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

For the MYs 2022-2026 analysis, NHTSA performed two exercises: first, the agency re-evaluated the statistical model used to determine the shape (i.e., slope, intercept, and cutpoints) of the target functions for passenger cars and light trucks. Then, based on its preferred choice of shape, NHTSA evaluated the compliance position of manufacturers in MYs 2022-2024 under alternative stringencies and compared results to the manufacturers achieved average fuel economy in these years. For both exercises, NHTSA relies on compliance data from manufacturer mid-year compliance reports. For its curve fitting analysis, NHTSA uses vehicle model level data on vehicle attributes, including footprint, HP, CW, and 2-cycle fuel economy. NHTSA also uses mid-year estimates of model sales from manufacturer compliance data. NHTSA's curve fitting analysis is described in greater detail in Final TSD Chapter 1. For NHTSA's comparison of achieved fuel economy and finalized standards levels, the agency uses compliance data at the model level for vehicle footprint, 2-cycle fuel economy, and mid-year estimates of vehicle sales.

For MYs 2022-2024, NHTSA uses each standard to calculate vehicle model target function values for each vehicle model in the standard- setting fleet.\79\ Consistent with past rulemakings, the agency uses piecewise linear functions of vehicle footprint, which map to a target value of fuel consumption rate in gallons per mile.\80\ NHTSA determines a vehicle's target fuel economy level in mpg for a given set of standards and then takes the reciprocal of this value. NHTSA determines the CAFE standards for each manufacturer at the regulatory class level under each alternative by taking the sales-weighted harmonic mean of the relevant models produced by the manufacturer in each regulatory class in each model year. The agency repeats these calculations for each model year under consideration to determine a single value for each regulatory class in which the manufacturer produced vehicles.

\79\ Per 49 U.S.C. 32902(h), dedicated alternative fueled vehicles, such as EVs, are excluded from this analysis. For dual- fueled vehicles, the analysis uses a fuel economy value for the vehicles operating only on gasoline or diesel fuel. Id.

\80\ See Chapter 1.2 of the Final TSD discussing footprint functions.

NHTSA also computes the MDPCS for each model year by taking the sales-weighted harmonic mean of the model-level target function values for all vehicles in the passenger car fleet in that model year and multiplying the value by 92 percent.\81\

\81\ 49 U.S.C. 32902(b)(4).

NHTSA determines each manufacturer's achieved fuel economy in mpg separately for each regulatory class using the sales-weighted average of the 2-cycle fuel economy values of all models produced by the manufacturer in the relevant regulatory class. NHTSA then compares this achieved value to the corresponding regulatory class standard for each manufacturer in each model year to determine whether the fleet of vehicles to which it corresponds would comply with each standard. To determine the total number of vehicles out of compliance, NHTSA determines compliance for each manufacturer's regulatory fleet in each model year under each finalized alternative. If a fleet is determined to be out of compliance, the agency sums the total number of vehicles sold in the non-compliant fleet.

As discussed in more detail in Section IV, NHTSA analyzes the difference between each manufacturer's fleet CAFE compliance value and the standard. NHTSA considered using the CAFE Model to simulate behavior for the MYs 2022-2026 compliance period to estimate how manufacturers and consumers could have responded to different CAFE standards. However, for MYs 2022-2026, production is closed or is in process at the time of this final rule's publishing. This type of analysis overestimates the ability of manufacturers to optimize in response to the finalized standards for these years and likely leads to different results from the actual outcomes. Thus, simulating a response and any monetized costs or benefits deriving from that response do not represent real economic effects from the final change in policy. 2. What inputs does the compliance analysis require for 2027-2031?

For the MYs 2027-2031 amendment analysis, NHTSA used the CAFE Model to simulate manufacturers' potential responses to new CAFE standards and to estimate the various impacts of those responses on manufacturers and society. The Model considers various inputs, such as technology effectiveness data, technology costs, and other relevant factors, and uses those inputs to generate output predictions.

NHTSA attempts to ensure that the technology inputs and assumptions that go into the CAFE Model are based on sound science and reliable data and that NHTSA's reasons for using those inputs and assumptions are transparent and understandable to stakeholders. This section and the following section discuss at a high level how the agency generates the technology inputs and assumptions that the CAFE Model uses for the compliance simulation.\82\ The

Final TSD, CAFE Model Documentation, CAFE Analysis Autonomie Documentation,\83\ and other technical reports supporting this final rule discuss the agency's technology inputs and assumptions in more detail.

\82\ As explained throughout this section, a NHTSA input is a specific number or datapoint used by the Model, and NHTSA's assumptions are based on judgment after careful consideration of available evidence. An assumption can be an underlying reason for the use of a specific datapoint, function, or modeling process. For example, an input might be the fuel economy value of the Ford Mustang, whereas the assumption is that the Ford Mustang's fuel economy value reported in Ford's CAFE compliance data should be used in NHTSA's modeling.

\83\ The Argonne report is titled “Vehicle Simulation Process to Support the Analysis for MY 2027 and Beyond CAFE and MY 2030 and Beyond HDPUV FE Standards.” However, for ease of use and consistency with the Final TSD it is referred to as “CAFE Analysis Autonomie Documentation.”

NHTSA incorporates technology inputs and assumptions either directly in the CAFE Model or in the CAFE Model's various input files. The compliance simulation algorithm is at the heart of the CAFE Model's approach on applying technologies to a manufacturer's vehicles to project how the manufacturer could meet CAFE standards. The compliance simulation algorithm consists of several equations that direct the Model to apply fuel economy-improving technologies to vehicles in a way that simulates how manufacturers might apply those technologies to their vehicles in the real world. The compliance simulation algorithm projects a cost-effective pathway for manufacturers to comply with different levels of CAFE standards, considering the technology present on manufacturers' vehicles now and what technology could be applied to their vehicles in the future. Embedded in the CAFE Model is the universe of technology options that the Model can consider and rules about the order in which it can consider those options, as well as estimates of how effective fuel economy-improving technology is on different types of vehicles (e.g., sedan or pickup truck).

Technology inputs and assumptions are also located in all four of the CAFE Model Input Files. The Market Data Input File is a spreadsheet file that characterizes the fleet of vehicles used as the starting point for the CAFE Model. There is one row describing each vehicle model and model configuration manufactured for the United States market in a model year (or years) and input and assumption data that links those vehicles to technology and economic, environmental, and safety inputs and assumptions. The Technologies Input File identifies 69 technologies the agency uses in the analysis, along with information used to inform the compliance simulation and effects estimates, including phase-in caps to identify when and how widely each technology can be applied to specific types of vehicles, most of the technology costs (hybrid vehicle battery costs are provided in a separate file), and the fuel share percentage for PHEV to capture the charge sustaining operation. The Scenarios Input File provides the coefficient values defining the standards for each regulatory alternative \84\ and other relevant information applicable to modeling each regulatory scenario.\85\ Finally, the Parameters Input File contains mainly economic and environmental data.\86\

\84\ The coefficient values are defined in RIA Chapter 3 for the CAFE standard.

\85\ This file also includes information about the amount of fuel consumption improvement values a manufacturer currently generates for compliance purposes under EPA's regulations and information on EPA's regulatory limits on generating FCIVs for each model year in the analysis. For this analysis the FCIVs will go to 0 in MY 2028 for the regulatory alternatives, as discussed in preamble Section II.D.8.

\86\ See CAFE Model Documentation for a detailed discussion of what inputs are held in each of the input data files.

NHTSA generates these technology inputs and assumptions in several ways, including using data submitted by vehicle manufacturers pursuant to their CAFE reporting obligations; public data on vehicle models from manufacturer websites, press materials, marketing brochures, and other publicly available information; collaborative research, testing, and modeling with other Federal agencies, like Argonne; and research, testing, and modeling with independent organizations, like IAV GmbH Ingenieurgesellschaft Auto und Verkehr (IAV), Southwest Research Institute (SwRI), National Academy of Sciences (NAS), and FEV North America. NHTSA also considers the work done to develop inputs and assumptions for prior rules to the extent it is still relevant and applicable; feedback from stakeholders on prior rules and from meetings conducted before the commencement of this final rule; and NHTSA's own engineering judgment. NHTSA uses the term “engineering judgment” throughout this rulemaking to refer to decisions made by a team of NHTSA engineers and analysts. This judgment is based on their experience working in the automotive industry and other relevant fields and assessment of all the data sources described above. Most importantly, the agency uses engineering judgment to assess how best to represent vehicle manufacturers' potential responses to different levels of CAFE standards within the boundaries of the agency's modeling tools, as “a model is meant to simplify reality in order to make it tractable.” \87\ In other words, NHTSA uses engineering judgment to concentrate potential technology inputs and assumptions from millions of discrete data points from hundreds of sources into four external input files and three datasets integrated into the CAFE Model. How the CAFE Model decides to apply technology (i.e., the compliance simulation algorithm) has been developed using engineering judgment considering factors that manufacturers consider when they add technology to vehicles in the real world. The specific technology inputs and assumptions are discussed in more detail in the following sections and in the associated technical documentation.

\87\ Chem. Mfrs. Ass'n v. EPA, 28 F.3d 1259, 1264-65 (D.C. Cir. 1994) (citing Milton Friedman, in Friedman, M., The Methodology of Positive Economics, in Essays in Positive Economics 3, University of Chicago Press: Chicago, IL, pp. 14-15 (1953), available at: https://www.wiwiss.fu-berlin.de/fachbereich/bwl/pruefungs-steuerlehre/loeffler/Lehre/bachelor/investition/Friedman_the_methology_of_positive_economics.pdf (accessed: May 28, 2026)).

NHTSA begins the compliance analysis by defining the range of fuel economy-improving technologies that the CAFE Model could add to a manufacturer's vehicles in the U.S. market. These are technologies that the agency believes are representative of what vehicle manufacturers currently use on their vehicles, and that vehicle manufacturers could use on their vehicles in the timeframe for the finalized standards (MYs 2027-2031). The technology options include engines, transmissions, hybridization, and road load technologies, which include mass reduction, aerodynamic improvement (aerodynamic drag technology (AERO)), and tire rolling resistance (ROLL) reduction technologies.\88\

\88\ Final TSD Chapter 3 contains discussion on the technology tree and technologies available.

Adding a technology to the range of options that the CAFE Model can consider requires several data elements, including a broadly applicable technology definition, estimates of how effective that technology is at improving fuel economy on different vehicle types (e.g., sedan or pickup truck), and the cost to apply that technology to each. Each technology the agency selects is designed to be representative of a wide range of specific technology applications used in the automotive industry. Some manufacturers' systems may perform better or worse than NHTSA's modeled systems, and some

may cost more or less than NHTSA's modeled systems. However, selecting representative technology definitions for the agency's analysis ensures the agency captures a reasonable level of costs and benefits that would result from any manufacturer applying the technology.

NHTSA has been refining the technology options it considers since first developing the CAFE Model in 2002. In this context, “refining” means both adding and removing technology options depending on current technology availability and projected future availability in the U.S. market, while balancing a reasonable amount of modeling and analytical complexity. In recent years, the agency has refined internal combustion engine (ICE) technology options, particularly the TURBO and high compression ratio (HCR) pathways, to reflect better the diversity of engines in the current fleet. The agency includes several hybrid technologies to represent appropriately the diversity of current and anticipated future technology options while ensuring NHTSA's analysis remains consistent with statutory limitations prohibiting the consideration of EVs in establishing standards and considering only the gas or diesel operation of dual-fueled automobiles.

The technology options do not include technologies NHTSA has determined will not be available in the rulemaking timeframe. As with past analyses, the agency does not include technologies unlikely to be feasible in the rulemaking timeframe, engine technologies designed for markets other than the United States market or required to use unique gasoline,\89\ or technologies for which appropriate data are not available for the range of vehicles that the agency models in the analysis (i.e., technologies that are still in the research and development phase and not ready for mass-market production). Each technology section below and Chapter 3 of the Final TSD discuss these modeling decisions in detail.

\89\ In general, most vehicles produced for sale in the United States have been designed to use “regular” gasoline, or 87 octane. See EIA, Gasoline Explained: What is octane?, last revised: Nov. 17, 2022, available at: https://www.eia.gov/energyexplained/gasoline/octane-in-depth.php (accessed: May 25, 2026).

The CAFE Model does not dictate or predict the technologies manufacturers must use to comply; rather, the CAFE Model outlines a technology pathway that manufacturers could use to meet the standards in a cost-effective way. While NHTSA estimates the costs and benefits for different levels of CAFE standards based on a simulation of the technology manufacturers could apply in the rulemaking timeframe, it is entirely possible and reasonable that manufacturers may use different technology options to meet the agency's standards in the real world and may even use technologies that NHTSA does not include in the analysis. This is because NHTSA's standards do not mandate the application of any technology. Rather, NHTSA's standards are performance-based: manufacturers can and do use a range of compliance solutions that include technology application and encouraging sales shifts from one vehicle model or trim level to another.\90\ The agency has determined that the 69 technology options included in the analysis strike a reasonable balance between representing the diversity of technology used by the entire industry and simplifying reality to make modeling workable.\91\

\90\ Manufacturers could increase their production of one type of vehicle with higher fuel economy, like the hybrid version of a conventional vehicle model, to meet the standards. For example, Ford has conventional and hybrid versions of its F-150 pickup truck, and Toyota has conventional, hybrid, and plug-in hybrid versions of its RAV4 sport utility vehicle.

\91\ For each technology option, the analysis includes distinct technology cost and effectiveness values for 10 different types of vehicles, resulting in nearly half a million different technology effectiveness and cost data points.

