Coordinating the electric vehicle transition and electricity grid decarbonization in the U.S. is not essential to achieving substantial long-term carbon dioxide emissions reductions

How quickly the US can decarbonize light-duty vehicle (LDV) transportation depends on the rates of change of electric vehicle (EV) sales, stock turnover, and grid decarbonization. We build a stock turnover model to assess how sensitive achieving 2050 LDV decarbonization targets is to these rates. We estimate carbon dioxide (CO2) reductions of 70%–85% by 2050, including emissions from vehicles and upstream electricity generation, provided that new vehicle sales transition to 100% EVs and substantial grid decarbonization are accomplished by 2050. This result is robust to continuation of long-term trends of increasing vehicle longevity, and to whether the timing of EV sales growth and grid decarbonization are coordinated. If the two key goals are met, the annual contribution of EV electricity use to CO2 emissions will be small over the entire period.


Introduction
To address the growing threat of climate change, the United States has established a goal for reducing its own greenhouse gas (GHG) emissions over the current decade.In 2021, the Biden administration announced a goal of reducing economy-wide net emissions by 43% in 2030 relative to 1990 levels (White House 2021a).This goal shares a similar ambition to the goal in the most recent IPCC report for limiting warming to 1.5 degrees C, which requires reducing global emissions by 40% below 1990 levels by 2030 and net-zero emissions by 2050 (White House 2021b).The transportation sector accounts for 28% of all U.S. GHG emissions, which is the largest source of greenhouse gas (GHG) emissions among all sectors in the U.S. Light-duty vehicle (LDV) emissions represent the largest share of transportation emissions (EPA 2022a(EPA , 2022b)).Achieving cuts in GHG emissions necessary to meet the U.S. goals therefore requires substantial emissions reductions from passenger vehicles.
Transitioning the fleet to electric vehicles (EVs) is regarded as a promising method for reducing passenger vehicle emissions (Leard and McConnell 2020, Carey 2023).We use carbon dioxide (CO 2 ) emissions from vehicles and electricity generation as a proxy for LDV GHG emissions because CO 2 comprises 97% of LDV tailpipe GHG emissions, and hydrofluorocarbons, which account for 2%, are no longer permitted to be used in new vehicle air conditioners, and because the grid decarbonization scenarios we use reflect only CO 2 emissions (Gagnon et al 2022).Furthermore, U.S. electricity generation generates other GHGs, but CO 2 accounts for almost 98.9% of CO 2 -equivalent emissions from U.S. electricity generation (EPA 2023, tables 2-11) and emissions of other gases such as N 2 O and CH 4 will also decrease as the grid decarbonizes.In a full lifecycle analysis of LDV GHG emissions, Zhu et al (2021) found that CO 2 emissions accounted for 91%-95% of CO 2equivalent emissions in all scenarios they considered.We do not account for emissions embodied in vehicles which are about 7% and 2.6% of lifecycle emissions of BEV and gasoline-fueled vehicles, respectively (e.g.Logan et al 2021, Rasbash et al 2023).We assume that passenger cars will comprise about 30% of LDV sales, pickup trucks 17%, with SUVs and vans making up the remainder, continuing a long-term trend of declining passenger car market share in the U.S. As the SUVs' fleet share increases in the future, embodied emissions will become a larger factor in total emissions.We also omit emissions associated with charging infrastructure, which were found by Bohnes et al (2017) to have a negligible effect on lifecycle battery EV emissions.Nor do we account for end of life emissions, which are a fraction of 1% for both vehicle types (Zhu et al 2021, Rasbash et al 2023).These emissions in future years will also decrease with the decarbonization of electricity generation and other U.S. sectors, and also depend on the economies of other nations that supply commodities to the U.S., which are outside the scope of our analysis.
The carbon intensity of the grid has been shown to be by far the most important determinant of lifecycle assessments of BEV GHG emissions (Marmiroli et al 2018).Our modeling is at the national level which averages across large regional differences in emissions from electricity generation and EV adoption, especially during the early phases of grid decarbonization.In this respect, to date 40% of U.S. EV sales have been in California where grid emissions are less than half the U.S. average (CARB 2023).In terms of CO 2 emissions per vehicle, the geography of EV sales and the timing of EV charging matter (Woody et al 2022, Vega-Perkins et al 2023).However, as we show below, the integration of low EV sales when the grid is relatively dirty with high EV sales when it is relatively clean means that EV's upstream plus tailpipe emissions will be a small fraction of total LDV CO 2 emissions through 2050, provided that the transition to EVs and deep decarbonization of the U.S. grid are accomplished before 2050.
The achievement of national goals for decarbonizing LDV transportation critically depends on three rates of change: (1) the rate at which EVs' share of new vehicle sales increases, (2) the rate at which the stock of passenger vehicles turns over and, (3) the rate at which the electricity grid decarbonizes.The central issue we address is how sensitive accomplishing the national goals for carbon emissions reduction is to the timing and rates at which the three factors change.We also evaluate the near-term importance of vehicle and upstream EV emissions in the context of an urgent and eventually successful transition to EVs.
Although the energy efficiency of electricity production varies depending on the grid mix and how renewable energy is accounted for, under any reasonable assumptions most of the energy used to power EVs occurs upstream (USDOE 2023).As a consequence, the benefits of electrifying cars and light trucks depends on GHG emissions from electricity generation.Numerous studies have addressed this issue and documented the importance of factors such as the marginal impact of additional electricity use, the location and timing of EV charging, and the importance of full lifecycle GHG emissions (e.g.Leard 2023).Increasing longevity of U.S. passenger cars and light trucks has been observed in studies dating back to 1996, using data sets spanning the years from 1958 to 2020.Greene and Leard (2023) show that longevity, as measured by the age at which half of a given vintage of vehicles are expected to be still on the road and half to have been scrapped, has been increasing over time.Greene and Leard (2023) summarize results from prior studies that use data spanning over 1958-2020, which suggest longevity has increased by an average of 0.5% to 1%, and calculate a new vehicle sales-weighted annual rate of 0.67% per year based on relevant U.S.-based studies.
We consider two possible projections for future vehicle longevity.The first is based on recent observed scrappage and survival frequencies in 2019-2020 that we denote as the static scrappage scenario.It assumes no change in the rate of survival over time.Our second projection accounts for the observed trend summarized in Greene and Leard (2023), where vehicles are scrapped less frequently and last longer.We call this scenario dynamic scrappage.We plot the implied survivability curves for the scenarios in figure 1.The horizontal axis in each subfigure is the age of a vehicle, and the vertical axis measures the survival probability, which quantifies how likely a vehicle is to remain in operation.We present vehicle survivability probabilities separately for cars, SUVs and vans, and pickup trucks.We can see that the dynamic scrappage scenarios show increasing survival probabilities across all ages and vehicle types.For example, under the static scrappage scenario, a new pickup sold in 2020 has about a 50% probability of being in operation by the age of 25.For the dynamic scrappage scenario, a new pickup truck sold in 2035 has about a 60% probability of being in operation by the age of 25.Adopting the dynamic scrappage scenario implies that the historical trend in stock turnover slowing down will continue through mid-century, increasing vehicle longevity and reducing the rate at which the LDV stock turns over.

