The Safer Affordable Fuel-Efficient (SAFE) Vehicles Rule III for Model Years 2022 to 2031 Passenger Cars and Light Trucks

Tire rolling resistance burns additional fuel when driving. As a car or truck tire rolls, at the point the tread touches the pavement, the tire flattens out to create what tire engineers call the contact patch. The rubber in the contact patch deforms to mold to the tiny peaks and valleys of the pavement. The interlock between the rubber and these tiny peaks and valleys creates grip. Every time the contact patch leaves the road surface as the tire rotates, it must recover to its original shape, and then as the tire goes all the way around, it must create a new contact patch that molds to a new piece of road surface. However, this molding and repeated re-molding action takes energy. Just like stretching a rubber band requires work, so does deforming the rubber and the tire to form the contact patch. When thinking about the efficiency of driving a car down the road, this means that not all the energy produced by a vehicle's engine can go into propelling the vehicle forward. Instead, some small, but appreciable, amount goes into deforming the tire and creating the contact patch repeatedly. This also explains why tires with low pressure have higher rolling resistance than properly inflated tires. When the tire pressure is low, the tire deforms more to create the contact patch, which is the same as stretching the rubber farther in the analogy above. Larger deformations consume even more energy, which results in worse fuel economy. Low rolling resistance tires have characteristics that reduce frictional losses associated with the energy dissipated mainly in the deformation of the tires under load, thereby improving fuel economy.

NHTSA uses three levels of low rolling resistance tire technology for the

light-duty analysis. Each level of low rolling resistance tire technology reduces rolling resistance by 10 percent from an industry- average rolling resistance coefficient (RRC) value of 0.009.\359\ RRC data from a NHTSA-sponsored study shows that similar vehicles across the light-duty vehicle categories have been able to achieve similar RRC improvements. Chapter 3.6 of the Final TSD presents more information on this comparison. Final TSD Chapter 3.6.1 shows the light-duty low rolling resistance technology options and their associated RRC.

\359\ See Technical Analysis of Vehicle Load Reduction by CONTROLTEC for California Air Resources Board (Apr. 29, 2015). NHTSA determined the industry-average baseline RRC using a CONTROLTEC study prepared for the CARB in addition to considering CBI submitted by vehicle manufacturers prior to the 2018 light-duty NPRM analysis. The RRC values used in this study were a combination of manufacturer information, estimates from coast-down tests for some vehicles, and application of tire RRC values across other vehicles on the same platform. The average RRC from surveying 1,358 vehicle models by the CONTROLTEC study is 0.009. The CONTROLTEC study compared the findings of their survey with values provided by the U.S. Tire Manufacturers Association for original equipment tires. The average RRC from the data provided by the U.S. Tire Manufacturers Association is 0.0092, compared to the average of 0.009 from CONTROLTEC.

NHTSA has been using ROLL10 and ROLL20 in the last several CAFE Model analyses. NHTSA has only recently included ROLL30 due to lack of widespread commercial adoption of ROLL30 tires in the fleet within past rulemaking timeframes, despite commenters' argument on availability of the technology on current vehicle models and the possibility that there would be additional tire improvements over the next decade.\360\ NHTSA has received comments in previous CAFE rules that also reflect the application of ROLL30 by OEMs, though they discourage considering the technology due to high cost and possible wet traction reduction. With increasing use of ROLL30 application by OEMs,\361\ and material selection making it possible to design low rolling resistance independent of tire wet grip (discussed in detail in Chapter 3.6 of the Final TSD), NHTSA considers ROLL30 as a viable future technology during this rulemaking period. NHTSA believes that the tire industry is in the process of moving automotive manufacturers towards higher levels of low rolling resistance technology in the vehicle fleet. NHTSA believes that, at this time, the emerging tire technologies that would achieve 30-percent improvement in rolling resistance, like changing tire profile, stiffening tire walls, employing novel synthetic rubber compounds, or adopting improved tires along with active chassis control, among other technologies, may be available for commercial adoption in the fleet during this rulemaking timeframe.

\360\ See The Safer Affordable Fuel-Efficient (SAFE) Vehicles Rule for Model Years 2021-2026 Passenger Cars and Light Trucks, Docket No. NHTSA-2018-0067-11985.

\361\ See NHTSA, Evaluation of Rolling Resistance and Wet Grip Performance of OEM Stock Tires Obtained from NCAP Crash Tested Vehicles Phase One and Two, Memorandum, NHTSA: Washington, DC, Docket No. NHTSA-2021-0053 (2021), available at: https://downloads.regulations.gov/NHTSA-2021-0053-0010/attachment_3.pdf (accessed: May 28, 2026); CONTROLTEC, LLC, Technical Analysis of Vehicle Load Reduction, California Air Resources Board: Sacramento, CA, Docket No. NHTSA-2021-0053-0010 (2015), available at: https://ww2.arb.ca.gov/sites/default/files/classic/research/apr/past/13-313.pdf (accessed: May 28, 2026); Evans, L. et al., NHTSA Tire Fuel Efficiency Consumer Information Program Development: Phase 2-- Effects of Tire Rolling Resistance Levels on Traction, Treadwear, and Vehicle Fuel Economy, DOT HS 811 154, NHTSA: Washington, DC, Docket No NHTSA-2008-0121-0035 (2009), available at: https://downloads.regulations.gov/NHTSA-2008-0121-0035/attachment_1.pdf (accessed: May 25, 2026).

Assigning low rolling resistance tire technology to the analysis fleet is difficult because RRC data are not part of tire manufacturers' publicly released specifications, and because vehicle manufacturers often offer multiple wheel and tire packages for the same nameplate. Consistent with previous rules, NHTSA uses a combination of CBI, data from a NHTSA-sponsored ROLL study, and assumptions about parts-sharing to assign tire technology in the analysis fleet. A slight majority of vehicles (54.9 percent) in the analysis fleet do not use any ROLL improvement technology (ROLL0), while 13.0 percent of vehicles use ROLL10, and 28.4 percent of vehicles use ROLL20. Only 3.7 percent of vehicles in the analysis fleet use ROLL30.

Lucid commented that the Draft TSD assumes all vehicles sharing a nameplate use identical rolling resistance technology in the fuel economy model. Lucid stated its view that this assumption is inaccurate, noting this approach “skews modeled fleet performance toward higher resistance values and undermines accuracy in estimating standards compliance.” \362\ NHTSA has determined that the approach used in the analysis is appropriate given the available data discussed above in this section and in Final TSD Chapter 3.6.2, and that the analysis reasonably captures the state of and application of low rolling resistance technology and its impact on fuel economy. Attempting to gather or generate specific rolling resistance values for each tire make and model for each vehicle in the analysis fleet would be unreasonably time consuming and expensive without contributing significantly to the robustness of the analysis. Lucid also commented, “NHTSA's analysis exaggerates cost burdens and fails to distinguish between low cost RRC improvements and wheel and tire size changes.” \363\ Consistent with previous rules, ROLL10 costs are in line with the supporting studies and reports discussed in Final TSD Chapter 3.6. Second, NHTSA concludes that using percent reductions in tire rolling resistance and the costs associated with the different levels of ROLL technology best represents the available technology and how the fleet could incorporate that technology. For this final rule, NHTSA has not made changes to the approach or costs for tire rolling resistance technologies.

\362\ Lucid, Docket No. NHTSA-2025-0491-6043, at 10.

\363\ Lucid, Docket No. NHTSA-2025-0491-6043, at 10.

The CAFE Model can apply ROLL technology at either a vehicle refresh or redesign. NHTSA recognizes that some vehicle manufacturers prefer to use higher RRC tires on some performance cars and SUVs. Since many performance cars have higher torque, to avoid tire slip, OEMs prefer to use higher RRC tires for these vehicles. Like the aerodynamic technology improvements discussed above, NHTSA applies ROLL technology adoption features based on vehicle HP and body style. As explained in Final TSD Chapter 3.6.3, all light-duty vehicles under 350 HP can adopt ROLL technology, and as vehicle HP increases, fewer vehicles can adopt the highest levels of ROLL technology. Final TSD Chapter 3.6 shows how effective the different levels of ROLL technology are at improving vehicle fuel consumption.

DMCs and learning rates for ROLL10 and ROLL20 are the same as prior analyses \364\ but are updated to the

dollar-year used in this analysis. In the absence of ROLL30 DMCs from tire manufacturers, vehicle manufacturers, or studies, NHTSA extrapolated the DMCs from ROLL10 and ROLL20 to develop the DMC for ROLL30. NHTSA believes that the added cost of each tire technology accurately represents the price difference that would be experienced by the different fleets. ROLL technology costs are discussed in detail in Chapter 3.6 of the Final TSD, and ROLL technology costs for all vehicle technology classes can be found in the Technologies Input File.

\364\ See Transportation Research Board, Tires and Passenger Vehicle Fuel Economy: Informing Consumers, Improving Performance, Special Report 286, The National Academies Press: Washington, DC (2006), available at: https://nap.nationalacademies.org/catalog/11620/tires-and-passenger-vehicle-fuel-economy-informing-consumers-improving-performance (accessed: May 28, 2026); NHTSA, Corporate Average Fuel Economy for MY 2011 Passenger Cars and Light Trucks, Final Regulatory Impact Analysis, NHTSA: Washington, DC (2009), available at: https://www.nhtsa.gov/sites/nhtsa.gov/files/cafe_final_rule_my2011_fria.pdf (accessed: May 28, 206); EPA and NHTSA, Joint Technical Support Document: Rulemaking to Establish Light-Duty Vehicle Greenhouse Gas Emission Standards and Corporate Average Fuel Economy Standards, EPA-420-R-10-901, EPA and NHTSA: Washington, DC, p. 3-77 (2010), available at: https://www.nhtsa.gov/sites/nhtsa.gov/files/final_joint_tsd.pdf (accessed: May 28, 2026); EPA and NHTSA, Draft Technical Assessment Report: Midterm Evaluation of Light-Duty Vehicle Greenhouse Gas Emission Standards and Corporate Average Fuel Economy Standards for Model Years 2022-2025, EPA-420-D-16-900, EPA and NHTSA: Washington, DC, pp. 5-153 and 154, 5-419 (2016), available at: https://www.nhtsa.gov/sites/nhtsa.gov/files/draft-tar-final.pdf (accessed: May 28, 2026). In brief, the estimates for ROLL10 are based on the incremental $5 value for four tires and a spare tire in the NAS/NRC Special Report and confidential manufacturer comments that provided a wide range of cost estimates. The estimates for ROLL20 are based on incremental interpolated ROLL10 costs for four tires (as NHTSA and EPA believed that ROLL20 technology would not be used for the spare tire) and are seen to be fairly consistent with CBI suggestions by tire suppliers.

Under EPA's current procedures for determining fleet average fuel economy for CAFE compliance, manufacturers may generate FCIVs, which improve their fuel economy values. Manufacturers may generate FCIVs for the addition of OC and AC efficiency technologies, which can provide fuel economy benefits in real-world vehicle operation that are not fully captured using the 2-cycle test procedures (e.g., FTP and HFET) used to measure fuel economy.\365\ Starting in MY 2027, only automobiles powered by ICEs are eligible to generate FCIVs, and the OC FCIV program is currently being phased out between MYs 2031-2033, with manufacturers no longer being able to generate OC FCIVs for MY 2033 and beyond. OC technologies can include, but are not limited to, thermal control technologies, high-efficiency alternators, and high-efficiency exterior lighting. As an example, manufacturers can generate FCIVs for the addition of thermal control technologies like active seat ventilation and solar reflective surface coating, which help to regulate the temperature within the vehicle's cabin--making it more comfortable for the occupants and reducing the use of low-efficiency heating, ventilation, and air-conditioning (HVAC) systems. AC efficiency technologies are technologies that reduce the operation of or the loads on the compressor, which pressurizes AC refrigerant. The less the compressor operates or the more efficiently it operates, the less load the compressor places on the engine or battery storage system, resulting in better fuel efficiency. AC efficiency technologies can include, but are not limited to, blower motor controls, internal heat exchangers, and improved condensers/evaporators.