Chapter 3 of the Final TSD and Section II.D below describe the technologies that NHTSA uses for the analysis. Each technology has a name that loosely corresponds to its real-world technology equivalent. NHTSA abbreviates the name to a short signifier for the CAFE Model to read. The agency organizes those technologies into groups based on technology type: basic and advanced engines, transmissions, hybridization, and road load technologies, which include mass reduction, aerodynamic improvement, and low rolling resistance tire technologies.

NHTSA then organizes the groups into pathways. The pathways instruct the CAFE Model how and in what order to apply technology. In other words, the pathways define mutually exclusive technologies (i.e., those that cannot be applied at the same time) and define the direction in which vehicles can advance as the Model evaluates which technologies to apply. The respective technology chapters in the Final TSD and Section 4 of the CAFE Model Documentation include a visual of each technology pathway. In general, the paths are tied to ease of implementation of additional technology and how closely the technologies are related.

As an example, NHTSA's “Turbo Engine Path” consists of five different engine technologies that employ different levels of turbocharging technology. A turbocharger is essentially a small turbine driven by exhaust gases produced by the engine. As these gases flow through the turbocharger, they spin the turbine, which in turn spins a compressor that pushes more air into an engine's cylinders. Having more air in the engine's cylinders allows the engine to burn more fuel, which then creates more power, without needing a physically larger engine. In the agency's analysis, an engine that is turbocharged “downsizes,” or becomes smaller. Choosing to turbocharge an engine allows a manufacturer to maintain similar levels of performance to a larger, non-turbocharged engine with a smaller engine that uses less fuel to do the same amount of work. Allowing basic engines to be downsized and turbocharged instead of just turbocharged keeps the vehicle's utility and performance constant so that NHTSA can measure the costs and benefits of different levels of fuel economy improvements, rather than the change in different vehicle attributes. This concept of performance neutrality is discussed further below.

The Model allows only forward movement along the technology pathways, adding more advanced technology as the Model moves through the technology tree. This ensures that a vehicle that uses a more advanced technology cannot downgrade to a less advanced version of the technology and ensures that a vehicle does not switch to technology that is significantly technically different. This progressive order also realistically represents how manufacturers often start with the lowest and most cost-effective technologies and generally advance along particular technology pathways. As an example, if a vehicle in the compliance simulation begins with a TURBOD engine--a turbocharged engine with cylinder deactivation--it cannot adopt a TURBO0 engine.\92\ Similarly, this vehicle with a TURBOD engine cannot adopt an advanced cylinder deactivation on a dual-overhead camshaft (ADEACD) engine.\93\ As an example of NHTSA's rationale for ordering technologies on the technology

tree, an engine could potentially be changed from TURBO0 to TURBO2 without redesigning the engine block or requiring significantly different expertise to design and implement. A change to ADEACD likely would require a different engine block that might not fit in the engine bay of the vehicle without a complete redesign and different technical expertise requiring years of research and development. This change, which would strand capital and impact parts sharing, is why the advanced engine paths restrict most movement between them. The concept of stranded capital is discussed further in Section II.C.2.f.

\92\ TURBO0 is the baseline turbocharged engine and TURBOD is TURBO0 with the addition of cylinder deactivation (DEAC). Chapter 3 of the Final TSD provides more discussion on engine technologies.

\93\ ADEACD is a dual-overhead camshaft engine with advanced cylinder deactivation. Chapter 3 of the Final TSD provides more discussion on engine technologies.

NHTSA also considers two categories of technology that the agency could not simulate as part of the CAFE Model's technology pathways for the regulatory alternatives for some of standard-setting years. “Off- cycle” and AC efficiency are two types of technologies that improve vehicle fuel economy but are not accounted for using 2-cycle testing. To account for the benefits of these technologies, EPA has allowed manufacturers to generate FCIVs when they add these technologies, which are used to improve a manufacturers' fleet average fuel economy used for complying with the CAFE standards. As an example, manufacturers can generate FCIVs for technology like active seat ventilation and solar reflective surface coatings that make the cabin of a vehicle more comfortable for the occupants without using less efficient accessories like heat or AC. Instead of including OC and AC efficiency technologies in the technology pathways, NHTSA includes the improvement as a defined benefit that gets applied to a manufacturer's entire fleet in applicable model years instead of to individual vehicles. The defined benefit that each manufacturer receives in the analysis for using OC and AC efficiency technology on their vehicles is located in the Market Data Input File. Chapter 3.7 of the Final TSD provides more discussion on how OC and AC efficiency technologies are developed and modeled. As discussed further in preamble Section II.D.8, NHTSA is removing consideration of FCIVs from its standard-setting analysis beginning with MY 2028. Preamble Section VI contains discussion of how manufacturers generate FCIVs under the limits for FCIVs under EPA's regulations.

To illustrate how NHTSA simulates technology application, throughout this section NHTSA follows the hypothetical vehicle mentioned above that begins the compliance simulation with a TURBOD engine. The agency's hypothetical vehicle, Generic Motors' Ravine Runner F Series, is a roomy, top-of-the-line SUV. The Ravine Runner F Series starts the compliance simulation with technologies from most technology pathways; specifically, after looking at Generic Motors' website and marketing materials, the agency determines that it has technology that loosely fits within the following technologies that the agency considers in the CAFE Model: it has a turbocharged engine with cylinder deactivation, a fairly advanced 10-speed automatic transmission, a 12V start-stop system, the least advanced tire technology, a fairly aerodynamic vehicle body, and it employs a fairly advanced level of mass reduction. NHTSA tracks the technologies on each vehicle using a “technology key,” which is the string of technology abbreviations for each vehicle. The vehicle technologies and their abbreviations that the agency considers in this analysis are shown in Final TSD Chapter 2. The technology key for the Ravine Runner F Series is “TURBOD; AT10L2; SS12V; ROLL0; AERO5; MR3.” b. Defining Manufacturers' Current Technology Positions in the Analysis Fleet

The Market Data Input File is one of four Excel input files that the CAFE Model uses for compliance and effects simulation. The Market Data Input File's “Vehicles” tab (or worksheet) houses one of the most significant compilations of technology inputs and assumptions in the analysis, which is a characterization of the fleet of vehicle models each manufacturer produced for sale in the United States for MY 2024. This provides the starting point from which the CAFE Model adds fuel economy-improving technology. NHTSA calls this fleet the “analysis fleet.” The analysis fleet includes a number of inputs necessary for the Model to add fuel economy-improving technology to each vehicle for the compliance analysis and to calculate the resulting impacts for the effects analysis.

The “Vehicles” tab contains a separate row for each vehicle model. Vehicle models are vehicles that share the same fuel economy value and vehicle footprint based on EPA's regulations for calculating fuel economy. This means that vehicle “trims” with different configurations that affect the vehicle's certification fuel economy value are considered unique models distinguished in separate rows in the Vehicles tab. For example, the agency's Ravine Runner example vehicle comes in three different configurations--the Ravine Runner FWD, Ravine Runner AWD, and Ravine Runner F Series--which would be reported separately under EPA's regulations for compliance purposes and would therefore result in three separate rows in the “Vehicles” tab.

In each row, NHTSA also designates a vehicle's engine, transmission, and platform codes.\94\ Vehicles that have the same engine, transmission, or platform code are deemed to “share” that component in the CAFE Model. Parts sharing helps manufacturers achieve economies of scale, deploy capital efficiently, and make the most of shared research and development expenses, while still presenting a wide array of consumer choices to the market. The CAFE Model has been developed to treat vehicles, platforms, engines, and transmissions as separate entities, which allows the modeling system to evaluate technology improvements on multiple vehicles that may share a common component concurrently. Sharing also enables realistic propagation, or “inheriting,” of previously applied technologies from an upgraded component down to the vehicle “users” of that component that have not yet realized the benefits of the upgrade. Section 2.1 and Section 4.4 of the CAFE Model Documentation contain additional information about the initial state of the fleet, as well as technology evaluation and inheriting within the CAFE Model.

\94\ Each numeric engine, transmission, or platform code designates important information about that vehicle's technology; for example, a vehicle's 6-digit transmission code includes information about the manufacturer, the vehicle's drive configuration (e.g., front-wheel drive, all-wheel drive, 4WD, or rear-wheel drive), transmission type, number of gears (i.e., a 6- speed transmission has 6 gears), and the transmission variant.

Figure II-1 below shows how an example of how the different configurations of the hypothetical Ravine Runner would be separated. NHTSA sees by the Platform Codes that these Ravine Runners all share the same platform, but only the Ravine Runner FWD and Ravine Runner AWD share an engine. Even so, all three fuel economy values are different, which is common for vehicles that differ in drive type (drive type meaning whether the vehicle has AWD, 4-wheel drive (4WD), front-wheel drive (FWD), or rear-wheel drive (RWD)). Though it is simpler to aggregate vehicles by model, ensuring that NHTSA captures model variants at the level they would be reported for compliance improves the accuracy of the analysis and the potential that estimated costs and benefits from different levels of standards are appropriate. NHTSA includes information about other vehicle

technologies at the farthest right side of the Vehicles tab, and in the “Engines,” “Transmissions,” and “Platforms” worksheets, as discussed further below. [GRAPHIC] [TIFF OMITTED] TR30SE26.069

Moving from left to right on the Vehicles tab, after including general information about vehicles and their compliance fuel economy value, NHTSA includes sales and manufacturer's suggested retail price (MSRP) data, regulatory class information (e.g., domestic passenger automobile, import passenger automobile, or non-passenger automobile), and information about how NHTSA classifies vehicles for the effectiveness and safety analyses. Each of these data points is important to different parts of the compliance and effects analysis, so that the CAFE Model can accurately average the technologies required across a manufacturer's regulatory fleet to meet its CAFE standard or estimate the impacts of higher fuel economy standards on vehicle sales.

\95\ Note that not all data columns are shown in this example for brevity.

Next, NHTSA includes vehicle information necessary for applying different types of technology; for example, designating a vehicle's body style allows NHTSA to apply aerodynamic technology appropriately, and designating starting CW values allows the agency to apply mass reduction technology more accurately. Importantly, this section also includes vehicle footprint data, which is needed because NHTSA sets footprint-based standards.

NHTSA also sets product design cycles, which are the years in which the CAFE Model can apply technologies to vehicles. Manufacturers often introduce fuel-saving technologies at a “redesign” of their product or adopt technologies at “refreshes” in between product redesigns. As an example, the redesigned third generation Chevrolet Silverado was released for MY 2019 and featured a new platform, updated drivetrain, increased towing capacity, reduced weight, improved safety, and expanded trim levels, to name a few improvements. For MY 2022, the Chevrolet Silverado received a refresh (or facelift as it is commonly called), with an updated interior, infotainment, and front-end appearance.\96\ Setting these product design cycles provides realistic durations of product stability and ensures that the CAFE Model simulates the opportunities manufacturers have to apply technologies in line with refresh and redesign cycles.

\96\ GM Authority, 2022 Chevy Silverado, last revised: 2022, available at: https://gmauthority.com/blog/gm/chevrolet/silverado/2022-chevrolet-silverado/ (accessed: May 28, 2026).

During modeling, all improvements from technology application are initially realized on a component and then propagated (or inherited) down to the vehicles that share that component. As such, new component- level technologies are initially evaluated and applied to a platform, engine, or transmission during their respective redesign or refresh years. Any vehicles that share the same redesign or refresh schedule as the component apply these technology improvements during the same model year. The rest of the vehicles inherit technologies from the component during their refresh or

redesign year (for engine- and transmission-level technologies) or during a redesign year only (for platform-level technologies). Section 4.4 of the CAFE Model Documentation contains additional information about technology evaluation and inheriting within the CAFE Model.

The CAFE Model also considers the potential safety effect of mass reduction technologies and crash compatibility of different vehicle types. Mass reduction technologies lower the vehicle's CW, which may change crash compatibility and safety, depending on the type of vehicle. NHTSA assigns each vehicle in the Market Data Input File a “safety class” that best aligns with the CAFE Model's analysis of vehicle mass, size, and safety, and include the vehicle's starting CW.97 98

\97\ Vehicle curb weight is the weight of the vehicle with all fluids and components but without the drivers, passengers, or cargo.

\98\ Preamble Section II.H.1 and Final TSD Chapter 7.3 provides more in depth discussion on the impacts of mass reduction on safety.

The CAFE Model includes procedures to consider the direct labor impacts of manufacturers' responses to CAFE regulations, considering the assembly location of vehicles, engines, and transmissions; the percent U.S. content (based on the percent U.S. and Canadian content, as reported by manufacturers to NHTSA); and the dealership employment associated with new vehicle sales. Estimated labor information, by vehicle, is included in the Market Data Input File. Sales volumes included in and adapted from the market data also influence total estimated direct labor projected in the analysis. Chapter 6.2.5 of the Final TSD contains additional discussion of the labor utilization analysis.

NHTSA then assigns the technologies to individual vehicles. This initial linkage of vehicle technologies is how the CAFE Model knows how to advance a vehicle down each technology pathway. Assigning CAFE Model technologies to individual vehicles is dependent on the mix of information the agency has about any particular vehicle and trends about how a manufacturer has added technology to that vehicle in the past, equations and models that translate real-world technologies to their counterparts in NHTSA's analysis (e.g., drag coefficients and body styles can be used to determine a vehicle's AERO level), and the agency's engineering judgment.

As discussed further below, the agency uses information directly from manufacturers to populate some fields in the Market Data Input File, like vehicle HP ratings and vehicle weight. NHTSA also uses manufacturer data as an input to various other models that calculate how a manufacturer's real-world technology equates to a technology level in the agency's model. For example, the agency calculates initial mass reduction, aerodynamic drag reduction, and ROLL levels by looking at industry-wide trends and calculating--through models or equations-- levels of improvement for each technology. The models and algorithms that the agency uses are described further below and in detail in Chapter 3 of the Final TSD. Other fields, like vehicle refresh and redesign years, are projected forward based on historic trends.