New EV sales and electricity grid decarbonization
We consider a range of new EV sales and electricity grid decarbonization.We consider two key pathways for new vehicle sales: one where the market share for new EVs reaches 100% by 2035, and another where the market share for new EVs reaches 50% by 2035 and 100% by 2045.In each scenario, we assume the split between battery EV and plug-in hybrid EV sales are 70% and 30% of total new EV sales, which reflects market shares of 2021 sales.We chose the 100% by 2035 scenario because several large automakers including Ford and General Motors have made public announcements of their plans to phase out gasoline vehicle sales by 2035 and California regulators have proposed a zero emission vehicle mandate scheduled to ban the sale of gasoline and diesel vehicles by 2035 (Becker 2022).We chose the 50% by 2035 scenario because various industry sources forecast this rate of EV adoption (Muratori et al 2021).The scenarios appear in figure 2(a).A critical difference between the scenarios is that in 2035, the slower rate of EV adoption scenario has an EV market share of 50%, or half of that of the fast EV adoption scenario.In each scenario, we assume an initial growth in hybrid vehicles followed by a phaseout of hybrid vehicles, based on recent statements made by automobile manufacturers (Jin 2022).
We consider four scenarios for grid decarbonization: a 100% reduction in carbon dioxide (CO 2 ) per kilowatt hour (kWh) by 2035, a 95% reduction in CO 2 per kWh by 2050 relative to 2020, a current policy with no nascent technologies scenario, and a 65% reduction in CO 2 per kWh by 2050.These scenarios are based on National Renewable Energy Laboratory 2021 and 2022 standard scenarios reports and encompass nearly the entire range of projections from the NREL 2022 report (Cole and Carag 2021, Gagnon et al 2022).
To understand the implications of stock turnover trends, possible future paths for new EV sales and grid decarbonization schedules for decarbonizing LDVs, we build a stock turnover and GHG emissions model.The model tracks the evolution of 12 vehicle types uniquely defined by combinations of vehicle body style (car, SUV/van, pickup truck) and powertrain (gasoline, hybrid, plug-in hybrid, and battery EV).We calibrate the model to alternative scrappage schedules based on prior studies and a robust set of assumptions for vehicle scrappage in the spirit of Keith et al (2019).We merge data on vehicle characteristics such as fuel economy, battery range, vehicle miles traveled