\365\ See 49 U.S.C 32904(c) (“The Administrator shall measure fuel economy for each model and calculate average fuel economy for a manufacturer under testing and calculation procedures prescribed by the Administrator . . . [T]he Administrator shall use the same procedures for passenger automobiles the Administrator used for model year 1975 (weighted 55 percent urban cycle and 45 percent highway cycle), or procedures that give comparable results.”).

Since EPA first proposed allowing manufacturers to earn FCIVs for AC efficiency and OC technologies, NHTSA has not modeled AC efficiency and OC technologies in the CAFE Model like other vehicle technologies, for several reasons. Each time NHTSA adds a technology option to the CAFE Model's technology pathways, the agency increases the number of Autonomie simulations by approximately a hundred thousand. This means that adding just five AC efficiency and five OC technology options would double the agency's Autonomie simulations to around 2 million total simulations. Instead, for applicable model years, the CAFE Model applies predetermined AC efficiency and OC benefits to each manufacturer's fleet after the CAFE Model applies traditional technology pathway options. The CAFE Model attempts to apply pathway technologies and AC efficiency and OC technologies in a way that both minimizes cost and allows the manufacturer to meet a given CAFE standard without over-or under-complying. The predetermined benefits that the CAFE Model applies for AC efficiency and OC technologies are based on manufacturers' MY 2024 mid-model year CBI compliance reports.

NHTSA uses manufacturers' MY 2024 AC efficiency and OC FCIVs as a starting point for each regulatory class, then holds those values constant from MYs 2024-2031 for the No-Action Alternative and through MY 2027 for action alternatives. Unlike previous versions of this analysis, NHTSA does not extrapolate the MY 2024 values to future model years. Instead, the CAFE Model assumes that FCIVs for MY 2027 will be the same as they were for MY 2024. Manufacturers have been able to settle in on a level of AC efficiency and OC technologies that maximize their ROI; therefore, NHTSA does not anticipate a significant increase in manufacturers' AC efficiency and OC FCIVs between MYs 2024-2027 for any regulatory category. Additional details about how NHTSA determines AC efficiency and OC technology application rates are discussed Chapter 3.7 of the Final TSD.

Because the CAFE Model applies AC efficiency and OC technology benefits independent of the technology pathways, NHTSA must account for the costs of those technologies independently, as well. NHTSA generates costs for these technologies on a dollars per gram of CO2 per mile ($ per g/mi) basis, as AC efficiency and OC technology benefits are applied in the CAFE Model on a gram per-mile basis (as in the regulations). NHTSA updates the AC efficiency and OC technology costs by implementing an updated calculation methodology and converting the DMCs to 2024 dollars. The AC efficiency costs are based on data from EPA's 2010 FRIA and the 2010 and 2012 Joint NHTSA/EPA TSDs.366 367 368 NHTSA has used data from EPA's 2016 Proposed Determination TSD \369\ to develop the updated OC costs that were used for the 2022 final rule and now this final rule.

\366\ EPA, Final Rule for Model Year 2012--2016 Light-Duty Vehicle Greenhouse Gas Emission Standards and Corporate Average Fuel Economy Standards, last revised: Apr. 23, 2026, available at: https://www.epa.gov/regulations-emissions-vehicles-and-engines/final-rule-model-year-2012-2016-light-duty-vehicle (accessed: May 28, 2026) (hereinafter, “Final Rulemaking MYs 2012-2016”).

\367\ Final Rulemaking MYs 2012-2016.

\368\ EPA and NHTSA, Joint Technical Support Document: Final Rulemaking for 2017-2025 Light-Duty Vehicle Greenhouse Gas Emission Standards and Corporate Average Fuel Economy Standards, EPA-420-R- 12-901, EPA and NHTSA: Washington, DC (2012), available at: https://www.nhtsa.gov/sites/nhtsa.gov/files/joint_final_tsd.pdf (accessed: May 28, 2026).

\369\ EPA, Proposed Determination on the Appropriateness of the Model Year 2022-2025 Light-Duty Vehicle Greenhouse Gas Emissions Standards under the Midterm Evaluation: Technical Support Document, EPA-420-R-16-020, EPA: Washington, DC (2016), available at: https://downloads.regulations.gov/EPA-HQ-OAR-2022-0829-0230/attachment_1.pdf (accessed: May 28, 2026).

In the NPRM and for this final analysis, NHTSA removed FCIVs from its standard-setting analysis starting with MY 2028, which is the first year in which a removal of FCIVs could go into effect.\370\ NHTSA initially made this change in the NPRM to align with its conclusion that technology-specific incentives should not be considered when running the compliance simulation that informs its consideration of maximum feasible standards. For MY 2028 and beyond, NHTSA's analysis is based on simulating compliance based on 2-cycle testing. To simulate compliance

pathways using the CAFE Model without AC efficiency and OC technologies, NHTSA sets the maximum allowable FCIV to 0 g CO2 (CO2)/mi in the Scenarios Input File. Section VI contains a more detailed discussion of how AC efficiency and OC benefits affect compliance with NHTSA's fuel economy standards.\371\

\370\ 49 U.S.C. 32904(d).

\371\ Compliance with NHTSA's fuel economy standards is determined in accordance with EPA's calculation procedures at 40 CFR 600.512.

In addition, NHTSA believes that the FCIVs generated under the OC and AC efficiency programs are no longer representative of actual fuel savings. The values for adding such technologies were estimated from emission-reduction assessments performed on MY 2008 automobiles. As fuel economy has improved in the model years since these assessments were performed, the FCIVs for adding OC technologies have increasingly represented a larger percentage improvement in fuel economy values. As a result, the values for FCIVs have become less representative of actual fuel savings and have created market distortions by incentivizing the addition of technology that does not provide commensurate fuel savings. NHTSA sought comment on this determination. Additional details and assumptions used for AC efficiency and OC costs are discussed in Chapter 3.7.2 of the Final TSD.

Comments from auto manufacturers and supplier groups supported the exclusion of AC efficiency and OC values from the standard-setting analysis. The Alliance \372\ agreed that technology-specific incentives should not be used to justify higher standards. Historically, including these values has increased stringency of the standards, effectively requiring their application. Ford Motor Company (Ford), Hyundai, Nissan, Porsche Cars North America (Porsche), and Stellantis \373\ support removing AC efficiency and FCIVs from the analysis used to set standards starting in MY 2028. MEMA, The Vehicle Suppliers Association and Specialty Equipment Market Association (SEMA) \374\ support the decision to base standards exclusively on gas and diesel vehicle fuel economy, excluding adjustments for OC technologies during the standard- setting phase.

\372\ The Alliance, Docket No. NHTSA-2025-0491-5707-A2, at 40.

\373\ Ford, Docket No. NHTSA-2025-0491-5821-A1, at 6; Hyundai, Docket No. NHTSA-2025-0491-4972-A1, at 2; Nissan, Docket No. NHTSA- 2025-0491-5857-A1, at 5; Porsche, Docket No. NHTSA-2025-0491-0044, at 3; Stellantis, Docket No. NHTSA-2025-0491-5968-A1, at 11.

\374\ MEMA, Docket No. NHTSA-2025-0491-5859-A1, at 2; SEMA, Docket No. NHTSA-2025-0491-5891-A1, at 1-2.

NHTSA did not receive any comments requesting a different modeling approach for AC and OC FCIVs, nor any comments arguing that NHTSA should continue to include the simulation of these technologies and their FCIVs in the analysis. Therefore, NHTSA has made no changes to the analysis from the NPRM regarding AC efficiency and OC technology FCIVs. As described above, FCIVs are removed from the regulatory alternatives starting in MY 2028 and remain in the No-Action Alternative per the limits and phase-out schedule in EPA's regulations.

The preceding subsections of Section II discussed how manufacturers might respond to the final standards. While the technology analysis outlined different compliance strategies available to manufacturers, the costs and benefits that would accrue because of the final standards are dependent on how consumers respond to manufacturers' compliance decisions. The following subsections describe how the agency models potential consumer response to changes in vehicle prices and attributes caused by manufacturer compliance decisions, as simulated by the CAFE Model. 1. Macroeconomic and Consumer Behavior Assumptions

Most of the economic effects simulated within the analysis are influenced by macroeconomic conditions outside the agency's influence. For example, fuel prices are determined mainly by global petroleum supply and demand, yet they affect how much fuel efficiency-improving technology U.S. manufacturers would apply to their vehicles, how much more consumers would be willing to pay to purchase vehicles offering higher fuel economy, how much buyers would drive those vehicles, and the value of each gallon of fuel saved from improved fuel efficiency. Forecasting the consequences of CAFE standards requires robust projections of demographic and macroeconomic variables that span the full timeframe of the analysis, including real GDP, consumer confidence, U.S. population, and real disposable personal income.

The analysis presented with the final rule utilizes fuel price projections developed by EIA, an agency within DOE that collects, analyzes, and disseminates independent and impartial energy information to promote sound policy-making and public understanding of energy. EIA uses its National Energy Modeling System (NEMS) to produce its AEO, which presents projections of future fuel prices (among many other economic and energy-related variables). The agency's analysis for the final rule uses AEO's 2026 Alternative Transportation and Electricity case projections of U.S. population, GDP, disposable personal income, GDP deflator, and fuel prices.\375\ NHTSA uses AEO's 2026 Alternative Transportation and Electricity case because this case is intended to reflect recently finalized changes to policy and therefore provides a more informed analysis of conditions that will affect fuel prices than the Reference Baseline case, especially in the near-term. The analysis also relies on the S&P Global forecasts of the University of Michigan's Consumer Sentiment Index from its fall 2025 U.S. Macroeconomic Forecast, which EIA also uses to develop the projections it reports in its AEO.\376\

\375\ Similarly, for the proposal NHTSA used the Alternative Transportation case from the 2025 AEO for projections related to macroeconomic variables and fuel prices, and its Alternative Electricity case for electricity price forecasts. For the 2026 AEO, the EIA combined these two cases into the case that NHTSA is using in this final rule.

\376\ NHTSA sourced the data from IHS-Polk. S&P Global purchased IHS Markit and rights to this data in 2022.

These macroeconomic assumptions are important inputs to the analysis, but they are also uncertain, particularly over the long lifetimes of the vehicles affected by this final rule. To reflect the effects of this uncertainty, the agency also uses the 2026 AEO's Low Oil and Gas Supply and High Oil and Gas Supply side cases, and the 2025 AEO's Low Oil Price and High Oil Price side cases to analyze the sensitivity of its analysis to alternative fuel price projections.\377\ The purpose of the sensitivity analysis, which is discussed in greater detail in Chapter 9 of the FRIA, is to measure the degree to which different assumptions about fuel prices can change simulated outcomes. NHTSA similarly uses low and high economic growth cases from the 2026 AEO forecast as bounding cases for the macroeconomic variables in its analysis.

\377\ The 2026 AEO did not include either a Low or High Oil Price side case.

NHTSA sought and received comments on these data sources. Prime Mover Institute (PMI) criticized past price forecasts from the AEO, which overestimated the long-term level of prices, but supported NHTSA's choice in the proposal to rely on the Alternative Transportation case from the 2025 AEO because of the mix of policies

included in this case as compared to the Reference case.\378\ In contrast, an individual commenter criticized the use of the Alternative Transportation case forecast, claiming that this artificially deflated the level of fuel prices and thus the benefits from higher CAFE standards.\379\

\378\ PMI, Docket No. NHTSA-2025-0491-5001-A2, at 47.

\379\ Anonymous, Docket No. NHTSA-2025-0491-5040.