Recall the Ravine Runner F Series example with the technology key “TURBOD; AT10L2, SS12V; ROLL0; AERO5; MR3.” For this example, Generic Motor's publicly available specification sheet for the Ravine Runner F Series says that it uses Generic Motor's Turbo V6 engine with proprietary Adaptive Cylinder Management Engine (ACME) technology. Generic Motor's ACME improves fuel economy and lowers emissions by operating the engine using only three of the engine's cylinders in most conditions and using all six engine cylinders when more power is required. Based on this information, NHTSA would conclude that this engine is turbocharged and uses a form of cylinder deactivation, meaning it would be appropriately classified as TURBOD. Generic Motors uses this engine in several of their vehicles, and the specifications of the engine can be found in the Engines Tab of the Market Data Input File, under a six-digit engine code.\99\

\99\ Like the transmission codes discussed above, the engine codes include information identifying the manufacturer, engine displacement (how many liters the engine is), whether the engine is naturally aspirated or force-inducted (turbocharged), and other unique engine attributes.

This is a relatively easy engine to assign based on publicly available specification sheets, but some technologies are more difficult to assign. Manufacturers use different trade names or terms for different technology, and the way that the agency assigns the technology in the agency's analysis may not necessarily line up with how a manufacturer describes the technology. NHTSA must use some engineering judgment to determine how discrete technologies in the market best fit the technology options that the agency considers in the agency's analysis. The agency discusses factors used to assign each vehicle technology in the individual technology subsections below.

In addition to the Vehicles Tab that houses the analysis fleet, the Market Data Input File includes information that affects how the CAFE Model might apply technology to vehicles in the compliance simulation. Specifically, the Market Data Input File's “Manufacturers” tab includes a list of vehicle manufacturers considered in the analysis and several pieces of information about their economic and compliance behaviors. For this analysis, the compliance simulation assumes that manufacturers continue to apply technology to the extent practicable to reach compliance. This modeling change is made by indicating in the “Manufacturers” tab that all manufacturers will comply with NHTSA's standards and is consistent with the recent amendment to EPCA that set civil penalties (i.e., fines) to $0 effective for MY 2022 vehicles and beyond.\100\ The CAFE Model's compliance simulation algorithm is discussed in Section II.C.2.f.

\100\ See Public Law 119-21, 139 Stat. 72, sec. 40006 (July 4, 2025).

Finally, NHTSA designates a “payback period” for each manufacturer. The payback period represents an assumption that consumers are willing to buy vehicles with more fuel economy technology because the fuel economy technology saves them money on gas in the long run. For the past several rulemaking analyses using the CAFE Model the agency has assumed that in the absence of CAFE or other regulatory standards, manufacturers apply technology that “pays for itself”--by saving the consumer money on fuel--in 30 months, or 2.5 years. NHTSA has updated the agency's payback period for this rulemaking to assume a full 3-year payback period based on an examination of empirical economics literature. This is discussed in detail in Section II.E.1.a below, and in the Final TSD and FRIA.

Before the agency begins building the Market Data Input File for any analysis, NHTSA must consider what model year vehicles comprise the analysis fleet. There is an inherent time delay in the data the agency can use for any analysis because NHTSA receives compliance data after a model year has been completed.

For this rulemaking, NHTSA uses data from manufacturers' 2024 mid- model year compliance reports. Though the agency possesses a limited amount of more recent data, NHTSA is not using

that data for this rulemaking because the dataset is not complete.

At the time NHTSA starts building the analysis fleet, data received from vehicle manufacturers \101\ offers the best snapshot of vehicles for sale in the United States in a model year. The mid-model year reports include information about individual vehicles at the vehicle configuration level. NHTSA uses the vehicle configuration, certification fuel economy, sales, regulatory class, and additional technology data from these reports as the starting point to build a “row” (i.e., a vehicle model, with all necessary information about the vehicle) in the Market Data Input File's Vehicles Tab. Additional technology data comes from publicly available information, including vehicle specification sheets, manufacturer press releases, owner's manuals, and websites. NHTSA also generates some assumptions in the Market Data Input File for data fields where there is limited data, like refresh and redesign cycles for future model years, and technology levels for certain road load reduction technologies like mass reduction and aerodynamic drag reduction.

\101\ 49 U.S.C. 32907(a)(2) and 49 CFR part 537.

For this analysis, the light-duty analysis fleet consists of every vehicle model in MY 2024 in nearly every configuration that has a different compliance fuel economy value. This results in nearly 4,000 individual rows in the Vehicles Tab of the Market Data Input File.

The next section discusses how the agency's analysis evaluates how effectively adding technology to a vehicle in the analysis fleet improves that vehicle's fuel economy value. c. Technology Effectiveness Values

The CAFE Model uses technology effectiveness values to allow it to know which technologies to apply. Without these values, it does not know how effective any particular technology is at improving a vehicle's fuel economy value. Accurate technology effectiveness estimates require information about (1) the vehicle type and size; (2) other technologies on the vehicle or being added to the vehicle at the same time; and (3) and how the vehicle is driven. Any oversimplification of these complex factors could make the effectiveness estimates less accurate.

To build a database of technology effectiveness estimates that includes these factors, NHTSA partners with Argonne. Argonne has developed and maintains a modeling and simulation tool called Autonomie that generates technology effectiveness estimates for the CAFE Model. The Autonomie Model is a mathematical representation of an entire vehicle, including its individual technologies (such as the engine and transmission), overall vehicle characteristics (such as mass and aerodynamic drag), and environmental conditions (such as ambient temperature and barometric pressure). The Autonomie Model simulates vehicle behavior over time.

NHTSA simulates a vehicle model's behavior over the two-cycle tests used to measure vehicle fuel economy.\102\ The two-cycle test is carried out by operating a vehicle on a dynamometer. Using a dynamometer is like running a car on a treadmill following a program-- or more specifically, two programs. The programs are the Federal Test Procedure (FTP) and the Highway Fuel Economy Test (HFET). The FTP and HFET are also commonly referred to as the urban cycle and highway cycle, respectively. For the FTP drive cycle, the vehicle meets certain speeds at certain times during the test, or in technical terms, the vehicle must follow a designated speed trace.\103\ The FTP is meant to simulate stop-and-go city driving, and the HFET is meant to simulate steady flowing highway driving at about 50 miles per hour (mph). The agency also uses Society of Automotive Engineers (SAE) recommended practices to simulate hybridized drive cycles,\104\ which involves the test cycles mentioned above as well as additional test cycles to measure battery energy consumption and range. For PHEVs, this analysis utilizes only the gasoline (charge-sustaining) mode for the drive cycles.

\102\ NHTSA is statutorily required to use the two-cycle tests to measure vehicle fuel economy in the CAFE program. See 49 U.S.C. 32904(c) (“Testing and calculation procedures. . . . [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.”).

\103\ EPA, Emissions Standards Reference Guide: EPA Federal Test Procedure (FTP), last revised: Mar. 13, 2025, available at: https://www.epa.gov/emission-standards-reference-guide/epa-federal-test-procedure-ftp (accessed: May 28, 2026).

\104\ SAE, Recommended Practice for Measuring the Exhaust Emissions and Fuel Economy of Hybrid-Electric Vehicles, Including Plug-in Hybrid Vehicles, SAE Standard J1711_202302, SAE International: Warrendale, PA (2023), available at: https://www.sae.org/standards/j1711_202302-recommended-practice-measuring-exhaust-emissions-fuel-economy-hybrid-electric-vehicles-including-plug-hybrid-vehicles (accessed: May 28, 2026); SAE, Battery Electric Vehicle Energy Consumption and Range Test Procedure, SAE Standard J1634_202104, SAE International: Warrendale, PA (2021), available at: https://www.sae.org/standards/content/j1634_202104/ (accessed: May 28, 2026).

Measuring every vehicle's fuel economy value by using the same test cycles ensures that the fuel economy certification results are repeatable for each vehicle model and comparable across all the different vehicle models. When performing physical vehicle cycle testing, sophisticated test and measurement equipment is calibrated according to strict industry standards, which ensures repeatability and comparability of the results. Testing variables can include dynamometers, environmental conditions, types and locations of measurement equipment, and precise testing procedures. These physical tests provide the benchmarking empirical data used to develop and verify Autonomie's vehicle control algorithms and simulation results. Autonomie's inputs are discussed in more detail later in this section.

Full-vehicle modeling and simulation are also essential to measuring how all technologies on a vehicle interact. For example, if technology A improves a particular vehicle's fuel economy by 5 percent and technology B improves a particular vehicle's fuel economy by 10 percent, an analysis using single or limited point estimates may erroneously assume that applying both of these technologies together would achieve a simple additive fuel economy improvement of 15 percent. Single point estimates generally do not provide accurate effectiveness values because they do not capture complex relationships among technologies. Technology effectiveness often differs significantly depending on the vehicle type (e.g., sedan or pickup truck) and the way in which the technology interacts with other technologies on the vehicle, as different technologies may provide different incremental levels of fuel economy improvement if implemented alone or in combination with other technologies. Any oversimplification of these complex factors could lead to less accurate technology effectiveness estimates.

In addition, because manufacturers often add several fuel-saving technologies simultaneously when redesigning a vehicle, it is difficult to isolate the effect of adding any one individual technology to the full-vehicle system. Modeling and simulation offer the opportunity to isolate the effects of individual technologies by using a single or small number of initial vehicle configurations and incrementally adding technologies to those configurations. This provides a consistent reference point for the incremental effectiveness estimates for each technology and for combinations of technologies for each vehicle type.

Vehicle modeling also reduces the potential for overcounting or undercounting technology effectiveness.

Argonne does not build an individual vehicle model for every single-vehicle configuration in NHTSA's light-duty Market Data Input File. This would be nearly impossible, because Autonomie requires very detailed data on hundreds of different vehicle attributes (e.g., the weight of the vehicle's fuel tank, the weight of the vehicle's transmission housing, the weight of the engine, or the vehicle's 0-60 mph time) to build a vehicle model. For practical reasons, NHTSA cannot acquire 4,000 vehicles and obtain these measurements every time the agency promulgates a new rule, and the agency cannot acquire vehicles that have not yet been built. Rather, Argonne builds a discrete number of vehicle models representative of the most popular vehicles on sale in the current fleet. The agency refers to the vehicle model's type and performance level as the vehicle's “technology class.” By assigning each vehicle in the Market Data Input File a “technology class,” NHTSA can connect it to the Autonomie effectiveness estimate that best represents how effective the technology would be on the vehicle, accounting for vehicle characteristics like body style (e.g., sedan or pickup truck) and performance metrics. Because each vehicle technology class has unique characteristics, the effectiveness of technologies and combinations of technologies is different for each technology class.

There are 10 technology classes for this analysis: small car (SmallCar), small performance car (SmallCarPerf), medium car (MedCar), medium performance car (MedCarPerf), small SUV (SmallSUV), small performance SUV (SmallSUVPerf), medium SUV (MedSUV), medium performance SUV (MedSUVPerf), pickup truck (Pickup), and high towing pickup truck (PickupHT).

NHTSA uses a two-step process that involves two algorithms to give vehicles a “fit score” that determines which vehicles best fit into each technology class. At the first step, the agency determines the vehicle's size. At the second step, NHTSA determines the vehicle's performance level. Both algorithms consider several metrics about the individual vehicle and compare that vehicle to other vehicles in the analysis fleet. This process is discussed in detail in Final TSD Chapter 2.2.

Consider NHTSA's example Ravine Runner F Series, which is a medium- sized performance SUV. The exact same combination of technologies on the Ravine Runner F Series operate differently in a compact car or pickup truck because they are different vehicle sizes. The example Ravine Runner F Series also achieves slightly better performance metrics than other medium-sized SUVs in the analysis fleet. By “performance metrics,” the agency means power, acceleration, handling, braking, and so on. For the performance versus standard technology classification, the agency considers the vehicle's estimated 0-60 mph time compared to an average 0-60 mph time for the vehicle's technology class. Accordingly, the “technology class” for the Ravine Runner F Series in the agency's analysis is “MedSUVPerf,” because it meets the criteria of a “performance” 0-60 mph acceleration time.

Table II-2 shows how vehicles in different technology classes that use the exact same fuel economy technology have very different absolute fuel economy values. Note that the Autonomie absolute fuel economy values are not used directly in the CAFE Model; NHTSA calculates the ratio between two Autonomie absolute fuel economy values (one for each technology key for a specific technology class) and applies that ratio to an analysis fleet vehicle's starting fuel economy value. [GRAPHIC] [TIFF OMITTED] TR30SE26.070

Depending on the technology, when two technologies are added to the vehicle together, they may not result in an additive fuel economy improvement. This is an important concept to understand because in Section II.D, NHTSA presents technology effectiveness estimates for every single combination of technology that could be applied to a vehicle. In some cases, technology effectiveness estimates show that a combined technology has a different effectiveness estimate than if the individual technologies were added together individually. However, this is expected and not an error.

