic to assume that a
manufacturer can change major technology on all of its products every
year, the CAFE model can easily be operated with every model year
designated as a redesign year for every product, but as discussed
throughout this document, the agencies consider this to be extremely
unrealistic. While this means the CAFE model can be run without a year-
by-year representation that carries forward technologies between model
years, doing so would be patently unrealistic (as reflected in some
stakeholders' comments in 2002 on the first version of the CAFE model).
Conversely, the OMEGA model cannot be operated in a way that accounts
for what the agencies consider to be very real product planning
considerations.
 However, having also considered Mr. Rykowski's comments about the
CAFE model's ``effective cost'' metric, and having conducted side-by-
side testing documented in the accompanying FRIA, the agencies are
satisfied that an alternative ``cost per credit'' metric is also a
reasonable metric to use for estimating how manufacturers might
selected among available options to add specific fuel-saving
technologies to specific vehicles.\158\ Therefore, NHTSA has revised
the CAFE model accordingly, as discussed below in Section VI.A.4.
---------------------------------------------------------------------------
 \158\ As discussed in the FRIA, results vary with model inputs,
among manufacturers, and across model years, but compared to the
NPRM's ``effective cost'' metric, the ``cost per credit'' metric
appears to more frequently produce less expensive solutions than
more expensive solutions, at least when simulating compliance with
CO2 standards. Differences are more mixed when simulating
compliance with CAFE standards, and even when simulating compliance
with CO2 standards, results simulating ``perfect'' trading of
CO2 compliance credits are less intuitive when the ``cost
per credit metric.'' Nevertheless, and while less expensive
solutions are not necessarily ``optimal'' solutions (e.g., if
gasoline costs $7 per gallon and electricity is free, expensive
electrification could be optimal), the agencies consider it
reasonable to apply the ``cost per credit'' metric for the analysis
supporting today's rulemaking.
---------------------------------------------------------------------------
 Section VI.C.1 also addresses Mr. Rogers's comments on engine maps
used as estimates to full vehicle simulation. In any event, because
engine maps are inputs to full vehicle modeling and simulation, the
relative merits of specific maps provide no basis to prefer one vehicle
simulation modeling system over another. Similarly, Section VI.B.3 also
addresses Mr. Duleep's comments preferring EPA's prior approach, using
ALPHA, of effectively assuming that a manufacturer would incur no
additional cost by reoptimizing every powertrain to extract the full
fuel economy potential of even the smallest incremental changes to
aerodynamic drag and tire rolling resistance. Mr. Duleep implies that
Autonomie is flawed because the NPRM analysis did not apply Autonomie
in a way that makes such assumptions. The agencies discuss powertrain
sizing and calibration in Section VI.B.3, and note here that such
assumptions are not inherent to
[[Page 24230]]
Autonomie; like engine maps, these are inputs to full vehicle
simulation. Therefore, neither of these comments by Mr. Rogers and Mr.
Duleep lead the agencies to find reason not to use Autonomie.
 None of this is to say that Autonomie and the CAFE model as
developed and applied for the NPRM left no room for improvement. In the
NPRM and RIA, the agencies discussed plans to continue work in a range
of specific technical areas, and invited comment on all aspects of the
analysis. As discussed below in Chapter VI, the agencies received
extensive comment on the published model, inputs, and analysis, both in
response to the NPRM and, for newly-introduced modeling capabilities
(estimation of sales, scrappage, and employment effects), in response
to additional peer review conducted in 2019. The agencies have
carefully considered these comments, refined various specific technical
aspects of the CAFE model (like the ``effective cost'' metric mentioned
above), and have also updated inputs to both Autonomie and the CAFE
model. Especially given these refinements and updates, as discussed
throughout this rule, EPA maintains that for CO2 rulemaking
analysis, Autonomie and the CAFE model have advantages that warrant
relying on them rather than on EPA's ALPHA and OMEGA models. Some
examples of such advantages include: A longer history of ongong
development and application for rulemaking, including by EPA;
documentation and model operation stakeholders have found to be
comparatively clear and enabling of independent replication of agency
analyses; a mechanism to explicitly reflect the fact that
manufacturers' product decisions are likely to be informed by fuel
prices; better integration of various model functions, enabling
efficient sensitivity analysis; and an annual time step that makes it
possible to conduct report results on both a calendar year and model
year basis, to estimate accruing impacts on vehicle sales and
scrappage, and to account for the fact that not every vehicle can be
designed in every model year; and other advantages discussed throughout
today's notice. Therefore, recognizing that models inform but do not
make regulatory decisions, EPA has elected to rely solely on the
Autonomie and CAFE models to produce today's analysis of regulatory
alternatives for CO2 standards.
 The following sections provide a brief technical overview of the
CAFE model, including changes NHTSA made to the model since 2012, and
differences between the current analysis, the analysis for the 2016
Draft TAR and for the 2017 Proposed Determination/2018 Final
Determination, and the 2018 NPRM, before discussing inputs to the model
and then diving more deeply into how the model works. For more
information on the latter topic, see the CAFE model documentation,
available in the docket for this rulemaking and on NHTSA's website.
1. What assumptions have changed since the 2012 final rule?
 Any analysis of regulatory actions that will be implemented several
years in the future, and whose benefits and costs accrue over decades,
requires a large number of assumptions. Over such time horizons, many,
if not most, of the relevant assumptions in such an analysis are
inevitably uncertain.\159\ The 2012 CAFE/CO2 rule considered
regulatory alternatives for model years through MY 2025 (17 model years
after the 2008 market information that formed the basis of the
analysis) that accrued costs and benefits into the 2060s. Not only was
the new vehicle market in 2025 unlikely to resemble the market in 2008,
but so, too, were fuel prices. It is natural, then, that each
successive CAFE/CO2 analysis should update assumptions to
reflect better the current state of the world and the best current
estimates of future conditions.\160\ However, beyond the issue of
unreliable projections about the future, a number of agency assertions
have proven similarly problematic. In fact, Securing America's Future
Energy (SAFE) stated in their comments on the NPRM:
---------------------------------------------------------------------------
 \159\ As often stated, ``It's difficult to make predictions,
especially about the future.'' See, e.g., https://quoteinvestigator.com/2013/10/20/no-predict/.
 \160\ See, e.g., 77 FR 62785 (Oct. 15, 2012) (``If EPA initiates
a rulemaking [to revise standards for MYs 2022-2025], it will be a
joint rulemaking with NHTSA. . . . NHTSA's development of its
proposal in that later rulemaking will include the making of
economic and technology analyses and estimates that are appropriate
for those model years and based on then-current information.'').
 Although the agencies argue ``circumstances have changed'' and
``analytical methods and inputs have been updated,'' a thorough
analysis should provide a side-by-side comparison of the changing
circumstances, methods, and inputs used to arrive at this
determination . . . They represent a rapid, dramatic departure from
the agencies' previous analyses, without time for careful review and
consideration.\161\
---------------------------------------------------------------------------
 \161\ Securing America's Energy Future, NHTSA-2018-0067-12172,
at 39.
 We describe in detail (below) the changes to critical assumptions,
perspectives, and modeling techniques that have created substantive
differences between the current analysis and the analysis conducted in
2012 to support the final rule. To the greatest extent possible, we
have calculated the impacts of these changes on the 2012 analysis.
a) The Value of Fuel Savings
 The value of fuel savings associated with the preferred alternative
in the 2012 final rule is primarily a consequence of two assumptions:
\162\ The fuel price forecast and the assumed growth in fuel economy in
the baseline alternative against which savings are measured. Therefore,
as the value of fuel savings accounted for nearly 80 percent of the
total benefits of the 2012 rule, each of these assumptions is
consequential. With a lower fuel price projection and an expectation
that new vehicle buyers respond to fuel prices, the 2012 rule would
have shown much smaller fuel savings attributable to the more stringent
standards. Projected fuel prices are considerably lower today than in
2012, the agencies now understand new vehicle buyers to be at least
somewhat responsive to fuel prices, and the agencies have therefore
updated corresponding model inputs to produce an analysis the agencies
consider to be more realistic.
---------------------------------------------------------------------------
 \162\ The value of fuel savings is also affected by the rebound
effect assumption, assumed lifetime VMT accumulation, and the
simulated penetration of alternative fuel technologies. However,
each of these ancillary factors is small compared to the impact of
the two factors discussed in this subsection.
---------------------------------------------------------------------------
 The first of these assumptions, fuel prices, was simply an artifact
of the timing of the rule. Following recent periodic spikes in the
national average gasoline price and continued volatility after the
great recession, the fuel price forecast then produced by EIA (as part
of AEO 2011) showed a steady march toward historically high, sustained
gasoline prices in the United States. However, the actual series of
fuel prices has skewed much lower. As it has turned out, the observed
fuel price in the years between the 2012 final rule and this rule has
frequently been lower than the ``Low Oil Price'' sensitivity case in
the 2011 AEO, even when adjusted for inflation. The following graph
compares fuel prices underlying the 2012 final rule to fuel prices
applied in the analysis reported in today's notice, expressing both
projections in 2010 dollars. The differences are clear and significant:
[[Page 24231]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.042
 The discrepancy in fuel prices is important to the discussion of
differences between the current rule and the 2012 final rule, because
that discrepancy leads in turn to differences in analytical outputs and
thus to differences in what the agencies consider in assessing what
levels of standards are reasonable, appropriate, and/or maximum
feasible. As an example, the agencies discuss in Sections VI.D.3
Simulating Environmental Impacts of Regulatory Alternatives and
VIII.A.3 EPA's Conclusion that the Final CO2 Standards are
Appropriate and Reasonable that fuel price projections from the 2012
rule were one assumption, among others, that could have led to
overestimates of the health benefits that resulted from reducing
criteria pollutant emissions. Yet the agencies caution readers not to
interpret this discrepancy as a reflection of negligence on the part of
the agencies, or on the part of EIA. Long-term predictions are
challenging and the fuel price projections in the 2012 rule were within
the range of conventional wisdom at the time. However, it does suggest
that fuel economy and tailpipe CO2 regulations set almost
two decades into the future are vulnerable to surprises, in some ways,
and reinforces the value of being able to adjust course when critical
assumptions are proven inaccurate. This value was codified in
regulation when EPA bound itself to the mid-term evaluation process as
part of the 2012 final rule.\163\
---------------------------------------------------------------------------
 \163\ See 40 CFR 86-1818-12(h).
---------------------------------------------------------------------------
 To illustrate this point clearly, substituting the current (and
observed) fuel price forecast for the forecast used in the 2012 final
rule creates a significant difference in the value of fuel savings.
Even under identical discounting methods (see Section 2, below), and
otherwise identical inputs in the 2012 version of the CAFE Model, the
current (and historical) fuel price forecast reduces the value of fuel
savings by $150 billion--from $525 billion to $375 billion (in 2009
dollars).
 The second assumption employed in the 2012 (as well as the 2010)
final rule, that new vehicle fuel economy never improves unless
manufacturers are required to increase fuel economy in the new vehicle
market by increasingly stringent regulations, is more problematic.
Despite the extensive set of recent academic studies showing, as
discussed in Section VI.D.1.a)(2), that consumers value at least some
portion, and in some studies nearly all, of the potential fuel savings
from higher levels of fuel economy at the time they purchase vehicles,
the agencies assumed in past rulemakings that buyers of new vehicles
would never purchase, and manufacturers would never supply, vehicles
with higher fuel economy than those in the baseline (MY 2016 in the
2012 analysis), regardless of technology cost or prevailing fuel prices
in future model years. In calendar year 2025, the 2012 final rule
assumed gasoline would cost nearly $4.50/gallon in today's dollars, and
continue to rise in subsequent years. Even recognizing that higher
levels of fuel economy would be achieved under the augural/existing
standards than without them, the assertion that fuel economy and
CO2 emissions would not improve beyond 2016 levels in the
presence of nearly $5/gallon gasoline is not supportable. This is
highlighted by the observed increased consumer demand for higher-fuel-
economy vehicles during the gas price spike of 2008, when average U.S.
prices briefly broke $4/gallon. In the 2012 final rule, this
assumption--that fuel economy and emissions would never improve absent
regulation--created a persistent gap in fuel economy between
[[Page 24232]]
the baseline and augural standards that grew to 13 mpg (at the industry
average, across all vehicles) by MY 2025. In the 2016 Draft TAR,
NHTSA's analysis included the assumption that manufacturers would
deploy, and consumers would demand, any technology that recovered its
own cost in the first year of ownership through avoided fuel costs.
However, in both the Draft TAR and the Proposed and Final Determination
documents, EPA's analysis assumed that the fuel economy levels achieved
to reach compliance with MY 2021 standards would persist indefinitely,
regardless of fuel prices or technology costs.
 By substituting the conservative assumption that consumers are
willing to purchase fuel economy improvements that pay for themselves
with avoided fuel expenditures over the first 2.5 years \164\
(identical to the assumption in this final rule's central analysis) the
gap in industry average fuel economy between the baseline and augural
scenarios narrows from 13 mpg in 2025 to 6 mpg in 2025. As a corollary,
acknowledging that fuel economy would continue to improve in the
baseline under the fuel price forecast used in the final rule erodes
the value of fuel savings attributable to the preferred alternative.
While each gallon is still worth as much as was assumed in 2012, fewer
gallons are consumed in the baseline due to higher fuel economy levels
in new vehicles. In particular, the number of gallons saved by the
preferred alternative selected in 2012 drops from about 180 billion to
50 billion once we acknowledge the existence of even a moderate market
for fuel economy.\165\ The value of fuel savings is similarly eroded,
as higher fuel prices lead to correspondingly higher demand for fuel
economy even in the baseline--reducing the value of fuel savings from
$525 billion to $190 billion.
---------------------------------------------------------------------------
 \164\ Greene, D.L. and Welch, J.G., ``Impacts of fuel economy
improvements on the distribution of income in the U.S.,'' Energy
Policy, Volume 122, November 2018, pp. 528-41 (``Four nationwide
random sample surveys conducted between May 2004 and January 2013
produced payback period estimates of approximately three years,
consistent with the manufacturers' perceptions.'') (The 2018 article
succeeds Greene and Welch's 2017 publication titled ``The Impact of
Increased Fuel Economy for Light-Duty Vehicles on the Distribution
of Income in the U.S.: A Retrospective and Prospective Analysis,''
Howard H. Baker Jr. Center for Public Policy, March 2017, which
Consumers Union, CFA, and ACEEE comments include as Attachment 4,
Docket NHTSA-2018-0067-11731).
 \165\ Readers should note that this is not an estimate of the
total amount of fuel that will be consumed or not consumed by the
fleet as a whole, but simply the amount of fuel that will be
consumed or not consumed as a direct result of the regulation. As
illustrated in Section VII, light-duty vehicles in the U.S. would
continue to consume considerable quantities of fuel and emit
considerable quantities of CO2 even under the baseline/
augural standards, and agencies' analysis shows that the standards
finalized today will likely increase fuel consumption and
CO2 emissions by a small amount.
---------------------------------------------------------------------------
 The magnitude of the fuel economy improvement in the baseline is a
consequence of both the fuel prices assumed in the 2012 rule (already
discussed as being higher than both subsequent observed prices and
current projections) and the assumed technology costs. In 2012, a
number of technologies were assumed to have negative incremental
costs--meaning that applying those technologies to existing vehicles
would both improve their fuel economy and reduce the cost to produce
them. Asserting that the baseline would experience no improvement in
fuel economy without regulation is equivalent to asserting that
manufacturers, despite their status as profit maximizing entities,
would not apply these cost-saving technologies unless forced to do so
by regulation. While this issue is discussed in greater detail in
Section VI.B the combination of inexpensive (or free) technology and
high fuel prices created a logically inconsistent perspective in the
2012 rule--where consumers never demanded additional fuel economy,
despite high fuel costs, and manufacturers never supplied additional
fuel economy, despite the availability of inexpensive (or cost saving)
technology to do so.
 Many commenters on earlier rules supported the assumption that fuel
economy would not improve at all in the absence of standards. In fact,
some commenters still support this position. For example, EDF commented
to the NPRM that, ``NHTSA set the Volpe model to project that, with
CAFE standards remaining flat at MY 2020 levels through MY 2026,
automakers would over-comply with the MY 2020 standards by 9 grams/mile
of CO2 for cars and 15 g/mi of CO2 for light
trucks during the 2029-2032 timeframe, plus 1%/year improvements beyond
MY 2032. This assumption unreasonably obscures the impacts of the
rollback and is not reflective of historical compliance performance.''
\166\
---------------------------------------------------------------------------
 \166\ EDF, NHTSA-2018-0067-11996, Comments to DEIS, at 4.
---------------------------------------------------------------------------
 EDF is mistaken in two different ways: (1) By acknowledging the
existence of a well-documented market for fuel economy, rather than
erroneously inflating the benefits associated with increasing
standards, this assumption serves to isolate the benefits actually
attributable to each regulatory alternative, and (2) it is, indeed,
reflective of historical compliance performance. While the agencies
rely on the academic literature (and comments from companies that build
and sell automobiles) to defend the assertion that a market for fuel
economy exists, the industry's historical CAFE compliance performance
is a matter of public record.\167\ As shown in Figure IV-3, Figure IV-
4, and Figure IV-5 for more than a decade, the industry average CAFE
has exceeded the standard for each regulatory class--by several mpg
during periods of high fuel prices.
---------------------------------------------------------------------------
 \167\ Data from CAFE Public Information Center (PIC), https://one.nhtsa.gov/cafe_pic/CAFE_PIC_Home.htm, last accessed 10/08/2019.
---------------------------------------------------------------------------
BILLING CODE 4910-59-P
[[Page 24233]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.043
[[Page 24234]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.044
BILLING CODE 4910-59-C
 While this rulemaking has shown the impact of deviations from the
2012 rule assumptions individually, these two assumptions affect the
value of fuel savings jointly. Replacing the fuel price forecast with
the observed historical and current projected prices, and including any
technology that pays for itself in the first 2.5 years of ownership
through avoided fuel expenditures, reduces the value of fuel savings
from $525 billion in the 2012 rule to $140 billion, all else equal.
Interestingly, this reduction in the value of fuel savings is smaller
than the result when assuming only that the desired payback period is
nonzero. While it may seem counterintuitive, it is entirely consistent.
 The number of gallons saved under the preferred alternative is
actually higher when modifying both assumptions, compared to only
modifying the payback period. Updating both assumptions leads to about
100 billion gallons saved by the preferred alternative in 2012,
compared to only 50 billion from changing only the payback period, and
180 billion in the 2012 analysis. This occurs because the fuel economy
in the baseline is lower when updating both the fuel price and the
payback period--the gap between the augural standards and the baseline
grows to 9 mpg, rather than only 6 mpg when updating only the payback
period. Despite the existence of inexpensive
[[Page 24235]]
technology in both cases, with lower fuel prices there are fewer
opportunities to apply technology that will pay back quickly. As a
consequence, the number of gallons saved by the preferred alternative
in 2012 increases--but each gallon saved is worth less because the
price of fuel is lower.
b) Technology Cost
 While the methods used to identify cost-effective technologies to
improve fuel economy in new vehicles have continuously evolved since
2012 (as discussed further in Section IV.B.1), as have the estimated
cost of individual technologies, the inclusion of a market response in
all scenarios (including the baseline) has changed the total technology
cost associated with a given alternative. As also discussed in Section
VI.B, acknowledging the existence of a market for fuel economy leads to
continued application of the most cost-effective technologies in the
baseline--and in other less stringent alternatives--up to the point at
which there are no remaining technologies whose cost is fully offset by
the value of fuel saved in the first 30 months of ownership. The
application of this market-driven technology has implications for fuel
economy levels under lower stringencies (as discussed earlier), but
also for the incremental technology cost associated with more stringent
alternatives. As lower stringency alternatives (including the 2012
baseline) accrue more technology, the incremental cost of more
stringent alternatives decreases.
 By including a modest market for fuel economy, and preserving all
other assumptions from the 2012 final rule, the incremental cost of
technology attributable to the preferred alternative decreases from
about $140 billion to about $72 billion. This significant reduction in
technology cost is somewhat diminished by the associated reduction in
the value of fuel savings (a decrease of $385 billion) when
acknowledging the existence of a market for fuel economy. Another
consequence of these changes is that the incremental cost of fuel
economy technology is responsive to fuel price, as it should be. Under
higher prices (as were assumed in 2012), consumers demand higher fuel
economy in the new vehicle market. Under lower prices (as have occurred
since the 2012 rule) consumers demand less fuel economy than would have
been consistent with the fuel price assumptions in 2012.\168\ Including
a market response in the analysis ensures that, in each case, the cost
of fuel economy technology within an alternative is consistent with
those assumptions. Using the same fuel price forecast that supports
this rule, and the same estimate of market demand for fuel economy, the
incremental cost of technology in the preferred alternative would rise
back up to about $110 billion.
---------------------------------------------------------------------------
 \168\ This is why dozens of studies examining the ability of
fuel taxes (and carbon taxes, which produce the same result for
transportation fuels) to reduce CO2 emissions have found
cost-effective opportunities available for those pricing mechanisms.
---------------------------------------------------------------------------
c) The Social Cost of Carbon (SCC) Emissions
 As discussed extensively in the NPRM, the agencies' perspective
regarding the social cost of carbon has narrowed in focus. While the
2012 final rule considered the net present value of global damages
resulting from carbon emitted by vehicles sold in the U.S. between MY
2009 and MY 2025, the NPRM (and this final rule) consider only those
damages that occur to the United States and U.S. territories. As a
result of this change in perspective, the value of estimated damages
per-ton of carbon is correspondingly smaller. Had the 2012 final rule
utilized the same perspective on the social cost of carbon, the
benefits associated with the preferred alternative would have been
about $11 billion, rather than $53 billion. However, the savings
associated with carbon damages are a consequence of both the assumed
cost per-ton of damages and the number of gallons of fuel saved. As
discussed above, the gallons saved in the 2012 final rule were likely
inflated as a result of both fuel price forecasts and the assumption
that no market exists for fuel economy improvements. Correcting the
estimate of gallons saved from the preferred alternative in the 2012
rule and considering only the domestic social cost of carbon further
reduces the savings in carbon damages to $6 billion.
d) Safety Neutrality
 In the 2012 final rule, the agencies showed a ``safety neutral''
compliance solution; that is, a compliance solution that produced no
net increase in on-road fatalities for MYs 2017-2025 vehicles as a
result of technology changes associated with the preferred alternative.
In practice, safety neutrality was achieved by expressly limiting the
availability of mass reduction technology to only those vehicles whose
usage causes fewer fatalities with decreased mass. This result was
discussed as one possible solution, where manufacturers chose
technology solutions that limited the amount of mass reduction applied,
and concentrated the application on vehicles that improve the safety of
other vehicles on the roads (primarily by reducing the mass
differential in collisions). However, it implicitly assumed that each
and every manufacturer would leave cost-effective technologies unused
on entire market segments of vehicles in order to preserve a safety
neutral outcome at the fleet level for a given model year (or set of
model years) whose useful lives stretched out as far as the 2060s.
Removing these restrictions tells a different story.
 When mass reduction technology, determined in the model to be a
cost-effective solution (particularly in later model years, when more
advanced levels of mass reduction were expected to be possible), is
unrestricted in its application, the 2012 version of the CAFE Model
chooses to apply it to vehicles in all segments. This has a small
effect on technology costs, increasing compliance costs in the earliest
years of the program by a couple billion dollars, and reducing
compliance costs for MYs 2022--2025 by a couple billion dollars.
However, the impact on safety outcomes is more pronounced.
 Also starting with the model and inputs used for the 2012 final
rule (and, as an example, focusing on that rule's 2008-based market
forecast), removing the restrictions on the application of mass
reduction technology results in an additional 3,400 fatalities over the
full lives of MYs 2009-2025 vehicles in the baseline,\169\ and another
6,900 fatalities over those same vehicle lives under the preferred
alternative. The result, a net increase of 3,500 fatalities under the
preferred alternative relative to the baseline, also produces a net
social cost of $18 billion. The agencies' current treatment of both
mass reduction technology, which can greatly improve the effectiveness
of certain technology packages by reducing road load, and estimated
fatalities and now account for both general exposure (omitted in the
2012 final rule modeling) and fatality risk by age of the vehicle,
further changes the story around mass reduction technology application
for compliance and its relationship to on-road safety.
---------------------------------------------------------------------------
 \169\ Relative to the continuation of vehicle mass from the 2008
model year carried forward into the future.
---------------------------------------------------------------------------
2. What methods have changed since the 2012 final rule?
 Simulating how manufacturers might respond to CAFE/CO2
standards
[[Page 24236]]
requires information about existing products being offered for sale, as
well as information about the costs and effectiveness of technologies
that could be applied to those vehicles to bring the fleets in which
they reside into compliance with a given set of standards. Following
extensive additional work and consideration since the 2012 analysis,
both agencies now use the CAFE Model to simulate these compliance
decisions. This has several practical implications which are discussed
in greater detail in Section VI.A. Briefly, this change represents a
shift toward including a number of real-world production constraints--
such as component sharing across a manufacturer's portfolio--and
product cadence, where only a subset of vehicles in a given model year
are redesigned (and thus eligible to receive fuel economy technology).
Furthermore, the year-by-year accounting ensures a continuous evolution
of a manufacturer's product portfolio that begins with the market data
of an initial model year (model year 2017, in this analysis) and
continues through the last year for which compliance is simulated.
Finally, the modeling approach has migrated from one that relied on the
simple product of single values to estimate technology effectiveness to
a model that relies on full vehicle simulation to determine the
effectiveness of any combination of fuel economy technologies. The
combination of these changes has greatly improved the realism of
simulated vehicle fuel economy for combinations of technologies across
vehicle systems and classes.
 In addition to these changes to the portions of the analysis that
represent the supply of fuel economy (by manufacturer, fleet, and model
year) in the new vehicle market, this analysis contains changes to the
representation of consumer demand for fuel economy. One such measure
was discussed above--the notion that consumers will demand some amount
of fuel economy improvement over time, consistent with technology costs
and fuel prices. However, another deviation from the 2012 final rule
analysis reflects overall demand for new vehicles. Across ten
alternatives, ranging from the baseline (freezing future standards at
2016 levels) to scenarios that increased stringency by seven percent
per year, from 2017 through 2025, the 2012 analysis showed no response
in new vehicle sales, down to the individual model level. This implied
that, regardless of changes to vehicle cost or attributes driven by
stringency increases, no fewer (or possibly more) units of any single
model would be sold in any year, in any alternative. Essentially, that
analysis asserted that the new vehicle market does not respond, in any
way, to average new vehicle prices across the alternatives--regardless
of whether the incremental cost is $1,600/vehicle (as it was estimated
to be under the preferred alternative) or nearly $4,000/vehicle (as it
was in under the 7 percent alternative). Both the NPRM and this final
rule, while not employing pricing models or full consumer choice models
to address differentiated demand within brands or manufacturer
portfolios, have incorporated a modeled sales response that seeks to
quantify what was not quantified in previous rulemakings.
 An important accounting method has also changed since the 2012
final rule was published. At the time of that rule, the agencies used
an approach to discounting that combined attributes of a private
perspective and a social perspective in their respective benefit cost
analyses. This approach was logically inconsistent, and further
reinforced some of the exaggerated estimates of fuel savings, social
benefits (from reduced externalities), and technology costs described
above. The old method discounted the value of all incremental
quantities, whether categorized as benefits or costs, to the model year
of the vehicle to which they accrued. This approach is largely
acceptable for use in a private benefit cost analysis, where the costs
and benefits accrue to the buyer of a new vehicle (in the case of this
policy) who weighs their discounted present values at the time of
purchase. However, the private perspective would not include any costs
or benefits that are external to the buyer (e.g., congestion or the
social cost of carbon emissions). For an analysis that compares
benefits and costs from the social perspective, attempting to estimate
the relative value of a policy to all of society rather than just
buyers of new vehicles, this approach is more problematic.
 The discounting approach in the 2012 final rule was particularly
distortionary for a few reasons. The fact that benefits and costs
occurred over long time periods in the 2012 rule, and the standards
isolated the most aggressive stringency increases in the latter years
of the program, served to allow benefits that occurred in 2025 (for
example) to enter the accounting without being discounted, provided
that they accrued to the affected vehicles during their first year of
ownership. In a setting where numerous inputs (e.g., fuel price and
social cost of carbon) increase over time, benefits were able to grow
faster than the discount rate in some cases--essentially making them
infinite. The interpretation of discounting for externalities was
equally problematic. For example, the discounting approach in the 2012
final rule would have counted a ton of CO2 not emitted in CY
2025 in multiple ways, despite the fact that the social cost of carbon
emissions was inherently tied to the calendar year in which the
emissions occurred. Were the ton avoided by a MY 2020 vehicle, which
would have been five years old in CY 2025, the value of that ton would
have been the social cost of carbon times 0.86, but would have been
undiscounted if that same ton had been saved by a MY 2025 vehicle in
its initial year of usage.
 This approach was initially updated in the 2016 Draft TAR to be
consistent with common economic practice for benefit-cost analysis, and
this analysis continues that approach. In the social perspective, all
benefits and costs are discounted back to the decision year based on
the calendar year in which they occur. Had the agencies utilized such
an approach in the 2012 final rule, net benefits would have been
reduced by about 20 percent, from $465 billion to $374 billion--not
accounting for any of the other adjustments discussed above.
3. How have conditions changed since the 2012 final rule was published?
 The 2012 final rule relied on market and compliance information
from model year 2008 to establish standards for model years 2017-2025.
However, in the intervening years, both the market and the industry's
compliance positions have evolved. The industry has undergone a
significant degree of change since the MY 2008 fleet on which the
2012FR was based. Entire brands (Pontiac, Oldsmobile, Saturn, Hummer,
Mercury, etc.) and companies (Saab, Suzuki, Lotus) have exited the U.S.
market, while others (most notably Tesla) have emerged. Several dozen
nameplates have been retired and dozens of other created in that time.
Overall, the industry has offered a diverse set of vehicle models that
have generally higher fuel economy than the prior generation, and an
ever-increasing set of alternative fuel powertrains.
 As Table IV-1 shows, alternative powertrains have steadily
increased under CAFE/CO2 regulations. Under the standards
between 2011 and 2018, the number of electric vehicle offerings in the
market has increased from 1 model to 57 models (inclusive of all plug-
in vehicles that are rated for use on the highway), and hybrids (like
the Toyota Prius) have increased from 20 models to
[[Page 24237]]
43 models based on data from DOE's Alternative Fuels Data Center. Fuel
efficient diesel vehicles have similarly been on the rise in that
period, more than doubling the number of offerings. Flexible fuel
vehicles (FFVs), capable of operating on both gasoline and E85 remain
readily available in the market, but have been excluded from the table
due to both their lower fuel economy and demonstrated consumer
reluctance to operate FFVs on E85. They have historically been used to
improve a manufacturer's compliance position, rather than other
alternative fuel systems that reduce fuel consumption and save buyers
money.
[GRAPHIC] [TIFF OMITTED] TR30AP20.045
 Not only have alternative powertrain options proliferated since the
2012 FR, the average fuel economy of new vehicles within each body
style has increased. However, the more dramatic effect may lie in the
range of fuel economies available within each body style. Figure IV-6
shows the distribution of new vehicle fuel economy (in miles per gallon
equivalent) by body style for MYs 2008, 2016, and 2020 (simulated).
Each box represents the 25th and 75th percentiles, where 25 and 75
percent of new models offered are less fuel efficient than that level.
Not only has the median fuel economy improved (the median shows the
point at which 50 percent of new models are less efficient) under the
CAFE/CO2 programs, but the range of available fuel economies
(determined by the length of the boxes and their whiskers) has
increased as well. For example, the 25th percentile of pickup truck
fuel economy in 2020 is expected to be significantly more efficient
than 75 percent of the pickups offered in 2008. In MY 2008, there were
only a few SUVs offered with rated fuel economies above 34MPG. By MY
2020 almost half of the SUVs offered will have higher fuel economy
ratings--with almost 20 percent of offerings exceeding 40MPG.
 The improvement in passenger car styles has been no less dramatic.
As with the other styles, the range of available fuel economies has
increased under the CAFE/CO2 programs and the distribution
of available fuel economies skewed higher--with 40 percent of MY 2020
models exceeding 40MPG. The attribute-based standards are designed to
encourage manufacturers to improve vehicle fuel economy across their
portfolios, and they have clearly done so. Not only have the higher
ends of the distributions increased, the lower ends in all body styles
have improved as well, where the least fuel efficient 25 percent of
vehicles offered in MY 2016 (and simulated in 2020) are more fuel
efficient than the most efficient 25 percent of vehicles offered in MY
2008.
BILLING CODE 4910-59-P
[[Page 24238]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.046
BILLING CODE 4910-59-C
 Some commenters have argued that consumers will be harmed by any
set of standards lower than the baseline (augural) standards because
buyers of new vehicles will be forced to spend more on fuel than they
would have under the augural standards. However, as Figure IV-6 shows,
the range of fuel economies available in the new market is already
sufficient to suit the needs of buyers who desire greater fuel economy
rather than interior volume or some other attributes. Full size pickup
trucks are now available with smaller turbocharged engines paired with
8 and 10-speed transmissions and some mild electrification. Buyers
looking to transport a large family can choose to purchase a plug-in
hybrid minivan. There were 57 electric models available in 2018, and
hybrid powertrains are no longer limited to compact cars (as they once
were). Buyers can choose hybrid SUVs with all-wheel and four-wheel
drive. While these kinds of highly efficient options were largely
absent from some body styles in MY 2008, this is no longer the case.
Given that high-MPG vehicles are widely available, consumers must also
value other vehicle attributes (e.g., acceleration and load-carrying
capacity) that can can also be improved with the same technologies that
can be used to improve fuel economy.
---------------------------------------------------------------------------
 \170\ Circles represent specific outlying vehicle models.
---------------------------------------------------------------------------
 Manufacturers have accomplished a portfolio-wide improvement by
improving the combustion efficiency of engines (through direct
injection and
[[Page 24239]]
turbocharging), migrating from four and five speed transmissions to 8
and 10 speed transmissions, and electrifying to varying degrees. All of
this has increased both production costs and fuel efficiency during a
period of economic expansion and low energy prices. While the vehicles
offered for sale have increased significantly in efficiency between MY
2008 and MY 2020, the sales-weighted average fuel economy has achieved
less improvement. Despite stringency increases of about five percent
(year-over-year) between 2012 and 2016, the sales-weighted average fuel
economy increased marginally. Figure IV-7 shows an initial increase in
average new vehicle fuel economy (the heavy solid line, shown in mpg as
indicated on the left y axis), followed by relative stagnation as fuel
prices (the light dashed lines, shown in dollars per gallon as
indicated on the right y axis) fell and remained low.\171\ It is worth
noting that average new vehicle fuel economy observed a brief spike
during the year that the Tesla Model 3 was introduced (as a consequence
of strong initial sales volumes, as pre-orders were satisfied, and fuel
economy ratings that are significantly higher than the industry
average), and settled around 27.5 MPG in Fall 2019. Average fuel
economy receded further over the next several months to 26.6 MPG in
February 2020.\172\
---------------------------------------------------------------------------
 \171\ Ward's Automotive, https://www.wardsauto.com/industry/fuel-economy-index-shows-slow-improvement-april. Last accessed Dec.
13, 2019.
 \172\ Ward's Automotive, https://wardsintelligence.informa.com/WI964622/Fuel-Economy-Slightly-Down-in-February. Last accessed Mar.
9, 2020.
[GRAPHIC] [TIFF OMITTED] TR30AP20.047
 In their NPRM comments, manufacturers expressed concern that CAFE
standards had already increased to the point where the price increases
necessary to recoup manufacturers' increased costs for providing
further increases in fuel economy outweigh the value of fuel
savings.173 174 The agencies do not agree that this point
has already been reached by previous stringency increases, but
acknowledge the reality of diminishing marginal returns to improvements
in fuel economy. A driver with a 40MPG vehicle uses about 300 gallons
of fuel per year. Increasing the fuel economy of that vehicle to 50MPG,
a 25 percent increase, would likely be over $1000 in additional
technology cost. However, that driver would only save 25 percent of
their annual fuel consumption, or 75 gallons out of 300 gallons. Even
at $3/gallon, higher than the current national average, that represents
$225 per year in fuel savings. That means that the buyer's $1000
investment in additional fuel economy pays back in just under 4.5 years
(undiscounted). The agencies' respective programs have created greater
access to high MPG vehicles in all classes and encouraged the
proliferation of alternative fuels and powertrains. But if the value of
the fuel savings is insufficient to motivate buyers to invest in ever
greater levels of fuel economy, manufacturers will face challenges in
the market.
---------------------------------------------------------------------------
 \173\ NHTSA-2018-0067-12064-25.
 \174\ NHTSA-2018-0067-12073-2.
---------------------------------------------------------------------------
 While Figure IV-3 through Figure IV-5 illustrate the trends in
historical CAFE compliance for the entire industry, the figures contain
another relevant fact. After several consecutive years of increasing
standards, the achieved and required levels converge. When the
standards began increasing again for passenger cars in 2011, the prior
year had industry CAFE levels 5.6 mpg and 7.7 mpg in excess of their
standards for domestic cars and imported cars, respectively. Yet, by
2016, the consecutive year-over-year increases had eroded the levels of
over-compliance. Light trucks similarly exceeded their standard prior
to increasing standards, which began in 2005. Yet, by 2011, after
several consecutive years of stringency increases, the industry light-
truck average CAFE was merely compliant with the rising standard.
 This is largely due to the fact that stringency requirements have
increased at a faster rate than achieved fuel
[[Page 24240]]
economy levels for several years. The attribute-based standards took
effect in 2011 for all regulatory classes, although light truck CAFE
standards had been increasing since 2005. Since 2004, light truck
stringency has increased an average of 2.7 percent per year, while
light truck's compliance fuel economy has increased by an average of
1.7 percent over the same period.\175\ For the passenger classes, a
similar story unfolds over a shorter period of time. Year over year
stringency increases have averaged 4.7 percent per year for domestic
cars (though increases in the first two years were about 8 percent--
with lower subsequent increases), but achieved fuel economy increases
averaged only 2.2 percent per year over the same period. Imported
passenger cars were similar to domestic cars, with average annual
stringency increases of 4.4 percent but achieved fuel economy levels
increasing an average of only 1.4 percent per year from 2011 through
2017. Given that each successive percent increase in stringency is
harder to achieve than the previous one, long-term discrepancies
between required and achieved year-over-year increases cannot be offset
indefinitely with existing credit banks, as they have been so far.
---------------------------------------------------------------------------
 \175\ Both the standards and these calculations are defined in
consumption space--gallons per mile--which also translates directly
into CO2 based on the carbon content of the fuel
consumed.
---------------------------------------------------------------------------
 With the fuel price increases fresh in the minds of consumers, and
the great recession only recently passed, the CAFE stringency increases
that began in 2011 (and subsequent CAFE/CO2 stringency
increases after EPA's program was first enforced in MY 2012) had
something of a head start. As Figure IV-3 through Figure IV-5
illustrate, the standards were not binding in MY 2011--even
manufacturers that had historically paid civil penalties were earning
credits for overcompliance. It took two years of stringency increase to
catch up to the CAFE levels already present in MY 2011. However, seven
consecutive years of increases for passenger cars and a decade of
increases for light trucks has changed the credit situation. Figure IV-
8 shows CAFE credit performance for regulated fleets--the solid line
represents the number of fleets generating shortfalls and the dashed
line represents the number of fleets earning credits in each model
year.
[GRAPHIC] [TIFF OMITTED] TR30AP20.048
 Fewer than half as many fleets earned surplus credits for over-
compliance in MY 2017 compared to MY 2011--and this trend is
persistent. The story varies from one manufacturer to another, but it
seems sufficient to state the obvious--when the agencies conducted the
analysis to establish standards through MY 2025 back in 2012, most (if
not all) manufacturers had healthy credit positions. That is no longer
the case, and each successive increase requires many fleets to not only
achieve the new level from the resulting increase, but to resolve
deficits from the prior year as well. The large sums of credits, which
last five years under both programs, have allowed most manufacturers to
resolve shortfalls. But the light truck fleet, in particular, has a
dwindling supply of credits available for purchase or trade. The
CO2 program has a provision that allows credits earned
during the early years of over-compliance to be applied through MY
2021. This has reduced the compliance burden in the last several years,
as intended, but will not mitigate the compliance challenges some OEMs
would face if the baseline standards remained in place and energy
prices persisted at current levels.
BILLING CODE 4910-59-P
[[Page 24241]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.049
BILLING CODE 4910-59-C
 Table IV-2 shows the credits earned by each manufacturer over
time.\176\ As the table shows, when the agencies considered future
standards in 2012, most manufacturers were earning credits in at least
one fleet. However, the bold values show years with deficits and even
some manufacturers who started out in strong positions, such as Ford's
passenger car fleet, have seen growing deficits in recent years. While
[[Page 24242]]
the initial banks for early-action years eases the burden of
CO2 compliance for many OEMs, the year-to-year compliance
story is similar to CAFE, see Table IV-3.
---------------------------------------------------------------------------
 \176\ MY 2017 values represent estimated earned credits based on
MY 2017 final compliance data.
[GRAPHIC] [TIFF OMITTED] TR30AP20.050
 Credit position and shortfall rates clearly illustrate
manufacturers' fleet performance relative to the standards. Recognizing
that manufacturers plan compliance over several model years at any
given time, sporadic shortfalls may not be evidence of undue
difficulty, but sustained, widespread, growing shortfalls should
probably be viewed as evidence that standards previously believed to be
manageable might no longer be so. While NHTSA is prohibited by statute
from considering availability of credits (and thus, size of credit
banks) in determining maximum feasible standards, it does consider
shortfalls as part of its determination. EPA has no such prohibition
under the CAA and is free to consider both credits and shortfalls.
 These increasing credit shortfalls are occurring at a time that the
industry is deploying more technology than the agencies anticipated
when establishing future standards in 2012. The agencies' projections
of transmission technologies were mixed. While the agencies expected
the deployment of 8-speed transmissions to about 25 percent of the
market by MY 2018, transmissions with eight or more gears account for
almost 30 percent of the market. However, the agencies projected no CVT
transmissions in future model years, instead projecting high
penetration of DCTs. However, CVTs currently make up more than 20
percent of new transmissions. The tradeoff between advanced engines and
electrification was also underestimated. While the agencies projected
penetration rates of turbocharged engines that are higher than we've
observed in the market (45 percent compared to 30 percent), the
estimated penetration of electric technologies was significantly lower.
The agencies projected a couple percent of strong hybrids--which we've
seen--but virtually no PHEVs or EVs. While the volumes of those
vehicles are still only a couple percent of the new vehicle market,
they are heavily credited under both programs and can significantly
improve compliance positions even at smaller volumes. Even lower-level
electrification technologies, like stop-start systems, are
significantly more prevalent than we anticipated (stop-start systems
were projected to be in about 2 percent of the market, compared to over
20 percent in the 2018 fleet). Despite technology deployment that is
comparable to 2012 projections, and occasionally more aggressive,
passenger car and light truck fleets have slightly lower fuel economy
than projected. As fleet volumes have shifted along the footprint
curve, the standards have decreased as well (relative to the
expectation in 2012), but less so. While compliance deficits have been
modest, they have been accompanied by record sales for several years.
This has not only depleted existing credit banks, but created
significant shortfalls that may be more difficult to overcome if sales
recede from record levels.
V. Regulatory Alternatives Considered
 Agencies typically consider regulatory alternatives in proposals as
a way of evaluating the comparative effects of different potential ways
of accomplishing their desired goal. NEPA
[[Page 24243]]
requires agencies (in this case, NHTSA, but not EPA) to compare the
potential environmental impacts of their proposed actions to those of a
reasonable range of alternatives. Executive Orders 12866 and 13563 and
OMB Circular A-4 also encourage agencies to evaluate regulatory
alternatives in their rulemaking analyses. Alternatives analysis begins
with a ``no-action'' alternative, typically described as what would
occur in the absence of any regulatory action. This final rule, like
the proposal, includes a no-action alternative, described below, as
well as seven ``action alternatives.'' The final standards may, in
places, be referred to as the ``preferred alternative,'' which is NEPA
parlance, but NHTSA and EPA intend ``final standards'' and ``preferred
alternative'' to be used interchangeably for purposes of this
rulemaking.
 In the proposal, NHTSA and EPA defined the different regulatory
alternatives (other than the no-action alternative) in terms of
percent-increases in CAFE and CO2 stringency from year to
year. Percent increases in stringency referred to changes in the
standards year over year--as in, standards that become 1 percent more
stringent each year. Readers should recognize that those year-over-year
changes in stringency are not measured in terms of mile per gallon or
CO2 gram per mile differences (as in, 1 percent more
stringent than 30 miles per gallon in one year equals 30.3 miles per
gallon in the following year), but in terms of shifts in the footprint
functions that form the basis for the actual CAFE and CO2
standards (as in, on a gallon or gram per mile basis, the CAFE and
CO2 standards change by a given percentage from one model
year to the next). Under some alternatives, the rate of change was the
same for both passenger cars and light trucks; under others, the rate
of change differed. Like the no-action alternative, all of the
alternatives considered in the proposal were more stringent than the
preferred alternative.
 Alternatives considered in the proposal also varied in other
significant ways. Alternatives 3 and 7 in the proposal involved a
gradual discontinuation of CAFE and average CO2 adjustments
reflecting the use of technologies that improve air conditioner
efficiency or otherwise improve fuel economy under conditions not
represented by long-standing fuel economy test procedures (off-cycle
adjustments, described in further detail in Section IX, although the
proposal itself would have retained these flexibilities. Commenters
responding to the request for comment on phasing out these
flexibilities generally supported maintaining the existing program, as
proposed. Some commenters suggested changes to the existing program
that were not discussed in the NPRM. Such changes would be beyond the
scope of this rulemaking and were not considered. Section IX contains a
more thorough summary of these comments and the issues they raise, as
well as the agencies' responses. Consistent with the decision to retain
these flexibilities in the final rule, alternatives reflecting their
phase-out have not been considered in this final rule.
 Additionally, in the NPRM for this rule, EPA proposed to exclude
the option for manufacturers partially to comply with tailpipe
CO2 standards by generating CO2-equivalent
emission adjustments associated with air conditioning refrigerants and
leakage after MY 2020. This approach was proposed in the interest of
improved harmonization between the CAFE and tailpipe CO2
emissions programs because this optional flexibility cannot be
available in the CAFE program.\177\ Alternatives 1 through 8 excluded
this option. EPA requested comment ``on whether to proceed with [the]
proposal to discontinue accounting for A/C leakage, methane emissions,
and nitrous oxide emissions as part of the CO2 emissions
standards to provide for better harmony with the CAFE program, or
whether to continue to consider these factors toward compliance and
retain that as a feature that differs between the programs.'' \178\ EPA
stated that if EPA were to proceed with excluding A/C refrigerant
credits as proposed, ``EPA would consider whether it is appropriate to
initiate a new rulemaking to regulate these programs independently . .
. .'' \179\ EPA also stated that ``[i]f the agency decides to retain
the A/C leakage . . . provisions for CO2 compliance, it
would likely re-insert the current A/C leakage offset and increase the
stringency levels for CO2 compliance by the offset amounts
described above (i.e., 13.8 g/mi equivalent for passenger cars and 17.2
g/mi equivalent for light trucks). EPA received comments from a wide
range of stakeholders, most of whom opposed the elimination of these
flexibility provisions.
---------------------------------------------------------------------------
 \177\ For the CAFE program, carbon-based tailpipe emissions
(including CO2, HC, and CO) are measured, and fuel
economy is calculated using a carbon balance equation. EPA also uses
carbon-based emissions (CO2, HC, and CO, the same as for
CAFE) to calculate tailpipe CO2 for use in determining
compliance with its standards. In addition, under the no-action
alternative for the proposal and under all alternatives in the final
rule, in determining compliance, EPA includes on a CO2
equivalent basis (using Global Warming Potential (GWP) adjustment)
A/C refrigerant leakage credits, at the manufacturer's option, and
nitrous oxide and methane emissions. EPA also has separate emissions
standards for methane and nitrous oxides. The CAFE program does not
include or account for A/C refrigerant leakage, nitrous oxide and
methane emissions because they do not impact fuel economy. Under
Alternatives 1-8 in the proposal, the standards were closely aligned
for gasoline powered vehicles because compliance with the fleet
average standard for such vehicles is based on tailpipe
CO2, HC, and CO for both programs and not emissions
unrelated to fuel economy, although diesel and alternative fuel
vehicles would have continued to be treated differently between the
CAFE and CO2 programs. While such an approach would have
significantly improved harmonization between the programs, standards
would not have been fully aligned because of the small fraction of
the fleet that uses diesel and alternative fuels (as described in
the proposal, such vehicles made up approximately four percent of
the MY 2016 fleet), as well as differences involving EPCA/EISA
provisions EPA has not adopted, such as minimum standards for
domestic passenger cars and limits on credit transfers between
regulated fleets. The proposal to eliminate flexibilities associated
with A/C refrigerants and leakage was not adopted for this final
rule, and the reasons for and implications of that decision are
discussed further below.
 \178\ 83 FR at 43193 (Aug. 24, 2018).
 \179\ Id. at 43194.
---------------------------------------------------------------------------
 Specifically, the two major trade organizations of auto
manufacturers, as well as some individual automakers, supported
retaining these provisions. Global Automakers commented that ``[a]ir
conditioning refrigerant leakage . . . should be included for
compliance with the EPA standards for all MYs, even if it means a
divergence from the NHTSA standards.'' \180\ Global provides several
detailed reasons for their comments, including that the existing
provisions are ``. . . important to maintaining regulatory flexibility
through real [CO2] emission reductions and would prevent the
potential for additional bifurcated, separate programs at the state
level.'' \181\ The Alliance similarly commented that it ``supports
continuation of the full air conditioning refrigerant leakage credits
under the [CO2] standards.'' \182\ Some individual
[[Page 24244]]
manufacturers, including General Motors,\183\ Fiat Chrysler,\184\ and
BMW,\185\ also commented in support of maintaining the current A/C
refrigerant and leakage credits.
---------------------------------------------------------------------------
 \180\ Global, NHTSA-2018-0067-12032, Appendix A at A-5.
 \181\ Id. Global also stated that excluding A/C leakage credits
would ``. . . greatly limit the ability [of manufacturers] to select
the most cost-effective approach for technology improvements and
result in a costlier, separate set of regulations that actually
relate to the overall GHG standards.'' Global also expressed concern
that issuing separate regulations for A/C leakage could take too
long and create a gap in which States might act to separately
regulate or even ban refrigerants, and supported continued inclusion
of A/C leakage and refrigerant regulation in EPA's GHG program to
avoid risk of an ensuing patchwork. Global argued that manufacturers
had already invested to meet the existing program, and that ``the
proposed phase-out also creates another risk that manufacturers will
have stranded capital in technologies that are not fully
amortized.'' Global Automakers, EPA-HQ-OAR-2018-0283-5704,
Attachment A, at A.43-44.
 \182\ Alliance, NHTSA-2018-0067-12073, Full Comment Set, at 12.
Alliance also expressed concern about stranded capital and risk of
patchwork of state regulation if MAC direct credits were not
retained in the Federal GHG program. Id. at 80-81.
 \183\ General Motors, NHTSA-2018-0067-11858, Appendix 4, at 1
(``General Motors supports the extensive comments from the Alliance
of Automobile Manufacturers regarding flexibility mechanisms, and
incorporates them by reference. In particular, the Alliance cites
the widening gap between the regulatory standards and actual
industry-wide new vehicle average fuel economy that has become
evident since 2016, despite the growing use of improvement `credits'
from various flexibility mechanisms, such as off-cycle technology
credits, mobile air conditioner efficiency credits, mobile air
conditioner refrigerant leak reduction credits and credits from
electrified vehicles.'')
 \184\ FCA, NHTSA-2018-0067-11943, at 8. FCA also expressed
concern about patchwork in the absence of a federal rule. Id.
 \185\ BMW, EPA-HQ-OAR-2018-4204, at 3. BMW stated that ``Today's
rules allow flexibilities to be used by the motor vehicle
manufacturers for fuel saving technologies and efficiency gains
which are not covered in the applicable test procedures. To enhance
the future use of these technologies and to reward motor vehicle
manufacturer's investments taken for future innovations, the
agencies should consider the continuation of current flexibilities
for the model years 2021 to 2026.''
---------------------------------------------------------------------------
 Auto manufacturing suppliers who addressed A/C refrigerant and
leakage credits also generally supported retaining the existing
provisions. MEMA commented that ``It is essential for supplier
investment and jobs, and continuous innovation and improvements in the
technologies that the credit programs continue and expand to broaden
the compliance pathways. MEMA urges the agencies to continue the
current credit and incentives programs . . . . '' \186\ DENSO also
supported maintaining the current provisions.\187\ However, BorgWarner
supported the proposed removal of A/C refrigerant credits ``for
harmonization reasons,'' while encouraging EPA to regulate A/C
refrigerants and leakage separately from the CO2
standards.\188\
---------------------------------------------------------------------------
 \186\ MEMA, available at https://www.mema.org/sites/default/files/resource/MEMA%20CAFE%20and%20GHG%20Vehicle%20Comments%20FINAL%20with%20Appendices%20Oct%2026%202018.pdf, comment at p. 2. MEMA also expressed
concern about stranded capital investments by suppliers and supplier
jobs if the direct MAC credits were not available; stated that the
credits were an important compliance flexibility and ``one of the
highest values of any credit offered in the EPA program;'' and
stated that ``Harmonizing the programs does not require making them
identical or equivalent. Rather, harmonization can be achieved by
better coordinating the two programs to the extent feasible while
allowing each agency to implement its separate and distinct
mandate.'' Id. at 15-16.
 \187\ DENSO, NHTSA-2018-0067-11880, at 8.
 \188\ BorgWarner, NHTSA-2018-0067-11895, at 10.
---------------------------------------------------------------------------
 The two producers of a lower GWP refrigerant, Chemours and
Honeywell, commented extensively in support of continuing to allow A/C
refrigerant and leakage credits for CO2 compliance, making
both economic and legal arguments. Both Chemours and Honeywell stated
that A/C refrigerant and leakage credits were a highly cost-effective
way for OEMs to comply with the CO2 standards,\189\ with
Chemours suggesting that OEM compliance strategies are based on the
assumption that these credits will be available for CO2
compliance \190\ and that any increase in stringency above the proposal
effectively necessitates that the credits remain part of the
program.\191\ Honeywell stated that all OEMs (and a variety of other
parties) supported retaining the credits for CO2
compliance,\192\ and Chemours, Honeywell, and CBD et al. all noted that
OEMs are already using the credits for low GWP refrigerants in more
than 50 percent of the MY 2018 vehicles produced for sale in the
U.S.\193\ The American Chemistry Council also stated that the ``auto
industry widely supports the credits, and U.S. chemical manufacturers
are at a loss as to why EPA would propose to eliminate such a
successful flexible compliance program.'' \194\ In response to NPRM
statements expressing concern that the A/C refrigerant and leakage
credits could be market distorting, both Chemours and Honeywell
disagreed,\195\ arguing that the credits were simply a highly cost-
effective means of complying with the CO2 standards,\196\
and that removal of the credits at this point would, itself, distort
the market for refrigerants. Honeywell argued that eliminating the A/C
credit program from CO2 compliance would put the U.S. at a
competitive disadvantage with other countries, and would risk U.S.
jobs.\197\
---------------------------------------------------------------------------
 \189\ Chemours at 1 (``MVAC credits many times offer the `least
cost' approach to compliance . . .'') and 9; Honeywell at 6.
 \190\ Chemours at 6-7; both Chemours and Honeywell expressed
concern about OEM reliance on the expectation that HFC credits would
continue to be part of the CO2 program (Chemours at 31;
Honeywell at 16-20) and that investments in alternative refrigerants
would be stranded (Chemours at 1, 3, 4-6; Honeywell at 2, 7-8).
 \191\ Chemours at 7.
 \192\ Honeywell at 8-11.
 \193\ Chemours at 4; Honeywell at 6-7; CBD et al. at 46-47.
 \194\ American Chemistry Council, EPA-HQ-OAR-2018-0283-1415, at
9-10 (comments similar to Chemours and Honeywell).
 \195\ Chemours at 1; Honeywell at 13.
 \196\ Chemours at 29-30; Honeywell at 13-14.
 \197\ Honeywell at 20-21.
---------------------------------------------------------------------------
 Regarding the NPRM's statements that removing the A/C refrigerant
and leakage credits from CO2 compliance would promote
harmonization with the CAFE program, these commenters argued that
harmonization was not a valid basis for that aspect of the proposal.
Chemours, Honeywell, and CBD et al. all argued that Section 202(a)
creates no obligation to harmonize the [CO2] program with
the CAFE program.\198\ Chemours further argued that to the extent
disharmonization between the programs existed, it should be addressed
via stringency changes (i.e., reducing CAFE stringency relative to
CO2 stringency) rather than ``dropping low-cost compliance
options.'' \199\
---------------------------------------------------------------------------
 \198\ Chemours at 23-24; Honeywell at 11-12; CBD et al. at 47.
 \199\ Chemours at 9-11.
---------------------------------------------------------------------------
 These commenters also expressed concern that the proposal
constituted an EPA decision not to regulate HFC emissions from motor
vehicles at all. Commenters argued that the NPRM provided no legal
analysis or reasoned explanation for stopping regulation of HFCs,\200\
and that Massachusetts v. EPA requires any final rule to regulate all
greenhouse gases from motor vehicles and not CO2 alone,\201\
suggesting that there was a high likelihood of a lapse in regulation
because EPA had not yet proposed a new way of regulating HFC
emissions.\202\ Because the NPRM provided no specific information about
how EPA might regulate non-CO2 emissions separately,
commenters argued that the lack of clarity was inherently disruptive to
OEMs.\203\ CBD et al. argued that any lapse in regulation is ``illegal
on its face'' and that even creating a separate standard for HFC
emissions would be ``illegal'' because it ``would increase costs to
manufacturers and result in environmental detriment by removing any
incentive to use the most aggressive approaches to curtail emissions of
these highly potent GHGs.'' \204\
---------------------------------------------------------------------------
 \200\ Chemours at 1-2; Honeywell at 11.
 \201\ Chemours at 18-19; Honeywell at 14-16.
 \202\ Chemours at 6; Honeywell at 16.
 \203\ Chemours at 21; Honeywell at 16; ICCT at I-39.
 \204\ CBD et al. at 46.
---------------------------------------------------------------------------
 Environmental organizations,\205\ other NGOs, academic
institutions, consumer organizations, and state governments \206\ also
commented in support of continuing the existing provisions.
---------------------------------------------------------------------------
 \205\ ICCT, NHTSA-2018-0067-11741, Full Comments, at 4
(describing ``air conditioning GHG-reduction technologies [as]
available, cost-effective, and experiencing increased deployment by
many companies due to the standards.''); CBD et al., Appendix A, at
45-47.
 \206\ CARB, NHTSA-2018-0067-11873, Detailed Comments, at 120-
121; Washington State Department of Ecology, NHTSA-2018-0067-11926,
at 6 (HFCs are an important GHG; compliance flexibility is
important).
---------------------------------------------------------------------------
 EPA has considered its proposed approach to A/C refrigerant and
leakage
[[Page 24245]]
credits in light of these comments. EPA believes that maintaining this
element of its program is consistent with EPA's authority under Section
202(a) to establish standards for reducing emissions from LDVs. Thus,
maintaining the optional HFC credit program is appropriate. In
addition, EPA recognizes the value of regulatory flexibility and
compliance options, and has concluded that the advantages from
retaining the existing A/C refrigerant/leakage credit program and
associated offset between the CO2 and CAFE standards--in
terms of providing for a more-comprehensive regulation of emissions
from light-duty vehicles--outweigh the disadvantages resulting from the
lack of harmonization.
 Regarding the comment from BorgWarner about how having a separate
A/C refrigerant and leakage regulation would allow for better
harmonization between the programs, the agencies accept this to be an
accurate statement, but believe the benefits of continued refrigerant
regulation as an option for CO2 compliance outweigh the
problems associated with lack of harmonization with the CAFE program.
 For these reasons, EPA is not finalizing the proposed provisions,
and is making no changes in the A/C refrigerant and leakage-related
provisions of the current program. In light of this conclusion, EPA
does not need to address the legal arguments made by CBD et al. and
CARB about regulating refrigerant-related emissions separately, or
potential lapses in regulation of refrigerant emissions while such a
program could be developed.
 As with A/C refrigerant and leakage credits, EPA proposed to
exclude nitrous oxide and methane from average performance calculations
after model year 2020, thereby removing these optional program
flexibilities. Alternatives 1 through 8 excluded this option. EPA
sought comment on whether to remove those aspects of the program that
allow a manufacturer to use nitrous oxide and methane emissions
reductions for compliance with its CO2 average fleet
standards because such a flexibility is not allowed in the NHTSA CAFE
program, or whether to retain the flexibilities as a feature that
differs between the programs. Further, EPA sought comment on whether to
change the existing methane and nitrous oxide standards. Specifically,
EPA requested information from the public on whether the existing
standards are appropriate, or whether they should be revised to be less
stringent or more stringent based on any updated data.
 The Alliance in its comments may have misunderstood EPA's proposal
to mean that EPA was proposing to eliminate regulation of methane and
nitrous oxide emissions altogether. The Alliance commented in support
of such a proposal as they understood it, to eliminate the standards to
provide better harmony between the two compliance programs.\207\ The
Alliance commented that ``[n]ot only is emission of these two
substances from vehicles a relatively minor contribution to GHG
emissions, the Alliance has continuing concern regarding measurement
and testing technologies for nitrous oxide.'' \208\ The Alliance
commented further that if ``EPA decides instead to continue to regulate
methane and nitrous oxide, the Alliance recommends that EPA re-assess
whether the levels of the standards remain appropriate and to retain
the current compliance flexibilities. Furthermore, in this scenario,
the Alliance also recommends that methane and nitrous oxide standards
be assessed as a fleet average and as the average of FTP and HFET test
cycles.'' \209\ Several individual manufacturers submitted similar
comments, including Ford,\210\ FCA,\211\ Volvo,\212\ and Mazda.\213\
Ford also commented that it does not support the proposal to maintain
the existing N2O/CH4 standards while removing the
program flexibilities.\214\
---------------------------------------------------------------------------
 \207\ Alliance, NHTSA-2018-0067-12073, Full Comment Set, at 13.
 \208\ Id.
 \209\ Id.
 \210\ Ford, EPA-HQ-OAR-2018-0283-5691, at 4.
 \211\ FCA, NHTSA-2018-0067-11943, at 9.
 \212\ Volvo, NHTSA-2018-0067-12036, at 5.
 \213\ Mazda, NHTSA-2018-0067-11727, at 3 (``In reality, these
emissions are at deminimis levels and have very little, if any,
impact on global warming. So, the need to regulate these emissions
as part of the GHG program, or separately, is unclear. Although most
current engines can comply with the existing requirements, there are
some existing and upcoming new technologies that may not be able to
fully comply. These technologies can provide substantial
CO2 reductions.'').
 \214\ Ford, at 4 (``Finally, without the ability to incorporate
exceedances into CREE, each vehicle will need to employ hardware
solutions if they do not comply. We do not believe it was EPA's
intent in the original rulemaking to require additional after-
treatment, with associated cost increases, explicitly for the
control and reduction of an insignificant contributor to GHG
emissions. Therefore, we do not support the proposal to maintain the
existing N2O/CH4 standards while removing the
CREE exceedance pathway.'').
---------------------------------------------------------------------------
 The Alliance further commented that ``data from the 2016 EPA report
on light-duty vehicle emissions supports the position that
CH4 and N2O have minimal impact on total GHG
emissions, reporting only 0.045 percent in exceedance of the standard.
This new information makes it apparent that CH4 and
N2O contribute a de minimis amount to GHG emissions.
Additionally, gasoline CH4 and N2O performance is
within the current standards. Finally, the main producers of
CH4 and N2O emissions are flex fuel (E85) and
diesel vehicles, and these vehicles have been declining in sales as
compared to gasoline-fueled vehicles.'' \215\ The Alliance also
commented that CH4 and N2O have minimal
opportunities to be catalytically treated, as N2O is
generated in the catalyst and CH4 has a low conversion
efficiency compared to other emissions. EPA did not intend that
additional hardware should be required to comply with the
CH4 or N2O standards on any vehicle.'' \216\
---------------------------------------------------------------------------
 \215\ Alliance, NHTSA-2018-0067-12073, Full Comment Set, at 43.
 \216\ Id. at 44.
---------------------------------------------------------------------------
 Global Automakers commented in support of continuing inclusion of
nitrous oxide and methane emissions standards for all MYs, even if it
means a divergence from the NHTSA standards for these program elements
in the regulations, ``because they are complementary to EPA's program,
and are better managed through a coordinated federal policy. They are
also important to maintaining regulatory flexibility through real
[CO2] emission reductions and would prevent the potential
for additional bifurcated, separate programs at the state level.''
\217\ Global Automakers recommended that they remain in place per the
existing program but continued to support that the N2O
testing is not necessary. Global Automakers commented that it
``strongly recommends reducing the need for N2O testing or
eliminating these test requirements in their entirety. It should be
sufficient to allow manufacturers to attest to compliance with the
N2O capped standards based upon good engineering judgment,
development testing, and correlation to NOX emissions. EPA
could, however, maintain the option to request testing to be performed
for new technologies only, which could have unknown impacts on
N2O emissions.'' \218\ Hyundai \219\ and Kia \220\ submitted
similar comments.
---------------------------------------------------------------------------
 \217\ Global, NHTSA-2018-0067-12032, at 4, 5.
 \218\ Global, Appendix A, NHTSA-2018-0067-12032, at A-44, fn.
89.
 \219\ Hyundai, EPA-HQ-OAR-2018-0283-4411, at 7.
 \220\ Kia, EPA-HQ-OAR-2018-0283-4195, at 8-9.
---------------------------------------------------------------------------
 Others commented in support of retaining the existing program. MECA
commented that it supports the existing standards for methane and
nitrous oxide because catalyst technologies provided by MECA members
that reduce these climate forcing gases are readily
[[Page 24246]]
available and cost-effective.\221\ MECA also commented that the ability
to trade reductions in these pollutants in exchange for CO2
gives vehicle manufacturers the flexibilities they need to comply with
the emission limits by the most cost-effective means.\222\ CBD et al.
commented that the alternative compliance mechanisms currently
available in the program exist to provide cost-effective options for
compliance, and were considered by manufacturers to be a necessary
element of the program for certain types of vehicles.\223\ CBD et al.
further argued that ``[e]liminating these flexibilities consequently
imposes costs on manufacturers without discernible environmental
benefits,'' and suggested that harmonization with the CAFE program was
not a relevant decision factor for EPA.\224\ Several other parties
commented generally in support of retaining the existing program for A/
C leakage credits, discussed above, and N2O and
CH4 standards.\225\
---------------------------------------------------------------------------
 \221\ MECA, NHTSA-2018-0067-11994, at 12.
 \222\ Id.
 \223\ CBD et al. at 48.
 \224\ Id.
 \225\ Washington State Department of Ecology, NHTSA-2018-0067-
11926, at 6.
---------------------------------------------------------------------------
 After considering these comments, EPA is retaining the regulatory
provisions related to the N2O and CH4 standards
with no changes, specifically including the existing flexibilities that
accompany those standards. EPA is not adopting its proposal to exclude
nitrous oxide and methane emissions from average performance
calculations after model year 2020 or any other changes to the program.
The standards continue to serve their intended purpose of capping
emissions of those pollutants and providing for more-comprehensive
regulation of emissions from light-duty vehicles. The standards were
intended to prevent future emissions increases, and these standards
were generally not expected to result in the application of new
technologies or significant costs for manufacturers using current
vehicle designs.\226\ The program flexibilities are working as intended
and all manufacturers are successfully complying with the standards.
Most vehicle models are well below the standards and for those that are
above the standards, manufacturers have used the flexibilities to
offset exceedances with CO2 improvements to demonstrate
compliance. EPA did not receive any data in response to its request for
comments supporting potential alternative levels of stringency.
---------------------------------------------------------------------------
 \226\ 77 FR 62624, at 62799 (Oct 15, 2012).
---------------------------------------------------------------------------
 While the Alliance and several individual manufacturers recommended
eliminating the standards altogether, EPA did not propose to eliminate
the standards, but to eliminate the optional flexibilities, and
solicited comment on adjusting the standards to be more or less
stringent. Thus, EPA does not believe it would be appropriate to
eliminate completely the standards in this final rule without providing
an opportunity for comment on that idea. Furthermore, as noted above,
EPA believes the standards are continuing to serve their intended
purpose of capping emissions and remain appropriate. Manufacturers have
been subject to the standards for several years, and the Alliance
acknowledges in their comments that the exceedance of the standards,
which is offset by manufacturers using compliance flexibilities, is
very small and that most vehicles meet the standards. Regarding the
Alliance comments that the standards should be based on a fleet average
approach, EPA notes that the purpose of the standards is to cap
emissions, not to achieve fleet-wide reductions.\227\ The fleet average
emissions for N2O and CH4 are well below the
numerical level of the cap standards and therefore the existing cap
standards would not be an appropriate fleet average standard. Adopting
a fleet average approach using the same numerical level as the
established cap standards would not achieve the intended goal of
capping emissions at current levels. If technologies lead to
exceedances of the caps, automakers have the opportunity to apply
appropriate flexibilities under the current program to achieve GHG
emission neutrality. EPA is not aware of any manufacturer that has been
prevented from bringing a technology to the marketplace because of the
current cap levels or approach. EPA believes it would need to consider
all options further, with an opportunity for public comment, before
adopting such a significant change to the program.
---------------------------------------------------------------------------
 \227\ Relatedly, the Alliance and Global Automakers raised
concerns in their comments regarding N2O measurement and
testing burden. EPA did not propose any changes in testing
requirements and at this time EPA is not adopting any changes.
Manufacturers have been measuring N2O emissions and have
successfully certified vehicles to the N2O standards for
several years and EPA does not believe N2O measurement is
an issue needing regulatory change. EPA continues to believe direct
measurement is the best way for manufacturers to demonstrate
compliance with the N2O standards and is more appropriate
than an engineering statement without direct measurement.
---------------------------------------------------------------------------
 As explained above, the agencies have changed the alternatives
considered for the final rule, partly in response to comments. The
basic form of the standards represented by the alternatives--footprint-
based, defined by particular mathematical functions--remains the same
and as described in the NPRM. For the EPA program, EPA has chosen in
this final rule to retain the existing program for regulation of A/C
refrigerant leakage, nitrous oxide, and methane emissions as part of
the CO2 standard. This allows manufacturers to continue to
rely on this flexibility which they describe as extremely important for
compliance, although it results in continued differences between EPA's
and NHTSA's programs. This approach also avoids the possibility of gaps
in the regulation of HFCs, CH4, and N2O while EPA
developed a different way of regulating the non-CO2
emissions as part of or concurrent with the NPRM, and thereby allows
EPA to continue to regulate GHE emissions from light-duty vehicles on a
more-comprehensive basis. Thus, all alternatives considered in this
final rule reflect inclusion of CH4, N2O, and HFC
in EPA's overall ``CO2'' (more accurately, CO2-
equivalent, or CO2e) requirements. Besides this change, the
alternatives considered for the final rule differ from the NPRM in two
additional ways: First, alternatives reflecting the phase-out of the A/
C efficiency and off-cycle programs have been dropped in response to
certain comments and in recognition of the potential real-world
benefits of those programs. And second, the preferred alternative for
this final rule reflects a 1.5 percent year-over-year increase for both
passenger cars and light trucks. These changes will be discussed
further below, following a brief discussion of the form of the
standards.
A. Form of the Standards
 As in the CAFE and CO2 rulemakings in 2010 and 2012,
NHTSA and EPA proposed in the NPRM to set attribute-based CAFE and
CO2 standards defined by a mathematical function of vehicle
footprint, which has observable correlation with fuel economy and
vehicle emissions. EPCA, as amended by EISA, expressly requires that
CAFE standards for passenger cars and light trucks be based on one or
more vehicle attributes related to fuel economy and be expressed in the
form of a mathematical function.\228\ While the CAA includes no
specific requirements regarding CO2 regulation, EPA has
chosen to adopt attribute-based CO2 standards consistent
with NHTSA's EPCA/EISA requirements in the interest of harmonization
and simplifying compliance. Such an approach is permissible under
section 202(a) of the
[[Page 24247]]
CAA, and EPA has used the attribute-based approach in issuing standards
under analogous provisions of the CAA. Thus, both the proposed and
final standards take the form of fuel economy and CO2
targets expressed as functions of vehicle footprint (the product of
vehicle wheelbase and average track width). Section V.A.2 below
discusses the agencies' continued reliance on footprint as the relevant
attribute.
---------------------------------------------------------------------------
 \228\ 49 U.S.C. 32902(a)(3)(A).
---------------------------------------------------------------------------
 Under the footprint-based standards, the function defines a
CO2 or fuel economy performance target for each unique
footprint combination within a car or truck model type. Using the
functions, each manufacturer thus will have a CAFE and CO2
average standard for each year that is almost certainly unique to each
of its fleets,\229\ based upon the footprints and production volumes of
the vehicle models produced by that manufacturer. A manufacturer will
have separate footprint-based standards for cars and for trucks. The
functions are mostly sloped, so that generally, larger vehicles (i.e.,
vehicles with larger footprints) will be subject to lower CAFE mpg
targets and higher CO2 grams/mile targets than smaller
vehicles. This is because, generally speaking, smaller vehicles are
more capable of achieving higher levels of fuel economy/lower levels of
CO2 emissions, mostly because they tend not to have to work
as hard (and therefore require as much energy) to perform their driving
task. Although a manufacturer's fleet average standards could be
estimated throughout the model year based on the projected production
volume of its vehicle fleet (and are estimated as part of EPA's
certification process), the standards to which the manufacturer must
comply are determined by its final model year production figures. A
manufacturer's calculation of its fleet average standards as well as
its fleets' average performance at the end of the model year will thus
be based on the production-weighted average target and performance of
each model in its fleet.\230\
---------------------------------------------------------------------------
 \229\ EPCA/EISA requires NHTSA to separate passenger cars into
domestic and import passenger car fleets whereas EPA combines all
passenger cars into one fleet.
 \230\ As discussed in prior rulemakings, a manufacturer may have
some vehicle models that exceed their target and some that are below
their target. Compliance with a fleet average standard is determined
by comparing the fleet average standard (based on the production-
weighted average of the target levels for each model) with fleet
average performance (based on the production-weighted average of the
performance of each model).
---------------------------------------------------------------------------
 For passenger cars, consistent with prior rulemakings, NHTSA is
defining fuel economy targets as follows:
[GRAPHIC] [TIFF OMITTED] TR30AP20.051
where:
TARGETFE is the fuel economy target (in mpg) applicable to a
specific vehicle model type with a unique footprint combination,
a is a minimum fuel economy target (in mpg),
b is a maximum fuel economy target (in mpg),
c is the slope (in gallons per mile per square foot, or gpm, per
square foot) of a line relating fuel consumption (the inverse of
fuel economy) to footprint, and
d is an intercept (in gpm) of the same line.
 Here, MIN and MAX are functions that take the minimum and maximum
values, respectively, of the set of included values. For example,
MIN[40,35] = 35 and MAX(40, 25) = 40, such that MIN[MAX(40, 25), 35] =
35.
 For light trucks, also consistent with prior rulemakings, NHTSA is
defining fuel economy targets as follows:
[GRAPHIC] [TIFF OMITTED] TR30AP20.052
where:
TARGETFE is the fuel economy target (in mpg) applicable to a
specific vehicle model type with a unique footprint combination,
a, b, c, and d are as for passenger cars, but taking values specific
to light trucks,
e is a second minimum fuel economy target (in mpg),
f is a second maximum fuel economy target (in mpg),
g is the slope (in gpm per square foot) of a second line relating
fuel consumption (the inverse of fuel economy) to footprint, and
h is an intercept (in gpm) of the same second line.
 Although the general model of the target function equation is the
same for each vehicle category (passenger cars and light trucks) and
each model year, the parameters of the function equation differ for
cars and trucks. For MYs 2020-2026, the parameters are unchanged,
resulting in the same stringency in each of those model years.
 Mathematical functions defining the CO2 targets are
expressed as functions that are similar, with coefficients a-h
corresponding to those listed above.\231\ For passenger cars, EPA is
defining CO2 targets mathematically equivalent to the
following:
---------------------------------------------------------------------------
 \231\ EPA regulations use a different but mathematically
equivalent approach to specify targets. Rather than using a function
with nested minima and maxima functions, EPA regulations specify
requirements separately for different ranges of vehicle footprint.
Because these ranges reflect the combined application of the listed
minima, maxima, and linear functions, it is mathematically
equivalent and more efficient to present the targets as in this
Section.
---------------------------------------------------------------------------
TARGETCO2 = MIN[b, MAX[a, c x FOOTPRINT + d]]
where:
TARGETCO2 is the is the CO2 target (in grams per mile, or
g/mi) applicable to a specific vehicle model configuration,
a is a minimum CO2 target (in g/mi),
b is a maximum CO2 target (in g/mi),
c is the slope (in g/mi, per square foot) of a line relating
CO2 emissions to footprint, and
d is an intercept (in g/mi) of the same line.
 For light trucks, CO2 targets are defined as follows:
TARGETCO2 = MIN[MIN[b, MAX[a, c x FOOTPRINT + d]], MIN[f, MAX[e, g x
FOOTPRINT + h]]
[[Page 24248]]
where:
TARGETCO2 is the is the CO2 target (in g/mi) applicable
to a specific vehicle model configuration,
a, b, c, and d are as for passenger cars, but taking values specific
to light trucks,
e is a second minimum CO2 target (in g/mi),
f is a second maximum CO2 target (in g/mi),
g is the slope (in g/mi per square foot) of a second line relating
CO2 emissions to footprint, and
h is an intercept (in g/mi) of the same second line.
 To be clear, as has been the case since the agencies began
establishing attribute-based standards, no vehicle need meet the
specific applicable fuel economy or CO2 targets, because
compliance with either CAFE or CO2 standards is determined
based on corporate average fuel economy or fleet average CO2
emission rates. In this respect, CAFE and CO2 standards are
unlike, for example, safety standards and traditional vehicle emissions
standards. CAFE and CO2 standards apply to the average fuel
economy levels and CO2 emission rates achieved by
manufacturers' entire fleets of vehicles produced for sale in the U.S.
Safety standards apply on a vehicle-by-vehicle basis, such that every
single vehicle produced for sale in the U.S. must, on its own, comply
with minimum FMVSS. Similarly, criteria pollutant emissions standards
are applied on a per-vehicle basis, such that every vehicle produced
for sale in the U.S. must, on its own, comply with all applicable
emissions standards. When first mandating CAFE standards in the 1970s,
Congress specified a more flexible averaging-based approach that allows
some vehicles to ``under comply'' (i.e., fall short of the overall flat
standard, or fall short of their target under attribute-based
standards) as long as a manufacturer's overall fleet is in compliance.
 The required CAFE level applicable to a given fleet in a given
model year is determined by calculating the production-weighted
harmonic average of fuel economy targets applicable to specific vehicle
model configurations in the fleet, as follows:
[GRAPHIC] [TIFF OMITTED] TR30AP20.053
where:
CAFErequired is the CAFE level the fleet is required to achieve,
i refers to specific vehicle model/configurations in the fleet,
PRODUCTIONi is the number of model configuration i produced for sale
in the U.S., and
TARGETFE,i the fuel economy target (as defined above) for model
configuration i.
 Similarly, the required average CO2 level applicable to
a given fleet in a given model year is determined by calculating the
production-weighted average (not harmonic) of CO2 targets
applicable to specific vehicle model configurations in the fleet, as
follows:
[GRAPHIC] [TIFF OMITTED] TR30AP20.054
where:
CO2required is the average CO2 level the fleet is
required to achieve,
i refers to specific vehicle model/configurations in the fleet,
PRODUCTIONi is the number of model configuration i produced for sale
in the U.S., and
TARGETCO2,i is the CO2 target (as defined above) for
model configuration i.
 Section VI.A.1 describes the advantages of attribute standards,
generally. Section VI.A.2 explains the agencies' specific decision to
use vehicle footprint as the attribute over which to vary stringency
for past and current rules. Section VI.A.3 discusses the policy
considerations in selecting the specific mathematical function. Section
VI.A.4 discusses the methodologies used to develop current attribute-
based standards, and the agencies' current proposal to continue to do
so for MYs 2021-2026. Section VI.A.5 discusses the methodologies used
to reconsider the mathematical function for the proposed standards.