In response to these comments, NHTSA acknowledges that it is important to rely on a forecast that uses up-to-date assumptions about the most likely effects of policy on energy markets and the economy. The forecast NHTSA used in the proposal for gasoline prices was higher than the Reference case forecast, and this remains the case in the final rule. For 2031-2050, crude oil prices are also higher in the Alternative Transportation case that NHTSA is using than in the other AEO cases.

The analysis presented for this final rulemaking uses a 2024 base year, consistent with the use of vehicle data for MY 2024, and data for that year represents actual observations rather than estimates to the extent possible. Chapter 4.1 of the Final TSD discusses macroeconomic forecasts and assumptions NHTSA uses in this analysis.

Another key assumption used throughout the agency's analysis is how much consumers are willing to pay for improved fuel economy, including how much they are willing to pay when these fuel savings are accompanied by sacrifices in other vehicle features. The payback period assumption also has important implications for other regulatory analysis results, including the effect of standards on sales and the use of new vehicles, as well as the number and use of older, used vehicles. The NPRM proposed to change the payback period assumption from 30 months of fuel savings used in prior analyses to 36 months. The agency has updated its review of the academic literature on willingness to pay as part of its analysis of this final rule, which is discussed in Final TSD Chapter 4.21 and FRIA Chapter 2.1.2. As noted in previous rulemakings, the range of estimates presented in the literature is wide. Some of the studies conclude that consumers value much of the potential savings in fuel costs from driving higher mpg vehicles, while others conclude that consumers significantly undervalue expected fuel savings. The more recent studies suggest that consumers value somewhere between 24 months and the full lifetime value of undiscounted fuel savings, which is also supported by several of the older studies.\380\

\380\ See Final TSD Chapter 4.21 and FRIA Chapter 2.1.2 for literature references.

Manufacturers have repeatedly informed the agency that they believe consumers only value between two to three years of fuel savings when choosing among competing models to purchase,\381\ and the plurality of consumers when surveyed about their payback preferences have stated they are willing to pay for technology that repays the upfront cost within 24 months.\382\

\381\ See, e.g., 87 FR 25710, 25856 (May 2, 2022).

\382\ Some survey data such as Consumer Reports shows consumers with lower payback periods (around 24 months). However, the methodology employed by surveys like Consumer Reports are less rigorous than the revealed preferences data from the other sources, which is why Circular A-4 directs the agencies to attempt to use studies that rely on revealed preferences when feasible.

The agency also performed a retrospective analysis using the CAFE Model with reference fleets created to support prior rules. For the NPRM, the agency modeled how the 2020 reference fleet (used for the 2022 final rule), similarly projected forward, compared with the 2022 reference fleet (used in the 2024 final rule) and how the 2022 reference fleet (used for the 2024 final rule) projected forward with different payback assumptions compared with the 2024 reference fleet. These simulations provided model predictions about the technology penetration rates under different assumptions about the length of the payback period and under different projections of future fuel prices and technology costs. By comparing these to actual penetration rates, NHTSA could assess the Model's ability to predict technology adoption under each payback assumption. NHTSA found that the payback assumption that predicted technology adoption most accurately is 36 months, followed by 30 months. Both longer and shorter payback periods create a larger divergence; that is, when NHTSA tested longer and shorter payback periods, the model's predictions become less accurate, compared to the actual use of technology observed in MY 2024.

After weighing the results from the academic literature, previous statements from manufacturers, and the agency's retrospective analysis, NHTSA used a 36-month payback assumption for the analysis supporting the proposal. In coming to this determination, NHTSA was persuaded by its retrospective analysis, which found that a 36-month payback period assumption yielded greater consistency between the modeled 2024 fleet and observed 2024 technology application and achieved compliance. This estimate is also consistent with the literature and what manufacturers have consistently relayed to the agency. While this estimate represents a longer payback period assumption than was applied in the analysis of the previous three CAFE rules, the agency stated in the proposal its belief that the preponderance of the evidence suggests that 36 months is appropriate.

NHTSA sought comments on whether this is an appropriate representation of consumer willingness to pay higher upfront vehicle prices for future fuel savings. Several commenters criticized the assumption of a 36-month payback period. Consumer Reports argued that this was too short and at odds with survey evidence, while ZETA and other commenters argued 36 months was at odds with NHTSA's survey of the literature regarding consumer willingness to pay for fuel economy, and criticized NHTSA's survey as too selective.\383\ Another commenter suggested that NHTSA should use a longer payback period, noting that the average vehicle age is now 12.8 years.\384\ Several commenters disputed NHTSA's assumption that automakers will add technologies with a 36-month payback voluntarily in the absence of standards, stating this has no historical basis and artificially inflates the costs and pollution associated with the proposed rule.\385\ UCS argued that, because there were not significant changes in achieved fuel economy during the period in which standards were not increasing, modeling voluntary adoption of fuel-saving technologies that pay back within 36 months was inappropriate.\386\ UCS argued that no voluntary over- compliance should be assumed, which is equivalent to a 0 month payback period. They further argued that this assumption is inconsistent with NHTSA's inclusion of the implicit opportunity cost (IOC), because in principle consumers may value performance improvements more.\387\ Another commenter suggested that if

industry truly believed consumers valued fuel economy improvements with a 36-month payback, they would invest in efficiency without Federal mandates.\388\ PMI supported NHTSA's choice to model a 36 month payback period in conjunction with its inclusion of the IOC, arguing that this was a reasonable proxy for buyers' preference for other vehicle characteristics, as opposed to being the result of myopia.\389\

\383\ Consumer Reports, Docket No. NHTSA-2025-0491-5926-A1, at 1, 4; ZETA, Docket No. NHTSA-2025-0491-6039-A2, at 27-29; UCS, Docket No. NHTSA-2025-0491-6027-A1, at 20-23; Attorneys General, Docket No. NHTSA-2025-0491-6064-A4, at 27-29.

\384\ Diana Furchtgott-Roth, Docket No. NHTSA-2025-0491-5765-A1, at 14.

\385\ ICCT, Docket No. NHTSA-2025-0491-5240-A2, at 13-16; ACEEE, Docket No. NHTSA-2025-0491-5943-A1, at 2-3; UCS, Docket No. NHTSA- 2025-0491-6027-A1, at 41-45.

\386\ UCS, Docket No. NHTSA-2025-0491-6027, Technical Appendix at 41-44.

\387\ UCS, Docket No. NHTSA-2025-0491-6027, Technical Appendix at 45-47.

\388\ Diana Furchtgott-Roth, Docket No. NHTSA-2025-0491-5765-A1, at 14.

\389\ PMI, Docket No. NHTSA-2025-0491-5001-A2, at 48-49.

The range of comments NHTSA received is consistent with the range of findings in the literature surrounding how much consumers value fuel savings. As noted in Final TSD Chapter 4.2.1.1, NHTSA's updated survey of the literature found a wide range of estimates that vary based on factors such as how the value was elicited (stated preference surveys vs. revealed preference surveys and market sales data). Including additional sources in NHTSA's review of the literature did not change NHTSA's findings about the approximate range of estimates. Indeed, one large survey found that, across different methodologies and datasets, the range of average estimates lined up closely with NHTSA's findings.\390\ Ultimately, NHTSA must choose a set of assumptions to make about consumer behavior in its modeling. For this reason, to determine an appropriate payback period assumption, NHTSA relied on retrospective analysis comparing results obtained from its model using a range of different payback periods. This allows NHTSA to use a payback period that produces realistic results in its simulations. NHTSA found that using a payback period of 36 months provided the most accurate representation of consumer behavior across the range of technologies modeled. Using significantly shorter or longer payback period assumptions appears to produce less representative results when predicting technology adoption.

\390\ Greene, D. et al., Consumer willingness to pay for vehicle attributes: What do we know?, Transportation Research Part A: Policy and Practice, Vol. 118(c): pp. 258-79 (2018), available at: https://www.sciencedirect.com/science/article/abs/pii/S0965856417308546 (accessed: May 28, 2026).

NHTSA's retrospective analysis separately examined the technology adoption by manufacturers for vehicle fleets that over-complied with the MY 2024 standards and found that its assumption of 36 months outperformed shorter payback periods.\391\ This contradicts the contention by commenters that manufacturers will not apply fuel-saving technologies in the absence of higher standards. Therefore, eliminating cost-effective technology adoption in NHTSA's analysis would bias downward the achieved fuel economy levels in the absence of standards, and would bias upwards the technology costs of the standards. NHTSA also disagrees with UCS's argument that inclusion of the IOC precludes voluntary over-compliance. The technologies adopted under voluntary over-compliance are those whose fuel savings most significantly exceed upfront costs. Crucially, the fuel savings of vehicles with these technologies by definition are large enough that the portion of their fuel savings that consumers are assumed to value exceeds the sum of both the upfront additional costs of the technology and NHTSA's estimate of the IOC.

\391\ This included individual regulatory-class-fleets for BMW, Mercedes-Benz, Stellantis, Honda, Hyundai, Kia, JLR, Nissan, Toyota, Volvo, and VWA.

The contention that consumer valuation of fuel savings supports a longer payback period might not consider potential associated trade- offs in vehicle attributes. For example, survey evidence may indeed support the idea that consumers are willing to pay more for a vehicle with greater fuel economy if the question used in the survey does not include the caveat that the vehicle will come equipped with a less advanced set of features, less interior space, less power, or worse performance than expected. Similarly, consumers may indeed show a willingness to pay more for an individual vehicle model when fuel prices are higher (and projected fuel savings are greater) versus when fuel prices are lower. However, this does not address directly how a consumer would react to a combination of changes that includes new fuel-saving technology, changes to other vehicle features, and a higher sticker price.

Accordingly, NHTSA has concluded that increasing the payback period from 30 months in previous rulemakings to 36 months for this rulemaking is appropriate. Recognizing the consequences of the payback assumption in the agency's regulatory analysis, NHTSA also includes sensitivity cases to examine the impacts of longer and shorter payback periods in Chapter 9 of the FRIA. These concepts are explored more thoroughly in Chapter 4.2.1.1 of the Final TSD and Chapter 2.1 of the FRIA. 2. Fleet Composition

The composition of the on-road fleet--and how it changes in response to standards--determines many of the costs and benefits of the final rule. For example, how much fuel is consumed depends on the number and efficiency of new vehicles sold and how rapidly older, less efficient, less safe vehicles are retired. Reducing the stringency of the CAFE standards would lower the price of new vehicles compared to the No-Action Alternative and would lead to a relative increase in sales of newer, safer vehicles, which in turn would decrease the price of used vehicles leading to the quicker retirement of the oldest, least safe, and less fuel-efficient vehicles on the road.

The analysis accompanying the proposal and now this final rule simulates changes in the vehicle fleet's size, composition, and usage as manufacturers and consumers respond to regulatory alternatives, fuel prices, and macroeconomic conditions. The analysis of fleet composition examines two primary drivers: how sales of new vehicles and their integration into the existing fleet change in response to each regulatory alternative, and how economic and regulatory factors influence the retirement of used vehicles from the fleet (scrappage). NHTSA models sales and scrappage independently.

CAFE standards have been rising every year for nearly two decades. This constant increase in standards has been accompanied by a rise in both the cost of new vehicles and the age of the on-road fleet. The average selling price for new cars and light trucks rose nearly 50 percent between 2012 and 2024 and now approaches $50,000,\392\ while average U.S. household income increased only about half as much over that same period.\393\ Meanwhile, the total number of cars and light trucks in use rose by about 30 million, with the entire increase representing used vehicles, while their average age rose

from 10.6 to 12.6 years.\394\ Below are brief descriptions of how the agency models sales and scrappage; for full explanations, readers should refer to Chapter 4.2.1 of the Final TSD.