Continuing NHTSA's example from above, turbocharging technology and dynamic cylinder deactivation (DEAC) technology both improve fuel economy by reducing the engine displacement and accordingly burning less fuel. Turbocharging allows a manufacturer to use a smaller engine that can offer performance equivalent to a larger naturally aspirated engine, and its fuel efficiency improvements are, in part, due to the reduced displacement. DEAC effectively makes an engine with a particular displacement intermittently offer some of the fuel economy benefits of a smaller displacement engine by deactivating cylinders when the work demand does not require the full engine displacement and reactivating them as-needed to meet higher work demands; the greater the displacement of the deactivated cylinders, the greater the fuel economy benefit. Therefore, a manufacturer upgrading to an engine that uses both a turbocharger and DEAC technology, like the TURBOD engine in the example above, would not see the full combined fuel economy improvement from that specific combination of technologies. Table II-3 shows a vehicle's fuel economy value when using the first-level DEAC technology and when using the first-level turbocharging technology, compared to the agency's example vehicle that uses both of those

technologies combined with a TURBOD engine. [GRAPHIC] [TIFF OMITTED] TR30SE26.071

As expected, the percent improvement in Table II-3 between the first and second rows is 1.7 percent and between the third and fourth rows is 0.3 percent, even though the only difference within the two sets of technology keys is the DEAC technology (note that the agency only compares technology keys within the same technology class). This is because there are complex interactions between all fuel economy- improving technologies. The agency models these individual technologies and groups of technologies to reduce the uncertainty and improve the accuracy of the CAFE Model outputs.

Some technologies that NHTSA discusses in Section II.D include advanced engine and hybrid powertrain technology--with some combinations that do not pair well. As an example, NHTSA does not see a particularly high effectiveness improvement from applying advanced engines to existing parallel strong hybrid (e.g., P2) architectures.\105\ In this instance, the P2 powertrain improves fuel economy, in part, by allowing the engine to spend more time operating at efficient engine speed and load conditions. This reduces the advantage of adding advanced engine technologies, which also improve fuel economy, by broadening the range of speed and load conditions for the engine to operate at high efficiency. This redundancy in fuel- saving mechanisms results in a lower effectiveness when the technologies are added to each other. Again, NHTSA expects that different combinations of technologies will provide different effectiveness improvements on different vehicle types. These examples all illustrate relationships observed using only full-vehicle modeling and simulation.

\105\ A parallel strong hybrid powertrain is fundamentally similar to a conventional powertrain but adds one electric motor to improve efficiency. Final TSD Chapter 3 shows all of the parallel strong hybrid powertrain options that NHTSA has modeled in this analysis.

Just as NHTSA's CAFE Model analysis requires a large set of technology inputs and assumptions, the Autonomie modeling uses a large set of technology inputs and assumptions. Figure II-2 below shows the suite of fuel consumption input data used in the Autonomie modeling to generate the fuel consumption input data NHTSA uses in the CAFE Model.

[GRAPHIC] [TIFF OMITTED] TR30SE26.072

As shown in Figure II-2 above, full-vehicle benchmarking is a major source of data for the Autonomie model. For full-vehicle benchmarking, vehicles are instrumented with sensors and tested on both the road and chassis dynamometers (i.e., the full-vehicle treadmills used to exercise the vehicle to provide means to calculate a vehicle's fuel economy values) under different conditions and duty-cycles. Vehicles are selected for benchmarking with the goal of selecting a mix of vehicles most representative of vehicle fleet and available technologies, taking into account sales volume, cost, and availability. Some examples of full-vehicle benchmark testing performed in conjunction with the agency's partners at Argonne include a 2019 Chevrolet Silverado, a 2021 Toyota Rav4 Prime, and a 2022 Hyundai Sonata Hybrid.\106\ NHTSA has produced a report for each vehicle benchmarked, which can be found in the docket. As discussed further below, full-vehicle benchmarking data are used as inputs to the engine modeling and Autonomie full-vehicle simulation modeling. Component benchmarking is like full-vehicle benchmarking, but instead of testing a full vehicle, the agency instruments a single production component or prototype component with sensors and tests it on a similar duty-cycle as a full vehicle. Examples of components NHTSA benchmarks include engines, transmissions, axles, electric motors, and batteries. Component benchmarking data are used as an input to component modeling, where a production or prototype component is changed in fit, form, or function and modeled in the same scenario. As an example, NHTSA might model a decrease in the size of holes in fuel injectors to see the fuel atomization impact or see how it affects the fuel spray angle.

\106\ For all Argonne full-vehicle benchmarking reports, see Docket No. NHTSA-2023-0022-0010.

NHTSA uses a range of models to perform component modeling. As shown in Figure II-2, battery pack modeling using Argonne's BatPaC Model \107\ and engine modeling are two of the most significant component models used to generate data for the Autonomie modeling. NHTSA discusses BatPaC in detail in Section II.D, but briefly, BatPaC is the battery pack modeling tool used to estimate the cost of vehicle battery packs for all hybridized vehicles, which is based on the materials chemistry, battery design, and manufacturing design of the plants manufacturing the battery packs.

\107\ CAFE Analysis Autonomie Documentation chapter titled “Battery Performance and Cost Model--BatPac Examples From Existing Vehicles in the Market.”

Engine modeling is used to generate engine fuel map models that define the fuel consumption rate for an engine equipped with specific technologies when operating over a variety of engine load and engine speed conditions. Some performance metrics captured in engine modeling include power, torque, airflow, volumetric efficiency, fuel consumption, turbocharger performance and matching, pumping losses, and more. Each engine map model has been developed ensuring the engine will still operate under real-world constraints using a suite of other models. Some examples of these models that ensure the engine map models capture real-world operating constraints include simulating heat release through a predictive combustion model, simulating knock characteristics through a kinetic fit knock model,\108\ and using physics-based heat flow and friction models, among others. NHTSA simulates these constraints using data gathered from component benchmarking as well as engineering and physics calculations.

\108\ Engine knock occurs when combustion of some of the air/ fuel mixture in the cylinder does not result from propagation of the flame front ignited by the spark plug; rather one or more pockets of air/fuel mixture explode outside of the envelope of the normal combustion front. Engine knock can result in unsteady operation and damage to the engine.

IAV and SwRI developed the engine map models, using the GT- POWER(copyright) modeling tool (GT-POWER), by creating a base, or root, engine map and then modifying that root map, incrementally, to isolate the effects of the added technologies. The engine maps are based on real-world engine designs. An important feature of the engine maps is that they use a knock model. As noted above, a knock model ensures that any engine size or specification that the agency models in the analysis does not result in engine knock, which could damage engine components in a real-world vehicle. Though the same engine map models are used for all vehicle technology classes, the effectiveness varies based on the characteristics of each class. For example, as discussed above, a compact car with a turbocharged engine has a different effectiveness value than a pickup truck with the same engine technology type. The engine map model development and specifications are discussed further in Chapter 3 of the Final TSD.

Argonne also compiles a database of vehicle attributes and characteristics reasonably representative of the vehicles in that technology class used to build the vehicle models. Relevant vehicle attributes may include a vehicle's fuel efficiency, HP, 0-60 mph acceleration time, and stopping distance, among others, while vehicle characteristics may include whether the vehicle has AWD, 18-inch wheels, summer tires, and so on. Argonne has identified representative vehicle attributes and characteristics for the light-duty fleet from publicly available information and automotive benchmarking databases, such as A2Mac1,\109\ Argonne's Downloadable Dynamometer Database (D\3\),\110\ EPA compliance and fuel economy data,\111\ EPA guidance on 2-cycle tests,\112\ and industry partnerships.\113\ The resulting vehicle technology class baseline assumptions and characteristics database consists of over 100 different attributes like vehicle height and width and weights for individual vehicle parts.

\109\ A2Mac1: Automotive Benchmarking (proprietary data), available at: https://www.a2mac1.com (accessed: May 28, 2026). A2Mac1 is subscription-based benchmarking service that conducts vehicle and component teardown analyses. Annually, A2Mac1 removes individual components from production vehicles, such as oil pans, electric machines, engines, and transmissions, among many other components. These components are weighed and documented for key specifications, which are then available to subscribers.

\110\ Argonne National Laboratory, Downloadable Dynamometer Database, last revised: 2025, available at: https://www.anl.gov/taps/downloadable-dynamometer-database (accessed: May 28, 2026).

\111\ EPA, Compliance and Fuel Economy Data: Data on Cars Used for Testing Fuel Economy, last revised: May 19, 2025, available at: https://www.epa.gov/compliance-and-fuel-economy-data/data-cars-used-testing-fuel-economy (accessed: May 28, 2026).

\112\ 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, pp. 2-265--2-266 (2016), available at: https://downloads.regulations.gov/EPA-HQ-OAR-2022-0829-0230/attachment_1.pdf (accessed: May 28, 2026).

\113\ North American Council for Freight Efficiency, Research & Analysis Are Fundamental, last revised: 2025, available at: https://www.nacfe.org/research/overview (accessed: May 28, 2026).

Argonne then assigns “reference” technologies to each vehicle model. The reference technologies are the technologies on the first step of each CAFE Model technology pathway, and they closely (but not exactly) correlate to the technology abbreviations that NHTSA uses in the CAFE Model. As an example, the first Autonomie vehicle model in the MedSUVPerf technology class starts out with the least advanced engine, which is DOHC (a dual-overhead cam engine) in the CAFE Model, or eng01 in the Autonomie modeling. The vehicle has the least advanced transmission (AT5), the least advanced mass reduction level (MR0), the least advanced aerodynamic body style (AERO0), and the least advanced ROLL level (ROLL0). The first vehicle model is also defined by initial vehicle attributes and characteristics that consist of data from the suite of sources mentioned above. Again, these attributes are meant to represent the average of vehicle attributes found on vehicles in a certain technology class.

Then, just as a vehicle manufacturer tests its vehicles to ensure they meet specific performance metrics, Autonomie ensures that the built vehicle model meets its performance metrics. NHTSA includes quantitative performance metrics in the agency's Autonomie modeling to ensure that the vehicle models can meet real-world performance metrics that consumers observe and that are important for vehicle utility and customer satisfaction. The four performance metrics that NHTSA uses in the Autonomie modeling for light-duty vehicles are low-speed acceleration (the time required to accelerate from 0 to 60 mph), high- speed passing acceleration (the time required to accelerate from 50 to 80 mph), gradeability (the ability of the vehicle to maintain constant 65 mph speed on a 6-percent upgrade), and towing capacity for light- duty pickup trucks. The agency has been using these performance metrics for the last several CAFE Model analyses, and vehicle manufacturers have agreed that these performance metrics are representative of the metrics considered in the automotive industry.\114\ Argonne simulates the vehicle model driving the two-cycle tests (i.e., running its treadmill “programs”) to ensure that it meets its applicable performance metrics (i.e., NHTSA's MedSUVPerf does not have to meet the towing capacity performance metric because it is not a pickup truck). These metrics are based on commonly used metrics in the automotive industry, including SAE J2807 tow requirements.\115\ Additional details about how NHTSA sizes light-duty powertrains in Autonomie to meet defined performance metrics can be found in the CAFE Analysis Autonomie Documentation.

\114\ See NHTSA-2021-0053-1492, at 134 (“Vehicle design parameters are never static. With each new generation of a vehicle, manufacturers seek to improve vehicle utility, performance, and other characteristics based on research of customer expectations and desires, and to add innovative features that improve the customer experience. [NHTSA and EPA] have historically sought to maintain the performance characteristics of vehicles modeled with fuel economy- improving technologies. Auto Innovators encourages the agencies to maintain a performance-neutral approach to the analysis, to the extent possible. Auto Innovators appreciates that the agencies continue to consider high-speed acceleration, gradeability, towing, range, traction, and interior room (including headroom) in the analysis when sizing powertrains and evaluating pathways for road- load reductions. All of these parameters should be considered separately, not just in combination. (For example, we do not support an approach where various acceleration times are added together to create a single `performance' statistic. Manufacturers must provide all types of performance, not just one or two to the detriment of others.)”).

\115\ SAE, Performance Requirements for Determining Tow-Vehicle Gross Combination Weight Rating and Trailer Weight Rating, SAE Standard J2807_202411, SAE International: Warrendale, PA (2024), available at: https://doi.org/10.4271/J2807_202411 (accessed: May 28, 2026).

If the vehicle model does not initially meet one of the performance metrics, then Autonomie's powertrain sizing algorithm increases the vehicle's engine power. The increase in power is achieved by increasing engine displacement (which is the measure of the volume of all cylinders in an engine), which might involve an increase in the number of engine cylinders, which may lead to an increase in the engine weight. This iterative process then determines if the baseline vehicle with increased engine power and corresponding updated engine weight meets the required performance metrics. The powertrain sizing algorithm stops once all the baseline vehicle's performance requirements are met.

Some technologies require extra steps for performance optimization before the vehicle models are ready for simulation. Specifically, the sizing and optimization process is more complex for hybridized vehicles which include hybrid electric vehicle (HEVs) and PHEVs, compared to vehicles with only ICE engines, as discussed further in the Final TSD Chapter 3.3.4. As an example, a PHEV powertrain that can travel a certain number of miles on its battery energy alone (referred to as all-electric range (AER)), or as performing in electric-only mode) is also sized to ensure that it can meet the performance requirements of the SAE standardized drive cycles mentioned above in electric-only mode. Autonomie follows EPA's regulatory guidance and uses the SAE J1711 test procedure to model the incremental effectiveness of adding PHEV technology to a vehicle. The procedure from this guidance is divided into several phases that model “charge sustaining,” “charge depleting,” and “cold operation” \116\ calculations for different test cycles. This is described in detail in the CAFE Analysis Autonomie Documentation.\117\ Final TSD Chapter 3.3.4 and the CAFE Analysis Autonomie Documentation contain more information on PHEV effectiveness.

\116\ SAE J1711 cold test operation occurs in both Charge Sustaining and Charge Depleting modes.

\117\ Chapter “Vehicle Sizing Process” of the CAFE Analysis Autonomie Documentation.