1. Why attribute-based standards, and what are the benefits?
 Under attribute-based standards, every vehicle model has fuel
economy and CO2 targets, the levels of which depend on the
level of that vehicle's determining attribute (for the MYs 2021-2026
standards, footprint is the determining attribute, as discussed below).
The manufacturer's fleet average CAFE performance is calculated by the
harmonic production-weighted average of those targets, as defined
below:
[GRAPHIC] [TIFF OMITTED] TR30AP20.055
 Here, i represents a given model \232\ in a manufacturer's
fleet, Productioni represents the U.S. production of that model, and
Targeti represents the target as defined by the attribute-based
standards. This means no vehicle is required to meet its target;
instead, manufacturers are free to balance improvements however they
deem best within (and, given credit transfers, at least partially
across) their fleets.
 \232\ If a model has more than one footprint variant, here each
of those variants is treated as a unique model, i, since each
footprint variant will have a unique target.
---------------------------------------------------------------------------
 Because CO2 is on a gram per mile basis rather a mile
per gallon basis,
[[Page 24249]]
harmonic averaging is not necessary when calculating required
CO2 levels:
[GRAPHIC] [TIFF OMITTED] TR30AP20.056
 The idea is to select the shape of the mathematical function
relating the standard to the fuel economy-related attribute to reflect
the trade-offs manufacturers face in producing more of that attribute
over fuel efficiency (due to technological limits of production and
relative demand of each attribute). If the shape captures these trade-
offs, every manufacturer is more likely to continue adding fuel-
efficient technology across the distribution of the attribute within
their fleet, instead of potentially changing the attribute--and other
correlated attributes, including fuel economy--as a part of their
compliance strategy. Attribute-based standards that achieve this have
several advantages.
 First, assuming the attribute is a measurement of vehicle size,
attribute-based standards help to at least partially reduce the
incentive for manufacturers to respond to CAFE and CO2
standards by reducing vehicle size in ways harmful to safety, as
compared to ``flat,'' non-attribute based standards.\233\ Larger
vehicles, in terms of mass and/or crush space, generally consume more
fuel and produce more carbon dioxide emissions, but are also generally
better able to protect occupants in a crash.\234\ Because each vehicle
model has its own target (determined by a size-related attribute),
properly fitted attribute-based standards reduce the incentive to build
smaller vehicles simply to meet a fleet-wide average, because smaller
vehicles are subject to more stringent compliance targets.
---------------------------------------------------------------------------
 \233\ The 2002 NAS Report described at length and quantified the
potential safety problem with average fuel economy standards that
specify a single numerical requirement for the entire industry. See
Transportation Research Board and National Research Council. 2002.
Effectiveness and Impact of Corporate Average Fuel Economy (CAFE)
Standards, Washington, DC: The National Academies Press (``2002 NAS
Report'') at 5, finding 12, available at https://www.nap.edu/catalog/10172/effectiveness-and-impact-of-corporate-average-fuel-economy-cafe-standards (last accessed June 15, 2018). Ensuing
analyses, including by NHTSA, support the fundamental conclusion
that standards structured to minimize incentives to downsize all but
the largest vehicles will tend to produce better safety outcomes
than flat standards.
 \234\ Bento, A., Gillingham, K., & Roth, K. (2017). The Effect
of Fuel Economy Standards on Vehicle Weight Dispersion and Accident
Fatalities. NBER Working Paper No. 23340. Available at http://www.nber.org/papers/w23340 (last accessed June 15, 2018).
---------------------------------------------------------------------------
 Second, attribute-based standards, if properly fitted, provide
automakers with more flexibility to respond to consumer preferences
than do single-valued standards. As discussed above, a single-valued
standard encourages a fleet mix with a larger share of smaller vehicles
by creating incentives for manufacturers to use downsizing the average
vehicle in their fleet (possibly through fleet mixing) as a compliance
strategy, which may result in manufacturers building vehicles for
compliance reasons that consumers do not want. Under a size-related,
attribute-based standard, reducing the size of the vehicle for
compliance's sake is a less-viable strategy because smaller vehicles
have more stringent regulatory targets. As a result, the fleet mix
under such standards is more likely to reflect aggregate consumer
demand for the size-related attribute used to determine vehicle
targets.
 Third, attribute-based standards provide a more equitable
regulatory framework across heterogeneous manufacturers who may each
produce different shares of vehicles along attributes correlated with
fuel economy.\235\ An industry-wide single-value CAFE standard imposes
disproportionate cost burden and compliance challenges on manufacturers
who produce more vehicles with attributes inherently correlated with
lower fuel economy--i.e. manufacturers who produce, on average, larger
vehicles. As discussed above, retaining flexibility for manufacturers
to produce vehicles which respect heterogeneous market preferences is
an important consideration. Since manufacturers may target different
markets as a part of their business strategy, ensuring that these
manufacturers do not incur a disproportionate share of the regulatory
cost burden is an important part of conserving consumer choices within
the market.
---------------------------------------------------------------------------
 \235\ 2002 NAS Report at 4-5, finding 10.
---------------------------------------------------------------------------
 Industry commenters generally supported attribute-based standards,
while other commenters questioned their benefits. IPI argued that
preserving the current vehicle mix was not necessarily desirable or
necessary for consumer welfare, and suggested that some vehicle
downsizing in the fleet might be beneficial both for safety and for
compliance.\236\ IPI also argued that compliance credit trading would
``help smooth out any disproportionate impacts on certain
manufacturers'' and ``ensure that manufacturers with relatively
efficient fleets still have an incentive to continue improving fuel
economy (in order to generate credits)'' \237\ Similarly, citing Ito
and Sallee, Kathryn Doolittle commented that ``. . . Ito and Sallee
(2018) have found ABR [``attribute-based regulations''] inefficient in
cost when juxtaposed with flat standard with compliance trading.''
\238\
---------------------------------------------------------------------------
 \236\ IPI, NHTSA-2018-0067-12362, at 14-15.
 \237\ IPI, NHTSA-2018-0067-12362, at 14.
 \238\ Doolittle, K, NHTSA-2018-0067-7411. See also Ito, K and
Sallee, J. ``The Economics of Attribute-Based Regulation: Theory and
Evidence from Fuel Economy Standards.'' The Review of Economics and
Statistics (2018), 100(2), pp. 319-36.
---------------------------------------------------------------------------
 The agencies have considered these comments. IPI incorrectly
characterizes the agencies' prior statements as claims that it is
important to preserve the current vehicle mix. EPA and NHTSA have never
claimed, and are not today claiming that it is important to preserve
the current fleet mix. The agencies have said, and are today
reiterating, that it is reasonable to expect that reducing the tendency
of standards to distort the market should reduce at least part of the
tendency of standards to reduce consumer welfare. Or, more concisely,
it is better to work with the market than against it. Single-value (aka
flat) CAFE standards in place from the 1970s through 2010 were clearly
distortionary. Recognizing this, the National Academy of Sciences
recommended in 2002 that NHTSA adopt attribute-based CAFE standards.
NHTSA did so in 2006, for light trucks produced starting MY 2008. As
mentioned above, in 2007, Congress codified the requirement for
attribute-based passenger car and light truck CAFE standards. Agreeing
with this history, premise, and motivation, EPA has also adopted
attribute-based CO2 standards. None of this is to say the
agencies consider it important to hold fleet mix constant. Rather, the
agencies expect that, compared to flat standards, attribute-based
standards can allow the market--including fleet mix--to better
[[Page 24250]]
follow its natural course, and all else equal, consumer acceptance is
likely to be greater if the market does so.
 The agencies also disagree with comments implying that compliance
credit trading can address all of the market distortion that flat
standards would entail. Evidence thus far suggests that trading is
fragmented, with some manufacturers apparently willing to trade only
with some other specific manufacturers. The Ito and Sallee article
cited by one commenter is a highly idealized theoretical construction,
with the authors noting, inter alia, that their model ``assumes perfect
competition.'' \239\ Its findings regarding comparative economic
efficiency of flat- and attribute-based standards are, therefore,
merely hypothetical, and the agencies find little basis in recent
transactions to suggest the compliance credit trading market reflects
the authors' idealized assumptions. Even if the agencies did expect
credit trading markets to operate as in an idealized textbook example,
basing the structure of standards on the presumption of perfect trading
would not be appropriate. FCA commented that ``. . . when flexibilities
are considered while setting targets, they cease to be flexibilities
and become simply additional technology mandates,'' and the Alliance
commented, similarly, that ``the Agencies should keep `flexibilities'
as optional ways to comply and not unduly assume that each flexibility
allows additional stringency of footprint-based standards.'' \240\
Perhaps recognizing this reality, Congress has barred NHTSA from
considering manufacturers' ability to use compliance credits (even
credits earned and used by the same OEM, much less credits traded
between OEMs). As discussed further in Section VIII.A.2, EPA believes
that while credit trading may be a useful flexibility to reduce the
overall costs of the program, it is important to set standards in a way
that does not rely on credit purchasing availability as a compliance
mechanism.
---------------------------------------------------------------------------
 \239\ Ito and Sallee, op. cit., Supplemental Appendix, at A-15,
available at https://www.mitpressjournals.org/doi/suppl/10.1162/REST_a_00704/suppl_file/REST_a_00704-esupp.pdf (accessed October 29,
2019).
 \240\ FCA, NHTSA-2018-0067-11943, at 6; Alliance, NHTSA-2018-
0067-12073, Full Comment Set, at 40, fn. 82.
---------------------------------------------------------------------------
 Considering these comments and realities, considering EPCA's
requirement for attribute-based CAFE standards, and considering the
benefits of regulatory harmonization, the agencies are, again,
finalizing attribute-based CAFE and CO2 standards rather
than, for either program, finalizing flat standards.
Why footprint as the attribute?
 It is important that the CAFE and CO2 standards be set
in a way that does not unnecessarily incentivize manufacturers to
respond by selling vehicles that are less safe. Vehicle size is highly
correlated with vehicle safety--for this reason, it is important to
choose an attribute correlated with vehicle size (mass or some
dimensional measure). Given this consideration, there are several
policy and technical reasons why footprint is considered to be the most
appropriate attribute upon which to base the standards, even though
other vehicle size attributes (notably, curb weight) are more strongly
correlated with fuel economy and tailpipe CO2 emissions.
 First, mass is strongly correlated with fuel economy; it takes a
certain amount of energy to move a certain amount of mass. Footprint
has some positive correlation with frontal surface area, likely a
negative correlation with aerodynamics, and therefore fuel economy, but
the relationship is less deterministic. Mass and crush space
(correlated with footprint) are both important safety considerations.
As discussed below and in the accompanying PRIA, NHTSA's research of
historical crash data indicates that holding footprint constant, and
decreasing the mass of the largest vehicles, will result in a net
positive safety impact to drivers overall, while holding footprint
constant and decreasing the mass of the smallest vehicles will result
in a net decrease in fleetwide safety. Properly fitted footprint-based
standards provide little, if any, incentive to build smaller footprint
vehicles to meet CAFE and CO2 standards, and therefore help
minimize the impact of standards on overall fleet safety.
 Second, it is important that the attribute not be easily
manipulated in a manner that does not achieve the goals of EPCA or
other goals, such as safety. Although weight is more strongly
correlated with fuel economy than footprint, there is less risk of
artificial manipulation (i.e., changing the attribute(s) to achieve a
more favorable target) by increasing footprint under footprint-based
standards than there would be by increasing vehicle mass under weight-
based standards. It is relatively easy for a manufacturer to add enough
weight to a vehicle to decrease its applicable fuel economy target a
significant amount, as compared to increasing vehicle footprint, which
is a much more complicated change that typically takes place only with
a vehicle redesign.
 Further, some commenters on the MY 2011 CAFE rulemaking were
concerned that there would be greater potential for such manipulation
under multi-attribute standards, such as those that also depend on
weight, torque, power, towing capability, and/or off-road capability.
As discussed in NHTSA's MY 2011 CAFE final rule,\241\ it is anticipated
that the possibility of manipulation is lowest with footprint-based
standards, as opposed to weight-based or multi-attribute-based
standards. Specifically, standards that incorporate weight, torque,
power, towing capability, and/or off-road capability in addition to
footprint would not only be more complex, but by providing degrees of
freedom with respect to more easily adjusted attributes, they could
make it less certain that the future fleet would actually achieve the
projected average fuel economy and CO2 levels. This is not
to say that a footprint-based system eliminates manipulation, or that a
footprint-based system eliminates the possibility that manufacturers
will change vehicles in ways that compromise occupant protection, but
footprint-based standards achieve the best balance among affected
considerations.
---------------------------------------------------------------------------
 \241\ See 74 FR at 14359 (Mar. 30, 2009).
---------------------------------------------------------------------------
 Several stakeholders commented on whether vehicular footprint is
the most suitable attribute upon which to base standards. IPI commented
that ``. . . footprint-based standards may be unnecessary to respect
consumer preferences, may negatively impact safety, and may be overall
inefficient. Several arguments call into question the footprint-based
approach, but a particularly important one is that large vehicles can
impose a negative safety externality on other drivers.'' \242\ IPI
commented, further, that the agencies should consider the relative
merits of other vehicle attributes, including vehicle fuel type,
suggesting that it would be more difficult for manufacturers to
manipulate a flatter standard or one ``differentiated by fuel type.''
\243\ Similarly, Michalek and Whitefoot recommended ``that the agencies
reexamine automaker response to the footprint-based standards to
determine if adjustments should be made to avoid inducing increases to
vehicle size.'' \244\
---------------------------------------------------------------------------
 \242\ IPI, NHTSA-2018-0067-12362, at 12.
 \243\ IPI, NHTSA-2018-0067-12362, at 13 et seq.
 \244\ Michalek, J. and Whitefoot, K., NHTSA-2018-0067-11903, at
13.
---------------------------------------------------------------------------
[[Page 24251]]
 Conversely, ICCT commented that ``the switch to footprint-based
CAFE and [CO2] standards has been widely credited with
diminishing safety concerns with efficiency standards. Footprint
standards encourage larger vehicles with wider track width, which
reduces rollovers, and longer wheelbase, which increases the crush
space and reduces deceleration forces for both vehicles in a two-
vehicle collision.'' \245\ Similarly, BorgWarner commented that ``the
use of a footprint standard not only provides greater incentive for
mass reduction, but also encourages a larger footprint for a given
vehicle mass, thus providing increased safety for a given mass
vehicle,'' \246\ and the Aluminum Association commented footprint based
standards drive ``fuel-efficiency improvement across all vehicle
classes,'' ``eliminate the incentive to shift fleet volume to smaller
cars which has been shown to slightly decrease safety in vehicle-to-
vehicle collisions,'' and provide ``an incentive for reducing weight in
the larger vehicles, where weight reduction is of the most benefit for
societal safety,'' citing Ford's aluminum-intensive F150 pickup truck
as an example.\247\ NADA urged the agencies to continue basing
standards on vehicle footprint, as doing so ``serves both to require
and allow OEMs to build more fuel-efficient vehicles across the
broadest possible light-duty passenger car and truck spectrum,'' \248\
and UCS commented that footprint-based standards ``increase consumer
choice, ensuring that the vehicles available for purchase in every
vehicle class continue to get more efficient.'' \249\ Furthermore,
regarding concerns that footprint-based standards may be susceptible to
manipulation, the Alliance commented that ``the data above [from
Novation Analytics] shows there are no systemic footprint increases (or
any type of target manipulation) occurring.'' \250\ While FCA's
comments supported this Alliance comment, FCA commented further that,
lacking some utility-related vehicle attributes such as towing
capability, 4-wheel-drive, and ride height, ``it is clear the footprint
standard does not fully account for pickup truck capability and the
components needed such as larger powertrains, greater mass and frontal
area,'' and requested the agencies ``correct LDT standards to reflect
the current market preference for capability over efficiency, and
introduce mechanisms into the regulation that can adjust for efficiency
and capability tradeoffs that footprint standards currently ignore.''
\251\
---------------------------------------------------------------------------
 \245\ ICCT, NHTSA-2018-0067-11741, at B-4.
 \246\ BorgWarner, NHTSA-2018-0067-11893, at 10.
 \247\ Aluminum Association, NHTSA-2018-0067-11952, at 3.
 \248\ NADA, NHTSA-2018-0067-12064, at 13.
 \249\ UCS, UCS, NHTSA-2018-0067-12039, at 46.
 \250\ Alliance, NHTSA-2018-0067-12073, at 123.
 \251\ FCA, NHTSA-2018-0067-11943, at 49.
---------------------------------------------------------------------------
 When first electing to adopt footprint-based standards, NHTSA
carefully considered other alternatives, including vehicle mass and
``shadow'' (overall width multiplied by overall length). Compared to
both of these other alternatives, footprint is much less susceptible to
gaming, because while there is some potential to adjust track width,
wheelbase is more expensive to change, at least outside a planned
vehicle redesign. EPA agreed with NHTSA's assessment, nothing has
changed the relative merits of at least these three potential
attributes, and nothing in the evolution of the fleet demonstrates that
footprint-based standards are leading manufacturers to increase the
footprint of specific vehicle models by more than they would in
response to customer demand. Also, even if footprint-based standards
are encouraging some increases in vehicle size, NHTSA continues to
maintain, and EPA to agree, that such increases should tend to improve
overall highway safety rather than degrading it. Regarding FCA's
request that the agencies adopt an approach that accounts for a wider
range of vehicle attributes related to both vehicle fuel economy and
customer-facing vehicle utility, the agencies are concerned that doing
so could further complicate already-complex standards and also lead to
unintended consequences. For example, it is not currently clear how a
multi-attribute approach would appropriately balance emphasis between
vehicle attributes (e.g., how much relative fuel consumption should be
attributed to, respectively, vehicle footprint, towing capacity, drive
type, and ground clearance). Also, basing standards on, in part, ground
clearance would encourage manufacturers to increase ride height,
potentially increasing the frequency of vehicle rollover crashes.
Regarding IPI's recommendation that fuel type be included as a vehicle
attribute for attribute-based standards, the agencies note that both
CAFE and CO2 standards already account for fuel type in the
procedures for measuring fuel economy levels and CO2
emission rates, and for calculating fleet average CAFE and
CO2 levels.
 Therefore, having considered public comments on the choice of
vehicle attributes for CAFE and CO2 standards, the agencies
are finalizing standards that, as proposed, are defined in terms of
vehicle footprint.
3. What mathematical function should be used to specify footprint-based
standards?
 In requiring NHTSA to ``prescribe by regulation separate average
fuel economy standards for passenger and non-passenger automobiles
based on 1 or more vehicle attributes related to fuel economy and
express each standard in the form of a mathematical function,'' EPCA/
EISA provides ample discretion regarding not only the selection of the
attribute(s), but also regarding the nature of the function. The CAA
provides no specific direction regarding CO2 regulation, and
EPA has continued to harmonize this aspect of its CO2
regulations with NHTSA's CAFE regulations. The relationship between
fuel economy (and CO2 emissions) and footprint, though
directionally clear (i.e., fuel economy tends to decrease and
CO2 emissions tend to increase with increasing footprint),
is theoretically vague, and quantitatively uncertain; in other words,
not so precise as to a priori yield only a single possible curve.
 The decision of how to specify this mathematical function therefore
reflects some amount of judgment. The function can be specified with a
view toward achieving different environmental and petroleum reduction
goals, encouraging different levels of application of fuel-saving
technologies, avoiding any adverse effects on overall highway safety,
reducing disparities of manufacturers' compliance burdens, and
preserving consumer choice, among other aims. The following are among
the specific technical concerns and resultant policy tradeoffs the
agencies have considered in selecting the details of specific past and
future curve shapes:
 Flatter standards (i.e., curves) increase the risk that
both the size of vehicles will be reduced, potentially compromising
highway safety, and reducing any utility consumers would have gained
from a larger vehicle.
 Steeper footprint-based standards may create incentives to
upsize vehicles, potentially oversupplying vehicles of certain
footprints beyond what consumers would naturally demand, and thus
increasing the possibility that fuel savings and CO2
reduction benefits will be forfeited artificially.
 Given the same industry-wide average required fuel economy
or CO2 standard, flatter standards tend to place greater
compliance burdens on full-line manufacturers.
 Given the same industry-wide average required fuel economy
or CO2
[[Page 24252]]
standard, dramatically steeper standards tend to place greater
compliance burdens on limited-line manufacturers (depending of course,
on which vehicles are being produced).
 If cutpoints are adopted, given the same industry-wide
average required fuel economy, moving small-vehicle cutpoints to the
left (i.e., up in terms of fuel economy, down in terms of
CO2 emissions) discourages the introduction of small
vehicles, and reduces the incentive to downsize small vehicles in ways
that could compromise overall highway safety.
 If cutpoints are adopted, given the same industry-wide
average required fuel economy, moving large-vehicle cutpoints to the
right (i.e., down in terms of fuel economy, up in terms of
CO2 emissions) better accommodates the design requirements
of larger vehicles--especially large pickups--and extends the size
range over which downsizing is discouraged.
4. What mathematical functions have been used previously, and why?
 Notwithstanding the aforementioned discretion under EPCA/EISA, data
should inform consideration of potential mathematical functions, but
how relevant data is defined and interpreted, and the choice of
methodology for fitting a curve to that data, can and should include
some consideration of specific policy goals. This section summarizes
the methodologies and policy concerns that were considered in
developing previous target curves (for a complete discussion see the
2012 FRIA).
 As discussed below, the MY 2011 final curves followed a constrained
logistic function defined specifically in the final rule.\252\ The MYs
2012-2021 final standards and the MYs 2022-2025 augural standards are
defined by constrained linear target functions of footprint, as shown
below: \253\
---------------------------------------------------------------------------
 \252\ See 74 FR 14196, 14363-14370 (Mar. 30, 2009) for NHTSA
discussion of curve fitting in the MY 2011 CAFE final rule.
 \253\ The right cutpoint for the light truck curve was moved
further to the right for MYs 2017-2021, so that more possible
footprints would fall on the sloped part of the curve. In order to
ensure that, for all possible footprints, future standards would be
at least as high as MY 2016 levels, the final standards for light
trucks for MYs 2017-2021 is the maximum of the MY 2016 target curves
and the target curves for the give MY standard. This is defined
further in the 2012 final rule. See 77 FR 62624, at 62699-700 (Oct.
15, 2012).
[GRAPHIC] [TIFF OMITTED] TR30AP20.057
 Here, Target is the fuel economy target applicable to vehicles
of a given footprint in square feet (Footprint). The upper
asymptote, a, and the lower asymptote, b, are specified in mpg; the
reciprocal of these values represent the lower and upper asymptotes,
respectively, when the curve is instead specified in gallons per
mile (gpm). The slope, c, and the intercept, d, of the linear
portion of the curve are specified as gpm per change in square feet,
---------------------------------------------------------------------------
and gpm, respectively.
 The min and max functions will take the minimum and maximum values
within their associated parentheses. Thus, the max function will first
find the maximum of the fitted line at a given footprint value and the
lower asymptote from the perspective of gpm. If the fitted line is
below the lower asymptote it is replaced with the floor, which is also
the minimum of the floor and the ceiling by definition, so that the
target in mpg space will be the reciprocal of the floor in mpg space,
or simply, a. If, however, the fitted line is not below the lower
asymptote, the fitted value is returned from the max function and the
min function takes the minimum value of the upper asymptote (in gpm
space) and the fitted line. If the fitted value is below the upper
asymptote, it is between the two asymptotes and the fitted value is
appropriately returned from the min function, making the overall target
in mpg the reciprocal of the fitted line in gpm. If the fitted value is
above the upper asymptote, the upper asymptote is returned is returned
from the min function, and the overall target in mpg is the reciprocal
of the upper asymptote in gpm space, or b.
 In this way curves specified as constrained linear functions are
specified by the following parameters:
a = upper limit (mpg)
b = lower limit (mpg)
c = slope (gpm per sq.ft.)
d = intercept (gpm)
 The slope and intercept are specified as gpm per sq. ft. and gpm
instead of mpg per sq. ft. and mpg because fuel consumption and
emissions appear roughly linearly related to gallons per mile (the
reciprocal of the miles per gallon).
a) NHTSA in MY 2008 and MY 2011 CAFE (Constrained Logistic)
 For the MY 2011 CAFE rule, NHTSA estimated fuel economy levels by
footprint from the MY 2008 fleet after normalization for differences in
technology,\254\ but did not make adjustments to reflect other vehicle
attributes (e.g., power-to-weight ratios). Starting with the
technology-adjusted passenger car and light truck fleets, NHTSA used
minimum absolute deviation (MAD) regression without sales weighting to
fit a logistic form as a starting point to develop mathematical
functions defining the standards. NHTSA then identified footprints at
which to apply minimum and maximum values (rather than letting the
standards extend without limit) and transposed these functions
vertically (i.e., on a gallons-per-mile basis, uniformly downward) to
produce the promulgated standards. In the preceding rule, for MYs 2008-
2011 light truck standards, NHTSA examined a range of potential
functional forms, and concluded that, compared to other considered
forms, the constrained logistic form provided the expected and
appropriate trend (decreasing fuel economy as footprint increases), but
avoided creating ``kinks'' the agency was concerned would provide
distortionary incentives for vehicles with neighboring footprints.\255\
---------------------------------------------------------------------------
 \254\ See 74 FR 14196, 14363-14370 (Mar. 30, 2009) for NHTSA
discussion of curve fitting in the MY 2011 CAFE final rule.
 \255\ See 71 FR 17556, 17609-17613 (Apr. 6, 2006) for NHTSA
discussion of ``kinks'' in the MYs 2008-2011 light truck CAFE final
rule (there described as ``edge effects''). A ``kink,'' as used
here, is a portion of the curve where a small change in footprint
results in a disproportionally large change in stringency.
---------------------------------------------------------------------------
b) MYs 2012-2016 Standards (Constrained Linear)
 For the MYs 2012-2016 rule, potential methods for specifying
mathematical functions to define fuel economy and CO2
standards were reevaluated. These methods were fit to the same MY 2008
data as the MY 2011 standard. Considering these further specifications,
the constrained logistic form, if applied to post-MY 2011 standards,
would likely contain a steep mid-section that would provide undue
incentive to increase the footprint of midsize passenger cars.\256\ A
range of
[[Page 24253]]
methods to fit the curves would have been reasonable, and a minimum
absolute deviation (MAD) regression without sales weighting on a
technology-adjusted car and light truck fleet was used to fit a linear
equation. This equation was used as a starting point to develop
mathematical functions defining the standards. Footprints were then
identified at which to apply minimum and maximum values (rather than
letting the standards extend without limit). Finally, these
constrained/piecewise linear functions were transposed vertically
(i.e., on a gpm or CO2 basis, uniformly downward) by
multiplying the initial curve by a single factor for each MY standard
to produce the final attribute-based targets for passenger cars and
light trucks described in the final rule.\257\ These transformations
are typically presented as percentage improvements over a previous MY
target curve.
---------------------------------------------------------------------------
 \256\ 75 FR at 25362.
 \257\ See generally 74 FR at 49491-96; 75 FR at 25357-62.
---------------------------------------------------------------------------
c) MYs 2017 and Beyond Standards (Constrained Linear)
 The mathematical functions finalized in 2012 for MYs 2017 and
beyond changed somewhat from the functions for the MYs 2012-2016
standards. These changes were made both to address comments from
stakeholders, and to consider further some of the technical concerns
and policy goals judged more preeminent under the increased uncertainty
of the impacts of finalizing and proposing standards for model years
further into the future.\258\ Recognizing the concerns raised by full-
line OEMs, it was concluded that continuing increases in the stringency
of the light truck standards would be more feasible if the light truck
curve for MYs 2017 and beyond was made steeper than the MY 2016 truck
curve and the right (large footprint) cut-point was extended only
gradually to larger footprints. To accommodate these considerations,
the 2012 final rule finalized the slope fit to the MY 2008 fleet using
a sales-weighted, ordinary least-squares regression, using a fleet that
had technology applied to make the technology application across the
fleet more uniform, and after adjusting the data for the effects of
weight-to-footprint. Information from an updated MY 2010 fleet was also
considered to support this decision. As the curve was vertically
shifted (with fuel economy specified as mpg instead of gpm or
CO2 emissions) upwards, the right cutpoint was progressively
moved for the light truck curves with successive model years, reaching
the final endpoint for MY 2021.
---------------------------------------------------------------------------
 \258\ The MYs 2012-2016 final standards were signed April 1st,
2010--putting 6.5 years between its signing and the last affected
model year, while the MYs 2017-2021 final standards were signed
August 28th, 2012--giving just more than nine years between signing
and the last affected final standards.
---------------------------------------------------------------------------
5. Reconsidering the Mathematical Functions for Today's Rulemaking
a) Why is it important to reconsider the mathematical functions?
 By shifting the developed curves by a single factor, it is assumed
that the underlying relationship of fuel consumption (in gallons per
mile) to vehicle footprint does not change significantly from the model
year data used to fit the curves to the range of model years for which
the shifted curve shape is applied to develop the standards. However,
it must be recognized that the relationship between vehicle footprint
and fuel economy is not necessarily constant over time; newly developed
technologies, changes in consumer demand, and even the curves
themselves could influence the observed relationships between the two
vehicle characteristics. For example, if certain technologies are more
effective or more marketable for certain types of vehicles, their
application may not be uniform over the range of vehicle footprints.
Further, if market demand has shifted between vehicle types, so that
certain vehicles make up a larger share of the fleet, any underlying
technological or market restrictions which inform the average shape of
the curves could change. That is, changes in the technology or market
restrictions themselves, or a mere re-weighting of different vehicles
types, could reshape the fit curves.
 For the above reasons, the curve shapes were reconsidered in the
proposal using the newest available data from MY 2016. With a view
toward corroboration through different techniques, a range of
descriptive statistical analyses were conducted that do not require
underlying engineering models of how fuel economy and footprint might
be expected to be related, and a separate analysis that uses vehicle
simulation results as the basis to estimate the relationship from a
perspective more explicitly informed by engineering theory was
conducted as well. Despite changes in the new vehicle fleet both in
terms of technologies applied and in market demand, the underlying
statistical relationship between footprint and fuel economy has not
changed significantly since the MY 2008 fleet used for the 2012 final
rule; therefore, EPA and NHTSA proposed to continue to use the curve
shapes fit in 2012. The analysis and reasoning supporting this decision
follows.
b) What statistical analyses did EPA and NHTSA consider?
 In considering how to address the various policy concerns discussed
above, data from the MY 2016 fleet was considered, and a number of
descriptive statistical analyses (i.e., involving observed fuel economy
levels and footprints) using various statistical methods, weighting
schemes, and adjustments to the data to make the fleets less
technologically heterogeneous were performed. There were several
adjustments to the data that were common to all of the statistical
analyses considered.
 With a view toward isolating the relationship between fuel economy
and footprint, the few diesels in the fleet were excluded, as well as
the limited number of vehicles with partial or full electric
propulsion; when the fleet is normalized so that technology is more
homogenous, application of these technologies is not allowed. This is
consistent with the methodology used in the 2012 final rule.
 The above adjustments were applied to all statistical analyses
considered, regardless of the specifics of each of the methods,
weights, and technology level of the data, used to view the
relationship of vehicle footprint and fuel economy. Table V-1, below,
summarizes the different assumptions considered and the key attributes
of each. The analysis was performed considering all possible
combinations of these assumptions, producing a total of eight footprint
curves.
[[Page 24254]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.058
(1) Current Technology Level Curves
 The ``current technology'' level curves exclude diesels and
vehicles with electric propulsion, as discussed above, but make no
other changes to each model year fleet. Comparing the MY 2016 curves to
ones built under the same methodology from previous model year fleets
shows whether the observed curve shape has changed significantly over
time as standards have become more stringent. Importantly, these curves
will include any market forces which make technology application
variable over the distribution of footprint. These market forces will
not be present in the ``maximum technology'' level curves: By making
technology levels homogenous, this variation is removed. The current
technology level curves built using both regression types and both
regression weight methodologies from the MY 2008, MY 2010, and MY 2016
fleets, shown in more detail in Chapter 4.4.2.1 of the PRIA, support
the curve slopes finalized in the 2012 final rule. The curves built
from most methodologies using each fleet generally shift, but remain
very similar in slope. This suggests that the relationship of footprint
to fuel economy, including both technology and market limits, has not
significantly changed.
(2) Maximum Technology Level Curves
 As in prior rulemakings, technology differences between vehicle
models were considered to be a significant factor producing uncertainty
regarding the relationship between fuel consumption and footprint.
Noting that attribute-based standards are intended to encourage the
application of additional technology to improve fuel efficiency and
reduce CO2 emissions across the distribution of footprint in
the fleet, approaches were considered in which technology application
is simulated for purposes of the curve fitting analysis in order to
produce fleets that are less varied in technology content. This
approach helps reduce ``noise'' (i.e., dispersion) in the plot of
vehicle footprints and fuel consumption levels and identify a more
technology-neutral relationship between footprint and fuel consumption.
The results of updated analysis for maximum technology level curves are
also shown in Chapter 4.4.2.2 of the PRIA. Especially if vehicles
progress over time toward more similar size-specific efficiency,
further removing variation in technology application both better
isolates the relationship between fuel consumption and footprint and
further supports the curve slopes finalized in the 2012 final rule.
c) What other methodologies were considered?
 The methods discussed above are descriptive in nature, using
statistical analysis to relate observed fuel economy levels to observed
footprints for known vehicles. As such, these methods are clearly based
on actual data, answering the question ``how does fuel economy appear
to be related to footprint?'' However, being independent of explicit
engineering theory, they do not answer the question ``how might one
expect fuel economy to be related to footprint?'' Therefore, as an
alternative to the above methods, an alternative methodology was also
developed and applied that, using full-vehicle simulation, comes closer
to answering the second question, providing a basis either to
corroborate answers to the first, or suggest that further investigation
could be important.
 As discussed in the 2012 final rule, several manufacturers have
confidentially shared with the agencies what they described as
``physics-based'' curves, with each OEM showing significantly different
shapes for the footprint-fuel economy relationships. This variation
suggests that manufacturers face different curves given the other
attributes of the vehicles in their fleets (i.e., performance
[[Page 24255]]
characteristics) and/or that their curves reflected different levels of
technology application. In reconsidering the shapes of the proposed MYs
2021-2026 standards, a similar estimation of physics-based curves
leveraging third-party simulation work form Argonne National
Laboratories (Argonne) was developed. Estimating physics-based curves
better ensures that technology and performance are held constant for
all footprints; augmenting a largely statistical analysis with an
analysis that more explicitly incorporates engineering theory helps to
corroborate that the relationship between fuel economy and footprint is
in fact being characterized.
 Tractive energy is the amount of energy it will take to move a
vehicle.\259\ Here, tractive energy effectiveness is defined as the
share of the energy content of fuel consumed which is converted into
mechanical energy and used to move a vehicle--for internal combustion
engine (ICE) vehicles, this will vary with the relative efficiency of
specific engines. Data from Argonne simulations suggest that the limits
of tractive energy effectiveness are approximately 25 percent for
vehicles with internal combustion engines which do not possess
integrated starter generator, other hybrid, plug-in, pure electric, or
fuel cell technology.
---------------------------------------------------------------------------
 \259\ Thomas, J. ``Drive Cycle Powertrain Efficiencies and
Trends Derived from EPA Vehicle Dynamometer Results,'' SAE Int. J.
Passeng. Cars--Mech. Syst. 7(4):2014, doi:10.4271/2014-01-2562.
Available at https://www.sae.org/publications/technical-papers/content/2014-01-2562/ (last accessed June 15, 2018).
---------------------------------------------------------------------------
 A tractive energy prediction model was also developed to support
today's proposal. Given a vehicle's mass, frontal area, aerodynamic
drag coefficient, and rolling resistance as inputs, the model will
predict the amount of tractive energy required for the vehicle to
complete the Federal test cycle. This model was used to predict the
tractive energy required for the average vehicle of a given footprint
\260\ and ``body technology package'' to complete the cycle. The body
technology packages considered are defined in Table V-2, below. Using
the absolute tractive energy predicted and tractive energy
effectiveness values spanning possible ICE engines, fuel economy values
were then estimated for different body technology packages and engine
tractive energy effectiveness values.
---------------------------------------------------------------------------
 \260\ The mass reduction curves used elsewhere in this analysis
were used to predict the mass of a vehicle with a given footprint,
body style box, and mass reduction level. The `Body style Box' is 1
for hatchbacks and minivans, 2 for pickups, and 3 for sedans, and is
an important predictor of aerodynamic drag. Mass is an essential
input in the tractive energy calculation.
[GRAPHIC] [TIFF OMITTED] TR30AP20.059
 Chapter 6 of the PRIA show the resultant CAFE levels estimated for
the vehicle classes Argonne simulated for this analysis, at different
footprint values and by vehicle ``box.'' Pickups are considered 1-box,
hatchbacks and minivans are 2-box, and sedans are 3-box. These
estimates are compared with the MY 2021 standards finalized in 2012.
The general trend of the simulated data points follows the pattern of
the previous MY 2021 standards for all technology packages and tractive
energy effectiveness values presented in the PRIA. The tractive energy
curves are intended to validate the curve shapes against a physics-
based alternative, and the analysis suggests that the curve shapes
track the physical relationship between fuel economy and tractive
energy for different footprint values.
 Physical limitations are not the only forces manufacturers face;
their success is dependent upon producing vehicles that consumers
desire and will purchase. For this reason, in setting future standards,
the analysis will continue to consider information from statistical
analyses that do not homogenize technology applications in addition to
statistical analyses which do, as well as a tractive energy analysis
similar to the one presented above.
 The relationship between fuel economy and footprint remains
directionally discernable but quantitatively uncertain. Nevertheless,
each standard must commit to only one function. Approaching the
question ``how is fuel economy related to footprint'' from different
directions and applying different approaches has given EPA and NHTSA
confidence that the function applied here appropriately and reasonably
reflects the relationship between fuel economy and footprint.
 The agencies invited comments on this conclusion and the supporting
analysis. IPI raised concerns that ``. . . several dozen models (mostly
subcompacts and sports cars) fall in the 30-40 square feet range, which
are all subject to the same standards'' and that ``manufacturers of
these models may have an incentive to decrease footprints as a
compliance strategy, since doing so would not trigger more stringent
standards.'' \261\ NHTSA and EPA agree that, all else equal, downsizing
the smallest cars (e.g., Chevrolet Spark, Ford Fiesta, Mini Cooper,
Mazda MX-5, Porsche 911, Toyota Yaris) would most likely tend to
degrade overall highway safety. At the same time, as discussed above,
the agencies recognize that small vehicles do appear attractive to some
market segments (although obviously the Ford Fiesta and Porsche 911
compete in different segments).
[[Page 24256]]
Therefore, there is a tension between on one hand, avoiding standards
that unduly encourage safety-eroding downsizing and, on the other,
avoiding standards that unduly penalize the market for small vehicles.
The agencies examined this issue, and note that the market for the
smallest vehicles has not evolved at all as estimated in the analysis
supporting the 2012 final rule, and attribute this more to fuel prices
and consumer demand for larger vehicles than to attribute-based CAFE
and CO2 standards. For example, the market for vehicles with
footprints less than 40 square foot was about 45 percent smaller in MY
2017 than in MY 2010. The agencies also found that among the smallest
vehicle models produced throughout MYs 2010-2017, most have become
larger, not smaller. For example, while the Mazda MX-5's footprint
decreased by 0.1 square foot (0.3 percent) during that time, the MY
2017 versions of the Mini Cooper, Smart fortwo, Porsche 911, and Toyota
Yaris had larger footprints than in MY 2010. With the market for very
small vehicles shrinking, and with manufacturers not evidencing a
tendency to make the smallest vehicles even smaller, the agencies are
satisfied that it would be unwise to change the target functions such
that targets never stop becoming more stringent as vehicle footprint
becomes ever smaller, because doing so could further impede an already-
shrinking market.
---------------------------------------------------------------------------
 \261\ IPI, NHTSA-2018-0067-12362, p. 14.
---------------------------------------------------------------------------
B. No-Action Alternative
 As in the proposal, the No-Action Alternative applies the augural
CAFE and final CO2 targets announced in 2012 for MYs 2021-
2025.\262\ For MY 2026, this alternative applies the same targets as
for MY 2025. The carbon dioxide equivalent of air conditioning
refrigerant leakage credits, nitrous oxide, and methane emissions are
included for compliance with the EPA standards for all model years
under the no-action alternative.\263\
---------------------------------------------------------------------------
 \262\ https://www.govinfo.gov/content/pkg/CFR-2014-title40-vol19/pdf/CFR-2014-title40-vol19-sec86-1818-12.pdf
 \263\ EPA regulations use a different but mathematically
equivalent approach to specify targets. Rather than using a function
with nested minima and maxima functions, EPA regulations specify
requirements separately for different ranges of vehicle footprint.
Because these ranges reflect the combined application of the listed
minima, maxima, and linear functions, it is mathematically
equivalent and more efficient to present the targets as in this
Section.
[GRAPHIC] [TIFF OMITTED] TR30AP20.060
[[Page 24257]]
 In comments on the DEIS, CBD et al. indicated that it was
appropriate for NHTSA to use the augural CAFE standards as the baseline
No Action regulatory alternative.\264\ However, CARB commented that the
baseline regulatory alternative should include CARB's ZEV mandate, in
part because EPA must consider ``other regulations promulgated by EPA
or other government entities,'' and, according to CARB, there will be
much more vehicle electrification in the future as manufacturers
respond to market demand and also work to comply with the ZEV
mandate.\265\ Similarly, EPA's Science Advisory Board recommended--
despite the action taken in the One National Program Action--that the
baseline include state ZEV mandates ``to be consistent with policies
that would prevail in the absence of the rule change.'' \266\ EPA's
Science Advisory Board further recommended including sensitivity
analyses with different penetration rates of ZEVs.
---------------------------------------------------------------------------
 \264\ CBD et al., NHTSA-2018-0067-12123, Attachment 1, at 13.
 \265\ CARB, NHTSA-2018-0067-11873, at 124-125.
 \266\ SAB at 12 and 29-30.
---------------------------------------------------------------------------
 On the other hand, arguing for consideration of standards less
stringent than those proposed in the NPRM, Walter Kreucher commented
that rather than using the augural standards as the baseline, ``a
better approach would be to assume a clean sheet of paper and start
from the existing 2016MY fleet and its associated standards as the
baseline using 0%/year increases for both passenger cars and light
trucks for MYs 2017-2026.'' \267\ Similarly, AVE argued that because
previously-promulgated standards for MYs 2018-2021 already present a
significant challenge that ``will likely require almost every automaker
to continue using credits for compliance, . . . AVE believes this
rulemaking should reset . . . the current compliance baseline for cars
and light trucks at MY 2018 . . .'' \268\ BorgWarner commented
similarly that ``Beginning in MY 2018, standards should be reset to the
levels the industry actually achieved. For MY 2018 and beyond,
succeeding model year targets should be set with an annual rate of
improvement defined by the slope of improvement the industry has
achieved over the last six years. . . . Based on these data, our
analysis suggests the most reasonable and logical rate of improvement
falls between 2.0% to 2.6% for cars and trucks. Additionally, a single
rate of improvement for the combined fleet should be considered.''
\269\
---------------------------------------------------------------------------
 \267\ Kreucher, W., NHTSA-2018-0067-0444, at 8.
 \268\ AVE, NHTSA-2018-0067-11696, at 8-9.
 \269\ BorgWarner, NHTSA-2018-0067-11895, at 3, 6.
---------------------------------------------------------------------------
 The No-Action Alternative represents expectations regarding the
world in the absence of a proposal, accounting for applicable laws
already in place. Although manufacturers are already making significant
use of compliance credits toward compliance with even MY 2017
standards, the agencies are obligated to evaluate regulatory
alternatives against the standards already in place through MY 2025.
Similarly, even though manufacturers are already producing electric
vehicles, EPA and NHTSA appropriately excluded California's ZEV mandate
from the No-Action alternative for the NPRM, for several reasons.
First, the ZEV mandate is not Federal law; second, as described in the
proposal and subsequently finalized in regulatory text, the ZEV mandate
is expressly and impliedly preempted by EPCA; third, EPA proposed to
withdraw the waiver of CAA preemption in the NPRM and subsequently
finalized this withdrawal. Accordingly, the agencies have, therefore,
appropriately excluded the ZEV mandate from the No-Action alternative.
However, as discussed below, the agencies' analysis does account for
the potential that under every regulatory alternative, including the
No-Action Alternative, vehicle electrification could increase in the
future, especially if batteries become less expensive as gasoline
becomes more expensive.
C. Action Alternatives
1. Alternatives in Final Rule
 Table V-5 below shows the different alternatives evaluated in
today's notice.
[GRAPHIC] [TIFF OMITTED] TR30AP20.061
[[Page 24258]]
 With one exception, the alternatives considered in the NPRM
included the changes in stringency for the above alternatives.
Alternative 3, the preferred alternative, is newly included for today's
notice.\270\
---------------------------------------------------------------------------
 \270\ As the agencies indicated in the NPRM, they were
considering and taking comment ``on a wide range of alternatives and
have specifically modeled eight alternatives.'' 83 FR at 42990 (Aug.
24, 2018). The preferred alternative in this final rule was within
the range of alternatives considered in the proposal, although it
was not specifically modeled at that time. This issue is discussed
in further detail below.
---------------------------------------------------------------------------
 Regulations regarding implementation of NEPA requires agencies to
``rigorously explore and objectively evaluate all reasonable
alternatives, and for alternatives which were eliminated from detailed
study, briefly discuss the reasons for their having been eliminated.''
\271\ This does not amount to a requirement that agencies evaluate the
widest conceivable spectrum of alternatives. For example, a State
considering adding a single travel lane to a preexisting section of
highway would not be required to consider adding three lanes, or to
consider dismantling the highway altogether.
---------------------------------------------------------------------------
 \271\ 40 CFR 1502.14.
---------------------------------------------------------------------------
 Among thousands of individual comments that mentioned the proposed
standards very generally, some comments addressed the range and
definition of these regulatory alternatives in specific terms, and
these specific comments include comments on the stringency, structure,
and particular provisions defining the set of regulatory alternatives
under consideration.
 As discussed throughout today's notice, the agencies have updated
and otherwise revised many aspects of the analysis. The agencies have
also reconsidered whether the set of alternatives studied in detail
should be expanded to include standards less stringent than the
proposal's preferred alternative, or to include standards more
stringent than the proposal's no-action alternative. On one hand,
comments from Walter Kreucher and AVE cited above indicate the agencies
should consider relaxing standards below MY 2020 levels, and CEI
challenged the agencies' failure to include less-stringent alternatives
in the following comments on this question:
 DOT failed to consider the possibility of freezing CAFE at an
even more lenient standard than currently exists, nor did it
consider making its proposed freeze take effect sooner than MY 2020.
However, as DOT's own analysis strongly indicates, doing so would
lead to even greater benefits and an even greater reduction in CAFE-
related deaths and injuries. In short, DOT's failure to consider
this possibility is arbitrary and capricious. It has an opportunity
to remedy this in its final rule, and it should do so by selecting a
standard that is even more lenient than the one it proposed. . . .
It should have gone beyond its original set of alternatives and
examined less stringent ones as well--until it found one that, for
some reason or another, failed to produce greater safety benefits or
failed to meet the statutory factors.\272\
---------------------------------------------------------------------------
 \272\ CEI, NHTSA-2018-0067-12015, at 1.
 On the other hand, a coalition of ten environmental advocacy
organizations stated that the agencies should consider alternatives
more stringent than those defining the baseline no action alternative,
arguing that in light of CEQ guidance and the 2018 IPCC report on
climate change, ``the increasing danger, increasing urgency, and
increasing importance of vehicle emissions all rationally counsel for
strengthening emission standards.'' \273\ CBD et al. observe that
``none of these alternatives [considered in the NPRM] increases fuel
economy in comparison with the No Action Alternative, none conserves
energy . . .'' and go on to assert that ``none represents maximum
feasible CAFE standards.'' \274\ Similarly, EDF commented that ``. . .
given its clear statutory directive to maximize fuel savings, NHTSA
should have considered a range of alternatives that would be more
protective than the existing standards,'' \275\ and three State
agencies in Minnesota commented that ``more stringent standards are
consistent with EPCA's purpose of energy conservation and the CAA's
purpose of reducing harmful air pollutants.'' \276\ The North Carolina
Department of Environmental Quality acknowledged the agencies'
determination in the proposal that alternatives beyond the augural
standards might be economically impracticable, but nevertheless argued
that ``alternatives that exceed the stringency of the current standards
are consistent with EPCA's purpose'' \277\ In oral testimony before the
agencies, the New York State Attorney General also indicated that the
agencies should consider alternatives more stringent than the augural
standards.\278\ A coalition of States and cities commented that ``at a
minimum, the existing standards should be left in place, but EPA should
also consider whether to make the standards more stringent, not less,
just as it has done in prior proposals.'' \279\ More specifically,
through International Mosaic, some individuals commented that the
agencies must ``fully and publicly consider a few options that require
at least a seven annual percent [sic] improvement in vehicle fleet
mileage.'' \280\ In comments on the DEIS, CBD, et al. went further,
commenting that ``NHTSA's most stringent alternative must be set at no
lower than a 9 percent improvement per year.'' \281\ Most manufacturers
who commented on stringency did not identify specific regulatory
alternatives that the agencies should consider, although Honda
suggested that standards be set to increase in stringency at 5 percent
annually for both passenger cars and light trucks throughout model
years 2021-2026.282 283
---------------------------------------------------------------------------
 \273\ CBD, et al., NHTSA-2018-0067-12057 p. 10. Also, see
comments from Senator Tom Carper, NHTSA-2018-0067-11910, at 8-9, and
from UCS, NHTSA-2018-0067-12039, at 3.
 \274\ CBD, et al., NHTSA-2018-0067-12123, at 12-13.
 \275\ EDF, NHTSA-2018-0067-11996, at 20.
 \276\ Minnesota Pollution Control Agency, Department of
Transportation, and Department of Health, NHTSA-2018-0067-11706, at
5.
 \277\ North Carolina Department of Environmental Quality, NHTSA-
2018-0067-12025, at 37-38.
 \278\ New York State Attorney General, Testimony of Austin
Thompson, NHTSA-2018-0067-12305, at 13.
 \279\ NHTSA-2018-0067-11735, at 49.
 \280\ International Mosaic NHTSA-2018-0067-11154, at 1
 \281\ CBD, et al., NHTSA-2018-0067-12123, at 17.
 \282\ Honda, NHTSA-2018-0067-12019, EPA-HQ-OAR-2018-0283, at 54.
 \283\ In model year 2021, the baseline standards for passenger
cars and light trucks increase by about 4% and 6.5%, respectively,
relative to standards for model year 2020. Depending on the
composition of the future new vehicle fleet (i.e., the footprints
and relative market shares of passenger cars and light trucks), this
amounts to an overall average stringency increase of about 5.5%
relative to model year 2020.
---------------------------------------------------------------------------
 The agencies carefully considered these comments to expand the
range of stringencies to be evaluated as possible candidates for
promulgation. To inform this consideration, the agencies used the CAFE
model to examine a progression of stringencies extending outside the
range presented in the proposal and draft EIS, and as a point of
reference, using a case that reverts to MY 2018 standards starting in
MY 2021. Scenarios included in this initial screening exercise ranged
as high as increasing annually at 9.5 percent during MYs 2021-2026,
reaching average CAFE and CO2 requirements of 66 mpg and 120
g/mi, respectively. Results of this analysis are presented in the
following tables and charts. Focusing on MY 2029, the tables show
average required and achieved CAFE (as mpg) and CO2 (as g/
mi) levels for each scenario, along with average per-vehicle costs (in
2018 dollars, relative to retaining MY 2017 technologies). The proposed
(0%/0%), final (1.5%/1.5%), and baseline augural standards are shown in
bold type. The charts present
[[Page 24259]]
the same results on a percentage basis, relative to values shown below
for the scenario that reverts to MY 2018 standards starting in MY 2021.
 For example, reverting to the MY 2018 CAFE standards starting in MY
2021 yields an average CAFE requirement of 35 mpg by MY 2029, with the
industry exceeding that standard by 5 mpg at an average cost of $1,255
relative to MY 2017 technology. Under the augural standards, the MY
2029 requirement increases to 47 mpg, the average compliance margin
falls to 1 mpg, and the average cost increases to $2,770. In other
words, compared to the scenario that reverts to MY 2018 stringency
starting in MY 2021, the augural standards increase stringency by 34
percent (from 35 to 47 mpg), increase average fuel economy by 20
percent (from 40 to 48 mpg), and increase costs by 121 percent (from
$1,255 to $2,770).
 As indicated in the following two charts, the reality of
diminishing returns clearly applies in both directions. On one hand,
relaxing stringency below the proposed standards by reverting to MY
2018 or MY 2019 standards reduces average MY 2029 costs by only modest
amounts ($54-$121). As discussed in Section VIII, the agencies' updated
analysis indicates that the proposed standards would not be maximum
feasible considering the EPCA/EISA statutory factors, and would not be
appropriate under the CAA after considering the appropriate factors. If
further relaxation of standards appeared likely to yield more
significant cost reductions, it is conceivable that such savings could
outweigh further foregoing of energy and climate benefits. However,
this screening analysis does not show dramatic cost reductions.
Therefore, the agencies did not include these two less stringent
alternatives in the detailed analysis presented in Section VII.
 On the other hand, increases in stringency beyond the baseline
augural standards show relative costs continuing to accrue much more
rapidly than relative CAFE and CO2 improvements. As
discussed below in Section VIII, even the no action alternative is
already well beyond levels that can be supported under the CAA and
EPCA. If further stringency increases appeared likely to yield more
significant additional energy and environmental benefits, it is
conceivable that these could outweigh these significant additional cost
increases. However, this screening analysis shows no dramatic relative
acceleration of energy and environmental benefits. Therefore, the
agencies did not include stringencies beyond the augural standards in
the detailed analysis presented in Section VII.
BILLING CODE 4910-59-P
[GRAPHIC] [TIFF OMITTED] TR30AP20.062
[[Page 24260]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.063
[GRAPHIC] [TIFF OMITTED] TR30AP20.064
[[Page 24261]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.065
BILLING CODE 4910-59-C
 Specific to model year 2021, some commenters argued that EPCA's
lead time requirement prohibits NHTSA from revising CAFE standards for
model year 2021.\284\ Regarding the revision of standards for model
year 2021, NHTSA did consider EPCA's lead time requirement, and
determined that while the agency would need to finalize a stringency
increase at least 18 months before the beginning of the first affected
model year, the agency can finalize a stringency decrease closer (or
even after) the beginning of the first affected model year. The
agency's reasoning is explained further in Section VIII. Therefore,
NHTSA did not change regulatory alternatives to avoid any relaxation of
stringency in model year 2021.
---------------------------------------------------------------------------
 \284\ State of California, et al., NHTSA-2018-0067-11735, at
78.; CBD, et al., NHTSA-2018-0067-12000, Appendix A, at 66.;
National Coalition for Advanced Transportation, NHTSA-2018-0067-
11969, at 46.
---------------------------------------------------------------------------
 The Auto Alliance stated that ``the truck increase rate should be
no greater than the car rate of increase and should be the `equivalent
task' per fleet.'' \285\ Supporting these Alliance comments, FCA
elaborated by commenting that ``(1) in MY2017, the latest data we have
available, most trucks have a larger gap to standards than cars, and
(2) all of the truck segments are challenged because consumers are
placing a greater emphasis on capability than fuel economy.'' \286\
Similarly, Ford commented that ``. . . the rates of increase in the
stringency of the standards should remain equivalent between passenger
cars and light duty trucks.'' \287\ Other commenters expressed general
support for equalizing the rates at which the stringencies of passenger
car and light truck standards increase.\288\
---------------------------------------------------------------------------
 \285\ Alliance, NHTSA-2018-0067-12073, at 7-8
 \286\ FCA, NHTSA-2018-0067-11943, at 46-47.
 \287\ Ford, NHTSA-2018-0067-11928, at 3.
 \288\ See, e.g., Global, NHTSA-2018-0067-12032, at 4; NADA,
NHTSA-2018-0067-12064, at 13; BorgWarner, NHTSA-2018-0067-11895, at
6.
---------------------------------------------------------------------------
 For the final rule, the agencies have added an alternative in which
stringency for both cars and trucks increases at 1.5 percent. This is
consistent with comments received requesting that both fleets'
standards increase in stringency by the same amount, and 1.5 percent
represents a rate of increase within the range of rates of increase
considered in the NPRM.
 Throughout the NPRM, the agencies described their consideration as
covering a range of alternatives.\289\ The preferred alternative for
this final rule, an increase in stringency of 1.5 percent for both cars
and trucks, falls squarely
[[Page 24262]]
within the range of alternatives proposed by the agencies.
---------------------------------------------------------------------------
 \289\ 83 FR at 42986 (Aug. 24, 2018) (explaining, in ``Summary''
section of NPRM, that ``comment is sought on a range of alternatives
discussed throughout this document''); id. at 42988 (stating that
the agencies are ``taking comment on a wide range of alternatives,
including different stringencies and retaining existing
CO2 standards and the augural CAFE standards''); 42990
(``As explained above, the agencies are taking comment on a wide
range of alternatives and have specifically modeled eight
alternatives (including the proposed alternative) and the current
requirements (i.e., baseline/no action).''); 43197 (``[T]oday's
notice also presents the results of analysis estimating impacts
under a range of other regulatory alternatives the agencies are
considering.''); 43229 (explaining that ``technology availability,
development and application, if it were considered in isolation, is
not necessarily a limiting factor in the Administrator's selection
of which standards are appropriate within the range of the
Alternatives presented in this proposal.''); 43369 (``As discussed
above, a range of regulatory alternatives are being considered.'').
---------------------------------------------------------------------------
 The NPRM alternatives were bounded on the upper end by the
baseline/no action alternative, and the proposed alternative on the
lower end (0 percent per year increase in stringency for both cars and
trucks). For passenger cars, the agencies considered a range of
stringency increases between 0 percent and 2 percent per year for
passenger cars, in addition to the baseline/no action alternative. For
light trucks, the agencies considered a range of stringency increases
between 0 percent and 3 percent per year, in addition to the baseline/
no action alternative.
 The agencies considered the same range of alternatives for this
final rule. As with the proposal, the alternatives for stringency are
bounded on the upper end by the baseline/no action alternative and on
the lower end by 0 percent per year increases for both passenger cars
and light trucks. Consistent with the proposal, for this final rule,
the agencies considered stringency increases of between 0 and 2 percent
per year for passenger cars and between 0 and 3 percent per year for
light trucks, in addition to the baseline/no action alternative.
 While it was not specifically modeled in the NPRM, the new
preferred alternative of an increase in stringency of 1.5 percent for
both cars and trucks was well within the range of alternatives
considered. The proposal described the alternatives specifically
modeled as options for the agencies, but also gave notice that they did
not limit the agencies in selecting from among the range of
alternatives under consideration.\290\
---------------------------------------------------------------------------
 \290\ See, e.g., 83 FR at 43003 (Aug. 24, 2018) (``These
alternatives were examined because they will be considered as
options for the final rule. The agencies seek comment on these
alternatives, seek any relevant data and information, and will
review responses. That review could lead to the selection of one of
the other regulatory alternatives for the final rule or some
combination of the other regulatory alternatives (e.g., combining
passenger cars standards from one alternative with light truck
standards from a different alternative).''); id. at 43229
(describing a factor relevant to ``the Administrator's selection of
which standards are appropriate within the range of the Alternatives
presented in this proposal'').
---------------------------------------------------------------------------
 The agencies explained in the proposal that they were ``taking
comment on a wide range of alternatives and have specifically modeled
eight alternatives.'' \291\ As with the proposal, for the final rule,
the agencies specifically modeled the upper and lower bounds of the
baseline/no action alternative and 0 percent per year stringency
increases for both passenger cars and light trucks. In both the
proposal and the final rule, the agencies also modeled a stringency
increase of 2 percent per year for passenger cars and 3 percent per
year for light trucks, as well as a variety of other specific increases
between 0 and 2 percent for passenger cars and 0 and 3 percent for
light trucks.
---------------------------------------------------------------------------
 \291\ 83 FR at 42990 (Aug. 24, 2018).
---------------------------------------------------------------------------
 The specific alternatives the agencies modeled for the final rule
reflect their consideration of public comments. As discussed above,
multiple commenters expressed support for equalizing the rates at which
the stringencies of passenger car and light truck standards increase.
To help the agencies evaluate alternatives that include the same
stringency increase for passenger cars and light trucks, three of the
seven alternatives (in addition to the baseline/no action alternative)
that the agencies specifically modeled for the final rule included the
same stringency increase for passenger cars and light trucks. This
includes the new preferred alternative of an increase in stringency of
1.5 percent for both cars and trucks. This alternative, and all others
specifically modeled for the final rule, falls within the range of
alternatives for stringency considered by the agencies in the proposal.
 Beyond these stringency provisions discussed in the NPRM, the
agencies also sought comment on a number of additional compliance
flexibilities for the programs, as discussed in Section IX.
2. Additional Alternatives Suggested by Commenters
 Beyond the comments discussed above regarding the shapes of the
functions defining fuel economy and CO2 targets, regarding
the inclusion of non-CO2 emissions, and regarding the
stringencies to be considered, the agencies also received a range of
other comments regarding regulatory alternatives.
 Some of these additional comments involved how CAFE and
CO2 standards compare to one another for any given
regulatory alternative. With a view toward maximizing harmonization of
the standards, the Alliance, supported by some of its members'
individual comments, indicated that ``to the degree flexibilities and
incentives are not completely aligned between the CAFE and
[CO2] programs, there must be an offset in the associated
footprint-based targets to account for those differences. Some areas of
particular concerns are air conditioning refrigerant credits, and
incentives for advanced technology vehicles. The Alliance urges the
Agencies to seek harmonization of the standards and flexibilities to
the greatest extent possible. . . .'' \292\
---------------------------------------------------------------------------
 \292\ Alliance, NHTSA-2018-0067-12073, at 40. See also FCA,
NHTSA-2018-0067-11943, at 6-7.
---------------------------------------------------------------------------
 On the other hand, discussing consideration of compliance credits
but making a more general argument, the NYU Institute for Policy
Integrity commented that ``. . . EPA is not allowed to set lower
standards just for the sake of harmonization; to the contrary, full
harmonization may be inconsistent with EPA's statutory
responsibilities.'' \293\ Similarly, ACEEE argued that ``any
consideration of an extension or expansion of credit provisions under
the [carbon dioxide] or CAFE standards program should take as a
starting point the assumption that the additional credits will allow
the stringency of the standards to be increased.'' \294\
---------------------------------------------------------------------------
 \293\ IPI, NHTSA-2018-0067-12213, at 21.
 \294\ ACEEE, NHTSA-2018-0067-12122, at 3.
---------------------------------------------------------------------------
 EPCA's requirement that NHTSA set standards at the maximum feasible
levels is separate and ``wholly independent'' from the CAA's
requirement, per Massachusetts v. EPA, that EPA issue regulations
addressing pollutants that EPA has determined endanger public health
and welfare.\295\ Nonetheless, as recognized by the Supreme Court,
``there is no reason to think the two agencies cannot both administer
their obligations and yet avoid inconsistency.'' \296\ This conclusion
was reached despite the fact that EPCA has a range of very specific
requirements about how CAFE standards are to be structured, how
manufacturers are to comply, what happens when manufacturers are unable
to comply, and how NHTSA is to approach setting standards, and despite
the fact that the CAA has virtually no such requirements. This means
that while nothing about either EPCA or the CAA, much less the
combination of the two, guarantees ``harmonization'' defining ``One
National Program,'' the agencies are expected to be able to work out
the differences.
---------------------------------------------------------------------------
 \295\ Massachusetts v. EPA, 549 U.S. 497, 532 (2007).
 \296\ Id.
---------------------------------------------------------------------------
 Since tailpipe CO2 standards are de facto fuel economy
standards, the more differences there are between CO2 and
CAFE standards and compliance provisions, the more challenging it is
for manufacturers to plan year-by-year production that responses to
both, and the more difficult it is for affected stakeholders and the
general public to understand regulation in this space. Therefore, even
if the two statutes, taken together, do not guarantee ``full
harmonization,'' steps toward greater
[[Page 24263]]
harmonization help with compliance planning and transparency--and meet
the expectations set forth by the Supreme Court that the agencies avoid
inconsistencies.
 The agencies have taken important steps toward doing so. For
example, EPA has adopted separate footprint-based CO2
standards for passenger cars and light trucks, and has redefined CAFE
calculation procedures to introduce recognition for the application of
real-world fuel-saving technology that is not captured with traditional
EPA two-cycle compliance testing. Detailed aspects of both sets of
standards and corresponding compliance provisions are discussed at
length in Section IX. The agencies never set out with the primary goal
of achieving ``full harmonization,'' such that both sets of standards
would lead each manufacturer to respond in exactly the same way in
every model year.\297\ For example, EPA did not adopt the EPCA
requirement that domestic passenger car fleets each meet a minimum
standard, or the EPCA cap on compliance credit transfers between
passenger car fleets. On the other hand, EPA also did not adopt the
EPCA civil penalty provisions that have allowed some manufacturers to
pay civil penalties as an alternative method of meeting EPCA
obligations. These and other differences provide that even if CAFE and
CO2 standards are ``mathematically'' harmonized, for any
given manufacturer, the two sets of standards will not be identically
burdensome in each model year. Inevitably, one standard will be more
challenging than the other, varying over time, between manufacturers,
and between fleets. This means manufacturers need to have compliance
plans for both sets of standards.
---------------------------------------------------------------------------
 \297\ Full harmonization would mean that, for example, if Ford
would do some set of things over time in response to CAFE standards
in isolation, it would do exactly the same things on exactly the
same schedule in response to CO2 standards in isolation.
---------------------------------------------------------------------------
 In 2012, recognizing that EPCA provides no clear basis to address
HFC, CH4, or N2O emissions directly, the agencies
``offset'' CO2 targets from fuel economy targets (after
converting the latter to a CO2 basis) by the amounts of
credit EPA anticipated manufacturers would, on average, earn in each
model years by reducing A/C leakage and adopting refrigerants with
reduced GWPs. In 2012, EPA assumed that by 2021, all manufacturers
would be earning the maximum available credit, and EPA's analysis
assumed that all manufacturers would make progress at the same rate.
However, as discussed above, data highlighted in comments by Chemours,
Inc., demonstrate that actual manufacturers' adoption of lower-GWP
refrigerants thus far ranges widely, with some manufacturers (e.g.,
Nissan) having taken no such steps to move toward lower-GWP
refrigerants, while others (e.g., JLR) have already applied lower-GWP
refrigerants to all vehicles produced for sale in the U.S. Therefore,
at least in practice, HFC provisions thus far continue to leave a gap
(in terms of harmonization) between the two sets of standards. The
proposal would have taken the additional step of decoupling provisions
regarding HFC (i.e., A/C leakage credits), CH4, and
N2O emissions from CO2 standards, addressing
these in separate regulations to be issued in a new proposal. As
discussed above, EPA did not finalize this proposal. Accordingly, for
the regulatory alternatives considered today, EPA has reinstated
offsets of CO2 targets from fuel economy targets, reflecting
the assumption that all manufacturers will be earning the maximum
available A/C leakage credit by MY 2021.
 In addition to general comments on harmonization, the agencies
received a range of comments on specific provisions--especially
involving ``flexibilities''--that may or may not impact harmonization.
With a view toward encouraging further electrification, NCAT proposed
that EPA extend indefinitely the exclusion of upstream emissions from
electricity generation, and also extend and potentially restructure
production multipliers for PHEVs, EVs, and FCVs.\298\ On the other
hand, connecting its comments back to the stringency of standards, NCAT
also commented that ``. . . expansion of compliance flexibilities in
the absence of any requirement to improve [CO2] reduction or
fuel economy (as under the agencies' preferred option) could result in
an effective deterioration of existing [CO2] and fuel
economy performance, as well as little or no effective support for
advanced vehicle technology development or deployment.'' \299\ Global
Automakers indicated that the final rule ``should include a package of
programmatic elements that provide automakers with flexible compliance
options that promote the full breadth of vehicle technologies,'' such
options to include the extension of ``advanced technology'' production
multipliers through MY 2026, the indefinite exclusion of emissions from
electricity generation, the extension to passenger cars of credits
currently granted for the application of ``game changing'' technologies
(e.g., HEVs) only to full-size pickup trucks, an increase (to 15 g/mi)
of the cap on credits for off-cycle technologies, an updated credit
``menu'' of off-cycle technologies, and easier process for handling
applications for off-cycle credits.\300\ The Alliance also called for
expanded sales multipliers and a permanent exclusion of emissions from
electricity generation.\301\ Walter Kreucher recommended the agencies
consider finalizing the proposed standards but also keeping the augural
standards as ``voluntary targets'' to ``provide compliance with the
statutes and an aspirational goal for manufacturers.'' \302\
---------------------------------------------------------------------------
 \298\ NCAT, NHTSA-2018-0067-11969, at 3-5.
 \299\ Id.
 \300\ Global Automakers, NHTSA-2018-0067-12032, at 4 et seq.
 \301\ Alliance, NHTSA-2018-0067-12073, at 8.
 \302\ Kreucher, W., NHTSA-2018-0067-0444, at 9.
---------------------------------------------------------------------------
 The agencies have carefully considered these comments, and have
determined that the current suite of ``flexibilities'' generally
provide ample incentive more rapidly to develop and apply advanced
technologies and technologies that produce fuel savings and/or
CO2 reductions that would otherwise not count toward
compliance. The agencies also share some stakeholders' concern that
expanding these flexibilities could increase the risk of ``gaming''
that would make compliance less transparent and would unduly compromise
energy and environmental benefits. Nevertheless, as discussed in
Section IX, EPA is adopting new multiplier incentives for natural gas
vehicles. EPA is also finalizing some changes to procedures for
evaluating applications for off-cycle credits, and expects these
changes to make this process more accurate and more efficient. Also,
EPA is revising its regulations to not require manufacturers to account
for upstream emissions associated with electricity use for electric
vehicles and plug-in hybrid electric vehicles through model year 2026;
compliance will instead be based on tailpipe emissions performance only
and not include emissions from electricity generation until model year
2027. As discussed below, even with this change, and even accounting
for continued increases in fuel prices and reductions in battery
prices, BEVs are projected in this final rule analysis to continue to
account for less than 5 percent of new light vehicle sales in the U.S.
through model year 2026. To the extent that this projection turns out
to reflect reality, this means that the impact of upstream emissions
from electricity use on the projected CO2
[[Page 24264]]
reductions associated with these standards would likely remain small.
Regarding comments suggesting that the augural standards should be
finalized as ``voluntary targets,'' the agencies have determined that
having such targets exist alongside actual regulatory requirements
would be, at best, unnecessary and confusing.
 Beyond these additional proposals, some commenters' proposals
clearly fell outside authority provided under EPCA or the CAA. Ron
Lindsay recommended the agencies ``consider postponing the rule changes
until the U.S. can establish a legally binding national and
international carbon budget and a binding mechanism to adhere to it.''
\303\ EPCA requires NHTSA to issue standards for MY 2022 by April 1,
2020, and previously-issued EPA regulations commit EPA to revisiting MY
2021-2025 standards on a similar schedule. These statutory and
regulatory provisions do not include a basis to delay decisions pending
an international negotiation for which prospects and schedules are both
unknown.
---------------------------------------------------------------------------
 \303\ Ron Lindsay, EPA-HQ-OAR-2018-0283-1414, at 6.
---------------------------------------------------------------------------
 SCAQMD, supported by Shyam Shukla, indicated that the agencies
should consider an alternative that keeps the waiver for California's
CO2 standards in place.\304\ NCAT and the North Carolina DEQ
offered similar comments and CBD, et al. commented that ``among the set
of more stringent alternatives that NEPA requires the agency to
consider, NHTSA must include action alternatives that retain the
standards California and other states have lawfully adopted.'' \305\ As
discussed above, the agencies recently issued a final rule addressing
the issue of California's authority. NEPA does not require NHTSA to
include action alternatives that cannot be lawfully realized.
---------------------------------------------------------------------------
 \304\ SCAQMD, NHTSA-2018-0067-5666, at 1-2; Shyam Shukla, NHTSA-
2018-0067-5793, at 1-2.
 \305\ NCAT, NHTSA-2018-0067-11969, at 64; NCDEQ, NHTSA-2018-
0067-12025, at 38; CBD et al., NHTSA-2018-0067-12123, Attachment 1,
at 18.
---------------------------------------------------------------------------
 International Mosiac commented that NHTSA's DEIS ``is fatally
flawed . . . because it does not consider any market-based alternatives
(e.g., a `cap and trade' type option).'' \306\ While EPCA/EISA does
include very specific provisions regarding trading of CAFE compliance
credits, the statute provides no authority for a broad-based cap-and-
trade program involving other sectors. Similarly, Michalek, et al.
wrote that ``a more economically efficient approach of, taxing
emissions and fuel consumption at socially appropriate levels would
allow households to determine whether to reduce fuel consumption and
emissions by driving less, by buying a vehicle with more fuel saving
technologies, or by buying a smaller vehicle--or, alternatively, not to
reduce fuel consumption and emissions at all but rather pay a cost
based on the damages they cause. Forcing improvements only through one
mechanism (fuel-saving technologies) increases the cost of achieving
these outcomes.'' \307\ While some economists would agree with these
comments, Congress has provided no clear authority for NHTSA or EPA to
implement either an emissions tax or a broad-based cap-and-trade
program in which motor vehicles could participate.
---------------------------------------------------------------------------
 \306\ International Mosaic, NHTSA-2018-0067-11154, at 1-2.
 \307\ Michalek, et al., NHTSA-2018-0067-11903, at 13.
---------------------------------------------------------------------------
3. Details of Alternatives Considered in Final Rule
a) Alternative 1
 Alternative 1 holds the stringency of targets constant and MY 2020
levels through MY 2026.
[[Page 24265]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.066
b) Alternative 2
 Alternative 2 increases the stringency of targets annually during
MYs 2021-2026 (on a gallon per mile basis, starting from MY 2020) by
0.5 percent for passenger cars and 0.5 percent for light trucks.
[[Page 24266]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.067
c) Alternative 3
 Alternative 3; the final standards promulgated today, increases the
stringency of targets annually during MYs 2021-2026 (on a gallon per
mile basis, starting from MY 2020) by 1.5 percent for passenger cars
and 1.5 percent for light trucks.
[[Page 24267]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.068
d) Alternative 4
 Alternative 4 increases the stringency of targets annually during
MYs 2021-2026 (on a gallon per mile basis, starting from MY 2020) by
1.0 percent for passenger cars and 2.0 percent for light trucks.
[[Page 24268]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.069
e) Alternative 5
 Alternative 5 increases the stringency of targets annually during
MYs 2022-2026 (on a gallon per mile basis, starting from MY 2021) by
1.0 percent for passenger cars and 2.0 percent for light trucks.
[GRAPHIC] [TIFF OMITTED] TR30AP20.070
[[Page 24269]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.071
f) Alternative 6
 Alternative 6 increases the stringency of targets annually during
MYs 2021-2026 (on a gallon per mile basis, starting from MY 2020) by
2.0 percent for passenger cars and 3.0 percent for light trucks.
[GRAPHIC] [TIFF OMITTED] TR30AP20.072
[GRAPHIC] [TIFF OMITTED] TR30AP20.073
[[Page 24270]]
g) Alternative 7
 Alternative 7 increases the stringency of targets annually during
MYs 2022-2026 (on a gallon per mile basis, starting from MY 2021) by
2.0 percent for passenger cars and 3.0 percent for light trucks.
[GRAPHIC] [TIFF OMITTED] TR30AP20.074
[GRAPHIC] [TIFF OMITTED] TR30AP20.075
 EPCA, as amended by EISA, requires that any manufacturer's
domestically-manufactured passenger car fleet must meet the greater of
either 27.5 mpg on average, or 92 percent of the average fuel economy
projected by the Secretary for the combined domestic and non-domestic
passenger automobile fleets manufactured for sale in the U.S. by all
manufacturers in the model year, which projection shall be published in
the Federal Register when the standard for that model year is
promulgated in accordance with 49 U.S.C. 32902(b).\308\ Any time NHTSA
establishes or changes a passenger car standard for a model year, the
MDPCS for that model year must also be evaluated or re-evaluated and
established accordingly. Thus, this final rule establishes the
applicable MDPCS for MYs 2021-2026. Table V-22 lists the minimum
domestic passenger car standards.
---------------------------------------------------------------------------
 \308\ 49 U.S.C. 32902(b)(4).
 [GRAPHIC] [TIFF OMITTED] TR30AP20.076

[[Page 24271]]
VI. Analytical Approach as Applied to Regulatory Alternatives
A. Overview of Methods
 Like analyses accompanying the NPRM and past CAFE and CAFE/
CO2 rulemakings, the analysis supporting today's notice
spans a range of technical topics, uses a range of different types of
data and estimates, and applies several different types of computer
models. The purpose of the analysis is not to determine the standards,
but rather to provide information for consideration in doing so. The
analysis aims to answer the question ``what impacts might each of these
regulatory alternatives have?''
 Over time, NHTSA's and, more recently, NHTSA's and EPA's analyses
have expanded to address an increasingly wide range of types of
impacts. Today's analysis involves, among other things, estimating how
the application of various combinations of technologies could impact
vehicles' costs and fuel economy levels (and CO2 emission
rates), estimating how vehicle manufacturers might respond to standards
by adding fuel-saving technologies to new vehicles, estimating how
changes in new vehicles might impact vehicle sales and operation, and
estimating how the combination of these changes might impact national-
scale energy consumption, emissions, highway safety, and public health.
In addition, the EIS accompanying today's notice addresses impacts on
air quality and climate. The analysis of these factors informs and
supports both NHTSA's application of the statutory requirements
governing the setting of ``maximum feasible'' fuel-economy standards
under EPCA, including, among others, technological feasibility and
economic practicability, and EPA's application of the CAA requirements
for tailpipe emissions.
 Supporting today's analysis, the agencies have brought to bear a
variety of different types of data, a few examples of which include
fuel economy compliance reports, historical sales and average
characteristics of light-duty vehicles, historical economic and
demographic measures, historical travel demand and energy prices and
consumption, and historical measures of highway safety. Also supporting
today's analysis, the agencies have applied several different types of
estimates, a few examples of which include projections of the future
cost of different fuel-saving technologies, projections of future GDP
and the number of households, estimates of the ``gap'' between
``laboratory'' and on-road fuel economy, and estimates of the social
cost of CO2 emissions and petroleum ``price shocks.''
 With a view toward transparency, repeatability, and efficiency, the
agencies have used a variety of computer models to conduct the majority
of today's analysis. For example, the agencies have applied DOE/EIA's
National Energy Modeling System (NEMS) to estimate future energy
prices, EPA's MOVES model to estimate tailpipe emission rates for ozone
precursors and other criteria pollutants, DOE/Argonne's GREET model to
estimate emission rates for ``upstream'' processes (e.g., petroleum
refining), and DOE/Argonne's Autonomie simulation tool to estimate the
fuel consumption impacts of different potential combinations of fuel-
saving technology. In addition, the EIS accompanying today's notice
applies photochemical models to estimate air quality impacts, and
applies climate models to estimate climate impacts of overall emissions
changes.
 Use of these different types of data, estimates, and models is
discussed further below in the most closely relevant sections. For
example, the agencies' use of NEMS is discussed below in the portion of
Section VI that addresses the macroeconomic context, which includes
fuel prices, and the agencies use of Autonomie is discussed in the
portion of Section VI.B.3 that addresses the agencies' approach to
estimating the effectiveness of various technologies (in reducing fuel
consumption and CO2 emissions).
 Providing an integrated means to estimate both vehicle
manufacturers' potential responses to CAFE or CO2 standards
and, in turn, many of the different potential direct results (e.g.,
changes in new vehicle costs) and indirect impacts (e.g., changes in
rates of fleet turnover) of those responses, the CAFE Model plays a
central role in the agencies' analysis supporting today's notice. The
agencies used the specific models mentioned above to develop inputs to
the CAFE model, such as fuel prices and emission factors. Outputs from
the CAFE Model are discussed in Sections VII and VIII of today's
notice, and in the accompanying RIA. The EIS accompanying today's
notice makes use of the CAFE Model's estimates of changes in total
emissions from light-duty vehicles, as well as corresponding changes in
upstream emissions. These changes in emissions are included in the set
of inputs to the models used to estimate air quality and climate
impacts.
 The remainder of this overview focuses on the CAFE Model. The
purpose of this overview is not to provide a comprehensive technical
description of the model,\309\ but rather to give an overview of the
model's functions, to explain some specific aspects not addressed
elsewhere in today's notice, and to discuss some model aspects that
were the subject of significant public comment. Some model functions
and related comments are addressed in other parts of today's notice.
