Battery-powered bargains? Assessing electric vehicle resale value in the United States

The resale market will play a critical role in expanding plug-in electric vehicle (PEV) adoption to middle- and lower-income households. Understanding PEV depreciation trends in comparison to those of conventional gasoline vehicles (CVs) is critical for assessing PEV affordability and informing relevant policy, such as subsidies for used PEVs. We deliver comprehensive, high-resolution estimates of value retention rates at the make-model level in the United States for battery electric vehicles (BEVs), plug-in hybrid electric vehicles (PHEVs), hybrid electric vehicles (HEVs), and CVs using a database of 9 million used cars listed online between 2016 and 2022. While BEVs and PHEVs have depreciated at faster rates than CVs, this trend is changing, with newer model year BEVs and those with larger ranges have significantly higher retention rates than older model years with smaller ranges. Tesla BEVs are a notable exception following the opposite trend, with earlier model years holding their value better than newer model years. Subsidized BEVs in the new market are associated with lower prices for the same model in the resale market, with the $7500 federal subsidy translating to a 3% lower resale price on average. Finally, disruptions from the COVID-19 pandemic have affected affordability across all vehicles, with mean listing prices rising 37% and 39% for CVs and BEVs, respectively, from January 2020 to March 2022 in inflation-adjusted 2019 dollars.


Introduction
Plug-in electric vehicle (PEV) adoption is a cornerstone of plans to transition to a sustainable, lowcarbon future [1,2], but the success of PEVs as a substitute for gasoline-powered conventional vehicles (CVs) will depend on consumers' willingness to purchase them.While higher prices and limited driving ranges are frequently cited as barriers to adoption [3,4], researchers have found that uncertainty in their resale value (or 'resale anxiety' [5,6]) remains an important consideration when purchasing a technology with uncertain durability [7].PEV affordability in the resale market is also important for expanding adoption beyond wealthier households, which currently comprise the vast majority of PEV owners [8].Finally, accurate estimates of PEV depreciation rates are important inputs for many modeling efforts surrounding PEV development, such as total cost of ownership models [9][10][11][12].Given the significance of understanding PEV resale value, this study attempts to provide comprehensive estimates of PEV value retention rates in the used market, leveraging a detailed dataset of used vehicle listings.
Vehicle resale value is affected by a variety of factors, such as the vehicle make, model, trim, year, and mileage, as well as features related to the market, such as the location and overall supply of used vehicles.For PEVs, resale value can be affected by additional factors, such as the all-electric driving range and battery condition [13], as well as PEVspecific policies, such as tax incentives in the new market [14].Much of the prior research estimating PEV value retention rates has concluded that while hybrid vehicles (HEVs) have depreciated at similar rates as CVs, plug-in hybrid vehicles (PHEVs) and battery electric vehicles (BEVs) have depreciated faster [15][16][17][18].Table 1 summarizes the prior research on PEV resale value.
Although more rapid PEV depreciation is a consistent finding in prior studies, this outcome is not necessarily the same for every vehicle.Tesla BEVs, for example, have been found to have among the highest value retention rates of any vehicle [10,13], in part due to their desirable features, such as high driving ranges and well-established private charging infrastructure, but also due to supply constraints they have historically faced in the new market, leading to higher used vehicle prices [22].Furthermore, many of the studies concluding that PEVs depreciate more quickly than CVs are based on data from earlier PEV models between 2011 and 2016, a time when most BEVs had relatively low driving ranges and were first generation vehicles [13,19,20].More recent studies are finding some evidence that newer model PEVs are holding their value better [10].
In this study, we aim to improve upon prior estimates of PEV value retention using listing prices from a large, nationally representative dataset of used cars listed online between 2016 and 2022 in the United States.Whereas prior studies have used relatively smaller samples of pre-processed data, such as the Edmunds true market value [9,10,13], or snapshots of available listed vehicles [12,19,20], the data we use contains the raw, daily listing prices from 66 641 dealerships.The detailed data enables an unprecedented ability to quantify changes in value retention due to vehicle features (e.g, mileage), environment features (e.g.number of days the vehicle was listed), policy features (e.g.available subsidies), and changes over time (e.g.newer versus older model years), enabling greater insights into PEV value retention rates compared to prior studies.We combine the listings data with other vehicle features to quantify the retention rates of each car model in the dataset via regression.