Every time a vehicle model in Autonomie adopts a new technology, the vehicle weight is updated to reflect the weight of the new technology. For some technologies, the direct weight change is easy to assess. For example, when a vehicle is updated to a higher geared transmission, the weight of the original transmission is replaced with the corresponding transmission weight (e.g., the weight of a vehicle moving from a 6-speed automatic (AT6) to an 8-speed automatic (AT8) transmission is updated based on the 8-speed transmission weight). For other technologies, like engine technologies, calculating the updated vehicle weight is more complex. As discussed earlier, modeling a change in engine technology involves both the new technology adoption and a change in power (because the reduction in vehicle weight leads to lower engine loads and a resized engine). When a vehicle adopts new engine technology, the associated weight change to the vehicle is accounted for based on a regression analysis of engine weight versus power.\118\

\118\ Merriam-Webster, Definition: Regression analysis, last revised: 2026, available at: https://www.merriam-webster.com/dictionary/regression%20analysis (accessed: May 28, 2026) (“the use of mathematical and statistical techniques to estimate one variable from another especially by the application of regression coefficients, regression curves, regression equations, or regression lines to empirical data”). In this case, NHTSA is estimating engine weight by looking at the relationship between engine weight and engine power.

In addition to using performance metrics commonly used by automotive manufacturers, NHTSA instructs Autonomie to mimic real-world manufacturer decisions by resizing engines only at specific intervals in the analysis and in specific ways. When a vehicle manufacturer is making decisions about how to change a vehicle model to add fuel economy-improving technology, the manufacturer could entirely redesign the vehicle, or the manufacturer could refresh the vehicle with relatively more minor technology changes. NHTSA discusses how the agency's modeling captures vehicle refreshes and redesigns in more detail below, but the details are easier to understand if the agency starts by discussing some straightforward yet important concepts. First, most changes to a vehicle's engine happen when the vehicle is redesigned and not refreshed, as incorporating a new engine in a vehicle is a 10- to 15-year endeavor at a cost of $750 million to $1 billion.\119\ However, manufacturers will use that same basic engine, with only minor changes, across multiple vehicle models. NHTSA models engine “inheriting” from one vehicle to another in both the Autonomie modeling and the CAFE Model. During a vehicle refresh, one vehicle may inherit an already redesigned engine from another vehicle that shares the same platform. In the Autonomie modeling, when a new vehicle adopts fuel-saving technologies that are inherited, the engine is not resized (i.e., the properties from the reference vehicle are used directly). While this may result in a small change in vehicle performance, manufacturers have consistently told NHTSA that the high costs for redesign and the increased manufacturing complexity that would result from resizing engines for small technology changes preclude them from doing so. In addition, when a manufacturer applies mass reduction technology (i.e., makes the vehicle lighter), the vehicle can use a less powerful engine because there is less weight to move. However, Autonomie will use a resized engine only at certain mass reduction application levels, as a representation of how manufacturers update their engine technologies. Again, this is intended to reflect manufacturers' comments that it would be unreasonable and unaffordable to resize powertrains for every unique combination of technologies. NHTSA has determined that the agency's rules about performance neutrality and technology inheritance result in a fleet that is essentially performance neutral.

\119\ 2015 NAS Report, at p. 256. It is likely that manufacturers have made improvements in the product lifetime and development cycles for engines since this NAS report and the report that NAS relied on, but NHTSA does not have data on how much. NHTSA believes that it is still reasonable to conclude that generating an all-new engine or transmission design with little to no carryover from the previous generation would be a notable investment.

\120\ UCS, Docket No. NHTSA-2025-0491-6027-A1, at 62-64.

\121\ UCS, Docket No. NHTSA-2025-0491-6027 A1, at 63.

\122\ CAFE Analysis Autonomie Documentation.

With respect to performance neutrality, the Union of Concerned Scientists (UCS) commented that “. . . the Autonomie modeling is a one-way ratchet, requiring that each and every parameter must have equal or better performance.” \120\ This statement is not true, as shown by the data presented in Figure 15 of its comment, which show plots of 0-60 time vs. 2-cycle mpg for several vehicle technology classes based on the Autonomie data.\121\ In that figure, there are data points both above and below the target 0-60 time indicating that the Autonomie data do not merely apply equal or better vehicle acceleration when adding technology or resizing the powertrain. The Autonomie model does not act as a one-way ratchet in terms of vehicle performance improvement but instead aims for a target to minimize the performance difference during powertrain resizing and can either undershoot or overshoot the target by small margins. This fact is seen in the Autonomie data for any of the vehicle technology classes, some 0-60 mph performance times are better than the target and some are worse than the target. As a vehicle's weight is reduced, its performance will increase if vehicle power and gearing remain the same, which is the assumption the Autonomie model runs until the model reaches a 10-percent reduction in vehicle mass, at which point it runs a resizing loop to adjust power output until vehicle performance is within the specified target and tolerance range.\122\ There will naturally be some variation in vehicle 0-60 mph performance time as technology is added between resizing events. NHTSA monitors vehicle performance fluctuation for each rule making analysis and believes that performance neutrality is being achieved and is in line with industry behavior. NHTSA's approach to performance neutrality considers

technology pathways that manufacturers could take to maintain similar vehicle attributes while assessing the cost and benefits of fuel economy focused technology.\123\ Doing so simplifies the analysis when considering various scenarios to set stringencies by reducing the variability of vehicle attributes and their perceived value to consumers. NHTSA has made no changes to its performance neutrality approach for this final rule.

\123\ Performance neutrality is discussed in more detail in Final TSD Chapter 2.3.5.

NHTSA's analysis ensures that vehicle models maintain consistent performance levels to allow NHTSA to estimate the costs and benefits of different levels of fuel economy standards more accurately. For its analysis, NHTSA wants to capture only the costs and benefits that result from NHTSA changing its CAFE standards. For example, a manufacturer may add a turbocharger to its engine without downsizing the engine and then direct all the additional engine work to additional vehicle HP instead of vehicle fuel economy improvements. If NHTSA modeled increases or decreases in performance because of fuel economy- improving technology, then that increase in performance has a monetized benefit attached to it that is not specifically due to the agency's fuel economy standards. By ensuring that the agency's vehicle modeling remains performance neutral, NHTSA can better ensure that the agency is reasonably capturing the costs and benefits due only to potential changes in the fuel economy standards.

Autonomie then adopts one single fuel-saving technology to the initial vehicle model, keeping everything else the same except for that one technology and the attributes associated with it. Once one technology is assigned to the vehicle model and the new vehicle model meets its performance metrics, the vehicle model is used as an input to the full-vehicle simulation. This means that Autonomie simulates driving the optimized vehicle models for each technology class on the test cycles NHTSA described above. As an example, the Autonomie modeling could start with 10 initial vehicle models (one for each technology class in the analysis). Those 10 initial vehicle models use a 5-speed automatic transmission (AT5). Argonne then builds 10 new vehicle models; the only difference between the 10 new vehicle models and the first set of vehicle models is that the new vehicle models have a 6-speed automatic transmission (AT6). Replacing the AT5 with an AT6 would lead either to an increase or decrease in the total weight of the vehicle because each technology class includes different assumptions about transmission weight. Argonne then ensures that the new vehicle models with the 6-speed automatic transmission meet their performance metrics. Argonne has 20 different vehicle models that can be simulated on the two-cycle tests. This process is repeated for each technology option and for each technology class. This results in 10 separate datasets, each with over 100,000 results, which include information about a vehicle model made of specific fuel economy-improving technology and the fuel economy value that the vehicle model achieved by driving its simulated test cycles.

NHTSA condenses the million-or-so datapoints from Autonomie into three datasets used in the CAFE Model. These three datasets include (1) the fuel economy value that each modeled vehicle achieved while driving the test cycles, for every technology combination in every technology class (converted into “fuel consumption,” which is the inverse of fuel economy; fuel economy is mpg and fuel consumption is gallons per mile); (2) the fuel economy value for PHEVs driving those test cycles, when those vehicles drive on gasoline only; and (3) optimized battery sizing and associated costs for each vehicle that adopts some sort of hybridized powertrain (discussed in more detail below). NHTSA then uses these datapoints to produce the technology effectiveness values in the CAFE Model.

Technology effectiveness values allow the CAFE Model to simulate how manufacturers can improve fuel economy relative to a consistent reference point by adding technology and combinations of technologies. The effectiveness values represent the simulated relative improvement of fuel economy that can be applied to a vehicle when new technology is added. These values are calculated based on comparing the achieved fuel economies simulated using the Autonomie full-vehicle models.

NHTSA adds the technology effectiveness values to the CAFE Model as inputs. When the CAFE Model runs a simulation, the effectiveness values for that vehicle's class determine how much the vehicle's fuel economy improves with the application of each technology. The CAFE Model's compliance simulation begins with actual fuel economy values derived from compliance data. As the CAFE Model adds technology, the technology effectiveness values are applied to estimate the new fuel economy value for the vehicle, and the CAFE Model runs millions of combinations of technologies on different vehicles to find the most cost-effective means of compliance for each manufacturer and fleet.

Return to the Ravine Runner F Series example, which has a starting fuel economy value of just over 26 mpg and a starting technology key “TURBOD; AT10L2; SS12V; ROLL0; AERO5; MR3.” The equivalent Autonomie vehicle model has a starting fuel economy value of just over 30.8 mpg and is represented by the technology descriptors Mid-size SUV, Perfo, Micro Hybrid, eng38, AUp10, MR3, AERO1, or ROLL0. In MY 2028, the CAFE Model determines that Generic Motors needs to redesign the Ravine Runner F Series to reach Generic Motors' new CAFE standard. The Ravine Runner F Series now has new fuel economy-improving technology, a parallel strong HEV with a turbocharged engine with the addition of cooled exhausted recirculation (TURBOE), an integrated 8-speed automatic transmission, 30-percent improvement in ROLL, 20-percent aerodynamic drag reduction, and 10-percent lighter glider (i.e., mass reduction). Its new technology key is now P2TRBE, ROLL30, AERO20, MR3. Table II-4 shows how the incremental fuel economy improvement from the Autonomie simulations is applied to the Ravine Runner F Series' starting fuel economy value.

[GRAPHIC] [TIFF OMITTED] TR30SE26.073

Note that the fuel economy values NHTSA obtains from the Autonomie modeling are based on the city and highway test cycles (i.e., the two- cycle test) described above. This is because NHTSA's analysis is based on the EPA procedures used for calculating fuel economy for CAFE compliance, which uses two-cycle testing.\124\ In 2008, EPA introduced three additional test cycles to bring fuel economy “label” values from two-cycle testing in line with the efficiency values consumers were experiencing in the real world, particularly for hybrids. This is known as 5-cycle testing. Generally, the revised 5-cycle testing values have proven to be a good approximation of what consumers will experience while driving and are significantly more representative than the previous two-cycle test values of real-world fuel economy. Though the compliance modeling uses two-cycle fuel economy values, the agency uses the “on-road” fuel economy values, which are the ratio of 5- cycle to 2-cycle testing values (i.e., the CAFE compliance values to the “label” values) \125\ to calculate the value of fuel savings to the consumer in the effects analysis. This is because the 5-cycle test fuel economy values better represent fuel savings that consumers will experience from real-world driving. FRIA Chapter 4.3.1 and Section 5.3.2 of the CAFE Model Documentation contain more information about these calculations. NHTSA's discussion of the effects analysis is presented later in this section.

\124\ 49 U.S.C. 32904(c) (EPA “shall measure fuel economy for each model and calculate average fuel economy for a manufacturer under testing and calculation procedures prescribed by the Administrator. However, except under section 32908 of this title, the 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.”).

\125\ NHTSA applied a certain percentage difference between the 2-cycle test value and 5-cycle test value to represent the gap in compliance fuel economy and real-world fuel economy. See FRIA Chapter 4.3.1 for further discussion.

In sum, NHTSA uses Autonomie to generate modeling and simulation technology effectiveness estimates. These estimates ensure that the modeling captures differences in technology effectiveness due to (1) vehicle size and performance relative to other vehicles in the analysis fleet; (2) other technologies on the vehicle or being added to the vehicle at the same time; and (3) how the vehicle is driven. The modeling approach allows the isolation of technology effects in the analysis supporting an accurate assessment and comports with the NAS 2015 recommendation to use full-vehicle modeling supported by the application of lumped improvements at the sub-model level.\126\

\126\ 2015 NAS Report, at p. 292.

In NHTSA's analysis, “technology effectiveness values” are the relative difference between the fuel economy value for one Autonomie vehicle model driving the two-cycle tests and a second Autonomie vehicle model that uses new technology driving the two-cycle tests. NHTSA adds the difference between two Autonomie-generated fuel economy values to a vehicle in the Market Data Input File's CAFE compliance fuel economy value. NHTSA then calculates the costs and benefits of different levels of fuel economy standards using the incremental improvement required to bring an analysis fleet vehicle model's fuel economy value to a level that contributes to a manufacturer's fleet meeting its CAFE standard.

In the next section, Technology Costs, NHTSA describes the process of generating costs for the Technologies Input File. d. Technology Costs

NHTSA estimates present and future costs for fuel-saving technologies by taking into consideration the type of vehicle or type of engine when technology costs vary by application. These cost estimates are based on three main inputs. First, direct manufacturing costs (DMCs), or the component and labor costs of producing and assembling the physical parts and systems, are estimated assuming high- volume production. Second, NHTSA estimates indirect costs. DMCs generally do not include the indirect costs of tools, capital equipment, financing, engineering, sales, administrative support, or return on investment (ROI). NHTSA accounts for these indirect costs via a scalar markup of DMCs, which is termed the retail price equivalent (RPE). Finally, the costs for technologies may change over time as industry streamlines design and manufacturing processes. To model this, the agency estimates potential cost improvements with cost learning. The retail cost of equipment in any future year is estimated to be equal to the product of the DMC, RPE, and cost learning. Considering the retail cost of equipment, instead of merely DMCs, allows NHTSA to account for the real-world price effects of a technology as well as market realities. Each of these technology cost components is described briefly below and in the following individual technology sections as well as in detail in Chapters 2 and 3 of the Final TSD.