For example, the model's handling of Autonomie-based fuel consumption
estimates is addressed in the portion of Section VI.B.3 that discusses
the agencies' application of Autonomie. The model documentation
accompanying today's notice provides a comprehensive and detailed
description of the model's functions, design, inputs, and outputs.
---------------------------------------------------------------------------
 \309\ The CAFE Model is available at https://www.nhtsa.gov/corporate-average-fuel-economy/compliance-and-effects-modeling-system with documentation and all inputs and outputs supporting
today's notice.
---------------------------------------------------------------------------
1. Overview of CAFE Model
 The basic design of the CAFE Model is as follows: The system first
estimates how vehicle manufacturers might respond to a given regulatory
scenario, and from that potential compliance solution, the system
estimates what impact that response will have on fuel consumption,
emissions, and economic externalities. A regulatory scenario involves
specification of the form, or shape, of the standards (e.g., flat
standards, or linear or logistic attribute-based standards), scope of
passenger car and truck regulatory classes, and stringency of the CAFE
and CO2 standards for each model year to be analyzed.
 Manufacturer compliance simulation and the ensuing effects
estimation, collectively referred to as compliance modeling, encompass
numerous subsidiary elements. Compliance simulation begins with a
detailed user-provided initial forecast of the vehicle models offered
for sale during the simulation period. The compliance simulation then
attempts to bring each manufacturer into compliance with the standards
defined by the regulatory scenario contained within an input file
developed by the user. For example, a regulatory scenario may define
CAFE or CO2 standards that increase in stringency by 4
percent per year for 5 consecutive years.
 The model applies various technologies to different vehicle models
in each manufacturer's product line to simulate how each manufacturer
might make progress toward compliance with the specified standard.
Subject to a variety of user-controlled constraints, the model applies
technologies based on
[[Page 24272]]
their relative cost-effectiveness, as determined by several input
assumptions regarding the cost and effectiveness of each technology,
the cost of compliance (determined by the change in CAFE or
CO2 credits, CAFE-related civil penalties, or value of
CO2 credits, depending on the compliance program being
evaluated and the effective-cost mode in use), and the value of avoided
fuel expenses. For a given manufacturer, the compliance simulation
algorithm applies technologies either until the manufacturer runs out
of cost-effective technologies, until the manufacturer exhausts all
available technologies, or, if the manufacturer is assumed to be
willing to pay civil penalties, until paying civil penalties becomes
more cost-effective than increasing vehicle fuel economy. At this
stage, the system assigns an incurred technology cost and updated fuel
economy to each vehicle model, as well as any civil penalties incurred
by each manufacturer. This compliance simulation process is repeated
for each model year available during the study period.
 This point marks the system's transition between compliance
simulation and effects calculations. At the conclusion of the
compliance simulation for a given regulatory scenario, the system
contains multiple copies of the updated fleet of vehicles corresponding
to each model year analyzed. For each model year, the vehicles'
attributes, such as fuel types (e.g., diesel, electricity), fuel
economy values, and curb weights have all been updated to reflect the
application of technologies in response to standards throughout the
study period. For each vehicle model in each of the model year specific
fleets, the system then estimates the following: Lifetime travel, fuel
consumption, carbon dioxide and criteria pollutant emissions, the
magnitude of various economic externalities related to vehicular travel
(e.g., noise), and energy consumption (e.g., the economic costs of
short-term increases in petroleum prices). The system then aggregates
model-specific results to produce an overall representation of modeling
effects for the entire industry.
 Different categorization schemes are relevant to different types of
effects. For example, while a fully disaggregated fleet is retained for
purposes of compliance simulation, vehicles are grouped by type of fuel
and regulatory class for the energy, carbon dioxide, criteria
pollutant, and safety calculations. Therefore, the system uses model-
by-model categorization and accounting when calculating most effects,
and aggregates results only as required for efficient reporting.
2. Representation of the Market
 As a starting point, the model needs enough information to
represent each manufacturer covered by the program. As discussed below
in Section VI.B.1, the MY 2017 analysis fleet contains information
about each manufacturer's:
 Vehicle models offered for sale--their current (i.e., MY
2017) production volumes, manufacturer suggested retail prices (MSRPs),
fuel saving technology content and other attributes (curb weight, drive
type, assignment to technology class and regulatory class);
 Production considerations--product cadence of vehicle
models (i.e., schedule of model redesigns and ``freshenings''), vehicle
platform membership, degree of engine and/or transmission sharing (for
each model variant) with other vehicles in the fleet; and
 Compliance constraints and flexibilities--preference for
full compliance or penalty payment/credit application, willingness to
apply additional cost-effective fuel saving technology in excess of
regulatory requirements, projected applicable flexible fuel credits,
and current credit balance (by model year and regulatory class) in
first model year of simulation.
Representation of Fuel-Saving Technologies
 The modeling system defines technology pathways for grouping and
establishing a logical progression of technologies that can be applied
to a vehicle. Technologies that share similar characteristics form
cohorts that can be represented and interpreted within the CAFE Model
as discrete entities. The following Table VI-1 shows the technologies
available within the modeling system used for this final rule. Each
technology is discussed in detail below. However, an understanding of
the technologies considered and how they are defined in the model
(e.g., a 6-speed manual transmission is defined as ``MT6'') is helpful
for the following explanation of the compliance simulation and the
inputs required for that simulation.
BILLING CODE 4910-59-P
[[Page 24273]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.077
[[Page 24274]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.078
BILLING CODE 4910-59-C
 These entities are then laid out into pathways (or paths), which
the system uses to define relations of mutual exclusivity between
conflicting sets of technologies. For example, as presented in the next
section, technologies on the Turbo Engine path are incompatible with
those on the HCR Engine or the Diesel Engine paths. As such, whenever a
vehicle uses a technology from one pathway (e.g., turbo), the modeling
system immediately disables the incompatible technologies from one or
more of the other pathways (e.g., HCR and diesel).
 In addition, each path designates the direction in which vehicles
are allowed to advance as the modeling system evaluates specific
technologies for application. Enforcing this directionality within the
model ensures that a vehicle that uses a more advanced or more
efficient technology (e.g., AT8) is not allowed to ``downgrade'' to a
less efficient option (e.g., AT5). Visually, as portrayed in the charts
in the sections that follow, this is represented by an arrow leading
from a preceding technology to a succeeding one, where vehicles begin
at the root of each path, and traverse to each successor technology in
the direction of the arrows.
 The modeling system incorporates twenty technology pathways for
evaluation as shown below. Similar to individual technologies, each
path carries an intrinsic application level that denotes the scope of
applicability of all technologies present within that path, and whether
the pathway is evaluated on one vehicle at a time, or on a collection
of vehicles that share a common platform, engine, or transmission.
[[Page 24275]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.079
 Even though technology pathways outline a logical progression
between related technologies, all technologies available to the system
are evaluated concurrently and independently of each other. Once all
technologies have been examined, the model selects a solution deemed to
be most cost-effective for application on a vehicle. If the modeling
system applies a technology that resides later in the pathway, it will
subsequently disable all preceding technologies from further
consideration to prevent a vehicle from potentially downgrading to a
less advanced option. Consequently, the system skips any technology
that is already present on a vehicle (either those that were available
on a vehicle from the input fleet or those that were previously applied
by the model). This ``parallel technology'' approach, unlike the
``parallel path'' methodology utilized in the preceding versions of the
model, allows the system always to consider the entire set of available
technologies instead of foregoing the application of potentially more
cost-effective options that happen to reside further down the
pathway.\310\ This revised approach addresses comments summarized
below, and allows the system to analyze all available technology
options concurrently and independently of one other without having to
first apply one or more ``predecessor'' technologies. For example, if
model inputs are such that a 7-speed transmission is cost-effective,
but not as cost-effective as an 8-speed transmission, the revised
approach enables the model to skip over the 7-speed transmission
entirely, whereas the NPRM version of the model might first apply the
7-speed transmission and then consider whether to proceed immediately
to the 8-speed transmission. As such, the model's choices for
evaluation of new technology solutions becomes slightly less
restrictive, allowing it immediately to consider and apply more
advanced options, and increasing the likelihood that the a globally
optimum solution is selected.
---------------------------------------------------------------------------
 \310\ Previous versions of the CAFE Model followed a ``low-
cost'' first approach where the system would stop evaluating
technologies residing within a given pathway as soon as the first
cost-effective option within that path was reached.
---------------------------------------------------------------------------
 Some commenters supported the agencies' use of such pathways in the
simulation of manufacturers' potential application of technologies. As
one of a dozen examples of CAFE model design elements that lead to the
transparent representation of real-world factors, the Alliance
highlighted ``recognition of the need for manufacturers to follow
`technology' pathways that retain capital and implementation expertise,
such as specializing in one type of engine or transmission instead of
following an unconstrained optimization that would cause manufacturers
to leap to unrelated technologies and show overly optimistic costs and
benefits.'' \311\ Similarly, Toyota commented that ``the inertia of
capital investments and engineering expertise dedicated to one
compliance technology or set of technologies makes it unreasonable for
manufacturers to immediately switch to another technology path.'' \312\
---------------------------------------------------------------------------
 \311\ Alliance, NHTSA-2018-0067-12073, at 9.
 \312\ Toyota, NHTSA-2018-0067-12098, at 7.
---------------------------------------------------------------------------
 Other commenters cited the use of technology pathways as inherently
overly restrictive. For example, as an example of ``arbitrary model
constraints,'' a coalition of commenters cited the fact the model
``prohibit[s] manufacturers from switching vehicle technology
pathways.'' \313\ Also, EDF, UCS, and CARB cited the combination of
technology pathways, decision making criteria, and model inputs as
producing unrealistic results.\314\ Regarding the technology pathways,
specifically, EDF's consultant argued that the technology paths are not
[[Page 24276]]
transparent, and cited the potential that specific paths may not
necessarily be arranged in progression from least to most cost-
effective--that ``NHTSA ignores the cost of the technology when
developing this list.'' \315\ Relatedly, as EDF's consultant commented:
---------------------------------------------------------------------------
 \313\ CBD, et al., NHTSA-2018-0067-12057, at 3.
 \314\ EDF, NHTSA-2018-0067-12108, Appendix A, at 57 et seq.;
UCS, NHTSA-2018-0067-12039, Appendix, at 25 et seq.; Roush
Industries, NHTSA-2018-0067-11984, at 5.
 \315\ EDF, NHTSA-2018-0067-12108, Appendix B, at 69.
 [T]he Volpe Model is not designed to look backwards along its
technology paths. Thus, the opportunity to recover the expenditure
of inefficient technology is missed. NHTSA might argue that a
manufacturer will not invest in 10-speed transmissions, for example,
and then return to an older design. Whether or not this is true in
real life, such a view would put too much stake in the Volpe Model
projections. The model simply projects what could be done, not what
will be. Anyone examining the progression of technology and noting
the reversion of transmission technology could easily modify the
model inputs to avoid this. Also, if NHTSA evaluated combinations of
technologies prior to entering them in the model piecemeal, it would
automatically avoid such apparent problems.\316\
---------------------------------------------------------------------------
 \316\ Ibid., at 70.
 The agencies also received additional public comments on specific
paths and specific interactions between paths (e.g., involving engines
and hybridization). These comments are addressed below.
 The agencies have carefully considered these comments and the
approach summarized below reflects some corresponding revision. As
mentioned above, the CAFE model now approaches the technology paths in
a such way that, faced with two cost-effective technologies on the same
path, the model can proceed directly to the more advanced technology if
that technology is the more cost effective of the two.
 However, the agencies reject assertions that the model's use of
technology paths is not transparent. The agencies provided extensive
explanatory text, figures, model documentation, and model source code
specifically addressing these paths (and other model features). This
transparency appears evident in that commenters (sometimes while
claiming that a specific feature of the model is not transparent)
presented analytical results involving changes to corresponding inputs
that required a detailed understanding of that feature's operation.
 Regarding comments that the technology paths should be arranged in
order of cost-effectiveness, the agencies note that such comments
presume, without merit, that costs, fuel consumption impacts, and other
inputs (e.g., fuel prices) that logically impact manufacturers'
decision-making are not subject to uncertainty. These inputs are all
subject to uncertainty, and the CAFE Model's arrangement of
technologies into several paths is responsive to these uncertainties.
Nevertheless, the agencies maintain that some technologies do reflect a
higher level of advancement than others (e.g., 10-speed transmissions
vs. 5-speed transmissions), and while manufacturers may, in practice,
occasionally revert to less advanced technologies, it is appropriate
and reasonable to conduct the agencies' analysis in a manner that
assumes manufacturers will continue to make forward progress. As
observed by EDF's consultant's remarks, the CAFE Model ``simply
projects what could be done, not what will be.'' While no model, much
less any model relying on information that can be made publicly
available, can hope to represent precisely each manufacturers' actual
detailed constrains related to product development and planning, such
constraints are real and important. The agencies agree that the CAFE
Model's representation of such constraints--including the Model's use
of technology paths--provides a reasonable means of accounting for
them.
4. Compliance Simulation
 The CAFE model provides a way of estimating how vehicle
manufacturers could attempt to comply with a given CAFE standard by
adding technology to fleets that the agencies anticipate they will
produce in future model years. This exercise constitutes a simulation
of manufacturers' decisions regarding compliance with CAFE or
CO2 standards.
 This compliance simulation begins with the following inputs: (a)
The analysis fleet of vehicles from model year 2017 discussed below in
Section VI.B.1, (b) fuel economy improving technology estimates
discussed below in Section VI.C, (c) economic inputs discussed below in
Section VI.D, and (d) inputs defining baseline and potential new CAFE
or CO2 standards discussed above in Section V. For each
manufacturer, the model applies technologies in both a logical sequence
and a cost-optimizing strategy in order to identify a set of
technologies the manufacturer could apply in response to new CAFE or
CO2 standards. The model applies technologies to each of the
projected individual vehicles in a manufacturer's fleet, considering
the combined effect of regulatory and market incentives while
attempting to account for manufacturers' production constraints.
Depending on how the model is exercised, it will apply technology until
one of the following occurs:
 (1) The manufacturer's fleet achieves compliance \317\ with the
applicable standard and adding additional technology in the current
model year would be attractive neither in terms of stand-alone (i.e.,
absent regulatory need) cost-effectiveness nor in terms of facilitating
compliance in future model years;
---------------------------------------------------------------------------
 \317\ When determining whether compliance has been achieved in
the CAFE program, existing CAFE credits that may be carried over
from prior model years or transferred between fleets are also used
to determine compliance status. For purposes of determining the
effect of maximum feasible CAFE standards, however, EPCA prohibits
NHTSA from considering these mechanisms for years being considered
(though it does so for model years that are already final) and the
agency runs the CAFE model without enabling these options. 49 U.S.C.
32902(h)(3).
---------------------------------------------------------------------------
 (2) The manufacturer ``exhausts'' available technologies; \318\ or
---------------------------------------------------------------------------
 \318\ In a given model year, it is possible that production
constraints cause a manufacturer to ``run out'' of available
technology before achieving compliance with standards. This can
occur when: (a) An insufficient volume of vehicles are expected to
be redesigned, (b) vehicles have moved to the ends of each
(relevant) technology pathway, after which no additional options
exist, or (c) engineering aspects of available vehicles make
available technology inapplicable (e.g., secondary axle disconnect
cannot be applied to two-wheel drive vehicles).
---------------------------------------------------------------------------
 (3) For manufacturers assumed to be willing to pay civil penalties
(in the CAFE program), the manufacturer reaches the point at which
doing so would be more cost-effective (from the manufacturer's
perspective) than adding further technology.
 The model accounts explicitly for each model year, applying
technologies when vehicles are scheduled to be redesigned or freshened
and carrying forward technologies between model years once they are
applied (until, if applicable, they are superseded by other
technologies). The model then uses these simulated manufacturer fleets
to generate both a representation of the U.S. auto industry and to
modify a representation of the entire light-duty registered vehicle
population. From these fleets, the model estimates changes in physical
quantities (gallons of fuel, pollutant emissions, traffic fatalities,
etc.) and calculates the relative costs and benefits of regulatory
alternatives under consideration.
 The CAFE model accounts explicitly for each model year, in turn,
because manufacturers actually ``carry forward'' most technologies
between model years, tending to concentrate the application of new
technology to vehicle redesigns or mid-cycle ``freshenings,'' and
design cycles vary widely among manufacturers and specific products.
[[Page 24277]]
Comments by manufacturers and model peer reviewers strongly support
explicit year-by-year simulation. Year-by-year accounting also enables
accounting for credit banking (i.e., carry-forward), as discussed
above, and at least four environmental organizations recently submitted
comments urging the agencies to consider such credits, citing NHTSA's
2016 results showing impacts of carried-forward credits.\319\ Moreover,
EPCA/EISA requires that NHTSA make a year-by-year determination of the
appropriate level of stringency and then set the standard at that
level, while ensuring ratable increases in average fuel economy through
MY 2020. The multi-year planning capability, simulation of ``market-
driven overcompliance,'' and EPCA credit mechanisms (again, for
purposes of modeling the CAFE program) increase the model's ability to
simulate manufacturers' real-world behavior, accounting for the fact
that manufacturers will seek out compliance paths for several model
years at a time, while accommodating the year-by-year requirement. This
same multi-year planning structure is used to simulate responses to
standards defined in grams CO2/mile, and utilizing the set
of specific credit provisions defined under EPA's program.
---------------------------------------------------------------------------
 \319\ Comment by Environmental Law & Policy Center, Natural
Resources Defense Council (NRDC), Public Citizen, and Sierra Club,
Docket ID EPA-HQ-OAR-2015-0827-9826, at 28-29.
---------------------------------------------------------------------------
 After the light-duty rulemaking analysis accompanying the 2012
final rule that finalized NHTSA's standards through MY 2021, NHTSA
began work on changes to the CAFE model with the intention of better
reflecting constraints of product planning and cadence for which
previous analyses did not account. This involves accounting for
expected future schedules for redesigning and ``freshening'' vehicle
models, and accounting for the fact that a given engine or transmission
is often shared among more than one vehicle model, and a given vehicle
production platform often includes more than one vehicle model. These
real product planning considerations are explained below.
 Like earlier versions, the current CAFE model provides the
capability for integrated analysis spanning different regulatory
classes, accounting both for standards that apply separately to
different classes and for interactions between regulatory classes.
Light vehicle CAFE and CO2 standards are specified
separately for passenger cars and light trucks. However, there is
considerable sharing between these two regulatory classes, where a
single engine, transmission, or platform can appear in both the
passenger car and light truck regulatory class. For example, some
sport-utility vehicles are offered in 2WD versions (classified as
passenger cars for compliance purposes) and 4WD versions (classified as
light trucks for compliance purposes). Integrated analysis of
manufacturers' passenger car and light truck fleets provides the
ability to account for such sharing and reduces the likelihood of
finding solutions that could involve introducing impractical and
unrealistic levels of complexity in manufacturers' product lines. In
addition, integrated fleet analysis provides the ability to simulate
the potential that manufacturers could earn CAFE and CO2
credits by over complying with the standard in one fleet and use those
credits toward compliance with the standard in another fleet (i.e., to
simulate credit transfers between regulatory classes).\320\
---------------------------------------------------------------------------
 \320\ Note, however, that EPCA prohibits NHTSA from considering
the availability of such credit trading when setting maximum
feasible fuel economy standards. 49 U.S.C. 32902(h)(3).
---------------------------------------------------------------------------
 The CAFE model also accounts for EPCA's requirement that compliance
be determined separately for fleets of domestic passenger cars and
fleets of imported passenger cars. The model accounts for all three
CAFE regulatory classes simultaneously (i.e., in an integrated way) yet
separately: Domestic passenger cars, imported passenger cars, and light
trucks. The model further accounts for two related specific statutory
requirements specifically involving this distinction between domestic
and imported passenger cars. First, EPCA/EISA requires that any given
fleet of domestic passenger cars meet a minimum standard, irrespective
of any available compliance credits. Second, EPCA/EISA requires
compliance with the standards applicable to the domestic passenger car
fleet without regard to traded or transferred credits.\321\
---------------------------------------------------------------------------
 \321\ 49 U.S.C. 32903(f)(2) and (g)(4).
---------------------------------------------------------------------------
 However, the CAA has no such limitation regarding compliance by
domestic and imported vehicles; EPA did not adopt provisions similar to
the aforementioned EPCA/EISA requirements and is not doing so today.
Therefore, the CAFE model determines compliance for manufacturers'
overall passenger car and light truck fleets for EPA's program.
 Each manufacturer's regulatory requirement represents the
production-weighted harmonic mean of their vehicle's targets in each
regulated fleet. This means that no individual vehicle has a
``standard,'' merely a target, and each manufacturer is free to
identify a compliance strategy that makes the most sense given its
unique combination of vehicle models, consumers, and competitive
position in the various market segments. As the CAFE model provides
flexibility when defining a set of regulatory standards, each
manufacturer's requirement is dynamically defined based on the
specification of the standards for any simulation and the distribution
of footprints within each fleet.
 Given this information, the model attempts to apply technology to
each manufacturer's fleet in a manner that, given product planning and
engineering-related considerations, optimizes the selected cost-related
metric. The metric supported by the NPRM version of the model is termed
``effective cost.'' The effective cost captures more than the
incremental cost of a given technology; it represents the difference
between their incremental cost and the value of fuel savings to a
potential buyer over the first 30 months of ownership.\322\ In addition
to the technology cost and fuel savings, the effective cost also
includes the change in CAFE civil penalties from applying a given
technology and any estimated welfare losses associated with the
technology (e.g., earlier versions of the CAFE model simulated low-
range electric vehicles that produced a welfare loss to buyers who
valued standard operating ranges between re-fueling events). Comments
on this metric are discussed below, as are model changes responding to
these comments.
---------------------------------------------------------------------------
 \322\ The length of time over which to value fuel savings in the
effective cost calculation is a model input that can be modified by
the user. This analysis uses 30 months' worth of fuel savings in the
effective cost calculation, using the price of fuel at the time of
vehicle purchase.
---------------------------------------------------------------------------
 This construction allows the model to choose technologies that both
improve a manufacturer's regulatory compliance position and are most
likely to be attractive to its consumers. This also means that
different assumptions about future fuel prices will produce different
rankings of technologies when the model evaluates available
technologies for application. For example, in a high fuel price regime,
an expensive but very efficient technology may look attractive to
manufacturers because the value of the fuel savings is sufficiently
high both to counteract the higher cost of the technology and,
implicitly, to satisfy consumer demand to balance price increases with
reductions in operating cost.
[[Page 24278]]
 In general, the model adds technology for several reasons but
checks these sequentially. The model then applies any ``forced''
technologies. Currently, only variable valve timing (VVT) is forced to
be applied to vehicles at redesign since it is the root of the engine
path and the reference point for all future engine technology
applications.\323\ The model next applies any inherited technologies
that were applied to a leader vehicle on the same vehicle platform and
carried forward into future model years where follower vehicles (on the
shared system) are freshened or redesigned (and thus eligible to
receive the updated version of the shared component). In practice, very
few vehicle models enter without VVT, so inheritance is typically the
first step in the compliance loop. Next, the model evaluates the
manufacturer's compliance status, applying all cost-effective
technologies regardless of compliance status.\324\ Then the model
applies expiring overcompliance credits (if allowed to do so under the
perspective of either the ``unconstrained'' or ``standard setting''
analysis, for CAFE purposes).\325\ At this point, the model checks the
manufacturer's compliance status again. If the manufacturer is still
not compliant (and is unwilling to pay civil penalties, again for CAFE
modeling), the model will add technologies that are not cost-effective
until the manufacturer reaches compliance. If the manufacturer exhausts
opportunities to comply with the standard by improving fuel economy/
reducing emissions (typically due to a limited percentage of its fleet
being redesigned in that year), the model will apply banked CAFE or
CO2 credits to offset the remaining deficit. If no credits
exist to offset the remaining deficit, the model will reach back in
time to alter technology solutions in earlier model years.
---------------------------------------------------------------------------
 \323\ As a practical matter, this affects very few vehicles.
More than 95 percent of vehicles in the market file either already
have VVT present or have surpassed the basic engine path through the
application of hybrids or electric vehicles.
 \324\ For further explanation of how the CAFE model considers
the effective cost of applying different technologies see the CAFE
Model Documentation for the final rule, at S5.3 Compliance
Simulation Algorithm.
 \325\ As mentioned above, EPCA prohibits consideration of
available credits when setting maximum feasible fuel economy
standards. 49 U.S.C. 32902(h)(3).
---------------------------------------------------------------------------
 The CAFE model implements multi-year planning by looking back,
rather than forward. When a manufacturer is unable to comply through
cost-effective (i.e., producing effective cost values less than zero)
technology improvements or credit application in a given year, the
model will ``reach back'' to earlier years and apply the most cost-
effective technologies that were not applied at that time and then
carry those technologies forward into the future and re-evaluate the
manufacturer's compliance position. The model repeats this process
until compliance in the current year is achieved, dynamically
rebuilding previous model year fleets and carrying them forward into
the future, and accumulating CAFE or CO2 credits from over-
compliance with the standard wherever appropriate.
 In a given model year, the model determines applicability of each
technology to each vehicle platform, model, engine, and transmission.
The compliance simulation algorithm begins the process of applying
technologies based on the CAFE or CO2 standards specified
during the current model year. This involves repeatedly evaluating the
degree of noncompliance, identifying the next ``best'' technology
(ranked by the effective cost discussed earlier) available on each of
the parallel technology paths described above and applying the best of
these. The algorithm combines some of the pathways, evaluating them
sequentially instead of in parallel, to ensure appropriate incremental
progression of technologies.
 The algorithm first finds the best next applicable technology in
each of the technology pathways and then selects the best among these.
For CAFE purposes, the model applies the technology to the affected
vehicles if a manufacturer is either unwilling to pay penalties or if
applying the technology is more cost-effective than paying penalties.
Afterwards, the algorithm reevaluates the manufacturer's degree of
noncompliance and continues application of technology. Once a
manufacturer reaches compliance (i.e., the manufacturer would no longer
need to pay penalties), the algorithm proceeds to apply any additional
technology determined to be cost-effective (as discussed above).
Conversely, if a manufacturer is assumed to prefer to pay penalties,
the algorithm only applies technology up to the point where doing so is
less costly than paying penalties. The algorithm stops applying
additional technology to this manufacturer's products once no more
cost-effective solutions are encountered. This process is repeated for
each manufacturer present in the input fleet. It is then repeated for
each model year. Once all model years have been processed, the
compliance simulation algorithm concludes. The process for
CO2 standard compliance simulation is similar, but without
the option of penalty payment, such that technologies are applied until
compliance (accounting for any modeled application of credits) is
achieved. For both CAFE and CO2 standards, the model also
applies any additional (i.e., beyond required for compliance)
technology that ``pays back'' within a specified period (for the NPRM
and today's analysis, 30 months).
 Some commenters argued that the CAFE model applies constraints that
excessively limit options manufacturers have to add technology, causing
the model to overestimate costs to achieve a given level of
improvement.\326\ Some of these commenters further argued that the
agencies should assume greater potential to apply technologies that
contribute to compliance by improving air conditioner efficiency or
otherwise reducing ``off cycle'' fuel consumption and CO2
emissions.\327\ Other commenters argued that such constraints, while
warranting some refinements, help the model to simulate manufacturers'
decision making realistically and to estimate technology effectiveness
and costs reasonably.328 329
---------------------------------------------------------------------------
 \326\ NHTSA-2018-0067-12057, CBD, et. al, p. 3.
 \327\ NHTSA-2018-0067-11741, ICCT, Attachment 2, p. 4.
 \328\ NHTSA-2018-0067-12073, Alliance of Automobile
Manufacturers, pp. 134-36.
 \329\ American Honda Motor Co., ``Honda Comments on the NPRM and
various proposals contained therein--Prepared for NHTSA, EPA and
ARB,'' October 17, 2018, pp. 12-16.
---------------------------------------------------------------------------
 Some commenters questioned the ``effective cost'' metric the model
uses to decide among available options, claiming that the metric also
causes the model to avoid selection of pathways that are not always
economically optimal.\330\ One of these commenters recommended the
agencies modify the effective cost metric for CO2 compliance
by removing the term placing a monetary value on progress toward
compliance, and instead dividing the remaining net cost (i.e., the
increase in technology costs minus a portion of the fuel outlays
expected to be avoided) by the additional CO2 credits
earned.\331\ Another of these commenters claimed on one hand, that the
effective cost metric ``does not include a measurement of the
technology's reduction in fuel consumption or CO2
emissions'' and, on the other, that the metric inappropriately places a
value on avoided fuel consumption.\332\
---------------------------------------------------------------------------
 \330\ NHTSA-2018-0067-11741, ICCT, Attachment 3, p. I-62.
 \331\ NHTSA-2018-0067-12039, UCS, Technical Appendix, pp. 28-32.
 \332\ NHTSA-2018-0067-12108, EDF, Appendix B, p. 67.
---------------------------------------------------------------------------
 One commenter claimed that the model inappropriately allows earned
[[Page 24279]]
credits (including CO2 program credits for which EPA has
granted a one-time exemption from carry-forward limits) to expire while
also showing undue degrees overcompliance with standards, and further
proposed that the model be modified to simulate both credit ``carry
back'' (aka ``borrowing'') and credit trading between
manufacturers.\333\
---------------------------------------------------------------------------
 \333\ NHTSA-2018-0067-12039, UCS, Technical Appendix, pp. 36-40.
---------------------------------------------------------------------------
 In addition, some commenters indicated that the agencies' analysis
(impliedly, its modeling) should account for some States' mandates that
manufacturers sell minimum quantities of ``Zero Emission Vehicles''
(ZEVs).334 335
---------------------------------------------------------------------------
 \334\ NHTSA-2018-0067-12036, Volvo, p. 5.
 \335\ NHTSA-2018-0067-11813, South Coast AQMD, Attachment 1, p.
4 and EIS comments, p. 9.
---------------------------------------------------------------------------
 Regarding the model's representation of engineering and product
planning constraints, the agencies maintain that having such
constraints produces more realistic potential (as mentioned above, not
``predicted'') pathways forward from manufacturers' current fleets than
would be the case were these constraints removed. For example, while
manufacturers' product plans are protected as confidential business
information (CBI), some manufacturers' public comments demonstrate
year-by-year balancing such as the CAFE model emulates.\336\ Also, even
manufacturers that have invested in technologies such as hybrid
electric powertrains and Atkinson cycle engines have commented that a
manufacturers' past investments will constrain the pathways it can
practicably take.\337\ Therefore, the agencies have retained the
model's basic structural constraints, have updated and expanded the
model's technology paths (and, as discussed, the model's logic for
approaching these paths), and have updated inputs defining the range of
manufacturer-, technology-, and product-specific constraints. These
updates are discussed below at greater length.
---------------------------------------------------------------------------
 \336\ See, e.g., FCA, pp. 5-6.
 \337\ Toyota, Attachment 1, p. 10.
---------------------------------------------------------------------------
 The agencies have also reconsidered opportunities manufacturers may
have to expand the application of technologies that contribute to
compliance by improving air conditioner efficiency or otherwise
reducing ``off cycle'' fuel consumption and CO2 emissions,
or to earn credit toward CO2 compliance by using
refrigerants with lower global warming potential (GWP) or reducing the
potential for refrigerant leaks. The version of the model used for the
proposal accommodates inputs that, for each of these adjustments or
credits, applies the same value to every model year. The agencies have
revised the model to accommodate inputs that specify the degree of
adjustment or credit separately for each model year, and have applied
inputs that assume manufacturers will increase application of these
improvements to the highest levels reported within the industry.
 Regarding comments on the effective cost metric the model uses to
compare and select among available options to add technology, the
agencies have considered changes such as those mentioned above. Given
the myriad of factors that manufacturers can consider, any weighing to
be conducted using publicly-available information will constitute a
simplified representation. Nevertheless, within the model's context, it
is obvious that any weighing of options should, at a minimum, consider
some measure of each option's costs and benefits. Since this aspect of
the model involves simulating manufacturers' decisions, it is also
clearly appropriate that these costs and benefits be considered from a
manufacturer perspective rather than a social perspective.
 The effective cost metric used for the NPRM version of the model
represents the cost of a given option as the cost to apply a given
technology to a given set of vehicles, and represents the benefit of
the same option as the extent to which the manufacturer might expect
buyers would be willing to pay for fuel economy (as represented by a
portion of the projected fuel savings), combined with any reduction in
CAFE civil penalties that the manufacturer might ultimately need to
pass along to buyers. The reduction in CAFE civil penalties places a
value on progress made toward compliance with CAFE standards. The CAA
provides no direction regarding CO2 standards, so the model
accepts inputs specifying an analogous basis for valuing changes in the
quantity of CO2 credits earned from (or required by) a
manufacturer's fleet. Because each of these three components
(technology cost, fuel benefit, and compliance benefit) is expressed in
dollars, subtracting benefits from costs produces a net cost, and after
dividing net costs by the number of affected vehicles, it is logical
to, at each step, select the option that produces the most negative net
unit cost. This approach can be interpreted as maximizing net benefits
(to the manufacturer).
 As an alternative, the agencies considered a simpler metric that
considers only the cost of the option and the extent to which the
option increases the quantity of earned credits, and does not require
input assumptions regarding how to value progress toward compliance.
Such a metric is expressed in dollars per ton or dollars per gallon
such that seeking options that produce the smallest (positive) values
can be interpreted as maximizing cost effectiveness (of progress toward
compliance). However, simply comparing technology costs to
corresponding compliance improvements would implicitly assume that
manufacturers do not respond at all to fuel prices. This assumption is
clearly unrealistic. For example, if diesel fuel costs $5 per gallon
and gasoline costs $2 per gallon, manufacturers will be reluctant to
respond to stringent CAFE or CO2 standards by replacing
gasoline engines with diesel engines. Manufacturers' comments credibly
assert that fuel prices matter, and in the agencies' judgment,
simulations of decisions between available options should continue to
account for avoided fuel outlays.
 On the other hand, while any metric should incorporate some measure
of progress toward compliance, it is not obvious that this progress
must be expressed in monetary terms. While the CAFE civil penalty
provisions provide a logical basis for doing so with respect to CAFE,
the recently-introduced (through EISA) option to trade credit between
manufacturers adds an alternative basis that is undefined and
uncertain, in part because terms of past trades are not known to the
agencies. Also, as mentioned above, EPCA/EISA's civil penalty
provisions are not applicable to noncompliance with CO2
standards.
 Therefore, for the purpose of selecting among available options to
add technology, the agencies consider it reasonable to use the degree
of compliance improvement in ``raw'' (i.e., not monetized) form, and to
divide net costs (i.e., technology costs minus a portion of expected
avoided fuel outlays) by this improvement. Under a range of side-by-
side tests, this change to the effective cost metric most frequently
produced lower overall estimates of compliance costs. However,
differences vary among manufacturers, model years, and regulatory
alternatives, and also depend on other model inputs. For example, at
high fuel prices, the new metric tends to select more expensive
pathways than the NPRM's metric, and with the new metric, a case
simulating ``perfect trading'' of CO2 compliance credits
tends to show such trading increasing compliance costs rather than, as
expected, decreasing such costs.
 The version of the model used for the proposal simulates the
potential that, for
[[Page 24280]]
a given fleet in a given model year, a manufacturer might be able to
use credits from an earlier model year or a different fleet. This
version of the model did not explicitly simulate the potential that,
for a given fleet in a given model year, a manufacturer might be to use
credits from a future model year or a different manufacturer. However,
the agencies did apply model inputs that reflected assumptions
regarding possible trading of credits actually earned prior to model
year 2016 (the earliest represented in detail in the agencies'
analysis), and the agencies did examine a case (included in the
sensitivity analysis) involving hypothetical ``perfect'' trading of
CO2 credits among manufacturers by treating the industry as
a single ``manufacturer.'' Although past versions of the CAFE Model had
included code under development with a view toward eventually
simulating one or both of these provisions, this code had never
proceeded beyond preliminary experimentation, and had never been the
focus of peer reviews or application in published analyses.
 Nevertheless, the agencies considered expanding the model to
simulate credit ``carry back'' (or ``borrowing'') and trading
(explicitly, rather than in an idealized hypothetical way). The
agencies closely examined the corresponding model revisions proposed by
UCS and determined that such methods would not produce repeatable
results. This is because the approach proposed by UCS ``randomly swaps
items in list to minimize trading bias.'' \338\
---------------------------------------------------------------------------
 \338\ UCS, NHTSA-2018-0067-12039, Technical Appendix, at 84-87.
---------------------------------------------------------------------------
 Even if such revisions could be modified to produce non-random
results, including credit banking and trading would introduce highly
speculative elements into the agencies' analysis. While manufacturers
have occasionally indicated plans to carry back credits from future
model years, those plans have sometimes backfired when projected
credits have failed to materialize, e.g., by misjudging consumer demand
for more efficient vehicles. In the agencies' judgment, it would be
inappropriate to set standards based on an analysis that relies on the
type of borrowing that has been known to fail. To rely also on credit
trading during the model years included in the analysis would compound
this undue speculation. For example, including credit borrowing and
trading throughout the analysis, as some commenters proposed, would
lead to an analysis that depends on the potential that, in order to
comply with the MY 2022 standard for light trucks, FCA could use
credits it expects to be able to buy from another manufacturer in MY
2025. Even if the agencies' analysis had knowledge of and made use of
manufacturers' actual product plans, expectations about the ability to
borrow others' unearned credits would necessarily be considered risky
and unreliable. Within an analysis that, to provide for public
disclosure, extrapolates forward many years from the most recent
observed fleet, such transactions would add an unreasonable level of
speculation. Therefore, the agencies have declined to introduce credit
borrowing and trading into the model's logic.
 The analysis presented in the proposal applied inputs reflecting
potential application of credits earned earlier than the first year
modeled explicitly. However, as observed by some commenters, those
inputs did not fully account for the one-time exemption from the 5-year
limit on the extent to which manufacturers may carry forward
CO2 credits. The agencies have updated the analysis fleet to
MY 2017 and, in doing so, have updated inputs specifying how credits
earned to MY 2017 might be applied. These updates implement a
reasonably full accounting of these ``legacy'' credits, including of
the one-time exemption from the credit life limit.
 As mentioned above, some commenters also indicated that the model
is unrealistically ``reluctant'' to apply credits carried forward from
early model years. As explained in the proposal and in the model
documentation, the model's application of carried-forward credits is
partially controlled by model inputs, which, for the proposal, were set
to assume that manufacturers would tend to retain credits as long as
possible. This assumption is entirely consistent with manufacturers'
past practice and logical in a context wherein the stringency of
standards is generally increasing over time. Even though using credits
in some model years might seem initially advantageous, doing so means
foregoing actual improvements likely to be needed in later model years.
 Regarding the model's treatment of mandates and credits for the
sale of ZEVs, as indicated in the model documentation accompanying the
proposal, these capabilities were experimental in that version of the
model. The reference case analysis for today's notice, like that for
the proposal, does not simulate compliance with ZEV mandates.\339\
---------------------------------------------------------------------------
 \339\ The agencies note their finalization of the One National
Program Final Action, in which EPA partially withdrew a waiver of
CAA preemption previously granted to the State of California
relating to its ZEV mandate, and NHTSA finalized regulations
providing that State ZEV mandates are impliedly and expressly
preempted by EPCA. This joint action is available at 84 FR 51310.
---------------------------------------------------------------------------
 For the NPRM, the CAFE model was exercised with inputs extending
this explicit simulation of technology application through MY 2032, as
the agencies anticipated this was sufficiently beyond MY 2026 that
nearly all multiyear planning attributable to MY 2026 standards should
be accounted for, and any compliance credits carried forward from MY
2026 would have expired. The analysis met this expectation, and the
agencies presented analysis of the resultant estimated impacts over the
useful lives of vehicles produced prior to MY 2030. The agencies
invited comment on all aspects of the analysis, and relevant to this
aspect of the analysis--i.e., its perspective and temporal span--EDF
stated that that these led the agencies to overstate the proposal's
positive impacts on safety, in part because by explicitly representing
vehicle model years only through 2032, the agencies had failed to
account for the impact of distant model years prices and fuel economy
levels on the retention and scrappage of vehicles produced through MY
2029.\340\ For example, some vehicles produced in MY 2026 will likely
still be on the road during calendar years (CY) 2033-2050 and the rates
at which these MY 2026 vehicles will be scrapped during CYs 2033-2050
will be impacted by the prices and fuel economy levels of vehicles
produced during MYs 2033-2050.
---------------------------------------------------------------------------
 \340\ EDF, NHTSA-2018-0067-12108, Attachment A at 11 and
Attachment B at 11-28.
---------------------------------------------------------------------------
 The agencies have addressed this comment by expanding model inputs
to extend the explicit simulation of technology application through MY
2050. Most of these expanded model inputs involve the analysis fleet
and inputs defining the cost and availability of various fuel-saving
technologies. These inputs are discussed below. The agencies also made
minor modifications to the model in order to extend model outputs to
cover this wider span and to carry forward each regulatory
alternative's standards automatically through the last year to be
modeled (e.g., extending standards without change from MY 2032 through
MY 2050). The model documentation discusses these
[[Page 24281]]
minor changes.\341\ In addition, although the agencies published
detailed model output files documenting all estimated annual impacts
through calendar year 2089, the notice and PRIA both emphasized the
above-mentioned ``model year'' perspective, as in past regulatory
analyses supporting CAFE and CO2 standards. Recognizing that
an alternative ``calendar year'' perspective is of interest to EDF and,
perhaps other stakeholders, the agencies have expanded the presentation
of results in today's notice and FRIA by presenting some physical
impacts (e.g., fuel consumption and CO2 emissions) as well
as monetized benefits, costs, and net benefits for each of CYs 2017-
2050. All of these results appear in the model output files published
with today's notice, as do corresponding results for more specific
impacts (e.g., year-by-year components of monetized social costs).\342\
---------------------------------------------------------------------------
 \341\ The model and documentation are available at https://www.nhtsa.gov/corporate-average-fuel-economy/compliance-and-effects-modeling-system.
 \342\ Detailed model inputs and outputs are available at https://www.nhtsa.gov/corporate-average-fuel-economy/compliance-and-effects-modeling-system.
---------------------------------------------------------------------------
5. Calculation of Physical Impacts
 Once it has completed the simulation of manufacturers' potential
application of technology in response to CAFE/CO2 standards
and fuel prices, the CAFE Model calculates impacts of the resultant
changes in new vehicle fuel economy levels and prices. This involves
several steps.
 The model calculates changes in the total quantity of new vehicles
sold in each model year as well as the relative shares passenger cars
and light trucks comprise of the overall new vehicle market. The
agencies received many comments on the estimation of sales impacts, and
as discussed below, today's analysis applies methods and corresponding
estimates that reflect careful consideration of these comments. Related
to these calculations, the model now operates in an iterated fashion
with a view toward obtaining sales impacts that are balanced with
changes in vehicle prices and fuel economy levels. This involves
solving for compliance, calculating sales impacts, re-solving for
compliance, and repeating these steps as many times as specified in
model inputs. For today's analysis, the agencies operated the model
with four iterations, as early testing suggested three iterations
should be sufficient for fleetwide results to converge between
iterations. The model documentation describes the procedures for
iteration in detail.
 The impacts on outlays for new vehicles occur coincident with the
sale of these vehicles so the model can simply calculate and record
these for each model year included in the analysis. However, virtually
all other impacts result from vehicle operation that extends long after
a vehicle is produced. Like other models (including, e.g., NEMS), the
CAFE Model includes procedures (sometimes referred to as ``stock
models'' or as models of fleet turnover) to estimate annual rates at
which new vehicles are used and subsequently scrapped. The agencies
received many comments on procedures for estimating vehicle scrappage
and on procedures for estimating annual quantities of highway travel,
accounting for the elasticity of travel demand with respect to per-mile
costs for fuel. Below, Section VI.D.1 discusses these comments and
reviews procedures and corresponding estimates that also reflect
careful consideration of these comments.
 For each vehicle model in each model year, these procedures result
in estimates of the number of vehicles remaining in service in each
calendar year, as well as the annual mileage accumulation (i.e.,
vehicle miles traveled, or VMT) in each calendar year. As mentioned
above, most of the physical impacts of interest derive from this
vehicle operation. Also discussed above, the simulated application of
technology results in ``initial'' and ``final'' estimates of the cost,
fuel type, fuel economy, and fuel share (for, in particular, PHEVs that
can run on gasoline or electricity) applicable to each vehicle model in
each model year. Together with quantities of travel, and with estimates
of the ``gap'' between ``laboratory'' and ``on-road'' fuel economy,
these enable calculation of quantities of fuel consumed in each year
during the useful life of each vehicle model produced in each model
year.\343\ The model documentation provides specific procedures and
formulas implementing these calculations.
---------------------------------------------------------------------------
 \343\ The agencies have applied the same estimates of the ``on
road gap'' as applied for the analysis supporting the NPRM. For
operation on gasoline, diesel, E85, and CNG, this gap is 20 percent;
for electricity and hydrogen, 30 percent.
---------------------------------------------------------------------------
 As for the NPRM, the model calculates emissions of CO2
and other air pollutants, reporting emissions both from vehicle
tailpipes and from upstream processes (e.g., petroleum refining)
involved in producing and supplying fuels. Section VI.D.3 below reviews
methods, models, and estimates used in performing these calculations.
The model also calculates impacts on highway safety, accounting for
changes in travel demand, changes in vehicle mass, and continued past
and expected progress in vehicle safety (through, e.g., the application
of new crash avoidance systems). Section VI.D.2 discusses methods, data
sources, and estimates involved in estimating safety impacts, comments
on the same, and changes included in today's analysis. In response to
the NPRM, some comments urged the agencies also to quantify different
types of health impacts from changes in air pollution rather than only
accounting for such impacts in aggregate estimates of the social costs
of air pollution. Considering these comments, the agencies added such
calculations to the model, as discussed in Section VI.D.3.
6. Calculation of Benefits and Costs
 Having estimated how technologies might be applied going forward,
and having estimated the range of resultant physical impacts, the CAFE
Model calculates a variety of private and social benefits and costs,
reporting these from the consumer, manufacturer, and social
perspectives, both in undiscounted and discounted present value form
(given inputs specifying the corresponding discount rate and present
year). Estimates of regulatory costs are among the direct outputs of
the simulation of manufacturers' potential responses to new standards.
Other benefits and costs are calculated based on the above-mentioned
estimates of travel demand, fuel consumption, emissions, and safety
impacts. The agencies received many comments on the NPRM's calculation
of benefits and costs, and Section VI.D.1 discusses these comments and
presents the methods, data sources, and estimates used in calculating
benefits and costs reported here.
7. Structure of Model Inputs and Outputs
 All CAFE Model inputs and outputs described above are specified in
Microsoft Excel format, and the user can define and edit all inputs to
the system. Table VI-3 describes (non-exhaustively) which inputs are
contained within each input file and Table VI-4 describes which outputs
are contained in each output file. This is important for three reasons:
(1) Each file is discussed throughout the following sections; (2)
several commenters conflated aspects of the model with its inputs; and
(3) several commenters seemed confused about where to find specific
information in the output files. This information was described in
detail in the NPRM CAFE Model Documentation, but is reproduced here for
quick reference. When specifically referencing the input
[[Page 24282]]
or output file used for the NPRM or final rule in the following
discussion, NPRM or FRM, respectively, will precede the file name.
[GRAPHIC] [TIFF OMITTED] TR30AP20.080
[GRAPHIC] [TIFF OMITTED] TR30AP20.081
 A catalog of the Argonne National Laboratory Autonomie fuel economy
technology effectiveness value output files are reproduced in the
following Table VI-5 as well. The left column shows the terminology
used in this text to refer to the file, while the right column
describes each file. NPRM or FRM, respectively, may precede the
terminology in the text as appropriate.
[[Page 24283]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.082
 Finally, Table VI-6 lists the terminologies used to refer to other
model-related documents which are referred to frequently throughout the
text. NPRM or FRM, respectively, may precede the terminology in the
text as appropriate.
[[Page 24284]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.083
B. What inputs does the compliance analysis require?
1. Analysis Fleet
 The starting point for the evaluation of the potential feasibility
of different stringency levels for future CAFE and CO2
standards is the analysis fleet, which is a snapshot of the recent
vehicle market. The analysis fleet provides a baseline from which to
project what and how additional technologies could feasibly be applied
to vehicles in a cost-effective manner to raise those vehicles' fuel
economy and lower their CO2 emission levels.\344\ The fleet
characterization also provides a reference point with data for other
factors considered in the analysis, including environmental effects and
effects estimated by the economic modules (i.e., sales, scrappage, and
labor utilization). When the scope of the analysis widens, another
piece of data must be included for each vehicle in the analysis fleet
to map a given element of the fleet appropriately onto an analysis
module.
---------------------------------------------------------------------------
 \344\ The CAFE model does not generate compliance paths a
manufacturer should, must, or will deploy. It is intended as a tool
to demonstrate a compliance pathway a manufacturer could choose. It
is almost certain all manufacturers will make compliance choices
differing from those projected by the CAFE model.
---------------------------------------------------------------------------
 For the analysis presented in this final rule, the analysis fleet
includes information about vehicles that is essential for each analysis
module. The first part of projecting how additional technologies could
be applied to vehicles is knowing which vehicles are produced by which
manufacturers, the fuel economies of those vehicles, how many of each
are sold, whether they are passenger cars or light trucks, and their
footprints. This is important because it improves understanding of the
overall impacts of different levels of CAFE and CO2
standards; overall impacts that result from industry's response to
standards, and industry's response, is made up of individual
manufacturer responses to the standards in light of the overall market
and their individual assessment of consumer acceptance. Establishing an
accurate representation of manufacturers' existing fleets (and the
vehicle models in them) that will be subject to future standards helps
in predicting potential individual manufacturer responses to those
future standards in addition to potential changes in those standards.
 Another part of projecting how additional fuel economy improving
technologies could be applied to vehicles is knowing which fuel saving
technologies manufacturers have equipped on which vehicles. In many
cases, the agencies also collect and reference additional information
on other vehicle attributes to help with this process.\345\ Accounting
for technologies already applied to vehicles helps avoid ``double-
counting'' the value of those technologies, by assuming they are still
available to be applied to improve fuel economy and reduce
CO2 emissions. It also promotes more realistic
determinations of what additional technologies can feasibly be applied
to those vehicles: If a manufacturer has already started down a
technological path to fuel economy or performance improvements, the
agencies do not assume it will completely abandon that path because
doing so would be unrealistic and fails to represent accurately
manufacturer responses to standards. Each vehicle model (and
configurations of each model) in the analysis fleet, therefore, has a
comprehensive list of its technologies, which is important because
different configurations may have different technologies applied to
them.\346\ In addition, to properly account for technology costs, the
agencies assign each vehicle to a technology class and an engine class.
Technology classes reference each vehicle to a set of full vehicle
simulations, so that the agencies may project fuel efficiency with
combinations of additional fuel saving equipment and hybrid and
electric vehicle battery costs.
---------------------------------------------------------------------------
 \345\ For instance, curb weight, horsepower, drive
configuration, pickup bed length, oil type, body style, aerodynamic
drag coefficients, and rolling resistance coefficients, and (if
applicable) battery sizes are all required to assign technology
content properly.
 \346\ Considering each vehicle model/configuration also improves
the ability to consider the differential impacts of different levels
of potential standards on different manufacturers, since all vehicle
model/configurations ``start'' at different places, in terms of
technologies already used and how those technologies are used.
---------------------------------------------------------------------------
 Yet another part of projecting which vehicles might exist in future
model years is developing reasonable real-world assumptions about when
and how manufacturers might apply certain technologies to vehicles. The
analysis fleet accounts for links between vehicles, recognizing vehicle
platforms will share technologies, and the vehicles that make up that
platform should receive (or not receive) additional technological
improvements together. Shared engines, shared transmissions, and shared
vehicle platforms for mass reduction technology are considered. In
addition, each vehicle model/configuration in the analysis fleet also
has information about its redesign
[[Page 24285]]
schedule, i.e., the last year it was redesigned and when the agencies
expect it to be redesigned again. Redesign schedules are a key part of
manufacturers' business plans, as each new product can cost more than
$1B, and involve a significant portion of a manufacturer's scarce
research, development, and manufacturing and equipment budgets and
resources.\347\ Manufacturers have repeatedly told the agencies that
sustainable business plans require careful management of resources and
capital spending, and that the length of time each product remains in
production is crucial to recouping the upfront product development and
plant/equipment costs, as well as the capital needed to fund the
development and manufacturing equipment needed for future products.
Because the production volume of any given vehicle model varies within
a manufacturer's product line, and varies among different
manufacturers, redesign schedules typically vary for each model and
manufacturer. Some (relatively few) technological improvements are
small enough that they can be applied in any model year; a few other
technological improvements may be applied during a refreshening (when a
few additional changes are made, but well short of a full redesign),
but others are major enough that they can only be cost-effectively
applied at a vehicle redesign, when many other things about the vehicle
are already changing. Ensuring the CAFE model makes technological
improvements to vehicles only when it is feasible to do so also helps
the analysis better represent manufacturer responses to different
levels of standards.
---------------------------------------------------------------------------
 \347\ Shea, T., Why Does It Cost So Much For Automakers To
Develop New Models? Autoblog (Jul. 27, 2010), https://www.autoblog.com/2010/07/27/why-does-it-cost-so-much-for-automakers-to-develop-new-models/.
---------------------------------------------------------------------------
 Finally, the agencies restrict the applications of some
technologies on some vehicles upon determining the technology is not
compatible with the functional and performance requirements of the
vehicle, or if the manufacturers are unlikely to apply a specific
technology to a specific vehicle for reasons articulated with
confidential business information that the agencies found credible.
 Other data important for the analysis that are referenced to the
analysis fleet include baseline economic, environmental, and safety
information. Vehicle fuel tank size is required to estimate range and
refueling benefit while curb weights and safety class assignments help
the agencies consider how changes in vehicle mass may affect safety.
The agencies identify the final assembly location for each vehicle,
engine, and transmission, as well as the percent of U.S. content to
support the labor impact analysis. In addition, the aforementioned
accounting for first-year vehicle production volumes (i.e., the number
of vehicles of each new model sold in MY 2017, for this analysis) is
the foundation for estimating how future vehicle sales might change in
response to different potential standards.
 The input file for the CAFE model characterizing the analysis
fleet, referred to as the ``market inputs'' file or ``market data''
file, accordingly includes a large amount of data about vehicles, their
technological characteristics, the manufacturers and fleets to which
they belong, and initial prices and production volumes, which provide
the starting points for projection (by the sales model) to ensuing
model years. In the Draft TAR (which utilized a MY 2015 analysis fleet)
and NPRM (which utilized a MY 2016 analysis fleet), the agencies needed
to populate about 230,000 cells in the market data file to characterize
the fleet. For this final rule (which utilized a MY 2017 analysis
fleet), the agencies populated more than 400,000 cells to characterize
the fleet. While the fleet is not actually much more heterogeneous in
reality,\348\ the agencies have provided and collected more data to
justify the characterization of the analysis fleet, and to support the
functionality of modules in the CAFE model.
---------------------------------------------------------------------------
 \348\ The expansion of cells is primarily due to (1) considering
more technologies, and (2) listing trim levels separately, which
often yields more precise curb weights and more accurate
manufacturer suggested retail prices.
---------------------------------------------------------------------------
 A solid characterization of a recent model year as an analytical
starting point helps realistically estimate ways manufacturers could
potentially respond to different levels of standards, and the modeling
strives to simulate realistically how manufacturers could progress from
that starting point. While manufacturers can respond in many ways
beyond those represented in the analysis (e.g., applying other
technologies, shifting production volumes, changing vehicle footprint),
such that it is impossible to predict with any certainty exactly how
each manufacturer will respond, it is still important to establish a
solid foundation from which to estimate potential costs and benefits of
potential future standards. The following sections discuss aspects of
how the analysis fleet was built for this analysis, and includes
discussion of the comments on fleet that the agencies received on the
proposed rule.
a) Principles on Data Sources Used To Populate the Analysis Fleet
 The source data for vehicles in the analysis fleet and their
technologies is a central input for the analysis. The sections below
discuss pros and cons of different potential sources and what the
agencies used for this analysis, and responds to comments the agencies
received on data sources in the proposal.
(1) Use of Confidential Business Information Versus Publicly-Releasable
Sources
 Since 2001, CAFE analysis has used either confidential, forward-
estimating product plans from manufacturers, or publicly available data
on vehicles already sold as a starting point for determining what
technologies can be applied to what vehicles in response to potential
different levels of standards. The use of either data source requires
certain tradeoffs. Confidential product plans comprehensively represent
what vehicles a manufacturer expects to produce in coming years,
accounting for plans to introduce new vehicles and fuel-saving
technologies and, for example, plans to discontinue other vehicles and
even brands. This information can be very thorough and can improve the
accuracy of the analysis, but cannot be publicly released. This makes
it difficult for public commenters to reproduce the analysis for
themselves as they develop their comments. Some non-industry commenters
have also expressed concern about manufacturers having an incentive in
the submitted plans to underestimate (deliberately or not) their future
fuel economy capabilities and overstate their expectations about, for
example, the levels of performance of future vehicle models in order to
affect the analysis. Accordingly, since 2010, EPA and NHTSA have based
analysis fleets almost exclusively on information from commercial and
public sources, starting with CAFE compliance data and adding
information from other sources.
 An analysis fleet based primarily on public sources can be released
to the public, solving the issue of commenters being unable to
reproduce the overall analysis. However, industry commenters have
argued such an analysis fleet cannot accurately reflect manufacturers'
actual plans to apply fuel-saving technologies (e.g., manufacturers may
apply turbocharging to improve not just fuel economy, but also to
improve vehicle performance) or manufacturers' plans to change product
offerings by introducing some vehicles and brands and discontinuing
other
[[Page 24286]]
vehicles and brands, precisely because that information is typically
confidential business information (CBI). A fully-publicly-releasable
analysis fleet holds vehicle characteristics unchanged over time and
lacks some level of accuracy when projected into the future. For
example, over time, manufacturers introduce new products and even
entire brands. On the other hand, plans announced in press releases do
not always ultimately bear out, nor do commercially available third-
party forecasts. Assumptions could be made about these issues to
improve the accuracy of a publicly releasable analysis fleet, but
concerns include that this information would either be largely
incorrect, or, if the assumptions were correct, information would be
released that manufacturers would consider CBI.
 Furthermore, some technologies considered in the rulemaking are
difficult to observe in the analysis fleet without expensive teardown
study and time-consuming benchmarking. Not giving credit for these
technologies puts the analysis at significant risk of double-counting
the effectiveness of these technologies, as manufacturers cannot equip
technologies twice to the same vehicle for double the fuel economy
benefit. As discussed in the Draft TAR, the agencies assigned little
(if any) technology application in the baseline fleet for some of these
technologies.\349\ For the NPRM MY 2016 fleet development process, the
agencies again offered the manufacturers the opportunity to volunteer
CBI to the agencies to help inform the technology content of the
analysis fleet, and many manufacturers did. The agencies were able to
confirm that many manufacturers had already included many hard-to-
observe technologies in the MY 2016 fleet (which they were not properly
given credit for in the characterization of the MY 2014 and MY 2015
fleets presented in Draft TAR) so the agencies reflected this new
information in the NPRM analysis and in the analysis presented today.
---------------------------------------------------------------------------
 \349\ These technologies include low rolling resistance
technology (incorrectly applied to zero baseline vehicles in Draft
TAR), low-drag brakes (incorrectly applied to zero baseline vehicles
in Draft TAR), electric power steering (incorrectly applied to too
few vehicles in Draft TAR), accessory drive improvements
(incorrectly applied to zero baseline vehicles in Draft TAR), engine
friction reduction (previously named LUBEFR1, LUBEFR2, and LUBEFR3),
secondary axle disconnect and transmission improvements.
---------------------------------------------------------------------------
 In addition, many manufacturers provided confidential comment on
the potential applicability of fuel-saving technologies to their fleet.
In particular, many manufacturers confidentially identified specific
engine technologies that they will not use in the near term, either on
specific vehicles, or at all. Reasons varied: Some manufacturers cited
intellectual property concerns, and others stated functional
performance concerns for some engine types on some vehicles. Other
manufacturers shared forward-looking product plans, and explained that
it would be cost prohibitive to scrap significant investments in one
technology in favor of another. This topic is discussed in more detail
in Section VI.B.1.b)(6), below.
 The agencies sought comment on how to address this issue going
forward, recognizing both the competing interests involved and the
typical timeframes for CAFE and CO2 standards rulemakings.
 Many commenters expressed concern with the agencies using any CBI
as part of the rulemaking process. Some commenters expressed concern
that use of CBI would make the CAFE model subject to inaccuracies
because manufacturers would only provide additional information in
situations in which a correction to the agencies' baseline assumptions
would favor the manufacturers.\350\ The agencies recognize this as a
reasonable concern, but the analysis presented in the Draft TAR
consistently assumed very little (if any) technology had been applied
in the baseline. In addition, many manufacturers shared information on
advanced technologies that were not yet in production in MY 2017, but
could be used in the future; manufacturer contributions helped the
agencies better model many advanced engine technologies and to include
them in today's analysis, and inclusion of these technologies (and
costs) in the analysis sometimes lowered the projected cost of
compliance for stringent alternatives. Other commenters expressed
concern that automakers would supply false or incomplete information
that would unduly restrict what technologies can be deployed.\351\ When
possible, the agencies sought independently to verify manufacturer CBI
(or claims made by other stakeholders) through lab testing and
benchmarking.\352\ The agencies found no evidence of misrepresentation
of engineering specifications in the MY 2017 fleet in manufacturer CBI;
instead, the agencies were able to verify independently many CBI
submissions, and confirm the credibility of information provided from
those sources.
---------------------------------------------------------------------------
 \350\ NHTSA-2018-0067-12039, Union of Concerned Scientists.
 \351\ NHTSA-2018-0067-11741, ICCT.
 \352\ For instance, the agencies continue to evaluate tire
rolling resistance on production vehicles via independent lab
testing, and the agencies bench-marked the operating behavior and
calibration of many engines and transmissions.
---------------------------------------------------------------------------
 Some commenters requested that more CBI be used in the analysis.
For instance, some commenters suggested that the agencies should return
to the use of product plans and announcements regarding future fleets
because manufacturers had already committed investments to bring
announced products to market.\353\ However, if the agencies were to
assume that these commitments were already in the baseline, the
agencies would underestimate the cost of compliance for stringent
alternatives. Moreover, while upfront investments to bring technologies
to market are significant, the total marginal costs of components are
typically large in comparison over the entire product life-cycle, and
these costs have not yet been realized in vehicles not yet produced.
---------------------------------------------------------------------------
 \353\ NHTSA-2018-0067-11956, PA Department of Environmental
Protection.
---------------------------------------------------------------------------
 The agencies did make use of some forward-looking CBI in the
analysis. The agencies received many comments from manufacturers on the
technological feasibility, or functional applicability of some fuel
saving technologies to certain vehicles, or certain vehicle
applications, and the agencies took this information into consideration
when projecting compliance pathways. These cases are discussed
generally in Section VI.B.1.b)(6), below, and specifically for each
technology in those technology sections. Some commenters expressed that
the use of CBI for future product plans would be acceptable, but only
if the agencies disclosed the CBI affecting all vehicles through MY
2025 at the time of publication.\354\ Functionally, this is not
possible. Manufacturer's confidential product plans cannot be made
public, as prohibited under NHTSA's regulations at 49 CFR part 512, and
if the information meets the requirements of section 208(c) of the
Clean Air Act. If the agencies disclosed confidential information, it
would not only violate the terms on which the agencies obtained the
CBI, but it is unlikely that manufacturers would continue to offer CBI,
which in turn would likely degrade the quality of the analysis. The
agencies believe that the use of CBI in the NPRM and final rule
analysis--to confirm, reference, or to otherwise modify aspects of the
analysis that can be made public--threads the needle between a more
accurate but less transparent analysis (using more CBI) and a less
accurate but more transparent analysis (using less CBI).
---------------------------------------------------------------------------
 \354\ NHTSA-2018-0067-11741.
---------------------------------------------------------------------------
[[Page 24287]]
(2) Source Data and Vintage Used in the Analysis
 Based on the assumption that a publicly-available analysis fleet
continued to be desirable, manufacturer compliance submissions to EPA
and NHTSA were used as a starting point for the NPRM and final rule
analysis fleets. Generally, manufacturer compliance submissions break
down vehicle fuel economy and production volume by regulatory class,
and include some very basic product information (typically including
vehicle nameplate, engine displacement, basic transmission information,
and drive configuration). Many different trim levels of a product are
typically rolled up and reported in an aggregated fashion, and these
groupings can make decomposition of different fuel-saving, road load
reducing technologies extremely difficult. For instance, vehicles in
different test weight classes, with different tires or aerodynamic
profiles may be aggregated and reported together.\355\ A second portion
of the compliance submission summarizes production volume by vehicle
footprints (a key compliance measure for standard setting) by
nameplate, and includes some basic information about engine
displacement, transmission, and drive configuration. Often these
production volumes by footprint do not fit seamlessly together with the
production volumes for fuel economy, so the agencies must reconcile
this information.
---------------------------------------------------------------------------
 \355\ Some fuel-economy compliance information for pickup trucks
span multiple cab and box configurations, but manufacturers reported
these disparate vehicles together.
---------------------------------------------------------------------------
 Information from the MY 2016 fleet was chosen as the foundation for
the NPRM analysis fleet because, at the time the rulemaking analysis
was initiated, the 2016 fleet represented the most up-to-date
information available in terms of individual vehicle models and
configurations, production technology levels, and production volumes.
If MY 2017 data had been used while this analysis was being developed,
the agencies would have needed to use product planning information that
could not be made available to the public until a later date.
 The NPRM analysis fleet was initially developed with 2016 mid-model
year compliance data because final compliance data was not available at
that time, and the timing provided manufacturers the opportunity to
review and comment on the characterization of their vehicles in the
fleet. With a view toward developing an accurate characterization of
the 2016 fleet to serve as an analytical starting point, corrections
and updates to mid-year data (e.g., to production estimates) were
sought, in addition to corroboration or correction of technical
information obtained from commercial and other sources (to the extent
that information was not included in compliance data), although future
product planning information from manufacturers (e.g., future product
offerings, products to be discontinued) was not requested, as most
manufacturers view such information as CBI. Manufacturers offered a
range of corrections to indicate engineering characteristics (e.g.,
footprint, curb weight, transmission type) of specific vehicle model/
configurations, as well as updates to fuel economy and production
volume estimates in mid-year reporting. After following up on a case-
by-case basis to investigate significant differences, the analysis
fleet was updated.
 Sales, footprint, and fuel economy values with final compliance
data were also updated if that data was available. In a few cases,
final production and fuel economy values were slightly different for
specific MY 2016 vehicle models and configurations than were indicated
in the NPRM analysis; however, other vehicle characteristics (e.g.,
footprint, curb weight, technology content) important to the analysis
were reasonably accurate. While some commenters have, in the past,
raised concerns that non-final CAFE compliance data is subject to
change, the potential for change is likely not significant enough to
merit using final data from an earlier model year reflecting a more
outdated fleet. Moreover, even ostensibly final CAFE compliance data is
frequently subject to later revision (e.g., if errors in fuel economy
tests are discovered), and the purpose of the analysis was not to
support enforcement actions but rather to provide a realistic
assessment of manufacturers' potential responses to future standards.
 Manufacturers integrated a significant amount of new technology in
the MY 2016 fleet, and this was especially true for newly-designed
vehicles launched in MY 2016. While subsequent fleets will involve even
further application of technology, using available data for MY 2016
provided the most realistic detailed foundation for analysis that could
be made available publicly in full detail, allowing stakeholders to
reproduce the analysis presented in the proposal independently. Insofar
as future product offerings are likely to be more similar to vehicles
produced in 2016 than to vehicles produced in earlier model years,
using available data regarding the 2016 model year provided the most
realistic, publicly releasable foundation for constructing a forecast
of the future vehicle market for this proposal. Many comments
responding to the Draft TAR, EPA's Proposed Determination, EPA's 2017
Request for Comment, and the NPRM preceding today's notice stated that
the most up-to-date analysis fleet possible should be used, because a
more up-to-date analysis fleet will better capture how manufacturers
apply technology and will account better for vehicle model/
configuration introductions and deletions.356 357
---------------------------------------------------------------------------
 \356\ 82 FR 39551 (Aug. 21, 2017).
 \357\ For example, in 2016 comments to dockets EPA-HQ-OAR-2015-
0827 and NHTSA-2016-0068, the Alliance of Automobile Manufacturers
commented that ``the Alliance supports the use of the most recent
data available in establishing the baseline fleet, and therefore
believes that NHTSA's selection [of, at the time, model year 2015]
was more appropriate for the Draft TAR.'' Alliance at 82, Docket ID.
EPA-HQ-OAR-2015-0827-4089. Global Automakers commented that ``a one-
year difference constitutes a technology change-over for up to 20%
of a manufacturer's fleet. It was also generally understood by
industry and the agencies that several new, and potentially
significant, technologies would be implemented in MY 2015. The use
of an older, outdated baseline can have significant impacts on the
modeling of subsequent Reference Case and Control Case
technologies.'' Global Automakers at A-10, Docket ID EPA-HQ-OAR-
2015-0827-4009.
---------------------------------------------------------------------------
 On the other hand, some commenters suggested that because
manufacturers continue improving vehicle performance and utility over
time, an older analysis fleet should be used to estimate how the fleet
could have evolved had manufacturers applied all technological
potential to fuel economy rather than continuing to improve vehicle
performance and utility.\358\ Because manufacturers change and improve
product offerings over time, conducting analysis with an older analysis
fleet (or with a fleet using fuel economy levels and CO2
emissions rates that have been adjusted to reflect an assumed return to
levels of performance and utility typical of some past model year)
would miss this real-world trend. While such an analysis could project
what industry could do if, for example, manufacturers devoted all
technological improvements toward raising fuel economy and reducing
CO2 emissions (and if consumers decided to purchase these
vehicles), the agencies do not believe it would be consistent with a
transparent examination of what effects different levels of standards
would have
[[Page 24288]]
on individual manufacturers and the fleet as a whole.
---------------------------------------------------------------------------
 \358\ For example, in 2016 comments to dockets EPA-HQ-OAR-2015-
0827 and NHTSA-2016-0068, UCS stated ``in modeling technology
effectiveness and use, the agencies should use 2010 levels of
performance as the baseline.'' UCS at 4, Docket ID. EPA-HQ-OAR-2015-
0827-4016.
---------------------------------------------------------------------------
 All else being equal, using a newer analysis fleet will produce
more realistic estimates of impacts of potential new standards than
using an outdated analysis fleet. However, among relatively current
options, a balance must be struck between input freshness, and input
completeness and accuracy.\359\ During assembly of the inputs for the
NPRM analysis, final compliance data was available for the MY 2015
model year but not, in a few cases, for MY 2016. However, between mid-
year compliance information and manufacturers' specific updates
discussed above, a robust and detailed characterization of the MY 2016
fleet was developed. While information continued to develop regarding
the MY 2017 and, to a lesser extent MY 2018 and even MY 2019 fleets,
this information was--even in mid-2017--too incomplete and inconsistent
to be assembled with confidence into an analysis fleet for modeling
supporting deliberations regarding the NPRM analysis.
---------------------------------------------------------------------------
 \359\ Comments provided through a recent peer review of the CAFE
model recognize the competing interests behind this balance. For
example, referring to NHTSA's 2016 Draft TAR analysis, one of the
peer reviewers commented as follows: ``The NHTSA decision to use MY
2015 data is wise. In the TAR they point out that a MY 2016
foundation would require the use of confidential data, which is less
desirable. Clearly they would also have a qualitative vision of the
MY 2016 landscape while employing MY 2015 as a foundation. Although
MY 2015 data may still be subject to minor revision, this is
unlikely to impact the predictive ability of the model . . . A more
complex alternative approach might be to employ some 2016 changes in
technology, and attempt a blend of MY 2015 and MY 2016, while
relying of estimation gained from only MY 2015 for sales. This
approach may add some relevancy in terms of technology, but might
introduce substantial error in terms of sales.''
---------------------------------------------------------------------------
 Manufacturers requested that the baseline fleet supporting the
final rule incorporate the MY 2018 or most recent information
available.\360\ Other commenters expressed desire for multiple fleets
of various vintages to compare the updated model outputs with those of
previous rule-makings. Specifically, some commenters requested that
older fleet vintages (MY 2010, for instance) be developed in parallel
with the MY 2017 fleet so that those too may be used as inputs for the
model.\361\
---------------------------------------------------------------------------
 \360\ NHTSA-2018-0067-12150, Toyota North America.
 \361\ NHTSA-2018-0067-11741, ICCT.
---------------------------------------------------------------------------
 Between the NPRM and this final rule, manufacturers submitted final
compliance data for the MY 2017 fleet. When the agencies pulled
together information for the fleet for the final rule, the agencies
decided to use the highest-quality, most up-to-date information
available. Given that pulling this information together takes some
time, and given that ``final'' compliance submissions often lag
production by a few years, the agencies decided to use 2017 model year
as the base year for the analysis fleet, as the agencies stated in the
NPRM.\362\ While the agencies could have used preliminary 2018 data or
even very early 2019 data, this information was not available in time
to support the final rulemaking. Likewise, the agencies chose not to
revert to a previous model year (for instance 2016 or 2012) because
many manufacturers have incorporated fuel savings technologies over the
last few years, realized some benefits for fuel economy, and adjusted
the performance or sales mix of vehicles to remain competitive in the
market. Also, using an earlier model year would provide less accurate
projections because the analysis would be based on what manufacturers
could have done in past model years and would have estimated the fuel
economy improvements instead of using known information on the
technologies that were employed and the actual fuel economy that
resulted from applying those technologies.
---------------------------------------------------------------------------
 \362\ 83 FR 43006 (``If newer compliance data (i.e., MY 2017)
becomes available and can be analyzed during the pendency of this
rulemaking, and if all other necessary steps can be performed, the
analysis fleet will be updated, as feasible, and made publicly
available.'').
---------------------------------------------------------------------------
 Some additional information (about off-cycle technologies, for
instance) was often not reported by manufacturers in MY 2017 formal
compliance submissions in a way that provided clear information on
which technologies were included on which products. As part of the
formal compliance submission, some manufacturers voluntarily submitted
additional information (about engine technologies, for instance). While
this data was generally of very high quality, there were some mistakes
or inconsistencies with publicly available information, causing the
agencies to contact the manufacturers to understand and correct
identified issues. In most cases, however, the formal compliance data
was very limited in nature, and the agencies collected additional
information necessary to characterize fully the fleet from other
sources, and scrutinized additional information submitted by
manufacturers carefully, independently verifying when possible.
 Specifically, the agencies downloaded and reviewed numerous
marketing brochures and product launch press releases to confirm
information submitted by manufacturers and to fill in information
necessary for the analysis fleet that was not provided in the
compliance data. Product brochures often served as the basis for the
curb weights used in the analysis. This publicly available manufacturer
information sometimes also included aerodynamic drag coefficients,
information about steering architecture, start-stop systems, pickup bed
lengths, fuel tank capacities, and high-voltage battery capacities. The
agencies recorded vehicle horsepower, compression ratio, fuel-type, and
recommended oil weight rating from a combination of manufacturer
product brochures and owner's manuals. The product brochures, as well
as online references such as Autobytel, informed which combinations of
fuel saving technologies were available on which trim levels, and what
the manufacturer suggested retail price was for many products. Overall
this information proved helpful for assigning technologies to vehicles,
and for getting data (such as fuel tank size \363\) necessary for the
analysis. These reference materials have been included in the
rulemaking documentation.\364\
---------------------------------------------------------------------------
 \363\ The quality of data for today's analysis fleet is notably
improved for fuel tank capacity, which factors into the calculation
of refueling time benefits. In many previous analyses, fuel tank
sizes were often stated as estimates or proxies, and not sourced so
carefully.
 \364\ Publicly available data used to supplement analysis fleet
information is available in the docket.
---------------------------------------------------------------------------
 The agencies elected not to develop fleets of previous model year
vintages that could be used in parallel as an input to the CAFE model.
Developing a detailed characterization of the fleet of any vintage
would be a huge undertaking with few benefits. As the scope has
increased, and as additional modules are added, going back in time to
re-characterize a previous fleet in a format that works with CAFE model
updates can be time- and resource-prohibitive for the agencies, even if
that work is adapting a fleet that was used in previous rule-making
analysis. Doing so also offers little value in determining what
potential fuel saving technology can be added to a more recent fleet
during the rulemaking timeframe.
 The MY 2017 manufacturer-submitted data, verified and supplemented
by the agencies with publicly-available information, therefore
presented the fullest, most up-to-date data set that the agencies could
have used to support this analysis.
[[Page 24289]]
b) Characterizing Vehicles and Their Technology Content
 The starting point for projecting what additional fuel economy
improving technologies could feasibly be applied to vehicles is knowing
what vehicles are produced by which manufacturers and what technologies
exist on those vehicles. Rows in the market data file are the smallest
portion of the fleet to which technology may be applied as part of a
projected compliance pathway. For the analysis presented in this final
rule, the agencies, when possible, attempted to include vehicle trim
level information in discrete rows. A manufacturer, for example GM, may
produce one or more vehicle makes (or brands), for example Chevrolet,
Buick and others. Each vehicle make may offer one or more vehicle
models, for example Malibu, Traverse and others. And each vehicle model
may be available in one or more trim levels (or standard option
levels), for example ``RS,'' ``Premier'' and others, which have
different levels of standard options, and in some cases, different
engines and transmissions.
 Manufacturer compliance submissions, discussed above, were used as
a starting point to define working rows in the market data file;
however, often the rows needed to be further disaggregated to correctly
characterize vehicle information covered in the scope of the analysis,
and analysis fleet. Manufacturers often grouped vehicles with multiple
trim levels together because they often included the same fuel-saving
technologies and may be aggregated to simplify reporting. However, the
manufacturer suggested retail prices of different trim levels are
certainly different, and other features relevant to the analysis are
occasionally different.
 As a result of further disaggregating compliance information, the
number of rows in the market data file increased from 1,667 rows used
in the NPRM to 2,952 rows for this final rule analysis. The agencies do
not have data on sales volumes for each nameplate by trim level, and
used an approach that evenly distributed volume across offered trim
levels, within the defined constraints of the compliance data.\365\
Evenly distributing the volume across trim levels is a simplification,
but this action should (1) highlight some difficulties that could be
encountered when acquiring data for a full-vehicle consumer choice
model should the agencies pursue developing one in the future
(discussed further, below), and (2) lower the average sales volume per
row in the market data file, thereby allowing the application of very
advanced electrification technologies in smaller lumps. The latter
effect is responsive to comments (discussed below) that suggested
electrification technologies could be more cost-effectively deployed in
lower volumes, and that the CAFE model artificially constrains cost
effective technologies that may be deployed, resulting in higher costs
and large over-compliance.
---------------------------------------------------------------------------
 \365\ The sum of volumes by nameplate configuration, for fuel
economy value, and for footprint value remains the same.
---------------------------------------------------------------------------
(1) Assigning Vehicle Technology Classes
 While each vehicle in the analysis fleet has its list of observed
technologies and equipment, the ways in which manufacturers apply
technologies and equipment do not always coincide perfectly with how
the analysis characterizes the various technologies that improve fuel
economy and reduce CO2 emissions. To improve how the
observed vehicle fleet ``fits into'' the analysis, each vehicle model/
configuration is ``mapped'' to the full-vehicle simulation modeling by
Argonne National Laboratory that is used to estimate the effectiveness
of the fuel economy-improving/CO2 emissions-reducing
technologies considered. Argonne produces full-vehicle simulation
modeling for many combinations of technologies, on many types of
vehicles, but it did not simulate literally every single manufacturer's
vehicle model/configuration in the analysis fleet because it would be
impractical to assemble the requisite detailed information--much of
which would likely only be provided on a confidential basis--specific
to each vehicle model/configuration and because the scale of the
simulation effort would correspondingly increase by at least two orders
of magnitude. Instead, Argonne simulated 10 different vehicle types
corresponding to the ``technology classes'' generally used in CAFE
analysis over the past several rulemakings (e.g., small car, small
performance car, pickup truck, etc.). Each of those 10 different
vehicle types was assigned a set of ``baseline characteristics'' to
which Argonne added combinations of fuel-saving technologies and then
ran simulations to determine the fuel economy achieved when applying
each combination of technologies to that vehicle type given its
baseline characteristics.
BILLING CODE 4910-59-P
[[Page 24290]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.089
BILLING CODE 4910-59-C
 In the analysis fleet, inputs assign each specific vehicle model/
configuration to a technology class, and once there, map to the
simulation within that technology class most closely matching the
combination of observed technologies and equipment on that vehicle.