\392\ Cox Automotive, Kelley Blue Book Report: Annual New- Vehicle Price Gains Slow in April as Market Hits Headwinds, last revised: May 12, 2026, available at: https://mediaroom.kbb.com/2026-05-12-Kelley-Blue-Book-Report-Annual-New-Vehicle-Price-Gains-Slow-in-April-as-Market-Hits-Headwinds (accessed: May 7, 2026).

\393\ U.S. Census Bureau, Current Population Survey, 1981 to 2024 Annual Social and Economic Supplements (CPS ASEC), Table H-9, Type of Household-All Households by Median and Mean Income: 1980 to 2024, available at: https://www.census.gov/data/tables/time-series/demo/income-poverty/historical-income-households.html (accessed: May 7, 2026).

\394\ Parekh, N., & Campau T., Average Age of Vehicles Hits New Record in 2024, last revised: May 22, 2024, available at: https://www.spglobal.com/mobility/en/research-analysis/average-age-vehicles-united-states-2024.html (accessed: June 2, 2026).

By reducing the regulatory costs of complying with fuel economy standards, the final rule would lead to an increase in new vehicle sales relative to the No-Action Alternative. For purposes of regulatory evaluation, the relevant metric is the difference in the number of new vehicles sold between the baseline and each alternative rather than the absolute number of sales under any alternative. The agency's analysis of the response of new vehicle sales to different stringencies of fuel economy standards includes three components: a forecast of sales based exclusively on macroeconomic factors, which is used to determine the sales quantity for the No-Action Alternative; the assumed price elasticity of new vehicle demand, which interacts with estimated price increases under each alternative to create differences in sales relative to the No-Action Alternative; and a fleet share model that projects differences in the passenger and non-passenger automobile fleet shares under each alternative.

The first component of the sales response model is the nominal total new vehicle sales forecast, which is based on a small set of macroeconomic inputs that together determine the size of the new vehicle market in each future year under the baseline alternative. This statistical model is intended to provide only an initial forecast of light-duty vehicle sales; it does not incorporate the effect of prices on sales and is not intended to be used for analysis of the response to price changes in the new vehicle market. NHTSA's projection varies in the early model years before leveling off at around 15 million vehicles per year in the 2030s. This result is consistent with the continued response to sales volatility in the years following the coronavirus (COVID-19) public health emergency and the supply chain challenges immediately thereafter, as well as consumer sentiment in the near-term. NHTSA acknowledges that excluding the regulatory costs to comply with the baseline standards has the potential to underestimate the effect of prevailing conditions on vehicle sales; however, given that the macroeconomic assumptions used in the analysis take into account the effects of various regulatory policies and the fact that the relevant metric is the differences created by alternative CAFE stringencies, the agency has determined this approach is appropriate.

The agency's baseline sales forecast assumes that total new vehicle sales are driven primarily by conditions in the U.S. economy that are outside the influence of the automobile industry. Over time, new vehicle sales have been cyclical--rising when prevailing economic conditions are positive (periods of growth) and falling during periods of economic contraction. While changes to vehicles' designs and prices that occur as consequences of manufacturers' compliance with earlier standards (and with regulations on vehicles' features other than fuel economy) exert some influence on the volume of new vehicle sales, they are far less influential than macroeconomic conditions. The effects of compliance are not large enough to reverse broader cyclical trends in sales; instead, they produce marginal differences in sales among regulatory alternatives that the agency's sales module is designed to simulate. Increases in new vehicle prices caused by higher regulatory costs reduce sales below the cyclical trend, and slow fleet turnover, while decreases in prices have the opposite effect.

NHTSA is prohibited by statute from considering the fuel economy of dedicated automobiles (e.g., battery electric or hydrogen vehicles) and therefore has removed dedicated automobiles from the sales forecast it uses to analyze the final rule. NHTSA uses market penetration rates from the AEO 2026 Alternative Transportation and Electricity case to estimate the market share of the gasoline-powered fleet. The agency then applies this market share to the total light-duty forecast produced by the nominal forecast.\395\ An independent projection like the AEO 2026 Alternative Transportation and Electricity case is a reliable estimate of the future market share for gasoline-powered vehicles.

\395\ NHTSA also considers other approaches, such as assuming the full fleet in future model years would be composed of gasoline- powered vehicles or holding the current market penetration rate for dedicate automobiles constant. Final TSD Chapter 4.2.1.2 provides more discussion of the selected approach and alternatives considered.

The second component of the sales response model captures how price changes affect the number of vehicles sold. NHTSA estimates the change in sales from its initial forecast during future years under each regulatory alternative by applying an assumed price elasticity of new vehicle demand to the percent difference in average price between the regulatory alternatives and the No-Action Alternative. This price change does not represent an increase or decrease from the previous model year, but rather the percent difference in the average price of new vehicles between the baseline and each regulatory alternative for that model year. The average new vehicle price in the baseline is defined as the observed price in 2024 (the last historical data year before the simulation is run) plus the average regulatory cost associated with the No-Action Alternative for each future model year.\396\ The agency also subtracts any tax credits for which a PHEV may qualify from those regulatory costs to simulate sales.\397\

\396\ The CAFE Model currently operates as if all costs incurred by the manufacturer as a consequence of meeting regulatory requirements, whether those costs are the cost of additional technology applied to vehicles in order to improve fleetwide fuel economy or civil penalties paid when fleets fail to achieve their standard, are “passed through” to buyers of new vehicles in the form of price increases.

\397\ For additional details about how NHTSA models tax credits, see Section II.C.2.e above.

Within the CAFE Model's logic, there is an assumption that new vehicle models within the same regulatory class (e.g., passenger automobiles) are close substitutes for one another, including vehicles with differing powertrains.\398\ NHTSA recognizes that different vehicle attributes may alter the perceived value of vehicles. NHTSA implements several modeling constraints to prevent the CAFE Model from considering technologies for fuel economy that could adversely affect the utility of vehicles, such as maintaining performance neutrality, including phase-in caps, and defining technology pathways by using engineering judgement. The agency acknowledges that, even with these constraints, it is possible that CAFE standards may influence attributes other than price or fuel economy that are unaccounted for in the agency's sales analysis.

\398\ The CAFE Model does not assign different preferences between technologies, and outside the standard-setting restrictions, the Model will apply technology on a cost-effectiveness basis. Similarly, outside of the sales response to changes in regulatory costs, consumers are assumed to be indifferent to specific technology pathways and will demand the same vehicles despite any changes in technological composition.

NHTSA has previously invested considerable resources in developing a discrete choice model of the new automobile market that would (1) enable the agency to incorporate the effect of additional vehicle attributes on buyers' choices among competing models; (2)

reflect consumers' differing preferences for specific vehicle attributes; and (3) provide the capability to simulate responses, such as strategic pricing strategies by manufacturers intended to alter the mix of models they sell and enable them to comply with new CAFE standards. However, those efforts have not yet produced a satisfactory and operational model.\399\ Instead, NHTSA accounts for the possibility of decreased utility of vehicles because of CAFE standards outside of the sales module.

\399\ NHTSA's experience partly reflects the fact that these models are highly sensitive to their data inputs and estimation procedures, and even versions that fit well when calibrated to data from a single period--usually a cross section of vehicles and shoppers or actual buyers--often produce unreliable forecasts for future periods, which NHTSA's regulatory analyses invariably require. This occurs because they are often unresponsive to shifts in economic conditions or consumer preferences, and also because it is difficult to incorporate factors such as the introduction of new model offerings--particularly those utilizing advances in technology or vehicle design--or shifts in manufacturers' pricing strategies into their representations of choices and forecasts of future sales or market shares. For these reasons, most vehicle choice models have been better suited for analysis of the determinants of historical variation in sales patterns than for forecasting future sales, volumes and market shares of particular categories.

Because the price elasticity that is applied in the CAFE Model assumes no perceived change in the quality of the product, and the vehicles produced under different regulatory scenarios have inherently different operating costs, the price metric must account for this difference. The price change to which the elasticity is applied in the analysis represents the residual price difference between the baseline and each regulatory alternative after deducting the value of fuel savings over the first three years of each model year's lifetime.

The price elasticity is also specified as an input, and for the proposal, the agency assumed a value of -0.4, meaning that a 5-percent increase in the average price of a new vehicle produces a 2-percent decrease in total sales. NHTSA has used this same elasticity in prior rulemakings. NHTSA stated in the proposal that estimates of this parameter reported in published literature vary widely,\400\ and NHTSA believes that its choice is a reasonable one within this range, but NHTSA also presented sensitivity cases that explore higher and lower elasticities in the PRIA.

\400\ See Jacobsen, M. et al., The Effects of New-Vehicle Price Changes on New- and Used-Vehicle Markets and Scrappage, EPA-420-R- 21-019, EPA: Washington, DC (2021), available at: https://cfpub.epa.gov/si/si_public_record_Report.cfm?Lab=OTAQ&dirEntryId=352754 (accessed: June 2, 2026) (reporting a range of estimates, with a value of approximately -0.4 representing an upper bound of this range). NHTSA selects this point estimate for the central case and explores alternative values in the sensitivity analysis.

The agency sought comment on this sales elasticity assumption-- including whether NHTSA should consider applying separate short-run and long-run elasticity assumptions in the analysis. NADA supported NHTSA's use of -0.4 as its price elasticity since it is in the range supported by recent literature.\401\ Natural Resources Defense Council et al. (NRDC et al.) suggested that NHTSA revise its approach by using a lower elasticity value.\402\ This suggestion was based on a selection of estimates made since the year 2010, and NRDC et al. argued that the lower value better captures the long run dynamics of the new vehicle market in conjunction with scrappage. In support of its position, NRDC et al. included a literature review of papers from 1952 to 2020. The mean of all the studies was -0.6 and -1.15 for the long-run and short- term estimates, respectively. Narrowing the scope to only studies since 2000, NRDC et al. found that the median of all estimates for the elasticity falls around -0.35 to -0.4. NRDC et al. claimed that even further censoring its dataset to just estimates produced since 2010 lowers the average more and pointed to two recent papers and an amicus brief filed in litigation over the 2020 final rule. Finally, NRDC et al. argued that NHTSA's choice of elasticity was inconsistent with the large welfare losses it ascribes to sacrifices in other features for vehicles. IPI also argued that NHTSA should adjust sales using a lower magnitude elasticity due to interactions with the used vehicle market and potential substitution by car owners to outside options like public transit and ride shares.\403\

\401\ NADA, Docket No. NHTSA-2025-0490-0036-A1, at 9-10.

\402\ NRDC et al., Docket No. NHTSA-2025-0491-5928-A2, at 98- 102.

\403\ IPI, Docket No. NHTSA-2025-0491-6015-A2, at 62-64 and 70- 72.

NHTSA's review of the literature NRDC et al. submitted found that one of these papers (Leard (2021) \404\) estimated a long-term elasticity of -0.35, in line with results from the larger set of estimates, and very close to NHTSA's estimate. The amicus brief suggests a range between -0.03 and -0.61, which puts NHTSA's preferred estimate firmly in the middle. NHTSA also notes that Jacobsen et al. (2021) conducted a thorough review of the literature around this parameter, and chose a parameter estimate of -0.4 as the lowest magnitude value to use in simulations.\405\ NHTSA also chose -0.4, which, as demonstrated by Jacobsen (2021) and literature cited in comments is at the relative low-end of estimates, to represent a long- run “policy elasticity” meant to capture the long-run dynamics outlined by IPI. IPI argued that NHTSA should consider outside options like public transit and carpooling. NHTSA believes that these options also carry additional costs and the availability of these options varies across the country. Modeling the impact of these options directly, beyond the long-run impact of fuel prices on light-duty vehicle sales, is beyond the scope of the CAFE analysis and it is unclear whether broadening the analysis' scope would provide materials improvement.

\404\ Leard, B., Estimating Consumer Substitution Between New and Used Passenger Vehicles, Working Paper 19-01 (revised Aug. 2021), Resources for the Future: Washington, DC (2021), available at: https://media.rff.org/documents/WP_19-01_rev_2021.pdf (accessed: May 28, 2026).