DMCs are the component and assembly costs of the physical parts and systems that make up a complete vehicle. NHTSA uses agency-sponsored tear-down studies of vehicles and parts to estimate the DMCs of individual technologies in addition to independent tear-down studies, other publications,

and confidential business information (CBI). In the simplest cases, NHTSA sponsors studies to produce results that confirm or refute third- party industry estimates and to determine alignment with confidential information provided by manufacturers and suppliers. In cases where the tear-down study results differ significantly from credible independent sources, the agency scrutinizes the study assumptions and sometimes revises or updates the analysis accordingly.

Due to the variety of technologies and their applications and the cost and time required to conduct detailed tear-down analyses, NHTSA did not sponsor tear-down studies for every technology. In addition, the analysis includes some fuel-saving technologies that are pre- production or sold in very small pilot volumes, but for which appropriate data are available for the range of vehicles the agency models. For those technologies, NHTSA could not conduct a tear-down study to assess costs because the product is not yet in the marketplace for evaluation. In these cases, the agency relies upon third-party estimates and confidential information from suppliers and manufacturers; however, relying on CBI to estimate costs introduces several analytical challenges. First, the agency and the CBI source may use incongruent or incompatible baselines or reference points from which to measure costs. Second, sources may provide incomplete data or project DMCs far into the future based on overly optimistic production volumes--critical caveats that the agency must weigh carefully. Furthermore, a manufacturer's proprietary cost structure may be influenced by intellectual property rights or exclusive strategic partnerships. Because not all manufacturers can access these proprietary technologies at the same price point, replicating these costs within the CAFE Model could be difficult. Given these complexities, NHTSA spends significant resources scrutinizing all new data, particularly those concerning emerging technologies.

While costs for fuel-saving technologies reflect the best estimates available at the time of this analysis, technology cost estimates likely will change in the future as technologies are deployed, production is expanded, and nascent technologies mature. For emerging technologies, NHTSA uses the best information available at the time of the analysis and continues to update cost assumptions for any future analysis. Chapter 3 of the Final TSD discusses each category of technologies (e.g., engines, transmissions, or hybridization) and the cost estimates the agency uses for this analysis.

As discussed above, direct costs represent the costs associated with acquiring raw materials, fabricating parts, and assembling vehicles with the various technologies that manufacturers are expected to use to improve the fuel economy of their fleets. They include materials, labor, and variable energy costs required to produce and assemble the vehicle. However, direct costs do not include overhead costs required to develop and produce the vehicle, costs incurred by manufacturers or dealers to sell vehicles, or the profit manufacturers and dealers make from their investments. These items together contribute to the price consumers ultimately pay for the vehicle. Table II-5 illustrates how these components can affect retail prices. [GRAPHIC] [TIFF OMITTED] TR30SE26.074

To estimate total consumer costs (i.e., both direct and indirect costs), NHTSA multiplies a technology's DMCs by an indirect cost factor (the RPE) to represent the average price for fuel-saving technologies at retail. The RPE markup factor is based on an examination of historical financial data contained in 10-K reports filed by manufacturers with the Securities and Exchange Commission. It represents the ratio between the retail price of motor vehicles and the direct costs of all activities in which manufacturers engage.

For more than three decades, the retail price of motor vehicles has been, on average, roughly 50 percent above the direct cost expenditures of manufacturers. That is, the retail price is approximately 1.5 times the direct cost expenditures.\127\ This ratio has been consistent, averaging roughly 1.5 with minor variations from year to year over this period. At no point has the RPE markup based on 10-K reports exceeded 1.6 or fallen below 1.4, based on data from 1972-1997 and 2007.\128\ During this timeframe, the average annual increase in real direct costs was 2.5 percent, and the average annual increase in real indirect costs was also 2.5 percent. The RPE averages 1.5 across the lifetime of technologies of all ages, with a lower average in earlier years of a technology's life, and, because of learning effects on direct costs, a higher average in later years. Many automotive industry stakeholders have either endorsed the 1.5 markup or have estimated alternative RPE values. As seen in Table II-6, all estimates range between 1.4 and 2.0, and most are in the 1.4 to 1.7 range.\129\

\127\ Rogozhin, A. et al., Automobile Industry Retail Price Equivalent and Indirect Cost Multipliers, EPA-420-R-09-003, EPA: Ann Arbor, MI (2009), available at: https://nepis.epa.gov/Exe/ZyPDF.cgi/P100AGJ1.PDF?Dockey=P100AGJ1.PDF (accessed: May 28, 2026); Spinney, B. et al., Advanced Air Bag Systems Cost, Weight, and Lead Time Analysis Summary Report, National Highway Traffic Safety Administration: Washington, DC (1999).

\128\ Data are not available for intervening years, but results for 2007 seem to indicate no significant change in the historical trend.

\129\ See The Alliance, Docket No. EPA-HQ-OAR-2018-0283-6186, at 143 (Oct. 26, 2018) (“The Alliance supports the use of retail price equivalents in the compliance cost modeling”). [GRAPHIC] [TIFF OMITTED] TR30SE26.075

An RPE of 1.5 does not mean that manufacturers automatically mark up each vehicle by exactly 50 percent. Rather, it means that, over time, the competitive marketplace has resulted in pricing structures that average out to this relationship across the entire industry. Prices for any individual model may be marked up at a higher or lower rate depending on market demand. On average, over time and across the vehicle fleet, consumers spend about $1.50 for each dollar of direct costs incurred by manufacturers. Based on NHTSA's own evaluation and the widespread use and acceptance of the RPE by automotive industry stakeholders, the agency has determined that the RPE provides a reasonable indirect cost markup for use in the analysis. A detailed discussion of indirect cost methods and the basis for the agency's use of the RPE to reflect these costs, rather than other indirect cost markup methods, is available in the FRIA for the 2020 final rule.\131\

\130\ Duleep, K., Analysis of Technology Cost and Retail Price, Presentation to Committee on Assessment of Technologies for Improving LDV Fuel Economy, Detroit, MI (2008); Jack Faucett Associates, Update of EPA's Motor Vehicle Emission Control Equipment Retail Price Equivalent (RPE) Calculation Formula, Report No. 68-03- 3244, EPA: Ann Arbor, MI (1985), available at: https://nepis.epa.gov/Exe/ZyPURL.cgi?Dockey=940047LI.txt (accessed: May 28, 2026); McKinsey & Company, New Horizons: Multinational Company Investment in Developing Economies, Final version, Mckinsey Global Institute: San Francisco, CA, Preface to the Auto Sector Cases, p. 1 (2003), available at: https://www.mckinsey.com/~/media/McKinsey/ Business%20Functions/McKinsey%20Digital/Our%20Insights/ New%20horizons%20for%20multinational%20company%20investment/ MGI_Multinational_company_investment_in_developing_economies_Full_Rep ort.ashx (accessed: Apr. 6, 2026); Transportation Research Board and National Research Council, Effectiveness and Impact of Corporate Average Fuel Economy (CAFE) Standards, National Academies Press: Washington, DC, pp. 5, 12 (2002), available at: https://nap.nationalacademies.org/catalog/10172/effectiveness-and-impact-of-corporate-average-fuel-economy-cafe-standards (accessed: May 28, 2026); National Research Council, Assessment of Fuel Economy Technologies for Light-Duty Vehicles, National Academies Press: Washington, DC (2011), available at: https://nap.nationalacademies.org/catalog/12924/assessment-of-fuel-economy-technologies-for-light-duty-vehicles (accessed: May 28, 2026); National Research Council, Cost, Effectiveness, and Deployment of Fuel Economy Technologies in LDVs, National Academies Press: Washington, DC (2015); Sierra Research, Inc., Study of Industry- Average Mark-Up Factors Used to Estimate Changes in Retail Price Equivalent (RPE) for Automotive Fuel Economy and Emissions Control Systems, Sierra Research, Inc.: Sacramento, CA (2007); Vyas, A. et al., Comparison of Indirect Cost Multipliers for Vehicle Manufacturing, Center for Transportation Research: Argonne, IL (2000), available at: https://publications.anl.gov/anlpubs/2000/05/36074.pdf (accessed: May 28, 2026).

\131\ NHTSA and EPA, FRIA: The Safer Affordable Fuel-Efficient (SAFE) Vehicles Rule for Model Year 2021-2026 Passenger Cars and Light Trucks (2020), available at: https://www.nhtsa.gov/sites/nhtsa.gov/files/documents/final_safe_fria_web_version_200701.pdf (accessed: May 28, 2026).

IPI criticized the established application of RPE in NHTSA's analysis. IPI asserted that “NHTSA must update its indirect cost factor to reflect recent

data and to remove any transfer effects.” \132\ NHTSA disagrees with IPI's assertion and continues to use the best available information for RPE (indirect cost factor). NHTSA justifies its choice of an indirect cost factor as a central estimate of a long run markup factor. IPI provided neither evidence that the long-stable indirect cost factor (RPE) has changed substantively since the most recent comprehensive analysis nor examples of estimates of indirect cost factors or RPEs that are more recent than those cited in the proposal.\133\ IPI based its assertion on a structural model of implied firm markups, which is a categorically different measure than indirect cost factors or RPE \134\ (IPI likely conflated the two concepts because the term “markup” is used to describe each of these distinct measures). IPI also argued in its comment that NHTSA cannot consider vehicle price changes resulting from regulatory action due to OMB guidance in Circular A-4, but it misunderstands that document, since here the purpose of NHTSA's analysis is to measure the potential impacts of a regulatory action.

\132\ IPI, Docket No. NHTSA-2025-0491-6015, at 3.

\133\ National Research Council, Cost, Effectiveness, and Deployment of Fuel Economy Technologies for Light-Duty Vehicles, The National Academies Press: Washington, DC (2015), available at: https://nap.nationalacademies.org/catalog/21744/cost-effectiveness-and-deployment-of-fuel-economytechnologies-for-light-duty-vehicles (accessed: May 25, 2026).

\134\ Grieco, P. et al., The evolution of market power in the U.S. automobile industry, The Quarterly Journal of Economics, Vol. 139(2): pp. 1201-53 (2024), available at: https://doi.org/10.1093/qje/qjad047 (accessed: May 28, 2026).

IPI claimed that NHTSA is irrationally counting RPE as a cost when IPI believes that RPE represents a transfer. IPI asserted that original equipment manufacturers (OEMs) have pricing power and then incorrectly characterizes RPE as representing “monopoly profits.” IPI also argued that NHTSA cannot account for the impact of changing new vehicle prices on consumer welfare due to guidance in Circular A-4.

In response to IPI's comment, NHTSA notes that Chapter 2.4 in the Final TSD provides NHTSA's rationale for including indirect costs in technology costs. IPI's representation of indirect costs (and as a result technology costs associated with standard setting) conflates manufacturer profits with impacts on vehicle prices. RPE represents indirect costs such as the costs of tools, capital equipment, financing, engineering, sales, administrative support, or ROI associated with the application of new vehicle technology. The RPE assumes that, over time, the competitive marketplace results in pricing structures that average out to this relationship across the automotive industry.\135\

\135\ Final TSD Chapter 2.4.2.

In addition, Circular A-4 provides instruction and guidance to agencies in developing assessments of the costs and benefits of regulatory actions and alternatives to those actions pursuant to Section 6(a)(3)(C) of Executive Order 12866. Here, NHTSA's standard- setting analysis is prescribed by law to consider the four statutory factors set forth in EPCA: technological feasibility, economic practicability, the effect of other motor vehicle standards of the Government on fuel economy, and the need of the United States to conserve energy.\136\ NHTSA's analysis of technology costs and RPE are consistent with this statutory directive and previous rulemakings. Measuring the technology costs paid by consumers through higher new vehicle prices provides a transparent representation of standards' effects on consumers.

\136\ 49 U.S.C. 32902(f).

Finally, manufacturers make improvements to production processes over time, which often result in lower costs. “Cost learning” reflects the effect of experience and volume on the cost of production, which results generally in better utilization of resources, leading to higher and more efficient production. As manufacturers gain experience through production, they refine production techniques, raw material and component sources, and assembly methods to maximize efficiency and reduce production costs.

NHTSA estimates cost learning by considering methods established by T.P. Wright and later expanded upon by J.R. Crawford. Wright, examining aircraft production, found that every doubling of cumulative production of airplanes resulted in decreasing labor hours at a fixed percentage. This fixed percentage is commonly referred to as the progress rate or progress ratio, where a lower rate implies faster learning as cumulative production increases. J.R. Crawford expanded upon Wright's learning curve theory to develop a single unit cost model, which estimates the cost of the nth unit produced where the following information is known: (1) cost to produce the first unit; (2) cumulative production of n units; and (3) the progress ratio.

Consistent with Wright's learning curve, NHTSA uses the basic approach by Wright for most technologies in the CAFE Model, with NHTSA estimating technology cost reductions by applying a fixed percentage to the projected cumulative production of a given fuel economy technology in a given model year.\137\ The agency estimates the cost to produce the first unit of any given technology by identifying the DMC for a technology in a specific model year. As discussed in detail below, and in Chapter 3 of the Final TSD, NHTSA's technology DMCs come from studies, teardown reports, other publicly available data, and feedback from manufacturers and suppliers. Because different studies or cost estimates are based on costs in specific model years, the agency identifies the “base” model years for each technology where the learning factor is equal to 1.00. Then, the agency applies a progress ratio to back-calculate the cost of the first unit produced. The majority of technologies in the CAFE Model use a progress ratio (i.e., the slope of the learning curve, or the rate at which cost reductions occur with respect to cumulative production) of approximately 0.89, which is derived from average progress ratios researched in studies funded or identified by NHTSA.\138\ Many fuel economy technologies that have existed in vehicles for some time will have a gradual sloping learning curve implying that cost reductions from learning is moderate and eventually becomes less steep toward MY 2050. Conversely, newer technologies have a steeper learning curve initially, where cost

reduction occurs at a high rate. Mature technologies generally have a flatter curve and may not incur much cost reduction, if at all, from learning. Final TSD Chapter 2.4.4 provides an illustration showing various slopes of learning curves.