This mapping to a specific simulation result most closely representing
a given vehicle model/configuration's initial technology ``state''
enables the CAFE model to estimate the same vehicle model/
configuration's fuel economy after application of some other
combination of technologies, leading to an alternative technology
state.
(2) Assigning Vehicle Technology Content
 As explained above, the analysis fleet is defined not only by the
vehicles it contains, but also by the technologies on those vehicles.
Each vehicle in the analysis fleet has an associated list of observed
technologies and equipment that can improve fuel economy and reduce
CO2 emissions.\366\ With a portfolio of descriptive
technologies arranged by manufacturer and model, the analysis fleet can
be summarized and project how vehicles in that fleet may increase fuel
economy over time via the application of additional technology.
---------------------------------------------------------------------------
 \366\ These technologies are generally grouped into the
following categories: Vehicle technologies include mass reduction,
aerodynamic drag reduction, low rolling resistance tires, and
others. Engine technologies include engine attributes describing
fuel type, engine aspiration, valvetrain configuration, compression
ratio, number of cylinders, size of displacement, and others.
Transmission technologies include different transmission
arrangements like manual, 6-speed automatic, 10-speed automatic,
continuously variable transmission, and dual-clutch transmissions.
Hybrid and electric powertrains may complement traditional engine
and transmission designs or replace them entirely.
---------------------------------------------------------------------------
 In many cases, vehicle technology is clearly observable from the
2017 compliance data (e.g., compliance data indicates clearly which
vehicles have turbochargers and which have continuously variable
transmissions), but in some cases technology levels are less
observable. For the latter, like levels of mass reduction, the analysis
categorized levels of technology already used in a given vehicle.
Similarly, engineering judgment was used to determine if higher mass
reduction levels may be used practicably and safely for a given
vehicle.
 Either in mid-year compliance data for MY 2016, final compliance
data for MY 2017, or separately and at the agencies' invitation prior
to the NPRM or in comments in responses to the NPRM, most manufacturers
provided guidance on the technology already present in each of their
vehicle model/configurations. This information was not as complete for
all manufacturers' products as needed for the analysis, so, in some
cases, information was supplemented with publicly available data,
typically from manufacturer media sites. In limited cases,
manufacturers did not supply information, and
[[Page 24291]]
information from commercial and publicly available sources was used.
 The agencies continued to evaluate emerging technologies in the
analysis. In response to comments,\367\ and given recent product
launches for MY 2020, and some very recently announced future product
offerings, the agencies elevated some technologies that were discussed
in the NPRM to the compliance simulation. As a result, several
additional engine technologies, expanded levels of mass reduction
technology, and some additional combinations of engines with plug-in
hybrid, or strong hybrid technology are available in the compliance
pathways for the final rule analysis.
---------------------------------------------------------------------------
 \367\ NHTSA-2018-0067-11741.
---------------------------------------------------------------------------
 In addition, some redundant technologies, or technologies that were
inadvertently represented on the technology tree as being available to
be applied twice, have been consolidated. For instance, previous basic
versions of engine friction reduction were layered on top of basic
engine maps, but the efficiency in many modern engine maps already
include the benefits of that engine friction reduction technology. The
following Table VI-8 lists the technologies considered in the final
rule analysis, with the data sources used to map those technologies to
vehicles in the analysis fleet.
BILLING CODE 4910-59-P
[[Page 24292]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.090
[[Page 24293]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.091
[[Page 24294]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.092
BILLING CODE 4910-59-C
 Industry commenters generally stated the MY 2016 baseline
technology content presented in the NPRM as an improvement over
previous analyses because it more accurately accounted for technology
already used in the fleet.368 369 In contrast, some
commenters expressed preference for EPA's baseline technology
assignment assumptions presented in the Draft TAR for mass reduction,
tire rolling resistance, and aerodynamic drag because those assumptions
projected very few technology improvements were present in the baseline
fleet. In assessing the comments, the agencies found that
[[Page 24295]]
using the EPA Draft TAR approach would lead to projected compliance
pathways with overestimated fuel economy improvements and
underestimated costs.\370\
---------------------------------------------------------------------------
 \368\ NHTSA-2018-0067-12073, Alliance of Automobile
Manufacturers.
 \369\ NHTSA-2018-0067-12150, Toyota North America.
 \370\ NHTSA-2018-0067-11741, ICCT.
---------------------------------------------------------------------------
 Many of those assumptions were neither scientifically meritorious,
nor isolated examples. For instance, for the EPA Draft TAR and Proposed
Determination analyses, the BMW i3, a vehicle with full carbon fiber
bodysides and downsized, mass-reduced wheels and tires (some of the
most advanced mass reducing technologies commercialized in the
automotive industry), was assumed to have 1.0 percent mass reduction (a
very minor level of mass reduction). Similarly, previous analyses
assigned the Chevrolet Corvette, a performance vehicle that has long
been a platform for commercializing advanced weight saving
technologies,\371\ with zero mass reduction. For aerodynamic drag,
previous EPA analysis assumed that pickup trucks could achieve the
aerodynamic drag profile typical of a sedan, with little regard for
form drag constraints or frontal area (and headroom, or ground
clearance) considerations. These assumptions commonly led to
projections of a 20 percent improvement in mass, aerodynamic drag, and
tire rolling resistance, even when a large portion of those
improvements had either already been implemented, or were not
technologically feasible. On the other hand, in the Draft TAR, NHTSA
presented methodologies to evaluate content for mass reduction
technology, aerodynamic drag improvements, and rolling resistance
technologies that better accounted for the actual level of technologies
in the analysis fleet. Throughout the rulemaking process, the agencies
reconciled these differences, jointly presented improved approaches in
the NPRM similar to what NHTSA presented in the Draft TAR, and again
used those reconciled approaches in today's analysis.\372\
---------------------------------------------------------------------------
 \371\ See, e.g., Fiberglass to Carbon Fiber: Corvette's
Lightweight Legacy, GM (August 2012), https://media.gm.com/media/us/en/gm/news.detail.html/content/Pages/news/us/en/2012/Aug/0816_corvette.html.
 \372\ Because these road load technologies are no longer double
counted, the projected compliance pathway in the NPRM, and in
today's analysis for stringent alternatives, often requires more
advanced fuel saving technologies than previously projected,
including higher projected penetration rates of hybrid and electric
vehicle technologies.
---------------------------------------------------------------------------
 Many commenters correctly observed that the analysis fleet in the
NPRM recognized more technology content in the baseline than in the
Draft TAR (with higher penetration rates of tire rolling resistance and
aerodynamic drag improvements, for instance), but also that the fuel
economy values of the fleet had not improved all that much from the
previous year. Some commenters concluded that the NPRM baseline
technology assignment process was arbitrary and overstated the
technology content already present in the baseline
fleet.373 374 The agencies agree that there was a large
increase in the amount of road load technology credited in the baseline
fleet between EPA's Draft TAR and the jointly produced NPRM, and
clarify that this change was largely due to a recognition of
technologies that were actually present in the fleet, but not properly
accounted for in previous analyses. The change in penetration rates of
road load technologies (after accounting for glider share updates,
which is discussed in more detail in the mass reduction technology
section) between the NPRM and today's analysis is relatively small.
---------------------------------------------------------------------------
 \373\ NHTSA-2018-0067-11741, ICCT.
 \374\ NHTSA-2018-0067-12039, Union of Concerned Scientists.
---------------------------------------------------------------------------
 Many commenters noted that the different baseline road load
assumptions (and other technology modeling) materially affect
compliance pathways, and projected costs.\375\ ICCT commented that the
agencies should conduct sensitivity analyses assuming every vehicle in
the analysis fleet is set to zero percent road load technology
improvement, to demonstrate how the technology content of the analysis
fleet affected the compliance scenarios.\376\
---------------------------------------------------------------------------
 \375\ NHTSA-2018-0067-11928, Ford Motor Company.
 \376\ NHTSA-2018-0067-11741, ICCT.
---------------------------------------------------------------------------
 While the agencies have clearly described the methods by which
initial road load technologies are assigned in Section VI.C.4 Mass
Reduction, Section VI.C.5 Aerodynamics, and Section VI.C.6 Tire Rolling
Resistance below, the agencies considered a sensitivity case that
assumed no mass reduction, rolling resistance, or aerodynamic
improvements had been made to the MY 2017 fleet (i.e., setting all
vehicle road levels to zero--MRO, AERO and ROLL0). While this is an
unrealistic characterization of the initial fleet, the agencies
conducted a sensitivity analysis to understand any affect it may have
on technology penetration along other paths (e.g. engine and hybrid
technology). Under the CAFE program, the sensitivity analysis shows a
slight decrease in reliance on engine technologies (HCR engines,
turbocharge engines, and engines utilizing cylinder deactivation) and
hybridization (strong hybrids and plug-in hybrids) in the baseline
(relative to the central analysis). The consequence of this shift to
reliance on lower-level road load technologies is a reduction in
compliance cost in the baseline of about $300 per vehicle (in MY 2026).
As a result, cost savings in the preferred alternative are reduced by
about $200 per vehicle. Under the CO2 program, the general
trend in technology shift is less dramatic (though the change in BEVs
is larger) than the CAFE results. The cost change is also comparable,
but slightly smaller ($200 per vehicle in the baseline) than the CAFE
program results. Cost savings under the preferred alternative are
further reduced by about $100. With the lower technology costs in all
cases, the consumer payback periods decreased as well. These results
are consistent with the approach taken by manufacturers who have
already deployed many of the low-level road load reduction
opportunities to improve fuel economy.
 Some commenters preferred that the agencies develop a different
methodology based on reported road load coefficients (``A,'' ``B'' and
``C'' coastdown coefficients) to estimate levels of aerodynamic drag
improvement and rolling resistance in the baseline fleet that did not
rely on CBI.\377\ The agencies considered this, but determined that
using CBI to assign baseline aerodynamic drag levels and rolling
resistance values was more accurate and appropriate. Estimating
aerodynamic drag levels and rolling resistance levels from coastdown
coefficients is not straightforward, and to do it well would require
information the agencies do not have (much of which is also CBI). For
instance, rotational inertias of wheel, tire, and brake packages can
affect coastdown, so mass of the vehicle is not sufficient. The frontal
area of the vehicles, a key component for calculating aerodynamic drag,
is rarely known, and often requires manufacturer input to get an
accurate value. Other important vehicle features like all-wheel-drive
should also be accounted for, and the agencies would struggle to
correctly identify improvements in rolling resistance, low-drag brakes,
and secondary axle disconnect, because all of these technologies would
present similar signature on a coast down test. All of these
technologies are represented as technology pathways in today's
analysis. Manufacturers acknowledged the possibility of using road load
coefficients to estimate rolling resistance and aerodynamic features,
but warned that the process ``required
[[Page 24296]]
various assumptions and is not very accurate,'' and stated that the use
of CBI to assess aerodynamic and rolling resistance technologies is an
``accurate and practical solution'' to assign these difficult to
observe technologies.\378\
---------------------------------------------------------------------------
 \377\ NHTSA-2018-0067-11741, ICCT.
 \378\ NHTSA-2018-0067-12073, Alliance of Automobile
Manufacturers.
---------------------------------------------------------------------------
(3) Assigning Engine Configurations
 Engine technology costs can vary significantly by the configuration
of the engine. For instance, adding variable valve lift to each
cylinder on an engine would cost more for an engine with eight
cylinders than an engine with four cylinders. Similarly, the cost of
adding a turbocharger to an engine and downsizing the engine would be
different going from a naturally aspirated V8 to a turbocharged V6 than
going from a naturally aspirated V6 to a turbocharged I4. As discussed
in detail in the engine technology section of this document, the cost
files for the CAFE model account for instances such as these examples.
 Information in the analysis fleet enables the CAFE model to
reference the intended engine costs. The ``Engine Technology Class
(Observed)'' lists the architecture of the observed engine. Notably,
the analysis assumes that nearly all turbo charged engines take
advantage of downsizing to optimize fuel efficiency, minimize the cost
of turbo charging, and to maintain performance (to the extent
practicable) with the naturally aspirated counterpart engine.
Therefore, engines observed in the fleet that have already been down-
sized must reference costs for a larger basic engine, which assumes
down-sizing with the application of turbo technology. In these cases,
the ``Engine Technology Class'' which is used to reference costs will
be larger than the ``Engine Technology Class (Observed).''
 This is the same process agencies used in the NPRM, and it corrects
a previous error in the Draft TAR analysis, which incorrectly
underestimated turbocharged engine costs.\379\ Some commenters
expressed confusion and disagreement with this correction, with some
even commenting that the analysis baselessly inflated costs of
turbocharging technologies between the Draft TAR and the NPRM.\380\ To
be clear, this was a correction so that the costs used to calculate
turbocharged engine costs accurately reflected the total costs for a
turbocharged engine.
---------------------------------------------------------------------------
 \379\ For instance, the Draft TAR engine costs would map an
observed V6 Turbo engine to I4 Turbo engine costs, by referencing a
4C1B engine cost.
 \380\ NHTSA-2018-0067-11741, ICCT.
---------------------------------------------------------------------------
(4) Characterizing Shared Vehicle Platforms, Engines, and Transmissions
 Another aspect of characterizing vehicle model/configurations in
the analysis fleet is based on whether they share a ``platform'' with
other vehicle model/configurations. A ``platform'' refers to engineered
underpinnings shared on several differentiated products. Manufacturers
share and standardize components, systems, tooling, and assembly
processes within their products (and occasionally with the products of
another manufacturer) to manage complexity and costs for development,
manufacturing, and assembly.
 The concept of platform sharing has evolved over time. Years ago,
manufacturers rebadged vehicles and offered luxury options only on
premium nameplates (and manufacturers shared some vehicle platforms in
limited cases). Today, manufacturers share parts across highly
differentiated vehicles with different body styles, sizes, and
capabilities that may share the same platform. For instance, the Honda
Civic and Honda CR-V share many parts and are built on the same
platform. Engineers design chassis platforms with the ability to vary
wheelbase, ride height, and even driveline configuration. Assembly
lines can produce hatchbacks and sedans to cost-effectively utilize
manufacturing capacity and respond to shifts in market demand. Engines
made on the same line may power small cars or mid-size sport utility
vehicles. In addition, although the agencies' analysis, like past CAFE
analyses, considers vehicles produced for sale in the U.S., the agency
notes these platforms are not constrained to vehicle models built for
sale in the U.S.; many manufacturers have developed, and use, global
platforms, and the total number of platforms is decreasing across the
industry. Several automakers (for example, General Motors and Ford)
either plan to, or already have, reduced their number of platforms to
less than 10 and account for the overwhelming majority of their
production volumes on that small number of platforms.
 Vehicle model/configurations derived from the same platform are so
identified in the analysis fleet. Many manufacturers' use of vehicle
platforms is well documented in the public record and widely recognized
among the vehicle engineering community. Engineering knowledge,
information from trade publications, and feedback from manufacturers
and suppliers was also used to assign vehicle platforms in the analysis
fleet.
 When the CAFE model is deciding where and how to add technology to
vehicles, if one vehicle on the platform receives new technology, other
vehicles on the platform also receive the technology as part of their
next major redesign or refresh.\381\ Similar to vehicle platforms,
manufacturers create engines that share parts. For instance,
manufacturers may use different piston strokes on a common engine
block, or bore out common engine block castings with different
diameters to create engines with an array of displacements. Head
assemblies for different displacement engines may share many components
and manufacturing processes across the engine family. Manufacturers may
finish crankshafts with the same tools to similar tolerances. Engines
on the same architecture may share pistons, connecting rods, and the
same engine architecture may include both six and eight cylinder
engines. One engine family may appear on many vehicles on a platform,
and changes to that engine may or may not carry through to all the
vehicles. Some engines are shared across a range of different vehicle
platforms. Vehicle model/configurations in the analysis fleet that
share engines belonging to the same platform are also identified as
such.
---------------------------------------------------------------------------
 \381\ The CAFE model assigns mass reduction technology at a
platform level, but many other technologies may be assigned and
shared at a vehicle nameplate or vehicle model level.
---------------------------------------------------------------------------
 It is important to note that manufacturers define common engines
differently. Some manufacturers consider engines as ``common'' if the
engines shared an architecture, components, or manufacturing processes.
Other manufacturers take a narrower definition, and only assume
``common'' engines if the parts in the engine assembly are the same. In
some cases, manufacturers designate each engine in each application as
a unique powertrain. For example, a manufacturer may have listed two
engines separately for a pair that share designs for the engine block,
the crank shaft, and the head because the accessory drive components,
oil pans, and engine calibrations differ between the two. In practice,
many engines share parts, tooling, and assembly resources, and
manufacturers often coordinate design updates between two similar
engines. Engine families, designated in the analysis using ``engine
codes,'' for each manufacturer were tabulated and assigned based on
data-driven criteria. If engines shared a common cylinder count and
configuration, displacement, valvetrain, and fuel type, those engines
[[Page 24297]]
may have been considered together. In addition, if the compression
ratio, horsepower, and displacement of engines were only slightly
different, those engines were considered the same for the purposes of
redesign and sharing.
 Vehicles in the analysis fleet with the same engine family will,
therefore, adopt engine technology in a coordinated fashion.
Specifically, if such vehicles have different design schedules (i.e.,
refresh and redesign schedules), and a subset of vehicles using a given
engine add engine technologies during of a redesign or refresh that
occurs in an early model year (e.g., 2018), other vehicles using the
same engine ``inherit'' these technologies at the soonest ensuing
refresh or redesign. This is consistent with a view that, over time,
most manufacturers are likely to find it more practicable to shift
production to a new version of an engine than to continue production of
both the new engine and a ``legacy'' engine indefinitely. By grouping
engines together, the CAFE model controls future engine families to
ensure reasonable powertrain complexity. This means, however, that for
manufacturers that submitted highly atomized engine and transmission
portfolios, there is a practical cap on powertrain complexity and the
ability of the manufacturer to optimize the displacement of (i.e.,
``right size'') engines perfectly for each vehicle configuration. This
concept is discussed further in Section VI.B.4.a), below.
 Like with engines, manufacturers often use transmissions that are
the same or similar on multiple vehicles. Manufacturers may produce
transmissions that have nominally different machining to castings, or
manufacturers may produce transmissions that are internally identical,
except for the final gear ratio. In some cases, manufacturers sub-
contract with suppliers that deliver whole transmissions. In other
cases, manufacturers form joint ventures to develop shared
transmissions, and these transmission platforms may be offered in many
vehicles across manufacturers. Manufacturers use supplier and joint-
venture transmissions to a greater extent than they do with engines. To
reflect this reality, shared transmissions were considered for
manufacturers as appropriate. Transmission configurations are referred
to in the analysis as ``transmission codes.'' Like the inheritance
approach outlined for engines, if one vehicle application of a shared
transmission family upgraded the transmission, other vehicle
applications also upgraded the transmission at the next refresh or
redesign year. To define common transmissions, the agencies considered
transmission type (manual, automatic, dual-clutch, continuously
variable), number of gears, and vehicle architecture (front-wheel-
drive, rear-wheel-drive, all-wheel-drive based on a front-wheel drive
platform, or all-wheel-drive based on a rear-wheel-drive platform). If
vehicles shared these attributes, these transmissions were grouped for
the analysis. Vehicles in the analysis fleet with the same transmission
configuration will adopt transmission technology together, as described
above.
 Having all vehicles that share a platform (or engines that are part
of a family) adopt fuel economy-improving/CO2 emissions-
reducing technologies together, subject to refresh/redesign
constraints, reflects the real-world considerations described above,
but also overlooks some decisions manufacturers might make in the real
world in response to market pull. Accordingly, even though the analysis
fleet is incredibly complex, it is also over-simplified in some
respects compared to the real world. For example, the CAFE model does
not currently attempt to simulate the potential for a manufacturer to
shift the application of technologies to improve performance rather
than fuel economy. Therefore, the model's representation of the
``inheritance'' of technology can lead to estimates a manufacturer
might eventually exceed fuel economy standards as technology continues
to propagate across shared platforms and engines. While the agencies
have previously seen examples of extended periods during which some
manufacturers exceeded one or both CAFE and/or CO2
standards, in plenty of other examples, manufacturers chose to
introduce (or even reintroduce) technological complexity into their
vehicle lineups in response to buyer preferences. Going forward, and
recognizing the recent trend for consolidating platforms, it seems
likely manufacturers will be more likely to choose efficiency over
complexity in this regard; therefore, the potential should be lower
that today's analysis turns out to be oversimplified compared to the
real world.
 Manufacturers described shared engines, transmissions, and vehicle
platforms as ``standard business practice'' and they were encouraged
that the NHTSA analysis in the Draft TAR, and the jointly issued NPRM
placed realistic limits on the number of unique engines and
transmissions in a powertrain portfolio.\382\ In previous rulemakings,
stakeholders pointed out that shared parts and portfolio complexity
should be considered (but were not), and that the proliferation of
unique technology combinations resulting from unconstrained compliance
pathways would jeopardize economies of scale in the real world.\383\
---------------------------------------------------------------------------
 \382\ NHTSA-2018-0067-12150, Toyota North America.
 \383\ Alliance of Automobile Manufacturers, EPA-HQ-OAR-0827 and
NHTSA-2016-0068.
---------------------------------------------------------------------------
 HD Systems acknowledged that previous rulemakings did not
appropriately consider part sharing, but contended that in today's
global marketplace, manufacturers have flexibility to compete in new
ways that break old part sharing rules.\384\ The agencies acknowledge
that some transmissions are now sourced through suppliers, and that
economies of scale could, in the future be achieved at an industry
level instead of a manufacturer level; however, even when manufacturers
outsource a transmission, recent history suggests they apply that
transmission to multiple vehicles to control assembly plant and service
parts complexity, as they would if they were making the transmission
themselves. Similarly, even for global platforms, or global
powertrains, there is little evidence that manufacturers fragment
powertrain line-ups for a vehicle, or a set of vehicles that have
typically used the same engine. The agencies will continue to consider
how to capture more accurately the ways vehicles share engines,
transmissions, and platforms in future rulemakings, but the part-
sharing and modeling approach presented in the NPRM and this final rule
represents a marked improvement over previous analysis.
---------------------------------------------------------------------------
 \384\ NHTSA-2018-0067-11985, HD Systems.
---------------------------------------------------------------------------
(5) Characterizing Production Design Cycles
 Another aspect of characterizing vehicles in the analysis fleet is
based on when they can next be refreshed or redesigned. Redesign
schedules play an important role in determining when new technologies
may be applied. Many technologies that improve fuel economy and reduce
CO2 emissions may be difficult to incorporate without a
major product redesign. Therefore, each vehicle model in the analysis
fleet has an associated redesign schedule, and the CAFE model uses that
schedule to implement significant advances in some technologies (like
major mass reduction) to redesign years, while allowing manufacturers
to include minor advances (such as improved tire rolling
[[Page 24298]]
resistance) during a vehicle ``refresh,'' or a smaller update made to a
vehicle, which can happen between redesigns. In addition to refresh and
redesign schedules associated with vehicle model/configurations,
vehicles that share a platform subsequently have platform-wide refresh
and redesign schedules for mass reduction technologies.
 To develop the refresh/redesign cycles used for the NPRM vehicles
in the analysis fleet, information from commercially available sources
was used to project redesign cycles through MY 2022, as was done for
NHTSA's analysis for the 2016 Draft TAR.\385\ Commercially available
sources' estimates through MY 2022 are generally supported by detailed
consideration of public announcements plus related intelligence from
suppliers and other sources, and recognize that uncertainty increases
considerably as the forecasting horizon is extended. For MYs 2023-2035,
in recognition of that uncertainty, redesign schedules were extended
considering past pacing for each product, estimated schedules through
MY 2022, and schedules for other products in the same technology
classes. As mentioned above, potentially confidential forward-looking
information was not requested from manufacturers; nevertheless, all
manufacturers had an opportunity to review the estimates of product-
specific redesign schedules. A few manufacturers provided related
forecasts and, for the most part, that information corroborated the
estimates.
---------------------------------------------------------------------------
 \385\ In some cases, data from commercially available sources
was found to be incomplete or inconsistent with other available
information. For instance, commercially available sources identified
some newly imported vehicles as new platforms, but the international
platform was midway through the product lifecycle. While new to the
U.S. market, treating these vehicles as new entrants would have
resulted in artificially short redesign cycles if carried forward,
in some cases. Similarly, commercially available sources labeled
some product refreshes as redesigns, and vice versa. In these
limited cases, the data was revised to be consistent with other
available information or typical redesign and refresh schedules for
CAFE modeling. In these limited cases, the forecast time between
redesigns and refreshes was updated to match the observed past
product timing.
---------------------------------------------------------------------------
 Some commenters suggested supplanting these estimated redesign
schedules with estimates applying faster cycles (e.g., four to five
years), and this approach was considered for the analysis. Some
manufacturers tend to operate with faster redesign cycles and may
continue to do so, and manufacturers tend to redesign some products
more frequently than others. However, especially considering that
information presented by manufacturers largely supports estimates
discussed above, applying a ``one size fits all'' acceleration of
redesign cycles would not improve the analysis; instead, assuming a
fixed, shortened redesign schedule across the industry would likely
reduce consistency with the real world, especially for light trucks,
which are redesigned, on average, no less than every six years (see
Table VI-9, below). Moreover, if some manufacturers accelerate
redesigns in response to new standards, doing so would likely involve
costs (greater levels of stranded capital, reduced opportunity to
benefit from ``learning''-related cost reductions) greater than
reflected in other inputs to the analysis.
 As discussed in the NPRM, manufacturers use diverse strategies with
respect to when, and how often they update vehicle designs. While most
vehicles have been redesigned sometime in the last five years, many
vehicles have not. In particular, vehicles with lower annual sales
volumes tend to be redesigned less frequently, perhaps giving
manufacturers more time to recoup the investment needed to bring the
product to market. In some cases, manufacturers continue to produce and
sell vehicles designed more than a decade ago.
BILLING CODE 4910-59-P
[GRAPHIC] [TIFF OMITTED] TR30AP20.093
[[Page 24299]]
 Each manufacturer may use different strategies throughout their
product portfolio, and a component of each strategy may include the
timing of refresh and redesign cycles. Table VI-10 summarizes the
average time between redesigns, by manufacturer, by vehicle technology
class. Dashes mean the manufacturer has no volume in that vehicle
technology class in the MY 2017 analysis fleet. Across the industry,
manufacturers average 6.6 years between product redesigns.
[GRAPHIC] [TIFF OMITTED] TR30AP20.094
 Trends on redesign schedules identified in the NPRM remain in place
for today's analysis. Pick-up trucks have much longer redesign
schedules than small cars. Some manufacturers redesign vehicles often,
while other manufacturers redesign vehicles less often. Even if two
manufacturers have similar redesign cadence, the model years in which
the redesigns occur may still be different and dependent on where each
of the manufacturer's products are in their life cycle.
 Table VI-11 summarizes the average age of manufacturers' offering
by vehicle technology class. A value of ``0.0'' means that every
vehicle for a manufacturer in the vehicle technology class, represented
by the MY 2017 analysis fleet was new in MY 2017. Across the industry
manufacturers redesigned MY 2017 vehicles an average of 3.5 years
earlier, meaning the average MY 2017 vehicle was last redesigned in
approximately MY 2013, also on average near a midpoint in their product
lifecycle.
[[Page 24300]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.095
BILLING CODE 4910-59-C
 Some commenters cited examples of vehicles in the NPRM analysis
fleet where the redesign years were off by a year here or there in the
2017-2022 timeframe relative to the most recent public announcements,
or that the extended forecasts were too rigid.\386\ The CAFE model
structurally requires an input for the redesign years, and the agencies
worked to make these generally representative without disclosing
precise CBI product plans. Many of the redesign schedules were carried
over from the NPRM, with a few minor updates.
---------------------------------------------------------------------------
 \386\ NHTSA-2018-0067-11723, Natural Resources Defense Council.
---------------------------------------------------------------------------
 Some commenters contended that the agencies should not look at the
historical data to project the timing between redesigns (``business as
usual''), but should instead adopt a ``policy case'' with an
accelerated pace of redesigns and refreshes.\387\ Some commenters
suggested that the agencies use a standard 5 or 6 year redesign
schedule for all manufacturers and all products as a way to lower
projected costs.\388\ Other stakeholders commented that the entire
industry should be modeled with the ability to redesign everything at
one time in the near term because that would not presuppose precisely
how manufacturers may adjust their fleet.\389\
---------------------------------------------------------------------------
 \387\ NHTSA-2018-0067-11723, Natural Resources Defense Council.
 \388\ NHTSA-2018-0067-11985, HD Systems.
 \389\ NHTSA-2018-0067-12039, Union of Concerned Scientists.
---------------------------------------------------------------------------
 If the agencies were to implement any such approaches, the agencies
would need to more precisely account for tooling costs, research and
development costs, and product lifecycle marketing costs, or risk
missing ``hidden costs'' of a shortened cadence. To account properly
for these, the CAFE model would require major changes, and would
require specific inputs that are currently covered generically under
the retail price equivalency (RPE) factor.\390\ The agencies considered
these comments, and decided the process for refresh and redesign
outlined in the NPRM was a reasonable and realistic approach to
characterize product changes. The agencies conducted sensitivity
analysis with compressed redesign and refresh schedules, though these
ignore the resulting compressed amortization schedules, missing
important costs that are incorporated in the current RPE assumptions.
---------------------------------------------------------------------------
 \390\ Shorter redesign schedules are likely to put upward
pressure on RPE, as the manufacturers would have less time to recoup
investments.
---------------------------------------------------------------------------
 Some commenters claimed that the agency had extraordinarily
extended redesign schedule of 17.7 years for FCA between 2021-2025, and
an average redesign time of 25.8 years for Ford between 2022-2025.\391\
The agencies found these claims inaccurate and without basis. Table VI-
10, ``Summary of Sales Weighted Average Time
[[Page 24301]]
between Engineering Redesigns, by Manufacturer, by Vehicle Technology
Class'' summarizes the data used in today's analysis (which is very
similar to the information used in the NPRM, with some minor
adjustments and updates to the fleet), and the detailed information
vehicle-by-vehicle is reported in the ``market data'' file. The
agencies recognize that the natural sequence of redesigns for some
manufacturers and some products is not ideal to meet stringent
alternatives, which is part of the consideration for economic
practicability and technological feasibility. Manufacturers commented
supportively on the idea of vehicle specific redesign schedules, and
the redesign cadence used in the NPRM, as these contribute to realistic
assessments of new technology penetration within the fleet, and
acknowledge the heterogeneity in the product development approaches and
business practices for each manufacturer.\392\ One commenter recognized
that redesign and refresh schedules represented a vast improvement over
phase-in caps to model the adoption of mature technologies.\393\
---------------------------------------------------------------------------
 \391\ NHTSA-2018-0067-11723, Natural Resources Defense Council.
 \392\ NHTSA-2018-0067-11928, Ford Motor Company.
 \393\ NHTSA-2018-0067-0444, Walter Kreucher.
---------------------------------------------------------------------------
 Other commenters argued that the structural construct of
technologies only being available at redesign or at refresh (via
inheritance) did not reflect real world actions and was not supported
by any actual data.\394\ Other commenters acknowledged the inheritance
of engine and transmission technologies at refresh as an important,
positive feature of the CAFE model.\395\ HD Systems argued that an
engine or transmission package available in other markets on a global
platform could be imported to the U.S. market during refresh, and did
not require a ``leader'' at redesign in the U.S. market to seed
adoption. HDS cited a few examples where manufacturers have introduced
strong hybrid powertrains on an existing vehicle a year or two after
the product launch, not associated with any particular vehicle redesign
or refresh.
---------------------------------------------------------------------------
 \394\ NHTSA-2018-0067-11985, HD Systems.
 \395\ NHTSA-2018-0067-11723, Natural Resources Defense Council.
---------------------------------------------------------------------------
 The agencies carefully considered these comments, and observed that
some relatively low volume hybrid options may appear after launch, or
that some transmissions were quickly replaced shortly after a major
redesign. In many of these cases, launch delays, warranty claims, or
other external factors contributed to, at least in part, an atypically
timed introduction of fuel saving technology to the fleet.\396\ At this
point, this does not appear to be a mainstream, or preferred industry
practice. However, the agencies will continue to evaluate this. For
future rulemaking, the agencies may consider engine refresh and
redesign cycles for engines and transmissions. These may be separate
from vehicle redesign and refresh schedules because the powertrain
product lifecycles may be longer on average than the typical vehicle
redesign schedules. This approach, if researched and implemented in
future analysis, could provide some opportunity for manufacturers to
introduce new powertrain technologies independent of the vehicle
redesign schedules, in addition to inheriting advanced powertrain
technology as refresh as already modeled in the NPRM and today's
analysis.
---------------------------------------------------------------------------
 \396\ Such instances are observable in detailed CAFE and
CO2 compliance data submitted to EPA and NHTSA.
---------------------------------------------------------------------------
 For today's analysis, the agencies, with a few exceptions based on
updated publicly available information, carried over redesign cadences
for each vehicle nameplate as presented in the NPRM. The agencies do
not claim that the projected redesign years will perfectly match what
industry does--notably because refresh and redesign information is CBI
and the agencies have applied more generalized schedules to protect the
CBI. Also, what any individual manufacturer may choose to do today
could be completely different than what it chooses to do tomorrow due
to changing business circumstances and plans--but the agencies have
worked to ensure the timing of redesigns will be roughly correct
(especially in the near term), and that the time between redesigns will
continue forward for each manufacturer as it has based on recent
history. The agencies have also increased the frequency of refreshes in
response to comments about the proliferation of some engine and
transmission families through manufacturers' product portfolios.
 Also for today's analysis, the agencies now explicitly model CAFE
compliance pathways out through 2050. For the model to work as
intended, the agencies must project refresh and redesign schedules out
through 2050. The agencies recognize that the accuracy of predictions
about the distant future, particularly about refresh and redesign
cycles through the 2030-2050 timeframe, are likely to be poor. If
historical evolution of the industry continues, many of the nameplates
carried forward in the fleet are likely to be out of production, and
new nameplates not considered in the analysis are sure to emerge.
Still, carrying forward the MY 2017 fleet with the current refresh and
redesign cadences is consistent with the current analysis, and imposing
an alternative schedule on the fleet, or making up new nameplates and
retiring older nameplates without a clear basis, would lack proper
foundation.
BILLING CODE 4910-59-P
[[Page 24302]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.096
BILLING CODE 4910-59-C
(6) Defining Technology Adoption Features
 In some circumstances, the agencies may reference full vehicle
simulation effectiveness data for technology combinations that are not
able to be, or are not likely to be applied to all vehicles. In some
cases, a specific technology as modeled only exists on paper, and
questions remain about the technological feasibility of the efficiency
characterization.\397\ Or, a technology may perform admirably on the
test cycle, but fail to meet all functional, or performance
requirements for certain vehicles.\398\ In other cases, the
intellectual property landscape may make commercialization of one
technology risky for a manufacturer without the consent of the
intellectual property owner.\399\ In such cases, the agencies may not
allow a technology to be applied to a certain vehicle. The agencies
designate this in the ``market data'' file with a ``SKIP'' for the
technology and vehicle. The logic is explained technology by technology
in this document, as the logic was explained in the PRIA for this rule.
---------------------------------------------------------------------------
 \397\ High levels of aerodynamic drag reduction for some body
styles, or EPA's previous, speculative characterization of ``HCR2''
engines, for example.
 \398\ Examples of applications that are unsuitable for certain
technologies include low end torque requirements for HCR engines on
high load vehicles, or towing and trailering applications,
continuously variable transmissions in high torque applications, and
low rolling resistance tires on vehicles built for precision
cornering and high lateral forces, or instant acceleration from a
stand still.
 \399\ Variable compression ratio engines, for example.
---------------------------------------------------------------------------
 Some commenters argued that the restrictions of technologies on a
case-by-case basis required case-by-case explanation (and not objective
specification defined cut-offs), and that the use of CBI for
performance considerations was unacceptable unless fully
disclosed.\400\ As discussed above, the agencies are not able to
disclose CBI. Stakeholders have had plenty of opportunities to comment
on the applicability of technologies, including the few that have used
SKIP logic restrictions for a portion of the fleet.
---------------------------------------------------------------------------
 \400\ NHTSA-2018-0067-11741, ICCT.
---------------------------------------------------------------------------
 Other commenters suggested an optimistic and wholly unfounded
approach to manufacturer innovation, arguing that costs would continue
to come down (beyond what is currently modeled with cost learning), and
the list of fuel-saving technologies would continually regenerate
itself (even if the technological mechanism for fuel saving
technologies was not yet identified).\401\ Therefore, the argument goes
that people will figure out new ways to improve fuel saving
technologies to increase their applicability, and the current
technology characterization should be enabled for selection with no
restriction--not because the commenter knows how the technology will be
adapted, but that the commenter believes the technology could,
eventually, within the timeline of the rulemaking, be adapted, brought
to market, and be accepted by consumers. While the agencies recognize
the improvements that many manufacturers
[[Page 24303]]
have achieved in fuel saving technologies, some of which were difficult
to foresee, the agencies have an obligation under the law to be
judicious and specific about technological feasibility, and to avoid
speculative conclusions about technologies to justify the rulemaking.
---------------------------------------------------------------------------
 \401\ NHTSA-208-0067-12122-33, American Council for an Energy-
Efficient Economy.
---------------------------------------------------------------------------
c) Other Analysis Fleet Data
(1) Safety Classes
 The agencies referenced the mass-size-safety analysis to project
the effects changes in weight may have on crash fatalities. That
analysis, discussed in more detail in Section VI.D.2, considers how
weight changes may affect safety for cars, crossover utility vehicles
and sport utility vehicles, and pick-up trucks. To consider these
effects, the agencies mapped each vehicle in the analysis fleet to the
appropriate ``Safety Class.''
(2) Labor Utilization
 The analysis fleet summarizes components of direct labor for each
vehicle considered in the analysis. The labor is split into three
components: (1) Dealership hours worked on sales functions per vehicle,
(2) direct assembly labor for final assembly, engine, and transmission,
and (3) percent U.S. content.
 In the MY 2016 fleet for the NPRM, the agencies catalogued
production locations and plant employment, reviewed annual reports from
the North American Dealership Association to estimate dealership
employment (27.8 hours per vehicle sold), and estimated the industry
average labor hours for final assembly of vehicles (30 hours per
vehicle produced), engine machining and assembly (4 hours per engine
produced), and transmission production (5 hours per transmission
produced).
 Today's analysis fleet carries over the estimated labor
coefficients for sales and production, but references the most recent
Part 583 American Automobile Labeling Act Report for percent U.S.
content and for the location of vehicle assembly, engine assembly, and
transmission assembly.\402\
---------------------------------------------------------------------------
 \402\ Part 583 American Automobile Labeling Act Report,
available at https://www.nhtsa.gov/part-583-american-automobile-labeling-act-reports.
---------------------------------------------------------------------------
(3) Production Volumes for Sales Analysis
 A final important aspect of projecting what vehicles will exist in
future model years and potential manufacturer responses to standards is
estimating how future sales might change in response to different
potential standards. If potential future standards appear likely to
have major effects in terms of shifting production from cars to trucks
(or vice versa), or in terms of shifting sales between manufacturers or
groups of manufacturers, that is important for the agencies to
consider. For previous analyses, the CAFE model used a static forecast
contained in the analysis fleet input file, which specified changes in
production volumes over time for each vehicle model/configuration. This
approach yielded results that, in terms of production volumes, did not
change between scenarios or with changes in important model inputs. For
example, very stringent standards with very high technology costs would
result in the same estimated production volumes as less stringent
standards with very low technology costs. For this analysis, as in the
proposal, the CAFE model begins with the first-year production volumes
(i.e., MY 2017 for today's analysis) and adjusts ensuing sales mix year
by year (between cars and trucks, and between manufacturers)
endogenously as part of the analysis, rather than using external
forecasts of future car/truck split and future manufacturer sales
volumes. This leads the model to produce different estimates of future
production volumes under different standards and in response to
different inputs, reflecting the expectation that regulatory standards
and other external factors will, in fact, impact the market.
(4) Comments on Other Analysis Fleet Data
 Some commenters suggest that the CAFE model should run as a full
consumer choice model (and this idea is discussed in more detail in
Section VI.D.1). While this sounds like a reasonable request on the
surface, such an approach would place enormous new demands on the data
characterized in the fleet (and preceding fleets, which may be needed
to calibrate a model properly). For instance, some model concepts may
depend on a bevy of product features, such as interior cargo room,
artistic appeal of the design, and perceived quality of the vehicle.
But product features alone may not be sufficient. Additional
information about dealership channels, product awareness and
advertising effectiveness, and financing terms also may be required.
Such information could dramatically increase the scope of work needed
to characterize the analysis fleet for future rulemakings. As described
in Section VI.D.1.b)(2)(d) Using Vehicle Choice Models in Rulemaking
Analysis. Accordingly, the agencies decided not to develop such a model
for this rulemaking.
2. Treatment of Compliance Credit Provisions
 Today's final rule involves a variety of provisions regarding
``credits'' and other compliance flexibilities. Some recently
introduced regulatory provisions allow a manufacturer to earn
``credits'' that will be counted toward a vehicle's rated
CO2 emissions level, or toward a fleet's rated average
CO2 or CAFE level, without reference to required levels for
these average levels of performance. Such flexibilities effectively
modify emissions and fuel economy test procedures, or methods for
calculating fleets' CAFE and average CO2 levels. Such
provisions are discussed below in Section VI.B.2. Other provisions (for
CAFE, statutory provisions) allow manufacturers to earn credits by
achieving CAFE or average CO2 levels beyond required levels;
these provisions may hence more appropriately be termed ``compliance
credits.''
 EPCA has long provided that, by exceeding the CAFE standard
applicable to a given fleet in a given model year, a manufacturer may
earn corresponding ``credits'' that the same manufacturer may, within
the same regulatory class, apply toward compliance in a different model
year. EISA amended these provisions by providing that manufacturers
may, subject to specific statutory limitations, transfer compliance
credits between regulatory classes, and trade compliance credits with
other manufacturers. The CAA provides EPA with broad standard-setting
authority for the CO2 program, with no specific directives
regarding either CO2 standards or CO2 compliance
credits.
 EPCA also specifies that NHTSA may not consider the availability of
CAFE credits (for transfer, trade, or direct application) toward
compliance with new standards when establishing the standards
themselves.\403\ Therefore, this analysis, like that presented in the
NPRM, considers 2020 to be the last model year in which carried-forward
or transferred credits can be applied for the CAFE program. Beginning
in model year 2021, today's ``standard setting'' analysis for NHTSA's
program is conducted assuming each fleet must comply with the CAFE
standard separately in every model year.
---------------------------------------------------------------------------
 \403\ 49 U.S.C. 32902(h)(3).
---------------------------------------------------------------------------
 The ``unconstrained'' perspective acknowledges that these
flexibilities exist as part of the program, and, while not considered
by NHTSA in setting standards, are nevertheless important to consider
when attempting to estimate the real impact of any alternative. Under
[[Page 24304]]
the ``unconstrained'' perspective, credits may be earned, transferred,
and applied to deficits in the CAFE program throughout the full range
of model years in the analysis. The Final Environmental Impact Analysis
(FEIS) accompanying today's final rule, like the corresponding Draft
EIS analysis, presents results of ``unconstrained'' modeling. Also,
because the CAA provides no direction regarding consideration of any
CO2 credit provisions, today's analysis, like the NPRM
analysis, includes simulation of carried-forward and transferred
CO2 credits in all model years.
 Some commenters took issue broadly with this treatment of
compliance credits. Michalek and Whitefoot wrote that ``we find this
requirement problematic because the automakers use these flexibilities
as a common means of complying with the regulation, and ignoring them
will bias the cost-benefit analysis to overestimate costs.'' \404\
---------------------------------------------------------------------------
 \404\ Michalek, J. and Whitefoot, K., NHTSA-2018-0067-11903, at
10-11.
---------------------------------------------------------------------------
 Counter to the above general claim, the CAFE model does provide
means to simulate manufacturers' potential application of some
compliance credits, and both the analysis of CO2 standards
and the NEPA analysis of CAFE standards do make use of this aspect of
the model. As discussed above, NHTSA does not have the discretion to
consider the credit program--in fact, the agency is prohibited by
statute from doing so--in establishing maximum feasible standards.
Further, as discussed below, the agencies also continue to find it
appropriate for the analysis largely to refrain from simulating two of
the mechanisms allowing the use of compliance credits.
 The model's approach to simulating compliance decisions accounts
for the potential to earn and use CAFE credits as provided by EPCA/
EISA. The model similarly accumulates and applies CO2
credits when simulating compliance with EPA's standards. Like past
versions, the current CAFE model can be used to simulate credit carry-
forward (a.k.a. banking) between model years and transfers between the
passenger car and light truck fleets but not credit carry-back (a.k.a.
borrowing) from future model years or trading between manufacturers.
 Regarding the potential to carry back compliance credits, UCS
commented that, although past versions of the CAFE model had
``considered this flexibility in its approach to multiyear modeling,''
NHTSA had, without explanation, ``abruptly discontinued support of this
method of compliance,'' such that ``manufacturers are generally
incentivized to over comply, regardless of whether carrying forward a
deficit to be compensated by later overcompliance would be a more cost-
effective method of compliance.'' \405\ Citing the potential that
manufacturers could make use of carried back credits in the future, UCS
also stated that ``NHTSA's decision to constrain it in the model is
unreasonable and arbitrary.'' \406\ UCS effectively implies that the
agencies should base standards on analysis that presumes manufacturers
will take full theoretical advantage of provisions allowing credits to
be borrowed.
---------------------------------------------------------------------------
 \405\ UCS, NHTSA-2018-0067-12039, Technical Appendix, at 44.
 \406\ UCS, op. cit., at 77.
---------------------------------------------------------------------------
 The agencies have carefully considered these comments, and while
EPA's decisions regarding CO2 standards can consider the
potential to carry back compliance credits from later to earlier model
years, and NHTSA's ``unconstrained'' evaluation could also do so, past
examples of failed attempts to carry back CAFE credits (e.g., a MY2014
carry back default leading to a civil penalty payment) underscore the
riskiness of such ``borrowing.'' Recent evidence indicates
manufacturers are disinclined to take such risks,\407\ and both
agencies find it reasonable and prudent to refrain from attempting to
simulate such ``borrowing'' in rulemaking analysis.
---------------------------------------------------------------------------
 \407\ Section IX, below, reviews data regarding manufacturers'
use of CAFE compliance credit mechanism during MYs 2011-2016, and
shows that the use of ``carry back'' credits is, relative to the use
of other compliance credit mechanisms, too small to discern.
---------------------------------------------------------------------------
 Unlike past versions, the NPRM and current versions of CAFE model
provide a basis to specify (in model inputs) CAFE credits available
from model years earlier than those being explicitly simulated. For
example, with this analysis representing model years 2017-2050
explicitly, credits earned in model year 2012 are made available for
use through model year 2017 (given the current five-year limit on
carry-forward of credits). The banked credits are specific to both the
model year and fleet in which they were earned.
 In addition to the above-mentioned comments, UCS also cited as
``errors'' that ``the model does not accurately reflect the one-time
exemption from the EPA 5-year credit life for credits earned in the MY
2010-2015 timeframe'' and ``NHTSA assumes that there will be absolutely
no credit trading between manufacturers.''
 As discussed below, in the course of updating the analysis fleet
from MY 2016 to MY 2017, the agencies have updated and expanded the
manner in which the model accounts for credits earned prior to MY 2017,
including credits earned as early as MY 2009. In order to increase the
realism with which the model transitions between the early model year
(MYs 2017-2020) and the later years that are the subject of this
action, the agencies have accounted for the potential that some
manufacturers might trade some of these pre-MY 2017 credits to other
manufacturers. However, as with the NPRM, the analysis refrains from
simulating the potential that manufacturers might continue to trade
credits during and beyond the model years covered by today's action.
The agencies remain concerned that any realistic simulation of such
trading would require assumptions regarding which specific pairs of
manufacturers might actually trade compliance credits, and the evidence
to date makes it clear that the credit market is far from fully
``open.'' With respect to the FCA comment cited above, the agencies
also remain concerned that to set standards based on an analysis that
presumes the use of program flexibilities risks making the
corresponding actions mandatory. Some flexibilities--credit carry-
forward (banking) and transfers between fleets in particular--involve
little risk, because they are internal to a manufacturer and known in
advance. As discussed above, credit carry-back involves significant
risk, because it amounts to borrowing against future improvements,
standards, and production volume and mix--and anticipated market demand
for fuel efficient vehicles often fail to materialize. Similarly,
credit trading also involves significant risk, because the ability of
manufacturer A to acquire credits from manufacturer B depends not just
on manufacturer B actually earning the expected amount of credit, but
also on manufacturer B being willing to trade with manufacturer A, and
on potential interest by other manufacturers. Manufacturers' compliance
plans have already evidenced cases of compliance credit trades that
were planned and subsequently aborted, reinforcing the agencies'
judgment that, like credit banking, credit trading involves too much
risk to be included in an analysis that informs decisions about the
stringency of future standards. Nevertheless, recognizing that some
manufacturers have actually been trading credits, the agencies have, as
in the NPRM, included in the sensitivity analysis a case that simulates
``perfect'' trading of compliance credits, focusing
[[Page 24305]]
on CO2 standards to illustrate the hypothetical maximum
potential impact of trading. The FRIA summarizes results of this and
other cases included in the sensitivity analysis.
 As discussed in the CAFE model documentation, the model's default
logic attempts to maximize credit carry-forward--that is, to ``hold
on'' to credits for as long as possible. If a manufacturer needs to
cover a shortfall that occurs when insufficient opportunities exist to
add technology in order to achieve compliance with a standard, the
model will apply credits. Otherwise the manufacturer carries forward
credits until they are about to expire, at which point it will use them
before adding technology that is not considered cost-effective. The
model attempts to use credits that will expire within the next three
years as a means to smooth out technology application over time to
avoid both compliance shortfalls and high levels of over-compliance
that can result in a surplus of credits. Although it remains impossible
precisely to predict manufacturer's actual earning and use of
compliance credits, and this aspect of the model may benefit from
future refinement as manufacturers and regulators continue to gain
experience with these provisions, this approach is generally consistent
with manufacturers' observed practices.
 NHTSA introduced the CAFE Public Information Center to provide
public access to a range of information regarding the CAFE
program,\408\ including manufacturers' credit balances. However, there
is a data lag in the information presented on the CAFE PIC that may not
capture credit actions across the industry for as much as several
months. Furthermore, CAFE credits that are traded between manufacturers
are adjusted to preserve the gallons saved that each credit
represents.\409\ The adjustment occurs at the time of application
rather than at the time the credits are traded. This means that a
manufacturer who has acquired credits through trade, but has not yet
applied them, may show a credit balance that is either considerably
higher or lower than the real value of the credits when they are
applied. For example, a manufacturer that buys 40 million credits from
Tesla may show a credit balance in excess of 40 million. However, when
those credits are applied, they may be worth only 1/10 as much--making
that manufacturer's true credit balance closer to 4 million than 40
million.
---------------------------------------------------------------------------
 \408\ CAFE Public Information Center, http://www.nhtsa.gov/CAFE_PIC/CAFE_PIC_Home.htm (last visited June 22, 2018).
 \409\ CO2 credits for EPA's program are denominated
in metric tons of CO2 rather than gram/mile compliance
credits and require no adjustment when traded between manufacturers
or fleets.
---------------------------------------------------------------------------
 For the NPRM, the agencies reviewed then-recent credit balances,
estimated the potential that some manufacturers could trade credits,
and developed inputs that make carried-forward credits available in
each of model years 2011-2015, after subtracting credits assumed to be
traded to other manufacturers, adding credits assumed to be acquired
from other manufacturers through such trades, and adjusting any traded
credits (up or down) to reflect their true value for the fleet and
model year into which they were traded.\410\ For today's analysis, an
additional model year's data was available in mid-2019, and the
agencies updated these inputs, as summarized in Table VI-12, Table VI-
13, and Table VI-14. While the CAFE model will transfer expiring
credits into another fleet (e.g., moving expiring credits from the
domestic car credit bank into the light truck fleet), some of these
credits were moved into the initial banks to improve the efficiency of
application and both to reflect better the projected shortfalls of each
manufacturer's regulated fleets and to represent observed behavior. For
context, a manufacturer that produces one million vehicles in a given
fleet, and experiences a shortfall of 2 mpg, would need 20 million
credits, adjusted for fuel savings, to offset the shortfall completely.
---------------------------------------------------------------------------
 \410\ The adjustments, which are based upon the CAFE standard
and model year of both the party originally earning the credits and
the party applying them, were implemented assuming the credits would
be applied to the model year in which they were set to expire. For
example, credits traded into a domestic passenger car fleet for MY
2014 were adjusted assuming they would be applied in the domestic
passenger car fleet for MY 2019.
---------------------------------------------------------------------------
BILLING CODE 4910-59-P
[[Page 24306]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.097
[GRAPHIC] [TIFF OMITTED] TR30AP20.098
[[Page 24307]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.099
BILLING CODE 4910-59-C
 In addition to the inclusion of these existing credit banks, the
CAFE model also updated its treatment of credits in the rulemaking
analysis. EPCA requires that NHTSA set CAFE standards at maximum
feasible levels for each model year without consideration of the
program's credit mechanisms. However, as recent NHTSA CAFE/EPA tailpipe
CO2 emissions rulemakings have evaluated effects of
standards over longer time periods, the early actions taken by
manufacturers required more nuanced representation. Accordingly, the
CAFE model now provides for a setting to establish a ``last year to
consider credits.'' This adjustment is set at the last year for which
new standards are not being considered (MY 2020 in this analysis). This
allows the model to replicate the practical application of existing
credits toward compliance in early years but also to examine the impact
of proposed standards based solely on fuel economy improvements in all
years for which new standards are being considered.
 Regarding the model's simulation of manufacturers' potential
earning and application of compliance credits, UCS commented that the
model ``inexplicably lets credits expire'' because ``all technologies
which pay for themselves within the assumed payback period are applied
to all manufacturers, regardless of credit status.'' UCS also claimed
that ``NHTSA did not accurately reflect unique attributes of EPA's
credit bank,'' that ``credits are not traded between manufacturers,''
and that ``NHTSA does not model credit carryback for compliance.''
\411\ Relatedly, as discussed above, UCS attributes modeling outcomes
to the ``effective cost'' metric used to select from among available
fuel-saving technologies.\412\ As discussed in Section VI.B.1, the
agencies expect that manufacturers are likely to improve fuel economy
voluntarily insofar as doing so ``pays back'' economically within a
short period (30 months), and the agencies note that periods of
regulatory stability have, in fact, been marked by CAFE levels
exceeding requirements. As discussed above, the agencies have excluded
simulation of credit trading (except in MYs prior to those under
consideration, aside from an idealized case presented in the
sensitivity analysis) and likewise excluded simulation of potential
``carryback'' provisions. The agencies have excluded modeling these
scenarios not just because of the analytical complexities involved (and
rejecting, for example, the random number generator analysis suggested
by UCS), but also because the agencies agree that the actual provisions
regarding trading and borrowing of compliance credits create too much
risk to be used in the analysis underlying consideration of standards.
However, as discussed above, the agencies have revised the ``metric''
used to prioritize available options to apply fuel-saving technologies.
As discussed below, the agencies have revised model inputs to include
the large quantity of ``legacy'' compliance credits EPA has made
available under its CO2 standards.
---------------------------------------------------------------------------
 \411\ UCS, NHTSA-2018-0067-12039, Technical Appendix, at 35-46.
 \412\ UCS, NHTSA-2018-0067-12039, Technical Appendix, at 28-30.
---------------------------------------------------------------------------
 The CAFE model has also been modified to include a similar
representation of existing credit banks in EPA's CO2
program. While the life of a CO2 credit, denominated in
metric tons of CO2, has a five-year life, matching the
lifespan of CAFE credits, such credits earned in the early MY 2009-2011
years of the EPA program, may be used through MY 2021.\413\ The CAFE
model was not modified to allow
[[Page 24308]]
exceptions to the life-span of compliance credits, and, to reflect
statutory requirements, treated them as if they may be carried forward
for no more than five years, so the initial credit banks were modified
to anticipate the years in which those credits might be needed. MY 2016
was simulated explicitly in the NPRM analysis to prohibit the inclusion
of banked credits in MY 2016 (which could be carried forward from MY
2016 to MY 2021), and thus underestimated the extent to which
individual manufacturers, and the industry as a whole, could rely on
these early credits to comply with EPA standards between MY 2016 and MY
2021. However, as indicated in the NPRM, the final rule's model inputs
updated the analysis fleet's basis to MY 2017, such that these
additional banked credits can be included. The credit banks with which
the simulations in this analysis were conducted are presented in the
following Tables:
---------------------------------------------------------------------------
 \413\ In the 2010 rule, EPA placed limits on credits earned in
MY 2009, which expired prior to this rule. However, credits
generated in MYs 2010-2011 may be carried forward, or traded, and
applied to deficits generated through MY 2021.
---------------------------------------------------------------------------
BILLING CODE 4910-59-P
[GRAPHIC] [TIFF OMITTED] TR30AP20.100
[[Page 24309]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.101
BILLING CODE 4910-59-C
 While the CAFE model does not simulate the ability to trade credits
between manufacturers, it does simulate the strategic accumulation and
application of compliance credits, as well as the ability to transfer
credits between fleets to improve the compliance position of a less
efficient fleet by leveraging credits earned by a more efficient fleet.
The model prefers to hold on to earned compliance credits within a
given fleet, carrying them forward into the future to offset potential
future deficits. This assumption is consistent with observed strategic
manufacturer behavior dating back to 2009.
 From 2009 to present, no manufacturer has transferred CAFE credits
into a fleet to offset a deficit in the same year in which they were
earned. This has occurred with credits acquired from other
manufacturers via trade but not with a manufacturer's own credits.
Therefore, the current representation of credit transfers between
fleets--where the model prefers to transfer expiring, or soon-to-be-
expiring credits rather than newly earned credits--is both appropriate
and consistent with observed industry behavior.
 This may not be the case for CO2 standards, though it is
difficult to be certain at this point. The CO2 program
seeded the industry with a large quantity of early compliance credits
(earned in MYs 2009-2011) \414\ prior to the existence formal
CO2 standards. Early credits from MYs 2010 and 2011,
however, do not expire until 2021. Thus, for manufacturers looking to
offset deficits, it is more sensible to exhaust credits that were
generated during later model years (which are set to expire within the
next five years), rather than relying on the initial bank of credits
from MYs 2010 and 2011. The first model year for which earned credits
outlive the initial bank is MY 2017, for which final manufacturer
CO2 performance data (and hence, banked credits) has not yet
been released. However, considering that under the CO2
program manufacturers simultaneously comply with passenger car and
light truck fleets, to more accurately represent the CO2
credit system the CAFE model allows (and encourages) intra-year
transfers between regulated fleets for the purpose of simulating
compliance with the CO2 standards.
---------------------------------------------------------------------------
 \414\ In response to public comment, EPA eliminated the possible
use of credits earned in MY 2009 for future model years. However,
credits earned in MY 2010 and MY 2011 remain available for use.
---------------------------------------------------------------------------
a) Off-Cycle and A/C Efficiency Adjustments to CAFE and Average
CO2 Levels
 In addition to more rigorous accounting of CAFE and CO2
credits, the model now also accounts for air conditioning efficiency
and off-cycle adjustments. NHTSA's program considers those adjustments
in a manufacturer's compliance calculation starting in MY 2017, and the
NPRM version of the model used the adjustments claimed by each
manufacturer in MY 2016 as the starting point for all future years.
Because air conditioning efficiency and off-cycle adjustments are not
credits in NHTSA's program, but rather adjustments to compliance fuel
economy (much like the Flexible Fuel Vehicle adjustments due to phase
out in MY 2019), they may be included under either a ``standard
setting'' or ``unconstrained'' analysis perspective.
 The manner in which the CAFE model treats the EPA and CAFE A/C
efficiency and off-cycle credit programs is similar, but the model also
accounts for A/C leakage (which is not part of NHTSA's program). When
determining the compliance status of a
[[Page 24310]]
manufacturer's fleet (in the case of EPA's program, PC and LT are the
only fleet distinctions), the CAFE model weighs future compliance
actions against the presence of existing (and expiring) CO2
credits resulting from over-compliance with earlier years' standards,
A/C efficiency credits, A/C leakage credits, and off-cycle credits.
 Another aspect of credit accounting, implemented in the NPRM
version of the CAFE model, involved credits related to the application
of off-cycle and A/C efficiency adjustments, which manufacturers earn
by taking actions such as special window glazing or using reflective
paints that provide fuel economy improvements in real-world operation
but do not produce measurable improvements in fuel consumption on the
2-cycle test.
 NHTSA's inclusion of off-cycle and A/C efficiency adjustments began
in MY 2017, while EPA has collected several years' worth of submissions
from manufacturers about off-cycle and A/C efficiency technology
deployment. Currently, the level of deployment can vary considerably by
manufacturer, with several claiming extensive Fuel Consumption
Improvement Values (FCIV) for off-cycle and A/C efficiency
technologies, and others almost none. The analysis of alternatives
presented here (and in the NPRM) does not attempt to project how future
off-cycle and A/C efficiency technology use will evolve or speculate
about the potential proliferation of FCIV proposals submitted to the
agencies. Rather, this analysis uses the off-cycle credits submitted by
each manufacturer for MY 2017 compliance, and, with a few exceptions,
carries these forward to future years. Several of the technologies
described below are associated with A/C efficiency and off-cycle FCIVs.
In particular, stop-start systems, integrated starter generators, and
full hybrids are assumed to generate off-cycle adjustments when applied
to vehicles to improve their fuel economy. Similarly, higher levels of
aerodynamic improvements are assumed to include active grille shutters
on the vehicle, which also qualify for off-cycle FCIVs.
 The NPRM analysis assumed that any off-cycle FCIVs that are
associated with actions outside of the technologies discussed in
Section VI.C (either chosen from the pre-approved ``pick list,'' or
granted in response to individual manufacturer petitions) remained at
the levels claimed by manufacturers in MY 2017. Any additional A/C
efficiency and off-cycle adjustments that accrued as the result of
explicit technology application calculated dynamically in each model
year for each alternative. The NPRM version of the CAFE model also
represented manufacturers' credits for off-cycle improvements, A/C
efficiency improvements, and A/C leakage reduction in terms of values
applicable across all model years.
 Recognizing that application of these improvements thus far varies
considerably among manufacturers, such that some manufacturers have
opportunities to earn significantly more of the corresponding
adjustments over time, the agencies have expanded the CAFE model's
representation of these credits to provide for year-by-year
specification of the amounts of each type of adjustment for each
manufacturer, denominated in grams CO2 per mile,\415\ as
summarized in the following table:
---------------------------------------------------------------------------
 \415\ For estimating their contribution to CAFE compliance, the
grams CO2/mile values in Table VI-1711 are converted to
gallons/mile and applied to a manufacturer's 2-cycle CAFE
performance. When calculating compliance with EPA's CO2
program, there is no conversion necessary (as standards are also
denominated in grams/mile).
 \416\ These values are specified in the ``market_ref.xlsx''
input file's ``Credits and Adjustments'' worksheet. The file is
available with the archive of model inputs and outputs posted at
https://www.nhtsa.gov/corporate-average-fuel-economy/compliance-and-effects-modeling-system.
---------------------------------------------------------------------------
BILLING CODE 4910-59-P
[[Page 24311]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.102
[[Page 24312]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.103
[[Page 24313]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.104
BILLING CODE 4910-59-C
 In addition to these refinements to the estimation of the
quantities of adjustments earned over time by each manufacturer, the
agencies revised the
[[Page 24314]]
CAFE model to apply estimates of the corresponding costs. For today's
analysis, the agencies applied estimates developed previously by EPA,
adjusting these values to 2019 dollars. The following table summarizes
inputs through model year 2030:
[GRAPHIC] [TIFF OMITTED] TR30AP20.105
 The model currently accounts for any off-cycle adjustments
associated with technologies that are included in the set of fuel-
saving technologies explicitly simulated as part of this proposal (for
example, start-stop systems that reduce fuel consumption during idle or
active grille shutters that improve aerodynamic drag at highway speeds)
and accumulates these adjustments up to the 10 g/mi cap. As a practical
matter, most of the adjustments for which manufacturers are claiming
off-cycle FCIV exist outside of the technology tree, so the cap is
rarely reached during compliance simulation. The agencies have
considered the potential to model their application explicitly.
However, doing so would require data regarding which vehicle models
already possess these improvements as well as the cost and expected
value of applying them to other models in the future. Such data is
currently too limited to support explicit modeling of these
technologies and adjustments.
b) Alternative Fuel Vehicles
 When establishing maximum feasible fuel economy standards, NHTSA is
prohibited from considering the availability of alternatively fueled
vehicles,\417\ and credit provisions related to AFVs that significantly
increase their fuel economy for CAFE compliance purposes. Under the
``standard setting'' perspective, these technologies (pure battery
electric vehicles and fuel cell vehicles) \418\ are not available in
the compliance simulation to improve fuel economy. Under the
``unconstrained'' perspective, such as is documented in the DEIS and
FEIS, the CAFE model considers these technologies in the same manner as
other available technologies, and may apply them if they represent
cost-effective compliance pathways. However, under both perspectives,
the analysis continues to include dedicated AFVs that already exist in
the MY 2017 fleet (and their projected future volumes). Also, because
the CAA provides no direction regarding consideration of alternative
fuels, the final rule's analysis includes simulation of the potential
that some manufacturers might introduce new AFVs in response to
CO2 standards. To represent the compliance benefit from such
a response fully, NHTSA modified the CAFE model to include the specific
provisions related to AFVs under the CO2 standards. In
particular, the CAFE model now carries a full representation of the
production multipliers related to electric vehicles, fuel cell
vehicles, plug-in hybrids, and CNG vehicles, all of which vary by year
through MY 2021.
---------------------------------------------------------------------------
 \417\ 49 U.S.C. 32902(h).
 \418\ Dedicated compressed natural gas (CNG) vehicles should
also be excluded in this perspective but are not considered as a
compliance strategy under any perspective in this analysis.
---------------------------------------------------------------------------
 EPCA also provides that CAFE levels may, subject to limitations, be
adjusted upward to reflect the sale of flexible fuel vehicles (FFVs).
Although these adjustments end after model year 2020, the final rule's
analysis, like the NPRM's, includes estimated potential use through MY
2019, as summarized below:
[[Page 24315]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.106
 For its part, EPA has provided that manufacturers selling
sufficient numbers of PHEVs, BEVs, and FCVs may, when calculating fleet
average CO2 levels, ``count'' each unit of production as
more than a single unit. The CAFE model accounts for these
``multipliers.'' As for the NPRM, the final rule's analysis applies the
following multipliers:
[GRAPHIC] [TIFF OMITTED] TR30AP20.107
 For example, under EPA's current regulation, when calculating the
average CO2 level achieved by its MY 2019 passenger car
fleet, a manufacturer may treat each 1,000 BEVs as 2,000 BEVs. When
calculating the average level required of this fleet, the manufacturer
must use the actual production volume (in this example, 1,000 units).
Similarly, the manufacturer must use the actual production volume when
calculating compliance credit balances.
 There were no natural gas vehicles in the baseline fleet, and the
analysis did not apply natural gas technology due to cost
effectiveness. The application of a 2.0 multiplier for natural gas
vehicles for MYs 2022-2026 would have no impact on the analysis because
given the state of natural gas vehicle refueling infrastructure, the
cost to equip vehicles with natural gas tanks, the outlook for
petroleum prices, and the outlook for battery prices, we have little
basis to project more than an inconsequential response to this
incentive in the foreseeable future.
 For the final rule's analysis, the CAFE model can be exercised in a
manner that simulates these current EPA requirements, or that simulates
two alternative approaches. The first includes the above-mentioned
multipliers in the calculation of average requirements, and the second
also includes the multipliers in the calculation of credit balances.
The central analysis reflects current regulations. The sensitivity
analysis presented in the FRIA includes a case
[[Page 24316]]
applying multipliers to the calculation of achieved and required
average CO2 levels, and calculation of credit balances.
c) Civil Penalties
 Throughout the history of the CAFE program, some manufacturers have
consistently achieved fuel economy levels below applicable standards,
electing instead to pay civil penalties as specified by EPCA. As in
previous versions of the CAFE model, the current version allows the
user to specify inputs identifying such manufacturers and to consider
their compliance decisions as if they are willing to pay civil
penalties for non-compliance with the CAFE program. As with the NPRM,
the civil penalty rate in the current analysis is $5.50 per 1/10 of a
mile per gallon, per vehicle manufactured for sale.
 NHTSA notes that treating a manufacturer as if it is willing to pay
civil penalties does not necessarily mean that it is expected to pay
penalties in reality. Doing so merely implies that the manufacturer
will only apply fuel economy technology up to a point, and then stop,
regardless of whether or not its corporate average fuel economy is
above its standard. In practice, the agencies expect that many of these
manufacturers will continue to be active in the credit market, using
trades with other manufacturers to transfer credits into specific
fleets that are challenged in any given year, rather than paying
penalties to resolve CAFE deficits. The CAFE model calculates the
amount of penalties paid by each manufacturer, but it does not simulate
trades between manufacturers. In practice, some (possibly most) of the
total estimated penalties may be a transfer from one OEM to another.
 Although EPCA, as amended in 2007 by the Energy Independence and
Security Act (EISA), prescribes these specific civil penalty provisions
for CAFE standards, the Clean Air Act (CAA) does not contain similar
provisions. Rather, the CAA's provisions regarding noncompliance
prohibit sale of a new motor vehicle that is not covered by an EPA
certificate of conformity, and in order to receive such a certificate
the new motor vehicle must meet EPA's Section 202 regulations,
including applicable emissions standards. Therefore, inputs regarding
civil penalties--including inputs regarding manufacturers' potential
willingness to treat civil penalty payment as an economic choice--apply
only to simulation of CAFE standards. On the other hand, some of the
same manufacturers recently opting to pay civil penalties instead of
complying with CAFE standards have also recently led adoption of lower-
GWP refrigerants, and the ``A/C leakage'' credits count toward
compliance only with CO2 standards, not CAFE standards. The
model accounts for this difference between the programs.
 When considering technology applications to improve fleet fuel
economy, the model will add technology up to the point at which the
effective cost of the technology (which includes technology cost,
consumer fuel savings, consumer welfare changes, and the cost of
penalties for non-compliance with the standard) is less costly than
paying civil penalties or purchasing credits. Unlike previous versions
of the model, the current implementation further acknowledges that some
manufacturers experience transitions between product lines where they
rely heavily on credits (either carried forward from earlier model
years or acquired from other manufacturers) or simply pay penalties in
one or more fleets for some number of years. The model now allows the
user to specify, when appropriate for the regulatory program being
simulated, on a year-by-year basis, whether each manufacturer should be
considered as willing to pay penalties for non-compliance. This
provides additional flexibility, particularly in the early years of the
simulation. As discussed above, this assumption is best considered as a
method to allow a manufacturer to under-comply with its standard in
some model years--treating the civil penalty rate and payment option as
a proxy for other actions it may take that are not represented in the
CAFE model (e.g., purchasing credits from another manufacturer, carry-
back from future model years, or negotiated settlements with NHTSA to
resolve deficits).
 For the NPRM, NHTSA relied on past compliance behavior and
certified transactions in the credit market to designate some
manufacturers as willing to pay CAFE penalties in some model years. The
full set of NPRM assumptions regarding manufacturer behavior with
respect to civil penalties is presented in Table VI-21, which shows all
manufacturers were assumed to be willing to pay civil penalties prior
to MY 2020. This was largely a reflection of either existing credit
balances (which manufacturers will use to offset CAFE deficits until
the credits reach their expiration dates) or inter-manufacturer trades
assumed likely to happen in the near future, based on previous
behavior. The manufacturers in the table whose names appear in bold all
had at least one regulated fleet (of three) whose CAFE was below its
standard in MY 2016. Because the NPRM analysis began with the MY 2016
fleet, and no technology could be added to vehicles that are already
designed and built, all manufacturers could generate civil penalties in
MY 2016. However, once a manufacturer is designated as unwilling to pay
penalties, the CAFE model will attempt to add technology to the
respective fleets to avoid shortfalls.
[[Page 24317]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.108
 Several of the manufacturers in Table VI-21 that were presumed to
be willing to pay civil penalties in the early years of the program
have no history of paying civil penalties. However, several of those
manufacturers have either bought or sold credits--or transferred
credits from one fleet to another to offset a shortfall in the
underperforming fleet. As the CAFE model does not simulate credit
trades between manufacturers, providing this additional flexibility in
the modeling avoids the outcome where the CAFE model applies more
technology than needed in the context of the full set of compliance
flexibilities at the industry level. By statute, NHTSA cannot consider
credit flexibilities when setting standards, so most manufacturers
(those without a history of civil penalty payment) are assumed to
comply with their standards through fuel economy improvements for the
model years being considered in this analysis. The notable exception to
this assumption is Fiat Chrysler Automobiles (FCA), which could still
satisfy the requirements of the program through a combination of credit
application and civil penalties through MY 2025 before eventually
complying exclusively through fuel economy improvements in MY 2026.
 As mentioned above, the CAA does not provide civil penalty
provisions similar to those provisions specified in EPCA/EISA, and the
above-mentioned corresponding inputs apply only to simulation of
compliance with CAFE standards.
 Some stakeholders offering comments related to the analytical
treatment of civil penalties indicated that NHTSA should tend toward
assuming manufacturers will take advantage of this EPCA provision as an
economically attractive alternative to compliance. Other commenters
implied that NHTSA should tend toward not relying on compliance
flexibilities in the analysis used to determine the maximum feasible
stringency of CAFE standards. For example, New York University's
Institute for Policy Integrity (IPI) offered the following comments:
 NHTSA assumes that most manufacturers will be unwilling to pay
penalties based in part on the fact that most manufacturers have not
paid penalties in recent years. The Proposed Rule cites the
statutory prohibition on NHTSA considering credit trading as a
reason to assume manufacturers without a history of paying penalties
will comply through technology alone, whatever the cost. But this is
an arbitrary assumption and is in no way dictated by the statute.
NHTSA knows as much, since elsewhere in the proposed rollback, the
agency explains ``EPCA is very clear as to which flexibilities are
not to be considered'' and NHTSA is allowed to consider off-cycle
adjustments because they are not specifically mentioned. But
considering penalties are not mentioned as off-limits for NHTSA in
setting the standards either. Instead, the prohibition focuses on
credit trading and transferring. The penalty safety valve has
existed in EPCA for decades, and Congress clearly would have known
how to add penalties to the list of trading and transferring. The
fact that Congress did not bar NHTSA from considering penalties as a
safety valve means that NHTSA must consider manufacturer's efficient
use of penalties as a cost minimizing compliance option. Besides,
NHTSA does consider penalties for some of the manufacturers making
its statutory justification even less rational.\419\
---------------------------------------------------------------------------
 \419\ Institute for Policy Integrity, NHTSA-2018-0067-12213, at
24.
 On the other hand, in more general comments about NHTSA's
analytical treatment of program flexibilities, FCA stated that ``when
flexibilities are considered while setting targets, they cease to be
flexibilities and become simply additional technology mandates.'' \420\
---------------------------------------------------------------------------
 \420\ FCA, Docket #NHTSA-2018-0067-11943, at 6.
---------------------------------------------------------------------------
 NHTSA agrees with IPI that EPCA does not expressly prohibit NHTSA,
when conducting analysis supporting determinations of the maximum
feasible stringency of future CAFE standards, from including
manufacturers' potential tendency to pay civil penalties rather than
complying with those standards. However, EPCA also does not require
NHTSA to include this tendency in its analysis. NHTSA also notes, as
does IPI, that EPCA does prohibit NHTSA from including credit trading,
transferring, or the availability of credits in such
[[Page 24318]]
analysis (although NHTSA interprets this prohibition to apply only to
the model years for which standards are being set). This statutory
difference is logical based on the way credits and penalties function
differently under EPCA. Because credits help manufacturers achieve
compliance with CAFE standards, absent the statutory prohibition,
credits would be relevant to the feasibility of a standard.\421\
Penalties, on the other hand, do not enable a manufacturer to comply
with an applicable standard; penalties are for noncompliance.\422\ When
Congress added credit trading provisions to EPCA in 2007, NHTSA
anticipated that competitive considerations would make manufacturers
reluctant to engage in such trades. Since that time, manufacturers
actually have demonstrated otherwise, although the reliance on
trading--especially between specific pairs of OEMs--appears to vary
widely. At this time, NHTSA considers it most likely that manufacturers
will shift away from paying civil penalties and toward compliance
credit trading. Consequently, for NHTSA to include civil penalty
payment in its analysis would increasingly amount to using civil
penalty payment as an analytical proxy for credit trading. Having
further considered the question, NHTSA's current view is, therefore,
that including civil penalty payment beyond MY 2020 would effectively
subvert EPCA's prohibition against considering credit trading.
Therefore, for today's announcement, NHTSA has modified its analysis to
assume that BMW, Daimler, FCA, JLR, and Volvo would consider paying
civil penalties through MY 2020, and that all manufacturers would apply
as much technology as would be needed in order to avoid paying civil
penalties after MY 2020.
---------------------------------------------------------------------------
 \421\ See 49 U.S.C. 32911(b) (``Compliance is determined after
considering credits available to the manufacturer . . . . '').
 \422\ See id.
---------------------------------------------------------------------------
3. Technology Effectiveness Values
 The next input required to simulate manufacturers' decision-making
processes for the year-by-year application of technologies to specific
vehicles is estimates of how effective each technology would be at
reducing fuel consumption. In the NPRM, the agencies used full-vehicle
modeling and simulation to estimate the fuel economy improvements
manufacturers could make to a fleet of vehicles, considering those
vehicles' technical specifications and how combinations of technologies
interact. Full-vehicle modeling and simulation uses computer software
and physics-based models to predict how combinations of technologies
perform as a full system under defined conditions.
 A model is a mathematical representation of a system, and
simulation is the behavior of that mathematical representation over
time. In this analysis, the model is a mathematical representation of
an entire vehicle,\423\ including its individual components such as the
engine and transmission, overall vehicle characteristics such as mass
and aerodynamic drag, and the environmental conditions, such as ambient
temperature and barometric pressure. The agencies simulated the model's
behavior over test cycles, including the 2-cycle laboratory compliance
tests (or 2-cycle tests),\424\ to determine how the individual
components interact. 2-cycle tests are test cycles that are used to
measure fuel economy and emissions for CAFE and CO2
compliance, and therefore are the relevant test cycles for determining
technology effectiveness when establishing standards. In the
laboratory, 2-cycle testing involves sophisticated test and measurement
equipment, carefully controlled environmental conditions, and precise
procedures to provide the most repeatable results possible with human
drivers. Measurements using these structured procedures serve as a
yardstick for fuel economy and CO2 emissions.
---------------------------------------------------------------------------
 \423\ Our full vehicle model was composed of sub-models, which
is why the full vehicle model could also be referred to as a full
system model, composed of sub-system models.
 \424\ EPA's compliance test cycles are used to measure the fuel
economy of a vehicle. For readers unfamiliar with this process, it
is like running a car on a treadmill following a program--or more
specifically, two programs. The ``programs'' are the ``urban
cycle,'' or Federal Test Procedure (abbreviated as ``FTP''), and the
``highway cycle,'' or Highway Fuel Economy Test (abbreviated as
``HFET''), and they have not changed substantively since 1975. Each
cycle is a designated speed trace (of vehicle speed versus time)
that all certified vehicles must follow during testing. The FTP is
meant roughly to simulate stop and go city driving, and the HFET is
meant roughly to simulate steady flowing highway driving at about 50
mph. For further details on compliance testing, see the discussion
in Section VI.B.3.a)(7).
---------------------------------------------------------------------------
 Full-vehicle modeling and simulation was initially developed to
avoid the costs of designing and testing prototype parts for every new
type of technology. For example, if a truck manufacturer has a concept
for a lightweight tailgate and wants to determine the fuel economy
impact for the weight reduction, the manufacturer can use physics-based
computer modeling to estimate the impact. The vehicle, modeled with the
proposed change, can be simulated on a defined test route and under a
defined test condition, such as city or highway driving in warm ambient
temperature conditions, and compared against the baseline reference
vehicle. Full-vehicle modeling and simulation allows the consideration
and evaluation of different designs and concepts before building a
single prototype. In addition, full vehicle modeling and simulation is
beneficial when considering technologies that provide small incremental
improvements. These improvements are difficult to measure in laboratory
tests due to variations in how vehicles are driven over the test cycle
by human drivers, variations in emissions measurement equipment, and
variations in environmental conditions.\425\
---------------------------------------------------------------------------
 \425\ Difficulty with controlling for such variability is
reflected, for example, in 40 CFR 1065.210, Work input and output
sensors, which describes complicated instructions and
recommendations to help control for variability in real world (non-
simulated) test instrumentation set up.