Data
We use vehicle listing data licensed from marketcheck.com, a market research firm that collects vehicle listing data from individual dealership websites daily.The same dataset is used in a related study on PEV mileage [23].Since the PEV market has a limited number of SUV models, we limit our analyses to only car models.While the dataset covers vehicles listed between January 2016 and March 2022, we only use listings up until the end of 2019 to avoid the pricing disruptions experienced due to supply shortages during and after the COVID-19 pandemic, which began in early 2020 (tables A5 and A6 show results from vehicles listed between 2020 and 2022, and figure 5 shows prices through March 2022 for different vehicle types for comparison).We also limit our dataset to vehicle ages between 1 and 8 as fewer BEV listings are available outside of this range (fewer vehicles are listed used within 1 year of being new, and few used BEVs are older than 8 years old as of December 2019).Finally, we only include vehicle models that comprised at least 1% of the listings within each powertrain as a compromise between including a representative sample of the vehicle market while remaining computationally reasonable.Of all the vehicle listings in the dataset, this 1% cutoff decision included 96% of the BEVs (13 models), 97% of the PHEVs (7 models), 95% of the HEVs (19 models), and 59% of the CVs (25 models).The lower inclusion rate (59%) of CV listings is because the CV market contains many models with few sales, such as exotic cars (the remaining 41% of listings is comprised of 852 vehicle models).The final dataset includes 9015 324 unique used car listings from 66 641 dealerships.Table 2 summarizes the dataset by powertrain, with Tesla and Non-Tesla BEVs separated given Tesla's historically higher retention rates and significant size in the U.S. market, and table A1 in the supplementary information summarizes each car model included in our analyses.
To control for other factors that can affect retention rates, other vehicle specifications were obtained and joined onto the listings data.We use manufacturer's suggested retail prices (MSRPs) from carsheet.io[24] and BEV and PHEV ranges as well as all vehicle efficiencies (miles per gallon for gasoline-powered vehicles and kWh per 100 miles for electricity-powered vehicles) from fueleconomy.gov[25], with a small number of missing values added from carsheet.io.Both listing prices and MSRPs were inflation-adjusted to constant 2019 dollars using the consumer price index [26].Vehicle operating cost (in cents per mile) was computed for every vehicle model using the vehicle efficiencies and monthly gasoline prices [27] and annual average electricity prices [28] where the vehicle was listed from the US energy information administration, using utilization factors (0-1) from fueleconomy.gov to compute the electric and gas portions of operating costs for PHEVs.Finally, we also include data on the federal and state subsidies available for new PEVs at the time used vehicles were listed.Data on these incentives between 2016 and 2018 were taken from Wee et al [29] and manually updated through 2020 primarily from the Alternative Fuels Data Center at https://afdc.energy.gov/.

Modeling
To establish a comparable metric across different vehicles from different price ranges, we compute a value retention rate for every vehicle model, calculated as the listing price divided by the original MSRP based on the vehicle make, model, trim, and model year.While some prior studies subtract off available federal and state subsidies from the MSRP before computing retention rates [10,13], we chose not to  make this adjustment because doing so is a nonlinear transformation of the retention rate that affects both the intercept and slope terms of the resulting depreciation.We believe this to be inconsistent with the effect that a subsidy in the new market should have on retained value in the resale market, which is to have an immediate effect lowering the price (a lower intercept term) but not necessarily affect the rate by which the value drops from that price (the slope term).In contrast, by subtracting the subsidy off the MSRP when computing the retention rate, the resulting depreciation curves will have larger intercept terms and larger depreciation slopes, which implicitly assumes that subsidized vehicles (BEVs and PHEVs) should depreciate as faster rates and have higher initial retention rates than unsubsidized vehicles.Finally, the subsidy available when a used vehicle was initially purchased is arguably less relevant to its resale value; indeed, it is the subsidy available for the same vehicle model in the new market at the time the used vehicle is listed that should impact resale value.For example, the maximum price a buyer should be willing to pay for a used Nissan Leaf should be the price for a new Nissan Leaf minus available subsidies, otherwise the buyer could simply buy the new version for the same price.Thus, in our models we include the total subsidy (state plus federal) available for the equivalent new model at the time the used vehicle is listed as a separate covariate to assess any potential effect of new PEV subsidies on listing prices in the resale market.
We develop an exponential decay model to estimate the potential effects of various factors, such as the powertrain, make, model, model year, mileage, and available new market subsidies, on the retention rate of different vehicles.The basic exponential model follows that of other related studies [9,10]: where r is the retention rate, x is a matrix of model covariates, β is a vector of coefficients for the covariates, and α represents the initial retention rate upon immediate sale in the new market.Taking the log of both sides of this equation yields a model that can be estimated using linear regression: Interpreting the estimated β coefficients in a meaningful way (e.g. the change in r based on a change in any one covariate in x) requires the transformation e β − 1.We present the transformed coefficients in tables 3 and 4 and the estimated (nontransformed) coefficients in tables A2 and A3 in the supplementary information.