\137\ NHTSA uses statically projected cumulative volume production estimates because the CAFE Model does not support dynamic projections of cumulative volume at this time.

\138\ Simons, J., Cost and Weight Added by the Federal Motor Vehicle Safety Standards for MY 1968-2012 Passenger Cars and LTVs, Report No. DOT HS 812 354, NHTSA: Washington D.C., pp. 30-33 (2017), available at: https://downloads.regulations.gov/NHTSA-2021-0053-1643/attachment_44.pdf (accessed: May 28, 2026); Argote, L. et al., The Acquisition and Depreciation of Knowledge in a Manufacturing Organization--Turnover and Plant Productivity, Working Paper, Graduate School of Industrial Administration, Carnegie Mellon University (1997); Benkard, C., Learning and forgetting: the dynamics of aircraft production, The American Economic Review, Vol. 90(4): pp. 1034-54 (2000), available at: https://www.aeaweb.org/articles?id=10.1257/aer.90.4.1034 (accessed: May 28, 2026); Epple, D. et al., Organizational learning curves: a method for investigating intra-plant transfer of knowledge acquired through learning by doing, Organization Science, Vol. 2(1): pp. 58-70 (1991), available at: https://www.jstor.org/stable/2634939 (accessed: May 28, 2026); Epple, D. et al., An empirical investigation of the microstructure of knowledge acquisition and transfer through learning by doing, Operations Research, Vol. 44(1): pp. 77-86 (1996), available at: https://ideas.repec.org/a/inm/oropre/v44y1996i1p77-86.html (accessed: May 28, 2026); Levitt, S. et al., Toward an understanding of learning by doing: evidence from an automobile assembly plant, Journal of Political Economy, Vol. 121(4): pp. 643-81 (2013), available at: https://www.nber.org/papers/w18017 (accessed: May 28, 2026).

The agency assigns groups of similar technologies or technologies of similar complexity to each learning curve. Though the grouped technologies differ in operating characteristics and design, NHTSA chooses to group them based on market availability, complexity of technology integration, and production volume of the technologies that can be implemented by manufacturers and suppliers. In general, the agency considers most basic engine and transmission technologies to be mature technologies that do not experience any additional improvements in design or manufacturing. Other basic engine technologies, like variable valve lift (VVL), stoichiometric gasoline direct injection (SGDI), and DEAC, decrease in costs through around MY 2036, because those were introduced into the market more recently. All advanced engine technologies follow the same general pattern of a gradual reduction in costs until MY 2036, when they plateau and remain flat. NHTSA expects the cost to decrease as production volumes increase, manufacturing processes are improved, and economies of scale are achieved. The agency has assigned advanced engine technologies based on a singular preceding technology to the same learning curve as that preceding technology. Similarly, the more advanced transmission technologies experience a gradual reduction in costs through MY 2031, when they plateau and remain flat. Lastly, the agency estimates that the learning curves for road load technologies, with the exception of the most advanced mass reduction level (which decreases at a fairly steep rate through MY 2040, as discussed further below and in Chapter 3.4 of the Final TSD), will decrease through MY 2036 and then remain flat.

For technologies that have been in production for many years, like some engine and transmission technologies, this approach produces reasonable estimates that NHTSA can compare against other studies and publicly available data. Generating the learning curve for battery packs for hybrid vehicles in future model years is significantly more complicated, and NHTSA discusses how the agency generated those learning curves in detail in Chapter 3.3 of the Final TSD. NHTSA's battery pack learning curves recognize that there are many factors that could potentially lower battery pack costs over time outside of cost reductions from improvements in manufacturing processes due to knowledge gained through experience in production.

Table II-7 shows how some of the technologies on the example MY 2024 Ravine Runner F Series decrease in cost over several years. Note that these costs are specifically applicable to the MedSUVPerf class, and other technology classes may have different costs for the same technologies. These costs are pulled directly from the Technologies Input File, meaning that they include the DMC, RPE, and learning. [GRAPHIC] [TIFF OMITTED] TR30SE26.076

The Inflation Reduction Act (IRA) included several tax credits intended to encourage the adoption of clean vehicles.\139\ OB3 amended these credits through a combination of stricter eligibility requirements and earlier phase out of certain credits.\140\ Consistent with prior rulemakings, NHTSA assumes that non-plug-in hybrids do not qualify for the tax credits because their battery size is below the minimum thresholds set within the credit provisions. As noted throughout this preamble, NHTSA is statutorily prohibited from considering the fuel economy of dedicated automobiles and therefore has excluded dedicated vehicles from the analysis. The agency considers the fuel-based efficiency of dual-fueled vehicles, such as PHEVs, which are the only vehicles the agency models that are eligible for tax credits.

\139\ Public Law 117-169, 136 Stat. 1818 (Aug. 16, 2025).

\140\ Enacted as Public Law 119-21, 139 Stat. 72 (July 4, 2025).

NHTSA models three provisions of the IRA--the Advanced manufacturing production credit (“AMPC” or “45X”), the Clean vehicle credit (“30D”), and the Credit for qualified commercial clean vehicles (“45W”). The AMPC, which provides a $35 per kWh tax credit for manufacturers of battery cells and an additional $10 per kWh for manufacturers of battery modules (all applicable to manufacture in the United States), is modeled through its phase out in 2032 in this final rule.\141\ The 30D and 45W credits (collectively, the Clean Vehicle Credits or “CVCs”) are modeled through their sunset, which for modeling purposes is assumed to be MY 2025.142 143

\141\ 26 U.S.C. 45X. If a manufacturer produces a battery module without battery cells, it is eligible to claim up to $45 per kWh for the battery module. Two other provisions of the AMPC are not modeled at this time; (1) a credit equal to 10 percent of the manufacturing cost of electrode active materials and (2) a credit equal to 10 percent of the manufacturing cost of critical minerals for battery production. NHTSA is not modeling these credits directly because of how battery costs are estimated, and to avoid the potential to double-count the tax credits if they are included into other analyses that feed into NHTSA's inputs. For a full account of the credit and any limitations, please refer to the statutory text.

\142\ 26 U.S.C. 45W. For a full account of the credit and any limitations, please refer to the statutory text.

\143\ 26 U.S.C. 30D. For a full account of the credit and any limitations, please refer to the statutory text.

The 30D credit provided up to $7,500 toward the purchase of new clean vehicles with critical minerals either extracted or processed in the United States or a country with which the United States has a free trade agreement or recycled in North America and battery components manufactured or assembled in North America.\144\ In

contrast to 30D, the 45W credit did not have the same critical minerals and production restraints, but instead the credit value is the lesser of the incremental cost to purchase a comparable ICE vehicle or 15 percent of the cost basis for PHEVs up to $7,500 for vehicles with a gross weight vehicle rating (GVWR) less than 14,000 pounds. The Department of the Treasury set safe harbors for claiming credits based on DOE's Incremental Purchase Cost Methodology and Results for Clean Vehicles report.\145\ The safe harbors were set at the full amount of $7,500 for most vehicles except for smaller PHEV sedans, which were capped just shy of the maximum value.\146\ Given the relatively small difference between the full credit value and the cap for smaller PHEV sedans, and the few vehicle lines categorized as smaller sedans, the agency assumed that all vehicles could qualify for the full credit as a modeling simplification. The 45W credit was also available only to commercial purchasers; however, the Department of the Treasury determined that leased vehicles could be eligible for the 45W credit where the financing company was the owner of the vehicle for Federal income tax purposes. NHTSA jointly models the CVC tax credits. Both credits are available at the time of sale and provided up to $7,500 towards the purchase of light-duty vehicles acquired on or before September 30, 2025. Because only one of the CVCs could be claimed for each vehicle purchase, NHTSA models them jointly.

\144\ Vehicle price and consumer income limitations apply to section 30D credits, as well. See Congressional Research Service, Tax Provisions in the Inflation Reduction Act of 2022 (H.R. 5376) (2022), available at: https://www.congress.gov/crs-product/R47202 (accessed: May 28, 2026).

\145\ See Internal Revenue Service, Frequently Asked Questions Related to New, Previously-Owned and Qualified Commercial Clean Vehicle Credits, FS-2022-42, Media Relations Office: Washington, DC, Q4 and Q8 (2022), available at: https://www.irs.gov/pub/taxpros/fs-2022-42.pdf (accessed: May 28, 2026).

\146\ See Internal Revenue Service, Section 45W Commercial Clean Vehicles and Incremental Cost for 2023, IRS Notice 2023-9 (2023), available at: https://www.irs.gov/pub/irs-drop/n-23-09.pdf (accessed: May 28, 2026); Internal Revenue Service, Section 45W Commercial Clean Vehicles and Incremental Cost for 2024, IRS Notice 2024-5 (2024), available at: https://www.irs.gov/pub/irs-drop/n-24-05.pdf (accessed: May 28, 2026); Internal Revenue Service, Section 45W Credit for Qualified Commercial Clean Vehicles and Incremental Cost for 2025, IRS Notice 2025-9 (2025), available at: https://www.irs.gov/pub/irs-drop/n-25-09.pdf (accessed: May 28, 2026).

The CAFE Model projects vehicles in model year cohorts rather than on a calendar year basis. Given that model years and calendar years can be misaligned (e.g., a MY 2024 vehicle could be sold in CYs 2023, 2024, or even 2025), choosing which calendar year a model year falls into is important for assigning tax credits that are phased out during the analytical period. NHTSA analyzed the timing of new vehicle sales and new vehicle registrations and determined that, for this final rule, it is appropriate to assume that credits available in a given calendar year are available to all vehicles sold in the same model year.

In the NPRM, the agency elected not to model the AMPC past MY 2025 because of the more stringent requirements imposed by OB3 restricting usage of materials from prohibited foreign entities (PFE) (i.e., constrained eligibility for the tax credit based on materials sources) and uncertainty about whether manufacturers could meet the non-PFE component threshold percentages. NHTSA conducted a sensitivity analysis for the NPRM where the AMPC extended through its statutory sunset in MY 2032.

Attorneys General from Multiple States (Attorneys General) argued that NHTSA failed to include the cost-reducing effects of the 45X credit to inflate artificially the price of hybrid vehicles, and that NHTSA incorrectly assumed that automakers would not meet the required domestic component thresholds and foreign entity constraints.\147\ The Attorneys General countered that significant recent onshoring means the industry can take advantage of the 45X credit. BlueGreen Alliance (BGA) commented that while Congress recently eliminated other EV incentives, it intentionally preserved the 45X credit.\148\ They argue that maintaining strong CAFE standards works hand-in-hand with the 45X credit and President Trump's executive actions to support onshore domestic battery manufacturing and secure America's critical mineral supply chain. The Alliance acknowledged the sunsetting of the 30D and 45W tax credits and noted that 45X eligibility was also tightened.\149\ The agency took these comments into consideration and updated the 45X phase out years for the final rule.

\147\ Attorneys General, Docket No. NHTSA-2025-0491-6064-A2, at 62.

\148\ BGA, Docket No. NHTSA-2025-0491-5931-A1, at 2-3.

\149\ The Alliance, Docket No. NHTSA-2025-0491-5707-A1, at 3.

The agency assumes that manufacturers and consumers each capture half of the dollar value of the AMPC and CVCs. The agency assumes that manufacturers' shares of both credits will offset part of the cost to supply models eligible for the credits--PHEVs, specifically. The subsidies reduce the costs of eligible vehicles and increase their attractiveness to buyers. Because the AMPC credit scales with battery capacity, NHTSA determines battery energy capacity separately for each vehicle based on Argonne simulation outputs. Final TSD Chapter 2.3.2 contains a detailed discussion of these assumptions. NHTSA accounts for all the eligibility requirements of 30D and the AMPC, such as the location of final assembly and battery production, the origin of critical minerals, and the income restrictions of 30D through the credit schedules constructed in part based off of these factors and allows all PHEVs produced and sold during the timeframe that tax credits are offered to be eligible for those credits subject to the MSRP restrictions discussed below.\150\

\150\ See 88 FR 56179 (Aug. 17, 2023) for a more detailed explanation of the process used for the previous proposal.

To account for the agency's inability to model dynamically sourcing requirements and income limits for 30D, NHTSA uses projected values of the average value of 30D and the AMPC for the final rule. The projections increase throughout the analysis due to the expectation that gradual improvements in supply chains over time would allow more vehicles to qualify for the credits.

NHTSA uses a DOE report that provides combined values of the CVCs.\151\ These values consider the latest information of PHEV penetration rates, PHEV retail prices, the share of United States PHEV sales that meet the critical minerals and battery component requirements, the share of vehicles that exclude suppliers that are “Foreign Entities of Concern,” and lease rates for vehicles that qualify for the 45W CVC. The DOE projections are the most detailed and rigorous projections of credit availability that NHTSA is aware of at this time, and DOE has not released updated projections that reflect the enactment of OB3. Final TSD Chapter 2.5.3 includes more information on the average AMPC credit per kWh that NHTSA uses in this final rule.

\151\ U.S. Department of Energy, Estimating Federal Tax Incentives for Heavy Duty Electric Vehicle Infrastructure and for Acquiring Electric Vehicles Weighing Less Than 14,000 Pounds, Memorandum (2024), available at: https://downloads.regulations.gov/EERE-2021-VT-0033-0056/content.pdf (May 28, 2026).