Results
We simulated the stock turnover model under alternative scenarios to project changes in stock composition.In figure 3, we plot powertrain registration shares, defined as the number of registered vehicles  3 show the evolution of powertrain shares for the scenario where the new EV market share reaches 100% by 2045.In this scenario, hybrid vehicle sales increase until around 2035, when they begin to be phased out and replaced by EV sales.Therefore, we see the hybrid stock peaking around 40% in 2040, then gradually falling after that year.The stock of EVs takes longer to increase, hitting a 50% market share in 2045.
Figure 4 shows changes in vehicle miles traveled (VMT) aggregated by vehicle powertrain as a percentage of total VMT.While this figure looks qualitatively similar to figure 3, its key distinction is that the VMT shares for EVs approach 100% faster than the registration shares.In panel (a), VMT among EVs reaches 90% of total VMT by 2050.In panel (a) of figure 3, the share of EVs on the road is projected to be 80%.The reason that VMT share approaches 100% faster is that newer vehicles tend to be driven more than older vehicles and the stock transition to electric is defined by a rapidly rising EV new vehicle market share.
We present our main findings for CO 2 reductions in figure 5.This figure contains a range of assumptions for scrappage rates, new EV sales, and grid decarbonization.The plots show percentage changes in CO 2 emissions relative to 2020.In Panel (a), we present results where we use static scrap rates and assume the sales share for EVs reaches 100% by 2035.
Several key takeaways emerge from the figure.CO 2 emissions fall by about 20%-25% in 2030 relative to 2020 levels.Because of rate of stock turnover, in 2030, a large percentage of LDV emissions are baked in as gasoline vehicles remain on the road for many years, which is why we see only a moderate amount of By 2050, we project that the stock will have had enough time to turn over to achieve substantial emissions reductions of about 75%-85% relative to 2020.This result is consistent with Wissell et al (2022) for scenarios with a nationwide EV policy, whereby the market share of EVs reaches 100% by 2035.By this date, a majority of the 2020 stock has been replaced by new vehicles, where an increasing share of these vehicles are EVs.
Coordination of future grid decarbonization with EV market penetration does not appear to have a significant effect on LDV emissions reductions for the range of scenarios that we consider.According to Panel (a) in figure 5, this is clearly the case for 2030, where the projected reductions are nearly overlapping at around 20%-25%.In 2030, EVs have not had much time to percolate through the vehicle stock.Beyond 2030, the choice of grid decarbonization scenario matters but only for the slowest grid decarbonization scenario.This is because gasoline vehicles remain a significant share of stock, even during 2030-2040 (see figure 3).The on-road share of EVs picks up significantly just as grid decarbonization kicks in for all three grid decarbonization scenarios.The timing creates an interaction between the evolution of the vehicle stock and electricity grid emissions intensity, where the choice of alternative grid decarbonization scenario has only a minor effect on LDV emissions.
These results are robust to alternative assumptions about projections of EV sales and scrap rates.In Panel (b) of figure 5, we show projections where the new EV sales share reaches 100% by 2035.Emissions reductions by 2030 are about 33% in this scenario, which is significantly more than those seen in Panel (a).The static scrappage schedules lead to faster stock turnover, which accelerates the rate at which old gasoline vehicles are removed from the stock.Therefore, we see greater short run emissions reductions in this set of projections.By 2050, emissions reductions are 80%-90%, which is similar to the results presented in Panel (a).By this year, the stock will have had enough time to nearly completely turnover for either assumed set of scrappage schedules; therefore, emissions reductions in Panels  see that the new EV sales share assumption has little effect on 2030 emissions reductions.By 2050, projected reductions are expected to be 70%-80%, which is somewhat less than those seen in Panel (a).Increased hybrid sales in this scenario mitigate the effect of slower EV sales growth, reducing total LDV CO 2 emissions in 2050 by about 25%.The 10 year delay in achieving a 100% new EV sales share still provides the stock with enough time to retire most old, fuel inefficient gasoline vehicles by 2050.
In each panel, we break out CO 2 electricity use emissions from EVs as a percentage of benchmark year emissions from all LDVs.These appear as dashed lines in each panel.A few notable findings appear.First, for the grid decarbonization scenarios that have the grid significantly decarbonized by 2050 (all of those except the 65% by 2050 scenario), EV emissions increase, then decline.Moreover, the maximum amount of EV emissions is tiny, peaking around five percent of benchmark LDV emissions.Second, EV emissions remain low even for the slowest grid decarbonization scenario.For this scenario, EV emissions in 2050 represent about 10% of benchmark 2020 LDV emissions.In summary, EVs appear to contribute little to total LDV emissions through 2050 in every scenario.This result reduces concerns about the pollution output from EVs relative to gasoline vehicles, especially in the initial decade of transition.