---------------------------------------------------------------------------
 Full-vehicle modeling and simulation requires detailed data
describing the individual technologies and performance-related
characteristics. Those specifications generally come from design
specifications, laboratory measurements, and other subsystem
simulations or modeling. One example of data used as an input to the
full vehicle simulation are engine maps for each engine technology that
define how much fuel is consumed by the engine technology across its
operating range.
 Using full-vehicle modeling and simulation to estimate technology
efficiency improvements has two primary advantages over using single or
limited point estimates. An analysis using single or limited point
estimates may assume that, for example, one fuel economy improving
technology with an effectiveness value of 5 percent by itself and
another technology with an effectiveness value of 10 percent by itself,
when applied together achieve an additive improvement of 15 percent.
Single point estimates generally do not provide accurate effectiveness
values because they do not capture complex relationships among
technologies. Technology effectiveness often differs significantly
depending on the vehicle type (e.g., sedan versus pickup truck) and how
the technology interacts with other technologies on the vehicle, as
different technologies may provide different incremental levels of fuel
economy improvement if implemented alone or in tandem with other
technologies. Any oversimplification of these complex interactions
leads to less accurate and often overestimated effectiveness estimates.
 In addition, because manufacturers often implement several fuel-
saving
[[Page 24319]]
technologies simultaneously when redesigning a vehicle, it is difficult
to isolate the effect of individual technologies using laboratory
measurement of production vehicles alone. Modeling and simulation
offers the opportunity to isolate the effects of individual
technologies by using a single or small number of baseline vehicle
configurations and incrementally adding technologies to those baseline
configurations. This provides a consistent reference point for the
incremental effectiveness estimates for each technology and for
combinations of technologies for each vehicle type. Vehicle modeling
also reduces the potential for overcounting or undercounting technology
effectiveness.
 An important feature of this analysis is that the incremental
effectiveness of each technology and combinations of technologies be
accurate and relative to a consistent baseline vehicle. The absolute
fuel economy values of the full vehicle simulations are used only to
determine incremental effectiveness and are never used directly to
assign an absolute fuel economy value to any vehicle model or
configuration for the rulemaking analysis.
 For this analysis, absolute fuel economy levels are based on the
individual fuel economy values from CAFE compliance data for each
vehicle in the baseline fleet. The incremental effectiveness from the
full vehicle simulations performed in Autonomie, a physics-based full-
vehicle modeling and simulation software developed and maintained by
the U.S. Department of Energy's Argonne National Laboratory, are
applied to baseline fuel economy to determine the absolute fuel economy
of applying the first technology change. For subsequent technology
changes, incremental effectiveness is applied to the absolute fuel
economy level of the previous technology configuration.
 For example, if a Ford F150 2-wheel drive crew cab and short bed in
the baseline fleet has a fuel economy value of 30 mpg for CAFE
compliance, 30 mpg will be considered the reference absolute fuel
economy value. A similar full vehicle model in the Autonomie simulation
may begin with an average fuel economy value of 32 mpg, and with
incremental addition of a specific technology X its fuel economy
improves to 35 mpg, a 9.3 percent improvement. In this example, the
incremental fuel economy improvement (9.3 percent) from technology X
would be applied to the F150's 30 mpg absolute value.
 For this analysis, the agencies determined the incremental
effectiveness of technologies as applied to the 2,952 unique vehicle
models in the analysis fleet. Although, as mentioned above, full-
vehicle modeling and simulation reduces the work and time required to
assess the impact of moving a vehicle from one technology state to
another, it would be impractical--if not impossible--to build a unique
vehicle model for every individual vehicle in the analysis fleet.
Therefore, as explained further below, vehicle models are built in a
way that maintains similar attributes to the analysis fleet vehicles,
which ensures key components are reasonably represented.
 We received a wide array of comments regarding the full-vehicle
modeling and simulation performed for the NPRM, but there was general
agreement that full-vehicle modeling and simulation was the appropriate
method to determine technology effectiveness.\426\ Stakeholders
commented on other areas, such as full vehicle simulation tools,
inputs, and assumptions, and these comments will be discussed in the
following sections. For this final rule, the agencies continued to use
the same full-vehicle simulation approach to estimate technology
effectiveness for technology adoption in the rulemaking timeframe. The
next sections will discuss the details of the explicit input
specifications and assumptions used for the final rule analysis.
---------------------------------------------------------------------------
 \426\ See NHTSA-2018-0067-12039; NHTSA-2018-0067-12073. UCS and
AAM both agreed that full vehicle simulation can significantly
improve the estimates of technology effectiveness.
---------------------------------------------------------------------------
a) Why This Rulemaking Used Autonomie Full-Vehicle Modeling and
Simulation To Determine Technology Effectiveness
 The NPRM and final rule analysis use effectiveness estimates for
technologies developed using Autonomie, a physics-based full-vehicle
modeling and simulation software developed and maintained by the U.S.
Department of Energy's Argonne National Laboratory.\427\ Autonomie was
designed to serve as a single tool to meet requirements of automotive
engineering throughout the vehicle development process, and has been
under continuous improvement by Argonne for over 20 years. Autonomie is
commercially available and widely used in the automotive industry by
suppliers, automakers, and academic researchers (who publish findings
in peer reviewed academic journals).\428\ DOE and manufacturers have
used Autonomie and its ability to simulate a large number of powertrain
configurations, component technologies, and vehicle-level controls over
numerous drive cycles to support studies on fuel efficiency, cost-
benefit analysis, and carbon dioxide emissions,\429\ and other topics.
---------------------------------------------------------------------------
 \427\ More information about Autonomie is available at https://www.anl.gov/technology/project/autonomie-automotive-system-design
(last accessed June 21, 2018). As mentioned in the preliminary
regulatory impact analysis (PRIA) for this rule, the agencies used
Autonomie version R15SP1, the same version used for the 2016 Draft
TAR.
 \428\ Rousseau, A. Shidore, N. Karbowski, D. Sharer, ``Autonomie
Vehicle Validation Summary.'' https://www.nhtsa.gov/sites/nhtsa.dot.gov/files/anl-autonomie-vehicle-model-validation-1509.pdf.
 \429\ Delorme et al. 2008, Rousseau, A, Sharer, P, Pagerit, S.,
& Das, S. ``Trade-off between Fuel Economy and Cost for Advanced
Vehicle Configurations,'' 20th International Electric Vehicle
Symposium (EVS20), Monaco (April 2005); Elgowainy, A., Burnham, A.,
Wang, M., Molburg, J., & Rousseau, A. ``Well-To-Wheels Energy Use
and Greenhouse Gas Emissions of Plug-in Hybrid Electric Vehicles,''
SAE 2009-01-1309, SAE World Congress, Detroit, April 2009.
---------------------------------------------------------------------------
 Autonomie has also been used to provide the U.S. government with
data to make decisions about future research, and is used by DOE for
analysis supporting budget priorities and plans for programs managed by
its Vehicle Technologies Office (VTO), and to support decision making
among competing vehicle technology research and development
projects.\430\ In addition, Autonomie is the primary vehicle simulation
tool used by DOE to support its U.S. DRIVE program, a government-
industry partnership focused on advanced automotive and related energy
infrastructure technology research and development.\431\
---------------------------------------------------------------------------
 \430\ U.S. DOE Benefits & Scenario Analysis publications is
available at https://www.autonomie.net/publications/fuel_economy_report.html (last accessed September 11, 2019).
 \431\ For more information on U.S. Drive, see https://www.energy.gov/eere/vehicles/us-drive.
---------------------------------------------------------------------------
 Autonomie is a MathWorks-based software environment and framework
for automotive control-system design, simulation, and analysis.\432\ It
is designed for rapid and easy integration of models with varying
levels of detail (low to high fidelity), abstraction (from subsystems
to systems and entire architectures), and processes (e.g., calibration,
validation). By building models automatically, Autonomie allows the
quick simulation of many component technologies and powertrain
configurations, and, in this case, to assess the energy consumption of
advanced powertrain technologies. Autonomie simulates subsystems,
[[Page 24320]]
systems, or entire vehicles; evaluates and analyzes fuel efficiency and
performance; performs analyses and tests for virtual calibration,
verification, and validation of hardware models and algorithms;
supports system hardware and software requirements; links to
optimization algorithms; and supplies libraries of models for
propulsion architectures of conventional powertrains as well as hybrid
and electric vehicles.
---------------------------------------------------------------------------
 \432\ Halbach, S. Sharer, P. Pagerit, P., Folkerts, C. &
Rousseau, A. ``Model Architecture, Methods, and Interfaces for
Efficient Math-Based design and Simulation of Automotive Control
Systems,'' SAE 2010-01-0241, SAE World Congress, Detroit, April,
2010.
---------------------------------------------------------------------------
 With hundreds of pre-defined powertrain configurations along with
vehicle level control strategies developed from dynamometer test data,
Autonomie is a highly capable tool for analyzing advantages and
drawbacks of applying different technology options within each
technology family, including conventional, parallel hybrid, power-split
hybrid electric vehicles (HEVs), plug-in hybrid electric vehicles
(PHEVs), battery electric vehicles (BEV) and fuel cell vehicles (FCVs).
Autonomie also allows users to evaluate the effect of component sizing
on fuel consumption for different powertrain technologies as well as to
define component requirements (e.g., power, energy) to maximize fuel
displacement for a specific application.\433\ To evaluate properly any
powertrain-configuration or component-sizing influence, vehicle-level
control models are critical, especially for electric drive vehicles
like hybrids and plug-in hybrids. Argonne has extensive expertise in
developing vehicle-level control models based on different approaches,
from global optimization to instantaneous optimization, rule-based
optimization, and heuristic optimization.\434\
---------------------------------------------------------------------------
 \433\ Nelson, P., Amine, K., Rousseau, A., & Yomoto, H. (EnerDel
Corp.), ``Advanced Lithium-ion Batteries for Plug-in Hybrid-electric
Vehicles,'' 23rd International Electric Vehicle Symposium (EVS23),
Anaheim, CA, (Dec. 2007); Karbowski, D., Haliburton, C., & Rousseau,
A. ``Impact of Component Size on Plug-in Hybrid Vehicles Energy
Consumption using Global Optimization,'' 23rd International Electric
Vehicle Symposium (EVS23), Anaheim, CA, (Dec. 2007).
 \434\ Karbowski, D., Kwon, J., Kim, N., & Rousseau, A.,
``Instantaneously Optimized Controller for a Multimode Hybrid
Electric Vehicle,'' SAE paper 2010-01-0816, SAE World Congress,
Detroit, April 2010; Sharer, P., Rousseau, A., Karbowski, D., &
Pagerit, S. ``Plug-in Hybrid Electric Vehicle Control Strategy--
Comparison between EV and Charge-Depleting Options,'' SAE paper
2008-01-0460, SAE World Congress, Detroit (April 2008); and
Rousseau, A., Shidore, N., Carlson, R., & Karbowski, D. ``Impact of
Battery Characteristics on PHEV Fuel Economy,'' AABC08.
---------------------------------------------------------------------------
 Autonomie has been developed to consider real-world vehicle metrics
like performance, hardware limitations, utility, and drivability
metrics (e.g., towing capability, shift busyness, frequency of engine
on/off transitions), which are important to producing realistic
estimates of fuel economy and CO2 emission rates. This
increasing realism has, in turn, steadily increased confidence in the
appropriateness of using Autonomie to make significant investment
decisions. Autonomie has also been validated for a number of powertrain
configurations and vehicle classes using Argonne's Advanced Mobility
Technology Laboratory (AMTL) (formerly Advanced Powertrain Research
Facility, or APRF) vehicle test data.\435\
---------------------------------------------------------------------------
 \435\ Jeong, J., Kim, N., Stutenberg, K., Rousseau, A.,
``Analysis and Model Validation of the Toyota Prius Prime.'' SAE
2019-01-0369, SAE World Congress, Detroit, April 2019; Kim, N,
Jeong, J. Rousseau, A. & Lohse-Busch, H. ``Control Analysis and
Thermal Model Development of PHEV,'' SAE 2015-01-1157, SAE World
Congress, Detroit, April 2015; Kim, N., Rousseau, A. & Lohse-Busch,
H. ``Advanced Automatic Transmission Model Validation Using
Dynamometer Test Data,'' SAE 2014-01-1778, SAE World Congress,
Detroit, Apr. 14; Lee, D. Rousseau, A. & Rask, E. ``Development and
Validation of the Ford Focus BEV Vehicle Model,'' 2014-01-1809, SAE
World Congress, Detroit, Apr. 14; Kim, N., Kim, N., Rousseau, A., &
Duoba, M. ``Validating Volt PHEV Model with Dynamometer Test Data
using Autonomie,'' SAE 2013-01-1458, SAE World Congress, Detroit,
Apr. 13; Kim, N., Rousseau, A., & Rask, E. ``Autonomie Model
Validation with Test Data for 2010 Toyota Prius,'' SAE 2012-01-1040,
SAE World Congress, Detroit, Apr. 12; Karbowski, D., Rousseau, A,
Pagerit, S., & Sharer, P. ``Plug-in Vehicle Control Strategy--From
Global Optimization to Real Time Application,'' 22th International
Electric Vehicle Symposium (EVS22), Yokohama, (October 2006).
---------------------------------------------------------------------------
 Argonne has spent several years developing, applying, and expanding
the means to use distributed computing to exercise its Autonomie full-
vehicle simulation tool over the scale necessary for realistic analysis
to provide data for CAFE and CO2 standards rulemaking. The
NPRM and PRIA detailed how Argonne used Autonomie to estimate the fuel
economy impacts for roughly a million combinations of technologies and
vehicle types.436 437 Argonne developed input parameters for
Autonomie to represent every combination of vehicle, powertrain, and
component technologies considered in this rulemaking. The sequential
addition of more than 50 fuel economy-improving technologies to ten
vehicle types generated more than 140,000 unique technology and vehicle
combinations. Running the Autonomie powertrain sizing algorithms to
determine the appropriate amount of engine downsizing needed to
maintain overall vehicle performance when vehicle mass reduction is
applied and for certain engine technology changes (discussed further,
below) increased the total number of simulations to more than one
million. The result of these simulations is a useful dataset
identifying the impacts of combinations of vehicle technologies on
energy consumption--a dataset that can be referenced as an input to the
CAFE model for assessing regulatory compliance alternatives.
---------------------------------------------------------------------------
 \436\ As part of the Argonne simulation effort, individual
technology combinations simulated in Autonomie were paired with
Argonne's BatPAC model to estimate the battery cost associated with
each technology combination based on characteristics of the
simulated vehicle and its level of electrification. Information
regarding Argonne's BatPAC model is available at http://www.cse.anl.gov/batpac/.
 \437\ Additionally, the impact of engine technologies on fuel
consumption, torque, and other metrics was characterized using GT
POWER simulation modeling in combination with other engine modeling
that was conducted by IAV Automotive Engineering, Inc. (IAV). The
engine characterization ``maps'' resulting from this analysis were
used as inputs for the Autonomie full-vehicle simulation modeling.
Information regarding GT Power is available at https://www.gtisoft.com/gt-suite-applications/propulsion-systems/gt-power-engine-simulation-software.
---------------------------------------------------------------------------
 The following sections discuss the full-vehicle modeling and
simulation inputs and data assumptions, and comments received on the
NPRM analysis. The discussion is necessarily technical, but also
important to understand the agencies' decisions to modify (or not) the
Autonomie analysis for the final rule.
(1) Full-Vehicle Modeling, Simulation Inputs and Data Assumptions
 The agencies provided extensive documentation that quantitatively
and qualitatively described the over 50 technologies considered as
inputs to the Autonomie modeling.438 439 These inputs
consisted of engine technologies, transmission technologies, powertrain
electrification, light-weighting, aerodynamic improvements, and tire
rolling resistance improvements.\440\ The PRIA provided an overview of
the sub-models for each technology, including the internal combustion
engine model, automatic transmission model, and others.\441\ The
Argonne NPRM model documentation expanded on these sub-models in detail
to show the interaction of each sub-model input and output.\442\
[[Page 24321]]
For example, as shown in Figure VI-2, the input for Autonomie's driver
model (i.e., the model used to approximate the driving behavior of a
real driver) is vehicle speed, and outputs are accelerator pedal, brake
pedal, and torque demand.
---------------------------------------------------------------------------
 \438\ NHTSA-2018-0067-12299. Preliminary Regulatory Impact
Analysis (July 2018).
 \439\ NHTSA-2018-0067-0007. Islam, E., S, Moawad, A., Kim, N,
Rousseau, A. ``A Detailed Vehicle Simulation Process To Support CAFE
Standards 04262018--Report'' ANL Autonomie Documentation. Aug 21,
2018. NHTSA-2018-0067-0004. ANL Autonomie Data Dictionary. Aug 21,
2018. NHTSA-2018-0067-0003. ANL Autonomie Summary of Main Component
Assumptions. Aug 21, 2018. NHTSA-2018-0067-0005. ANL Autonomie Model
Assumptions Summary. Aug 21, 2018. NHTSA-2018-0067-1692. ANL BatPac
Model 12 55. Aug 21, 2018.
 \440\ SAFE Rule for MY2021-2026 PRIA Chapter 6.2.3 Technology
groups in Autonomie simulations and CAFE model.
 \441\ PRIA at 189.
 \442\ NHTSA-2018-0067-0007. Islam, E., S, Moawad, A., Kim, N,
Rousseau, A. ``A Detailed Vehicle Simulation Process To Support CAFE
Standards 04262018--Report'' ANL Autonomie Documentation. Aug 21,
2018.
[GRAPHIC] [TIFF OMITTED] TR30AP20.109
 Effectiveness inputs for the NPRM and the final rule analysis were
specifically developed to consider many real world and compliance test
cycle constraints, to the extent a computer model could capture them.
Examples include the advanced engine knock model discussed below, in
addition to other constraints like allowing cylinder deactivation to
occur in ways that would not negatively impact noise-vibration-
harshness (NVH), and similarly optimizing the number of engine on/off
events (e.g., from start/stop 12V micro hybrid systems) to balance
between effectiveness and NVH.
 One major input used in the effectiveness modeling that the
agencies provided key specifications for in the PRIA are engine fuel
maps that define how an engine equipped with specific technologies
operates over a variety of engine load (torque) and engine speed
conditions. The engine maps used as inputs to the Autonomie modeling
portion of the analysis were developed by starting with a base map and
then modifying that base map, incrementally, to model the addition of
engine technologies. These engine maps, developed using the GT-Power
modeling tool by IAV, were based off real-world engine designs.
Simulated operation of these engines included the application of an IAV
knock model, also developed from real-world engine
data.443 444 Using this process, which incorporated real-
world data, ensured that real-world constraints were considered for
each vehicle type. Although the same type of engine map is used for all
technology classes, the effectiveness varies based on the
characteristics of each vehicle type. For example, a compact car with a
turbocharged engine will have different fuel economy and performance
values than a pickup truck with the same engine technology type. The
engine map specifications are discussed further in Section VI.C.1 of
this preamble and Section VI of FRIA.
---------------------------------------------------------------------------
 \443\ Engine knock in spark ignition engines occurs when
combustion of some of the air/fuel mixture in the cylinder does not
result from propagation of the flame front ignited by the spark
plug, but one or more pockets of air/fuel mixture explodes outside
of the envelope of the normal combustion front.
 \444\ See IAV material submitted to the docket; IAV_20190430_Eng
22-26 Updated_Docket.pdf,
IAV_Engine_tech_study_Sept_2016_Docket.pdf, IAV_Study for 4 Cylinder
Gas Engines_Docket.pdf.
---------------------------------------------------------------------------
 The agencies also provided key details about input assumptions for
various vehicle specifications like transmission gear ratios, tire
size, final drive ratios, and individual component weights.\445\ Each
of these assumptions, to some extent, varied between the ten technology
classes to capture appropriately real-world vehicle specifications like
wheel mass or fuel tank mass. These specific input assumptions were
developed based on the latest test data and current market fleet
information.\446\ The agencies relied on default assumptions developed
by the Autonomie team, based on test data and technical publication
review, for other model inputs required by Autonomie, such as throttle
time response and shifting strategies for different transmission
technologies. The Autonomie modeling tool did not simulate vehicle
attributes determined to have minimal impacts, like whether a vehicle
had a sun roof or hood scoops, as those attributes would have trivial
impact in the overall analysis.
---------------------------------------------------------------------------
 \445\ ANL Autonomie Model Assumptions Summary. Aug 21, 2018,
NHTSA-2018-0067-0005. ANL--Summary of Main Component Performance and
Assumptions NPRM. Aug 21, 2018, NHTSA-2018-0067-0003.
 \446\ See further details in Section VI.B.1 Analysis Fleet.
---------------------------------------------------------------------------
 Because the agencies model ten different vehicle types to represent
the 2,952 vehicles in the baseline fleet, improper assumptions about an
advanced technology could lead to errors in estimating effectiveness.
Autonomie is a sophisticated full-vehicle modeling tool that requires
extensive technology characteristics based on both physical and
intangible data, like proprietary software. With a few technologies,
the agencies did not have publicly available data, but had received
confidential business information confirming such technologies
potential availability in the market during the rulemaking time frame.
For such technologies, including advanced cylinder deactivation, the
agencies adopted a method in the CAFE model to represent the
effectiveness of the technology, and did not explicitly simulate the
technologies in the Autonomie model. For this limited set of
technologies, the agencies determined that effectiveness could
reasonably be represented as a fixed value.\447\ Effectiveness values
for technologies not explicitly simulated in Autonomie are discussed
further in the individual technology sections of this preamble.
---------------------------------------------------------------------------
 \447\ For final rule, 9 out of 50 plus technologies use fixed
offset effectiveness values. The total effectiveness of these
technologies cannot be captured on the 2-cycle test or, like ADEAC,
they are a new technology where robust data that could be used as an
input to the technology effectiveness modeling does not yet exist.
Specifically, these nine technologies are LDB, SAX, EPS, IACC, EFR,
ADEAC, DSLI, DSLIAD and TURBOAD.
---------------------------------------------------------------------------
 The agencies sought comments on all effectiveness inputs and input
assumptions, including the specific data used to characterize the
technologies,
[[Page 24322]]
such as data to build the technology input, data representing operating
range of technologies, and data for variation among technology inputs.
The agencies also sought comment on the effectiveness values used for
technologies not explicitly defined in Autonomie.
 Meszler Engineering Services, commenting on behalf of the Natural
Resources Defense Council, and ICCT questioned the accuracy of the
effectiveness estimates in the Argonne database, and as an example
Meszler analyzed the fuel economy impacts of a 10-speed automatic
transmission relative to a baseline 8-speed automatic transmission,
concluding that the widely ranging effectiveness estimates were
unexpected. ICCT questioned the accuracy of the IAV engine maps that
serve as an input to the Autonomie effectiveness modeling, and asked
whether those could ``reasonably stand as a foundation for automotive
developments and technology combinations'' discussed elsewhere in their
comments. ICCT also questioned whether Autonomie realistically and
validly modeled synergies between technologies, using the effectiveness
values from CEGR and transmissions as an example. Meszler stated that
the agencies have an obligation to validate the Autonomie estimates
before using them to support the NPRM or any other rulemaking. The
agencies also received comments on the specific effectiveness estimates
generated by Autonomie; however, those comments will be discussed in
each individual technology section, below.
 Despite these criticisms, Meszler stated that the critiques of the
Autonomie technology database were not meant to imply that the
Autonomie vehicle simulation model used to develop the database was
fundamentally flawed, or that the model could not be used to derive
accurate fuel economy impact estimates. Meszler noted that, as with any
model, estimates derived with Autonomie are only valid for a given set
of modeling parameters and if those parameters are well defined, the
estimates should be accurate and reliable. Conversely, if those
parameters are not well defined, the estimates would be inaccurate and
unreliable. Meszler stated that the agencies must make the full set of
modeling assumptions used for the Autonomie database available for
review and comment.
 We agree with Meszler that, in general, when inputs to a model are
inaccurate, output effectiveness results may be too high or too low.
The technology effectiveness estimates from modeling results often vary
with the type of vehicle and the other technologies that are on that
vehicle.\448\ The Autonomie output database consists of permutations of
over 50 technologies for each of the ten technology classes simulated
by the CAFE model. A wide range of effectiveness is expected when going
from a baseline technology to an advanced technology across different
technology classes because there are significant differences in how
much power is required from the powertrain during 2-cycle testing
across the ten vehicle types. This impacts powertrain operating
conditions (e.g., engine speed and load) during 2-cycle testing. Fuel
economy improving technologies have different effectiveness at each of
those operating conditions so vehicles that have higher average power
demands will have different effectiveness than vehicles with lower
average power demands. Further, the differences in effectiveness at
higher power and lower power vary by technology so the overall
relationship is complex. Large-scale full-vehicle modeling and
simulation account for these interactions and complexities.
---------------------------------------------------------------------------
 \448\ The PRIA Chapter 6.2.2.1, Table 6-2 and Table 6-3 defined
the characteristics of the reference technology classes that
representative of the analysis fleet.
---------------------------------------------------------------------------
 Before conducting any full-vehicle modeling and simulation, the
agencies spent a considerable amount of time and effort developing the
specific inputs used for the Autonomie analysis. The agencies believe
that these technology inputs provide reasonable estimates for the
light-duty vehicle technologies the agencies expect to be available in
the market in the rulemaking timeframe. As discussed earlier, these
inputs vary in effectiveness due to how different vehicles, like
compact cars and pickup trucks, operate on the 2-cycle test and in the
real world. Some technologies, such as 10-speed automatic transmissions
(AT10) relative to 8-speed automatic transmissions (AT8), can and
should have different effectiveness results in the analysis between two
different technology classes.\449\ These unique synergistic effects can
only be taken into account through conducting full-vehicle modeling and
simulation, which the agencies did here.
---------------------------------------------------------------------------
 \449\ Separately, the agencies modified specific transmission
modeling parameters for the final rule after additional review,
including a thorough review of public comments, and this review is
discussed in detail in Section VI.C.2.
---------------------------------------------------------------------------
 With regards to Meszler's comment that the agencies have an
obligation to validate the Autonomie estimates before using them to
support the NPRM or any other rulemaking, the agencies would like to
point Meszler to the description of the Argonne Autonomie team's robust
process for vehicle model validation that was contained in the
PRIA.\450\ To summarize, the NPRM and final rule analysis leveraged
extensive vehicle test data collected by Argonne National
Laboratory.\451\ Over the past 20 years, the Argonne team has developed
specific instrumentation lists and test procedures for collecting
sufficient information to develop and validate full vehicle models. In
addition, the agencies described the Argonne team's efforts to validate
specific component models as well, such as the advanced automatic
transmission and dual clutch transmission models.\452\
---------------------------------------------------------------------------
 \450\ PRIA at 216-7. See also N. Kim, A. Rousseau, E. Rask,
``Autonomie Model Validation with Test Data for 2010 Toyota Prius,''
SAE 2012-01-1040, SAE World Congress, Detroit, Apr12. https://www.autonomie.net/docs/5%20-%20Presentations/Validation/SAE%202012-01-1040.pdf; Vehicle Validation Status, February 2010 https://www.autonomie.net/docs/5%20-%20Presentations/Validation/vehicle_validation_status.pdf; Tahoe HEV Model Development in PSAT,
SAE paper 2009-01-1307, April 2009 https://www.autonomie.net/docs/5%20-%20Presentations/Validation/tahoe_hev.pdf; PHEV Model
Validation, U.S.DOE Merit Review 2008 https://www.autonomie.net/docs/5%20-%20Presentations/Validation/phev_model_validation.pdf ;
PHEV HyMotion Prius model validation and control improvements, 23rd
International Electric Vehicle Symposium (EVS23), Dec. 2007 https://www.autonomie.net/docs/5%20-%20Presentations/Validation/phev_hymotion_prius.pdf; Integrating Data, Performing Quality
Assurance, and Validating the Vehicle Model for the 2004 Prius Using
PSAT, SAE paper 2006-01-0667, April 2006; https://www.autonomie.net/docs/5%20-%20Presentations/Validation/integrating_data.pdf.
 \451\ A list of the vehicles that have been tested at the APRF
can be found under http://www.anl.gov/energy-systems/group/downloadable-dynamometer-database.
 \452\ Kim, N., Rousseau, N., Lohse-Bush, H. ``Advanced Automatic
Transmission Model Validation Using Dynamometer Test Data,'' SAE
2014-01-1778, SAE World Congress, Detroit, April 2014; Kim, N.,
Lohse-Bush, H., Rousseau, A. ``Development of a model of the dual
clutch transmission in Autonomie and validation with dynamometer
test data,'' International Journal of Automotive Technologies, March
2014, Volume 15, Issue 2, pp 263-71.
---------------------------------------------------------------------------
 The agencies also described the process for validating inputs used
to develop the IAV engine maps,453 454 another input to the
Autonomie simulations. As discussed in the PRIA, IAV's engine model
development relied on a collection of sub-models that controlled
independent combustion characteristics such as heat release, combustion
knock, friction, heat flow, and other combustion optimization tools.
These sub-models and other
[[Page 24323]]
computational fluid dynamics models were utilized to convert test data
for use in the IAV engine map development. Specific combustion
parameters, like from test data for the coefficient of variation for
the indicated mean effective pressure (COV of IMEP), which is a common
variable for combustion stability in a spark ignited engine, was used
to assure final engine models were reasonable. The assumptions and
inputs used in the modeling and validation of engine model results
leveraged IAV's global engine database, which included benchmarking
data, engine test data, single cylinder test data and prior modeling
studies, and also technical publications and information presented at
conferences. The agencies referenced in the PRIA that engine maps were
validated with engine dynamometer test data to the maximum extent
possible.\455\ Because the NPRM and the final rule analysis considered
some technologies not yet in production, the agencies relied on
technical publications and engine modeling by IAV to develop and
corroborate inputs and input assumptions where engine dynamometer test
data was not available.
---------------------------------------------------------------------------
 \453\ See PRIA at 251.
 \454\ See IAV material submitted to the docket; IAV_20190430_Eng
22-26 Updated_Docket.pdf,
IAV_Engine_tech_study_Sept_2016_Docket.pdf, IAV_Study for 4 Cylinder
Gas Engines_Docket.pdf.
 \455\ See PRIA at 288.
---------------------------------------------------------------------------
 In addition, as described earlier in this section, the full set of
NPRM modeling assumptions used for the Autonomie database were
available for review and comment in the docket for this
rulemaking.\456\ The full set of modeling assumptions used for the
final rule are also available in the docket.\457\
---------------------------------------------------------------------------
 \456\ NHTSA-2018-0067-0007. Islam, E., S, Moawad, A., Kim, N,
Rousseau, A., ``A Detailed Vehicle Simulation Process To Support
CAFE Standards 04262018--Report'' ANL Autonomie Documentation. Aug
21, 2018. NHTSA-2018-0067-0004. ANL Autonomie Data Dictionary. Aug
21, 2018. NHTSA-2018-0067-0003. ANL Autonomie Summary of Main
Component Assumptions. Aug 21, 2018. NHTSA-2018-0067-0005. ANL
Autonomie Model Assumptions Summary. Aug 21, 2018. NHTSA-2018-0067-
1692. ANL BatPac Model 12 55. Aug 21, 2018. Preliminary Regulatory
Impact Analysis (July 2018). Posted July 2018 and updated August 23
and October 16, 2018.
 \457\ The CAFE Model is available at https://www.nhtsa.gov/corporate-average-fuel-economy/compliance-and-effects-modeling-system with documentation and all inputs and outputs supporting
today's notice.
---------------------------------------------------------------------------
 Both ICCT and Meszler also commented on the availability of
technologies within the Autonomie database, with Meszler stating that
with limited exceptions, technologies were not included in the NPRM
CAFE model if they were not included in the simulation modeling that
underlay the Argonne database, and accordingly if a combination of
technologies was not modeled during the development of the Argonne
database, that package (or combination) of technologies was not
available for adoption in the CAFE model. Meszler stated that these
constraints limited the slate of technologies available to respond to
fuel economy standards, and independently expanding the model to
include additional technologies or technology combinations is not
trivial.
 ICCT gave specific examples of key efficiency technologies that it
stated Autonomie did not include, like advanced DEAC, VCR, Miller
Cycle, e-boost, and HCCI. ICCT argued that this was especially
problematic as the agencies appeared to have available engine maps from
IAV on advanced DEAC, VCR, Miller Cycle, E-boost (and from advanced
DEAC, VCR, Miller Cycle, E-boost, HCCI from EPA) that Argonne or the
agencies have been unable to or opted not to include in their modeling.
ICCT stated that the agencies must disclose how Autonomie had been
updated to incorporate ``cutting edge'' 2020-2025 automotive
technologies to ensure they reflect available improvements.\458\
---------------------------------------------------------------------------
 \458\ ICCT also made the same request of EPA's ALPHA model, and
the agencies' response to that comment is discussed in Section
VI.C.1 Engine Paths, below.
---------------------------------------------------------------------------
 The agencies have updated the final rule analysis to include
additional technologies. In the NPRM, the agencies presented the engine
maps for all of the technologies that ICCT listed, except HCCI, and
sought comment on the engine maps, technical assumptions and the
potential use of the technologies for the final rule analysis. Based on
the available technical information and the ICCT and Meszler comments,
for the final rule analysis, VCR, Miller Cycle (VTG), and e-boost (VTGe
with 48V BISG) technologies have been added and included in the
Autonomie modeling and simulations, and advanced DEAC technology has
been added using fixed point effectiveness estimates in the CAFE model
analysis. The agencies disagree with ICCT's assessment of HCCI and do
not believe it will be available for wide-scale application in the
rulemaking timeframe, and therefore have not included it as a
technology. HCCI technology has been in the research phase for several
decades, and the only production applications to date use a highly-
limited version that restricts HCCI combustion to a very narrow range
of engine operating conditions.459 460 461 Additional
discussion of how Autonomie-modeled and non-modeled technologies are
incorporated into the CAFE Model is located in Section VI.B.3.c),
below.
---------------------------------------------------------------------------
 \459\ Mazda introduced Skyactiv-X in Europe with a mild hybrid
technology to assist the engine.
 \460\ Mazda News. ``Revolutionary Mazda Skyactiv-X engine
details confirmed as sales start,'' May 6, 2019. https://www.mazda-press.com/eu/news/2019/revolutionary-mazda-skyactiv-x-engine-details-confirmed-as-sales-start/. Last accessed Dec. 2, 2019.
 \461\ Confer. K. Kirwan, J. ``Ultra Efficient Light-Duty
Powertrain with Gasoline Low-Temperature Combustion.'' DOE Merit
Review. June 9, 2017. https://www.energy.gov/sites/prod/files/2017/06/f34/acs094_confer_2017_o.pdf. Last accessed Dec. 2, 2019.
---------------------------------------------------------------------------
 ICCT and Meszler also commented that the agencies overly limited
the availability of several technologies in the NPRM analysis. In
response, the agencies reconsidered the restrictions that were applied
in the NPRM analysis, and agree with the commenters for several
technologies and technology classes. Many technologies identified by
the commenters are now in production for the MY2017 as well as MY2018
and MY2019. The agencies also think that the baseline fleet compliance
data reflects adoption of many of these technologies. For the final
rule analysis, the agencies have expanded the availability of several
technologies. In the CAFE model, the agencies are now allowing parallel
hybrids (SHEVP2) to be adopted with high compression Atkinson mode
engines (HCR0 and HCR1). In addition, as mentioned above, the Autonomie
full-vehicle modeling included Variable Compression Ratio engine (VCR),
Miller Cycle Engine (VTG), E-boost (VTGe) technologies, and cylinder
deactivation technologies (DEAC) to be applied to turbocharged engines
(TURBO1). As these changes relate to the technology effectiveness
modeling, the CAFE model analysis now includes effectiveness estimates
based on full vehicle simulations for all of these technology
combinations.
 We disagree with comments stating the agencies should allow every
technology to be available to every vehicle class.\462\ Discussed
earlier in this section, Autonomie models key aspects of vehicle
operation that are most relevant to assessing fuel economy, vehicle
performance and certain aspects of drivability (like EPA 2-cycle tests,
EPA US06 cycle tests, gradability, low speed acceleration time from 0-
to-60 mph, passing acceleration time from 50 to 80 mph, and number of
transmission shifts). However, there are other critical aspects of
vehicle functionality and operation that the agencies considered beyond
those criteria, that cannot necessarily be reflected in the Autonomie
modeling. For example, a pickup truck can be modeled with a
[[Page 24324]]
continuously variable transmission (CVT) and show improvements on the
2-cycle tests. However, pickup trucks are designed to provide high load
towing utility.\463\ CVTs lack the torque levels needed to provide that
towing utility, and would fail mechanically if subject to high load
towing.\464\ The agencies provided discussions of some of these
technical considerations in the PRIA, and explained why the agencies
had limited technologies for certain vehicle classes, such as limiting
CVTs on pickups as in the example above. These and other limitations
are discussed further in the individual technology sections.
---------------------------------------------------------------------------
 \462\ NHTSA-2018-0067-11723. NRDC Attachment2 at p. 4.
 \463\ SAE J2807. ``Performance Requirements for Determining Tow-
Vehicle Gross Combination Weight Rating and Trailer Weight Rating.''
Feb. 4, 2016.
 \464\ PRIA at p. 223 and 340.
---------------------------------------------------------------------------
 The agencies also received a variety of comments that conflated
aspects of the Autonomie models with technology inputs and input
assumptions. For example, commenters expressed concern about the
transmission gear set and final drive values used for the NPRM
analysis, or more specifically, that the gear ratios were held constant
across applications.\465\ In this case, both the inputs (gear set and
final drive ratio) and input assumption (ratios held constant) were
discussed by the commenters. Because these comments are actually about
technology inputs to the Autonomie model, for these and similar cases,
the agencies are addressing the comments in the individual technology
sections which discuss the technology inputs and input assumptions that
impact the effectiveness values for those technologies.
---------------------------------------------------------------------------
 \465\ NHTSA-2018-0067-11873. Comments from Roush Industries,
Attachment 1, at p. 14-15. NHTSA-2018-0067-11873. Comments from
CARB, at p.110.
---------------------------------------------------------------------------
 For the NPRM analysis, the agencies prioritized using inputs that
were based on data for identifiable technology configurations and that
reflected practical real world constraints. The agencies provided
detailed information on the NPRM analysis inputs and input assumptions
in the NPRM Preamble, PRIA and Argonne model documentation for engine
technologies, transmission technologies, powertrain electrification,
light-weighting, aerodynamic improvements, tire rolling resistance
improvements, and other vehicle technologies. Comments and the
agencies' assessment of comments for each technology are discussed in
the individual technology sections below. Through careful consideration
of the comments, the agencies have updated analytical inputs associated
with several technologies, and as discussed above, have included
several advanced technologies for which technical information was
included in the NPRM. However, for most technologies, the agencies have
determined that the technology inputs and input assumptions that were
used in the NPRM analysis remain reasonable and the best available for
the final rule analysis.
(2) How The Agencies Defined Different Vehicle Types in Autonomie
 As described in the NPRM, Argonne produced full-vehicle models and
ran simulations for many combinations of technologies, on many types of
vehicles, but it did not simulate literally every single vehicle model/
configuration in the analysis fleet because it would be impractical to
assemble the requisite detailed information--much of which would likely
only be provided on a confidential basis--specific to each vehicle
model/configuration and because the scale of the simulation effort
would correspondingly increase by orders of magnitude. Instead, Argonne
simulated 10 different vehicle types, corresponding to the five
``technology classes'' generally used in CAFE analysis over the past
several rulemakings, each with two performance levels and corresponding
vehicle technical specifications (e.g., small car, small performance
car, pickup truck, performance pickup truck, etc.).
 Technology classes are a means of specifying common technology
input assumptions for vehicles that share similar characteristics.
Because each vehicle technology class has unique characteristics, the
effectiveness of technologies and combinations of technologies is
different for each technology class. Conducting Autonomie simulations
uniquely for each technology class provides a specific set of
simulations and effectiveness data for each technology class. Like the
Draft TAR analysis, there are separate technology classes for compact
cars, midsize cars, small SUVs, large SUVs, and pickup trucks. However,
new for the NPRM analysis and carried into this final rule analysis,
each of those vehicle types has been split into ``low'' (or
``standard'') performance and a ``high'' performance versions, which
represent two classes with similar body styles but different levels of
performance attributes (for a total of 10 technology classes). The
separate technology classes for high performance and low performance
vehicles better account for performance diversity across the fleet.
 NHTSA directed Argonne to develop a vehicle assumptions database to
capture vehicle attributes that would comprise the full vehicle models.
For each vehicle technology class, representative vehicle attributes
and characteristics were identified from publicly available information
and automotive benchmarking databases like A2Mac1,\466\ Argonne's
Downloadable Dynamometer Database (D\3\),\467\ and EPA compliance and
fuel economy data,\468\ EPA's guidance on the cold start penalty on 2-
cycle tests.\469\ The resulting vehicle assumptions database consists
of over 100 different attributes like vehicle frontal area, drag
coefficient, fuel tank weight, transmission housing weight,
transmission clutch weight, hybrid vehicle component weights, and
weights for components that comprise engines and electric machines,
tire rolling resistance, transmission gear ratios and final drive
ratio. Each of the 10 different vehicle types was assigned a set of
these baseline attributes and characteristics, to which combinations of
fuel-saving technologies were added as inputs for the Autonomie
simulations. For example, the characteristics of the MY 2016 Honda Fit
were considered along with a wide range of other compact cars to
identify representative characteristics for the Autonomie simulations
for the base compact car technology class. The simulations determined
the fuel economy achieved when applying each combination of
technologies to that vehicle type, given its baseline characteristics.
---------------------------------------------------------------------------
 \466\ A2Mac1: Automotive Benchmarking. (Proprietary data).
Retrieved from https://a2mac1.com.
 \467\ Downloadable Dynamometer Database (D\3\). ANL Energy
Systems Division. https://www.anl.gov/es/downloadable-dynamometer-database. Last accessed Oct. 31, 2019.
 \468\ Data on Cars used for Testing Fuel Economy. EPA Compliance
and Fuel Economy Data. https://www.epa.gov/compliance-and-fuel-economy-data/data-cars-used-testing-fuel-economy. Last accessed Oct.
31, 2019.
 \469\ EPA PD TSD at p.2-265--2-266.
---------------------------------------------------------------------------
 For each vehicle technology class and for each vehicle attribute,
Argonne estimated the attribute value using statistical distribution
analysis of publicly available data and data obtained from the A2Mac1
benchmarking database.\470\ Some
[[Page 24325]]
vehicle attributes were also based on test data and vehicle
benchmarking, like the cold-start penalty for the FTP test cycle and
vehicle electrical accessories load. The analysis of vehicle attributes
used in the NPRM was discussed in the Argonne model documentation,\471\
and values for each vehicle technology class were provided with the
NPRM for public review.\472\
---------------------------------------------------------------------------
 \470\ A2Mac1 is subscription-based benchmarking service that
conducts vehicle and component teardown analyses. Annually, A2Mac1
removes individual components from production vehicles such as oil
pans, electric machines, engines, transmissions, among the many
other components. These components are weighed and documented for
key specifications which is then available to their subscribers.
 \471\ NHTSA-2018-0067-0007, at 131. Islam, E., S, Moawad, A.,
Kim, N, Rousseau, A., ``A Detailed Vehicle Simulation Process To
Support CAFE Standards 04262018--Report'' ANL Autonomie
Documentation. Aug 21, 2018.
 \472\ NHTSA-2018-0067-0003. ANL Autonomie Summary of Main
Component Assumptions. Aug 21, 2018.
---------------------------------------------------------------------------
 The agencies did not believe it was appropriate to assign one
single engine mass for each vehicle technology class in the NPRM
analysis. To account for the difference in weight for different engine
types, Argonne performed a regression analysis of engine peak power
versus weight, based on attribute data taken from the A2Mac1
benchmarking database. For example, to account for weight of different
engine sizes like 4-cylinder versus 8-cylinder, Argonne developed a
relationship curve between peak power and engine weight based on the
A2Mac1 benchmarking data. For the NPRM analysis, this relationship was
used to estimate mass for all engine types regardless of technology
type (e.g., variable valve lift and direct injection). Secondary weight
reduction associated with changes in engine technology was applied by
using this linear relationship between engine power and engine weight
from the A2Mac1 benchmarking database. When a vehicle in the analysis
fleet with an 8-cylinder engine adopted a more fuel efficient 6-
cylinder engine, the total vehicle weight would reflect the updated
engine weight with two less cylinders based on the peak power versus
engine weight relationship. The impact of engine mass reduction on
effectiveness is accounted for directly in the Autonomie simulation
data through the application of the above relationship. Engine mass
reduction through downsizing is, therefore, appropriately not included
as part of vehicle mass reduction technology that is discussed in
Section VI.C.4 because doing so would result in double counting the
impacts. As discussed further below, for the final rule the agencies
improved upon the precision of engine weights by creating two curves to
separately represent naturally aspirated engine designs and
turbocharged engine designs.
 In addition, certain attributes were held at constant levels within
each technology class to maintain vehicle functionality, performance
and utility including noise, vibration, and harshness (NVH), safety,
performance and other utilities important for customer satisfaction.
For example, in addition to the vehicle performance constraints
discussed in Section VI.B.3.a)(6), the analysis does not allow the
frontal area of the vehicle to change, in order to maintain utility
like ground clearance, head-room space, and cargo space, and a cold-
start penalty is used to account for fuel economy degradation for
heater performance and emissions system catalyst light-off.\473\ This
allows us to capture the discrete improvement in technology
effectiveness while maintaining vehicle attributes that are important
vehicle utility, consumer acceptance and compliance with criteria
emission standards, and considering these constraints similar to how
manufacturers do in the real world.
---------------------------------------------------------------------------
 \473\ The catalyst light-off is the temperature necessary to
initiate the catalytic reaction and this energy is generated from
engine.
---------------------------------------------------------------------------
 The agencies sought comment on the analytical approach used to
determine vehicle attributes and characteristics for the Autonomie
modeling. In response, the agencies received a wide variety of comments
on vehicle attributes ranging from discussions of performance increase
from technology adoption (e.g., if a vehicle adopting an electrified
powertrain improved its time to accelerate from 0-60 mph), to comments
on vehicle attributes not modeled in Autonomie, like heated seats and
cargo space.
 Toyota and the Alliance commented that the inclusion of performance
vehicle classes addressed the market reality that some consumers will
purchase vehicles for their performance attributes and will accept the
corresponding reduction in fuel economy. Furthermore, Toyota commented
that some gain in performance is more realistic, and that ``dedicating
all powertrain improvements to fuel efficiency is inconsistent with
market reality.'' Toyota ``supports the agencies' inclusion of
performance classes in compliance modeling where a subset of certain
models is defined to have higher performance and a commensurate
reduction in fuel efficiency.'' \474\ Also, in support of the addition
of performance vehicle classes, the Alliance commented that ``vehicle
categories have been increased to 10 to better recognize the range of
0-60 performance characteristics within each of the 5 previous
categories, in recognition of the fact that many vehicles in the
baseline fleet significantly exceeded the previously assumed 0-60
performance metrics. This provides better resolution of the baseline
fleet and more accurate estimates of the benefits of technology.''
\475\
---------------------------------------------------------------------------
 \474\ Toyota, Attachment 1, Docket No. NHTSA-2018-0067-12098, at
p. 6.
 \475\ Alliance of Automobile Manufacturers, Attachment ``Full
Comment Set,'' Docket No. NHTSA-2018-0067-12073, at p.135.
---------------------------------------------------------------------------
 UCS commented that the CAFE model incorporates technology
improvements to each vehicle by applying the effectiveness improvement
of the average vehicle in the technology class, leading to discrete
``stepped'' effectiveness levels for technologies across the different
vehicle types. UCS stated that in contrast, the OMEGA model takes into
account a vehicle's performance characteristics through response-
surface modeling based on relative deviation from the class average
modeled in ALPHA.\476\
---------------------------------------------------------------------------
 \476\ NHTSA-2018-0067-12039, at p.24.
---------------------------------------------------------------------------
 Although differences between the ALPHA and Autonomie models are
discussed in more detail below, for the NPRM vehicle simulation
analysis the agencies expanded the number of vehicle classes from the
five classes used in the Draft TAR to ten classes, to represent better
the diversity of vehicle characteristics across the fleet. Each of
these ten vehicle technology classes are empirically built from
benchmarking data and other information from various sources, amounting
to hundreds of vehicle characteristics data points to develop each
vehicle class. The agencies expand on these vehicle classes and
characteristics in Section VI.B.3.(a)(2) Vehicle Types in Autonomie and
Section VI.B.3.(a)(3) How Vehicle Models are Built in Autonomie and
Optimized for Simulation. The agencies believe that the real-world data
used to define vehicle characteristics for each of the ten vehicle
classes, in addition to the ten vehicle technology classes themselves,
ensures the analysis reasonably accounts for the diversity in vehicle
characteristics across the fleet.
 The agencies believe that UCS's characterization of how technology
improvements are applied in the analysis is a misleading
oversimplification. While the analysis approach in the final rule uses
a representative effectiveness value, the value is not linked solely to
the vehicle technology class, as the UCS implies. The entire technology
combination, or technology key, which includes the vehicle technology
class, is used to
[[Page 24326]]
determine the value for the platform being considered. Within each
vehicle class, the interactions between the added technology and the
full vehicle system (including other technologies and substantial road
load characteristics) are considered in the effectiveness values
calculated for each technology during compliance modeling. As discussed
under each of the technology pathways sections, the effectiveness for
most technologies is reported as a range rather than a single value.
The range exists because the effectiveness for each technology is
adjusted based on the technologies it is coupled with and the major
road load characteristics of the full vehicle system. This approach, in
combination with using the baseline vehicle's initial performance
values as a starting point for performance improvement, results in a
widely variable level of improvement for the system, dependent on
individual vehicle platform characteristics. As a result, the
application of a response-surface approach would likely result in
minimal improvement in accuracy for the Autonomie and CAFE model
analysis approach.
 For the final rule analysis, the agencies used the same process to
obtain the vehicle attributes and characteristics for the vehicle
technology classes. Data was acquired from publicly available sources,
Argonne D\3\, EPA compliance and fuel economy data, and A2mac1
benchmarking data. Accordingly, the attributes and characteristics of
the modeled vehicles reflect actual vehicles that meet customer
expectations and automakers' capabilities to manufacture the vehicles.
In addition, for the final rule, the agencies improved the NPRM
analysis by updating some of the attribute values to account for
changes in the fleet. For example, the agencies have updated vehicle
electrical accessory load on the test cycle to reflect higher
electrical loads associated with contemporary vehicle features.
(3) How This Rulemaking Builds Vehicle Models for Autonomie and
Optimize Them for Simulation
 Before any simulation is initiated in Autonomie, Argonne must
``build'' a vehicle by assigning reference technologies and initial
attributes to the components of the vehicle model representing each
technology class.\477\ The reference technologies are baseline
technologies that represent the first step on each technology pathway
used in the analysis. For example, a compact car is built by assigning
it a baseline engine, a baseline 6-speed automatic transmission (AT6),
a baseline level of aerodynamic improvement (AERO0), a baseline level
of rolling resistance improvement (ROLL0), a baseline level of mass
reduction technology (MR0), and corresponding attributes from the
Argonne vehicle assumptions database like individual component
weights.\478\ A baseline vehicle will have a unique starting point for
the simulation and a unique set of assigned inputs and attributes,
based on its technology class.
---------------------------------------------------------------------------
 \477\ For the NPRM analysis, Chapter 8 Vehicle-Sizing Process in
the ANL Model Documentation had discussed this process in detail.
Further discussion of this process is located in Chapter 8 of the
ANL Model Documentation for this final rule.
 \478\ See Section VI.A.7.
---------------------------------------------------------------------------
 The next step in the process is to run a powertrain sizing
algorithm that ensures the built vehicle meets or exceeds defined
performance metrics, including low-speed acceleration (i.e., time
required to accelerate from 0-60 mph), high-speed passing acceleration
(time required to accelerate from 50-80 mph), gradeability (e.g. the
ability of the vehicle to maintain constant 65 miles per hour speed on
a six percent upgrade), and towing capacity. Together, these
performance criteria are widely used by industry as metrics to quantify
vehicle performance attributes that consumers observe and that are
important for vehicle utility and customer satisfaction.
 In the compact car example used above, the agencies assigned an
initial specific engine design and engine power, transmission, AERO,
ROLL, and MR technologies, and other attributes like vehicle weight. If
the built vehicle does not meet all the performance criteria in the
first iteration, then the engine power is increased to meet the
performance requirement. This increase in power is from higher engine
displacement, which could involve an increase in number of cylinders,
leading to an increase in the engine weight. The iterative process
continues to check whether the compact car with updated engine power,
and corresponding updated engine weight, meets its defined performance
metrics. The loop stops once all the metrics are met, and at this
point, a compact car technology class vehicle model becomes ready for
simulation. For further discussion of the vehicle performance metrics,
see Section VI.B.3.(a).
 Autonomie then adopts a single fuel saving technology to the
baseline vehicle model, keeping everything else the same except for
that one technology and the attributes associated with it. For example,
the model would apply an 8-speed automatic transmission in place of the
baseline 6-speed automatic transmission, which would lead to either an
increase or decrease in the total weight of the vehicle based on the
technology class assumptions. At this point, Autonomie confirms whether
performance metrics are met for this new vehicle model through the
previously discussed sizing algorithm. Once a technology has been
assigned to the vehicle model and the resulting vehicle meets its
performance metrics, those vehicle models will be used as inputs to the
full vehicle simulations. So, in the example of the 6-speed to 8-speed
automatic transmission technology update, the agencies now have the
initial ten vehicle models (one for each technology class), plus the
ten new vehicle models with the updated 8-speed automatic transmission,
which adds up to 20 different vehicle models for simulation. This
permutation process is conducted for each of the over 50 technologies
considered, and for all ten technology classes, which results in more
than one million optimized vehicle models.
 Figure VI-3 shows the process for building vehicles in Autonomie
for simulation.
[[Page 24327]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.110
 Some of the technologies require extra steps for optimization
before the vehicle models are built for simulation; for example, the
sizing and optimization process is more complex for the electrified
vehicles (i.e., HEVs, PHEVs) compared to vehicles with internal
combustion engines, as discussed further, below. Throughout the vehicle
building process, the following items are considered for optimization:
 Vehicle weight is decreased or increased in response to
switching from one type of technology to another for the technologies
for which the agencies consider weight, such as different engine and
transmission types;
 Vehicle performance is decreased or increased in response
to the addition of mass reduction technologies when switching from one
vehicle model to another vehicle model for the same engine;
 Vehicle performance is decreased or increased in response
to the addition of a new technology when switching from one vehicle
model to another vehicle model for the same hybrid electric machine;
and
 Electric vehicle battery size is decreased or increased in
response to the addition of mass, aero and/or tire rolling resistance
technologies when switching from one vehicle model to another vehicle
model.
 Every time a vehicle adopts a new technology, the vehicle weight is
updated to reflect the new component weight. For some technologies, the
direct weight change is easy to assess. For example, in the NPRM the
agencies designated weights for transmissions so, when a vehicle is
updated to a higher geared transmission, the weight of the original
transmission is replaced with the corresponding transmission weight
(e.g., the weight of a vehicle moving from a 5-speed automatic
transmission to an 8-speed automatic transmission will be updated based
on the 8-speed transmission weight).
 For other technologies, like engine technologies, assessing the
updated vehicle weight is much more complex. Discussed earlier,
modeling a change in engine technology involves both the new technology
adoption and a change in power (because the reduction in vehicle weight
leads to lower engine loads, and a resized engine). When a new engine
technology is adopted on a vehicle the agencies account for the
associated weight change to the vehicle based on the earlier discussed
regression analysis of weight versus power. For the NPRM engine weight
regression analysis, the agencies considered 19 different engine
technologies that consisted of unique components to achieve fuel
economy improvements. This regression analysis is technology agnostic
by taking the approach of using engine peak power versus engine weight
because it removed biases to any specific engine technology in the
analysis. Although the agencies do not estimate the specific weight for
each individual engine technology, such as VVT and SGDI, this process
provides a reasonable estimate of the weight differences among engine
technologies.
[[Page 24328]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.111
 For the final rule analysis, the agencies used the same process to
assign initial weights to the original 19 engines, plus the added
engines. However, the agencies improved upon precision of the weights
by creating two separate curves separately to represent naturally
aspirated engine designs and turbocharged engine designs.\479\ This
update resulted in two benefits. First, small naturally aspirated 4-
cylinder engines that adopted turbocharging technology reflected the
increased weight of associated components like ducting, clamps, the
turbocharger itself, a charged air cooler, wiring, fasteners, and a
modified exhaust manifold. Second, larger cylinder count engines like
naturally aspirated 8-cylinder and 6-cylinder engines that adopted
turbocharging and downsized technologies would have lower weight due to
having fewer engine cylinders. For example, a naturally aspirated 8-
cylinder engine that adopts turbocharging technology when downsized to
a 6-cylinder turbocharged engine appropriately reflects the added
weight of turbocharging components, and the lower weight of fewer
cylinders.
---------------------------------------------------------------------------
 \479\ ANL Model Documentation for the final rule analysis,
Chapter 5.2.9 Engine Weight Determination.
---------------------------------------------------------------------------
 As with conventional vehicle models, electrified vehicle models
were built from the ground up. For the NPRM analysis, Argonne used data
from the A2mac1 database and vehicle test data to define different
attributes like weights and power. Argonne used one electric motor
specific power for each type of hybrid and electric vehicle.\480\ For
MY2017, the U.S. market has an expanded number of available hybrid and
electric vehicle models. To capture appropriately the improvements for
electrified vehicles for the final rule analysis, the agencies applied
the same regression analysis process that considers electric motor
weight versus electric motor power for vehicle models that have adopted
electric motors. Benchmarking data for hybrid and electric vehicles
from the A2Mac1 database was analyzed to develop a regression curve of
electric motor peak power versus electric motor weight.\481\
---------------------------------------------------------------------------
 \480\ NHTSA-2018-0067-0005. ANL Autonomie Model Assumptions
Summary. Aug 21, 2018. Non_Vehicle_Attributes tab. Specific power
for PS and P2 HEVs was set to 2750 watts/kg, plug-in HEVs were set
to 375 watts/kg, and electric vehicles were set to 1400 watts/kg.
 \481\ ANL Model Documentation for the final rule analysis,
Chapter 5.2.10 Electric Machines System Weight.
---------------------------------------------------------------------------
(4) How Autonomie Sizes Powertrains for Full Vehicle Simulation
 The agencies maintain performance neutrality of the full vehicle
simulation analysis by resizing engines, electric machines, and hybrid
electric vehicle battery packs at specific incremental technology
steps. To address product complexity and economies of scale, engine
resizing is limited to specific incremental technology changes that
would typically be associated with a major vehicle or engine
redesign.\482\ Manufacturers have repeatedly told the agencies that the
high costs for redesign and the increased manufacturing complexity that
would result from resizing engines for small technology changes
preclude them from doing so. It would be unreasonable and unaffordable
to resize powertrains for every unique combination of technologies, and
exceedingly so for every unique combination of technologies across
every vehicle model due to the extreme manufacturing complexity that
would be required to do so. The agencies reiterated in the NPRM that
the analysis should not include engine resizing with the application of
every technology or for combinations of technologies that drive small
performance changes so that the analysis better reflects what is
feasible for manufacturers.\483\
---------------------------------------------------------------------------
 \482\ See 83 FR 43027 (Aug. 24, 2018).
 \483\ For instance, a vehicle would not get a modestly bigger
engine if the vehicle comes with floor mats, nor would the vehicle
get a modestly smaller engine without floor mats. This example
demonstrates small levels of mass reduction. If manufacturers
resized engines for small changes, manufacturers would have
dramatically more part complexity, potentially losing economies of
scale.
---------------------------------------------------------------------------
 When a powertrain does need to be resized, Autonomie attempts to
mimic manufacturers' development approaches to the extent possible.
Discussed earlier, the Autonomie vehicle building process is initiated
by building a baseline vehicle model with a baseline engine,
transmission, and other baseline vehicle technologies. This baseline
vehicle model (for each technology class) is sized to meet a specific
set of
[[Page 24329]]
performance criteria, including acceleration and gradeability.
 The modeling also accounts for the industry practice of platform,
engine, and transmission sharing to manage component complexity and the
associated costs.\484\ At a vehicle refresh cycle, a vehicle may
inherit an already resized powertrain from another vehicle within the
same engine-sharing platform that adopted the powertrain in an earlier
model year. In the Autonomie modeling, when a new vehicle adopts fuel
saving technologies that are inherited, the engine is not resized (the
properties from the baseline reference vehicle are used directly and
unchanged) and there may be a small change in vehicle performance. For
example, in Figure VI-3, Vehicle 2 inherits Eng01 from Vehicle 1 while
updating the transmission. Inheritance of the engine with new
transmission may change performance. This example illustrates how
manufacturers generally manage manufacturing complexity for engines,
transmissions, and electrification technologies.
---------------------------------------------------------------------------
 \484\ Ford EcoBoost Engines are shared across ten different
models in MY2019. https://www.ford.com/powertrains/ecoboost/. Last
accessed Nov. 05, 2019.
---------------------------------------------------------------------------
 Autonomie implements different powertrain sizing algorithms
depending on the type of powertrain being considered because different
types of powertrains contain different components that must be
optimized.\485\ For example, the conventional powertrain resizing
considers the reference power of the conventional engine (e.g., Eng01,
a basic VVT engine, is rated at 108 kilowatts and this is the starting
reference power for all technology classes) against the power-split
hybrid (SHEVPS) resizing algorithm that must separately optimize engine
power, battery size (energy and power), and electric motor power. An
engine's reference power rating can either increase or decrease
depending on the architecture, vehicle technology class, and whether it
includes other advanced technologies.
---------------------------------------------------------------------------
 \485\ ANL Model Documentation for the final rule Analysis,
Chapter 8.3.1 Conventional-Vehicle Sizing Algorithm; Chapter 8.3.2
Split-HEV Sizing Algorithm; 8.3.4 Blended PHEV sizing Algorithm;
8.3.5 Voltec PHEV (Extended Range) Vehicle Sizing Algorithm; Chapter
8.3.6 BEV Sizing Algorithm.
---------------------------------------------------------------------------
 Performance requirements also differ depending on the type of
powertrain because vehicles with different powertrain types may need to
meet different criteria. For example, a plug-in hybrid electric vehicle
(PHEV) powertrain that is capable of traveling a certain number of
miles on its battery energy alone (referred to as all-electric range,
or AER, or as performing in electric-only mode) is also sized to ensure
that it can meet the performance requirements of a US06 cycle in
electric-only mode.
 The powertrain sizing algorithm is an iterative process that
attempts to optimize individual powertrain components at each step. For
example, the sizing algorithm for conventional powertrains estimates
required power to meet gradeability and acceleration performance and
compares it to the reference engine power for the technology class. If
the power required to meet gradeability and acceleration performance
exceeds the reference engine power, the engine power is updated to the
new value. Similarly, if the reference engine power exceeds the
gradeability and acceleration performance power, it will be decreased
to the lower power rating. As the change in power requires a change
design of the engine, like increasing displacement (e.g., going from a
5.2-liter to 5.6-liter engine, or vice versa) or increasing cylinder
count (e.g., going from an I4 to a V6 or vice versa), the engine weight
will also change. The new engine power is used to update the weight of
the engine.
 Next, the conventional powertrain sizing algorithm enters an
acceleration algorithm loop to verify low-speed acceleration
performance (time it takes to go from 0 mph to 60 mph). In this step,
Autonomie adjusts engine power to maintain a performance attribute for
the given technology class and updates engine weight accordingly. Once
the performance criteria are met, Autonomie ends the low-speed
acceleration performance algorithm loop and enters a high-speed
acceleration (time it takes to go from 50 mph to 80 mph) algorithm
loop. Again, Autonomie might need to adjust engine power to maintain a
performance attribute for the given technology, and it exits this loop
once the performance criteria have been met. At this point, the sizing
algorithm is complete for the conventional powertrain based on the
designation for engine type, transmissions type, aero type, mass
reduction technology and low rolling resistance technology.
 Figure VI-5 below shows the sizing algorithm for conventional
powertrains.
[[Page 24330]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.112
 Depending on the type of powertrain considered, the sizing
algorithms may also size to meet different performance criteria in
different order. The powertrain sizing algorithms for electrified
vehicles are considerably more complex, and are discussed in further
detail in Section VI.C.3, below.
(5) How the Agencies Considered Maintaining Vehicle Attributes
 For this rulemaking analysis, consistent with past CAFE and
CO2 rulemakings, the agencies have analyzed technology
pathways manufacturers could use for compliance that attempt to
maintain vehicle attributes, utility, and performance. Using this
approach allows the agencies to assess costs and benefits of potential
standards under a scenario where consumers continue to get the similar
vehicle attributes and features, other than changes in fuel economy.
The purpose of constraining vehicle attributes is to simplify the
analysis and reduce variance in other attributes that consumers value
across the analyzed regulatory alternatives. This allows for a more
streamlined accounting of costs and benefits by not requiring the
values of other vehicle attributes that trade off with fuel economy.
 Several examples of vehicle attributes, utility and performance
that could be impacted by adoption of fuel economy improving technology
include the following.
[[Page 24331]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.113
 Consequences for the agencies not fully considering or accounting
for potential changes in vehicle attributes, utility, and performance
are degradation in vehicle attributes, utility, and performance that
lead to consumer acceptance issues without accounting for the
corresponding costs and/or not accounting for the costs of technology
designs that maintain vehicle attributes, utility, and performance. The
agencies incorporated changes in the NPRM analysis and that are carried
into this final rule that address deficiencies in past analyses,
including the Draft TAR and Proposed Determination analyses. These
changes were discussed in the NPRM and are repeated in the discussion
of individual technologies in this Preamble, the FRIA, and supporting
documents. The following are several examples of technologies that did
not maintain vehicle attributes, utility, and performance in the Draft
TAR and Proposed Determination analyses.
 For the EPA Draft TAR and Proposed Determination analyses, HCR
engine and downsized and turbocharged engine technologies effectiveness
was estimated using Tier 2 certification fuel, which has a higher
octane rating compared to regular octane fuel.486 487 This
does not maintain functionality because consumers would incur higher
costs for using premium fuel in order to achieve the modeled fuel
economy improvements, compared to baseline engines that were replaced,
which operated on lower cost regular octane fuel. By not maintaining
the fuel octane functionality and vehicle attributes, the EPA Draft TAR
and Proposed Determination analyses applied higher effectiveness for
these technologies than could be achieved had regular octane fuel been
assumed for the HCR and downsized turbocharged engines. The Draft TAR
and Proposed Determination analyses also did not account for the higher
costs that would be incurred by consumers to pay for high octane fuel.
These issues were addressed in the
[[Page 24332]]
NPRM and this final rule analysis, and account for some of the
effectiveness and cost differences between the Draft TAR/Proposed
Determination and the NPRM/final rule.\488\
---------------------------------------------------------------------------
 \486\ Tier 2 fuel has an octane rating of 93. Typical regular
grade fuel has an octane rating of 87 ((R+M)/2 octane.
 \487\ EPA Proposed Determination at 2-209 to 2-212.
 \488\ For more details, see Section VI.C.1 Engine Paths.
---------------------------------------------------------------------------
 Another example is mass reduction technology. As background, the
agencies characterize mass reduction as either primary mass reduction
or secondary mass reduction. Primary mass reduction involves reducing
mass of components that can be done independently of the mass of other
components. For example, the mass of a hood (e.g., replacing a steel
hood with an aluminum hood) or reducing the mass of a seat are examples
of primary mass reduction because each can be implemented
independently. When there is a significant level of primary mass
reduction, other components that are designed based on the mass of
primary components, may be redesigned and have lower mass. An example
of secondary mass reduction is the brake system. If the mass of primary
components is reduced sufficiently, the resulting lighter weight
vehicle could maintain braking performance and attributes, and safety
with a lighter weight brake system. Mass reduction in the brake system
is secondary mass reduction because it requires primary mass reduction
before it can be incorporated. For the EPA Draft TAR and Proposed
Determination analyses, secondary mass reduction was applied
exclusively based on cost, with no regard to whether sufficient primary
mass reduction was applied concurrently. The analyses did not account
for the degraded functionality of the secondary components and systems
and also understated the costs for lower levels of mass reduction.\489\
These issues were addressed in the NPRM and this final rule analysis,
and account for some of the cost differences between the Draft TAR/
Proposed Determination and the NPRM/final rule.
---------------------------------------------------------------------------
 \489\ For more details, see Section VI.C.4 Mass Reduction.
---------------------------------------------------------------------------
 The agencies note that for some technologies it is not reasonable
or practicable to match exactly the baseline vehicle's attributes,
utility, and performance. For example, when engines are resized to
maintain acceleration performance, if the agencies applied a criterion
that allowed no shift in performance whatsoever, there would be an
extreme proliferation of unique engine displacements. Manufacturers
have repeatedly and consistently told the agencies that the high costs
for redesign and the increased manufacturing complexity that would
result from resizing engines for small technology changes preclude them
from doing so. It would be unreasonable and unaffordable to resize
powertrains for every unique combination of technologies, and
exceedingly so for every unique combination technologies across every
vehicle model due to the extreme manufacturing complexity that would be
required to do so.\490\ For the NPRM and final rule analyses, engine
resizing is limited to specific incremental technology changes that
would typically be associated with a major vehicle or engine redesign
to address product complexity and economies of scale considerations.
The EPA Draft TAR and Proposed Determination analyses adjusted the
effectiveness of every technology combination assuming performance
could be held constant for every combination, and the analysis did not
recognize or account for the extreme complexity nor the associated
costs for that impractical assumption. The NPRM and final rule analyses
account for these real-world practicalities and constraints, and doing
so explains some of the effectiveness and cost differences between the
Draft TAR/Proposed Determination and the NPRM/final rule.
---------------------------------------------------------------------------
 \490\ For more details, see Section VI.B.3.a)(6) Performance
Neutrality.
---------------------------------------------------------------------------
 The subsections for individual technologies discuss the technology
assumptions and constraints that were considered to maintain vehicle
attributes, utility, and performance as closely as possible. The
agencies believe that any minimal remaining differences, which may
directionally either improve or degrade vehicle attributes, utility and
performance are small enough to have de minimis impact on the analysis.
(6) How the Agencies Considered Performance Neutrality
 The CAFE model examines technologies that can improve fuel economy
and reduce CO2 emissions. An improvement in efficiency can
be realized by improving the powertrain that propels the vehicle (e.g.,
replacing a 6-cylinder engine with a smaller, turbocharged 4-cylinder
engine), or by reducing the vehicle's loads or burdens (e.g., lowering
aerodynamic drag, reducing vehicle mass and/or rolling resistance).
Either way, these changes reduce energy consumption and create a range
of choices for automobile manufacturers. At the two ends of the range,
the manufacturer can choose either:
 (A) To design a vehicle that does same the amount of work as before
but uses less fuel.
 For example, a redesigned pickup truck would receive a turbocharged
V6 engine in place of the outgoing V8. The pickup would offer no
additional towing capacity, acceleration, larger wheels and tires,
expanded infotainment packages, or customer convenience features, but
would achieve a higher fuel economy rating (and correspondingly lower
CO2 emissions).
 (B) To design a vehicle that does more work and uses the same
amount of fuel as before.
 For example, a redesigned pickup truck would receive a turbocharged
V6 engine in place of the outgoing V8, but with engine efficiency
improvements that allow the same amount of fuel to do more work. The
pickup would offer improved towing capacity, improved acceleration,
larger wheels and tires, an expanded (heavier) infotainment package,
and more convenience features, while maintaining (not improving) the
fuel economy rating of the previous year's model.
 In other words, automakers weigh the trade-offs between vehicle
performance/utility and fuel economy, and they choose a blend of these
attributes to balance meeting fuel economy and emissions standards and
suiting the demands of their customers.
 Historically, vehicle performance has improved over the years. The
average horsepower is the highest that it has ever been; all vehicle
types have improved horsepower by at least 49 percent compared to the
1975 model year, and pickup trucks have improved by 141 percent.\491\
Since 1978, the 0-60 acceleration time of vehicles has improved by 39-
47 percent depending on vehicle type.\492\ Also, to gain consumer
acceptance of downsized turbocharged engines, manufacturers have stated
they often offer an increase in performance.\493\ Fuel economy has also
improved, but the horsepower and acceleration trends show that not 100
percent of technological improvements have been applied to fuel
savings. While future trends are uncertain, the past trends suggest
vehicle performance is unlikely to decrease, as it seems reasonable to
assume that customers
[[Page 24333]]
will at a minimum demand vehicles that offer the same utility as
today's fleet.
---------------------------------------------------------------------------
 \491\ The 2018 EPA Automotive Trends Report (EPA-420-R-19-002
March 2019) https://www.epa.gov/automotive-trends/download-automotive-trends-report.
 \492\ The 2018 EPA Automotive Trends Report (EPA-420-R-19-002
March 2019) https://www.epa.gov/automotive-trends/download-automotive-trends-report.
 \493\ Alliance of Automobile Manufacturers, Attachment
``Comment,'' Docket No. EPA-HQ-OAR-2015-0827-4089, at p. 122.
---------------------------------------------------------------------------
 For this rulemaking analysis, consistent with past CAFE and
CO2 rulemakings, the agencies have analyzed technology
pathways manufacturers could use for compliance that attempt to
maintain vehicle attributes, utility and performance. NHTSA's analysis
in the Draft TAR used the same approach for performance neutrality as
was used for the NPRM and is being carried into this final rule. This
approach is described throughout this section and further in FRIA
Section VI. For the Draft TAR and Proposed Determination, the EPA
analyses used an approach that maintained 0-60 mph acceleration time
for every technology package. However, that approach did not account
for the added development, manufacturing, assembly and service parts
complexity and associated costs that would be incurred by manufacturers
to produce the substantial number of engine variants that would be
required to achieve those CO2 improvements.\494\ Using the
NPRM approach, which is carried into this final rule, allows the
agencies to assess costs and benefits of potential standards under a
scenario where consumers continue to get the same vehicle attributes
and features, other than changes in fuel economy (approaching the
scenario in example ``A'' above). This approach also eliminates the
need to assess the value of changes in vehicle attributes and features.
As discussed later in this section, while some small level of
performance increase is unavoidable when conducting this type of
analysis, the added technology results almost exclusively in improved
fuel economy. This allows the cost of these technologies to reflect
almost entirely the cost of compliance with standards with nearly
neutral vehicle performance.
---------------------------------------------------------------------------
 \494\ Each variant would require a unique engine displacement,
requiring unique internal engine components, such as crankshaft,
connecting rods and others.
---------------------------------------------------------------------------
 The CAFE model maintains the initial performance and utility levels
of the analysis vehicle fleet, while considering real world constraints
faced by manufacturers.
 To maintain performance neutrality when applying fuel economy
technologies, it is first necessary to characterize the performance
levels of each of the nearly 3000 vehicle models in the MY 2017
baseline fleet. As discussed in Section VI.B.1.b) Assigning Vehicle
Technology Classes, above, each individual vehicle model in the
analysis fleet was assigned to one of ten vehicle ``technology
classes''--the class that is most similar to the vehicle model. The
technology classes include five standard class vehicles (compact car,
midsize car, small SUV, midsize SUV, pickup) plus five ``performance''
versions of these same body styles.\495\ Each vehicle class has a
unique set of attributes and characteristics, including vehicle
performance metrics, that describe the typical characteristics of the
vehicles in that class.
---------------------------------------------------------------------------
 \495\ Separate technology classes were created for high
performance and low performance vehicles to better account for
performance diversity across the fleet.
---------------------------------------------------------------------------
 The analysis used four criteria to characterize vehicle performance
attributes and utility:
 Low-speed acceleration (time required to accelerate from 0-60
mph)
 High-speed acceleration (time required to accelerate from 50-
80 mph)
 Gradeability (the ability of the vehicle to maintain constant
65 miles per hour speed on a six percent upgrade)
 Towing capacity
 Low-speed and high-speed acceleration target times are typical of
current production vehicles and range from 6 to 10 seconds depending on
the vehicle class; for example, the midsize SUV performance class has a
low- and high-speed acceleration target of 7 seconds.\496\ The
gradeability criterion requires that the vehicle, given its attributes
of weight, engine power, and transmission gearing, be capable of
maintaining a minimum of 65 mph while going up a six percent grade. The
towing criterion, which is applicable only to the pickup truck and
performance pickup truck vehicle technology classes, is the same as the
gradeability requirement but adds an additional payload/towing mass
(3,000 lbs. for pickups, or 4,350 lbs for performance pickups) to the
vehicle, essentially making the vehicle heavier.
---------------------------------------------------------------------------
 \496\ Note, for all vehicle classes, the low and high-speed
acceleration targets use the same value. See section VI.B.1.b)(1)
Assigning Vehicle Technology Classes for a list of low-speed
acceleration target by vehicle technology class.
---------------------------------------------------------------------------
 In addition, to maintain the capabilities of certain electrified
vehicles in the 2017 baseline fleet, the analysis required that those
vehicles be capable of achieving the accelerations and speeds of
certain standard driving cycles. The agencies use the US06 ``aggressive
driving'' cycle and the UDDS ``city driving'' cycle to ensure that core
capabilities of BEVs and PHEVs, such as driving certain speeds and/or
distances in electric-only mode, are maintained. In addition to the
four criteria discussed above, the following performance criteria are
applied to these electrified vehicles:
 Battery electric vehicles (BEV) are sized to be capable of
completing the US06 ``aggressive driving'' cycle.
 Plug-in hybrid vehicles with 50 mile all-electric range
(PHEV50) are sized to be capable of completing the US06 ``aggressive
driving'' cycle in electric-only mode.
 Plug-in hybrid vehicles with 20 mile all-electric range
(PHEV20) are sized to be capable of completing the UDDS ``city
driving'' cycle in electric-only (charge depleting) mode.\497\
---------------------------------------------------------------------------
 \497\ PHEV20's are blended-type plug-in hybrid vehicles, which
are capable of completing the UDDS cycle in charge depleting mode
without assistance from the engine. However, under higher loads,
this charge depleting mode may use supplemental power from the
engine.
---------------------------------------------------------------------------
 Together, these performance criteria are widely used by industry as
metrics to quantify vehicle performance attributes that consumers
observe and that are important for vehicle utility and customer
satisfaction.\498\
---------------------------------------------------------------------------
 \498\ Conlon, B., Blohm, T., Harpster, M., Holmes, A. et al.,
``The Next Generation ``Voltec'' Extended Range EV Propulsion
System,'' SAE Int. J. Alt. Power. 4(2):2015, doi:10.4271/2015-01-
1152. Kapadia, J., Kok, D., Jennings, M., Kuang, M., et al.,
``Powersplit or Parallel--Selecting the Right Hybrid Architecture,''
SAE Int. J. Alt. Power. 6(1):2017, doi:10.4271/2017-01-1154. Islam,
E., A. Moawad, N. Kim, and A. Rousseau, 2018a, An Extensive Study on
Vehicle Sizing, Energy Consumption and Cost of Advance Vehicle
Technologies, Report No. ANL/ESD-17/17, Argonne National Laboratory,
Lemont, Ill., Oct 2018.
---------------------------------------------------------------------------
 When certain fuel-saving technologies are applied that affect
vehicle performance to a significant extent, such as replacing a pickup
truck's V8 engine with a turbocharged V6 engine, iterative resizing of
the vehicle powertrain (engine, electric motors, and/or battery) is
performed in the Autonomie simulation such that the above performance
criteria is maintained. For example, if the aforementioned engine
replacement caused an improvement in acceleration, the engine may be
iteratively resized until vehicle acceleration performance is shifted
back to the initial target time for that vehicle technology class. For
the low and high-speed acceleration criteria, engine resizing
iterations continued until the acceleration time was within plus or
minus 0.2 seconds of the target time,499 500 which is judged
to balance
[[Page 24334]]
reasonably the precision of engine resizing with the number of
simulation iterations needed to achieve performance within the 0.2
second window, and the associated computer resources and time required
to perform the iterative simulations. Engine resizing is explained
further in Section VI.B.3.a)(4) How Autonomie Sizes Powertrains for
Full Vehicle Simulation and the Argonne Model Documentation for the
final rule analysis.
---------------------------------------------------------------------------
 \499\ For example, if a vehicle has a target 0-60 acceleration
time of 6 seconds, a time within 5.8-6.2 seconds was accepted.
 \500\ With the exception of a few performance electrified
vehicle types which, based on observations in the marketplace, use
different criteria to maintain vehicle performance without battery
assist. Performance PHEV20, and Performance PHEV50 resize to the
performance of a conventional six-speed automatic (CONV 6AU).