\405\ See Jacobsen, M. et al., The Effects of New-Vehicle Price Changes on New- and Used-Vehicle Markets and Scrappage, EPA-420-R- 21-019, EPA: Washington, DC, (2021), available at: https://cfpub.epa.gov/si/si_public_record_Report.cfm?Lab=OTAQ&dirEntryId=352754 (accessed: June 2, 2026).

In response to NRDC et al.'s comment that NHTSA's choice of elasticity was inconsistent with the large welfare losses it ascribes to sacrifices in other features for vehicles, NHTSA's choice of a three-year payback period operates under the assumption that after accounting for the value of sacrificed attributes, a new vehicle's fuel-saving technology must payback within its first three years of operation in order to be cost effective. This assumption is incorporated in the sales model, meaning that the agency's sales model is consistent with its accounting for the value of other vehicle features.

After consideration of comments and the relevant literature, NHTSA is continuing to use -0.4 as its sales elasticity parameter in this final rule. NHTSA presents sensitivity cases that explore the impact of higher and lower elasticities on the analysis in FRIA Chapter 9. Chapter 4.2.1.2 of the Final TSD further presents evidence that NHTSA believes supports its decision.

The third and final component of the sales model is the dynamic fleet share module (DFS). The analysis uses the DFS developed during the previous rulemaking. The baseline fleet share projection is derived from the agency's own compliance data for the 2024 fleet and from the 2026 AEO projections for subsequent model years. These shares are applied to the total industry sales derived in the first stage of the total sales model to estimate sales volumes of car and light truck body styles. NHTSA determines individual model sales using the following sequence: (1) individual

manufacturer shares of each regulatory class (either passenger cars or light trucks) are multiplied by total industry sales of vehicles in that regulatory class and then (2) each vehicle within a manufacturer's volume of that regulatory class is assigned the same percentage share of that manufacturer's sales as in MY 2024. This assumes that consumer preferences for particular styles of vehicles are determined in the aggregate (i.e., at the industry level), but that manufacturers' sales shares of those body styles are consistent with their MY 2024 sales. Within a given regulatory class, NHTSA assumes a manufacturer's sales shares of individual models are also constant over time.

This approach also assumes implicitly that manufacturers are pricing individual vehicle models within market segments in a way that maximizes their profit. Without more information about each manufacturer's true cost of production, including its fixed and variable components and its target profit margins for its individual vehicle models, there is no basis to assume that strategic shifts within a manufacturer's portfolio will occur in response to standards.

Similar to the second component of the sales module, the DFS applies an elasticity to the change in price between each regulatory alternative and the No-Action Alternative to determine the change in fleet share from its baseline value. NHTSA uses the net regulatory cost differential (costs minus fuel savings) in a logistic model to capture the changes in fleet share between passenger cars and light trucks, with a relative price coefficient of -0.000042. NHTSA selected this methodology and price coefficient based on a review of academic literature.\406\ When the total regulatory costs for passenger automobiles of meeting standards minus the value of the resulting fuel savings exceed that of non-passenger automobiles, the market share of non-passenger automobiles will rise relative to passenger automobiles. For example, a $100 net regulatory cost increase in passenger automobiles relative to light trucks would produce around a 0.1 percent shift in market share towards light trucks, assuming in the example that the latter initially represents 60 percent of the fleet.

\406\ NHTSA describes this literature review and the calibrated logit model in more detail in the accompanying docket memo “Calibrated Estimates for Projecting Light-Duty Fleet Share in the CAFE Model.”

As discussed in preamble Section VI, the agency is modifying its regulatory definitions for vehicle classification starting with MY 2030. The agency takes account of this reclassification after it simulates the aggregate sales and DFS responses to changes in vehicle prices. NHTSA assigns vehicles both an “initial” classification based on how they are classified under the current regulations and a “revised” classification for how they would be classified under the final regulations. The aggregate sales response is calculated at the fleetwide level, so regulatory classification affects changes in sales only insofar as a reclassified vehicle model incurs a different regulatory cost to comply with the requirements of its new regulatory class. For the DFS model, the regulatory costs are borne by a vehicle's “initial” classification, so an SUV that is reclassified from the light truck fleet to the passenger car fleet has its regulatory costs for the DFS analysis attributed to the light truck fleet throughout the analysis. This method assumes that each individual model's sales shares within the “initial” regulatory class remain constant. This may cause the counterintuitive effect of an increase in a vehicle's price, leading to an increase in that vehicle's sales. NHTSA considered applying its existing model to sales shares determined by the “revised” classification but decided against this due to the cross- elasticities used in the analysis being estimated based on the current classification system.

NHTSA sought comment on this approach and whether it is appropriate to apply the DFS's price coefficient to the “revised” regulatory classes, and if there is an alternative elasticity or methodology the agency could employ in its analysis. NHTSA did not receive any comments on this issue. On a related matter, AEG agreed with NHTSA that a discrete choice model would best capture the tradeoffs present in the new vehicle market and suggested using a sequential decision problem model setup.\407\ AEG argued that this framework would allow NHTSA to analyze substitution at a more granular and realistic level, within specific segments of the new vehicle market, rather than between regulatory classes. Likewise, IPI commented that sales should be projected by make of vehicle or even by model, allowing for compliance cost differences between manufacturers and vehicle models to influence market shares.\408\

\407\ AEG, Docket No. NHTSA-2025-0491-5981-A1, at 2-5.

\408\ IPI, Docket No. NHTSA-2025-0491-6015-A2, at 70-72.

In response to AEG's comments, NHTSA notes that it is continuing to explore options for improving the modeling of consumer choice for future rules. Regarding IPI's comment, to model sales accurately at the make or even model level would require modeling pricing strategies for manufacturers, which would involve capturing firm-level priorities and costs outside of the fuel economy technology adoption decision and goes well beyond the scope of this analysis. As a result, NHTSA has not changed its approach for this final rule. b. Scrappage

For potential car buyers, new and used vehicles act as alternatives, or substitutes, for each other within broad limits. A consumer surveying the market for a new vehicle may find the combination of price and features undesirable and elect to continue to use an existing vehicle or purchase a used vehicle that better meets the consumer's price and feature preferences. When the price of a good increases, so does the demand for its substitutes, causing the equilibrium price and quantity of substitutes supplied to rise. Because the final rule would lower the price of new vehicles, demand for used vehicles would decrease, causing the equilibrium market price for used vehicles to decrease and simultaneously increasing the rate at which used vehicles are retired. Because used vehicles are in existence, their supply only can be increased by keeping more of those that would otherwise be retired in use longer, which corresponds to a reduction in their scrappage or retirement rates. As older vehicles are used longer, the average age of the fleet rises and the safety risk to all road users likewise increases, because older vehicles are less safe than newer ones.

When new vehicles become more expensive, demand for used vehicles increases. Because used vehicles are more valuable in such circumstances, they are scrapped at a lower rate, and just as rising new vehicle prices push some prospective buyers into the used vehicle market, rising prices for used vehicles force some prospective buyers to acquire even older vehicles or models with fewer desired attributes. The effect of fuel economy standards on scrappage is partially dependent on how consumers value future fuel savings; NHTSA assumes that, after accounting for sacrifices in other vehicle attributes, consumers value only the first 36 months of fuel savings when making a purchasing decision.

Many competing factors influence the decision to scrap a vehicle, including the cost to maintain and operate it, the household's demand for VMT, the cost

of alternative means of transportation, and the value that can be attained through reselling or scrapping the vehicle for parts. In theory, a car owner will decide to scrap a vehicle when the value of the vehicle minus the cost to insure, register, maintain, and repair the vehicle is less than its value as scrap material; in other words, when the owner realizes more value from scrapping the vehicle than from continuing to drive it or from selling it. Typically, the owner that scraps the vehicle is not the original owner.

While scrappage decisions are made at the household level, NHTSA is unaware of sufficiently detailed household data to capture scrappage at that level. Instead, NHTSA uses aggregate data measures that capture broader market trends. In addition, the aggregate results are consistent with the rest of the CAFE Model, as the Model does not attempt to project manufacturers' pricing strategies; the Model assumes instead that all regulatory costs to make a particular vehicle compliant are passed on to the purchaser who buys the vehicle.

The dominant source of scrappage is “engineering scrappage,” which is largely determined by the age of a vehicle and the durability of the specific model year or vintage it represents. NHTSA uses proprietary vehicle registration data from S&P Global to estimate vehicle age and durability. Other factors affecting decisions to retire used vehicles or retain them in service include fuel economy and new vehicle prices; for historical data on new vehicle transaction prices, NHTSA uses NADA data.\409\ The data consists of the average transaction price of all light-duty vehicles; because the transaction prices are not broken down by body style, the scrappage module may miss unique trends within a particular vehicle body style. The transaction prices reflect the amount consumers paid for new vehicles and exclude any trade-in value credited towards the purchase. This may be relevant particularly for pickup trucks, which have experienced considerable changes in average price as luxury and high-end options entered the market over the past decade.

\409\ The data can be obtained from NADA. For reference, the data for MY 2024 may be found at https://www.nada.org/nada/research-data/nada-data (accessed: May 28, 2026).

Vehicle survival rates, which are determined over time by scrappage, follow a roughly logistic function with age--that is, when a vintage is young, few vehicles in the cohort are scrapped; as they age, more and more of the cohort are retired each year, and the annual rate at which vehicles are scrapped reaches a peak. Scrappage then declines as vehicles enter their later years, as fewer and fewer vehicles in the cohort remain on the road. The analysis uses a logistic function to capture this trend of vehicle scrappage with age. The data shows that the durability of successive model years generally increases over time; put another way, historically, newer vehicles last longer than older vintages. However, this trend is not constant across all vehicle ages-- the instantaneous scrappage rate of vehicles is lower generally for more recent vintages up to a certain age, but must increase thereafter so that the final share of vehicles remaining converges to a similar share remaining for historically observed vintages.\410\ NHTSA's scrappage model uses fixed effects to capture potential changes in durability across model years and ensures that vehicles approaching the ends of their lives are scrapped in the analysis.

\410\ Some possible reasons for why durability may have changed are new automakers entering the market or general changes to manufacturing practices like switching some models from a car chassis to a truck chassis.

The final source of vehicle scrappage is from cyclical effects, which the CAFE Model captures using forecasts of GDP and fuel prices. The macroeconomic conditions variables discussed above are included in the logistic model to capture cyclical effects. Finally, the change in new vehicle prices projected in the Model (technology costs minus 36 months of fuel savings and any tax credits passed through to the consumer) is included, and changes in this variable are the source of differing scrappage rates among regulatory alternatives.

NHTSA sought comment on the suitability of its scrappage modeling approach in the proposal. NADA and other commenters supported NHTSA's scrappage modeling and agreed with its finding that policies like CAFE standards cause some vehicle owners to forgo retiring used vehicles when new vehicle prices increase. NADA pointed to the rise of the average age of vehicles in the on-road fleet in recent years and the greater safety risks that older vehicles present.\411\ On the other hand, Consumer Reports supported NHTSA's modeling of scrappage effects but argued that the length of its payback period unduly influenced its results. NRDC et al. disagreed with NHTSA's assumption that manufacturers will pass technology costs onto consumers and suggested that this would cause NHTSA to overestimate the effect of high standards on scrappage rates.\412\ NRDC et al. also argued that, in the absence of higher standards, manufacturers may choose to apply technology to improve other vehicle features and will not lower prices in response.

\411\ NADA, Docket No. NHTSA-2025-0490-0036-A1, at 11-12; PMI, Docket No. NHTSA-2025-0491-5001-A2, at 52-53; AmFree and Corn Growers Associations, Docket No. NHTSA-2025-0491-6000-A1, at 8-9.

\412\ NRDC et al., Docket No. NHTSA-2025-0491-5928-A2, at 70-73.