Discussion
We draw two general conclusions from our analysis.First, substantial CO 2 emissions reductions are possible by 2050, even with the slower rate of stock turnover.This result is consistent with both Hill et al (2019) and Alarfaj et al (2020).However, 2050 emissions from LDVs are not zero, primarily due to the prolonged longevity of gasoline vehicles.EVs represent a small portion of total stock emissions, even for scenarios of slow grid decarbonization.Together, the results indicate that policies such as phasing out new plug-in hybrid electric vehicles (PHEV) sales, accelerated gasoline vehicle scrappage programs, deployment of low-carbon liquid fuels and changes in travel behavior and urban form (e.g.Muniz and Dominguez 2020, Dillman et al 2021) are likely be necessary to achieve LDV CO 2 emissions by 2050.
Second, our results are robust to our range of assumptions for future grid decarbonization, and the coordination of grid decarbonization with increasing EV market share appears to have little effect on projected LDV emissions, provided that the grid is substantially decarbonized by 2050.This is primarily due to the timing of stock turnover.The time required to accumulate a large share of EVs on the road because of stock turnover reduces the magnitude of the effect that grid emission rates have on LDV emissions in the early phase of transition.In essence, it is a matter of fortunate timing: the point in the future at which grid emissions rates have fallen significantly is when the share of on-road EVs will be the largest.
Our second finding provides a source of optimism for meeting 2050 decarbonization goals.Even if one or multiple decarbonization objectives are delayed, such as completely phasing out gasoline powered vehicles from the new vehicle market by 2035, substantial emissions reductions can still occur by 2050 under a broad range of possible future scenarios.The robustness of decarbonizing the transportation sector suggests that policy makers should not be discouraged if progress along a key dimension of the transition evolves more slowly than current expectations.