Performance SHEVP2, engines/electric-motors were resized if the 0-60
acceleration time was worse than the target, but not resized if the
acceleration time was better than the target time.
---------------------------------------------------------------------------
 The Autonomie simulation resizes until the least capable of the
performance criteria is met, to ensure the pathways do not degrade any
of the vehicle performance metrics. It is possible that as one
criterion target is reached after the application of a specific
technology or technology package, other criteria may be better than
their target values. For example, if the engine size is decreased until
the low speed acceleration target is just met, it is possible that the
resulting engine size would cause high speed acceleration performance
to be better than its target.\501\ Or, a PHEV50 may have an electric
motor and battery appropriately sized to operate in all electric mode
through the repeated accelerations and high speeds in the US06 driving
cycle, but the resulting motor and battery size enables the PHEV50
slightly to over-perform in 0-60 acceleration, which utilizes the power
of both the electric motor and combustion engine.
---------------------------------------------------------------------------
 \501\ The Autonomie simulation databases include all of the
estimated performance metrics for each combination of technology as
modeled.
---------------------------------------------------------------------------
 To address product complexity and economies of scale, engine
resizing is limited to specific incremental technology changes that
would typically be associated with a major vehicle or engine
redesign.\502\ Manufacturers have repeatedly and consistently told the
agencies that the high costs for redesign and the increased
manufacturing complexity that would result from resizing engines for
small technology changes preclude them from doing so. It would be
unreasonable and unaffordable to resize powertrains for every unique
combination of technologies, and exceedingly so for every unique
combination technologies across every vehicle model due to the extreme
manufacturing complexity that would be required to do so. Engine
displacements are further described in Section VI.C.1 Engine Paths.
---------------------------------------------------------------------------
 \502\ See 83 FR 43027 (Aug. 24, 2018).
---------------------------------------------------------------------------
 To address this issue, and consistent with past rulemakings, the
NPRM simulation allowed engine resizing when mass reductions of 7.1
percent, 10.7 percent, 14.2 percent (and 20 percent for the final rule
analysis) were applied to the vehicle curb weight,\503\ and when one
powertrain architecture was replaced with another architecture during a
redesign cycle.\504\ At its refresh cycle, a vehicle may also inherit
an already resized powertrain from another vehicle within the same
engine-sharing platform. The analysis did not re-size the engine in
response to adding technologies that have smaller effects on vehicle
performance. For instance, if a vehicle's curb weight is reduced by 3.6
percent (MR1), causing the 0-60 mile per hour time to improve slightly,
the analysis would not resize the engine. The criteria for resizing
used for the analysis better reflects what is feasible for
manufacturers to do.\505\
---------------------------------------------------------------------------
 \503\ These correspond, respectively, to reductions of 10%, 15%,
20%, and 28.2% of the vehicle glider mass. For more detail on glider
mass calculation, see section VI.C.4 Mass Reduction.
 \504\ Some engine and accessory technologies may be added to an
engine without an engine architecture change. For instance,
manufacturers may adapt, but not replace engine architectures to
include cylinder deactivation, variable valve lift, belt-integrated
starter generators, and other basic technologies. However, switching
from a naturally aspirated engine to a turbo-downsized engine is an
engine architecture change typically associated with a major
redesign and radical change in engine displacement.
 \505\ For instance, a vehicle would not get a modestly bigger
engine if the vehicle comes with floor mats, nor would the vehicle
get a modestly smaller engine without floor mats. This example
demonstrates small levels of mass reduction. If manufacturers
resized engines for small changes, manufacturers would have
dramatically more part complexity, potentially losing economies of
scale.
---------------------------------------------------------------------------
 Automotive manufacturers have commented that the CAFE model's
consideration of the constraints faced in relation to vehicle
performance and economies of scale are realistic.
 Industry associations and individual manufacturers widely supported
the use of the performance metrics used in the NPRM analysis, the use
of standard and higher performance technology classes, and the
representation in the analysis of the real-world manufacturing
complexity constraints and criteria for powertrain redesign.
 The Alliance of Automobile Manufacturers (Alliance), Ford, and
Toyota stated that the inclusion of additional performance metrics such
as gradeability are appropriate. Specifically in support of the
gradeability performance criteria, the Alliance commented that
``performance metrics related to vehicle operation in top gear are just
as critical to customer acceptance as are performance metrics such as
0-60 mph times that focus on performance in low-gear ranges.'' \506\
The Alliance also commented specifically on the relationship between
gradeability and downsized engines, stating that as ``engine downsizing
levels increase, top-gear gradeability becomes more and more
important,'' and further that the consideration of gradeability ``helps
prevent the inclusion of small displacement engines that are not
commercially viable and that would artificially inflate fuel savings.''
\507\
---------------------------------------------------------------------------
 \506\ Alliance of Automobile Manufacturers, Attachment ``Full
Comment Set,'' Docket No. NHTSA-2018-0067-12073, at 139.
 \507\ Alliance of Automobile Manufacturers, Attachment ``Full
Comment Set,'' Docket No. NHTSA-2018-0067-12073, at 135.
---------------------------------------------------------------------------
 Ford and Toyota similarly commented in support of the CAFE model's
consideration of multiple performance criteria. Ford stated that this
model ``takes a more realistic approach to performance modeling'' and
``better replicates OEM attribute-balancing practices.'' Ford stated
furthermore that ``OEMs must ensure that each individual performance
measure--and not an overall average--meets its customer's
requirements,'' and that, in contrast, previous analyses did ``not
align with product planning realities.'' \508\ Toyota commented in
support of including gradeability as a performance metric ``to avoid
underpowered engines and overestimated fuel savings.'' \509\
---------------------------------------------------------------------------
 \508\ Ford, Attachment 1, Docket No. NHTSA-2018-0067-11928, at
8.
 \509\ Toyota, Attachment 1, Docket No. NHTSA-2018-0067-12098, at
6.
---------------------------------------------------------------------------
 Toyota and the Alliance commented that the inclusion of performance
vehicle classes addressed the market reality that some consumers will
purchase vehicles for their performance attributes and will accept the
corresponding reduction in fuel economy. Furthermore, Toyota commented
that most consumers consider more than just fuel economy when
purchasing a vehicle, and that ``dedicating all powertrain improvements
to fuel efficiency is inconsistent with market reality.'' Toyota
``supports the agencies' inclusion of performance classes in compliance
modeling where a subset of certain models is defined to have higher
performance and a commensurate reduction in fuel efficiency.'' \510\
Also in support of the addition of performance vehicle classes, the
Alliance commented that ``vehicle categories have been increased to 10
to better recognize the range of 0-60 performance
[[Page 24335]]
characteristics within each of the 5 previous categories, in
recognition of the fact that many vehicles in the baseline fleet
significantly exceeded the previously assumed 0-60 performance metrics.
This provides better resolution of the baseline fleet and more accurate
estimates of the benefits of technology.'' \511\
---------------------------------------------------------------------------
 \510\ Toyota, Attachment 1, Docket No. NHTSA-2018-0067-12098, at
6.
 \511\ Alliance of Automobile Manufacturers, Attachment ``Full
Comment Set,'' Docket No. NHTSA-2018-0067-12073, at 135.
---------------------------------------------------------------------------
 Toyota also commented in support of various real-world
manufacturing complexity constraints employed in the analysis for
powertrain redesigns. Toyota commented that model parameters such as
redesign cycles and engine sharing across vehicle models place a more
realistic limit on the number of engines and transmissions that a
manufacturer is capable of introducing. Toyota also commented in
support of the constraints that the CAFE model placed on engine
resizing, stating that ``there are now more realistic limits placed on
the number of engines and transmissions in a powertrain portfolio which
better recognizes [how] manufacturers must manage limited engineering
resources and control supplier, production, and service costs.
Technology sharing and inheritance between vehicle models tends to
limit the rate of improvement in a manufacturer's fleet.'' Toyota
pointed out that this is in contrast to previous analyses in which
resizing was too unconstrained, which created an ``unmanageable number
of engine configurations within a vehicle platform'' and spawned cases
where ``engine downsizing and power reduction sometimes exceeded limits
beyond basic acceleration requirements needed for vehicle safety and
customer satisfaction.'' \512\
---------------------------------------------------------------------------
 \512\ Toyota, Attachment 1, Docket No. NHTSA-2018-0067-12098, at
6.
---------------------------------------------------------------------------
 The above comments from the Alliance, Ford, and Toyota support the
methodologies the agencies employed to conduct a performance neutral
analysis. These methodologies helped to ensure that multiple
performance criteria, including gradeability, are all individually
accounted for and maintained when a vehicle powertrain is resized, and
that real-world manufacturing complexity constraints are factored in to
the agencies' analysis of feasible pathways manufacturers could take to
achieve compliance with CAFE standards. The agencies continue to
believe this is a reasonable approach for the aforementioned reasons.
 Environmental advocacy groups and CARB criticized the CAFE model's
engine resizing constraints and how they affected the acceleration
performance criteria.
 CARB, The International Council on Clean Transportation (ICCT), the
Union of Concerned Scientists (UCS), and the American Council for an
Energy-Efficient Economy (ACEEE) commented that the CAFE model was not
performance neutral, allowing an improvement in performance which
reduced the effectiveness of applied fuel-saving technologies and/or
increased the cost of compliance. Specifically, ACEEE stated that there
appeared to be a shortfall in the fuel economy effectiveness of
technology packages, potentially resulting from the effectiveness being
``consumed'' by additional vehicle performance rather than improvement
of fuel economy. Several of these same commenters conducted analyses
attempting to quantify the magnitude of these changes in vehicle
performance for various vehicle technology classes.
 CARB commented on the performance shift of several vehicle types.
Analyzing the 0-60 acceleration for the medium car non-performance
technology class and looking at all cases with resized engines, CARB
claimed that ``effectively half of the simulations resulted in improved
performance.'' \513\ Focusing on electrified vehicles in that same
technology class, CARB stated that ``the data from the Argonne
simulations shows that 76 of the 88 strong electrified packages
(including P2HPV, SHEVPS, BEV, FCEV, PHEV), where Argonne purposely
resized the system to maintain performance neutrality, resulted in
notably faster 0 to 60 mph acceleration times and passing times.''
Specifically regarding parallel hybrid electric vehicles (SHEVP2), CARB
stated that all modeled packages resulted in improved performance.\514\
UCS commented that the NPRM analysis allowed too much change in vehicle
performance, stating that ``while some performance creep may be
reasonable'' many performance values show ``an overlap between
performance and non-performance vehicles'' within the compact car
technology class.\515\
---------------------------------------------------------------------------
 \513\ California Air Resources Board, Attachment 2, Docket No.
NHTSA-2018-0067-11873, at 180. Note that the target acceleration
time for medium car non-performance is in fact 9.0 seconds, as
indicated in ANL documentation, but was incorrectly reported as 9.4s
in NPRM table II-7 in the NPRM.
 \514\ California Air Resources Board, Attachment 2, Docket No.
NHTSA-2018-0067-11873, at 186.
 \515\ Union of Concerned Scientists, Attachment 2, Docket No.
NHTSA-2018-0067-12039, at 24.
---------------------------------------------------------------------------
 The agencies carefully considered these comments. For the NPRM
analysis, the SHEVP2 engines/electric-motors were resized if the 0-60
acceleration time was worse than the target, but not resized if the
acceleration time was better than the target. This approach maintained
vehicle performance with a depleted battery (without electric assist)
in order to maintain fully the performance and utility characteristics
under all conditions, and improved performance when electric assist was
available (when the battery is not depleted), such as during the 0-60
mph acceleration. The agencies found that this resulted in some
parallel hybrid vehicles having improved 0-60 acceleration times. This
approach was initially chosen for the NPRM because the resulting level
of improved performance was consistent with observations of how
industry had applied SHEVP2 technology. However, in assessing the CARB
comment, the agencies balanced the NPRM approach for SHEVP2 performance
with the agencies' criteria of maintaining vehicle functionality and
performance when technology is applied. Both could not be fully
achieved under all conditions for the case of the SHEVP2.
 The agencies concluded it is reasonable to maintain performance
including electric assist when SHEVP2 technology is applied to a
standard (non-performance) vehicle, and therefore the analysis for the
final rule allows upsizing and downsizing of the parallel hybrid
powertrain (SHEVP2) using the 0.2 seconds window around the
target.\516\ For performance vehicles, the agencies concluded that it
remains reasonable to maintain vehicle performance with a depleted
battery (without electric assist) in order to maintain fully the
performance characteristics under all conditions, and continued to use
the NPRM methodology.
---------------------------------------------------------------------------
 \516\ To represent marketplace trends better, the performance
class of SHEVP2's allow acceleration time below 0.2 seconds less
than the target, and PHEV20's and PHEV50's inherit combustion engine
size from the conventional powertrain they are replacing. Further
discussion of resizing targets can be found in Chapter 8 of the ANL
Model Documentation for the final rule analysis.
---------------------------------------------------------------------------
 The refinement for the standard performance SHEVP2 resolved the
electrified packages issue identified by CARB, and also addressed most
of the change in performance in the overall fleet, including with
compact cars as mentioned by UCS. As explained further below, the
agencies assessed performance among the alternatives for the final rule
analysis. That assessment showed that, with the final rule refinements,
245 out of 255 total resized vehicles (96 percent of vehicles) in the
medium non-performance class (same
[[Page 24336]]
class focused on by CARB), had 0-60 mph acceleration times within the
plus-or-minus 0.2 second window (8.8 to 9.2 seconds).\517\ The only
vehicles outside the window were certain strong electrified vehicles
which exceeded 0-60 the acceleration target as a result of achieving
other performance criteria, such as the US06 driving cycles in all-
electric-mode.\518\
---------------------------------------------------------------------------
 \517\ This includes 135 strong electrified vehicles.
 \518\ As noted earlier, electrified vehicles had to be capable
of successfully completing UDDS or US06 driving cycles in all-
electric mode, and in some cases the resulting motor size produced
improved acceleration times.
---------------------------------------------------------------------------
 The assessment also showed that for the small car class (mentioned
by UCS) the acceleration times of performance and non-performance
vehicles do not go beyond each other's targets. For example, the
vehicle in the small car class with the very best 0-60 mph time and a
conventional powertrain achieves an 8.38 second 0-60 mph time, which is
slower than the performance small car baseline of 8 seconds. This
vehicle had multiple incremental technologies applied, including for
example aerodynamic improvements, and has not reached the threshold for
engine resizing.\519\ After engine resizing, the ``fastest''
conventional small car has a 0-60 mph time of 9.9 seconds, only 0.1
seconds from the target of 10 seconds.\520\
---------------------------------------------------------------------------
 \519\ Discussion of engine resizing can be found in Section
VI.B.3.a)(5).
 \520\ See NPRM Autonomie simulation database for Small cars,
Docket ID NHTSA-2018-0067-1855.
---------------------------------------------------------------------------
 CARB also commented on the improvement of ``passing times,'' or 50-
80 mph high-speed acceleration times. As stated above, an improvement
in one or more of the performance criteria is an expected outcome when
using the rulemaking analysis methodology that resizes powertrains such
that there is no degradation in any of the performance metrics.
Consistent with past rulemakings, the agencies do not believe it is
appropriate for the rulemaking analysis to show pathways that degrade
vehicle performance or utility for one or more of the performance
criteria, as doing so would adversely impact functional capability of
the vehicle and could lead to customer dissatisfaction. The agencies
agree there is very small increase in passing performance for some
technology combinations, and believe this is an appropriate outcome.
High-speed acceleration is rarely the least-capable performance
criteria.
 CARB, ICCT, UCS, and H-D Systems (HDS), in an attempt to identify a
potential cause for changes in performance, commented that the CAFE
model should have placed fewer constraints on engine resizing. CARB and
ICCT commented that engine resizing should have been allowed even at
low levels of mass reduction. Comments from CARB, UCS, HDS, and ICCT
stated that engine resizing should also have been allowed for other
incremental technologies, and within their comments they conducted
performance analysis of non-resized cases.
 CARB claimed that requiring a minimum of 7.1 percent curb weight
reduction before engine resizing is a constraint that ``limits the
optimization of the technologies being applied.'' \521\ UCS stated that
``a significant share of the benefit of a few percent reduction in mass
has gone towards improved performance rather than improved fuel
economy, leaving a substantial benefit of mass reduction underutilized
and/or uncounted.'' \522\ ICCT also commented that ``when vehicle
lightweighting is deployed at up to a 7 percent mass reduction, the
engine is not resized even though less power would be needed for the
lighter vehicle, meaning any such vehicles inherently are higher
performance.'' \523\
---------------------------------------------------------------------------
 \521\ California Air Resources Board, Attachment 2, Docket No.
NHTSA-2018-0067-11873, at 178. Note, a 7.1% curb weight reduction
equates to the agencies' third level of mass reduction (MR3);
additional discussion of engine resizing for mass reduction can be
found in Section VI.B.3.a)(4) Autonomie Sizes Powertrains for Full
Vehicle Simulation] and in the ANL Model Documentation for the final
rule analysis.
 \522\ Union of Concerned Scientists, Attachment 2, Docket No.
NHTSA-2018-0067-12039, at 11.
 \523\ International Council on Clean Transportation, Attachment
3, Docket No. NHTSA-2018-0067-11741, at I-50.
---------------------------------------------------------------------------
 UCS and HDS commented on the lack of resizing for technologies
other than mass reduction, with HDS stating that ``the Agencies
incorrectly limited the efficacy of technologies that reduce tractive
load because their modeling does not re-optimize engine performance
after applying these technologies.'' \524\ CARB also commented that the
lack of resizing when a BISG or CISG system is added ``results in a
less than optimized system that does not take full advantage of the
mild hybrid system.'' Similarly, ICCT noted a case in which a Dodge RAM
``did not apply engine downsizing with the BISG system on that truck,
so there are also significant performance benefits that should be
accounted for, meaning that for constant-performance the fuel
consumption reduction would be even greater.'' \525\
---------------------------------------------------------------------------
 \524\ H-D Systems, Attachment 1, Docket No. NHTSA-2018-0067-
12395, at 4. For reference, technologies that reduce tractive road
load include mass reduction, aerodynamic drag reduction, and tire
rolling resistance reduction.
 \525\ International Council on Clean Transportation, Attachment
3, Docket No. NHTSA-2018-0067-11741, at I-24.
---------------------------------------------------------------------------
 CARB further commented on the performance improvement in cases
without engine resizing by stating that ``94 percent of the packages
modeled result in improved performance,'' and that for these non-
resized cases that were actually adopted by a vehicle in the
simulation, ``fewer than 20 percent maintained baseline performance
with gains of 2 percent or less in acceleration time.'' \526\ Referring
specifically to non-resized electrified vehicles, CARB also stated that
``44,878 of the 53,818 packages, or greater than 83 percent, result in
improved performance.'' \527\ CARB also commented that engine sharing
across different vehicles within a platform, which in some cases may
constrain resizing for a member of that platform, should not dictate
that these engines must remain identical in all aspects, and that
``this overly restrictive sharing of identical engines newly imposed in
the CAFE Model is not consistent with today's industry practices and
results in less optimal engine sizing and causes a systematic
overestimation of technology costs to meet the existing standards.''
\528\
---------------------------------------------------------------------------
 \526\ California Air Resources Board, Attachment 2, Docket No.
NHTSA-2018-0067-11873, at 183.
 \527\ California Air Resources Board, Attachment 2, Docket No.
NHTSA-2018-0067-11873, at 187.
 \528\ California Air Resources Board, Attachment 2, Docket No.
NHTSA-2018-0067-11873, at 185.
---------------------------------------------------------------------------
 The agencies note broadly, in response to these comments, that when
conducting an analysis which balances performance neutrality against
the realities faced by manufacturers, such as manufacturing complexity,
economies of scale, and maintaining the full range of performance
criteria, it is inevitable to observe at least some minor shift in
vehicle performance. For example, if a new transmission is applied to a
vehicle, the greater number of gear ratios helps the engine run in its
most efficient range which improves fuel economy, but also helps the
engine to run in the optimal ``power band'' which improves performance.
Thus, the technology can provide both improved fuel economy and
performance. Another example is applying a small amount of mass
reduction that improves both fuel economy and performance by a small
amount. Resizing the engine to maintain performance in these examples
would require a unique engine displacement that is only slightly
different than the baseline engine. While engine resizing in these
incremental cases could have some small benefit to fuel economy, the
[[Page 24337]]
gains may not justify the costs of producing unique niche engines for
each combination of technologies. If manufacturers were to produce
marginally downsized engines to complement every small increment of
mass reduction or technology, the resulting large number of engine
variants that would need to be manufactured would cause a substantial
increase in manufacturing complexity, and require significant changes
to manufacturing and assembly plants and equipment.\529\ The high costs
would be economically infeasible.
---------------------------------------------------------------------------
 \529\ For example, each unique engine would require unique
internal components such as crankshafts, pistons, and connecting
rods, as well as unique engine calibrations for each displacement.
Assembly plants would need to stock and feed additional unique
engines to the stations where engines are dressed and inserted into
vehicles.
---------------------------------------------------------------------------
 Also, as noted in the NPRM, the 2015 NAS report stated that ``[f]or
small (under 5 percent [of curb weight]) changes in mass, resizing the
engine may not be justified, but as the reduction in mass increases
(greater than 10 percent [of curb weight]), it becomes more important
for certain vehicles to resize the engine and seek secondary mass
reduction opportunities.'' \530\ In consideration of both the NAS
report and comments received from manufacturers, the agencies
determined it would be reasonable to allow allows engine resizing upon
adoption of 7.1 percent, 10.7 percent, 14.2 percent, and 20 percent
curb weight reduction, but not at 3.6 percent and 5.3 percent.\531\
Resizing is also allowed upon changes in powertrain type or the
inheritance of a powertrain from another vehicle in the same platform.
The increments of these higher levels of mass reduction, or complete
powertrain changes, more appropriately match the typical engine
displacement increments that are available in a manufacturer's engine
portfolio.
---------------------------------------------------------------------------
 \530\ National Research Council. 2011. Assessment of Fuel
Economy Technologies for Light-Duty Vehicles. Washington, DC--The
National Academies Press. http://nap.edu/12924.
 \531\ These curb weight reductions equate to the following
levels of mass reduction as defined in the analysis: MR3, MR4, MR5
and MR6, but not MR1 and MR2; additional discussion of engine
resizing for mass reduction can be found in Section VI.B.3.a)(6)
Autonomie Sizes Powertrains for Full Vehicle Simulation.
---------------------------------------------------------------------------
 The agencies point to the comments from manufacturers, discussed
further above, which support the agencies' assertion that the CAFE
model's resizing constraints are appropriate. As discussed previously,
Toyota commented that this approach better considers the constraints of
engineering resources and manufacturing costs and results in a more
realistic number of engines and transmissions.\532\ The Alliance also
commented on the benefit of constraining engine resizing, stating that
``the platform and engine sharing methodology in the model better
replicates reality by making available to each manufacturer only a
finite number of engine displacements, helping to prevent
unrealistically `over-optimized' engine sizing.'' \533\
---------------------------------------------------------------------------
 \532\ Toyota, Attachment 1, Docket No. NHTSA-2018-0067-12098, at
6.
 \533\ Alliance of Automobile Manufacturers, Attachment ``Full
Comment Set,'' Docket No. NHTSA-2018-0067-12073, at 140.
---------------------------------------------------------------------------
 Another comment from CARB stated that engine resizing ``was only
simulated for cases where those levels of mass reduction were applied,
in the absence of virtually all other technology or efficiency
improvements.'' \534\ The agencies do not agree that resizing should be
simulated in all cases which involve small incremental technologies. In
the final rule analysis, vehicles can have engines resized at four (out
of six) levels of mass reduction technology, during a vehicle redesign
cycle which changes powertrain architecture, and by inheritance during
a vehicle refresh cycle. As discussed previously, the application of
small incremental technologies such as reductions in aerodynamic drag
or rolling resistance does not justify the high cost and complexity of
producing additional varieties of engine sizes. Accordingly, for each
curb weight reduction level of 7.1 percent or above and for each
vehicle technology class, Autonomie sized a baseline engine by running
a simulation of a vehicle without incremental technologies applied;
then, those baseline engines were inherited by all other simulations
using the same levels of curb weight reduction, which also added any
variety of incremental technologies.\535\ For further clarification, in
any case in which a vehicle adopts a 7.1 percent or more curb weight
reduction, no matter what other technologies were already present or
are added to the vehicle in conjunction with the mass reduction, that
vehicle will receive an engine which has been appropriately sized for
the newly applied mass reduction level.\536\ This can be observed in
the Autonomie simulation databases by tracking the ``EngineMaxPower''
column (not the ``VehicleSized'' column).
---------------------------------------------------------------------------
 \534\ California Air Resources Board, Attachment 2, Docket No.
NHTSA-2018-0067-11873, at 178.
 \535\ In the Autonomie simulation database files, the
simulations which establish baseline sized engines are marked
``yes'' in the ``VehicleSized'' column, and the subsequent
simulations which use this engine and add other incremental
technologies are marked ``inherited.'' For a list of Autonomie
simulation database files, see Table VI-4 Autonomie Simulation
Database Output Files in Section VI.A.7 Structure of Model Inputs
and Outputs.
 \536\ For example, if a vehicle possesses MR2, AERO1, and ROLL1
and subsequently adopts MR3, AERO1, ROLL2, the vehicle will adopt
the lower engine power level associated with MR3. As a counter
example, if a vehicle possesses MR3, ROLL1, and AERO1 and
subsequently adopts MR3, ROLL1, AERO2, the engine will not be
resized and it will retain the power level associated with MR3.
---------------------------------------------------------------------------
 Finally, ICCT claimed that the agencies did not sufficiently report
performance-related vehicle information. ICCT commented that the output
files did not show data on ``engine displacement, the maximum power of
each engine, the maximum torque of each engine, the initial and final
curb weight of each vehicle (in absolute terms), and estimated 0-60 mph
acceleration.'' ICCT claimed that because this data was not found, the
agencies are ``showing that they have not even attempted to analyze
accurately the future year fleet for their performance'' and that ``the
agencies are intentionally burying a critical assumption, whereby their
future fleet has not been appropriately downsized, and it therefore has
greatly increased utility and performance characteristics.'' \537\
---------------------------------------------------------------------------
 \537\ International Council on Clean Transportation, Attachment
3, Docket No. NHTSA-2018-0067-11741, at I-74.
---------------------------------------------------------------------------
 In fact, for the NPRM, and again for this final rule, the agencies
did analyze vehicle performance and have made the data available to the
public. An indication of the actual engine displacement change is
available by noting the displacements used in Automonie simulation
database for each of the technology states. The displacements reported
in Autonomie are used by the full-vehicle-simulation within the
Autonomie model, and while they do not directly represent each specific
vehicle's actual engine sizes, they do fully reflect the relative
change in engine size that is applied to each vehicle. It is the
relative change in engine size that is relevant for the analysis.
Similarly, the vehicle power and torque used by the full vehicle
simulations are reported in the Autonomie simulation databases; their
values and relative change across an engine resizing event can be
observed. Initial and final curb weights for the analysis fleet are
reported in Vehicles Report output file column titled ``CW Initial''
and ``CW,'' respectively. The time required for 0-60 mph acceleration
is reported in the Autonomie simulation database files. A detailed
description of the engine resizing methodology is available in the
Argonne Model
[[Page 24338]]
Documentation, which explains how vehicle characteristics are used to
calculate powertrain size.\538\ These data and information that are
available in the Autonomie and CAFE model documentation provide the
information needed to analyze performance, and in fact, this is
evidenced by the statements of numerous commenters discussed in this
section. The agencies have conducted their own performance analysis,
which is discussed further below, using the same data documentation
mentioned here.
---------------------------------------------------------------------------
 \538\ See Chapter 8 of the ANL Model documentation for the final
rule analysis.
---------------------------------------------------------------------------
 Updates to the CAFE model have minimized performance shift over the
simulated model years, and have eliminated performance differences
between simulated standards.
 The Autonomie simulation updates, discussed previously, were
included in the final rule analysis, and have resulted in average
performance that is similar across the regulatory alternatives. Because
the regulatory analysis compares differences in impacts among the
alternatives, the agencies believe that having consistent performance
across the alternatives is an important aspect of performance
neutrality. If the vehicle fleet had performance gains which varied
significantly depending on the alternative, performance differences
would impact the comparability of the simulations. Using the NPRM CAFE
model data, the agencies analyzed the sales-weighted average 0-60
performance of the entire simulated vehicle fleet for MYs 2016 and
2029, and identified that the Augural standards had 4.7 percent better
0-60 mph acceleration time compared to the NPRM preferred alternative,
which had no changes in standards in MYs 2021-2026.\539\ This
assessment confirmed the observations of the various commenters. With
the refinements that were incorporated for the final rule, similar
analysis showed that the Augural standards had a negligible 0.1 percent
difference in 0-60 mph acceleration time compared to the NPRM preferred
alternative.\540\
---------------------------------------------------------------------------
 \539\ The agencies' analysis matched all MY 2016 and MY 2029
vehicles in the NPRM Vehicles Report output file, under both the
Augural standards and preferred alternative, with the appropriate 0-
60 mph acceleration time from the NPRM Autonomie simulation
databases. This was done by examining each vehicle's assigned
technologies, finding the Autonomie simulation with the
corresponding set of technologies, and extracting that simulation's
0-60 mph acceleration time. This process effectively assigned a 0-60
time to every vehicle in the fleet for four scenarios: (1) MY 2016
under augural standards, (2) MY 2016 under the preferred
alternative, (3) MY 2029 under augural standards, and (4) MY 2029
under the preferred alternative. For each scenario, an overall
fleet-wide weighted average 0-60 time was calculated, using each
vehicle's MY2016 sales volumes as the weight. For more information,
see the FRIA Section VI.
 \540\ This updated analysis used the FRM CAFE Model Vehicles
Report output file and the FRM Autonomie simulation databases. The
final rule analysis introduced an updated MY 2017 fleet as a
starting point, replacing the NPRM 2016MY fleet. For more
information, see the FRIA Chapter VI.
---------------------------------------------------------------------------
 The updates applied to the final rule Autonomie simulations also
resulted in further minimizing the performance change across model
years. As the agencies attempted to minimize this performance shift
occurring ``over time,'' it was also acknowledged that a small increase
would be expected and would be reasonable. This increase is attributed
to the analysis recognizing the practical constraints on the number of
unique engine displacements manufacturers can implement, and therefore
not resizing powertrains for every individual technology and every
combination of technologies when the performance impacts are small.
Perfectly equal performance with 0 percent change would not be
achievable while accounting for these real world resizing constraints.
The performance analysis in the 2011 NAS report shared a similar view
on performance changes, stating that ``truly equal performance involves
nearly equal values . . . within 5 percent.'' \541\ In response to
comments, using NPRM CAFE model data, the agencies analyzed the sales-
weighted average 0-60 performance of the entire simulated vehicle
fleet, and identified that the performance increase from MYs 2016 and
2029 was 7.5 percent under Augural Standards and 3.1 percent under the
NPRM preferred alternative standards. The agencies conducted a similar
analysis using final rule data and found the performance increase over
time from MYs 2017 to 2029 was 3.9 percent for Augural Standards and
4.0 percent for the NPRM preferred alternative standards. The agencies
determined this change in performance is reasonable and note it is
within the 5 percent bound in discussed by NAS in its 2011 report.
---------------------------------------------------------------------------
 \541\ National Research Council. 2011. Assessment of Fuel
Economy Technologies for Light-Duty Vehicles. Washington, DC--The
National Academies Press, at 62. http://nap.edu/12924.
---------------------------------------------------------------------------
 This assessment shows that for the final rule analysis, performance
is neutral across regulatory alternatives and across the simulated
model years allowing for fair, direct comparison among the
alternatives.
(7) How the Agencies Simulated Vehicle Models on Test Cycles
 After vehicle models are built for every combination of
technologies and vehicle classes represented in the analysis, Autonomie
simulates their performance on test cycles to calculate the
effectiveness improvement of the fuel-economy-improving technologies
that have been added to the vehicle. Discussed earlier, the agencies
minimize the impact of potential variation in determining effectiveness
by using a series of tests and procedures specified by federal law and
regulations under controlled conditions.
 Autonomie simulates vehicles in a very similar process as the test
procedures and energy consumption calculations that manufacturers must
use for CAFE and CO2 compliance.542 543 544
Argonne simulated each vehicle model on several test procedures to
evaluate effectiveness. For vehicles with conventional powertrains and
micro hybrids, Autonomie simulates the vehicles on EPA 2-cycle test
procedures and guidelines.\545\ For mild and full hybrid electric
vehicles and FCVs, Autonomie simulates the vehicles using the same EPA
2-cycle test procedure and guidelines, and the drive cycles are
repeated until the initial and final state of charge are within a SAE
J1711 tolerance. For PHEVs, Autonomie simulates vehicles in similar
procedures and guidelines as SAE J1711.\546\ For BEVs Autonomie
simulates vehicles in similar procedures and guidelines as SAE
J1634.\547\
---------------------------------------------------------------------------
 \542\ EPA, ``How Vehicles are Tested.'' https://www.fueleconomy.gov/feg/how_tested.shtml. Last accessed Nov 14,
2019.
 \543\ ANL model documentation for final rule Chapter 6. Test
Procedures and Energy Consumption Calculations.
 \544\ EPA Guidance Letter. ``EPA Test Procedures for Electric
Vehicles and Plug-in Hybrids.'' Nov. 14, 2017. https://www.fueleconomy.gov/feg/pdfs/EPA%20test%20procedure%20for%20EVs-PHEVs-11-14-2017.pdf. Last accessed Nov. 7, 2019.
 \545\ 40 CFR part 600.
 \546\ PHEV testing is broken into several phased based on SAE
J1711. Charge-Sustaining on the City cycle, Charge-Sustaining on the
HWFET cycle, Charge-Depleting on the City and HWFET cycles.
 \547\ SAE J1634. ``Battery Electric Vehicle Energy Consumption
and Range Test Procedure.'' July 12, 2017.
---------------------------------------------------------------------------
b) Selection of One Full-Vehicle Modeling and Simulation Tool
 The NPRM described tools that the agencies previously used to
estimate technology effectiveness. For the analysis supporting the 2012
final rule for MYs 2017 and beyond, the agencies used technology
effectiveness estimates from EPA's lumped parameter model (LPM). The
LPM was calibrated using data from vehicle simulation work performed by
Ricardo Engineering.\548\
[[Page 24339]]
The agencies also used full vehicle simulation modeling data from
Autonomie vehicle simulations performed by Argonne for mild hybrid and
advanced transmission effectiveness estimates.549 550
---------------------------------------------------------------------------
 \548\ Response to Peer Review of: Ricardo Computer Simulation of
Light-Duty Vehicle Technologies for Greenhouse Gas Emission
Reduction in the 2020-2025 Timeframe, EPA-420-R-11-021 (December
2011), available at https://nepis.epa.gov/Exe/ZyPDF.cgi/P100D5BX.PDF?Dockey=P100D5BX.PDF.
 \549\ Joint TSD: Final Rulemaking for 2017-2025 Light-Duty
Vehicle Greenhouse Emission Standards and Corporate Average Fuel
Economy Standards. August 2012. EPA-420-R-12-901.3.3.1.3 Argonne
National Laboratory Simulation Study p. 3-69.
 \550\ Moawad, A. and Rousseau, A., ``Impact of Electric Drive
Vehicle Technologies on Fuel Efficiency,'' Energy Systems Division,
Argonne National Laboratory, ANL/ESD/12-7, August 2012.
---------------------------------------------------------------------------
 For the 2016 Draft TAR analysis, EPA and NHTSA used two different
full system simulation programs for complementary but separate
analyses. NHTSA used Argonne's Autonomie tool, described in detail
above, with engine map inputs developed by IAV using GT-Power in 2014,
and updated in 2016.551 552 553 Argonne, in coordination
with NHTSA, developed a methodology for large scale simulation using
Autonomie and distributed computing, thus overcoming one of the
challenges to full vehicle simulation that the NAS committee outlined
in its 2015 report and implementing a recommendation that the agencies
use full-vehicle simulation to improve the analysis method of
estimating technology effectiveness.\554\ EPA used a limited number of
full-vehicle simulations performed using its ALPHA model, an EPA-
developed full-vehicle simulation model,\555\ to calibrate the LPM,
used to estimate technology effectiveness. EPA also used the same
modeling approach for its Proposed Determination analysis.\556\
---------------------------------------------------------------------------
 \551\ GT-Power Engine Simulation Software. https://www.gtisoft.com/gt-suite-applications/propulsion-systems/gt-power-engine-simulation-software/. Last accessed Oct. 10, 2019.
 \552\ 2016 Draft TAR Engine Maps by IAV Automotive Engineering
using GT-Power. https://www.nhtsa.gov/staticfiles/rulemaking/pdf/cafe/IAV_EngineMaps_Details.xlsx. Lass accessed Oct. 10, 2019.
 \553\ NHTSA-2018-0067-0003. ANL--Summary of Main Component
Performance Assumptions NPRM.
 \554\ See National Research Council. 2015. Cost, Effectiveness,
and Deployment of Fuel Economy Technologies for Light-Duty Vehicles.
Washington, DC: The National Academies Press [hereinafter ``2015 NAS
Report''] at p. 263, available at https://www.nap.edu/catalog/21744/cost-effectiveness-and-deployment-of-fuel-economy-technologies-for-light-duty-vehicles (last accessed June 21, 2018). See also A.
Moawad, A. Rousseau, P. Balaprakash, S. Wild, ``Novel Large Scale
Simulation Process to Support DOT's CAFE Modeling System,''
International Journal of Automotive Technology (IJAT), Paper No.
220150349, Nov 2015; Pagerit, S., Sharper, P., Rousseau, A., Sun, Q.
Kropinski, M. Clark, N., Torossian, J., Hellestrand, G., ``Rapid
Partitioning, Automatic Assembly and Multicore Simulation of
Distributed Vehicle Systems.'' ANL, General Motors, EST Embedded
Systems Technology. 2015. https://www.autonomie.net/docs/5%20-%20Presentations/VPPC2015_ppt.pdf. Last accessed Dec. 9, 2019.
 \555\ See Lee, B., S. Lee, J. Cherry, A. Neam, J. Sanchez, and
E. Nam. 2013. Development of Advanced Light-Duty Powertrain and
Hybrid Analysis Tool. SAE Technical Paper 2013-01-0808. doi:
10.4271/2013-01-0808.
 \556\ Proposed Determination on the Appropriateness of the Model
Year 2022-2025 Light-Duty Vehicle Greenhouse Gas Emissions Standards
under the Midterm Evaluation, EPA-420-R-16-020 (November 2016),
available at https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockey=P100Q3DO.pdf; Final Determination on the
Appropriateness of the Model Year 2022-2025 Light-Duty Vehicle
Greenhouse Gas Emissions Standards under the Midterm Evaluation,
EPA-420-R-17-001 (January 2017), available at https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockey=P100QQ91.pdf.
---------------------------------------------------------------------------
 In the subsequent August 2017 Request for Comment on
Reconsideration of the Final Determination of the Mid-Term Evaluation
of Greenhouse Gas Emissions Standards for MY 2022-2025 Light-Duty
Vehicles, the agencies requested comments on whether EPA should use
alternative methodologies and modeling, including the Autonomie full-
vehicle simulation tool and DOT's CAFE model, for the analysis that
would accompany its revised Final Determination.\557\ As discussed in
the NPRM, stakeholders questioned the efficacy of the combined outputs
and assumptions of the LPM and ALPHA,\558\ especially as the tools were
used to evaluate increasingly heterogeneous combinations of
technologies in the vehicle fleet.\559\
---------------------------------------------------------------------------
 \557\ 82 FR 39551 (Aug. 21, 2017).
 \558\ 83 FR 43022 (``At NHTSA-2016-0068-0082, p. 49, FCA
provided the following comments, ``FCA believes EPA is
overestimating the benefits of technology. As the LPM is calibrated
to those projections, so too is the LPM too optimistic.'' FCA also
shared the chart, `LPM vs. Actual for 8 Speed Transmissions.' '').
 \559\ 83 FR 43022 (referencing Automotive News ``CAFE math gets
trickier as industry innovates'' (Kulisch), March 26, 2018.).
---------------------------------------------------------------------------
 More specifically, the Auto Alliance noted that their previous
comments to the midterm evaluation, in addition to comments from
individual manufacturers, highlighted multiple concerns with EPA's
ALPHA model that were unresolved, but addressed in Autonomie.\560\
First, the Alliance expressed concern over ALPHA modeling errors
related to road load reductions, stating that an error derived from how
mass and coast-down coefficients were updated when mass, tire and aero
improvements were made resulted in benefits overstated by 3 percent to
11 percent for all vehicle types. Next, the Alliance repeated its
concern that EPA should consider top-gear gradeability as one of its
performance metrics to maintain functionality, noting that EPA had
acknowledged the industry's comments in the Proposed Determination,
``but generally dismissed the auto industry concerns.'' Additional
analysis by EPA in its Response to Comments document did not allay the
Alliance's concerns,\561\ as the Alliance concluded that ``[c]onsistent
with the National Academy of Sciences recommendation from 2011, EPA
should monitor gradeability to ensure minimum performance.''
---------------------------------------------------------------------------
 \560\ EPA-HQ-OAR-2015-0827-9194, at p. 36-44.
 \561\ The Alliance noted that in higher-gear-count
transmissions, like 8-speed automatics, modeled by ALPHA with an
expanded ratio spread to achieve fuel economy, are concerning for
gradeability. Additionally, infinite engine downsizing along with
expanded ratio spread transmission, in real world gradeability may
cause further deteriorate as modeled in ALPHA, which leads to
inflated effectiveness values for powertrains that would not meet
customer demands.
---------------------------------------------------------------------------
 Furthermore, the Alliance stated that ALPHA vehicle technology
walks provided in response to manufacturer comments on the Proposed
Determination did not correctly predict cumulative effectiveness when
compared to technologies in real world applications. The Alliance
stated that many of the individual technologies and assumptions used by
ALPHA overestimated technology effectiveness and were derived from
questionable sources. As an example, the Alliance referenced an engine
map used by EPA to represent the Honda L15B7 engine, where the engine
map data was collected by ``(1) taking a picture of an SAE document
containing an image of the engine map, and then (2) `digitizing' the
image by `tracing image contours' '' (citing EPA's ALPHA
documentation). The Alliance could not definitively state whether the
``digitization'' process, lack of detail in the source image, or
another factor were the reasons that some regions of overestimated
efficiency were observed in the engine map, but concluded that ``the
use of this map should be discontinued within ALPHA,'' and ``any
analysis conducted with it is highly questionable.'' Based on these
concerns and others, the Alliance recommended that Autonomie be used to
inform the downstream cost optimization models (i.e., the CAFE model
and/or OMEGA).
 Global Automakers argued that NHTSA's CAFE model, which
incorporates data from Autonomie simulations, provided a more
transparent and discrete step through each of the modeling
scenarios.\562\ Global pointed out that the LPM is ``of particular
concern due to its simplified technology projection processes,'' and it
``propagates fundamentally flawed
[[Page 24340]]
content into the ALPHA and OMEGA models and therefore cannot accurately
assess the efficacy of fuel economy technologies.'' Global did note
that EPA ``plans to abandon its reliance on LPM in favor of another
modeling approach,'' referring to the RSE,\563\ but stated that ``EPA
must provide stakeholders with adequate time to evaluate the updated
modeling approach, ensure it is analytically robust, and provide
meaningful feedback.'' Global Automakers concluded that EPA's engine
mapping and tear-down analyses have played an important role in
generating publicly-available information, and stated that the data
should be integrated into the Autonomie model.
---------------------------------------------------------------------------
 \562\ EPA-HQ-OAR-2015-0827-9728, at 14.
 \563\ See Moskalik, A., Bolon, K., Newman, K., and Cherry, J.
``Representing GHG Reduction Technologies in the Future Fleet with
Full Vehicle Simulation,'' SAE Technical Paper 2018-01-1273, 2018,
doi:10.4271/2018-01-1273. Since 2018, EPA has employed vehicle-
class-specific response surface equations automatically generated
from a large number of ALPHA runs to more readily apply large-scale
simulation results, which eliminated the need for manual calibration
of effectiveness values between ALPHA and the LPM.
---------------------------------------------------------------------------
 On the other hand, other stakeholders commented that EPA's ALPHA
modeling should continue to be used, for procedural reasons like,
``[i]t would appear arbitrary for EPA now, after five years of modeling
based on ALPHA, to declare it can no longer use its internally
developed modeling tools and must rely solely on the Autonomie model,''
and ``[t]he ALPHA model is inextricably built into the regulatory and
technical process. It will require years of new analysis to replace the
many ALPHA and OMEGA modeling inputs and outputs that permeate the
entire rulemaking process, should EPA suddenly decide to change its
models.'' \564\ Commenters also cited technical reasons to use ALPHA,
like EPA's progress benchmarking and validating the ALPHA model to over
fifteen various MY 2013-2015 vehicles,\565\ and that technologies like
the ``Atkinson 2'' engine technology were not considered in NHTSA's
compliance modeling.\566\ Commenters also cited that ALPHA was created
to be publicly available, open-sourced, and peer-reviewed, ``to allow
for transparency to both automakers and public stakeholders, without
hidden and proprietary aspects that are present in commercial modeling
products.'' \567\
---------------------------------------------------------------------------
 \564\ EPA-HQ-OAR-2015-9826, at 39-40.
 \565\ EPA-HQ-OAR-2015-9826, at 40.
 \566\ EPA-HQ-OAR-2015-9197, at 28.
 \567\ EPA-HQ-OAR-2015-9826, at 38.
---------------------------------------------------------------------------
 The agencies described in the NPRM that after having reviewed
comments about whether EPA should use alternative methodologies and
modeling, and after having considered the matter fully, the agencies
determined it was reasonable and appropriate to use Autonomie for full-
vehicle simulation.\568\ The agencies stated that nothing in Section
202(a) of the Clean Air Act (CAA) mandated that EPA use any specific
model or set of models for analysis of potential CO2
standards for light duty vehicles. The agencies also distinguished the
models and the inputs used to populate them; specifically, comments
presented as criticisms of the models, such as ``Atkinson 2'' engine
technology not considered in the compliance modeling, actually
concerned model inputs.\569\
---------------------------------------------------------------------------
 \568\ 83 FR 43001.
 \569\ 83 FR 43002.
---------------------------------------------------------------------------
 With regards to modeling technology effectiveness, the agencies
concluded that, although the CAFE model requires no specific approach
to developing effectiveness inputs, the National Academy of Sciences
recommended, and stakeholders have commented, that full-vehicle
simulation provides the best balance between realism and practicality.
As stated above, Argonne has spent several years developing, applying,
and expanding means to use distributed computing to exercise its
Autonomie full-vehicle simulation tool at the scale necessary for
realistic analysis of technologies that could be used to comply with
CAFE and CO2 standards, and this scalability and related
flexibility (in terms of expanding the set of technologies to be
simulated) makes Autonomie well-suited for developing inputs to the
CAFE model.
 In response to the NPRM, the Auto Alliance commented that NHTSA's
modeling and analysis tools are superior to EPA's, noting that NHTSA's
tools have had a significant lead in their development.\570\ The
Alliance pointed out that Autonomie was developed from the beginning to
address the complex task of combining two power sources in a hybrid
powertrain, while EPA's ALPHA model had not been validated or used to
simulate hybrid powertrains. While both models are physics-based
forward looking vehicle simulators, the Alliance commented that
Autonomie is fully documented with available training, while ALPHA
``has not been documented with any instructions making it difficult for
users outside of EPA to run and interpret the model.'' The Alliance
also mentioned specific improvements in the Autonomie simulations since
the Draft TAR, including expanded performance classes to better
consider vehicle performance characteristics, the inclusion of
gradeability as a performance metric, as recommended by the NAS, the
inclusion of new fuel economy technologies, and the removal of unproven
technologies.
---------------------------------------------------------------------------
 \570\ NHTSA-2018-0067-12073.
---------------------------------------------------------------------------
 The Alliance, Global Automakers, and other automakers writing
separately all stated that the agencies should use one simulation and
modeling tool for analysis.571 572 The Alliance stated that
since both the Autonomie and ALPHA modeling systems answer essentially
the same questions, using both systems leads to inconsistencies and
conflicts, and is inefficient and counterproductive.
---------------------------------------------------------------------------
 \571\ NHTSA-2018-0067-12073; NHTSA-2018-0067-12032. Comments of
the Association of Global Automakers, Inc. on the Safer Affordable
Fuel-Efficient Vehicles Rule Docket ID Numbers: NHTSA-2018-0067 and
EPA-HQ-OAR-2018-0283 October 26, 2018.
 \572\ NHTSA-2018-0067-11943. FCA Comments on The Safer
Affordable Fuel-Efficient (SAFE) Vehicles Rule for Model Years 2021-
2026 Passenger Cars and Light Trucks Notice of Proposed Rulemaking.
---------------------------------------------------------------------------
 The agencies agree with the Alliance that the fully developed and
validated Autonomie model fulfills the agencies' analytical needs for
full-vehicle modeling and simulation. The agencies also agree that it
is counterintuitive to have two separate models conducting the same
work.
 Some commenters stated that broadly, EPA was required to conduct
its own technical analysis and rely on its own models to do so.\573\
Those comments are addressed in Section IV.
---------------------------------------------------------------------------
 \573\ NHTSA-2018-0067-12000; NHTSA-2018-0067-12039.
---------------------------------------------------------------------------
 Regarding the merits of EPA's models, and based on previous inputs
and assumptions used to populate those models, ICCT commented that
``[b]ased on the ICCT's global analysis of vehicle regulations, the
EPA's physics-based ALPHA modeling offers the most sophisticated and
thorough modeling of the applicable technologies that has ever been
conducted.'' ICCT listed several reasons for this, including that the
EPA modeling is based on systematic modeling of technologies and their
synergies; it was built and improved upon by extensive modeling by and
with Ricardo (an engineering consulting firm); it incorporated National
Academies input at multiple stages; it has included many peer reviews
at many stages of the modeling and the associated technical reports
published by engineers in many technical journal articles and
conference proceedings; and EPA's Draft TAR analysis, which used ALPHA,
used state-of-the-art engine maps based on benchmarked high-efficiency
engines. ICCT concluded
[[Page 24341]]
that ``[d]espite these rigorous advances in vehicle simulation
modeling, it appears that the agencies have inexplicably abandoned this
approach, expressly disregarding the EPA benchmarked engines, ALPHA
modeling, and all its enhancements since the last rulemaking.''
 The hallmarks ICCT lists regarding the ALPHA modeling are equally
applicable to Autonomie.\574\ Autonomie is also based on systematic
modeling of technologies and their synergies when combined as packages.
The U.S. Department of Energy created Autonomie, and over the past two
decades, helped to develop and mature the processes and inputs used to
represent real-world vehicles using continuous feedback from the tool's
worldwide user base of vehicle manufacturers, suppliers, government
agencies, and other organizations. Moreover, using Autonomie brings the
agencies closer to the NAS Committee's stated goal of ``full system
simulation modeling for every important technology pathway and for
every vehicle class.'' \575\ While the NAS Committee originally thought
that full vehicle simulation modeling would not be feasible for the
thousands of vehicles in the analysis fleets because the technologies
present on the vehicles might differ from the configurations used in
the simulation modeling,\576\ Argonne has developed a process to
simulate explicitly every important technology pathway for every
vehicle class. Moreover, although separate from the Autonomie model
itself, the Autonomie modeling for this rulemaking incorporated other
NAS committee recommendations regarding full vehicle simulation inputs
and input assumptions, including using engine-model-generated maps
derived from a validated baseline map in which all parameters except
the new technology of interest are held constant.\577\
---------------------------------------------------------------------------
 \574\ See Theo LeSieg, Ten Apples Up On Top! (1961), at 4-32.
 \575\ 2015 NAS Report at 358.
 \576\ 2015 NAS Report at 359.
 \577\ NAS Recommendation 2.1.
---------------------------------------------------------------------------
 As discussed further below and in VI.C.1 Engine Paths, this is one
reason why the IAV maps were used instead of the EPA maps, and the
agencies instead referenced EPA's engine maps to corroborate the
Autonomie effectiveness results. The IAV maps are engine-model-
generated maps derived from a validated baseline map in which all
parameters except the new technology of interest are held constant.
While EPA's engine maps benchmarking specific vehicles' engines
incorporate multiple technologies, for example including improvements
in engine friction and reduction in accessory parasitic loads,
comparisons presented in Section VI.C.1 showed that engine maps
developed by IAV, while not exactly the same, are representative of
EPA's engine benchmarking data.
 In addition, both ALPHA and Autonomie have been used to support
analyses that have been published in technical journal articles and
conference proceedings, but those analyses differ fundamentally because
of the nature of the tools. ALPHA was developed as a tool to be used by
EPA's in-house experts.\578\ As EPA stated in the ALPHA model peer
review,\579\ ``ALPHA is not intended to be a commercial product or
supported for wide external usage as a development tool.'' \580\
Accordingly, EPA experts have published several peer-reviewed journal
articles using ALPHA and have presented the results of those papers at
conference proceedings.\581\
---------------------------------------------------------------------------
 \578\ ALPHA Peer Review, at 4-1.
 \579\ ICCT's comments intimate that ALPHA has been peer reviewed
at many stages of the modeling; although EPA has published several
peer-reviewed technical papers, the ALPHA model itself has been
subject to one peer review. See Peer Review of ALPHA Full Vehicle
Simulation Model, available at https://nepis.epa.gov/Exe/ZyPdf.cgi?Dockey=P100PUKT.pdf.
 \580\ ALPHA Peer Review, at 4-2.
 \581\ See, e.g., Dekraker, P., Kargul, J., Moskalik, A., Newman,
K. et al., ``Fleet-Level Modeling of Real World Factors Influencing
Greenhouse Gas Emission Simulation in ALPHA,'' SAE Int. J. Fuels
Lubr. 10(1):2017, doi:10.4271/2017-01-0899.
---------------------------------------------------------------------------
 To explore ICCT's comments on the importance of peer review
further, it is important to take the actual substantive content of the
ALPHA peer review into account.\582\ One reviewer raised significant
questions over the availability of ALPHA documentation, stating
``[t]here is an overall lack of detail on key technical features that
are new in the model,'' and ``[w]e were not able to find any
information on how the model handles component weight changes.''
Reviewers also raised questions related to model readiness, stating
``[a]ccording to the documentation review, ALPHA's stop/start modeling
appears to be very simplistic.'' Moreover, when running ALPHA
simulations, the reviewer noted the results ``strongly suggest that the
model has errors in the underlying equations or coding with respect to
all of the load reductions.'' Also, one reviewer said the following of
ALPHA: ``A specific simulation runtime is significantly high, more than
10 mins. without providing any indication to the user progress made so
far. A fairly more complicated model such as Autonomie available even
with enhanced capabilities is significantly faster today.'' \583\
---------------------------------------------------------------------------
 \582\ EPA. ``Peer Review of ALPHA Full Vehicle Simulation
Model.'' EPA-420-R-16-013. October 2016. https://nepis.epa.gov/Exe/ZyPdf.cgi?Dockey=P100PUKT.pdf. Last accessed Nov 18, 2019.
 \583\ Peer Review of ALPHA Full Vehicle Simulation Model, at C-
4, available at https://nepis.epa.gov/Exe/ZyPdf.cgi?Dockey=P100PUKT.pdf.
---------------------------------------------------------------------------
 The peer reviewer's assessment of Autonomie as a more complicated
model with enhanced capabilities is not surprising, given Autonomie's
history of development. Autonomie is a commercial tool with more than
275 worldwide organizational users, including vehicle manufacturers,
suppliers, government agencies, and nonprofit organizations having
licensed and used Autonomie. Both Autonomie's creators and user base
unaffiliated with Argonne have published over 100 papers, including
peer-reviewed papers in journals, related to Autonomie validation and
other studies.584 585 One could even argue that the tool has
been continuously peer reviewed by these thousands of experts over the
past two decades.
---------------------------------------------------------------------------
 \584\ At least 15 peer-reviewed papers authored by ANL experts
have been referenced throughout this Section, and others can be
found at SAE International's website, https://www.sae.org/, using
the search bar for ``Autonomie.''
 \585\ See, e.g., Haupt, T., Henley, G., Card, A., Mazzola, M. et
al., ``Near Automatic Translation of Autonomie-Based Power Train
Architectures for Multi-Physics Simulations Using High Performance
Computing,'' SAE Int. J. Commer. Veh. 10(2):483-488, 2017, https://doi.org/10.4271/2017-01-0267; Samadani, E., Lo, J., Fowler, M.,
Fraser, R. et al., ``Impact of Temperature on the A123 Li-Ion
Battery Performance and Hybrid Electric Vehicle Range,'' SAE
Technical Paper 2013-01-1521, 2013, https://doi.org/10.4271/2013-01-1521.
---------------------------------------------------------------------------
 In fact, in responding to a peer review comment on the ALPHA
model's underlying equations and coding with respect to road load
reductions, EPA noted that Autonomie had been used as a reference
system simulation tool to validate ALPHA model results.\586\
---------------------------------------------------------------------------
 \586\ Peer Review of ALPHA Full Vehicle Simulation Model, at 4-
14 and 4-15, available at https://nepis.epa.gov/Exe/ZyPdf.cgi?Dockey=P100PUKT.pdf.
---------------------------------------------------------------------------
 Outside of formal peer-reviewed studies, Autonomie has been used by
organizations like ICCT to support policy documents, position briefs,
and white papers assessing the potential of future efficiency
technologies to meet potential regulatory requirements,\587\
[[Page 24342]]
just as the agencies did in this rulemaking.
---------------------------------------------------------------------------
 \587\ See, e.g., Oscar Delgado and Nic Lutsey, Advanced Tractor-
Trailer Efficiency Technology Potential in the 2020-2030 Timeframe
(April 2015), available at https://theicct.org/sites/default/files/publications/ICCT_ATTEST_20150420.pdf; Ben Sharpe, Cost-
Effectiveness of Engine Technologies for a Potential Heavy-Duty
Vehicle Fuel Efficiency Regulation in India (June 2015), available
at https://theicct.org/sites/default/files/publications/ICCT_position-brief_HDVenginetech-India_jun2015.pdf; Ben Sharpe and
Oscar Delgado, Engines and tires as technology areas for efficiency
improvements for trucks and buses in India (working paper published
March 2016), available at https://theicct.org/sites/default/files/publications/ICCT_HDV-engines-tires_India_20160314.pdf.
---------------------------------------------------------------------------
 Similarly to ICCT, UCS stated that in contrast to Autonomie, ALPHA
had been thoroughly peer-reviewed and is constantly being updated to
reflect the latest technology developments based on work performed by
the National Vehicle and Fuel Emissions Laboratory.\588\ UCS also
stated that because EPA has direct control over the model and its
interface to OMEGA, EPA can better ensure that the inputs into OMEGA
reflect the most up-to-date data, unlike the Autonomie work, which
effectively has to be ``locked in'' before it can be deployed in the
CAFE model. UCS also stated that ALPHA is based on the GEM model (used
to simulate compliance with heavy-duty vehicle regulations) which was
been updated with feedback from heavy-duty vehicle manufacturers and
suppliers, and in fact, ``NHTSA has such confidence in the GEM model
that they accept its simulation-based results as compliance with the
heavy-duty fuel economy regulations.''
---------------------------------------------------------------------------
 \588\ NHTSA-2018-0067-12039 (UCS).
---------------------------------------------------------------------------
 Again, the agencies believe that it is important to note that
Autonomie not only meets, but also exceeds, UCS' listed metrics.
Autonomie's models, sub-models, and controls are constantly being
updated to reflect the latest technology developments based on work
performed by Argonne National Laboratory's Advanced Mobility Technology
Laboratory (AMTL) (formerly Advanced Powertrain Research Facility, or
ARPF).589 590 The Autonomie validation has included nine
validation studies with accompanying reports for software, six
validation studies and reports for powertrains, nine validation studies
and reports for advanced components, ten validation studies and reports
for advanced controls, and overall model validation using test data
from over 50 vehicles.\591\
---------------------------------------------------------------------------
 \589\ See NPRM PRIA. The agencies cited a succinctly-summarized
presentation of Autonomie vehicle validation procedures based on
AMTL test data in the NPRM ANL modeling documentation and PRIA
docket for stakeholders to review at NHTSA-2018-0067-1972 and NHTSA-
2018-0067-0007.
 \590\ Jeong, J., Kim, N., Stutenberg, K., Rousseau, A.,
``Analysis and Model Validation of the Toyota Prius Prime,'' SAE
2019-01-0369, SAE World Congress, Detroit, April 2019; Kim, N,
Jeong, J., Rousseau, A. & Lohse-Busch, H. ``Control Analysis and
Thermal Model Development of PHEV,'' SAE 2015-01-1157, SAE World
Congress, Detroit, April 15; Kim, N., Rousseau, A. & Lohse-Busch, H.
``Advanced Automatic Transmission Model Validation Using Dynamometer
Test Data,'' SAE 2014-01-1778, SAE World Congress, Detroit, Apr.
14.; Lee, D. Rousseau, A. & Rask, E. ``Development and Validation of
the Ford Focus BEV Vehicle Model,'' 2014-01-1809, SAE World
Congress, Detroit, Apr. 14; Kim, N., Kim, N., Rousseau, A., & Duoba,
M. ``Validating Volt PHEV Model with Dynamometer Test Data using
Autonomie,'' SAE 2013-01-1458, SAE World Congress, Detroit, Apr.
13.; Kim, N., Rousseau, A., & Rask, E. ``Autonomie Model Validation
with Test Data for 2010 Toyota Prius,'' SAE 2012-01-1040, SAE World
Congress, Detroit, Apr. 12; Karbowski, D., Rousseau, A, Pagerit, S.,
& Sharer, P. ``Plug-in Vehicle Control Strategy--From Global
Optimization to Real Time Application,'' 22th International Electric
Vehicle Symposium (EVS22), Yokohama, (October 2006).
 \591\ Rousseau, A. Moawad, A. Kim, Namdoo. ``Vehicle System
Simulation to Support NHTSA CAFE standards for the Draft Tar.''
https://www.nhtsa.gov/sites/nhtsa.dot.gov/files/anl-nhtsa-workshop-vehicle-system-simulation.pdf. Last accessed Nov 20, 2019.
---------------------------------------------------------------------------
 In fact, using Autonomie, which has validated data based on test
data from over 50 vehicles, alleviates other stakeholder concerns about
the level of model validation in past analyses. For example, Global
Automakers expressed concerns about whether the effectiveness values
used in past EPA analysis, generated from ALPHA full-vehicle model
simulations, were properly validated, stating that ``[a]lthough EPA
claims that the LPM was calibrated based on thorough testing and
modeling with the ALPHA model, the materials provided with the Proposed
and Final Determination only cover 18 percent of the projected vehicle
fleet with regards to specific combinations of powertrain technology
presented by EPA in the MY 2025 OMEGA pathway. It is unclear how EPA
calibrated the LPM for the remaining 82 percent of the projected
vehicles. EPA's failure to publicly share the data for such a large
percentage of vehicles raises questions about the quality of data.''
\592\ While simple modeled parameters like single dimensional linear
systems, such as engine dynamometer torque measurements can be
validated through other models,\593\ full vehicle systems are complex
multi-dimensional non-linear systems that need to be developed with
multiple data sets, and validated with other fully independent data
sets. Autonomie's models and sub-models have undergone extensive
validation that has proven the models' agreement with empirical data
and the principles of physics.
---------------------------------------------------------------------------
 \592\ Docket ID EPA-HQ-OAR-2015-0827-9728. Global later repeated
that ``only 18% of all vehicle data used as inputs to the ALPHA
modeling was made available in the EPA's public sources. Additional
data had to be specifically requested subsequent to the publication
of the Draft TAR and Proposed Determination. This lack of publicly
available data highlights transparency concerns, which Global
Automakers has raised on several previous occasions.''
 \593\ Section 89.307 Dynamometer calibration.
---------------------------------------------------------------------------
 In addition, the agencies disagree with UCS' comment that EPA's
direct control over its effectiveness modeling and interface to OMEGA
results in a more up-to-date analysis. Argonne's participation in
developing inputs for the rulemaking analysis allowed the agencies
access to vehicle benchmarking data from more vehicles than if the
agencies were limited by their own resources, and access to the Argonne
staff's extensive experience based on direct coordination with vehicle
manufacturers, suppliers, and researchers that all actively use
Autonomie for their own work. In addition to Autonomie's continuous
updates to incorporate the latest fuel-economy-improving technologies,
discussed throughout this section, the data supplied to and generated
by Autonomie for use in the CAFE model was continuously updated during
the analysis process. This is just one part of the iterative quality
assurance (QA) and quality check (QC) process that the agencies
developed when Argonne's large-scale simulation modeling based in
Autonomie was first used for the Draft TAR.
 In addition to Argonne's team constantly updating Autonomie,
Argonne's use of high performance computing (HPC) allowed for constant
update of the analysis during the rulemaking process. Argonne's HPC
platform allows a full set of simulations--over 750,000 modeled
vehicles that incorporate over 50 different fuel-economy-improving
technologies--to be simulated in one week. Subsets of the simulations
can be re-run should issues come up during QA/QC in a day or less.
Tools like the internet and high performance computers have allowed the
agencies to evaluate technology effectiveness with up-to-date inputs
without the proximity of the computers and the people running them
working as a detriment the analysis.
 Finally, GEM, ALPHA, and Autonomie were all developed in the MATLAB
computational environment as forward-looking physics-based vehicle
models. Just as ALPHA has roots in GEM, created in 2010 to accompany
the agencies' heavy-duty vehicle CO2 emissions and fuel
consumption standards, Autonomie has its origins in the software PSAT,
developed over 20 years ago. While this information is useful, as
implied by UCS' comment, the origin of the software was less important
than the capabilities the software could provide for today's analysis.
NHTSA's acceptance of GEM
[[Page 24343]]
results for compliance with heavy-duty fuel economy regulations had no
bearing on the decision to use Autonomie to assess the effectiveness of
light-duty fuel economy and CO2 improving technologies. GEM
was developed to serve as the compliance model for heavy-duty
vehicles,\594\ and GEM serves that limited scope very well.
---------------------------------------------------------------------------
 \594\ Newman, K., Dekraker, P., Zhang, H., Sanchez, J. et al.,
``Development of Greenhouse Gas Emissions Model (GEM) for Heavy- and
Medium-Duty Vehicle Compliance,'' SAE Int. J. Commer. Veh.
8(2):2015, doi:10.4271/2015-01-2771.
---------------------------------------------------------------------------
 UCS did comment that full vehicle simulation could significantly
improve the estimates of technology effectiveness, but thought it
critical that the process be as open and transparent as possible. UCS
pointed to ALPHA results published in peer-reviewed journals as an
example of how transparency has provided the ALPHA modeling effort with
significant and valuable feedback, and contrasted what they
characterized as Autonomie's ``black box'' approach, which they stated
``does not lend itself to similar dialog, nor does it make it easy to
assess the validity of the results.'' Specifically, UCS stated that it
is ``impossible to verify, replicate, or alter the work done by
Autonomie due to the expensive nature of the tools used and lack of
open source or peer-reviewed output.'' In contrast, UCS stated that
EPA's ALPHA model has been thoroughly peer reviewed, and is readily
``downloadable, editable, and accessible to anyone with a MATLAB
license.''
 The agencies responses on the merits of how ALPHA and Autonomie
were peer-reviewed are discussed above. Regarding UCS' comment that it
is impossible to verify, replicate, or alter the work done by
Autonomie, the agencies disagree. All inputs, assumptions, model
documentation--including of component models and individual control
algorithms--and outputs for the NPRM Autonomie modeling were submitted
to the docket for review.\595\ Commenters were able to provide a robust
analysis of Autonomie's technology effectiveness inputs, input
assumptions, and outputs, as shown by their comments on specific
vehicle technology effectiveness assumptions, discussed throughout this
section and in the individual technology sections below.
---------------------------------------------------------------------------
 \595\ NHTSA-2018-0067-1855. ANL Autonomie Compact Car Vehicle
Class Results. Aug 21, 2018. NHTSA-2018-0067-1856. ANL Autonomie
Performance Compact Car Vehicle Class Results. Aug 21, 2018. NHTSA-
2018-0067-1494. ANL Autonomie Midsize Car Vehicle Class Results. Aug
21, 2018. NHTSA-2018-0067-1487. ANL Autonomie Performance Pick-Up
Truck Vehicle Class Results. Aug 21, 2018. NHTSA-2018-0067-1663. ANL
Autonomie Performance Midsize Car Vehicle Class Results. Aug 21,
2018. NHTSA-2018-0067-1486. ANL Autonomie Small SUV Vehicle Class
Results. Aug 21, 2018 NHTSA-2018-0067-1662. ANL Autonomie
Performance Midsize SUV Vehicle Class Results. Aug 21, 2018. NHTSA-
2018-0067-1661. ANL Autonomie Pickup Truck Vehicle Class Results.
Aug 21, 2018. NHTSA-2018-0067-1485. ANL Autonomie Small Performance
SUV Vehicle Class Results. Aug 21, 2018 NHTSA-2018-0067-1492. ANL
Autonomie Midsize SUV Vehicle Class Results. Aug. 21, 2018. NHTSA-
2018-0067-0005. ANL Autonomie Model Assumptions Summary. Aug 21,
2018. NHTSA-2018-0067-0003. ANL Autonomie Summary of Main Component
Assumptions. Aug 21, 2018. NHTSA-2018-0067-0007. Islam, E. S,
Moawad, A., Kim, N, Rousseau, A. ``A Detailed Vehicle Simulation
Process To Support CAFE Standards 04262018--Report'' ANL Autonomie
Documentation. Aug 21, 2018. NHTSA-2018-0067-0004. ANL Autonomie
Data Dictionary. Aug 21, 2018. NHTSA-2018-0067-1692. ANL BatPac
Model 12 55. Aug 21, 2018. NHTSA-2018-0067-12299. Preliminary
Regulatory Impact Analysis (July 2018). Posted July 2018 and updated
August 23 and October 16, 2018.
---------------------------------------------------------------------------
 The agencies also disagree with UCS' assessment of Autonomie as
``expensive.'' While Autonomie is a commercial product, the biggest
financial barrier to entry for both ALPHA and Autonomie is the same: A
MathWorks license.596 597 Regardless, Argonne has made the
version of Autonomie used for this final rule analysis available upon
request, including the individual runs used to generate each technology
effectiveness estimate.\598\
---------------------------------------------------------------------------
 \596\ Autonomie. Frequently Asked Questions. ``Which version of
matlab can I use?'' https://www.autonomie.net/faq.html#faq2. Last
accessed Nov. 19, 2019.
 \597\ EPA ALPHA v2.2 Technology Walk Samples. ``Running this
version of ALPHA requires Matlab/Simulink with StateFlow 2016b.''
https://www.epa.gov/regulations-emissions-vehicles-and-engines/advanced-light-duty-powertrain-and-hybrid-analysis-alpha.
 \598\ Argonne Nationally Laboratory. Autonomie License
Information. https://www.autonomie.net/asp/LicenseRequest.aspx. Last
accessed Nov, 18, 2019.
---------------------------------------------------------------------------
 Next, ICCT supplanted its statement that the agencies
``inexplicably'' abandoned ALPHA, commenting that the agencies'
explanation and justification for relying on Autonomie rather than
ALPHA failed to discuss ALPHA in detail, and the agencies did not
compare and contrast the two models. ICCT continued, ``the EPA cannot
select its modeling tool arbitrarily, yet it appeared that the EPA has
whimsically shifted from an extremely well-vetted, up-to-date,
industry-grade modeling tool to a less-vetted, academic-grade framework
with outdated inputs without even attempt to scrutinize the change.''
ICCT also stated that the agencies are legally obligated to acknowledge
and explain when they change position, and ``cannot simply ignore that
EPA previously concluded that the ALPHA modeling accurately projected
real-world effects of technologies and technology packages.''
 The agencies disagree that a more in-depth discussion of ALPHA was
required in the NPRM. In acknowledging the transition to using
Autonomie for effectiveness modeling and the CAFE model for analysis of
regulatory alternatives,\599\ the agencies described several analytical
needs that using a single analysis from the CAFE model--with inputs
from the Autonomie tool--addressed. These included that Autonomie
produced realistic estimates of fuel economy levels and CO2
emission rates through consideration of real-world constraints, such as
the estimation and consideration of performance, utility, and
drivability metrics (e.g., towing capability, shift busyness, frequency
of engine on/off transitions).\600\ That EPA previously concluded the
ALPHA modeling accurately projected real-world effects of technologies
and technology packages has no bearing on Autonomie's ability to
fulfill the analytical needs that the agencies articulated in the NPRM,
including that Autonomie also accurately projects real-world effects of
technologies and technology packages.
---------------------------------------------------------------------------
 \599\ 83 FR 43000 (Aug. 24, 2018).
 \600\ 83 FR 43001 (Aug. 24, 2018).
---------------------------------------------------------------------------
 The agencies also disagree with ICCT's characterization of ALPHA as
``an extremely well-vetted, up-to-date, industry-grade modeling tool''
and Autonomie as a ``less-vetted, academic-grade framework with
outdated inputs.'' Again, Autonomie has been used by government
agencies, vehicle manufacturers (and by agencies and manufacturers
together in the collaborative government-industry partnership U.S.
DRIVE program), suppliers, and other organizations because of its
ability to simulate many powertrain configurations, component
technologies, and vehicle-level controls over numerous drive cycles.
Characterizing ALPHA as an ``industry-grade modeling tool'' contravenes
EPA's own description of its tool--an in-house vehicle simulation model
used by EPA, not intended to be a commercial product.\601\
---------------------------------------------------------------------------
 \601\ See, e.g., Overview of ALPHA Model, https://www.epa.gov/regulations-emissions-vehicles-and-engines/advanced-light-duty-powertrain-and-hybrid-analysis-alpha; ALPHA Effectiveness Modeling:
Current and Future Light-Duty Vehicle & Powertrain Technologies
(Jan. 20, 2016), available at https://www.epa.gov/sites/production/files/2016-10/documents/alpha-model-sae-govt-ind-mtg-2016-01-20.pdf
(``ALPHA is not a commercial product (e.g. there are no user
manuals, tech support hotlines, graphical user interfaces, or full
libraries of components).''). See also Peer Review of ALPHA Full
Vehicle Simulation Model, available at https://nepis.epa.gov/Exe/ZyPdf.cgi?Dockey=P100PUKT.pdf. While ALPHA peer reviewers found the
model to be a ``fairly simple transparent model . . . [t]he model
execution requires an expert MatLab/Simulink user since no user-
friendly interface currently exists.'' Indeed, EPA noted in response
to this comment that ``[a]s with any internal tool, EPA does not
have the need for a ``user-friendly interface'' like one that would
normally accompany a commercial product which is available for
purchase and fully supported for wide external usage.''
---------------------------------------------------------------------------
[[Page 24344]]
 That characterization also contravenes documentation from the
automotive industry indicating that manufacturers consider ALPHA to
generate overly optimistic effectiveness values, to be unrepresentative
of real-world constraints, and a difficult tool to
use.602 603 The Alliance commented to the MTE
reconsideration that ``[p]revious comments from the Alliance and
individual manufacturers to the MTE docket have highlighted multiple
concerns with EPA's ALPHA model. Many of these concerns remain
unresolved.'' \604\ Furthermore, the Alliance commented that ALPHA
``has not been documented with any instructions making it difficult for
users outside of EPA to run and interpret the model.'' \605\ Global
Automakers further stated that the ``lack of publicly available data
[related to inputs used in the ALPHA modeling] highlights transparency
concerns, which Global Automakers has raised on several previous
occasions.'' \606\ In fact, both the Alliance of Automobile
Manufacturers and Global Automakers, the two trade organizations that
represent the automotive industry, concluded that Autonomie should be
used to generate effectiveness inputs for the CAFE model.\607\
---------------------------------------------------------------------------
 \602\ See EPA-HQ-OAR-2015-0827-10125, at 7. As part of their
assessment that known technologies could not meet the original MY
2022-2025 standards, Toyota noted that the ALPHA conversion of
Toyota's MY 2015 to MY 2025 performance ``appears to yield overly
optimistic results because the powertrain efficiency curves
represent best-case targets and not the average vehicle, the imposed
performance constraints are unmarketable, and the generated credits
are out of sync with product cadence and design cycles.'' See also
NHTSA-2018-0067-12431, at 7. More recently, Toyota stated in their
comments to the NPRM that ``Toyota's position [on the efficacy of
the OMEGA and LPM models] has been clearly represented by comments
previously submitted by the Alliance of Automobile Manufacturers,
Global Automakers, and Novation Analytics. Those comments identify
the LPM and OMEGA models as sources of inaccuracy in EPA technology
evaluations and provide suggested improvements. Neither model is
transparent, intuitive, or user friendly.''
 \603\ EPA-HQ-OAR-2015-0827-9194.
 \604\ EPA-HQ-OAR-2015-0827-9194, at 33.
 \605\ EPA-HQ-OAR-2015-0827-9194.
 \606\ EPA-HQ-OAR-2015-0827-9728.
 \607\ EPA-HQ-OAR-2015-0827-9163 at 5. (``EPA should abandon the
lumped-parameter model and instead use NHTSA's Autonomie and Volpe
models to support the Revised Final Determination.''). See also EPA-
HQ-OAR-2015-0827-9728 at 15 (stating the EPA's engine mapping and
tear down analyses ``should be integrated into the Autonomie model,
which then feeds into the Volpe modeling process.''); EPA-HQ-OAR-
2015-0827-9194 at 33.
---------------------------------------------------------------------------
 In addition, Autonomie contains up-to-date sub-models to represent
the latest electrification and advanced transmission and advanced
engine technologies. As summarized by the Alliance, ``Autonomie was
developed from the start to address the complex task of combining 2
power sources in a hybrid powertrain.'' \608\ Autonomie has
continuously improved over the years by adopting new technologies into
its modeling framework. Even a small sampling of SAE papers shows how
Autonomie has been validated to simulate the latest fuel-economy-
improving technologies like hybrid vehicles and PHEVs.\609\
---------------------------------------------------------------------------
 \608\ Alliance, Docket ID NHTSA-2018-0067-12073 at 135.
 \609\ Jeong, J., Kim, N., Stutenberg, K., Rousseau, A.,
``Analysis and Model Validation of the Toyota Prius Prime,'' SAE
2019-01-0369, SAE World Congress, Detroit, April 2019; Kim, N,
Jeong, J. Rousseau, A. & Lohse-Busch, H. ``Control Analysis and
Thermal Model Development of PHEV,'' SAE 2015-01-1157.
---------------------------------------------------------------------------
 Moreover, Autonomie effectively considers other real-world
constraints faced by the automotive industry. Vehicle manufacturers and
suppliers spend significant time and effort to ensure technologies are
incorporated into vehicles in ways that will balance consumer
acceptance for attributes such as driving quality,\610\ noise-
vibration-harshness (NVH), and meeting other regulatory mandates, like
EPA's and CARB's On-Board Diagnostics (OBD) requirements,\611\ and
EPA's and CARB's criteria exhaust emissions standards.\612\ The
implementation of new fuel economy improving technologies have at times
raised consumer acceptance issues.\613\ As discussed earlier, there are
diminishing returns for modeling every vehicle attribute and tradeoff,
as each takes time and incurs cost; however, Autonomie sub-models are
designed to account for a number of the key attributes and tradeoffs,
so the resulting effectiveness estimates reflect these real world
constraints.
---------------------------------------------------------------------------
 \610\ An example of a design requirement is accommodating the
``lag'' in torque delivery due to the spooling of a turbine in a
turbocharged downsized engine. This affects real-world vehicle
performance, as well as the vehicle's ability to shift during normal
driving and test cycles.
 \611\ EPA adopted and incorporated by reference current OBD
regulations by the California ARB, effective for MY 2017, that cover
all vehicles except those in the heavier fraction of the heavy-duty
vehicle class.
 \612\ Tier 3 emission standards for light-duty vehicles were
proposed in March 2013 78 FR 29815 (May 21, 2013) and signed into
law on March 3, 2014 79 FR 23413 (June 27, 2014). The Tier 3
standards--closely aligned with California LEV III standards--are
phased-in over the period from MY2017 through MY2025. The regulation
also tightens sulfur limits for gasoline.
 \613\ Atiyeh, C. ``What you need to know about Ford's PowerShift
Transmission Problems'' Car and Driver. July 11, 2019. https://www.caranddriver.com/news/a27438193/ford-powershift-transmission-problems/.