Regarding Consumer Reports' comment about the interaction between scrappage effects and payback period, NHTSA continues to believe that 36 months is the most appropriate payback period for its new vehicle sales analysis as discussed in Section II.E.1. Because the payback period included in the scrappage model is meant to capture the effects of changes in new vehicle sales on scrappage, NHTSA believes that the payback period for the scrappage model should be the same as that used for the sales module.

In response to NRDC et al.'s comment, NHTSA does not model new vehicle pricing strategies because it is beyond the scope of NHTSA's analysis. NHTSA is unaware of any empirical basis to model new vehicle pricing strategies more comprehensively and notes that the data necessary to do so is almost certain to include proprietary business information. If a manufacturer's profit maximizing strategy involves cross-subsidizing vehicles with new technologies by raising the price of other vehicles, the analysis may indeed overestimate the predicted price of some individual new vehicle models while underestimating the price of others. However, NHTSA's analysis operates at higher, more aggregated levels, and its assumptions at this level are reasonable. If manufacturers do not pass on additional costs of technology to consumers at the fleet level, over the long term, they will go out of business. In the case of improvements to other vehicle features, NHTSA anticipates that manufacturers would add these features by choice because they are valued by consumers and would increase demand for new vehicles. As outlined in Final TSD Chapter 4.2, these types of changes in the new vehicle market would tend to increase scrappage rates. Therefore, NHTSA believes that its use of pass-through assumptions at the aggregate level used in the scrappage model is reasonable.

Finally, some commenters took issue with NHTSA's decision to continue using the same dataset for estimating its scrappage model in this rulemaking, and to continue using a reduced-form

model.\413\ They argued that a reduced-form model, which estimates the statistical relationship between scrappage rates and other economic variables, rather than a structural model which formally relates the variables through a theoretical economic model, was inappropriate for predicting scrappage rates out into the future. NHTSA considered updating its dataset, which includes observations for all vehicles present in the on-road fleet for CYs 1975-2017.\414\ However, new and used vehicle markets faced significant disruptions due to supply chain interruptions from 2020-2022, making these datapoints somewhat atypical outliers. Scrappage rates require two years of data because they represent a year-to-year change in the size of the on-road fleet, meaning that only one additional year of observations after 2020 would be available prior to the start of the rulemaking period. Because the dataset NHTSA uses covers an extensive time series, adding even several new model years of data would only contribute a small share of additional information to the dataset used to develop the estimates. The commenters also do not point to any reasons or evidence for the assertion that not using more recent data biases the results. Therefore, NHTSA continues to use the same dataset to estimate its scrappage model results for the final rule. However, the agency did update its estimates by adjusting variables denominated in dollars to constant 2024$ values and adjusting the calculation of the explanatory variable representing price net of payback period fuel savings. In prior rulemakings, this variable used a value of fuel savings equal to the value accumulated over the first 2.5 years of use. NHTSA updated its payback period for the proposal to the first 3 years, which is equivalent to the first 48,000 miles of use. Because this variable is meant to capture the effects of the new vehicle market on the used market, and thus scrappage, the CAFE Model projects it forward using the same quantity of miles. Accordingly, the coefficient used in the scrappage model must be estimated using this same calculation in order for the model to be consistent. This means that, in both the proposal and this final rule, coefficients differ from those used in the 2024 final rule, which was noted by commenters.\415\ The adjustments are discussed in Chapter 4.2 of both the Draft and Final TSD.

\413\ IPI, Docket No. NHTSA-2025-0491-6015-A2, at 64-66.

\414\ The analysis begins in 1975 as this is the earliest year all required input data were available.

\415\ Attorneys General, Docket No. NHTSA-2025-0491-6064-A4, at 54-56.

In addition to the variables included in the scrappage module, NHTSA considers several other potential variables that likely either directly or indirectly influence scrappage in the real world, including maintenance and repair costs, the value of scrapped metal, vehicle characteristics, the quantity of new vehicles purchased, higher interest rates, and unemployment. These variables are excluded from the scrappage module either because of difficulties in obtaining data to measure them accurately or other modeling constraints. Their exclusion from the module is not intended to diminish their importance but rather highlights the practical constraints of modeling intricate decisions like scrappage.

NHTSA sought comments on whether it should include any of these variables and, if so, requested specific methodologies that would produce robust and unbiased estimates that could be used in a regulatory analysis setting. Some commenters argued that NHTSA should use a single model to determine both new vehicle sales and scrappage rates jointly.\416\

\416\ IPI, Docket No. NHTSA-2025-0491-6015-A2, at 62-64; Attorneys General, Docket No. NHTSA-2025-0491-6064-A2, at 95-96.

While NHTSA uses separate modules to predict sales and scrappage, the two are connected through the inclusion of new vehicle prices, net of payback period fuel savings, in each. Modeling sales and scrappage jointly would require developing a dynamic model for scrappage that takes account of consumer expectations for the cost of continuing to own and operate used vehicles as well as expectations about the future cost of owning and operating new vehicles in future years. This would add significant modeling complexity and require additional data not currently used in the CAFE Model, including maintenance costs by age, body style, and powertrain for the entire on-road fleet, which is not possible within the rulemaking timeframe. This would also require more granular data for existing used vehicles, some of which is not available and thus would require significant assumptions that could introduce error to the model. NHTSA includes a set of sensitivity cases that vary parameters in the sales and scrappage models in Chapter 9 of the FRIA which addresses many of the alternative theories posited by commenters. The agency will consider development of a dynamic scrappage model for future rulemaking, but the existing approach provides a reasonable means by which the agency can evaluate the relationship between changes in new vehicle prices and scrappage rates.

Resetting the CAFE standards is expected to accelerate the retirement of older vehicles. Because the final standards reduces the regulatory burden on manufacturers and by extension the price of new vehicles, the demand and price for used vehicles should decrease, and incentivize households to replace the older vehicles that are costly to maintain with newer, cheaper options--including newer used vehicles. 3. Changes in Vehicle Miles Traveled

As described in the fleet turnover section, fuel economy standards influence the quantity of new vehicles sold and how quickly older vehicles are retired. Model years of different vintages possess distinguishable characteristics, with newer vehicles typically being more fuel efficient and safer than their older contingents. While the decision itself to buy a new vehicle or retire an older vehicle may confer certain costs and benefits to their owners, most of the effects are realized only through the use of those vehicles. The lower standards of the final rule will accelerate fleet turnover compared to the baseline, which results in more miles being driven in newer, safer vehicles compared to older, less safe vehicles. The agency anticipates that fewer miles will be driven in the oldest and least safe vehicles on the road, and the number of fatal accidents and serious injuries from highway crashes will decrease as result of the final rule.

Deciphering which vehicles are being driven is just as important as how many miles are being driven. Any shift in miles driven by older vehicles to newer vehicles creates a corresponding shift in societal benefits. To capture how CAFE standards influence the distribution of miles across the fleet, NHTSA estimates VMT based on the average use of vehicles at different ages, the total number of vehicles in use, and the composition of the fleet by ages. These three components--average vehicle usage, new vehicle sales, and older vehicle scrappage--jointly determine total VMT projections for each alternative.

VMT is determined by how much households want to drive and how much they can afford to do so. NHTSA believes that a significant portion of light-duty VMT is unaffected by fuel economy standards. Households have some basic level of travel demand that needs to be met such as driving to work or school, and those households will drive those miles regardless of the

imposition that fuel economy standards may impose. NHTSA's perspective is that the total demand for VMT should not vary excessively across alternatives. To prevent large differences from arising among the regulatory alternatives, the agency constrains the aggregate amount of VMT--besides VMT attributable to the “rebound effect”--across alternatives to be equal with the No-Action Alternative.

In prior rules, the agency used the Federal Highway Administration (FHWA) VMT Forecasting Model to project total VMT in future calendar years and then adjusted alternatives based on fleet composition. NHTSA employed this methodology because it used a reliable, external projection of annual VMT as a starting point. However, because the FHWA model includes miles that will be driven in dedicated automobiles, NHTSA amended its approach for the proposed rule, which the agency has elected to retain for the final rule.

The No-Action Alternative's projection of VMT uses the simulated projections of the gas-powered fleet produced by the sales and scrappage models and applies it to estimates of VMT per vehicle. Vehicles of different ages and body styles have different costs to own and operate, and usage changes across vehicle ages independent of CAFE standards. To account properly for the average value of consumer and societal costs and benefits associated with vehicle usage under various alternatives, it is necessary to partition miles by age and body type. Using S&P Global odometer data, NHTSA creates “mileage accumulation schedules” as an initial estimate of how much a vehicle is expected to drive at each age throughout its life. The mileage accumulation schedules also account for differences in driving habits based on body style. Multiplying the numbers of each vehicle projected to be in the fleet by the per-vehicle VMT estimates from the mileage accumulation schedules creates a forecast of VMT in each calendar year.

The methodology to allocate miles within the regulatory alternatives is similar. NHTSA uses the forecasts of the fleet produced by the sales and scrappage models and multiplies those by mileage accumulation schedules to create a total estimate of VMT. NHTSA then scales the alternative's VMT to match the No-Action Alternative's aggregate VMT, preserving the percentage of VMT driven by each model.

NHTSA sought comments on whether it should remove the VMT constraint and allow alternatives to have differing levels of VMT. NHTSA stated in the NPRM that, while most household VMT is likely inelastic, it may be reasonable to assume that fleets with differing sizes, age distributions, and inherent cost of operation may have marginally different annual VMT (even without considering VMT associated with rebound miles). In previous rules, NHTSA elected to continue to constrain VMT across alternatives in part because of the difficulty of determining whether VMT would shift to other modes of transportation and, if so, how to account for the impacts of any such mode shift. NHTSA sought comments on whether it is appropriate to consider mode shifts if the agency removes the VMT constraints and requested data or suggested modeling approaches that could assist the agency.

IPI commented that changes in the fleet composition and size should be accompanied by changes in non-rebound VMT.\417\ UCS by contrast agreed with NHTSA's assessment that household VMT is largely inelastic, and that changes in travel demand are captured by the modeling of rebound VMT.\418\

\417\ IPI, Docket No. NHTSA-2025-0491-6015-A2, at 72-76.

\418\ UCS, Docket No. NHTSA-2025-0491-6027, at Technical Appendix 58.

NHTSA agrees with UCS and is continuing to model a non-rebound VMT constraint in this final rule. While some households may choose not to own a vehicle, or might dramatically change their daily travel habits when they buy or sell one of their vehicles, it is more likely that most of the changes in fleet size caused by increased CAFE standards will result in compensating changes in the usage of other vehicles, primarily older vehicles being driven more. Even if CAFE standards do cause some changes in VMT, attempting to model the exact nature of this substitution requires a complex model of household travel demand, and alternative modes of travel would require capturing regional and local variation in household driving patterns, and the availability of alternatives like public transportation and carpooling. Modeling this type of decision-making adds little additional precision to NHTSA's incremental results and could instead introduce additional sources of error. Doing so would thus significantly expand the model beyond its current scope, for little additional gain in estimating the variables of interest for NHTSA's CAFE analysis.

The Attorneys General criticized NHTSA's approach to modeling non- rebound VMT in the standard-setting analysis.\419\ These commenters argued that, by not including BEVs and other dedicated AFVs, NHTSA failed to consider the costs and benefits related to changes in VMT driven by these vehicles. These commenters also pointed to inconsistencies between NHTSA's standard-setting projection for non- rebound VMT and other projections, including its EIS projection and projections produced by EIA in its 2025 AEO. Commenters pointed to periods where NHTSA's gas-powered fleet's projected non-rebound VMT exceeded the EIS projection, and the faster rate of decline in the gas- powered fleet's share of non-rebound VMT compared to the AEO. These commenters argued that these factors showed that NHTSA's projection model was not suitable for determining non-rebound VMT, and that it had material effects on the costs and benefits estimated in NHTSA's analysis.