Methods
We merge several datasets for our analysis.The primary data are LDV registrations from IHS Markit.These data are annual counts of vehicle registrations in the US, where counts are tabulated at the national level and are recorded in the first quarter of each calendar year.We observe registration counts for every calendar year between 2002 and 2020.
We merge the count data with fuel economy and electricity use per mile data from Wards Automotive and fueleconomy.gov.For calculating gasoline use and CO 2 emissions from gasoline, hybrid, and plugin hybrid vehicles driven using gasoline, we assign gallons per mile as the inverse of fuel economy.We merge the energy use data with the registrations using all available vehicle identifiers, including model year, make, model, trim, powertrain, drive type, body style, and engine size.To assign gasoline and electricity use for PHEV, we apply utility factors from the Society of Automotive Engineers (SAE) J2841 database.For 2020, we assume that 55% and 27% of PHEV miles traveled are in charge-depleting mode for cars and SUVs/vans/pickup trucks, respectively.We assume that the percentage of miles driven using the vehicle's battery for PHEVs increases over time in response to increases in battery range.By 2050, the fractions of miles traveled in charge-depleting mode for PHEVs increase to 88% and 77% for cars and SUVs/vans/pickups, respectively.
We aggregate registration and energy use per mile data to the calendar year-model year-powertrainbody style level by summing the detailed nameplate level counts and taking a registration-weighted average of fuel economy and electricity per mile.We aggregate counts over four powertrains: gasoline, hybrid, plug-in hybrid, and electric.For our body style category, we aggregate to three types: cars, SUVs and vans, and pickup trucks.
We obtained annual VMT schedules from the NHTSA database underlying the most recent analysis of federal fuel economy standards.These schedules are functions of vehicle age and are disaggregated by the three body types: car, SUV/van, and pickup truck.We assume that these schedules remain fixed in future years.
We The 2022 Standard Scenarios Report discusses a 100% grid decarbonization by 2035 scenario.We use this to represent the fastest possible rate of electricity grid decarbonization.The report also includes a current policy scenario, where no new decarbonization policies are adopted through 2050.This scenario incorporates features of the Inflation Reduction Act (IRA) such as the effects of IRA tax credits.
The reports provide projections of CO 2 emissions and electricity generation for every other year through 2050.We constructed an annual projection of these variables by linearly interpolating the in between missing years.We then computed a projected CO 2 rate by dividing CO 2 emissions by electricity generation in each year.
To convert gallons of gasoline used to CO 2 emissions, we use data from the EPA where each gallon of gasoline burned emits 8,887 grams of CO 2 , which we convert to 0.00887 metric tons for calculating emissions (EPA 2022c).In computing CO 2 emissions from electricity used in EVs, we account for the fact that electricity transmission and distribution losses mean that only 95.14% of the electricity generated makes it to the vehicle (USDOE 2023).This increases upstream emissions by about 5.1%.Likewise, to put the gasoline vehicle emissions on the same footing as EV emissions, we include the efficiency of gasoline refining, transport and distribution of 0.8659, which increases gasoline emissions by about 15.5% (USDOE 2023).
We assume new vehicle sales increase by 0.1% per year, following assumptions in the EIA Energy Outlook (EIA 2022a, table 38).We model sales and stocks of plug-in hybrid EVs and battery EVs separately.We assume that new battery EV sales represent 70% of all new plug-in hybrid EV and battery EV sales, which is consistent with 2021 new vehicle market shares.For each sales scenario, we assume that new hybrid vehicle sales initially increase, then decline to zero as the share of EVs approaches 100%.For the scenario where new EV sales represent 100% market share by 2035, we assume that the new hybrid market share reaches 30% in 2028, then declines to zero by 2035.The initial increase represents an effort to improve fuel economy of internal combustion engine vehicles as imposed by fuel economy and GHG standards (EIA 2022b).For the scenario where new EV sales represent 100% market share by 2045, we assume that new hybrid market share reaches 35% in 2028, then gradually declines to zero by 2045.In projections that we have made that are available upon request, we find that our results are insensitive to this assumption.We make several assumptions about fuel intensity (the inverse of fuel economy) and electricity use per mile projections.We assume that fuel intensity of gasoline and hybrid vehicles falls by two percent per year through 2022, then by four percent per year through 2026.After 2026, we assume that fuel intensity of gasoline and hybrid vehicles falls by 0.1% per year (EIA 2022a, table 40).For EVs, we assume that electricity use per mile falls by 0.5% per year.

Scrap rates
We assign scrap rates, which are defined as the percentage of a vehicle type that is removed from the on-road stock from one year to the next, using two alternative scrap rate schedules.We follow the scrappage literature and use logistic scrappage schedules (Parks 1977, Greene and Chen 1981, Engers et al 2009, Kolli et al 2010, Bento et al 2018).We adopt scrappage schedules based on logistic estimation of scrap rates and vehicle age from Greene and Leard (2023).We use predicted values for 2019 to assign scrap rates and denote this as a static scrappage schedule.The second alternative set of scrap rates is based on dynamic scrappage schedules projected in Greene and Leard (2023).We use the projected logistic survival curves and their implied scrappage probabilities through 2050 from this study.