The CAFE Model accounts for the statutory MSRP restrictions of 30D by assuming that the CVCs cannot be applied to cars with an MSRP above $55,000 or other vehicles with an MSRP above $80,000, which are ineligible for 30D. The 45W credit does not have the same MSRP restrictions; however, because NHTSA is unable to model the CVCs separately at this time, the agency has to choose whether to model the restriction for both CVCs or not to

model the restriction at all. NHTSA chooses to include the restriction for both CVCs to be conservative.\152\ Chapter 2.5.2 of the Final TSD contains additional details on how NHTSA implements tax credits.

\152\ Bureau of Transportation Statistics, New and Used Passenger Car and Light Truck Sales and Leases, last revised: 2025, available at: https://www.bts.gov/content/new-and-used-passenger-car-sales-and-leases-thousands-vehicles (accessed: May 28, 2026).

NHTSA uses real dollars for future costs and benefits, such as technology costs in future model years. Including the tax credits as nominal dollars instead of real dollars artificially raises the value of the credits in respect to other costs, so NHTSA converts the DOE projections to real dollars.

NHTSA does not model individual State tax credits or rebate programs. State clean vehicle tax credits and rebates vary from jurisdiction to jurisdiction and are subject to more uncertainty in funding availability and eligibility than their Federal counterparts.\153\ Tracking sales by jurisdiction and modeling each program's individual compliance program would require significant revisions to the CAFE Model while likely producing minimal changes in the net outputs of the analysis given constraints imposed by statute. NHTSA has not changed this approach for the final rule but will continue to monitor State programs for appropriateness for consideration in future rulemakings.

\153\ States have additional mechanisms to amend or remove tax incentives or rebates. Sometimes, even after these programs are enacted, uncertainty persists. See Farah, N., The Untimely Death of America's “Most Equitable” EV Rebate, last revised: Jan. 30, 2023, available at: https://www.eenews.net/articles/the-untimely-death-of-americas-most-equitable-ev-rebate/ (accessed: May 28, 2026).

As NHTSA describes above, the CAFE Model simulates cost-effective ways that vehicle manufacturers could comply with CAFE standards, subject to limits that ensure that the Model reasonably replicates manufacturers' decisions in the real world. This section describes the equations the CAFE Model uses to determine how to apply technology to vehicles, including whether technologies are cost effective, and why the agency believes the CAFE Model's calculation of potential compliance pathways reasonably represents manufacturers' decision- making. This section also gives a high-level overview of real-world limitations that vehicle manufacturers face when designing and manufacturing vehicles and how the agency includes those in the technology inputs and assumptions in the analysis.

For each manufacturer's fleet, the CAFE Model first determines whether any technology should be “inherited” from an engine, transmission, or platform that currently uses the technology and should be applied to a vehicle that is due for a refresh or redesign. NHTSA describes above how vehicle manufacturers use the same or similar engines, transmissions, and platforms across multiple vehicle models, and the agency tracks vehicle models that share technology by assigning Engine, Transmission, and Platform Codes to vehicles in the analysis fleet. As an example, variants of the Ford 10R80 10-speed transmission are currently used in the following Ford Motor Company vehicles: 2017- present Ford F-150, 2018-present Ford Mustang, 2018-present Ford Expedition/Lincoln Navigator, 2019-present Ford Ranger, and the 2020- present Ford Explorer/Lincoln Aviator. The 2WD variant of the 10R80, as applied to the CAFE Model, is shared by the 2WD Expedition models, 2WD F-150 models, and the Mustang, thus linking these models by the same Transmission Code. If one of these three vehicle model types receives a transmission upgrade, the other two would automatically receive the same upgrade at their next redesign or refresh.

After applying inherited technologies, the Model begins the process of evaluating what additional technologies could be applied to the manufacturer's vehicles. The CAFE Model applies the most cost-effective technology out of the universe of technology options that the Model could potentially apply. To determine whether a particular technology is cost effective, the Model calculates the “effective cost” of multiple technology options and chooses the option that results in the lowest “effective cost.” A technology that has an effective cost less than zero (Equation II-4 results in a negative number) is considered cost effective, as a negative effective cost implies that the technology “pays for itself.” The “effective cost” calculation is actually multiple calculations, but this section describes only the highest levels of that logic; interested readers can consult the CAFE Model Documentation for additional information on the calculation of effective cost. Equation II-4 shows the CAFE Model's effective cost calculation for this analysis. Equation II[dash]4: CAFE Model Effective Cost Calculation [GRAPHIC] [TIFF OMITTED] TR30SE26.077

Where:

TechCostTotal: the total cost of a candidate technology evaluated on a group of selected vehicles; TaxCreditsTotal: the cumulative value, if any, of additional vehicle and battery tax credits (or Federal incentives) resulting from application of a candidate technology evaluated on a group of selected vehicles; FuelSavingsTotal: the value of the reduction in fuel consumption (or fuel savings) resulting from application of a candidate technology evaluated on a group of selected vehicles; [Delta]Fines: the change in manufacturer's fines in the analysis year, if applicable; [Delta]ComplianceCredits: the change in manufacturer's CAFE compliance value in the analysis year (denominated in thousands of gallons); EffCost: the calculated effective cost attributed to application of a candidate technology evaluated on a group of selected vehicles.

The components of this “cost per credit” effective cost calculation are described further here. The CAFE Model considers the total cost of a technology (TechCost) that could be applied to a group of connected vehicles, just as a vehicle manufacturer might consider what new technologies it has ready for the market and which vehicles should and could receive the upgrade. Next, like the technology costs, the CAFE Model calculates the total value of Federal incentives (TaxCredits) available for a technology that could be applied to a group of vehicles and subtracts that total incentive from the total technology costs. The total fuel cost savings (FuelSavings) are the savings in fuel expense resulting from adding additional technology or switching from one technology to another. For this, the CAFE Model must calculate the total fuel cost for the vehicle before application of a

technology and subtract the total fuel cost for the vehicle after calculation of that technology. The total fuel cost for a given vehicle depends on both the price of gas (or gasoline equivalent fuel) and the number of miles that a vehicle is driven during the initial years of ownership, among other factors.\154\ As technology is applied to vehicles in groups, the fuel cost savings for the vehicle is then multiplied by the sales volume of a vehicle in a model year to equal total fuel cost savings, which is then subtracted in the numerator of the effective cost equation. Finally, in the numerator, the agency subtracts the change in a manufacturer's expected fines ([Delta]Fines), which are set at $0 for this analysis as a result of Public Law 119-21, before and after application of a specific technology, if any.\155\ This approach can be thought of as subtraction of the fines avoided by upgrading to a certain technology. Then, the result from the sequence above is divided by the change in compliance credits ([Delta]ComplianceCredits), which means the difference in a manufacturer's calculated fuel economy value in a compliance category before and after the application of a technology to a group of vehicles. This approach can be thought of as dividing the result by the gain in fuel economy performance resulting from upgrading to a certain technology.

\154\ This fuel cost savings is calculated using the miles driven over 3 years, based on the assumption that consumers are likely to buy vehicles with fuel economy-improving technology that pays for itself within 3 years.

\155\ See Section VI noting the value of civil penalties are set to $0 in this analysis.

After inherited technologies and cost-effective (effective cost is less than zero) technologies are applied, the CAFE Model determines whether the manufacturer's fleet meets its CAFE standard. If the manufacturer is still not in compliance, the Model applies non-cost- effective technologies (which have an effective cost greater than zero, or which do not “pay for themselves”) until it runs out of technology options.

The Model runs the compliance simulation successively and accounts for technology added during each previous model year by carrying forward technologies between model years once they are applied. The CAFE Model does this by mirroring real-world decisions of manufacturers to carry forward most technologies between model years, concentrating the application of new technology to vehicle redesigns or mid-cycle “freshenings,” and design cycles vary widely among manufacturers and specific products. Comments from manufacturers and Model peer reviewers for past CAFE rules have strongly supported explicit year-by-year simulation. In addition, the multi-year planning capability increases the Model's ability to simulate manufacturers' real-world behavior, accounting for the fact that manufacturers will seek out compliance paths for several model years at a time, while accommodating the year- by-year requirement.

In addition to the Model's technology application decisions pursuant to the compliance simulation algorithm, several technology inputs and assumptions work together to determine which technologies the CAFE Model can apply. The technology pathways, discussed in detail above, are one significant way that the agency instructs the CAFE Model to apply technology. The pathways define mutually exclusive technologies (i.e., those that cannot be applied at the same time) and define the direction in which vehicles can advance as the modeling system evaluates specific technologies for application. Then, the arrows between technologies instruct the Model on the order in which to evaluate technologies on a pathway, to ensure that a vehicle that uses a more fuel-efficient technology cannot downgrade to a less efficient option.

In addition to technology pathway logic, NHTSA uses several technology applicability rules to replicate better manufacturers' decision-making. The “skip” input--represented in the Market Data Input File as “SKIP” in the appropriate technology column corresponding to a specific vehicle model--is particularly important for accurately representing how a manufacturer applies technologies to their vehicles in the real world. This tells the Model not to apply a specific technology to a specific vehicle model. SKIP inputs are used to simulate manufacturer decisions, including: (1) parts and process sharing; (2) stranded capital; and (3) performance neutrality.

First, parts sharing includes the concepts of platform, engine, and transmission sharing, which are discussed in Section II.C.2.b. A “platform” refers to engineered underpinnings shared on several differentiated vehicle models and configurations. Manufacturers share and standardize components, systems, tooling, and assembly processes within their products (and occasionally with the products of another manufacturer) to manage complexity and costs for development, manufacturing, and assembly. Detailed discussion for this type of SKIP is provided in the “adoption features” section for different technologies, if applicable, in Chapter 3 of the Final TSD.

Similar to vehicle platforms, manufacturers create engines that share parts. For instance, manufacturers may use different piston strokes on a common engine block or bore out common engine block castings with different diameters to create engines with an array of displacements. Head assemblies for different displacement engines may share many components and manufacturing processes across the engine family. Manufacturers may finish crankshafts with the same tools to similar tolerances. Engines on the same architecture may share pistons and connecting rods, and the same engine architecture may include both 6- and 8-cylinder engines. One engine family may appear on many vehicles on a platform, and changes to that engine may or may not carry through to all the vehicles. Some engines are shared across a range of different vehicle platforms. Vehicle model/configurations in the analysis fleet that share engines belonging to the same platform are identified as such, and the agency also may apply a SKIP to a particular engine technology where it is known that a manufacturer shares an engine throughout several of their vehicle models and the engine technology is not appropriate for any of the platforms that share the same engine.

It is important to note that manufacturers can define a “common” engine platform in different ways. Some manufacturers consider engines as “common” if the engines share an architecture, components, or manufacturing processes. Other manufacturers take a narrower approach and consider engines “common” only if the parts in the engine assembly are the same. In some cases, manufacturers designate each engine in each application as a unique powertrain. For example, a manufacturer may have listed two engines separately for a pair that share designs for the engine block, the crankshaft, and the head because the accessory drive components, oil pans, and engine calibrations differ between the two. In practice, many engines share parts, tooling, and assembly resources, and manufacturers often coordinate design updates between two similar engines. NHTSA considers engines to be on a common platform (for purposes of coding, discussed in Section II.C.2 above, and for SKIP application) if the engines share a common cylinder count and configuration, displacement, valvetrain, and fuel type, or if the engines only differ slightly in

compression ratio (CR), HP, and displacement.

Parts sharing also includes the concept of sharing manufacturing lines (the systems, tooling, and assembly processes discussed above), because manufacturers are unlikely to build a new manufacturing line to build a completely new engine. A new engine designed to be mass manufactured on an existing production line has limits in number of parts used, type of parts used, weight, and packaging size due to the weight limits of the pallets, material handling interaction points, and conveyance line design to produce one unit of a product. The restrictions are reflected in the usage of a SKIP of engine technology that the manufacturing line would not accommodate.

SKIPs also relate to instances of stranded capital when manufacturers amortize research, development, and tooling expenses over many years, especially for engines and transmissions. The traditional production life cycles for transmissions and engines have been a decade or longer. If a manufacturer launches or updates a product with fuel- saving technology, and then later replaces that technology with unrelated or different fuel-saving technology before the equipment and research and development investments have been fully paid off, there will be unrecouped, or stranded, capital costs. Quantifying stranded capital costs accounts for such lost investments. One design where manufacturers take an iterative redesign approach, as described in a recent SAE paper,\156\ is the MacPherson strut suspension. It is a popular low-cost suspension design, and manufacturers use it across their fleets. As the agency observed previously, manufacturers may be shifting their investment strategies in ways that may alter how stranded capital could be considered. For example, some suppliers sell similar transmissions to multiple manufacturers. Such arrangements allow manufacturers to share in capital expenditures or amortize expenses more quickly. Manufacturers share parts on vehicles around the globe, achieving greater scale and greatly affecting tooling strategies and costs.

\156\ Pilla, S. et al., Parametric Design Study of McPherson Strut to Stabilizer Bar Link Bracket Weld Fatigue Using Design for Six Sigma and Taguchi Approach, SAE Technical Paper 2021-01-0235, SAE International (2021), available at: https://doi.org/10.4271/2021-01-0235 (accessed: May 28, 2026).

As a proxy for stranded capital, the CAFE Model accounts for platform and engine sharing and includes redesign and refresh cycles for significant and less significant vehicle updates. This analysis continues to rely on the CAFE Model's explicit year-by-year accounting for estimated refresh and redesign cycles, and shared vehicle platforms and engines, to moderate the cadence of technology adoption and thereby limit the implied occurrence of stranded capital and the need to account for it explicitly. In addition, limiting the specific advanced technology pathways manufacturers may pursue for certain models through technology adoption prevents additional capital being dedicated towards technologies that may be quickly replaced and therefore minimizes the amount of stranded capital indirectly. Adoption features specific to each technology are discussed in each technology section.