---------------------------------------------------------------------------
 Furthermore, aside from the fact that Autonomie represents the
structural state-of-the-art in full-vehicle modeling and simulation,
Autonomie can be populated with any inputs that could be populated in
the ALPHA model.\614\ The agencies chose to use specific inputs for
this rulemaking because, as discussed further in Sections VI.C below,
they best represent the technologies that manufacturers could
incorporate in the rulemaking timeframe, in a way that balanced
important concerns like consumer acceptance. Some other examples of how
Autonomie inputs have been updated with the latest vehicle technology
data specifically for this analysis include test data incorporated from
both Argonne and NHTSA-sponsored vehicle benchmarking, including an
updated automatic transmission skip-shifting feature,\615\ additional
application of cylinder deactivation for turbocharged downsized
engines, and as discussed above, new modeling and simulation that
includes variable compression ratio and Miller Cycle engines.
---------------------------------------------------------------------------
 \614\ For example, Autonomie used the HCR1 and HCR2 engine maps
used as inputs to ALHPA in the Draft TAR and Proposed Determination.
 \615\ NHTSA Benchmarking, ``Laboratory Testing of a 2017 Ford F-
150 3.5 V6 EcoBoost with a 10-speed transmission.'' DOT HS 812 520.
---------------------------------------------------------------------------
 Finally, ICCT commented that the agencies must conduct a systematic
comparison of the Autonomie modeling system and ALPHA modeling in
several respects, including the differences in technical inputs and
resulting efficiency estimates, to explain how the choice of model
altered the regulatory technology penetration and compliance cost
estimations, and the differences in modeling methodologies, including
regarding the relative level of experience of the teams conducting the
effectiveness modeling, to demonstrate that the choice to use Autonomie
was not ``due to convenience and easier access by the NHTSA research
team, rather than for any technical improvement.'' ICCT stated that
without performing this comparison, ``it otherwise appears that the
agencies switched from a better-vetted model and system of inputs with
more recent input data to a less-vetted model and system of inputs as a
way to bury many dozens of changes without transparency or expert
assessment (as illustrated in the
[[Page 24345]]
above errors and invalidated data on individual technologies).'' Each
issue is discussed below in turn.
 First, regarding technical inputs, technology pathways, and
resulting outputs, ICCT stated that the agencies must compare (1)
whether the models have been routinely strengthened by incorporating
cutting edge 2020-2025 automotive technologies to ensure they reflect
the available improvements; (2) every efficiency technology in the 2016
Draft TAR and original EPA TSD and Proposed and Final Determination
analysis against the NPRM; (3) all the major technology package
pathways (i.e., all combinations with high uptake in the Adopted and
Augural 2025 standards) in the current NPRM versus the 2016 Draft TAR
and the 2016 TSD and original Final Determination analysis; (4) each of
the major 2025 technology package synergies; (5) the modeling work of
EPA's, Ricardo's, and Argonne's 2014-2018 model year engine
benchmarking and modeling of top engine and transmission models; and
``defend why they appear to have chosen to dismiss the superior and
better vetted technology modeling approach.''
 ICCT stated that the agencies must make these comparisons because,
``[o]therwise, it seems obvious that the agencies have subjectively
decided to use the modeling that increases the modeled cost, providing
further evidence of a high degree of bias without an objective
accounting of the methodological differences and the sensitivity of the
results to their new decision.'' Moreover, ICCT stated that ``[b]ecause
ALPHA is the dominant, preferred, and better-vetted modeling and was
used in the original Proposed and Final Determination, the agencies are
responsible for assessing and describing how the use of the ALPHA
modeling would result in a different regulatory result for their
analysis of the 2017-2025 adopted [CO2] and Augural CAFE
standards.''
 The agencies do not believe that it is necessary to conduct a
retrospective comparison of ALPHA/LPM and Autonomie effectiveness for
every technology in the Draft TAR and Proposed Determination to the
NPRM and final rule analyses, between the two models for technologies
and packages used in the NPRM and final rule analysis, or to explain
where and why Autonomie provided different results from ALPHA and the
LPM, to assess and describe how the use of the ALPHA modeling would
result in a different regulatory result of CAFE and CO2
standards, per ICCT's request. While it is anticipated that different
values will be produced using different tools in an analysis, it is not
appropriate to select the tool for use based on preferred results. The
selection of an analysis tool should be based on an evaluation of the
tool's capabilities and appropriateness for the analysis task. The
analysis tool should support the full extent of the analysis and
support the level of input and output resolution required. To compare
the output of the two models for the purpose of selecting a tool for
the analysis would likely be biased and disingenuous to the purpose of
the analysis. In this case, Autonomie was selected for this analysis
for the reasons discussed throughout this section, and accordingly the
agencies believe that it was reasonable to consider effectiveness
estimates developed with Autonomie.
 That said, comparison of how the tools behave is discussed here to
further support the agencies' decision process. To demonstrate, in
addition to everything discussed previously in this section,
differences in how each model handles powertrain systems modeling with
specific examples are discussed below as a reference, and differences
between the agencies' approaches to effectiveness modeling for specific
technologies is discussed in Section VI.C where appropriate. While the
improved approach to estimating technology effectiveness estimates
certainly impacted the regulatory technology penetration, compliance
cost estimates, and ``major 2025 technology packages and synergies,''
how technologies are applied in the compliance modeling and the
associated costs of the technologies is equally as important to
consider when examining factors that might impact the regulatory
analysis; that consideration goes beyond the scope of simply
considering which full vehicle simulation model better performs the
functions required of this analysis.
 The agencies have discussed updates to the technologies considered
in the Autonomie modeling throughout this section, in addition to
Autonomie's models and sub-models that control advanced technologies
like hybrid and electrified powertrains. Autonomie's explicit models,
sub-models, and controls for hybrid and electric vehicles have been
continuously validated over the past several years,\616\ as Autonomie
was developed from the beginning to address the complex task of
combining two power sources in a hybrid powertrain.
---------------------------------------------------------------------------
 \616\ Karbowski, D., Kwon, J., Kim, N., & Rousseau, A.,
``Instantaneously Optimized Controller for a Multimode Hybrid
Electric Vehicle,'' SAE paper 2010-01-0816, SAE World Congress,
Detroit, April 2010; Sharer, P., Rousseau, A., Karbowski, D., &
Pagerit, S. ``Plug-in Hybrid Electric Vehicle Control Strategy--
Comparison between EV and Charge-Depleting Options,'' SAE paper
2008-01-0460, SAE World Congress, Detroit (April 2008); and
Rousseau, A., Shidore, N., Carlson, R., & Karbowski, D. ``Impact of
Battery Characteristics on PHEV Fuel Economy,'' AABC08; Jeong, J.,
Kim, N., Stutenberg, K., Rousseau, A., ``Analysis and Model
Validation of the Toyota Prius Prime,'' SAE 2019-01-0369, SAE World
Congress, Detroit, April 2019; Kim, N, Jeong, J. Rousseau, A. &
Lohse-Busch, H. ``Control Analysis and Thermal Model Development of
PHEV,'' SAE 2015-01-1157, SAE World Congress, Detroit, April 15;
Lee, D. Rousseau, A. & Rask, E. ``Development and Validation of the
Ford Focus BEV Vehicle Model,'' 2014-01-1809, SAE World Congress,
Detroit, Apr. 14; Kim, N., Kim, N., Rousseau, A., & Duoba, M.
``Validating Volt PHEV Model with Dynamometer Test Data using
Autonomie,'' SAE 2013-01-1458, SAE World Congress, Detroit, Apr.
13.; Kim, N., Rousseau, A., & Rask, E. ``Autonomie Model Validation
with Test Data for 2010 Toyota Prius,'' SAE 2012-01-1040, SAE World
Congress, Detroit, Apr. 12; Karbowski, D., Rousseau, A, Pagerit, S.,
& Sharer, P. ``Plug-in Vehicle Control Strategy--From Global
Optimization to Real Time Application,'' 22th International Electric
Vehicle Symposium (EVS22), Yokohama, (October 2006).
---------------------------------------------------------------------------
 Also regarding the inputs to both models, as highlighted in Section
VI.C.3.a), and discussed above, inputs and assumptions for the ALPHA
modeling used for the EPA Draft TAR and Proposed Determination analysis
were projected from benchmarking testing. While it is straightforward
to measure engine fuel consumption and create an engine fuel map, it is
extremely challenging to identify the specific technologies and levels
of technologies present on a benchmarking engine. Attributing changes
in the overall engine fuel consumption to the individual engine
technologies that make up the complete engine involves significant
uncertainty.
 The fixed-point model approach used by the ALPHA model does not
develop an effectiveness function and assigns a single value to a
technology. The single value is derived from benchmark testing, which
often does not isolate the effect of a single technology from the
effects of other technologies on the tested vehicle. To isolate a
single technology's effect for use in fixed point modeling properly,
the agencies would need to benchmark multiple versions of a single
vehicle, carefully controlling changes to the vehicles' fuel efficiency
technologies. This process would need to be repeated for a large
portion of the vehicle fleet and would require significant funding and
thousands of lab hours to complete. Without this level of data, fixed-
point effectiveness estimates tend to be too high, as they are unable
to account for synergetic effects of multiple technologies.
Specifically, when EPA benchmarks vehicles like the 2018 Toyota Camry,
the resulting fuel map captures the benefits of many
[[Page 24346]]
technologies associated with that engine. This data can be helpful when
developing controls and validating component operations in modeling,
but it is inaccurate to conclude that fuel consumption is directly
related to individual engine technologies, such as lubrication and
friction reduction, and geometric improvements in efficiency.
 Contrasted, the NPRM and final rule Autonomie analyses selected
specific base engine maps and applied technologies incrementally, both
individually and in known combinations, to better isolate the impacts
of the technologies. As discussed above, this also implemented NAS
Recommendation 2.1, to use engine-model-generated maps in the full
vehicle simulations derived from a validated baseline map in which all
parameters except the new technology of interest are held
constant.\617\ While the different methods are valid for different
purposes, the method selected for the analysis presented today was more
useful for measuring the incremental effectiveness increments as
opposed to the absolute values of technology effectiveness, e.g., that
could be measured by benchmarking a technology package.
---------------------------------------------------------------------------
 \617\ 2015 NAS Report at p. 82.
---------------------------------------------------------------------------
 To provide an example of another difference in behavior between the
simulation tools, a comparison between ALPHA and Autonomie
transmissions shifting behavior was conducted. The comparison
highlighted the differences in how each simulation tool approaches
transmission shift logic. The ALPHA simulation tool used ALPHAShift.
ALPHAShift is an optimization algorithm that uses numerous vehicle
characteristics to find a best shifting strategy. The primary inputs
for the algorithm includes the fuel consumption (or cost) map for the
vehicle engine.\618\ Although a public version of ALPHA is available
for evaluation, the ALPHAShift algorithm used by the tool is hard coded
with fixed values.619 620 This is an issue, because despite
peer reviewed documentation on how to tune the algorithm,\621\ no
documentation of how the algorithm logic works is available for review.
This is confounding for the use of the software, particularly when the
observed behavior of the model departs from expected behavior. Figure
VI-6 below shows simulated gear shift (left) versus actual gear shift
(right), demonstrating an unexpected shift to neutral before shifting
to the requested gear.
---------------------------------------------------------------------------
 \618\ Newman, K., Kargul, J., and Barba, D., ``Development and
Testing of an Automatic Transmission Shift Schedule Algorithm for
Vehicle Simulation,'' SAE Int. J. Engines 8(3):2015, doi:10.4271/
2015-01-1142.
 \619\ Aymeric, R. Islam, E. S. ``Analysis of EPA's ALPHA Shift
Model--ALPHAShift.'' ANL. March 9, 2020.
 \620\ ALPHA v2.2 Technology Walk Samples. EPA. January 2017.
https://www.epa.gov/sites/production/files/2017-01/alpha-20170112.zip. Last Accessed March 9, 2020.
 \621\ Newman, K., Kargul, J., and Barba, D., ``Development and
Testing of an Automatic Transmission Shift Schedule Algorithm for
Vehicle Simulation,'' SAE Int. J. Engines 8(3):2015, doi:10.4271/
2015-01-1142.
---------------------------------------------------------------------------
 By contrast, and discussed further in VI.C.2 Transmission Paths,
Autonomie uses a fully documented algorithm to develop a best shifting
strategy for each unique vehicle configuration. The algorithm develops
shifting strategies unique to each individual vehicle based on gear
ratio, final drive ratio, engine BSFC and other vehicle
characteristics. This is one example of model behavior, in addition to
the availability of more transparency on this behavior for greater
stakeholder review, that led the agencies to determine it was
reasonable and appropriate to use Autonomie for this analysis.
---------------------------------------------------------------------------
 \622\ ALPHA v2.2 Technology Walk Samples. Jan. 12, 2017. https://www.epa.gov/sites/production/files/2017-01/alpha-20170112.zip. Last
accessed Dec 9, 2019.
[GRAPHIC] [TIFF OMITTED] TR30AP20.114
 Regarding the technical expertise of the team conducting the
---------------------------------------------------------------------------
effectiveness modeling, ICCT commented:
 [T]he agencies should also disclose how much commercial business
is conducted by the Ricardo, IAV, and Argonne Autonomie teams that
underpin the modeling of EPA and NHTSA, respectively, including how
much related research they have done for auto industry clients over
the past ten years. We mention this because we strongly suspect that
Ricardo, upon which EPA built its ALPHA model, has done at least an
order of magnitude (in number of projects, person-hours, and budget)
more work with and for the automotive industry than the IAV and
Autonomie teams have in direct work for
[[Page 24347]]
automotive industry clients. A conventional government procurement
effort that competitively vets potential research expert teams would
presumably have selected for such automotive industry credentials
and experience, yet it appears that the agencies are wholly
deferring to Autonomie's less rigorous research-grade modeling
framework and data due to convenience and easier access by the NHTSA
research team, rather than for any technical improvement, and this
is to the detriment of showing clear understanding of real-world
automotive engineering developments (as demonstrated by many
erroneous technology combination results throughout these comments).
 First, NHTSA follows Federal Acquisition Regulation (FAR) to award
contracts and Interagency Agreements (IAAs),\623\ and any awarded
contracts and IAAs must follow the FAR requirements. Importantly, FAR
3.101-1 includes key aspects of conduct and ethics that NHTSA must
follow in awarding a contract or IAA:
---------------------------------------------------------------------------
 \623\ Federal Acquisition Regulation (FAR). https://www.acquisition.gov/.
 Government business shall be conducted in a manner above
reproach and, except as authorized by statute or regulation, with
complete impartiality and with preferential treatment for none.
Transactions relating to the expenditure of public funds require the
highest degree of public trust and an impeccable standard of
conduct. The general rule is to avoid strictly any conflict of
interest or even the appearance of a conflict of interest in
Government-contractor relationships. While many Federal laws and
regulations place restrictions on the actions of Government
personnel, their official conduct must, in addition, be such that
they would have no reluctance to make a full public disclosure of
their actions.\624\
---------------------------------------------------------------------------
 \624\ FAR 3.101-1.
 While some factors are more relevant than others in considering
whether to award a contract or enter into an IAA, the amount of work
that an organization has performed, characterized by projects, person-
hours, and budget, is only one of a multitude of factors that is
considered (if it is even considered at all--an agency might not
request this information and an organization might decline to provide
it because of contractual clauses or to protect commercial business
interests) when assessing whether an organization meets the agency's
needs for a specific task. Other factors, such as the federal budget,
also set boundaries for the scope of work that can be performed under
any competitive government procurement effort.
 As discussed throughout this section, the team at Argonne National
Laboratory behind Autonomie has developed and refined a state-of-the-
art tool that is used by the automotive industry, government agencies,
and research or other nongovernmental institutions around the world.
The tool has been and continues to be validated to production vehicles,
and updated to include models, sub-models, and controls representing
the state-of-the-art in fuel economy improving technology. To the
extent that ICCT believes that ``research done for auto industry
clients,'' ``work with and for the automotive industry,'' and
``automotive industry credentials and experience,'' are metrics upon
which to base this type of important decision, the agencies point ICCT
to the statements from the automotive industry, above, recommending
Autonomie be used for technology effectiveness modeling.
 ICCT concluded that ``[w]hile the agencies are in their process of
conducting a proper vetting of their NPRM's foundational Autonomie-
based modeling, we recommend that they rely on what appears to be the
superior and better vetted technology modeling approach with more
thorough and state-of-the-art advanced powertrain systems modeling and
engine maps from the EPA ALPHA modeling.''
 The agencies properly vetted the Autonomie modeling and decided
that Autonomie represented a reasonable and appropriate tool to provide
technology effectiveness estimates for this rulemaking. To the extent
that commenters' concerns were more about the effectiveness results
than the tools used to model technology effectiveness, modeling updates
detailed in the Section VI.B.3.c), below, address those comments. While
some commenters may still be dissatisfied with Autonomie's technology
effectiveness estimates, the agencies believe that the refinement of
inputs and input assumptions, and associated explanation of why those
refinements are appropriate and reasonable, have appropriately
addressed comments on these issues. Importantly, none of these
refinements have led either agency to reconsider using Autonomie for
this rulemaking analysis.
 Additional discussion of the agencies' decision to rely on one set
of modeling tools for this rulemaking is located in Section VI.A of
this preamble.
c) Technology Effectiveness Values Implementation in the CAFE Model
 While the Autonomie model produces a large amount of information
about each simulation run--for a single technology combination, in a
single technology class--the CAFE model only uses two elements of that
information: Battery costs and fuel consumption on the city and highway
cycles. The agencies combine the fuel economy information from the two
cycles to produce a composite fuel economy for each vehicle, on each
fuel. Plug-in hybrids, being the only dual-fuel vehicles in the
Autonomie simulation, require efficiency estimates of operation on both
gasoline and electricity--as well as an estimate of the utility factor,
or the number of miles driven on each fuel. The fuel economy
information for each technology combination, for each technology class,
is converted into a single number for use in the CAFE model.
 As described in greater detail below, each Autonomie simulation
record represents a unique combination of technologies, and the
agencies create a technology ``key'' or technology state vector that
describes all the technology content associated with a record. The 2-
cycle fuel economy of each combination is converted into fuel
consumption (gallons per mile) and then normalized relative to the
starting point for the simulations. In each technology class, the
combination with the lowest technology content is the VVT (only)
engine, with a 5-speed transmission, no electrification, and no body-
level improvements (mass reduction, aerodynamic improvements, or low
rolling resistance tires). This is the reference point (for each
technology class) for all the effectiveness estimates in the CAFE
model. The improvement factors that the model uses are a given
combination's fuel consumption improvement relative to the reference
vehicle in its technology class.
 For the majority of the technologies analyzed within the CAFE
Model, the fuel economy improvements were derived from the database of
Autonomie's detailed full-vehicle modeling and simulation results. In
addition to the technologies found in the Autonomie simulation
database, the CAFE modeling system also incorporated a handful of
technologies that were required for CAFE modeling, but were not
explicitly simulated in Autonomie. The total effectiveness of these
technologies either could not be captured on the 2-cycle test, or there
was no robust data that could be used as an input to the full-vehicle
modeling and simulation, like with emerging technologies such as
advanced cylinder deactivation (ADEAC). These additional technologies
are discussed further in Sections VI.B.3 Technology Effectiveness and
individual technologies sections. For calculating fuel economy
improvements attributable to these additional technologies, the model
used defined fuel consumption improvement factors that are constant
[[Page 24348]]
across all technology combinations in the database and scale
multiplicatively when applied together. The Autonomie-simulated and
additional technologies were then externally combined, forming a single
dataset of simulation results (referred to as the vehicle simulation
database, or simply, database), which may then be utilized by the CAFE
modeling system.
 To incorporate the results of the combined database of Autonomie-
simulated and additional technologies, while still preserving the basic
structure of the CAFE Model's technology subsystem, it was necessary to
translate the points in this database into corresponding locations
defined by the technology pathways. By recognizing that most of the
pathways are unrelated, and are only logically linked to designate the
direction in which technologies are allowed to progress, it is possible
to condense the paths into a smaller number of groups based on the
specific technology. In addition, to allow for technologies present on
the Basic Engine and Dynamic Road Load (DLR--i.e., MASS, AERO, and
ROLL) paths to be evaluated and applied in any given combination, a
unique group was established for each of these technologies.
 As such, the following technology groups are defined within the
modeling system: Engine cam configuration (CONFIG), VVT engine
technology (VVT), VVL engine technology (VVL), SGDI engine technology
(SGDI), DEAC engine technology (DEAC), non-basic engine technologies
(ADVENG), transmission technologies (TRANS), electrification and
hybridization (ELEC), low rolling resistance tires (ROLL), aerodynamic
improvements (AERO), mass reduction levels (MR), EFR engine technology
(EFR), electric accessory improvement technologies (ELECACC), LDB
technology (LDB), and SAX technology (SAX). The combination of
technologies along each of these groups forms a unique technology state
vector and defines a unique technology combination that corresponds to
a single point in the database for each technology class evaluated
within the modeling system.
 As an example, a technology state vector describing a vehicle with
a SOHC engine, variable valve timing (only), a 6-speed automatic
transmission, a belt-integrated starter generator, rolling resistance
(level 1), aerodynamic improvements (level 2), mass reduction (level
1), electric power steering, and low drag brakes, would be specified as
``SOHC; VVT; AT6; BISG; ROLL10; AERO20; MR1; EPS; LDB.'' \625\ By
assigning each unique technology combination a state vector such as the
one in the example, the CAFE Model can then assign each vehicle in the
analysis fleet an initial state that corresponds to a point in the
database.
---------------------------------------------------------------------------
 \625\ In the example technology state vector, the series of
semicolons between VVT and AT6 correspond to the engine technologies
which are not included as part of the combination, while the gap
between MR1 and EPS corresponds to EFR and the omitted technology
after LDB is SAX. The extra semicolons for omitted technologies are
preserved in this example for clarity and emphasis, and will not be
included in future examples.
---------------------------------------------------------------------------
 Once a vehicle is assigned (or mapped) to an appropriate technology
state vector (from one of approximately three million unique
combinations, which are defined in the vehicle simulation database as
CONFIG; VVT; VVL; SGDI; DEAC; ADVENG; TRANS; ELEC; ROLL; AERO; MR; EFR;
ELECACC; LDB; SAX), adding a new technology to the vehicle simply
represents progress from a previous state vector to a new state vector.
The previous state vector simply refers to the technologies that are
currently in use on a vehicle. The new state vector, however, is
computed within the modeling system by adding a new technology to the
combination of technologies represented by the previous state vector,
while simultaneously removing any other technologies that are
superseded by the newly added one.
 For example, consider the vehicle with the state vector described
as: SOHC; VVT; AT6; BISG; ROLL10; AERO20; MR1; EPS; LDB. Assume the
system is evaluating PHEV20 as a candidate technology for application
on this vehicle. The new state vector for this vehicle is computed by
removing SOHC, VVT, AT6, and BISG technologies from the previous state
vector,\626\ while also adding PHEV20, resulting in the following:
PHEV20; ROLL10; AERO20; MR1; EPS; LDB.
---------------------------------------------------------------------------
 \626\ For more discussion of how the CAFE Model handles
technology supersession, see Section VI.A.7.
---------------------------------------------------------------------------
 From here, it is relatively simple to obtain a fuel economy
improvement factor for any new combination of technologies and apply
that factor to the fuel economy of a vehicle in the analysis fleet. The
formula for calculating a vehicle's fuel economy after application of
each successive technology represented within the database is defined,
simply put, as the difference between the fuel economy improvement
factor associated with the technology state vector before application
of a candidate technology, and after the application of a candidate
technology.\627\ This is applied to the original compliance fuel
economy value for a discrete vehicle in the MY 2017 analysis fleet, as
discussed previously in Section VI.B.3 Technology Effectiveness.
---------------------------------------------------------------------------
 \627\ For more discussion of how the CAFE Model calculates a
vehicle's fuel economy where the vehicle switches from one type of
fuel to another, for example, from gasoline operation to diesel
operation or from gasoline operation to plug-in hybrid/electric
vehicle operation, see Section VI.A CAFE Model.
---------------------------------------------------------------------------
 The fuel economy improvement factor is defined in a way that
captures the incremental improvement of moving between points in the
database, where each point is defined uniquely as a combination of up
to 15 distinct technologies describing, as mentioned above, the
engine's cam configuration, multiple distinct combinations of engine
technologies, transmission, electrification type, and various vehicle
body level technologies.
 Unlike the preceding versions of the modeling system, the current
version of the CAFE Model relies entirely on the vehicle simulation
database for calculating fuel economy improvements resulting from all
technologies available to the system. The fuel economy improvements are
derived from the factors defined for each unique technology combination
or state vector. Each time the improvement factor for a new state
vector is added to a vehicle's existing fuel economy, the factor
associated with the old technology combination is entirely removed. In
that sense, application of technologies obtained from the Autonomie
database is ``self-correcting'' within the model. As such, special-case
adjustments defined by the previous version of the model are not
applicable to this one.
 Meszler Engineering Services, commenting on behalf of Natural
Resources Defense Council, commented that ``[w]ith very limited
exception, technology is not included in the NPRM CAFE model if it was
not included in the simulation modeling that underlies the Argonne
database,'' citing the ``add-on'' technologies and technologies with
fixed effectiveness values.\628\ Meszler continued, ``[t]his same
limitation controls the coupling of technologies, and by extension the
definition of the CAFE model technology pathways. If a combination of
technologies were not modeled during the development of the Argonne
database, that package (or combination) of technologies is not
available for adoption in the CAFE model. Both of these design
constraints serve to limit the slate of technologies available to
respond to fuel economy
[[Page 24349]]
standards. The slate of available technologies is basically constrained
to those included in NHTSA's research activity. If a technology or
technology combination was not in the NHTSA research planning process,
it is not available in the model.'' Finally, Meszler stated that
``because of the constrained model architecture and the reliance on the
Argonne database for impact estimates, independently expanding the
model to include additional technologies or technology combinations is
not trivial.''
---------------------------------------------------------------------------
 \628\ NHTSA-2018-0067-11723, at 4-5.
---------------------------------------------------------------------------
 We agree that expanding the database to include new technologies is
not trivial. However, it is possible. The set of available technologies
is part of the model code, and the code is made public upon each
release of the model. Many commenters made modifications to the model
code, conducted additional tests of their own, and presented their
results to the agencies in the form of public comments before the end
of the public comment period. A user could add the new technology,
identify the associated engineering restrictions that determine
combinations for which that technology should not be considered, and
add the relevant rows (representing possible technology combinations
that include the new technology) in the database (which exists locally
on every computer that runs the model). An enterprising user could also
take an existing technology along a given path and replace the
efficiency values with new values--presumably from their own full
vehicle simulations for each technology combination that contains the
technology in question. Given the length of time and computing power
required to simulate vehicle fuel economy on the test cycle for every
possible combination that could be considered by the CAFE model, using
a pre-defined database that represents a large ensemble of simulated
technology combinations is preferable to the alternative of fully
integrating a vehicle simulation model that would be required to run in
real-time during the compliance simulation to evaluate the
effectiveness of every combination considered (not just applied) by the
model.
4. Technology Costs
 In the proposal, the agencies estimated present and future costs
for fuel-saving technologies, taking into consideration the type of
vehicle, or type of engine if technology costs vary by application.
These cost estimates are based on three main inputs. First, the
agencies estimated direct manufacturing costs (DMCs), or the component
and labor costs of producing and assembling the physical parts and
systems, with estimated costs assuming high volume production. DMCs
generally do not include the indirect costs of tools, capital
equipment, financing costs, engineering, sales, administrative support
or return on investment. Second, the agencies accounted for these
indirect costs via a scalar markup of direct manufacturing costs (the
retail price equivalent, or RPE). Finally, costs for technologies may
change over time as industry streamlines design and manufacturing
processes. The agencies therefore estimated potential cost improvements
with learning effects (LE). The retail cost of equipment in any future
year is estimated to be equal to the product of the DMC, RPE, and LE.
Considering the retail cost of equipment, instead of merely direct
manufacturing costs, is important to account for the real-world price
effects of a technology, as well as market realities. Absent a
government mandate, motor vehicle manufacturers will not undertake
expensive development and production efforts to implement technologies
without realistic prospects of consumers being willing to pay enough
for such technology to allow for the manufacturers to recover their
investment.
a) Direct Manufacturing Costs
 Direct manufacturing costs (DMCs) are the component costs of the
physical parts and systems that make up a complete vehicle. The
analysis used agency-sponsored tear-down studies of vehicles and parts
to estimate the DMCs of individual technologies, in addition to
independent tear-down studies, other publications, and confidential
business information. In the simplest cases, the agency-sponsored
studies produced results that confirmed third-party industry estimates,
and aligned with confidential information provided by manufacturers and
suppliers. In cases with a large difference between the tear-down study
results and credible independent sources, study assumptions were
scrutinized, and sometimes the analysis was revised or updated
accordingly.
 Due to the variety of technologies and their applications, and the
cost and time required to conduct detailed tear-down analyses, the
agencies did not sponsor teardown studies for every technology. In
addition, many fuel-saving technologies were considered that are pre-
production, or sold in very small pilot volumes. For those
technologies, a tear-down study could not be conducted to assess costs
because the product is not yet in the marketplace for evaluation. In
these cases, the agencies relied upon third-party estimates and
confidential information from suppliers and manufacturers were relied
upon; however, there are some common pitfalls with relying on
confidential business information to estimate costs. The agencies and
the source may have had incongruent or incompatible definitions of
``baseline.'' The source may have provided DMCs at a date many years in
the future, and assumed very high production volumes, important caveats
to consider for agency analysis. In addition, a source, under no
contractual obligation to the agencies, may provide incomplete and/or
misleading information. In other cases, intellectual property
considerations and strategic business partnerships may have contributed
to a manufacturer's cost information and could be difficult to account
for in the model as not all manufacturer's may have access to
proprietary technologies at stated costs. The agencies carefully
evaluated new information in light of these common pitfalls, especially
regarding emerging technologies.
 Specifically, the analysis used third-party, forward-looking
information for advanced cylinder deactivation and variable compression
ratio engines. While these cost estimates may be preliminary (as is the
case with many emerging technologies prior to commercialization), the
agencies consider them to be reasonable estimates of the likely costs
of these technologies.
 While costs for fuel-saving technologies reflect the best estimates
available today, technology cost estimates will likely change in the
future as technologies are deployed and as production is expanded. For
emerging technologies, the best information available at the time of
the analysis was utilized, and cost assumptions will continue to be
updated for any future analysis. Below, discussion of each category of
technologies (e.g., engines, transmissions, electrification) summarizes
comments on corresponding direct cost estimates, and reviews estimates
the agencies have applied for today's analysis.
Indirect Costs
 As discussed above, direct costs represent the cost associated with
acquiring raw materials, fabricating parts, and assembling vehicles
with the various technologies manufacturers are expected to use to meet
future CAFE and CO2 standards. They include materials,
labor, and variable energy costs required to produce and assemble the
vehicle. However, they do not
[[Page 24350]]
include overhead costs required to develop and produce the vehicle,
costs incurred by manufacturers or dealers to sell vehicles, or the
profit manufacturers and dealers make from their investments. All of
these items contribute to the price consumers ultimately pay for the
vehicle. These components of retail prices are illustrated in Table VI-
23 below.
[GRAPHIC] [TIFF OMITTED] TR30AP20.115
 In addition to direct manufacturing costs, the agencies estimated
and considered indirect manufacturing costs. To estimate indirect
costs, direct manufacturing costs are multiplied by a factor to
represent the average price for fuel-saving technologies at retail.
 In the Draft TAR and preceding CAFE and safety rulemaking analyses,
NHTSA relied on a factor, referred to as the retail price equivalent
(RPE), to account for indirect manufacturing costs. The RPE accounts
for indirect costs like engineering, sales, and administrative support,
as well as other overhead costs, business expenses, warranty costs, and
return on capital considerations. In the Draft TAR (and subsequent
Determination) as well as the 2012 rulemaking analysis, EPA applied an
``Indirect Cost Multiplier'' (ICM) approach that it first applied in
the 2010 rulemaking regarding standards for MYs 2012-2016, which also
accounted for indirect manufacturing costs, albeit in a different way
than the RPE approach.
 Some commenters recommended the agencies rely on the ICM approach
for the current rulemaking, citing EPA's prior peer review and use of
this approach.\629\ Others supported the agencies' reliance on the RPE
approach, citing the National Research Council's observations in 2015
that the ICM approach lacks an empirical basis.\630\ The agencies have
carefully considered these comments, and conclude that while the ICM
approach has conceptual merit, its application requires a range of
specific estimates, and data to support such estimates is scant and, in
some cases, nonexistent. The agencies have, therefore, applied the RPE
approach for this final rule, as in the NPRM analysis and other
rulemaking analyses. The following sections discuss both approaches in
detail to explain why the RPE approach was chosen for this final rule.
---------------------------------------------------------------------------
 \629\ See, e.g., ICCT, NHTSA-2018-0067-11741, Attachment 3, at
I-83. See also CFA, NHTSA-2018-0067-12005, Attachment B, at p.189.
 \630\ See, e.g., Alliance, NHTSA-2018-0067-12073, at 143. See
also National Research Council, ``Cost, Effectiveness, and
Deployment of Fuel Economy Technologies for Light-Duty Vehicles,''
2015, available at https://www.nap.edu/catalog/21744/cost-effectiveness-and-deployment-of-fuel-economy-technologies-for-lightduty-vehicles (``. . . the empirical basis for such multipliers
is still lacking, and, since their application depends on expert
judgment, it is not possible for to determine whether the Agencies'
ICMs are accurate or not'').
---------------------------------------------------------------------------
(1) Retail Price Equivalent
 Historically, the method most commonly used to estimate indirect
costs of producing a motor vehicle has been the retail price equivalent
(RPE). The RPE markup factor is based on an examination of historical
financial data contained in 10-K reports filed by manufacturers with
the Securities and Exchange Commission (SEC). It represents the ratio
between the retail price of motor vehicles and the direct costs of all
activities that manufacturers engage in, including the design,
development, manufacturing, assembly,
[[Page 24351]]
and sales of new vehicles, refreshed vehicle designs, and modifications
to meet safety or fuel economy standards.
 Figure VI-7 indicates that for more than three decades, the retail
price of motor vehicles has been, on average, roughly 50 percent above
the direct cost expenditures of manufacturers. This ratio has been
remarkably consistent, averaging roughly 1.5 with minor variations from
year to year over this period. At no point has the RPE markup exceeded
1.6 or fallen below 1.4.\631\ During this time frame, the average
annual increase in real direct costs was 2.5 percent, and the average
annual increase in real indirect costs was also 2.5 percent. Figure VI-
7 illustrates the historical relationship between retail prices and
direct manufacturing costs.\632\
---------------------------------------------------------------------------
 \631\ Based on data from 1972-1997 and 2007. Data were not
available for intervening years, but results for 2007 seem to
indicate no significant change in the historical trend.
 \632\ Rogozhin, A., Gallaher, M., & McManus, W., 2009,
Automobile Industry Retail Price Equivalent and Indirect Cost
Multipliers. Report by RTI International to Office of Transportation
Air Quality. U.S. Environmental Protection Agency, RTI Project
Number 0211577.002.004, February, Research Triangle Park, N.C.
Spinney, B.C., Faigin, B., Bowie, N., & St. Kratzke, 1999, Advanced
Air Bag Systems Cost, Weight, and Lead Time analysis Summary Report,
Contract NO. DTNH22-96-0-12003, Task Orders--001, 003, and 005.
Washington, DC, U.S. Department of Transportation.
---------------------------------------------------------------------------
 An RPE of 1.5 does not imply that manufacturers automatically mark
up each vehicle by exactly 50 percent. Rather, it means that, over
time, the competitive marketplace has resulted in pricing structures
that average out to this relationship across the entire industry.
Prices for any individual model may be marked up at a higher or lower
rate depending on market demand. The consumer who buys a popular
vehicle may, in effect, subsidize the installation of a new technology
in a less marketable vehicle. But, on average, over time and across the
vehicle fleet, the retail price paid by consumers has risen by about
$1.50 for each dollar of direct costs incurred by manufacturers.
[GRAPHIC] [TIFF OMITTED] TR30AP20.116
 It is also important to note that direct costs associated with any
specific technology will change over time as some combination of
learning and resource price changes occurs. Resource costs, such as the
price of steel, can fluctuate over time and can experience real long-
term trends in either direction, depending on supply and demand.
However, the normal learning process generally reduces direct
production costs as manufacturers refine production techniques and seek
out less costly parts and materials for increasing production volumes.
By contrast, this learning process does not generally influence
indirect costs. The implied RPE for any given technology would thus be
expected to grow over time as direct costs decline relative to indirect
costs. The RPE for any given year is based on direct costs of
technologies at different stages in their learning cycles, and which
may have different implied RPEs than they did in previous years. The
RPE averages 1.5 across the lifetime of technologies of all ages, with
a lower average in earlier years of a technology's life, and, because
of learning effects on direct costs, a higher average in later years.
 The RPE has been used in all NHTSA safety and most previous CAFE
rulemakings to estimate costs. The National Academy of Sciences
recommends RPEs of 1.5 for suppliers and 2.0 for in-house production be
used to estimate total costs. The Alliance of Automobile Manufacturers
also advocates these values as appropriate markup factors for
estimating costs of technology changes. An RPE of 2.0 has also been
adopted by a coalition of environmental and research groups (NESCCAF,
ICCT, Southwest Research Institute, and TIAX-LLC) in a report on
reducing heavy truck emissions, and 2.0 is recommended by the U.S.
Department of Energy for estimating the cost of hybrid-electric and
automotive fuel cell costs ((see Vyas et al. (2000) in Table VI-24,
below).
 Table VI-24 below lists other estimates of the RPE. Note that all
RPE estimates vary between 1.4 and 2.0, with most in the 1.4 to 1.7
range.
[[Page 24352]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.117
 The RPE has thus enjoyed widespread use and acceptance by a variety
of governmental, academic, and industry organizations. The RPE has been
the most commonly used basis for indirect cost markups in regulatory
analyses. However, as noted above, the RPE is an aggregate measure
across all technologies applied by manufacturers and is not technology
specific. A more detailed examination of these technologies is possible
through an alternative measure, the indirect cost multiplier, which was
developed to focus more specifically on technologies used to meet CAFE
and CO2 standards.
---------------------------------------------------------------------------
 \633\ Duleep, K.G. ``2008 Analysis of Technology Cost and Retail
Price.'' Presentation to Committee on Assessment of Technologies for
Improving Light Duty Vehicle Fuel Economy, January 25, Detroit, MI.;
Jack Faucett Associates, September 4, 1985. Update of EPA's Motor
Vehicle Emission Control Equipment Retail Price Equivalent (RPE)
Calculation Formula. Chevy Chase, MD--Jack Faucett Associates;
McKinsey & Company, October 2003. Preface to the Auto Sector Cases.
New Horizons--Multinational Company Investment in Developing
Economies, San Francisco, CA.; NRC (National Research Council),
2002. Effectiveness and Impact of Corporate Average Fuel Economy
Standards, Washington, DC--The National Academies Press; NRC, 2011.
Assessment of Fuel Economy Technologies for Light Duty Vehicles.
Washington, DC--The National Academies Press; Sierra Research, Inc.,
November 21, 2007, Study of Industry-Average Mark-Up Factors used to
Estimate Changes in Retail Price Equivalent (RPE) for Automotive
Fuel Economy and Emissions Control Systems, Sacramento, CA--Sierra
Research, Inc.; Vyas, A. Santini, D., & Cuenca, R. 2000. Comparison
of Indirect Cost Multipliers for Vehicle Manufacturing. Center for
Transportation Research, Argonne National Laboratory, April.
Argonne, Ill.
---------------------------------------------------------------------------
(2) Indirect Cost Multiplier
 A second approach to accounting for indirect costs is the indirect
cost multiplier (ICM). ICMs specifically evaluate the components of
indirect costs likely to be affected by vehicle modifications
associated with environmental regulation. EPA developed the ICM concept
to enable the application of markups more specific to each technology.
For example, the indirect cost implications of using tires with better
rolling resistance would not be the same as those for developing an
entire new hybrid vehicle technology, which would require far more R&D,
capital investment, and management oversight. With more than 80
different technologies available to incrementally achieve fuel economy
improvements,\634\ a wide range of indirect cost effects might be
expected. ICMs attempt to isolate only those indirect costs that would
have to change to develop a specific technology. Thus, for example, if
a company were to hire additional staff to sell vehicles equipped with
fuel economy improving technology, or to search the technology
requirements of new CO2 or CAFE standards, the cost of these
staff would be included in ICMs. However, if these functions were
accomplished by existing staff, they would not be included. For
example, if an executive who normally devoted 10 percent of his time to
fuel economy standards compliance were to devote 50 percent of his time
in response to new more stringent requirements, his salary would not be
included in ICMs because he would be paid the same salary regardless of
whether he devoted his time to addressing CAFE requirements, developing
new performance technologies, or improving the company's market share.
ICMs thus do not account for the diverted resources required for
manufacturers to meet these standards, but rather for the net change in
costs manufacturers might experience because of hiring additional
personal or acquiring additional assets or services.
---------------------------------------------------------------------------
 \634\ There are roughly 40 different basic unique technologies,
but variations among these technologies roughly double the possible
number of different technology applications.
---------------------------------------------------------------------------
 For past rulemakings EPA developed both short-term and long-term
ICMs. Long-term ICMs are lower than short-term ICMs. This decline
reflects the belief that many indirect costs will decline over time.
For example, research is initially required to develop a new technology
and apply it throughout the vehicle fleet, but a lower level of
research will be required to improve, maintain, or adapt that new
technology to subsequent vehicle designs.
 While the RPE was derived from data in financial statements
(reflecting real-world operating and financial results), no similar
data sources were available to estimate ICMs. ICMs are based on the
RPE, broken into its components, as shown in Table VI-25. Adjustment
factors were then developed for those components, based on the
complexity and time frame of low-, medium-, and high-complexity
technologies. The adjustment factors were developed from two panels of
engineers with background in the automobile industry. Initially, a
group of engineers met and developed an estimate of ICMs for three
different technologies. This ``consensus'' panel examined one low
complexity technology, one medium complexity technology, and one high
complexity technology, with the initial intent of using these
technologies to represent ICM factors for all technologies falling in
those categories. At a later date, a second panel was convened to
examine three more technologies (one low, one medium, and one high
complexity), using a modified Delphi approach to estimate indirect cost
effects. The results from the second panel identified the same pattern
as those of the original report--the indirect cost multipliers increase
with the
[[Page 24353]]
complexity of the technology and decrease over time. The values derived
in process are higher than those in the RPE/IC Report by values ranging
from 0.09 (that is, the multiplier increased from 1.20 to 1.29) to 0.19
(the multiplier increased from 1.45 to 1.64). This variation may be due
to differences in the technologies used in each panel. The results are
shown in Figure VI-8, together with the historical average RPE.
[GRAPHIC] [TIFF OMITTED] TR30AP20.118
 In subsequent CAFE and CO2 analyses for MYs 2011, as
well as for the 2012-2016 rulemaking, a simple average of the two
resulting ICMs in the low and medium technology complexity categories
was applied to direct costs for all unexamined technologies in each
specific category. For high complexity technologies, the lower
consensus-based estimate was used for high complexity technologies
currently being produced, while the higher modified Delphi-based
estimate was used for more advanced technologies, such as plug-in
hybrid or electric vehicles, which had little or no current market
penetration. Note that ICMs originally did not include profit or
``return on capital,'' a fundamental difference from the RPE. However,
prior to the 2012-2016 CAFE analysis, ICMs were modified to include
provision for return on capital.
(3) Application of ICMs in the 2017-2025 Analysis
 For the model year 2017-2025 rulemaking analysis, NHTSA and EPA
revisited technologies evaluated by EPA staff and reconsidered their
method of application. The agencies were concerned that averaging
consensus and modified Delphi ICMs might not be the most accurate way
to develop an estimate for the larger group of unexamined technologies.
Specifically, there was concern that some technologies might not be
representative of the larger groups they were chosen to represent.
Further, the agencies were concerned that the values developed under
the consensus method were not subject to the same analytical discipline
as those developed from the modified Delphi method. As a result, the
agencies relied primarily on the modified Delphi-based technologies to
establish their revised distributions. Thus, for the MY 2017-2025
analysis, the agencies used the following basis for estimating ICMs:
 All low complexity technologies were estimated to equal
the ICM of the modified Delphi-based low technology-passive aerodynamic
improvements.
 All medium complexity technologies were estimated to equal
the ICM of the modified Delphi-based medium technology-engine turbo
downsizing.
 Strong hybrids and non-battery plug-in hybrid electric
vehicles (PHEVs) were estimated to equal the ICM of the high complexity
consensus-based high technology-hybrid electric vehicle.
 PHEVs with battery packs and full electric vehicles were
estimated to equal the ICM of the high complexity modified Delphi-based
high technology-plug-in hybrid electric vehicle.
 In addition to shifting the proxy basis for each technology group,
the agencies reexamined each technology's complexity designation in
light of the examined technologies that would serve as the basis for
each group. The resulting designations together with the associated
proxy technologies are shown in Table VI-25.
[[Page 24354]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.119
 Many basic technologies noted in Table VI-25 have variations
sharing the same complexity designation and ICM estimate. Table VI-26
lists each technology used in the CAFE model together with their ICM
category and the year through which the short-term ICM would be
applied. Note that the number behind each ICM category designation
refers to the source of the ICM estimate, with 1 indicating the
consensus panel and 2 indicating the modified Delphi panel.
BILLING CODE 4910-59-P
[[Page 24355]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.120
[[Page 24356]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.121
[[Page 24357]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.122
[[Page 24358]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.123
BILLING CODE 4910-59-C
 An additional adjustment was made to ICMs to account for the fact
that they were derived from the RPE analysis for a specific year
(2007). The agencies believed it would be more appropriate to base ICMs
on the expected long-term average RPE rather than that of one specific
year. To account for this, ICMs were normalized to an average RPE
multiplier level of 1.5.
 Table VI-27 lists values of ICMs by technology category used in the
previous MY 2017-2025 rulemaking. As noted previously, the Low 1 and
Medium 1 categories, which were derived using the initial consensus
panel, are not used. Short-term values applied to CAFE technologies
thus range from 1.24 for Low complexity technologies, 1.39 for Medium
complexity technologies, 1.56 for High1 complexity technologies, and
1.77 for High2 complexity technologies. When long-term ICMs are applied
in the year following that noted in the far-right column of Table VI-
27, these values will drop to 1.19 for Low, 1.29 for Medium, 1.35 for
High1 and 1.50 for High2 complexity technologies.
[GRAPHIC] [TIFF OMITTED] TR30AP20.124
 Note that ICMs for warranty costs are listed separately in Table
VI-27. This was done because warranty costs are treated differently
than other indirect costs. In some previous analyses (prior to MY 2017-
2025), learning was applied directly to total costs. However, the
agencies believe learning curves are more appropriately applied only to
direct costs, with indirect costs established up front based on the ICM
and held constant while direct costs are reduced by learning.
Warranties are an exception to this because warranty costs involve
future replacement of defective parts, and the cost of these parts
would reflect the effect of learning. Warranty costs were thus treated
as being subject to learning along with direct costs.\635\
---------------------------------------------------------------------------
 \635\ Note that warranty costs also involve labor costs for
installation. This is typically done at dealerships, and it is
unlikely labor costs would be subject to learning curves that affect
motor vehicle parts or assembly costs. However, the portion of these
costs that is due to labor versus that due to parts is unknown, so
for this analysis, learning is applied to the full warranty cost.
---------------------------------------------------------------------------
 The effect of learning on direct costs, together with the eventual
substitution of lower long-term ICMs, causes the effective markup from
ICMs to differ from the initial ICM on a yearly basis. An example of
how this occurs is provided in Table VI-28.\636\ This table, which was
originally developed for the MY 2017-2025 analysis, traces the effect
of learning on direct costs and its implications for both total costs
and the ICM-based markup. Direct costs are assigned a value
(proportion) of 1 to facilitate analysis on the same basis as ICMs (in
an ICM markup factor, the proportion of direct costs is represented by
1 while the proportion of indirect costs is represented by the fraction
of 1 to the right of the decimal.) Table VI-28 examines the effects of
these factors on turbocharged downsized engines, one of the more
prevalent CAFE technologies.
---------------------------------------------------------------------------
 \636\ Table VI-22 illustrates the learning process from the base
year consistent with the direct cost estimate obtained by the
agencies. It is a mature technology well into the flat portion of
the learning curve. Note that costs were actually applied in this
rulemaking example beginning with MY 2017.
---------------------------------------------------------------------------
[[Page 24359]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.125
 The second column of Table VI-28 lists the learning schedule
applied to turbocharged downsized engines. Turbocharged downsized
engines are a mature technology, so the learning schedule captures the
relatively flat portion of the learning curve occurring after larger
decreases have already reduced direct costs. The cost basis for
turbocharged downsized engines in the analysis was effective in 2012,
so this is the base year for this calculation when direct costs are set
to 1. The third column shows the progressive decline in direct costs as
the learning schedule in column 2 is applied to direct costs. Column 4
contains the value of all indirect costs except warranty. Turbocharged
downsized engines are a medium-complexity technology, so this value is
taken from the Medium2 row of Table VI-27. The initial value in 2012 is
the short-term value, which is used through 2018. During this time,
these indirect costs are not affected by learning, and they remain
constant. Beginning in 2019, the long-term ICM from Table VI-27 is
applied.
 The fifth column contains warranty costs. As previously mentioned,
these costs are considered to be affected by learning like direct
costs, so they decline steadily until the long-term ICM is applied in
2019, at which point they drop noticeably before continuing their
gradual decline. In the sixth column, direct and indirect costs are
totaled. Results indicate a decline in total costs of roughly 30
percent during this 14-year period. The last column shows the effective
ICM-based markup, which is derived by dividing total costs by direct
costs. Over this period, the ICM-based markup rose from the initial
short-term ICM level of 1.39 to 1.45 in 2018. It then declined to 1.35
in 2019 when the long-term ICM was applied to the 2019 direct cost.
Over the remaining years, it gradually rises back up to 1.41 as
learning continues to degrade direct costs.
 There are thus two somewhat offsetting processes affecting total
costs derived from ICMs. The first is the learning curve, which reduces
direct costs, which raises the effective ICM-based markup. As noted
previously, learning reflects learned efficiencies in assembly methods
as well as reduced parts and materials costs. The second is the
application of a long-term ICM, which reduces the effective ICM-based
markup. This represents the reduced burden needed to maintain new
technologies once they are fully developed. In this case, the two
processes largely offset one another and produce an average real ICM
over this 14-year period that roughly equals the original short-term
ICM.
 Figure VI-9 illustrates this process for each of the 4 technologies
used to represent the universe of fuel economy and CO2
improving technologies. As with the turbocharged engines, aerodynamic
improvements and mild hybrid vehicles show a gradual increase in the
effective ICM-based markup through the point where the long-term ICM is
applied. At that time, the ICM-based markup makes an abrupt decline
before beginning a gradual rise. The decline due to application of
long-term ICMs is particularly pronounced in the case of the mild
hybrid--even more so than for the advanced hybrid. The advanced hybrid
ICM behaves somewhat differently because it is shown through its
developing stages when more radical learning is applied, but only every
few years. This produces a significant step-up in ICM levels concurrent
with each learning
[[Page 24360]]
application, followed by a sharp decline when the long-term ICM is
applied. After that, it begins a gradual rise as more moderate learning
is applied to reflect its shift to a mature technology. Note that as
with the turbocharged downsized engine example above, for the
aerodynamic improvements and mild hybrid technologies, the offsetting
processes of learning and long-term ICMs result in an average ICM over
the full time frame that is roughly equal to the initial short-term
ICM. However, the advanced hybrid ICM rose to a level significantly
higher than the initial ICM. This is a direct function of the rapid
learning schedule applied in the early years to this developing
technology. Brand new technologies might thus be expected to have
effective lifetime ICM markups exceeding their initial ICMs, while more
mature technologies are more likely to experience ICMs over their
remaining life span that more closely approximate their initial ICMs.
[GRAPHIC] [TIFF OMITTED] TR30AP20.126
 ICMs for these 4 technologies would drive the indirect cost markup
rate for the analysis. However, the effect on total costs is also a
function of the relative incidence of each of the 50+ technologies
shown in Table VI-26 which are assumed to have ICMs similar to one of
these 4 technologies. The net effect on costs of these ICMs is also
influenced by the learning curve appropriate to each technology,
creating numerous different and unique ICM paths. The average ICM
applied by the model is also a function of each technology's direct
cost and because ICMs are applied to direct costs, the measured
indirect cost is proportionately higher for any given ICM when direct
costs are higher. The average ICM applied to the fleet for any given
model year is calculated as follows:
[GRAPHIC] [TIFF OMITTED] TR30AP20.127
where:
D = direct cost of each technology
A = application rate for each technology
ICM = average ICM applied to each technology
and n = 1, 2 . . . . 88
 The CAFE model predicts technology application rates assuming
manufacturers will apply technologies
[[Page 24361]]
to meet standards in a logical fashion based on estimated costs and
benefits. The application rates will thus be different for each model
year and for each alternative scenario examined. For the MY 2017-2025
FRIA, to illustrate the effects of ICMs on total technology costs,
NHTSA calculated the weighted average ICM across all technologies for
the preferred alternative.\637\ This was done separately for each
vehicle type and then aggregated based on predicted sales of each
vehicle type used in the model. Results are shown in Table VI-29.
---------------------------------------------------------------------------
 \637\ For each alternative, this rulemaking examined numerous
scenarios based on different assumptions, and these assumptions
could influence the relative frequency of selection of different
technologies, which in turn could affect the average ICM. The
scenario examined here assumed a 3 percent discount rate, a 1-year
payback period, real world application of expected civil penalties,
and reflects expected voluntary over-compliance by manufacturers.
[GRAPHIC] [TIFF OMITTED] TR30AP20.128
 The ICM-based markups in Table VI-29 were derived in a manner
consistent with the way the RPE is measured, that is, they reflect
combined influences of direct cost learning and changes in indirect
cost requirements weighted by both the incidence of each technology's
adaptation and the relative direct cost of each technology. The results
indicate generally higher ICMs for passenger cars than for light
trucks. This is a function of the technologies estimated to be adopted
for each respective vehicle type, especially in later years when
hybrids and electric vehicles become more prevalent in the passenger
car fleet. The influence of these advanced vehicles is driven primarily
by their direct costs, which greatly outweigh the costs of other
technologies. This results in the application of much more weight to
their higher ICMs. This is most notable in MYs 2024 and 2025 for
passenger cars, when electric vehicles begin to enter the fleet. The
average ICM increased 0.013 in 2024 primarily because of these
vehicles. It immediately dropped 0.017 in 2025 because both an
additional application of steep (20 percent) learning is applied to the
direct cost of these vehicles (which reduces their relative weight),
and the long-term ICM becomes effective in that year (which decreases
the absolute ICM factor). Both influences occur one year after these
vehicles begin to enter the fleet because of CAFE requirements.
 ICMs also change over time, again, reflecting the different mix of
technologies present during earlier years but that are often replaced
with more expensive technologies in later years. Across all model
years, the wide-ranging application of diverse technologies required to
meet CAFE and CO2 standards produced an average ICM-based
markup (or RPE equivalent) of approximately 1.34, applying only 67
percent of the indirect costs found in the RPE and implying total costs
11 percent below those predicted by the RPE-based calculation.
(4) Uncertainty
 As noted above, the RPE and ICM assign different markups over
direct manufacturing costs, and thus imply different total cost
estimates for CAFE and CO2 technologies. While there is a
level of uncertainty associated with both markups, this uncertainty
stems from different issues. The RPE is derived from financial
statements and is thus grounded in historical data. Although
compilation of this data is subject to some level of interpretation,
the two independent researchers who derived RPE estimates from these
financial reports each reached essentially identical conclusions,
placing the RPE at roughly 1.5. All other estimates of the RPE fall
between 1.4 and 2.0, and most are between 1.4 and 1.7. There is thus a
reasonable level of consistency among researchers that RPEs are 1.4 or
greater. In addition, the RPE is a measure of the cumulative effects of
all operations manufacturers undertake in the course of producing their
vehicles, and is thus not specific to individual technologies, nor of
CAFE or CO2 technologies in particular. Because this
provides only a single aggregate measure, using the RPE multiplier
results in the application of a common incremental markup to all
technologies. This assures the aggregate cost effect across all
technologies is consistent with empirical data, but it does not allow
for indirect cost discrimination among different technologies or over
time. Because it is applied across all changes, this implies the markup
for some technologies is likely to be understated, and for others it is
likely to be overstated.
 By contrast, the ICM process derives markups specific to several
CAFE and CO2 technologies, but these markups
[[Page 24362]]
have no basis in empirical data. They are based on informed judgment of
a panel of engineers with auto industry experience regarding cost
effects of a small sample (roughly 8 percent) of the 50+ technologies
applied to achieve compliance with CAFE and CO2 standards.
Uncertainty regarding ICMs is thus based both on the accuracy of the
initial assessments of the panel on the examined technologies and on
the assumption that these 4 technologies are representative of the
remaining technologies that were not examined. Both agencies attempted
to categorize these technologies in the most representative way
possible. However, while this represented the best judgment of EPA and
NHTSA's engineering staffs at that time, the actual effect on indirect
costs remains uncertain for most technologies. As with RPEs, this means
that even if ICMs were accurate for the specific technologies examined,
indirect cost will be understated for some technologies and overstated
for others.
 There was considerable uncertainty demonstrated in the ICM panel's
assessments, as illustrated by the range of estimates among the 14
modified Delphi panel members surrounding the central values reported
by the panel. These ranges are shown in Table VI-30 and Figure VI-10,
Figure VI-11, and Figure VI-12 below. For the low complexity
technology, passive aerodynamic improvements, panel responses ranged
from a low of basically no indirect costs (1.001 short term and 1.0
long term), to a high of roughly a 40 percent markup (1.434 and 1.421).
For the medium complexity technology, turbo charged and downsized
engines, responses ranged from a low estimate implying almost no
indirect cost (1.018 and 1.011), to a high estimate implying that
indirect costs for this technology would roughly equal the average RPE
(1.5) for all technologies (1.527 and 1.445). For the high complexity
technology, plug-in hybrid electric vehicles, responses ranged from a
low estimate that these vehicles would require significantly less
indirect cost than the average RPE (1.367 and 1.121) to a high estimate
implying they would require more indirect costs than the average RPE
(2.153 and 1.691). There was considerable diversity of opinion among
the panel members.\638\ This is apparent in Figure VI-10, Figure VI-11,
and Figure VI-12, which show the 14 panel members' final estimates for
short-term ICMs as scatter plots.
---------------------------------------------------------------------------
 \638\ Sample confidence intervals, which mitigate the effect of
outlying opinions, indicate a less extreme but still significant
range of ICMs. Applying mean ICMs helps mitigate these potential
differences, but there is clearly a significant level of uncertainty
regarding indirect costs. A t-distribution is used to estimate
confidence intervals because of the small sample size (14 panel
members).
[GRAPHIC] [TIFF OMITTED] TR30AP20.129
[[Page 24363]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.130
[GRAPHIC] [TIFF OMITTED] TR30AP20.131
[[Page 24364]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.132
 Although these results were based on modified Delphi panel
techniques, it is apparent the goal of the Delphi process, an eventual
consensus or convergence of opinion among panel experts, was not
achieved. Given this lack of consensus and the divergence of ICM-based
results from the only available empirical measure (the RPE), there is
considerable uncertainty that current ICM estimates provide a realistic
basis of estimating indirect costs. ICMs have not been validated
through a direct accounting of actual indirect costs for individual
technologies, and they produce results that conflict with the only
available empirical evidence of indirect cost markups. Further, they
are intended to represent indirect costs specifically associated with
the most comprehensive redesign effort ever undertaken by the auto
industry, with virtually every make/model requiring ground-up design
modifications to comply. This includes entirely new vehicle design
concepts, extensive material substitution, and complete drivetrain
redesigns, all of which require significant research efforts and
assembly plant redesign. Under these circumstances, one might expect
indirect costs to equal or possibly increase above the historical
average, but not to decrease, as implied by estimated ICMs. For
regulations, such as the CAFE and CO2 emission standards
under consideration, that drive changes to nearly every vehicle system,
the overall average indirect costs should align with the RPE value.
Applying RPE to the cost for each technology assures that alignment.
 In the 2015 NAS study, the Committee stated a conceptual agreement
with the ICM method because ICM takes into account design challenges
and the activities required to implement each technology. However,
although endorsing ICMs as a concept, the NAS Committee stated ``the
empirical basis for such multipliers is still lacking, and, since their
application depends on expert judgment, it is not possible to determine
whether the Agencies' ICMs are accurate or not.'' \639\ NAS also stated
``the specific values for the ICMs are critical because they may affect
the overall estimates of costs and benefits for the overall standards
and the cost effectiveness of the individual technologies.'' \640\ The
Committee encouraged continued research into ICMs given the lack of
empirical data for them to evaluate ICMs used by the agencies in past
analyses. On balance, and considering the relative merits of both
approaches for realistically estimating indirect costs, the agencies
consider the RPE method to be a more reliable basis for estimating
indirect costs.
---------------------------------------------------------------------------
 \639\ National Research Council of the National Academies
(2015). Cost, Effectiveness, and Deployment of Fuel Economy
Technologies for Light-Duty Vehicles. https://www.nap.edu/resource/21744/deps_166210.pdf.
 \640\ Ibid.
---------------------------------------------------------------------------
(5) Using RPE To Evaluate Indirect Costs in This Analysis
 To ensure overall indirect costs in the analysis align with the
historical RPE value, the primary analysis has been developed based on
applying the RPE value of 1.5 to each technology. As noted previously,
the RPE is the ratio of aggregate retail prices to aggregate direct
manufacturing costs. The ratio already reflects the mixture of learned
costs of technologies at various stages of maturity. Therefore, the RPE
is applied directly to the learned direct cost for each technology in
each year. This was previously done in the MY 2017-2025 FRIA for the
preferred alternative for that rulemaking, used in the above analysis
of average ICMs. Results are shown in Table VI-31.
 Recognizing there is uncertainty in any estimate of indirect costs,
a sensitivity analyses of indirect costs has also been conducted by
applying a lower RPE value as a proxy for the ICM approach. This value
was derived from a direct comparison of incremental technology costs
determined in the MY 2017-2025 FRIA.\641\ This analysis is summarized
in Table VI-31 below. From this table, total costs were estimated to be
roughly 18 percent lower using ICMs compared to the RPE. As previously
mentioned, there are two different reasons for these differences. The
first is the direct effect of applying a higher retail markup. The
second is an indirect effect resulting from the influence these
differing markups have on the order of the selection of technologies in
the CAFE model, which can change as different direct cost levels
interact with altered retail markups, shifting their relative overall
effectiveness.
---------------------------------------------------------------------------
 \641\ See Table 5-9a in Final Regulatory Impact Analysis,
Corporate Average Fuel Economy for MY 2017-MY 2025 Passenger Cars
and Light Trucks.
---------------------------------------------------------------------------
 The relative effects of ICMs may vary somewhat by scenario, but in
this case, the application of ICMs produces total
[[Page 24365]]
technology cost estimates roughly 18 percent lower than those that
would result from applying a single RPE factor to all technologies, or,
conversely, the RPE produces estimates that averaged 21 percent higher
than the ICM. Under the CAFE model construct, which will apply an
alternate RPE to the same base technology profile to represent ICMs,
this implies an RPE equivalent of 1.24 would produce similar net
impacts [1.5/(1 + x) = 1.21, x = 0.24]. This value is applied for the
ICM proxy estimate. Additional values were also examined over a range
of 1.1-2.0. The results, as well as the reference case using the 1.5
RPE, are summarized in Table VI-32.
[GRAPHIC] [TIFF OMITTED] TR30AP20.133
[GRAPHIC] [TIFF OMITTED] TR30AP20.134
 Several responders submitted comments on the issue of indirect
costs. The International Council on Clean Transportation (ICCT) stated
that ``The agencies abandoned their previously-used indirect cost
multiplier method for estimating total costs, which was vetted with
peer review, and more complexly handled differing technologies with
different supply chain and manufacturing aspects. The agencies have, at
this point, opted to use a simplistic retail price equivalent method,
which crudely assumes all technologies have a 50 percent markup from
the direct manufacturing technology cost. We recommend the agencies
revert back to the previously-used and better substantiated ICM
approach.'' \642\
---------------------------------------------------------------------------
 \642\ NHTSA-2018-0067-11741.
---------------------------------------------------------------------------
 A private commenter, Thomas Stephens, noted that ``In Section II.
Technical Foundation for NPRM Analysis, under 1. Data Sources and
Processes for Developing Individual Technology Assumptions, the
agencies state that indirect costs are estimated using a Retail Price
Equivalent (RPE) factor. Concerns with RPE factors and the difficulty
of accounting for differences in indirect costs of different
technologies when using this approach were identified by the EPA
(Rogozhin et al., Using indirect cost multipliers to estimate the total
cost of adding new technology in the automobile industry, International
Journal of Production Economics 124, 360-368, 2010), which suggested
using indirect cost (IC) multipliers instead of RPE factors. The EPA
developed and updated IC multipliers for relevant vehicle technologies
with automotive industry input and review. The agencies should consider
using these IC multipliers to estimate indirect manufacturing costs
instead of RPE factors.'' \643\
---------------------------------------------------------------------------
 \643\ NHTSA-2018-0067-12067.
---------------------------------------------------------------------------
 By contrast, the Alliance of Automobile Manufacturers (The
Alliance) ``supports the use of retail
[[Page 24366]]
price equivalents in the compliance cost modeling to estimate the
indirect costs associated with the additional added technology required
to meet a given future standard. The alternative indirect cost
multiplier (``ICM'') approach is not sufficiently developed for use in
rulemaking. As noted by the National Research Council, the indirect
cost multipliers previously developed by EPA have not been validated
with empirical data.\644\ Furthermore, in reference to the memorandum
documenting the development of ICMs previously used by EPA, Exponent
Failure Analysis Associates found that,
---------------------------------------------------------------------------
 \644\ Cost, Effectiveness, and Development of Fuel Economy
Technologies for Light-Duty Vehicles, pages 248-49, National
research Council, the National Academies Press (2015).
---------------------------------------------------------------------------
Past Toyota Comments on Atkinson-Cycle Benefits Have Addressed Only
Those Derived From Variable Valve Timing
 In response to these comments the agencies continue to find the RPE
approach preferable to the ICM approach, at least at this stage in the
development ICM estimates, for the reasons discussed both above and
previously in the NPRM. The agencies note that the concerns are not
with the concept of ICMs, but rather with the judgment-based values
suggested for use as ICMs, which have not been validated, and which
conflict with the empirically derived RPE value. The agencies will
continue to monitor any developments in ICM methodologies as part of
future rulemakings.
c) Stranded Capital Costs
 Past analyses accounted for costs associated with stranded capital
when fuel economy standards caused a technology to be replaced before
its costs were fully amortized. The idea behind stranded capital is
that manufacturers amortize research, development, and tooling expenses
over many years, especially for engines and transmissions. The
traditional production life-cycles for transmissions and engines have
been a decade or longer. If a manufacturer launches or updates a
product with fuel-saving technology, and then later replaces that
technology with an unrelated or different fuel-saving technology before
the equipment and research and development investments have been fully
paid off, there will be unrecouped, or stranded, capital costs.
Quantifying stranded capital costs accounts for such lost investments.
 In the Draft TAR and NPRM analyses, only a few technologies for a
few manufacturers were projected to have stranded capital costs. As
more technologies are included in this analysis, and as the CAFE model
has been expanded to account for platform and engine sharing and
updated with redesign and refresh cycles, accounting for stranded
capital has become increasingly complex. Separately, manufacturers may
be shifting their investment strategies in ways that may alter how
stranded capital calculations were traditionally considered. For
example, some suppliers sell similar transmissions to multiple
manufacturers. Such arrangements allow manufacturers to share in
capital expenditures, or amortize expenses more quickly.
 Manufacturers share parts on vehicles around the globe, achieving
greater scale and greatly affecting tooling strategies and costs. Given
these trends in the industry and their uncertain effect on capital
amortization, and given the difficulty of handling this uncertainty in
the CAFE model, this analysis does not account for stranded capital.
The agencies' analysis continues to rely on the CAFE model's explicit
year-by-year accounting for estimated refresh and redesign cycles, and
shared vehicle platforms and engines, to moderate the cadence of
technology adoption and thereby limit the implied occurrence of
stranded capital and the need to account for it explicitly. The
agencies will monitor these trends to assess the role of stranded
capital moving forward.
d) Cost Learning
 Manufacturers make improvements to production processes over time,
which often result in lower costs. ``Cost learning'' reflects the
effect of experience and volume on the cost of production, which
generally results in better utilization of resources, leading to higher
and more efficient production. As manufacturers gain experience through
production, they refine production techniques, raw material and
component sources, and assembly methods to maximize efficiency and
reduce production costs. Typically, a representation of this cost
learning, or learning curves, reflect initial learning rates that are
relatively high, followed by slower learning as additional improvements
are made and production efficiency peaks. This eventually produces an
asymptotic shape to the learning curve, as small percent decreases are
applied to gradually declining cost levels. These learning curve
estimates are applied to various technologies that are used to meet
CAFE standards.
 For the NPRM and this final rule, the agencies estimated cost
learning by considering methods established by T.P. Wright \645\ and
later expanded upon by J.R. Crawford. Wright, examining aircraft
production, found that every doubling of cumulative production of
airplanes resulted in decreasing labor hours at a fixed percentage.
This fixed percentage is commonly referred to as the progress rate or
progress ratio, where a lower rate implies faster learning as
cumulative production increases. J.R. Crawford expanded upon Wright's
learning curve theory to develop a single unit cost model,\646\ that
estimates the cost of the nth unit produced given the following
information is known: (1) Cost to produce the first unit; (2)
cumulative production of n units; and (3) the progress ratio.
---------------------------------------------------------------------------
 \645\ Wright, T.P., Factors Affecting the Cost of Airplanes.
Journal of Aeronautical Sciences, Vol. 3 (1936), pp. 124-125.
Available at http://www.uvm.edu/pdodds/research/papers/others/1936/wright1936a.pdf.
 \646\ Crawford, J.R., Learning Curve, Ship Curve, Ratios,
Related Data, Burbank, California-Lockheed Aircraft Corporation
(1944).
---------------------------------------------------------------------------
 As pictured in Figure VI-13, Wright's learning curve shows the
first unit is produced at a cost of $1,000. Initially cost per unit
falls rapidly for each successive unit produced. However, as production
continues, cost falls more gradually at a decreasing rate. For each
doubling of cumulative production at any level, cost per unit declines
20 percent, so that 80 percent of cost is retained. The CAFE model uses
the basic approach by Wright, where cost reduction is estimated by
applying a fixed percentage to the projected cumulative production of a
given fuel economy technology.
[[Page 24367]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.135
 The analysis accounts for learning effects with model year-based
cost learning forecasts for each technology that reduce direct
manufacturing costs over time. The agencies evaluated the historical
use of technologies, and reviewed industry forecasts to estimate future
volumes for the purpose of developing the model year-based technology
cost learning curves.
 The following section discusses the agencies' development of model
year-based cost learning forecasts, including how the approach has
evolved from the 2012 rulemaking for MY 2017-2025 vehicles, and how the
progress ratios were developed for different technologies considered in
the analysis. Finally, the agencies discuss how these learning effects
are applied in the CAFE Model.
(1) Time Versus Volume-Based Learning
 For the 2012 joint CAFE/CO2 rulemaking, the agencies
developed learning curves as a function of vehicle model year.\647\
Although the concept of this methodology is derived from Wright's
cumulative production volume-based learning curve, its application for
CAFE and CO2 technologies was more of a function of time.
More than a dozen learning curve schedules were developed, varying
between fast and slow learning, and assigned to each technology
corresponding to its level of complexity and maturity. The schedules
were applied to the base year of direct manufacturing cost and
incorporate a percentage of cost reduction by model year declining at a
decreasing rate through the technology's production life. Some newer
technologies experience 20 percent cost reductions for introductory
model years, while mature or less complex technologies experience 0-3
percent cost reductions over a few years.
---------------------------------------------------------------------------
 \647\ CAFE 2012 Final Rule, NHTSA DOT, 77 FR 62624.
---------------------------------------------------------------------------
 In their 2015 report to Congress, the National Academy of Sciences
(NAS) recommended the agencies should ``continue to conduct and review
empirical evidence for the cost reductions that occur in the automobile
industry with volume, especially for large-volume technologies that
will be relied on to meet the CAFE/GHG standards.'' \648\
---------------------------------------------------------------------------
 \648\ Cost, Effectiveness, and Deployment of Fuel Economy
Technologies for Light-Duty Vehicles, National Research Council of
the National Academies (2015), available at https://www.nap.edu/resource/21744/deps_166210.pdf.
---------------------------------------------------------------------------
 In response, the agencies have incorporated statically projected
cumulative volume production data of fuel economy improving
technologies, representing an improvement over the previously used
time-based method. Dynamic projections of cumulative production are not
feasible with current CAFE model capabilities, so one set of projected
cumulative production data for most vehicle technologies was developed
for the purpose of determining cost impact. For many technologies
produced and/or sold in the U.S., historical cumulative production data
was obtained to establish a starting point for learning schedules.
Groups of similar technologies or technologies of similar complexity
may share identical learning schedules.
 The slope of the learning curve, which determines the rate at which
cost reductions occur, has been estimated using research from an
extensive literature review and automotive cost tear-down reports (see
below). The slope of the learning curve is derived from the progress
ratio of manufacturing automotive and other mobile source technologies.
(2) Deriving the Progress Ratio Used in This Analysis
 Learning curves vary among different types of manufactured
products. Progress ratios can range from 70 to 100 percent, where 100
percent indicates no learning can be achieved.\649\ Learning effects
tend to be greatest in operations where workers often touch the
product, while effects are less substantial in operations consisting of
more automated processes. As automotive manufacturing plant processes
become increasingly automated, a progress ratio towards the higher end
would seem more suitable. The agencies incorporated findings from
automotive cost-teardown studies with EPA's literature review of
learning-related studies to estimate a progress ratio used to determine
learning schedules of fuel economy improving technologies.
---------------------------------------------------------------------------
 \649\ Martin, J., ``What is a Learning Curve?'' Management and
Accounting Web, University of South Florida, available at: https://www.maaw.info/LearningCurveSummary.htm.
---------------------------------------------------------------------------
 EPA's literature review examined and summarized 20 studies related
to learning in manufacturing industries and mobile source
manufacturing.\650\
[[Page 24368]]
The studies focused on many industries, including motor vehicles,
ships, aviation, semiconductors, and environmental energy. Based on
several criteria, EPA selected five studies providing quantitative
analysis from the mobile source sector (progress ratio estimates from
each study are summarized in Table VI-33, below). Further, those
studies expand on Wright's Learning Curve function by using cumulative
output as a predictor variable, and unit cost as the response variable.
As a result, EPA determined a best estimate of 84 percent as the
progress ratio in mobile source industries. However, of those five
studies, EPA at the time placed less weight on the Epple et al. (1991)
study, because of a disruption in learning due to incomplete knowledge
transfer from the first shift to introduction of a second shift at a
North American truck plant. While learning may have decelerated
immediately after adding a second shift, the agencies note that unit
costs continued to fall as the organization gained experience operating
with both shifts. The agencies now recognize that disruptions are an
essential part of the learning process and should not, in and of
themselves, be discredited. For this reason, the analysis uses a re-
estimated average progress ratio of 85 percent from those five studies
(equally-weighted).
---------------------------------------------------------------------------
 \650\ Cost Reduction through Learning in Manufacturing
Industries and in the Manufacture of Mobile Sources, United States
Environmental Protection Agency (2015). Prepared by ICF
International and available at https://19january2017snapshot.epa.gov/sites/production/files/2016-11/documents/420r16018.pdf.
[GRAPHIC] [TIFF OMITTED] TR30AP20.136
 In addition to EPA's literature review, this progress ratio
estimate was informed based on NHTSA's findings from automotive cost-
teardown studies. NHTSA routinely performs evaluations of costs of
previously issued Federal Motor Vehicle Safety Standards (FMVSS) for
new motor vehicles and equipment. NHTSA's engages contractors to
perform detailed engineering ``tear-down'' analyses for representative
samples of vehicles, to estimate how much specific FMVSS add to the
weight and retail price of a vehicle. As part of the effort, cost and
production volume are examined for automotive safety technologies. In
particular, the agency estimated costs from multiple cost tear-down
studies for technologies with actual production data from the Cost and
weight added by the Federal Motor Vehicle Safety Standards for MY 1968-
2012 passenger cars and LTVs (2017).\656\
---------------------------------------------------------------------------
 \651\ Argote, L., Epple, D., Rao, R. D., & Murphy, K., The
acquisition and depreciation of knowledge in a manufacturing
organization--Turnover and plant productivity, Working paper,
Graduate School of Industrial Administration, Carnegie Mellon
University (1997).
 \652\ Benkard, C. L., Learning and Forgetting--The Dynamics of
Aircraft Production, The American Economic Review, Vol. 90(4), pp.
1034-54 (2000).
 \653\ Epple, D., Argote, L., & Devadas, R., Organizational
Learning Curves--A Method for Investigating Intra-Plant Transfer of
Knowledge Acquired through Learning by Doing, Organization Science,
Vol. 2(1), pp. 58-70 (1991).
 \654\ Epple, D., Argote, L., & Murphy, K., An Empirical
Investigation of the Microstructure of Knowledge Acquisition and
Transfer through Learning by Doing, Operations Research, Vol. 44(1),
pp. 77-86 (1996).
 \655\ Levitt, S. D., List, J. A., & Syverson, C., Toward an
Understanding of Learning by Doing--Evidence from an Automobile
Assembly Plant, Journal of Political Economy, Vol. 121 (4), pp. 643-
81 (2013).
 \656\ Simons, J. F., Cost and weight added by the Federal Motor
Vehicle Safety Standards for MY 1968-2012 Passenger Cars and LTVs
(Report No. DOT HS 812 354). Washington, DC--National Highway
Traffic Safety Administration (November 2017), at pp. 30-33.
---------------------------------------------------------------------------
 NHTSA chose five vehicle safety technologies with sufficient data
to estimate progress ratios of each, because these technologies are
large-volume technologies and are used by almost all vehicle
manufacturers. Table VI-34 below includes these five technologies and
yields an average progress rate of 92 percent:
[GRAPHIC] [TIFF OMITTED] TR30AP20.137
[[Page 24369]]
 For a final progress ratio used in the CAFE model, the five
progress rates from EPA's literature review and five progress rates
from NHTSA's evaluation of automotive safety technologies results were
averaged. This resulted in an average progress rate of approximately 89
percent. Equal weight was placed on progress ratios from all 10
sources. More specifically, equal weight was placed on the Epple et al.
(1991) study, because disruptions have more recently been recognized as
an essential part in the learning process, especially in an effort to
increase the rate of output. Further discussion of how the progress
ratios were derived for this analysis is located in FRIA Section 9.
 ICCT commented that the choice to use safety technology as a model
for fuel efficiency led to lower learning rates in the NPRM analysis
compared to prior analyses.\657\ ICCT stated that safety technologies
were chosen for the NPRM because they are used by almost every
manufacturer, in contrast to fuel efficiency technologies, where not
every manufacturer will use them, particularly when they are first
introduced. ICCT stated that to show the impact of changing learning
rates, the agencies should run a sensitivity analysis using the
learning rates in the TAR, as well as EPA's learning rates in its Final
Determination. ICCT concluded that ``[w]ithout doing so and without
conducting a peer review of the change in approach, it appears clear
the agencies have decided to switch to a new costing method that
affects all future costs, but without any significant research
justification, vetting, or review.''
---------------------------------------------------------------------------
 \657\ NHTSA-2018-0067-11741.
---------------------------------------------------------------------------
 The agencies' selection of a progress rate of 0.89 is based on an
average of findings across research and literature reviews conducted by
NHTSA and EPA. The EPA cited rates were derived from five studies
selected from a sample of 20 transportation modal learning studies that
were examined by an EPA contractor, ICF International.\658\ One of
these 5 studies (Benkard (2000) examines learning in the commercial
aircraft industry, which the author notes has many unique features that
influence marginal costs. It also has the lowest progress rate. The
agencies note that EPA regulates all mobile sources, and while the
inclusion of non-passenger vehicle studies in their report was
justified, it may have biased the estimate of learning attributable to
the motor vehicle industry. Notably, nearly all of the other studies
included in the ICF International study found progress rates higher
than the 0.84 rate selected by the authors at that time. In reviewing
the ICF study, NHTSA found many other studies not included in the
report, including many specific to the motor vehicle and environmental
technology industries. Over 90 percent of those studies indicated
higher progress ratios than ICF recommended.\659\ The agencies' current
approach includes a broader and more representative sample of these
studies rather than the narrow sample selected by ICF.