\419\ Attorneys General, Docket No. NHTSA-2025-0491-6064, at 21- 26.

NHTSA did not include these vehicles in its projection because doing so would violate the statutory restrictions that preclude consideration of such vehicles when setting standards. While NHTSA's projection of VMT differs from other projections given these statutory constraints, this does not make it unsuitable for the task for which it is used, namely producing a baseline forecast that is consistent with the projected fluctuations in the size and composition of the gas- powered on-road fleet NHTSA models in its standard-setting analysis. Per-vehicle VMT is determined using NHTSA's VMT accumulation schedule, which is documented in Final TSD Chapter 4.3, and represents a reasonable expectation for vehicle usage over the long term. The size and composition of the on-road fleet are consistent with the development of the fleet as predicted by NHTSA's sales and scrappage models. Finally, the purpose of this projection is to provide a baseline expected level of VMT that can be fixed across alternatives and used as a basis for determining rebound VMT. NHTSA's methodology ensures that the per-vehicle usage that rebound elasticities are applied to remains at a reasonable level across alternatives. Using an outside forecast not generated consistently with the projected fleet risks introducing a source of error into the analysis.

NHTSA analyzed sensitivity cases that used alternative sources for its non-rebound VMT constraint, including projections of gas powered VMT from

the 2026 AEO and the EIS. Results are discussed in PRIA Chapter 9.

A portion of household travel that is elastic is known as “rebound” mileage. The fuel economy rebound effect--a specific example of the well-documented energy efficiency rebound effect for energy-consuming capital goods--refers to motorists who choose to increase vehicle use (as measured by VMT) when fuel economy is improved and, as a result, the cost per mile (CPM) of driving declines. If fuel economy increases, the cost to drive additional miles decreases, causing vehicles with better fuel efficiency to be driven more. For the final rule, reducing the level of fuel economy required by government regulation would have the opposite effect, reducing the number of miles driven.

NHTSA has employed several different estimates of the rebound effect through the years. Until recently, the agency had historically used an estimate between 15 and 20 percent. The agency lowered its estimate in the 2022 final rule to 10 percent, a value that was also used in the 2024 final rule. For this rulemaking, NHTSA re-reviewed the literature related to the fuel economy rebound effect, which is extensive and covers multiple decades and geographic regions.\420\ The totality of evidence, without excluding certain studies based on arbitrary selection criteria, suggests that the plausible range for the rebound effect is quite wide, extending from 10 percent to perhaps as high as 40 percent. This range implies that, for example, a 10-percent reduction in vehicles' fuel CPM would lead to an increase of between 1 to 4 percent in the number of miles they are driven annually. The central tendency of this range appears to be at or slightly above its mid-point, which is 25 percent. Considering only those studies that NHTSA believes utilize robust and reliable data, employ identification strategies that are likely to prove effective at isolating the rebound effect, and apply rigorous estimation methods, suggests a range of approximately 10-35 percent, with most of the estimates falling in the 15-30 percent range.

\420\ Final TSD Chapter 4.3.4 provides more information.

When NHTSA reviewed the literature for both the 2022 and 2024 rules, the agency arrived at a similar conclusion but chose to use an estimate at the lowest end supportable by published research. NHTSA argued that both economic theory and empirical evidence suggested that the rebound effect was declining over time in response to factors such as increasing income (which increases the value of travelers' time), progressively smaller reductions in fuel costs from continuing increases in fuel economy, and slower growth in car ownership and the number of license holders. The agency also noted that some lower estimates of the rebound effect were associated with recently published studies that rely on U.S. data, measure vehicle use using actual odometer readings, control for the potential endogeneity of fuel economy, and--critically--estimate the response of vehicle use to variation in fuel economy itself rather than to fuel cost per distance driven or fuel prices. The agency gave greater weight to these studies, which suggested a rebound effect in the 5 to 15 percent range.

Consistent with NHTSA's surveys of the latest available data for each successive CAFE analysis, as discussed above, the agency reconsidered for this analysis its prior assumptions about the rebound effect discussed in the 2022 and 2024 final rules--in particular its assumption that the rebound effect is declining over time--and concluded that a rebound estimate of 15 percent is more appropriate. In particular, a meta-analysis of 74 recently published studies of the rebound effect noted that “the magnitude of the rebound effect in road transport can be considered to be, on average, in the area of 20 [percent],” and that the most likely long-run estimate was about 32 percent \421\--both significantly higher than the agency's prior 10 percent value, and also well above the 15 percent value employed in this analysis. The agency also believes that selecting a rebound estimate that is well-supported by the scientific consensus is more appropriate than speculating about potential future trends. NHTSA examines the sensitivity of estimated impacts to values of the rebound effect ranging from 10 to 20 percent to account for the uncertainty surrounding its exact value. For a more complete discussion of the rebound literature, refer to Final TSD Chapter 4.3.4.

\421\ Dimitropoulos, A. et al., The Rebound Effect in Road Transport: A Meta-analysis of Empirical Studies, OECD Environment Working Papers, No. 113, OECD Publishing: Paris, France (2016), available at: https://dx.doi.org/10.1787/8516ab3a-en (accessed: June 2, 2026).

Several commenters objected to NHTSA's decision to use a 15 percent rebound effect in its analysis supporting the NPRM, rather than the 10 percent value it has used in some previous rulemaking analyses. ICCT described the lower 10 percent value as “in alignment with prior analyses and updated research, especially in the context of a new fleet that is more efficient than ever and thus costs less to drive than ever [. . .] Restoring a 10 percent rebound is consistent with both historical precedent and recent literature.” ICCT and NRDC et al. both argued that NHTSA's review of the literature was focused on studies using changes in fuel price or fuel CPM rather than changes in fuel efficiency, and that the former generated a stronger response from consumers than the latter.\422\

\422\ ICCT, Docket No. NHTSA-2025-0491-5246 (citing Gillingham, K., Policy Brief: The Rebound Effect and the Rollback of Fuel Economy Standards (2018), available at: https://resources.environment.yale.edu/gillingham/Gillingham_ReboundFuelEconomyStds.pdf (accessed: June 2, 2026)); NRDC et al., Docket No. NHTSA-2025-0491-5928-A2, at 91-98.

NRDC et al. objected to NHTSA's consideration of all published estimates of the rebound effect, arguing that it should more heavily weight estimates identified by the authors of published studies as their preferred estimates, while discounting those the authors may view as less reliable.\423\ NRDC et al. noted that while authors such as Gillingham (2015) and Hymel and Small (2015) report wide ranges of estimates for the rebound effect, these authors prefer estimates toward the lower end of the ranges they report. NRDC et al. also cited some research arguing that the rebound effect is likely to decline over time in response to rising incomes, which raise the value of travel time while reducing the importance of fuel costs in drivers' decisions about how much to travel, thus reducing the magnitude of the fuel economy rebound effect.

\423\ NRDC et al., Docket No. NHTSA-2025-0491-5928-A2, at 91-98.

Similarly, the City of Cleveland (Cleveland) noted “Historically, NHTSA and U.S. EPA have used a 10 percent rebound effect--meaning that VMT will increase 10 percent for every doubling of fuel economy--which reflects the best available research. In this NPRM, NHTSA rejects that 10 percent standard and instead adopts a 15 percent rebound effect. This is an inappropriately high estimate that breaks with recent research and historical norms.” \424\ Attorneys General asserted that “. . . the best evidence on rebound driving available--from the United States, using odometer data from emissions or safety inspections, and from recent years--leads to a central case estimate for the rebound effect of 10 percent, or perhaps even lower,” \425\

\424\ Cleveland, Docket No. NHTSA-2025-0491-4840, at 13.

\425\ Attorneys General, Docket No. NHTSA-2025-0491-6064-A4, at 30.

In contrast, PMI argued that increasing the rebound effect to 15

percent is appropriate, observing that “at least some rebound effect is very likely, meaning the decrease in gasoline consumption will be muted . . . NHTSA's review of the literature suggests “a range of approximately 10-45 percent” for the rebound effect, “with most of the estimates falling in the 15-30 percent range. NHTSA previously assumed that the rebound effect was 10 percent, but NHTSA now assumes that it is 15 percent. Though still conservative, this change is appropriate.” \426\

\426\ PMI, Docket No. NHTSA-2025-0491-5001-A2, at 51.

While NHTSA acknowledges certain commenters' concern about the rebound effect estimate \427\ the agency concludes that the 15 percent value used in this analysis better represents the totality of evidence about the rebound effect's likely magnitude. As indicated above, the agency purposely chose the previous 10 percent value to be at the lower end of the range of values that could be supported by published research. After considering the issue in this rulemaking, however, NHTSA has concluded that it is more appropriate to select a value more accurately representing the full range of available evidence. The 15 percent value better serves this purpose than does 10 percent because the former lies closer to the central tendency of estimates reviewed in Final TSD Chapter 4.3.4. Specifically, the probability distributions of estimates of the rebound effect based on the response of vehicle use to variation in fuel cost per-mile driven and using different data and estimation approaches (shown in Final TSD Figure 4.20) suggest a most likely value close to 15 percent, while the distribution of estimates based on fuel efficiency (see Final TSD Figure 4.19) suggests a most likely value only slightly lower. Using a 15 percent rebound effect in the central analysis better represents the overall sense of the extensive research on its magnitude conducted over the past three decades, and conducting sensitivity analyses using values of 10 percent and 20 percent adequately encompasses the range of plausible values.

\427\ This is not the first time that the agency has re-examined the existing rebound literature. In the agency's 2020 final rule, the agency used an estimate of 20 percent. In its 2021 proposal, it lowered its estimate of the rebound to 15 percent. The agency then lowered its estimate yet again for the final rule to 10 percent. As shown in Final TSD Chapter 4.3.4, there were no studies published over that time period that influenced the agency's decision.

In order to calculate total VMT after allowing for the rebound effect, the CAFE Model applies the price elasticity of VMT (taken from the FHWA forecasting model) to the change in fuel CPM resulting from higher fuel economy and uses the result to adjust the initial estimate of each model's annual use accordingly. The CAFE Model applies this adjustment after the reallocation step described previously because that adjustment is intended to ensure that total VMT is identical among alternatives before considering the contribution of increased driving due to the rebound effect. Its contribution differs among regulatory alternatives because alternatives requiring higher fuel economy lead to larger reductions in the per-mile fuel cost of driving and thus to larger increases in vehicle use.

To summarize, because the finalized standards would lower the cost of newer vehicles, more of the base household travel demand will be satisfied by safer, newer vehicles, and simultaneously, newer vehicles will have lower fuel economy, leading to fewer miles being driven and resulting in a further reduction in fatalities and fuel expenditures.

Chapter 4.3 of the Final TSD provides more information on how NHTSA accounts for and models VMT. 4. Changes to Fuel Consumption

NHTSA uses fuel economy, age, and VMT estimates to determine changes in fuel consumption. NHTSA divides the expected vehicle use by the anticipated mpg to calculate the gallons consumed by each simulated vehicle, and when aggregated, the total fuel consumed in each alternative.

Changes in fuel consumption because of changes in CAFE standards (and resulting technology application) will result in changes in emissions of various pollutants.\428\ Vehicle-related emissions are computed by multiplying vehicle activity (e.g., miles traveled, hours operated, or gallons of fuel burned), population (or number of vehicles), and emission factors. An emission factor is a representative rate that attempts to relate the quantity of a pollutant released to the atmosphere per unit of activity. As in past rules, the CAFE Model generates vehicle activity levels (both miles traveled and fuel consumption), while emission factors have been adapted from models developed and maintained by other Federal agencies.

\428\ The various pollutants include carbon monoxide (CO), volatile organic compounds (VOCs), nitrogen oxides (NOX), sulfur oxides (SOX), particulate matter with a diameter of 2.5-micron ([micro]m) or less (PM2.5), carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O).