Stock turnover equations
We model how the stock of LDVs evolves with a simple set of survivability equations.Denoting q bf (a, t) as the number of registered vehicles of body style b, powertrain f, age a in year t, the stock evolves according to q bf (a + 1, t + 1) = ( 1 − y bf (a, t) ) q bf (a, t) , where y bf (a, t) is the scrap rate of vehicle body style b, powertrain f, age a in year t.

Figure 1 .
Figure 1.Implied survival probabilities for alternative scrappage scenarios and vehicle types.

Figure 2 .
Figure 2. New electric vehicle sales and electricity grid decarbonization scenarios.

Figure 3 .
Figure 3. Registration shares for alternative scenarios.Notes: In each figure, the vertical axis measures the registration share of on-road vehicles for each powertrain.Registration shares are calculated as the number of registrations of a particular powertrain divided by total registrations and sum to one.The EV share includes both plug-in hybrid vehicles and battery electric vehicles.The horizontal axis represents the year of the simulation.Panels (a) and (b) include assumptions for new EV sales to reach 100% market share by 2035.Panels (c) and (d) include assumptions for new EV sales to reach 100% market share by 2045.Panels (a) and (c) include assumptions for static scrap rate schedules reflecting 2019-2020 scrappage data.Panels (b) and (d) include assumptions for dynamic scrap rate schedules (see methods).

Figure 4 .
Figure 4. Vehicle miles traveled shares for alternative scenarios.Notes: In each figure, the vertical axis measures share of vehicle miles traveled (VMT) of on-road vehicles for each vehicle powertrain.VMT among the EV group includes VMT of both plug-in hybrid vehicles and battery electric vehicles.The horizontal axis represents the year of the simulation.Panels (a) and (b) include assumptions for new EV sales to reach 100% market share by 2035.Panels (c) and (d) include assumptions for new EV sales to reach 100% market share by 2045.Panels (a) and (c) include assumptions for static scrap rate schedules reflecting 2019-2020 scrappage data.Panels (b) and (d) include assumptions for dynamic scrap rate schedules (see methods).
(a) and (b) by 2050 are similar.In Panels (c) and (d) of figure 5, we show projections where the new EV sales share reaches 100% by 2045.By comparing Panel (c) with Panel (a), we

Figure 5 .
Figure 5. Emissions reductions for alternative scenarios.Notes: In each panel, the left vertical axis measures the percentage change in CO2 emissions relative to the year 2020 for all light-duty vehicles (LDVs).The right vertical axis measures emissions from EV electricity use, including electricity use by plug-in hybrid EVs and battery EVs, as a percentage of 2020 LDV emissions.The horizontal axis represents the year of the simulation.Panels (a) and (b) include assumptions for new EV sales to reach 100% market share by 2035.Panels (c) and (d) include assumptions for new EV sales to reach 100% market share by 2045.Panels (a) and (c) include assumptions for static scrap rate schedules reflecting 2019-2020 scrappage data.Panels (b) and (d) include assumptions for dynamic scrap rate schedules (see methods).The green lines indicate the future pathway for electricity grid decarbonization where the grid achieves a 100% reduction in CO2 per KWH by 2035 relative to 2020.The blue lines indicate the future pathway for electricity grid decarbonization where the grid achieves a 95% reduction in CO2 per KWH by 2050 relative to 2020.The gray lines represent future pathways given the current suite of decarbonization policies.The red lines indicate the future pathway for electricity grid decarbonization where the grid achieves a 65% reduction in CO2 per KWH by 2050 relative to 2020.The solid lines show total LDV emissions relative to base year and the dashed lines represent emissions from EV electricity use relative to base year LDV emissions.

. The slowdown of passenger vehicle stock turnover
Archsmith et al 2015, Tamayo et al 2015, Marmiroli et al 2018, Miller et al 2020, Milovanoff obtained electricity generation (denominated in kilowatt hours) and tons of CO 2 for two future scenarios from the National Renewable Energy Laboratory (NREL) 2021 Standard Scenarios Report two scenarios from the 2022 Standard Scenarios Report (Cole and Vincent Carag 2021, Gagnon et al 2022).The 2021 NREL report describes the scenarios