---------------------------------------------------------------------------
 \658\ Cost Reduction through Learning in Manufacturing
Industries and in the Manufacture of Mobile Sources. United States
Environmental Protection Agency. Prepared by ICF International and
available at: https://19january2017snapshot.epa.gov/sites/production/files/2016-11/documents/420r16018.pdf.
 \659\ See, for example, progress ratios of multiple technologies
referenced in The Carbon Productivity Challenge: Curbing Climate
Change and Sustaining Economic Growth, McKinsey Climate Change
Special Initiative, McKinsey Global Institute, June 2008 (quoting
from UC Berkeley Energy Resource Group, Navigant Consulting) and
Technology Innovation for Climate Mitigation and its Relation to
Government Policies, Edward S. Rubin, Carnegie Mellon University,
Presentation to the UNFCCC Workshop on Climate Change Mitigation,
Bonn, Germany, June 19, 2004.
---------------------------------------------------------------------------
 The agencies do not agree that safety technologies are adopted by
all manufacturers at an early stage. Most safety technologies are
initially offered as options or standard equipment on only a small
segment of the vehicle fleet, typically luxury vehicles. After a number
of years, these technologies may be adopted on less expensive vehicles,
and eventually they will become required equipment on all vehicles, but
the production process is gradual, as it is with fuel efficiency
technologies. FMVSS are necessarily established as performance
standards--and automakers are free to develop or choose from existing
technologies to achieve such performance requirements--much like
automakers can develop or choose from a number of established fuel
efficiency technologies to achieve fuel economy requirements. Further,
the derivation of progress ratios is based on the concept of a doubling
of cumulative production, not time. Therefore, even if production
continues at a different pace, it should not disqualify non-fuel
efficiency studies. Moreover, the derivation of the progress ratio used
in the TAR and Final Determination document were not confined to fuel
efficiency technologies. In fact, as noted above, they even included at
least one entirely unrelated study of the aircraft industry.
 Finally, the agencies note that the previous learning schedules
used in the TAR and EPA's Final Determination were only developed
through 2025, whereas this final rule projects learning through 2050.
The previous learning schedules are thus not directly compatible with
the analysis conducted in this Final Rule, making a sensitivity
analysis problematic.
(3) Obtaining Appropriate Baseline Years for Direct Manufacturing Costs
To Create Learning Curves
 Direct manufacturing costs for each fuel economy improving
technology were obtained from various sources, as discussed above. To
establish a consistent basis for direct manufacturing costs in the
rulemaking analysis, each technology cost is adjusted to MY 2018
dollars. For each technology, the DMC is associated with a specific
model year, and sometimes a specific production volume, or cumulative
production volume. The base model year is established as the MY in
which direct manufacturing costs were assessed (with learning factor of
1.00). With the aforementioned data on cumulative production volume for
each technology and the assumption of a 0.89 progress ratio for all
automotive technologies, the agencies can solve for an implied cost for
the first unit produced. For some technologies, the agencies used
modestly different progress ratios to match detailed cost projections
if available from another source (for instance, batteries for plug-in
hybrids and battery electric vehicles).
 This approach produced reasonable estimates for technologies
already in production, and some additional steps were required to set
appropriate learning rates for technologies not yet in production.
Specifically, for technologies not yet in production in MY 2017 (the
baseline analysis fleet), the cumulative production volume in MY 2017
is zero, because manufacturers have not yet produced the technologies.
For pre-production cost estimates in the NPRM, the agencies often
relied on confidential business information sources to predict future
costs. Many sources for pre-production cost estimates include
significant learning effects, often providing cost estimates assuming
high volume production, and often for a timeframe late in the first
production generation or early in the second generation of the
technology. Rapid doubling and re-doubling of a low cumulative volume
base with Wright's learning curves can provide unrealistic cost
estimates. In addition, direct manufacturing cost projections can vary
depending on the initial production volume assumed. Accordingly, the
agencies carefully examined direct costs with learning, and made
adjustments to the starting point for those technologies on the
learning curve to better align
[[Page 24370]]
with the assumptions used for the initial direct cost estimate.
(4) Cost Learning as Applied in the CAFE Model
 For the NPRM analysis, the agencies updated the manner in which
learning effects apply to costs. In the Draft TAR analysis, the
agencies had applied learning curves only to the incremental direct
manufacturing costs or costs over the previous technology on the
technology tree. In practice, two things were observed: (1) If the
incremental direct manufacturing costs were positive, technologies
could not become less expensive than their predecessors on the
technology tree, and (2) absolute costs over baseline technology
depended on the learning curves of root technologies on the technology
tree. For the NPRM and final rule analysis, the agencies applied
learning effects to the incremental cost over the null technology state
on the applicable technology tree. After this step, the agencies
calculated year-by-year incremental costs over preceding technologies
on the tech tree to create the CAFE model inputs. As discussed below,
for the final rule, the agencies revised the CAFE model to replace
incremental cost estimates with absolute estimates, each specified
relative to the null technology state on the applicable technology
tree. This change facilitated quality assurance and is expected to make
cost inputs more transparently relatable to detailed model output.
Likewise, this change made it easier to apply learning curves in the
course of developing inputs to the CAFE model.
 The agencies grouped certain technologies, such as advanced
engines, advanced transmissions, and non-battery electric components
and assigned them to the same learning schedule. While these grouped
technologies differ in operating characteristics and design, the
agencies chose to group them based on their complexity, technology
integration, and economies of scale across manufacturers. The low
volume of certain advanced technologies, such as hybrid and electric
technologies, poses a significant issue for suppliers and prevents them
from producing components needed for advanced transmissions and other
technologies at more efficient high scale production. The technology
groupings were carried over from the NPRM analysis for the final rule
analysis.\660\ Like the NPRM, this final rule analysis uses the same
groupings that considers market availability, complexity of technology
integration, and production volume of the technologies that can be
implemented by manufacturers and suppliers. For example, technologies
like ADEAC and VCR are grouped together; these technologies were not in
production or were only in limited introduction in MY 2017, and are
planned to be introduced in limited production by a few manufacturers.
The details of these technologies are discussed in Section VI.C.
---------------------------------------------------------------------------
 \660\ See PRIA Chapter 6 for technology groupings.
---------------------------------------------------------------------------
 In addition, for the final rule, as discussed in Section VI.A.4
Compliance Simulation, the agencies expanded model inputs to extend the
explicit simulation of technology application through MY 2050, in
response to comments on the NPRM. Accordingly, the agencies updated the
learning curves for each technology group to cover MYs through 2050.
For MYs 2017-2032, the agencies expect incremental improvements in all
technologies, particularly in electrification technologies because of
increased production volumes, labor efficiency, improved manufacturing
methods, specialization, network building, and other factors. While
these and other factors contribute to continual cost learning, the
agencies believe that many fuel economy improving technologies
considered in this rule will approach a flat learning level by the
early 2030s. Specifically, older and less complex internal combustion
engine technologies and transmissions will reach a flat learning curve
sooner when compared to electrification technologies, which have more
opportunity for improvement. For batteries and non-battery
electrification components, the agencies estimated a steeper learning
curve that will gradually flatten after MY 2040. For a more detailed
discussion of the electrification learning curves used for the final
rule analysis, see Section VI.C.3.e) Electrification Costs. The
following Table VI-35 and Table VI-36 show the learning curve schedules
for CAFE model technologies for MYs 2017-2033 and MYs 2034-2050.
BILLING CODE 4910-59-P
[[Page 24371]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.138
[[Page 24372]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.139
[[Page 24373]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.140
[[Page 24374]]
[GRAPHIC] [TIFF OMITTED] TR30AP20.141
BILLING CODE 4910-59-C
 Each technology in the CAFE Model is assigned a learning schedule
developed from the methodology explained previously. For example, the
[[Page 24375]]
following chart shows learning rates for several technologies
applicable to midsize sedans, demonstrating that while the agencies
estimate that such learning effects have already been almost entirely
realized for engine turbocharging (a technology that has been in
production for many years), the agencies estimate that significant
opportunities to reduce the cost of the greatest levels of mass
reduction (e.g., MR5) remain, and even greater opportunities remain to
reduce the cost of batteries for HEVs, PHEVs, BEVs. In fact, for
certain advanced technologies, the agencies determined that the results
predicted by the standard learning curves progress ratio was not
realistic, based on unusual market price and production relationships.
For these technologies, the agencies developed specific learning
estimates that may diverge from the 0.89 progress rate. As shown in
Figure VI-14, these technologies include: Turbocharging and downsizing
level 1 (TURBO1), variable turbo geometry electric (VTGE), aerodynamic
drag reduction by 15 percent (AERO15), mass reduction level 5 (MR5), 20
percent improvement in low-rolling resistance tire technology over the
baseline, and battery integrated starter/generator (BISG).
[GRAPHIC] [TIFF OMITTED] TR30AP20.142
(5) Potential Future Approaches to Considering Cost Learning in the
CAFE Model
 As discussed above, cost inputs to the CAFE model incorporate
estimates of volume-based learning. As an alternative approach, the
agencies have considered modifications to the CAFE model that would
calculate degrees of volume-based learning dynamically, responding to
the model's application of affected technologies. While it is intuitive
that the degree of cost reduction achieved through experience producing
a given technology should depend on the actual accumulated experience
(i.e., volume) producing that technology, such dynamic implementation
in the CAFE model is thus far infeasible. Insufficient data have been
available regarding manufacturers' historical application of specific
technology. Further, insofar as the agencies' estimates of underlying
direct manufacturing costs already make some assumptions about volume
and scale, insufficient information is currently available to determine
how to dynamically adjust these underlying costs. It should be noted
that if learning responds dynamically to volume, and volume responds
dynamically to learning, an internally consistent model solution would
likely require iteration of the CAFE model to seek a stable solution
within the model's representation of multiyear planning. As discussed
below, the CAFE model now supports iteration to balance vehicle
[[Page 24376]]
cost and fuel economy changes with corresponding changes in sales
volumes, but, this iteration is not yet implemented in a manner that
would necessarily support the balance of learning effects on a
multiyear basis. The agencies invited comment on the issue, seeking
data and methods that would provide the basis for a practicable
approach to doing so. Having reviewed comments on cost learning
effects, the agencies conclude it remains infeasible to calculate
degrees of volume-based learning in a manner that responds dynamically
to modeled technology application. The agencies will continue to
examine this issue for future development.
e) Cost Accounting
 The CAFE model applied for the NPRM analysis used an incremental
approach to specifying technology cost estimates, such that the cost
for any given technology was specified as an incremental value,
relative to the technology immediately preceding on the relevant
technology pathway. For example, the cost of a 7-speed transmission was
specified as an amount beyond the cost of a 6-speed transmission. This
approach necessitated careful dynamic accounting for the progressive
application of the technology as the model worked on a step-by-step
basis to ``build'' a technology solution. As discussed in the
corresponding model documentation, the model included complex logic to
``back out'' some of these costs carefully when, for example, replacing
a conventional powertrain with a hybrid-electric system.\661\
---------------------------------------------------------------------------
 \661\ The CAFE Model is available at https://www.nhtsa.gov/corporate-average-fuel-economy/compliance-and-effects-modeling-system with documentation and all inputs and outputs supporting
today's notice.
---------------------------------------------------------------------------
 To facilitate specification of detailed model inputs and review of
detailed model outputs, today's CAFE model replaces incremental cost
inputs with absolute cost inputs, such that the estimated cost of each
technology is specified relative to a common reference point for the
relevant technology pathway. For example, the cost of the above-
mentioned 7-speed transmission is specified relative to a 4-speed
transmission, as is the cost of every other transmission technology.
This change in the structure of cost inputs does not, by itself, change
model results, but it does make the connection between these inputs and
corresponding outputs more transparent. Model documentation
accompanying today's analysis presents details of the updated structure
for model cost inputs.
5. Other Inputs to the Agencies' Analysis
 CAFE Model input files described above defining the analysis fleet
and the fuel-saving technologies to be included in the analysis span
more than a million records, but deal with a relatively discrete range
of subjects (e.g., what vehicles are in the fleet, what are the key
characteristics of those vehicles, what fuel-saving technologies are
expected to be available, and how might adding those technologies
impact vehicles' fuel economy levels and costs). The CAFE Model makes
use of a considerably wider range of other types of inputs, and most of
these are contained in other model input files. The nature and function
of many of these inputs remains unchanged relative to the model and
input files applied for the analysis documented in the proposal that
preceded today's notice. The CAFE Model documentation accompanying
today's notice lists and describes all model inputs, and explains how
inputs are used by the model. Many commenters addressed not only the
model's function and design, but also specific inputs. Most input
values are discussed either above (e.g., the preceding subsection
addresses specific inputs regarding technology costs) or below, in
subsections discussing specific economic, energy, safety, and
environmental factors. The remainder of this subsection provides an
overview of the scope of different model input files. The overview is
organized based on CAFE Model file types, as in the model
documentation.
a) Market Data File
 The ``Market Data'' file contains the detailed description--
discussed above--of the vehicle models and model configurations each
manufacturer produces for sale in the U.S. The file also contains a
range of other inputs that, though not specific to individual vehicle
models, may be specific to individual manufacturers. The file contains
a set of specific worksheets, as follows:
 ``Manufacturers'' worksheet: Lists specific manufacturers,
indicates whether manufacturers are expected to prefer paying CAFE
fines to applying technologies that would not be cost-effective,
indicates what ``payback period'' defines buyers' willingness to pay
for fuel economy improvements, enumerates CAFE and CO2
credits banked from model years prior to those represented explicitly,
and indicates how sales ``multipliers'' are to be applied when
simulating compliance with CO2 standards.
 ``Credits and Adjustments'' worksheet: Enumerates estimates--
specific to each manufacturer and fleet--of expected CO2 and
CAFE adjustments reflecting improved AC efficiency, reduced AC
refrigerant leakage, improvements to ``off cycle'' efficiency, and
production of flexible fuel vehicles (FFVs). The model applies AC
refrigerant leakage adjustments only to CO2 levels, and
applies FFV adjustments only to CAFE levels.
 ``Vehicles'' worksheet: Lists vehicle models and model
configurations each manufacturer produces for sale in the U.S.;
identifies shared vehicle platforms; indicates which engine and
transmission is present in each vehicle model configuration; specifies
each vehicle model configuration's fuel economy level, production
volume, and average price; specifies several engineering
characteristics (e.g., curb weight, footprint, and fuel tank volume);
assigns each vehicle model configuration to a regulatory class,
technology class, engine class, and safety class; specifies schedules
on which specific vehicle models are expected to be redesigned and
freshened; specifies how much U.S. labor is involved in producing each
vehicle model/configuration; and indicates whether specific
technologies are already present on specific vehicle model
configurations, or, due to engineering or product planning
considerations, should be skipped.
 ``Engines'' worksheet: Identifies specific engines used by each
manufacturer and for each engine, lists a unique code (referenced by
the engine code specified for each vehicle model configuration and
identifies the fuel(s) with which the engine is compatible, the
valvetrain design (e.g., DOHC), the engine's displacement, cylinder
configuration and count, and the engine's aspiration type (e.g.,
naturally aspirated, turbocharged). The worksheet also indicates
whether specific technologies are already present on specific engines,
or, due to engineering or product planning considerations, should be
skipped.
 ``Transmissions'' worksheet: Similar to the Engines worksheet,
identifies specific transmissions used by each manufacturer and for
each transmission, lists a unique code (referenced by the transmission
code specified for each vehicle model configuration and identifies the
type (e.g., automatic or CVT) and number of forward gears. Also
indicates whether specific technologies are already present or, due to
engineering or product planning considerations, should be skipped.
[[Page 24377]]
b) Technologies File
 The Technologies file identifies about six dozen technologies to be
included in the analysis, indicates when and how widely each technology
can be applied to specific types of vehicles, provides most of the
inputs involved in estimating what costs will be incurred, and provides
some of the inputs involved in estimating impacts on vehicle fuel
consumption and weight. The file contains the following types of
worksheets:
 ``Parameters'' worksheet: Not to be confused with the
``Parameters'' file discussed below, this worksheet in the Technologies
file indicates, for each technology class, the share of the vehicle's
curb weight represented by the ``glider'' (the vehicle without the
powertrain).
 ``Technologies'' worksheet: For each named technology, specifies
the share of the entire fleet to which the technology may be
additionally applied in each model year.
 Technology Class worksheets: In a separate worksheet for each of
the 10 technology classes discussed above (and an additional 2--not
used for this analysis--for heavy-duty pickup trucks and vans),
identifies whether and how soon the technology is expected to be
available for wide commercialization, specifies the percentage of miles
a vehicle is expected to travel on a secondary fuel (if applicable, as
for plug-in hybrid electric vehicles), indicates a vehicle's expected
electric power and all-electric range (if applicable), specifies
expected impacts on vehicle weight, specifies estimates of costs in
each model year (and factors by which electric battery costs are
expected to be reduced in each model year), specifies any estimates of
maintenance and repair cost impacts, and specifies any estimates of
consumers' willingness to pay for the technology.
 Engine Type worksheets: In a separate worksheet for each of 28
initial engine types identified by cylinder count, number of cylinder
banks, and configuration (DOHC, unless identified as OHV or SOHC),
specifies estimates of costs in each model year, as well as any
estimates of impacts on maintenance and repair costs.
c) Parameters File
 The ``Parameters'' file contains inputs spanning a range of
considerations, such as economic and labor utilization impacts, vehicle
fleet characteristics, fuel prices, scrappage and safety model
coefficients, fuel properties, and emission rates. The file contains a
set of specific worksheets, as follows:
 Economic Values worksheet: Specifies a variety of inputs, including
social and consumer discount rates to be applied, the ``base year'' to
which to discount social benefits and costs (i.e., the reference years
for present value analysis), discount rates to be applied to the social
cost of CO2 emissions, the elasticity of highway travel with
respect to per-mile fuel costs (also referred to as the rebound
effect), the gap between test (for certification) and on-road (aka real
world) fuel economy, the fixed amount of time involved in each refuel
event, the share of the tank refueled during an average refueling
event, the value of travel time (in dollars per hour per vehicle), the
estimated average number of miles between mid-trip EV recharging events
(separately for 200 and 300-mile EVs), the rate (in miles of capacity
per hour of charging) at which EV batteries are recharged during such
events, the values (in dollars per vehicle-mile) of congestion and
noise costs, costs of vehicle ownership and operation (e.g., sales
tax), economic costs of oil imports, estimates of future macroeconomic
measures (e.g., GDP), and rates of growth in overall highway travel
(separately for low, reference, and high oil prices).
 Vehicle Age Data worksheet: Specifies nominal average survival
rates and annual mileage accumulation for cars, vans and SUVs, and
pickup trucks. These inputs are used only for displaying estimates of
avoided fuel savings and CO2 emissions while the model is
operating. Calculations reported in model output files reflect, among
other things, application of the scrappage model.
 Fuel Prices worksheet: Separately for gasoline, E85, diesel,
electricity, hydrogen, and CNG, specifies historical and estimated
future fuel prices (and average rates of taxation). Includes values
reflecting low, reference, and high estimates of oil prices.
 Scrappage Model Values worksheet: Specifies coefficients applied by
the scrappage model, which the CAFE Model uses to estimate rates at
which vehicles will be scrapped (removed from service) during the
period covered by the analysis.
 Historic Fleet Data worksheet: For model years not simulated
explicitly (here, model years through 2016), and separately for cars,
vans and SUVs, and pickup trucks, specifies the initial size (i.e.,
number new vehicles produced for sale in the U.S.) of the fleet, the
number still in service in the indicated calendar year (here, 2016),
the relative shares of different fuel types, and the average fuel
economy achieved by vehicles with different fuel types, and the
averages of horsepower, curb weight, fuel capacity, and price (when
new).
 Safety Values worksheet: Specifies coefficients used to estimate
the extent to which changes in vehicle mass impact highway safety. Also
specifies statistical value of highway fatalities, the share of
incremental risk (of any additional driving) internalized by drivers,
rates relating the cost of damages from non-fatal losses to the cost of
fatalities, and rates relating the occurrence of non-fatal injuries to
the occurrence of fatalities.
 Fatality Rates worksheet: Separately for each model year from 1975-
2050, and separately for each vehicle age (through 39 years) specifies
the estimated nominal number of fatalities incurred per billion miles
of travel by which to offset fatalities.
 Credit Trading Values worksheet: Specifies whether various
provisions related to compliance credits are to be simulated (currently
limited to credit carry-forward and transfers), and specifies the
maximum number of years credits may be carried forward to future model
years. Also specifies statutory (for CAFE only) limits on the quantity
of credit that may be transferred between fleets, and specifies amounts
of lifetime mileage accumulation to be assumed when adjusting the value
of transferred credits. Also accommodates a setting indicating the
maximum number of model years to consider when using expiring credits.
 Employment Values worksheet: Specifies the estimated average
revenue OEMs and suppliers earn per employee, the retail price
equivalent factor applied in developing technology costs, the average
quantity of annual labor (in hours) per employee, a multiplier to apply
to U.S. final assembly labor utilization in order to obtain estimated
direct automotive manufacturing labor, and a multiplier to be applied
to all labor hours.
 Fuel Properties worksheet: Separately for gasoline, E85, diesel,
electricity, hydrogen, and CNG, specifies energy density, mass density,
carbon content, and tailpipe SO2 emissions (grams per unit
of energy).
 Fuel Import Assumptions worksheet: Separately for gasoline, E85,
diesel, electricity, hydrogen, and CNG, specifies the extent to which
(a) changes in fuel consumption lead to changes in net imports of
finished fuel, (b) changes in fuel consumption lead to changes in
domestic refining output, (c) changes in domestic refining output lead
to changes in domestic crude oil production, and (d) changes in
domestic refining output lead to changes in net imports of crude oil.
[[Page 24378]]
 Emissions Health Impacts worksheet: Separately for NOX,
SO2 and PM2.5 emissions, separately for upstream
and vehicular emissions, and for each of calendar years 2016, 2020,
2025, and 2030, specifies estimates of various health impacts, such as
premature deaths, acute bronchitis, and respiratory hospital
admissions.
 Carbon Dioxide Emission Costs worksheet: For each calendar year
through 2080, specifies low, average, and high estimates of the social
cost of CO2 emissions, in dollars per metric ton.
Accommodates analogous estimates for CH4 and N2O.
 Criteria Pollutant Emission Costs worksheet: Separately for
NOX, SO2 and PM2.5 emissions,
separately for upstream and vehicular emissions, and for each of
calendar years 2016, 2020, 2025, and 2030, specifies social costs on a
per-ton basis.
 Upstream Emissions (UE) worksheets: Separately for gasoline, E85,
diesel, electricity, hydrogen, and CNG, and separately for calendar
years 2017, 2020, 2025, 2030, 2035, 2040, 2045, and 2050, and
separately for various upstream processes (e.g., petroleum refining),
specifies emission factors (in grams per million BTU) for each included
criteria pollutant (e.g., NOX) and toxic air contaminant
(e.g., benzene).
 Tailpipe Emissions (TE) worksheets: Separately for gasoline and
diesel, for each of model years 1975-2050, for each vehicle vintage
through age 39, specifies vehicle tailpipe emission factors (in grams
per mile) for CO, VOC, NOX, PM2.5,
CH4, N2O, acetaldehyde, acrolein, benzene,
butadiene, formaldehyde, and diesel PM10.
d) Scenarios File
 The CAFE Model represents each regulatory alternative as a discrete
scenario, identifying the first-listed scenario as the baseline
relative to which impacts are to be calculated. Each scenario is
described in a worksheet in the Scenarios input file, with standards
and related provisions specified separately for each regulatory class
(passenger car or light truck) and each model year. Inputs specify the
standards' functional forms and defining coefficients in each model
year. Multiplicative factors and additive offsets are used to convert
fuel economy targets to CO2 targets, the two being directly
mathematically related by a linear transformation. Additional inputs
specify minimum CAFE standards for domestic passenger car fleets,
determine whether upstream emissions from electricity and hydrogen are
to be included in CO2 compliance calculations, specify the
governing rates for CAFE civil penalties, specify estimates of the
value of CAFE and CO2 credits (for CAFE Model operating
modes applying these values), specify how flexible fuel vehicles (FFVs)
and PHEVs are to be accounted for in CAFE compliance calculations,
specific caps on adjustments reflecting improvements to off-cycle and
AC efficiency and emissions, specify any estimated amounts of average
Federal tax credits earned by HEVs, PHEVs, BEVs, and FCVs. The
worksheets also accommodate some other inputs, such those as involved
in analyzing standards for heavy-duty pickups and vans, not used in
today's analysis.
e) ``Run Time'' Settings
 In addition to inputs contained in the above-mentioned files, the
CAFE Model makes use of some settings selected when operating the
model. These include which standards (CAFE or CO2) are to be
evaluated; what model years the analysis is to span; when technology
application is to begin; what ``effective cost'' mode is to be used
when selecting among technologies; whether use of compliance credits is
to be simulated and, if so, until what model year; whether dynamic
economic models are to be exercised and, if so, how many sales model
iterations are to be undertaken and using what price elasticity;
whether low, average, or high estimates are to be applied for fuel
prices, the social cost of carbon, and fatality rates; by how much to
scale benefits to consumers; and whether to report an implicit
opportunity cost.
f) Simulation Inputs
 As mentioned above, the CAFE Model makes use of databases of
estimates of fuel consumption impacts and, as applicable, battery costs
for different combinations of fuel saving technologies. For today's
analysis, the agencies developed these databases using a large set of
full vehicle and accompanying battery cost model simulations developed
by Argonne National Laboratory. To be used as files provided separately
from the model and loaded every time the model is executed, these
databases are prohibitively large, spanning more than a million records
and more than half a gigabyte. To conserve space and speed model
operation, the agencies have integrated the databases into the CAFE
Model executable file. When the model is run, however, the databases
are extracted and placed in an accessible location on the user's disk
drive. The databases, each of which is in the form of a simple (if
large) text file, are as follows:
 ``FE1_Adjustments.csv:'' This is the main database of fuel
consumption estimates. Each record contains such estimates for a
specific indexed (using a multidimensional ``key'') combination of
technologies for each of the technology classes in the Market Data and
Technologies files. Each estimate is specified as a percentage of the
``base'' technology combination for the indicated technology class.
 ``FE2_Adjustments.csv:'' Specific to PHEVs, this is a database of
fuel consumption estimates applicable to operation on electricity,
specified in the same manner as those in the main database.
 ``Battery_Costs.csv:'' Specific to technology combinations
involving vehicle electrification (including 12V stop-start systems),
this is a database of estimates of corresponding base costs (before
learning effects) for batteries in these systems.
g) On Road Fuel Economy and CO2 Emissions Gap
 Rather than rely on the compliance values of fuel economy for
either historical vehicles or vehicles that go through the full
compliance simulation, the model applies an ``on-road gap'' to
represent the expected difference between fuel economy on the
laboratory test cycle and fuel economy under real-world operation. In
other words, all of the reported physical impacts analysis (including
emissions impacts) are based on actual real world fuel consumption and
emissions, not on values based on 2-cycle fuel economy ratings and
CO2 emission rates, nor on regulatory incentives such as
sales multipliers that treat a single vehicle as two vehicles, or that
set aside emissions resulting from generation of electricity to power
electric vehicles. This was a topic of interest in the recent peer
review of the CAFE model. While the model currently allows the user to
specify an on-road gap that varies by fuel type (gasoline, E85, diesel,
electricity, hydrogen, and CNG), it does not vary over time, by vehicle
age, or by technology combination. It is possible that the ``gap''
between laboratory fuel economy and real-world fuel economy has changed
over time, that fuel economy changes as a vehicle ages, or that
specific combinations of fuel-saving technologies have a larger
discrepancy between laboratory and real-world fuel economy than others.
For today's analysis, and considering data EPA collects from
manufacturers regarding vehicles' fuel economy and CO2 as
tested for both fuel economy and emissions compliance and for vehicle
fuel economy and emissions labeling
[[Page 24379]]
(labeling making use of procedures spanning a wider range of real-world
vehicle operating conditions), the agencies have determined that the
future gap is, at this time, best estimated using the same values
applied for the analysis documented in the NPRM. The agencies will
continue to assess such test data and any other available data
regarding real-world fuel economy and emissions and, as warranted, will
revise methods and inputs representing the gap between laboratory and
real-world fuel economy and CO2 emissions in future
rulemakings. The sensitivity analysis summarized in the FRIA
accompanying the final rule includes cases representing narrower and
wider gaps.
C. The Model Applies Technologies Based on a Least-Cost Technology
Pathway to Compliance, Given the Framework Above
 The CAFE model, discussed in detail above, is designed to simulate
compliance with a given set of CAFE or tailpipe CO2
emissions standards for each manufacturer that sells vehicles in the
United States. For the final rule analysis, the model began with a
representation of the MY 2017 vehicle model offerings for each
manufacturer that included the specific engines and transmissions on
each model variant, observed sales volumes, and all fuel economy
improving technology that is already present on those vehicles. From
there the model added technology, in response to the standards being
considered, in a way that minimized the cost of compliance and
reflected many real-world constraints faced by automobile
manufacturers. The model addressed fleet year-by-year compliance,
taking into consideration vehicle refresh and redesign schedules and
shared platforms, engines, and transmissions among vehicles.
 The agencies evaluated a wide array of technologies manufacturers
could use to improve the fuel economy of new vehicles, in both the
immediate future and during the timeframe of this rulemaking, to meet
the fuel economy and CO2 standards. The agencies evaluated
costs for these technologies, and looked at how costs may change over
time. The agencies also considered how fuel-saving technologies may be
used on many types of vehicles (ranging from small cars to trucks) and
how the technologies may perform in improving fuel economy and
CO2 emissions in combination with other technologies. With
cost and effectiveness estimates for technologies, the agencies
forecast how manufacturers may respond to potential standards and can
estimate the associated costs and benefits related to technology and
equipment changes. This assists the assessment of technological
feasibility and is a building block for the consideration of economic
practicability of the standards.
 The agencies described in the NPRM that the characterization of
current and anticipated fuel-saving technologies relied on portions of
the analysis presented in the Draft TAR, in addition to new information
that had been gathered and developed since conducting that analysis,
and the significant, substantive input that was received during the
Draft TAR comment period.\662\ The Draft TAR considered many
technologies previously assessed in the 2012 final rule; \663\ in some
cases, manufacturers have nearly universally adopted a technology in
today's new vehicle fleet (for example, electric power steering), but
in other cases, manufacturers only occasionally use a technology in
today's new vehicle fleet (like turbocharged engines). For a few
technologies considered in the 2012 rulemaking, manufacturers began
implementing the technologies but have since largely pivoted to other
technologies due to consumer acceptance issues (for instance,
drivability and performance feel issues associated with some dual
clutch transmissions without a torque converter) or limited commercial
success.
---------------------------------------------------------------------------
 \662\ 83 FR 43021-22 (Aug. 24, 2018).
 \663\ 77 FR 62624 (Oct. 15, 2012).
---------------------------------------------------------------------------
 In some cases, EPA and NHTSA presented different analytical
approaches in the Draft TAR. However, for the NPRM and final rule
analysis, the agencies harmonized their analytical approach to use one
set of effectiveness values (developed with one tool), one set of cost
assumptions, and one set of assumptions about the limitations of some
technologies. To develop these assumptions, the agencies evaluated many
sources of data, in addition to many stakeholder comments received on
the Draft TAR. The preferred approach was to harmonize on sources and
methodologies that were data-driven and reproducible for independent
verification, produced using tools utilized by OEMs, suppliers, and
academic institutions, and using tools that could support both CAFE and
CO2 analysis. As the agencies noted in the NPRM, a single
set of assumptions also facilitated and focused public comment by
reducing burden on stakeholders who sought to review all of the
supporting documentation surrounding the analysis.
 The agencies also identified a preference to use values developed
from careful review of commercialized technologies; however, in some
cases for technologies that are new, and are not yet for sale in any
vehicle, the analysis relied on information from other sources,
including CBI and third-party research reports and publications. The
agencies strived to keep the technology analysis as current as possible
in light of the ongoing technology development and implementation in
the automotive industry. Additional emerging technologies added for the
final rule analysis are described in further detail, below.
 The agencies' process to develop effectiveness assumptions is
described in detail in Section VI.B.3 Technology Effectiveness, and
summarized here. The NPRM and final rule analysis modeled combinations
of more than 50 fuel economy-improving technologies across 10 vehicle
types (an increase from five vehicle types in NHTSA's Draft TAR
analysis). Only 10 vehicle technology classes were used because large
portions of the production volume in the analysis fleet have similar
specifications, especially in highly competitive segments. For
instance, many mid-sized sedans, small SUVs, and large SUVs coalesce
around similar specifications, respectively. Baseline simulations have
been aligned around these modal specifications. Parametrically
combining these technologies generated more than 100,000 unique
combinations per vehicle class. Multiplying the unique technology
combinations by the 10 technology classes resulted in the simulation of
more than one million individual full-vehicle system models. Modeling
was also conducted to determine appropriate levels of engine downsizing
required to maintain baseline vehicle performance when advanced mass
reduction technology or advanced engine technology were applied.
Performance neutrality is discussed in detail in VI.B.3.
 Some baseline vehicle assumptions used in the simulation modeling
were updated since the Draft TAR based on public comments, and further
assessment of the NPRM and final rule analysis fleets. The agencies
updated assumptions about curb weight, as well as technology properties
like baseline rolling resistance, aerodynamic drag coefficients, and
frontal areas. Many of the assumptions are aligned with published
research from the Department of Energy and other independent
[[Page 24380]]
sources.\664\ Additional transmission technologies and more levels of
aerodynamic technologies than NHTSA presented in the Draft TAR analysis
were also added for the analysis. Having additional technologies in the
model allowed the agencies to assign baselines and estimate fuel-
savings opportunities with more precision.
---------------------------------------------------------------------------
 \664\ See, e.g., Islam, E., A. Moawad, N. Kim, and A. Rousseau,
2018a, An Extensive Study on Vehicle Sizing, Energy Consumption and
Cost of Advance Vehicle Technologies, Report No. ANL/ESD-17/17,
Argonne National Laboratory, Lemont, Ill., Oct 2018. https://www.autonomie.net/pdfs/ANL_BaSce_FY17_Report_10042018.pdf. Last
accessed March 18, 2020; Pannone, G. ``Technical Analysis of Vehicle
Load Reduction Potential for Advanced Clean Cars,'' April 29, 2015.
Available at https://www.arb.ca.gov/research/apr/past/13-313.pdf.
Last accessed December 28, 2019.
---------------------------------------------------------------------------
 To develop technology cost assumptions, the agencies estimated
present and future costs for fuel-saving technologies, taking into
consideration the type of vehicle, or type of engine if technology
costs vary by application. Since the 2012 final rule, many cost
assessments, including tear down studies, were funded and completed,
and presented as part of the Draft TAR analysis. These studies
evaluated transmissions, engines, hybrid technologies, and mass
reduction.\665\ The NPRM and final rule analyses use the 2016 Draft
TAR's cost estimates for many technologies. In addition to those
studies, the analysis also leveraged research reports from other
organizations to assess costs.\666\ Consistent with past analyses, this
analysis used BatPaC to provide estimates for future battery costs for
hybrids, plug-in hybrids, and electric vehicles, taking into account
the different battery design characteristics and taking into account
the size of the battery for different applications.\667\ The agencies
also updated technology costs for the NPRM to 2016 dollars, because, as
in many cases, technology costs were estimated several years ago, and
since then have further updated technology costs to 2018 dollars for
the final rule.
---------------------------------------------------------------------------
 \665\ FEV prepared several cost analysis studies for EPA on
subjects ranging from advanced 8-speed transmissions to belt
alternator starter, or Start/Stop systems. NHTSA also contracted
with Electricore and EDAG on teardown studies evaluating mass
reduction. The 2015 NAS report on fuel economy technologies for
light-duty vehicles also evaluated the agencies' technology costs
developed based on these teardown studies, and the technology costs
used in this proposal were updated accordingly.
 \666\ For example, the agencies relied on reports from the
Department of Energy's Office of Energy Efficiency & Renewable
Energy's Vehicle Technologies Office. More information on that
office is available at https://www.energy.gov/eere/vehicles/vehicle-technologies-office. Other agency reports that were relied on for
technology or other information are referenced throughout the NPRM
and accompanying PRIA, and this final rule and the accompanying
FRIA.
 \667\ For instance, battery electric vehicles with high levels
of mass reduction may use a smaller battery than a comparable
vehicle with less mass reduction technology and still deliver the
same range on a charge. See, e.g., Ward, J. & Gohlke, D. & Nealer,
Rachael. (2017). The Importance of Powertrain Downsizing in a
Benefit-Cost Analysis of Vehicle Lightweighting. JOM. 69.
---------------------------------------------------------------------------
 Cost and effectiveness values were estimated for each technology
included in the analysis. As mentioned above, more than 50 technologies
were considered in the NPRM and final rule analyses, and the agencies
evaluated many combinations of these technologies in many applications.
In the NPRM, the agencies identified overarching potential issues in
assessing technology effectiveness and cost, including:
 Baseline vehicle technology level assessed as too low, or
too high. Compliance information was extensively reviewed and
supplemented with available literature on the vehicle models considered
in the analysis fleet. Manufacturers could also review the baseline
technology assignments for their vehicles, and the analysis
incorporates feedback received from manufacturers.
 Technology costs too low or too high. Tear down cost
studies, CBI, literature, and the 2015 NAS study information were
referenced to estimate technology costs. In cases where one technology
appeared to exceed all other technologies on cost and effectiveness,
information was acquired from additional sources to confirm or reject
assumptions. Cost assumptions for emerging technologies were reassessed
in cases where new information became available.
 Technology effectiveness too high or too low in
combination with other vehicle technologies. Technology effectiveness
was evaluated using the Autonomie full-vehicle simulation modeling,
taking into account the impact of other technologies on the vehicle and
the vehicle type. Inputs and modeling for the analysis took into
account laboratory test data for production and some pre-production
technologies, technical publications, manufacturer and supplier CBI,
and simulation modeling of specific technologies. Evaluating recently
introduced production products to inform the technology effectiveness
models of emerging technologies was preferred; however, some
technologies that are not yet in production were considered using CBI.
Simulation modeling used carefully chosen baseline configurations to
provide a consistent, reasonable reference point for the incremental
effectiveness estimates.
 Vehicle performance not considered or applied in an
infeasible manner. Performance criteria, including low speed
acceleration (0-60 mph time), high speed acceleration (50-80 mph time),
towing, and gradeability (six percent grade at 65 mph) were also
considered. In the simulation modeling, resizing was applied to achieve
the same performance level as the baseline for the least capable
performance criteria but only with significant design changes. The
analysis struck a balance by employing a frequency of engine downsizing
that took product complexity and economies of scale into account.
 Availability of technologies for production application
too soon or too late. A number of technologies were evaluated that are
not yet in production. CBI was gathered on the maturity and timing of
these technologies and the cadence at which manufacturers could adopt
these technologies.
 Product complexity and design cadence constraints too low
or too high. Product platforms, refresh and redesign cycles, shared
engines, and shared transmissions were also considered in the analysis.
Product complexity and the cadence of product launches were matched to
historical values for each manufacturer.
 Customer acceptance under estimated or over estimated.
Resale prices for hybrid vehicles, electric vehicles, and internal
combustion engine vehicles were evaluated to assess consumer
willingness to pay for those technologies. The analysis accounts for
the differential in the cost for those technologies and the amount
consumers have actually paid for those technologies. Separately, new
dual-clutch transmissions and manual transmissions were applied to
vehicles already equipped with these transmission architectures.
 The agencies sought comments on all assumptions for fuel economy
technology costs, effectiveness, availability, and applicability to
vehicles in the fleet.
 Several commenters compared the technology effectiveness and cost
estimates from prior rulemaking actions to the NPRM, some commenting
that the NPRM analysis represented a better balance of input from all
stakeholders regarding the potential costs and benefits of future fuel
economy improving technologies,\668\ and some commenting that the NPRM
analysis represented a step back from the Draft TAR and EPA's Proposed
Determination in terms of both the analysis itself and the resulting
conclusions about the level of technology required to meet the
[[Page 24381]]
augural standards.\669\ Specifically, while some commenters stated that
the Draft TAR and subsequent EPA midterm review documents had recently
concluded that augural standards were achievable with very low levels
of electrification based on currently available information on
technology effectiveness and cost,\670\ other commenters reiterated
that conventional gasoline powertrains alone were insufficient to
achieve post-2021 model year targets.\671\
---------------------------------------------------------------------------
 \668\ See, e.g., NHTSA-2018-0067-11928.
 \669\ See, e.g., NHTSA-2018-0067-11873.
 \670\ See, e.g., NHTSA-2018-0067-11969.
 \671\ See, e.g., NHTSA-2018-0067-12150.
---------------------------------------------------------------------------
 Generally, the automotive industry supported the agencies' NPRM
analysis over previous analyses. In addition to the automotive
industry's support of the agencies' use of one modeling tool for
analysis, discussed in Section IV, above, the industry also commented
in support of specific technology effectiveness, cost, and adoption
assumptions used in the updated analysis.
 The Alliance commented in support of the NPRM modeling approach,
and referenced important technology-specific features of the modeling
process, including ``The acknowledgement and application of real-world
limitations on technology application including a limit on the number
of engine displacements available to any one manufacturer, application
of shared platforms, engines, and transmissions, and the reality that
improvements and redesigns of components are not only extended across
vehicles but sometimes constrained in implementation opportunity to
common vehicle redesign cycles; recognition of the need for
manufacturers to follow ``technology'' pathways that retain capital and
implementation expertise, such as specializing in one type of engine or
transmission instead of following an unconstrained optimization that
would cause manufacturers to leap to unrelated technologies and show
overly optimistic costs and benefits; the application of specific
instead of generic technology descriptions that allow for the above-
mentioned real-world constraints; [and] the need to accommodate for
intellectual property rights in that not all technologies will be
available to all manufacturers.'' \672\
---------------------------------------------------------------------------
 \672\ NHTSA-2018-0067-12073, at 9.
---------------------------------------------------------------------------
 More specifically, the Alliance commented that the analysis
appropriately restricted the application of some technologies, like the
application of low rolling resistance tires on performance vehicles,
and limited aerodynamic improvements for trucks and minivans.\673\
Similarly, the Alliance commented in support of the decision to exclude
HCR2 technology from the analysis, citing previous comments stating
that ``the inexplicably high benefits ascribed to this theoretical
combination of technologies has not been validated by physical
testing.''
---------------------------------------------------------------------------
 \673\ NHTSA-2018-0067-12073, at 134.
---------------------------------------------------------------------------
 Ford commented more broadly that ``[t]he previous analyses
performed by the Agencies too often selected technology benefits from
the high-end of the forecasted range, and cost from the lower-end, in
part because deference was given to supplier or other third-party
claims over manufacturers' estimates.'' \674\ Ford noted that,
``[m]anufacturer estimates, while viewed as conservative by some, are
informed by years of experience integrating new technologies into
vehicle systems in a manner that avoids compromising other important
attributes (NVH, utility, safety, etc.),'' continuing that ``[t]he need
to preserve these attributes often limits the actualized benefit of a
new technology, an effect insufficiently considered in projections from
most non-OEM sources.'' Ford concluded, as mentioned above, that the
NPRM analysis better balanced these considerations.
---------------------------------------------------------------------------
 \674\ NHTSA-2018-0067-11928.
---------------------------------------------------------------------------
 Toyota commented that the discrepancy between the automotive
industry and prior regulatory assessments stemmed from ``agency
modeling relying on overly optimistic assumptions about technology cost
effectiveness and deployment rates.'' \675\ Toyota pointed to a prior
analysis that projected compliance for Toyota's MY 2025 lineup using
the ALPHA model as an example of how ``the agency's analysis failed to
account for customer requirements (cost, power, weight-adding options,
etc.) that erode optimal fuel economy, and normal business
considerations that govern the pace of technology deployment.'' In
contrast, Toyota stated that the ``[m]odeled technology cost,
effectiveness, and compliance pathways in the proposed rulemaking rely
on more recent data as well as more realistic assumptions about the
level of technology already on the road today, the pace of technology
deployment, and trade-offs between vehicle efficiency and customer
requirements.''
---------------------------------------------------------------------------
 \675\ NHTSA-2018-0067-12150.
---------------------------------------------------------------------------
 Honda, in its feedback on the models used in the standard setting
process, commented that ``the current version of the CAFE model is
reasonably accurate in terms of technology efficiency, cost, and
overall compliance considerations, and reflects a notable improvement
over previous agency modeling efforts conducted over the past few
years.'' \676\
---------------------------------------------------------------------------
 \676\ NHTSA-2018-0067-11818.
---------------------------------------------------------------------------
 FCA commented in recognition of the CAFE model improvements over
the Draft TAR version, but noted they ``continue to believe that the
cost and benefits used as inputs to the model are overly optimistic.''
\677\ FCA used its updated Jeep Wrangler Unlimited and Ram 1500 pickup
models as examples of vehicles that ``provide real life examples of the
costs and benefits that can be achieved with fuel and weight saving
technology;'' however, ``after all of the real world concerns such as
emissions, drivability, OBD, and fuels are considered, the benefits
observed remain less than those derived by the Autonomie model and used
as inputs to the Volpe model.''
---------------------------------------------------------------------------
 \677\ NHTSA-2018-0067-11943.
---------------------------------------------------------------------------
 Conversely, environmental groups, consumer groups, and some States
and localities commented that the Draft TAR and subsequent EPA analyses
were more representative of the current state of vehicle technologies.
These groups all generally commented, in different terms, that the NPRM
analysis technology effectiveness was understated and technology costs
were overstated, and additional constraints the agencies placed on the
analysis, like excluding technologies already in production or
constraining technology pathways, also helped lead to that result.\678\
---------------------------------------------------------------------------
 \678\ NHTSA-2018-0067-11873; NHTSA-2018-0067-11984.
---------------------------------------------------------------------------
 ICCT commented that the agencies ``ignored their own rigorous 2015-
2017 technological assessment, and have adopted a series of invalid and
unsupportable decisions which artificially constrain the availability
and dramatically under-estimate levels of effectiveness of many
different fuel economy improvement and GHG-reduction technologies and
unreasonably increase modeled compliance costs.'' \679\ ICCT also
commented that the agencies ignored, suppressed, dismissed, or
restricted the use of work done to update technologies and technology
cost and effectiveness assessments since the 2012 final rule for MYs
2017-2025. ICCT stated that the ``invalid high cost result [of the
modeled augural standards in 2025] was created by the agencies by
making many dozens of unsupported changes in the technology
effectiveness and availability inputs, the technology cost inputs, and
the technology package constraints.''
[[Page 24382]]
ICCT stated that ``the agencies failed to capture the latest available
information and, as a result, their assessment incorrectly and
artificially overstates technology costs.''
---------------------------------------------------------------------------
 \679\ NHTSA-2018-0067-11741 full comments.
---------------------------------------------------------------------------
 CARB commented that the agencies did not present sufficient new
evidence to change previous technical findings, specifically in regards
to conventional vehicle technologies.\680\ CARB stated that instead of
relying on new information, as had been asserted as justification for
the proposal, the analysis was based on older data that did not reflect
current technology. Accordingly, CARB pointed out that previous
analysis by the agencies projected far less need for electrification
than what was required in the proposal, stating that the underlying
cause is a reduction in the assumed cumulative improvements for what
advanced gasoline technology is able to achieve.
---------------------------------------------------------------------------
 \680\ NHTSA-2018-0067-11873.
---------------------------------------------------------------------------
 A coalition of States and Cities similarly commented that ``[t]he
Agencies' conclusions regarding the technology necessary to meet the
2025 standards and the cost of that technology run counter to the
evidence before the agency, diverge from prior factual findings without
explanation and without transparency as to the source of data relied
on, and are unsupported by any reasoned analysis. Such analysis bears
many hallmarks of an arbitrary and capricious action.'' \681\
---------------------------------------------------------------------------
 \681\ NHTSA-2018-0067-11735 (citing State Farm, 463 U.S. at 43;
Fox Television, 556 U.S. at 515; Humane Soc. of U.S. v. Locke, 626
F.3d 1040, 1049 (9th Cir. 2010)).
---------------------------------------------------------------------------
 Roush Industries, commenting on behalf of CARB, commented that
``the 2018 PRIA projected average costs for technology implementation
to achieve the existing standards to be significantly overstated and in
conflict with the 2016 Draft TAR cost estimates generated by the
Agencies only two years earlier.'' \682\ Roush commented that the Draft
TAR analyses of cost and incremental fuel economy improvement necessary
to achieve the augural standards was consistent with Roush's own
estimates and other published data.
---------------------------------------------------------------------------
 \682\ NHTSA-2018-0067-11984.
---------------------------------------------------------------------------
 Similarly, H-D Systems (HDS), commenting on behalf of the
California DOJ, commented that ``the estimates in the 2016 TAR on
technology cost and effectiveness still represent the correct estimates
based on the latest available data.'' \683\ HDS, in its analysis of the
costs of technologies to meet different potential standards between the
Draft TAR and the NPRM, noted that ``costs for most conventional (i.e.,
non-electric) drivetrain technologies were similar in both reports in
that costs were within +5% of the average of the costs from the two
reports. The only exception was the cost estimate for the High CR
second generation Atkinson cycle or HCR2 engine which was estimated to
be much more expensive. Due to differences in nomenclature,
transmission technology costs could not be directly compared but were
similar at the highest efficiency level. In contrast, cost of hybrid
technology was estimated to be much higher in the PRIA and were 200 to
250% higher for strong hybrids. Costs of drag reduction, rolling
resistance reduction and auxiliary system technologies were also quite
similar but the cost of mass reduction was substantially higher in the
PRIA by a factor of 2 to 3. Costs of engine friction reduction appear
not to be included in the cost computation for the PRIA although the
technology appears to be integrated into some of the engine technology
packages analyzed in the PRIA to estimate effectiveness.''
---------------------------------------------------------------------------
 \683\ NHTSA-2018-0067-11985.
---------------------------------------------------------------------------
 CFA commented that ``[t]he overarching discussion of technology
developments that introduces the NHTSA analysis is fundamentally flawed
and infects the entire proposal,'' taking issue with the NPRM statement
that ``some options considered in the original order for the National
Program ha[d] not worked out as EPA/NHTSA anticipated.'' \684\ CFA
commented that the agencies failed to note that some technology options
have performed better than anticipated, and ``the fact that some
technologies have done better than expected is a basis for increasing
the standards, not in the context of a mid-term review that was
supposed to tweak the long-term program.''
---------------------------------------------------------------------------
 \684\ NHTSA-2018-0067-12005.
---------------------------------------------------------------------------
 NCAT commented that the ``inflation of projected technology costs
does not appear to be attributable primarily to the projected cost of
any given technology, but rather to modeling constraints on the
application of such technologies to vehicles. Many of these constraints
appear to be arbitrary and NHTSA's departure from prior analyses in
these respects is not adequately supported.'' \685\
---------------------------------------------------------------------------
 \685\ NHTSA-2018-0067-11969.
---------------------------------------------------------------------------
 Environmental groups and States also commented that the agencies
either should reincorporate all the Draft TAR or the EPA Proposed and
Final Determination analyses' technologies, technology effectiveness
values, and technology costs into the analysis, and/or compare the
final rule analysis with those prior analyses to show how the updated
assumptions changed the results from those prior analyses.
 For example, ICCT commented that ``[f]or the agencies to conduct a
credible regulatory assessment they must remove all the technology
availability constraints, re-incorporate and make available the full
portfolio of technology options as was available in EPA's analysis for
the original 2017 Final Determination, and include at least 15 g/mile
CO2 for off-cycle credits by 2025, to credibly reflect the
real-world technology developments in the auto industry.'' \686\ ICCT
also stated that ``[t]he agencies need to identify each and every
technology cost input used in their modeling, and provide a clear
engineering and evidence based justification for why that cost differs
from the costs employed in the extremely well documented and well
justified Draft TAR and in EPA's 2016 TSD and 2017 Final Determination,
taking into account the above discussion of significant new evidence
developed since those prior estimates were made. Absent such disclosure
and justification, the default assumption needs to be that the prior
costs estimated based on the most recent data are more appropriate than
the estimates used for the proposal.''
---------------------------------------------------------------------------
 \686\ NHTSA-2018-0067-11741 full comments.
---------------------------------------------------------------------------
 In addition, groups of commenters were equally split on the ability
of technologies to meet different compliance targets. For example, the
Alliance commented that ``the only technologies that have demonstrated
the improvements necessary to meet the MY 2025 standards are strong
hybrids, plug-in electric vehicles, and fuel cell electric vehicles.
The Agencies' analysis for this Proposed Rule predict the need for
significant growth in sales of electrified vehicles, a finding
consistent with third-party analyses.'' \687\ In contrast, UCS
commented that electrified powertrains ``are not especially relevant
for the MY 2022-2025 regulations.'' \688\
---------------------------------------------------------------------------
 \687\ NHTSA-2018-0067-Alliance at 15.
 \688\ NHTSA-2018-0067-UCS at 23.
---------------------------------------------------------------------------
 The agencies are aware that the prior analyses concluded that
compliance with the augural standards could largely be met through
advances in gasoline vehicle technologies, and with only very low
levels of strong hybrids and electric vehicles. As the agencies stated
in the NPRM, consistent with both agencies' statutes, the proposal was
entirely de novo, based on an entirely new analysis reflecting the best
and most up-to-date information available to the agencies at the time
of this rulemaking.\689\ As discussed in Section IV, Section VI.B, and
further below, the NPRM and final rule analyses reflect updates to
[[Page 24383]]
technology effectiveness estimates, technology costs, and the
methodology for applying technologies to vehicles that the agencies
believed better represent the state of technology and the associated
costs compared to prior analyses, that result in pathways to compliance
that look both similar and different to those in prior analyses.
---------------------------------------------------------------------------
 \689\ 83 FR 42897.
---------------------------------------------------------------------------
 That said, several of the effectiveness and cost values used in the
NPRM and final rule analysis were directly carried over from the 2012
rule for MYs 2017-2025, Draft TAR, and EPA Midterm Evaluation
analyses.\690\ Several others were carried over from the 2015 NAS
report,\691\ which the agencies heavily relied upon in past analyses
even if specific cost or effectiveness values were not used. Different
technology effectiveness estimates, cost estimates, or adoption
constraints were employed where the agencies had information, from
technical reports, manufacturers, or other stakeholders, indicating
that a technology could or could not be feasibly adopted in the
rulemaking timeframe, or a technology could or could not be adopted in
the way that the agencies had previously modeled it. Notably, most
differences in pathways to compliance are attributable to only a few
significant differences between this rulemaking analysis and prior
rulemaking analyses.
---------------------------------------------------------------------------
 \690\ See, e.g., PRIA at 449, 451, 452, 453, 458.
 \691\ See, e.g., PRIA at 358-360.
---------------------------------------------------------------------------
 For example, as discussed in Section VI.B.3 Technology
Effectiveness and Modeling and Section VI.C.1 Engine Paths, in the EPA
Draft TAR and Proposed Determination analyses, effectiveness of HCR
engine technologies and downsized turbocharged engine technologies were
estimated using Tier 2 certification fuel. Tier 2 certified fuel has a
higher octane rating compared to regular octane
fuel.692 693 694 As summarized by EPA in the PD TSD, ``EPA's
estimate of effectiveness for gasoline-fueled engines and engine
technologies was based on Tier 2 Indolene fuel although protection for
operation in-use on Tier 3 gasoline (87 AKI E10) was included in the
analysis of engine technologies considered both within the Draft TAR
and Proposed Determination. Additionally, in the technology assessment
for this Proposed Determination, EPA has considered the required engine
sizing and associated effectiveness adjustments when performance
neutrality is maintained on 87AKI gasoline typical of real-world use.''
\695\
---------------------------------------------------------------------------
 \692\ Draft TAR at 5-228.
 \693\ Tier 2 fuel has an octane rating of 93. Typical regular
grade fuel has an octane rating of 87 ((R+M)/2 octane.
 \694\ EPA Proposed Determination TSD at 2-209 to 2-212.
 \695\ EPA Proposed Determination TSD at 2-210.
---------------------------------------------------------------------------
 NHTSA's effectiveness analysis for the Draft TAR used some engine
maps also developed using premium octane gasoline. However, at the time
NHTSA stated the agency would ensure all future engine model
development will be performed with regular grade octane gasoline.\696\
Commenters like Ford stated the effectiveness estimates for turbo
downsized engine packages were too high, in part because of the use of
high octane fuel. However they also commented in appreciation of
NHTSA's acknowledgement that any subsequent analysis would be based on
fuel at an appropriate octane level, as they stated the impact of the
change needed to be reflected in future analyses.\697\
---------------------------------------------------------------------------
 \696\ Draft TAR at 5-504, 5-512.
 \697\ Ford Motor Company Response to the Draft TAR September 26,
2016 NHTSA-2016-0068-0048, at 4.
---------------------------------------------------------------------------
 Engine specifications used to create the engine maps for the NPRM
and the final rule analysis were developed using Tier 3 fuel to assure
the engines were capable of operating on real world regular octane (87
pump octane = (R+M/2)). The process was similar to what manufacturers
must do to ensure engines have acceptable noise, vibration, harshness,
drivability, performance, and will not fail prematurely when operated
on regular octane fuel. This eliminated the need for any adjustments
that were applied in the 2016 Draft TAR and PD TSD to account for Tier
2 to Tier 3 fuel properties. This accounts for some of the
effectiveness and cost differences for engine technologies between the
Draft TAR/Proposed Determination and the NPRM/final rule. For more
details, see Section VI.C.1 Engine Paths.
 The agencies believe ICCT's and other commenters' assertions that
the engine maps should reflect Tier 2 fuel and not be updated for Tier
3 fuel would ignore these important considerations, and would provide
engine maps that could not achieve the fuel economy improvements unless
operated on high octane fuel. Therefore, the agencies determined that
engine maps developed for the Draft TAR and EPA Proposed Determination
that were based on Tier 2 fuel should not be used for the NPRM and
final rule analyses for these technical reasons.
 As another related example, the agencies described that prior
analyses had relied heavily on the availability of the HCR2 (or ATK2)
``future'' Atkinson Cycle engine as a cost-effective pathway to
compliance for stringent alternatives, but many engine experts
questioned its technical feasibility and near-term commercial
practicability.\698\ The agencies explained that EPA staff began
theoretical development of this conceptual engine with a best-in-class
2.0L Atkinson cycle engine and then increased the efficiency of the
engine map further, through the theoretical application of additional
technologies in combination, including cylinder deactivation, engine
friction reduction, and cooled exhaust gas recirculation. While the
potential of such an engine is interesting, nevertheless the engine
remains entirely speculative. No production HCR2/ATK2 engine, as
outlined in the EPA SAE paper,\699\ has ever been commercially
produced. Furthermore, the engine map has not been validated with
hardware, bench data, or even on a prototype level (as no such engine
exists to test to validate the engine map).
---------------------------------------------------------------------------
 \698\ 83 FR 43038.
 \699\ Schenk, C. and Dekraker, P., ``Potential Fuel Economy
Improvements from the Implementation of cEGR and CDA on an Atkinson
Cycle Engine,'' SAE Technical Paper 2017-01-1016, 2017. Available at
https://doi.org/10.4271/2017-01-1016.
---------------------------------------------------------------------------
 Vehicle manufacturers also commented on EPA's effectiveness
assumptions and estimates of HCR2/ATK2 model's future penetration
levels in the Draft TAR, stating ``[t]he effectiveness values for the
`futured' ATK2 package--projected at 40% penetration in 2025MY and
includes cooled exhaust gas recirculation (CEGR) and cylinder
deactivation (DEAC)--are too high, primarily due to overtly-optimistic
efficiencies in the base engine map, insufficient accounting of CEGR
and DEAC integration losses, and no accounting of the impact of 91RON
Tier 3 test fuel,'' and that ``44% fleet-wide penetration of ATK2 in
2025MY is unrealistic given the limited number of powertrain refresh
cycles available before 2025MY. In addition, it is unreasonable to
assume that OEMs already heavily invested in different high-efficiency
powertrain pathways (e.g., turbo-downsizing) would be able to commit
the immense resources needed to reach these high ATK2 penetration
levels in such a short time.'' \700\
---------------------------------------------------------------------------
 \700\ Ford Motor Company Response to the Draft TAR September 26,
2016 NHTSA-2016-0068-0048, at 4.
---------------------------------------------------------------------------
 Accordingly, the agencies decided to not include HCR2 technology in
the NPRM and final rule analysis. The engine model was not used because
no observable physical demonstration of the speculative technology
combination model has yet been created. Further,
[[Page 24384]]
many questions remain about the model's practicability as specified,
especially in high load, low engine speed operating conditions. The
HCR2 model combines multiple technologies to provide cumulative
estimate of benefits without consideration the practical interaction of
technologies. This approach runs contrary to the modeling approach
attempted in the NPRM and final rule analysis. The approach the
agencies tried to follow restricted models to adding discrete advanced
technologies. This approach allowed an accounting of synergetic
effects, identified incremental benefits, and increased the precision
of cost estimates.
 As another example, further discussed in Section VI.B.1 Analysis
Fleet, the agencies had traditionally taken different approaches to
assigning baseline road load reduction technology assignments. For
analyzing baseline levels of mass reduction in an analysis fleet, NHTSA
had developed for the Draft TAR a regression model to summarize a
vehicle's weight savings using a relative performance approach and
accounting for vehicle content, using cost curves developed from
teardown studies of a MY 2011 Honda Accord and MY 2014 Chevrolet
Silverado pickup truck. EPA developed its own methodology that
classified vehicles based on weight reductions from a MY 2008 vehicle,
compared to the MY 2014 version of the same vehicle, using a cost curve
from a tear-down study of a MY 2010 Toyota Venza. In the EPA's mass
reduction technology costing approach, a cost reduction was applied
when mass reduction 1 technology was applied to a system at mass
reduction 0 technology level. NHTSA's approach, used in the NPRM and
final rule analysis, set baseline mass reduction assignments so costs
of implementing mass reduction technologies are fully applied as
vehicle platforms move along the mass reduction technology path.
 The agencies also included additional advanced powertrain
technologies and other vehicle-level technologies in the technology
pathways between the Draft TAR and NPRM, and between the NPRM and final
rule. However, manufacturers and suppliers have repeatedly told the
agencies that there are diminishing returns to increasing the
complexity of advanced gasoline engines, including in the amount of
fuel efficiency benefit that they can provide. For example, Toyota
commented, in response to the EPA SAE paper benchmarking the 2018 Camry
with the 2.5L Atkinson-cycle engine and ``futuring'' midsize exemplar
vehicles based on the generated engine map,\701\ that although EPA's
addition of cylinder deactivation to the hypothetical 2025 exemplar
vehicle is technically possible and would provide some fuel economy and
CO2 benefit, the primary function of cylinder deactivation
is to reduce engine pumping losses which the Atkinson cycle and EGR
already accomplish on the 2018 Camry.\702\ Toyota concluded, ``The
overlapping and redundant measures to reduce engine pumping losses
would add costs with diminishing efficiency returns.'' Similarly,
BorgWarner commented that they ``do not expect that variable
compression ratio (VCR) or homogeneous charge compression ignition
(HCCI) will see broad application in the short term, if ever. While
each of these technologies can offer marginal efficiency gains at some
engine speed-load conditions, the use of down-sized boosted engines
with 8-10 speed transmissions makes it possible to run engines at near
optimum conditions and effectively minimizes gains from VCR or HCCI.
VCR mechanisms result in additional mass, cost and complexity, and true
HCCI has yet to be demonstrated in a production vehicle. The agencies
do not believe that OEMs will judge these technologies to be cost
effective.'' \703\
---------------------------------------------------------------------------
 \701\ Kargul, J., Stuhldreher, M., Barba, D., Schenk, C. et al.,
``Benchmarking a 2018 Toyota Camry 2.5-Liter Atkinson Cycle Engine
with Cooled-EGR,'' SAE Technical Paper 2019-01-0249, 2019,
doi:10.4271/2019-01-0249.
 \702\ NHTSA-2018-0067-12431, at 8.
 \703\ NHTSA-2018-0067-11895.
---------------------------------------------------------------------------
 So, while previous analyses may have shown pathways to compliance
with increasingly complex advanced gasoline engines, the NPRM and final
rule analyses more appropriately reflect that the most complex gasoline
engine technologies will account for a smaller share of manufacturers'
products during the rulemaking timeframe. However, despite this fact,
the NPRM and final rule analysis include more advanced powertrain
technologies than previous analyses, in part to account for important
considerations like intellectual property and the fact that some
manufacturers have already started down the path of incorporating a
certain advanced engine technology in their product portfolio, and that
abrupt switching to another advanced engine technology would result in
unrealistic stranding of capital costs. In addition, greater precision
in how cumulative technologies applied to engines, as estimated through
the Autonomie effectiveness modeling, appropriately reflects the
diminishing returns to efficiency benefits that those advanced engines
can provide. Moreover, as identified by a wide range of commenters,
battery costs are projected to fall in the rulemaking timeframe to a
point where, in the compliance modeling, it becomes more cost effective
to add electrification technologies to vehicles than to apply other
advanced gasoline engine technologies.
 Finally, the agencies declined to incorporate some information and
data for the NPRM or final rule central analysis for reasons discussed
in the following sections. In general, the data produced by agencies or
submitted by commenters failed to isolate effectiveness impacts of
individual technologies (or in some cases a combination of two or
several technologies). The data included effects from additional
unaccounted and undocumented technologies. Because the effectiveness
improvement measured or claimed resulted from more than just the
reported sources, the actual effectiveness of the technology or
technologies is obfuscated and easily under or over predicted. Using
effectiveness values generated in this manner carries a high risk of
double counting effectiveness and undercounting costs.
 In many cases, this problem exists where data or information is
based on laboratory testing or on-road testing of production vehicles
or components including engines and transmissions. Production vehicles
and components usually include multiple technology improvements from
one redesign to the next, and rarely incorporate just a single
technology change. Furthermore, technology improvements on production
vehicles in some cases cannot be readily observed, such as the level of
mechanical friction in an engine, and isolation and identification of
the improvement attributable to each technology would be impractical
given the costs and time required to do so. That said, in some cases,
where possible to do so, the agencies used the data or information from
production vehicles to corroborate information from the Autonomie
simulations. However, the agencies declined to apply that data or
information directly in the analysis if the effectiveness improvement
attributable to a particular technology could not be isolated.
 The agencies made these updates from prior analyses not, as some
commenters have suggested, to ``artificially overstate technology
costs,'' \704\ or to ``ignore the knowledge and expertise of the EPA
engineering
[[Page 24385]]
and compliance staff,'' \705\ ``so that the model in many instances
selects more expensive, less fuel efficient technology while excluding
less expensive and more efficient alternatives,'' \706\ but because the
updates reflected the agencies' reasonable assessment of the current
state of vehicle technologies and their costs, and the state of future
vehicle technologies and costs in the rulemaking timeframe.
---------------------------------------------------------------------------
 \704\ NHTSA-2018-0067-11741 at 7.
 \705\ NHTSA-2018-0067-11741 at I-23.
 \706\ NHTSA-2018-0067-12123.
---------------------------------------------------------------------------
 Separate from the decision to update assumptions used for the NPRM
analysis from prior analyses, the agencies did refine some technology
effectiveness and cost assumptions from the NPRM to this final rule
analysis. In addition to being appropriate for technical reasons, this
should address some commenters' overarching concerns about understated
technology effectiveness and overstated technology costs. For example,
several commenters noted that the costs of BISG/CISG systems were
higher for small Cars/SUVs and medium cars than for medium SUVs and
pickup trucks, which the Alliance and FCA described as ``implausible''
and ``misaligned with industry understanding,'' and which ICCT
described as ``contrary to basic engineering logic, which holds that a
system which would be smaller and have lower energy and power
requirements would be less expensive, not more.'' \707\ The agencies
agree, and have made changes to address this issue, as described in
Section VI.C.3.a) Electrification.
---------------------------------------------------------------------------
 \707\ NHTSA-2018-0067-11741.
---------------------------------------------------------------------------
 After considering comments, the agencies also added several engine
technologies and technology combinations for the final rule analysis.
These included a basic high compression ratio Atkinson cycle engine, a
variable compression ratio engine, a variable turbo geometry engine,
and a variable turbo geometry with electric assist engine (VTGe). The
NPRM discussed and provided engine maps for each of these technologies.
The agencies also added new technology combinations including diesel
engines with cylinder deactivation, turbocharged engines with advanced
cylinder deactivation, diesel engines paired with manual transmissions,
and diesel engines paired with 12-volt start-stop technology.
Transmission revisions included updating the effectiveness of 6-speed
automatic transmissions, applying updated shift logic for 10-speed
automatic transmissions, and increasing the gear span for efficient 10-
speed automatic transmissions. Mass reduction technology was expanded
to include up to 20 percent curb weight reduction, compared with up to
10 percent for the NPRM. These changes, and the comments upon which
they were based, are described in further detail in the following
sections.
1. Engine Paths
 The internal combustion (IC) engine is a heat engine that converts
chemical energy in a fuel into mechanical energy. Chemical energy of
the fuel is first converted to thermal energy by means of combustion or
oxidation with air inside the engine. This thermal energy raises the
temperature and pressure of the gases within the engine, and the high-
pressure gas then expands against the internal mechanisms of the
engine. This expansion is converted by the mechanical linkages of the
engine to a rotating crankshaft, which is the output of the engine. The
crankshaft, in turn, is connected to a transmission to transmit the
rotating mechanical energy to the desired final use, particularly the
propulsion of vehicles.
 IC engines can be categorized in a number of different ways
depending upon which technologies are designed into the engine: By type
of ignition (e.g., spark ignition or compression ignition), by engine
cycle (e.g., Otto cycle or Atkinson cycle), by valve actuation (e.g.,
overhead valve (OHV), single overhead camshaft (SOHC), or dual overhead
camshaft (DOHC)), by basic design (e.g., reciprocating or rotary), by
configuration and number of cylinders (e.g., inline four-cylinder (I4)
or V-shaped six-cylinder (V6)), by air intake (e.g., forced induction
(turbo or super charging) or naturally aspirated), by method of fuel
delivery (e.g., port injection or direction injection), by fuel type
(e.g., gasoline or diesel), by application (e.g., passenger car or
light truck),or by type of cooling (e.g., air-cooled or water-cooled).
For each combination of technologies among the various categories,
there is a theoretical maximum efficiency for all engines within that
set. There are various metrics that can be used to compare engine
efficiency, and the four metrics the agencies use or discuss in this
preamble are:
 Brake specific fuel consumption (BSFC), which is the mass
of fuel consumed per unit of work output (amount of fuel used to
produce power);
 Brake thermal efficiency (BTE), which is the total fuel
energy released per unit of work output (percentage of fuel used to
produce power);
 Fuel consumption (gallons per mile), which looks at the
gallons of fuel consumed per unit of work output (mile travelled); and
 Fuel economy (in MPG), which is the amount of work output
(miles travelled) per unit (gallon) of fuel consumed.
 When comparing the efficiency of IC engines, it is important to
identify the metric(s) used and the test cycle for the measurement
because results vary widely when engines operate over different test
cycles. Two-cycle fuel economy tests used to certify vehicles'
compliance with the CAFE standards tend to overestimate the average
fuel economy motorists will typically achieve during on-road
operation.\708\ In the NPRM and for this final rule analysis, the
agencies considered technology effectiveness for the 2-cycle test
procedures and AC and off-cycle test procedures to evaluate how
technologies could be applied for manufacturers to comply with
standards. The agencies also considered real world operation beyond
these test procedures when considering IC engine technologies in order
to assure the technologies were configured and specified in a manner
that could be used in real world vehicle applications.
---------------------------------------------------------------------------
 \708\ 77 FR 62988.
---------------------------------------------------------------------------
a) Fuel Octane
 As mentioned in other sections of the Preamble, the agencies go to
great lengths to ensure engine technologies considered for potential
compliance pathways are feasible for real-world implementation and
effectiveness. An important facet of this evaluation are both the fuels
that are used for efficiency testing and also the fuels that consumers
may purchase in the marketplace.
 In the NPRM, the agencies included a general overview of fuel
octane (stability) level, including levels currently available, and the
potential impact of fuel octane on engines developed for the U.S.
market.\709\ The agencies described that a typical, overarching goal of
optimal spark-ignited engine design and operation is to maximize the
greatest amount of energy from the fuel available, without manifesting
detrimental impacts to the engine over expected operating conditions.
Design factors, such as compression ratio, intake and exhaust value
control specifications, and combustion chamber and piston
characteristics, among others, are all impacted by the octane of the
fuel consumers are anticipated to use.\710\
---------------------------------------------------------------------------
 \709\ PRIA at 253.
 \710\ In addition, PRIA Chapter 6 contains a brief discussion of
fuel properties, octane levels used for engine simulation and in
real-world testing, and how octane levels can impact performance
under these test conditions.
---------------------------------------------------------------------------
[[Page 24386]]
 The agencies also discussed potential challenges associated with
octane levels available currently, and how those octane levels may play
a role in potential vehicle fuel efficiency improvements. Vehicle
manufacturers typically develop their engines and engine control system
calibrations based on the fuel available to consumers. In many cases,
manufacturers may recommend a fuel grade for best performance and to
prevent potential damage. In some cases, manufacturers may require a
specific fuel grade for both best performance, to achieve advertised
power ratings, and/or to prevent potential engine damage.
 Consumers, though, may or may not choose to follow the
manufacturer's recommendation or requirement for a specific fuel grade
for their vehicle. As such, vehicle manufacturers often choose to
employ engine control strategies for scenarios where the consumer uses
a lower than recommended, or required, fuel octane level, as a way to
mitigate potential engine damage over the life of a vehicle. These
strategies limit the extent to which some efficiency improving engine
technologies can be implemented, such as increased compression ratio
and intake system and combustion chamber designs that increase burn
rates and rate of in-cylinder pressure rise. If the minimum octane
level available in the market were higher (especially the current sub-
octane regular grade in the mountain states), vehicle manufacturers
might not feel compelled to design vehicles sub-optimally to
accommodate such blends.
 When knock (also referred to as detonation) is encountered during
engine operation, at the most basic level, non-turbocharged engines can
adjust the timing of the spark that ignites the fuel, as well as the
amounts of fuel injected at each intake stroke (``fueling''). In
turbocharged applications, knocking is typically controlled by
adjusting boost levels along with spark timing and/or the amount of
fuel injected. Past rulemakings discussed other techniques that may be
employed to allow higher compression ratios, including optimizing spark
timing, and adding of cooled exhaust gas recirculation (EGR).
Regardless of the type of spark-ignition engine or technology employed,
efforts to reduce or prevent knock with the lower-octane fuels that are
available in the market result in the loss of potential power output,
creating a ``knock-limited'' constraint on performance and efficiency.
 The agencies noted that despite limits imposed by available fuel
grades, manufacturers continue to make progress in extracting more
power and efficiency from spark-ignited engines. Production engines are
safely operating with regular 87 AKI fuel with compression ratios and
boost levels once viewed as only possible with premium fuel. According
to the Department of Energy, the average gasoline octane level has
remained fundamentally flat starting in the early 1980's and decreased
slightly starting in the early 2000s. During this time, however, the
average compression ratio for the U.S. fleet has increased from 8.4 to
10.52, a more than 20 percent increase. As explained by the Department
of Energy, ``[t]here is some concern that in the future, auto
manufacturers will reach the limit of technological increases in
compression ratios without further increases in the octane of the
fuel.'' \711\ As such, manufacturers are still limited by the fuel
grades available to consumers and the need to safeguard the durability
of their products for all of the available fuels; thus, the potential
improvement in the design of spark-ignition engines continues to be
overshadowed by the fuel grades available to consumers.
---------------------------------------------------------------------------
 \711\ Fact of the Week, Fact #940: August 29, 2016 Diverging
Trends of Engine Compression Ratio and Gasoline Octane Rating, U.S.
Department of Energy, https://www.energy.gov/eere/vehicles/fact-940-august-29-2016-diverging-trends-engine-compression-ratio-and-gasoline-octane (last visited Mar. 21, 2018).
---------------------------------------------------------------------------
 EPA and NHTSA also described ongoing research and positions from
automakers and advocacy groups on fuel octane levels, including
comments received during past agency rulemakings and on the 2016 Draft
TAR regarding the potential for increasing octane levels in the U.S.
market. The agencies described arguments for adjusting to octane
levels, including making today's premium grade the base grade of fuel
available, which could enable low cost design changes to improve fuel
economy and reduce tailpipe CO2 emissions. Challenges
associated with this approach include the increased cost to consumers
who drive vehicles designed for current regular octane grade fuel, who
would not benefit from the use of the higher cost higher-octane fuel.
The costs of such a transition to higher-octane fuel would be high and
persist well into the future, since unless current regular octane fuel
were unavailable in the North American market, manufacturers would be
effectively unable to redesign their engines to operate on higher-
octane fuel. In addition, the full benefits of such a transition would
not be realized until vehicles with such redesigned engines were
produced for a sufficient number of model years largely to replace the
current on-road vehicle fleet. The transition to net positive benefits
would take many years.
 The agencies also described input received from renewable fuel
industry stakeholders and from the automotive industry supporting high-
octane gasoline fuel blends to enable fuel economy and CO2
improving technologies such as higher compression ratio engines.
Stakeholders suggested that mid-level (e.g., E30) high-octane ethanol
blends should be considered and that EPA should consider requiring that
mid-level blends be made available at service stations. Stakeholders
supporting higher-octane blends suggested that higher-octane gasoline
could provide auto manufacturers with more flexibility to meet more
stringent standards by enabling opportunities for use of lower tailpipe
CO2 emitting technologies (e.g., higher compression ratio
engines, improved turbocharging, optimized engine combustion).
 The agencies sought additional comment in the NPRM on various
aspects of current fuel octane levels and how fuel octane could play a
role in the future. More specifically, the agencies sought comment on
how increasing fuel octane levels could have an impact on product
offerings and engine technologies, as well as what improvements to fuel
economy and tailpipe CO2 emissions could result from higher-
octane fuels. The agencies sought comment on an ideal octane level for
mass-market consumption, and whether there were downsides with
increasing the available octane levels and, potentially, eliminating
lower-octane fuel blends. EPA also requested comment on whether and how
EPA could require the production and use of higher-octane gasoline
consistent with Title II of the Clean Air Act.
 The agencies received numerous, wide-ranging comments in response
to the NPRM discussion, and some direct responses to the agencies'
requests for comments. The commenters included fuel producers,
individual vehicle manufactures, environmental groups, vehicle
suppliers, fuel advocacy groups, and agricultural organizations, among
others. Commenters provided a broad range of comments ranging from
explication of the many challenges to increasing available octane
levels, to claims of the substantial efficiency
[[Page 24387]]
increases that could be easily obtained by requiring higher-octane
levels.
 Several ethanol industry stakeholders commented in support of
requiring higher-octane fuels using mid-level ethanol blends. The High-
Octane, Low Carbon (HOLC) Alliance commented that it believes ``NHTSA
and EPA have a critical opportunity to cost-effectively ensure progress
in fuel efficiency and CO2 emissions standards. Scientific
experts agree that high-octane, low-carbon fuel can yield greater fuel
economy and emissions benefits when paired with internal combustion
engines (ICEs). But, to realize such benefits, automobile manufacturers
require approval sooner rather than later to such fuels. Alternatively,
automobile manufacturers will be limited in their ability to maximize
the environmental performance of their vehicles until non-liquid fuel
engines become more readily available. In finalizing the Proposed Rule,
the HOLC Alliance strongly urges EPA and NHTSA to establish a pathway
forward toward incentivizing the production and adoption of higher-
octane, lower carbon fuels. By doing so, EPA and NHTSA can continue to
incrementally increase CO2 and fuel economy standards,
respectively.'' \712\
---------------------------------------------------------------------------
 \712\ HOLC Alliance, Detailed Comments, EPA-HQ-OAR-2018-0283-
4196.
---------------------------------------------------------------------------
 Renewable Fuels Associations (RFA) commented that ``it strongly
believes vehicles and fuels must be considered together as integrated
systems. As EPA has recognized in the past, a `systems approach enables
emission reductions that are both technologically feasible and cost
effective beyond what would be possible looking at vehicle and fuel
standards in isolation.' Because ethanol-based high-octane low-carbon
fuel blends would enable cost-effective gains in fuel economy and
carbon dioxide reductions, the agencies should take steps to support
[high-octane low-carbon] fuels in the final SAFE rule.'' \713\
---------------------------------------------------------------------------
 \713\ RFA, Detailed Comments, EPA-HQ-OAR-2018-0283-4409.
---------------------------------------------------------------------------
 RFA cited several studies indicating benefits are available from
raising the floor of fuel octane levels currently available, and,
particularly, ``[t]he results from the studies reviewed generally
support a main conclusion that splash blending ethanol is a highly
effective means of rais