This section provides a brief overview of how the agency estimates the resulting changes in emissions and associated effects from emissions of those pollutants.\429\ In this section, emissions that are generated between the initial point of oil extraction and delivering fuel to vehicles' fuel tanks or energy storage systems are referred to as “upstream” emissions, while “downstream” emissions refer to those emitted by vehicles' exhaust systems, and also include other emissions generated during vehicle refueling, use, and inactivity (called “soaking”), including hydrofluorocarbons leaked from vehicles' AC systems.\430\ Emissions also include particulate matter released into the atmosphere by BTW, as well as evaporation of volatile organic compounds from fuel pumps and vehicles' fuel storage systems during refueling and when parked.

\429\ While NHTSA considers the impacts of this rulemaking on the levels of various pollutant emissions, the main analysis does not include a monetization of any changes in levels of CO2, CH4, and N2O emissions. (An analysis using the domestic-only valuation of those emissions is included in a sensitivity case). Monetized changes in criteria pollutant emissions are discussed in the preamble Section II.G and Chapter 6.2.2 of the Final TSD.

\430\ Emissions from HFC leakage from air conditioner systems are not captured in the CAFE Model analysis due to limitations in the pollutants modeled by MOVES5.

For the final rule, the agency updated upstream petroleum emission factors using R&D GREET 2025, a lifecycle emissions model developed by Argonne.\431\ Several commenters supported the inclusion of upstream impacts in this final rule analysis.\432\ As in past analyses, the agency derived emission factors for the following four upstream emission processes for gasoline and diesel: (1) petroleum extraction; (2) petroleum transportation and storage; (3) petroleum refining; and (4) fuel transportation, storage, and distribution. A detailed description of how the agency used R&D GREET 2025 to generate upstream emission factors appears in Chapter 5 of the Final TSD. In this final rule, NHTSA uses a simplified parameterized economic model for estimating the response of domestic fuel production to changes in

U.S. fuel consumption because such responses also affect upstream emissions estimates. Using this model, NHTSA estimates that 20 percent of the reduction in fuel consumption will be translated into reductions in domestic fuel production. Though NHTSA has not made any methodological changes to calculating upstream emissions in this final rule, projections of non-criteria and criteria pollutant upstream emissions inventories in FRIA Chapter 8.5 are now reported separately for global and domestic emissions, in line with Circular A-4 (2003).\433\

\431\ Argonne National Laboratory, The Research and Development Greenhouse Gases, Regulated Emissions, and Energy Use in Technologies (R&D GREET) Model 2025 (2025), last revised: Dec. 2025, available at: https://greet.anl.gov/ (accessed: June 2, 2026).

\432\ CPAC Foundation Center for Regulatory Freedom (CPAC-CRF), Docket No. NHTSA-2025-0491-5054, at 3, 8-9; NACAA, Docket No. NHTSA- 2025-0491-5884, at 14-15; ME DEP, Docket No. NHTSA-2025-0490-0026, at 6.

\433\ Circular A-4.

The agency estimated downstream emission factors for gasoline and diesel fuels for the majority of pollutants using EPA's MOVES5 model, a regulatory highway emissions inventory model developed by that agency's National Vehicle and Fuel Emissions Laboratory.434 435

\434\ EPA, Latest Version of Motor Vehicle Emission Simulator (MOVES), last revised: Apr. 3, 2026, available at: https://www.epa.gov/moves/latest-version-motor-vehicle-emission-simulator-moves (accessed: July 28, 2026).

\435\ The one exception is that downstream CO2 emission factors were generated based on the carbon content and mass density per unit of each specific type of fuel assuming each fuel's entire carbon content is converted to CO2 emissions during combustion. See Final TSD Chapter 5.3 for further discussion.

In the proposal, NHTSA sought comments on the assumptions and methods used to project future emission inventories, which included estimated effects from Federal emissions standards for light-duty vehicles, including EPA's CO2 standards for MYs 2024-2026 and MYs 2027-2031. These effects were estimated prior to EPA publishing its proposal to rescind its action titled “Endangerment and Cause or Contribute Finding for Greenhouse Gases Under Section 202(a) of the Clean Air Act” (Endangerment Finding) and all resulting GHG emissions standards for light-, medium-, and heavy-duty vehicles and engines.\436\ There were no substantive comments on the policy assumptions included in MOVES5 used to estimate emission inventories in the proposal; however, to be consistent with EPA's final rule rescinding the Endangerment Finding, we have removed policy assumptions that are no longer relevant to estimating future non-criteria pollutant emission inventories for this final rule.\437\

\436\ Reconsideration of 2009 Endangerment Finding and Greenhouse Gas Vehicle Standards; Proposed Rule, 90 FR 36288 (2025), available at: https://www.federalregister.gov/documents/2025/08/01/2025-14572/reconsideration-of-2009-endangerment-finding-and-greenhouse-gas-vehicle-standards (accessed: June 3, 2026); see also Rescission of the Greenhouse Gas Endangerment Finding and Motor Vehicle Greenhouse Gas Emission Standards Under the Clean Air Act; Final Rule, 91 FR 7686 (2026), available at: https://www.federalregister.gov/documents/2026/02/18/2026-03157/rescission-of-the-greenhouse-gas-endangerment-finding-and-motor-vehicle-greenhouse-gas-emission (accessed: May 15, 2026).

\437\ See Final TSD Chapter 5.3.1 for more detailed information on downstream emissions modeling updates for this analysis.

In the proposal, NHTSA explored updating its methodology for applying downstream emission factors to vehicle classes within the CAFE Model and sought comment. MOVES regulatory classes may no longer map directly to the CAFE Model regulatory classes beginning in MY 2030, at which time NHTSA will subject vehicles to the amended vehicle classification definitions. However, because the CAFE Model applies downstream emission factors based on specific vehicle attributes, rather than regulatory classification, NHTSA's update to vehicle classification definitions does not affect how downstream emission inventories are estimated. Final TSD Chapter 5.3 contains additional details about how the agency generated the downstream emission factors used in this analysis, and Section VI presents additional information about NHTSA's finalized vehicle reclassification beginning in MY 2030.

As with downstream emission factors, the agency generated BTW emission factors using the latest version of EPA's MOVES5 model.\438\ NHTSA believes that compared to previous versions of MOVES, MOVES5's updated assumptions about brake pad composition and vehicle weights to estimate brake wear emissions that vary by model year, regulatory class, and fuel type present reasonable estimates for use in the agency's regulatory analysis. For further reading on BTW assumptions and how the agency employed those assumptions in the CAFE Model, please refer to Final TSD Chapter 5.3.3.4. NHTSA sought comments on this methodology and received no substantive comments or feedback. NHTSA continues to use the NPRM estimates of BTW emission factors for the analysis supporting this final rule. In terms of combined upstream, downstream, and BTW emissions, commenters noted that the action supported by this analysis would result in increased non-criteria, criteria, and air toxic emissions.\439\

\438\ EPA, Brake and Tire Wear Emissions from Onroad Vehicles in MOVES5, EPA-420-R-24-012, EPA: Washington, DC, pp. 1-69 (2024), available at: https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockey=P101CTUW.pdf (accessed: June 3, 2026).

\439\ The Climate Reality Project (Climate Reality), Docket No. NHTSA-2025-0491-4896, at 2; Puget Sound Clean Air Agency (PSCAA), Docket No. NHTSA-2025-0491-4918, at 2-3; Cleveland, Docket No. NHTSA-2025-0491-4840, at 2, 4-6; AVE, Docket No. NHTSA-2025-0490- 0033, at 5; Southern Environmental Law Center (SELC), Docket No. NHTSA-2025-0490-0035, at 5-11; NACAA, Docket No. NHTSA-2025-0491- 5884, at 1, 8-11; South Coast Air Quality Management District (South Coast AQMD), Docket No. NHTSA-2025-0490-0064, at 3-4; ME DEP, Docket No. NHTSA-2025-0490-0026, at 6; City of Madison, Wisconsin (Madison), Docket No. NHTSA-2025-0491-6063, at 1-2; NRDC et al., Docket No. NHTSA-2025-0491-5928, at 9-10, 15.

The CAFE Model computes select health impacts resulting from localized population exposure to PM2.5 and its precursor pollutants that are measured by the number of instances predicted to result from exposure to each ton of relevant pollutant.\440\ As in past CAFE analyses, NHTSA relied on publicly available scientific literature to estimate PM2.5-related effects for each upstream and downstream emissions source \441\ and employed certain assumptions to determine the most reasonable approach to incorporate estimates from literature into the Model.\442\ NHTSA includes additional discussion of the agency's approach to estimating these effects in Chapter 5.4 of the Final TSD. NHTSA received comments from individuals specifically calling for the agency to estimate direct health damages from air toxics.\443\

\440\ As the health incidences for the different source sectors are all based on the emission of 1 ton of the same pollutants, NOX, SOX, and directly emitted PM2.5, differences in the incidence per ton values arise from differences in the geographic distribution of each pollutant's emissions, which in turn affects the number of people exposed to the estimated concentrations of each pollutant.

\441\ EPA, Estimating the Benefit per Ton of Reducing PM2.5 Precursors from 17 Sectors, EPA: Washington, DC, pp. 1-108 (2018), available at: https://19january2017snapshot.epa.gov/benmap/estimating-benefit-ton-reducing-pm25-precursors-17-sectors_.html (accessed: June 5, 2026); Fann, N. et al., Assessing Human Health PM2.5 and Ozone Impacts from U.S. Oil and Natural Gas Sector Emissions in 2025, Environmental Science & Technology, Vol. 52(15), pp. 8095-103 (2018), available at: https://doi.org/10.1021/acs.est.8b02050 (accessed: June 5, 2026) (hereinafter, “Fann et al.”); Wolfe, P. et al., Monetized Health Benefits Attributable to Mobile Source Emission Reductions Across the United States in 2025, The Science of the Total Environment, Vol. 650 (Pt 2), pp. 2490-98 (2019), available at: https://doi.org/10.1016/j.scitotenv.2018.09.273 (accessed: June 5, 2026) (hereinafter, “Wolfe et al.”). Health incidence per ton values corresponding to this paper were sent by EPA staff.

\442\ Some CAFE Model upstream emissions components do not correspond to any single EPA source sector identified in available literature, so NHTSA determined the most reasonable approach was to use a weighted average of different source sectors to generate those values. NHTSA is also aware that EPA in 2023 updated its estimated benefits for reducing PM2.5 from several sources, but those do not include mobile sources (which include the vehicles subject to CAFE standards). NHTSA has thus retained the PM2.5 incidence per ton values from the previous CAFE analysis for consistency with the current mobile source emissions estimates.

\443\ Jana Milford, Docket No. NHTSA-2025-0491-4828, at 2.

In the absence of any published sources giving per-ton health incidence estimates for these pollutants, NHTSA is continuing to focus the CAFE analysis accounting of health damages from emissions on those stemming from NOX, SO2, and PM2.5. Commenters noted that there are updated damage estimates for criteria pollutants available for some, but not all, sectors.\444\ They encouraged NHTSA to update estimates for the sectors that are available and only retain the current estimates for sectors that have not been updated.\445\

\444\ Jana Milford, Docket No. NHTSA-2025-0491-4828, at 2.

\445\ IPI, Docket No. NHTSA-2025-0491-6015, Appendix at 15; NRDC et al., Docket No. NHTSA-2025-0491-5928, Appendix A at 58-59.

NHTSA has reviewed updated damage estimates for the available sectors and has determined that using these estimates would result in a lack of consistency and the inability to compare health damages across sectors. The available updated damage estimates use different specific health endpoints and including them would require a full restructuring of the health damages analysis that would leave many gaps remaining until the publication of updated damage estimates for all the sectors used in the CAFE analysis. Thus, NHTSA is maintaining the health damages estimates used in the NPRM for this final rule analysis.