Results
Figure 1 compares the retention rates of CVs (gray bands) with those of other powertrains (green bands), including HEVs, PHEVs, and BEVs, where the median (solid lines) and interquartile ranges (bands) of retention rates were computed for all listings in each month of age.In general, CVs and HEVs follow similar retention rate patterns over time while PHEVs initially retain slightly more value before dropping at steeper rater compared to CVs.BEVs in general depreciate faster than CVs, with the exception of Tesla BEVs (blue bands) which actual retain higher values in the first few years of age relative to CVs before falling to similar values with CVs.This aligns with prior research in this area which finds Tesla BEVs follow a different depreciation trend than other BEVs [13].
To quantify the differences in retention rates by powertrain, we estimate an exponential decay model of the log of retention rate on age interacted with powertrain type.Since the estimated coefficients (presented in table A2 in the supplementary information) are less intuitive to interpret, we present a transformation of those coefficients in table 3, interpreted as the initial sale retention rate (exponentiation of the intercept coefficients) and annual depreciation rates (exponentiation of the slope coefficients minus one).Models 1 and 2 are identical except for BEVs, which are pooled in Model 1 and separated by Tesla or Non-Tesla in Model 2.
Results are consistent with the trends shown in figure 1: the annual depreciation rate (change in retention rate with age) is approximately 9.7% and 10.0% for CVs and HEVs, respectively, and steeper for PHEVs and Non-Tesla BEVs at approximately 14.5% and 16.1%, respectively.Tesla BEVs depreciate at a slightly lower rate of 13.2% per year but start at a much higher intercept than any other vehicle type, retaining 97.8% of their MSRP upon initial sale compared to only 68.5% for Non-Tesla BEVs.CVs and HEVs retain 79.3% and 80.4%, respectively, of their value upon initial sale and PHEVs retain slightly more at 85.3%.It is important to note the distinction here between retention rate and prices; since BEVs tend to have higher MSRPs than CVs, a lower retention rate does not necessarily translate into lower relative prices.For example, 68.5% of $36 600 (the mean MSRP for Non-Tesla BEVs in our data) is $25 071; in comparison, 79.3% of $28 100 (the mean MSRP for Non-Tesla BEVs in our data) is $22 283.Thus, despite the much lower initial sale retention rate for Non-Tesla BEVs, they still often have higher prices than used CVs.
To understand in greater detail the factors associated with depreciation within each powertrain, we estimate additional models on all cars within each powertrain.Again, we present the transformed coefficients in table 4, with the estimated coefficients presented in table A3 in the supplementary information.These models included interactions effects between the vehicle age and each vehicle model to allow for different depreciation intercepts and rates by model, presented in table A4.
One of the notable observations from these models is that while Non-Tesla BEVs and PHEVs have in general depreciated faster than HEVs and CVs, this appears to be changing over time.Specifically, we observe significantly higher initial sale retention rates for these powertrains with each new model year, with Non-Tesla BEVs increasing by 11.23% and PHEVs increasing by 21.42% between 2018 and 2012 model years.The opposite trend has occurred for CVs and HEVs, falling by 5.35% and 1.29%, respectively, between 2018 and 2012 model years.Notably, Tesla BEV retention rates have fallen steeply over this period, with the 2018 model year initial sale retention rate being 47.42% lower than that of the 2012 model year.While this may seem surprising given Tesla's relative popularity, it is important to note that all of the pre-2018 models in this regression are from a single vehicle model: the Model S. Throughout this period, Tesla did not release any new car models (the Model X is an SUV) and continued to cut new Model S prices; as a result, used prices fell in anticipation of continued future lower prices.The pending release of the Model 3 in 2018 could also have contributed to waning demand for the Model S.
These results suggest a more competitive market is emerging across alternative powertrain vehicles, with PHEVs and Non-Tesla BEVs performing increasingly better in terms of value retention in the resale market, though not yet quite as good as HEVs and CVs. Figure 2 illustrates this trend, comparing the predicted two-year old retention rates of a 2018 versus 2014 model year vehicle for every vehicle model in our analyses.While not all vehicle models were available in both years, for those that were we observe a steep increase in the predicted retention rates of PHEVs and Non-Tesla BEVs and a steep decline for Tesla BEVs.
Table 4 also shows other important findings regarding how retention rates vary with different  vehicle characteristics.We observe that increased mileage has a negative impact on retention rate across all powertrains, with Tesla BEVs having the lowest sensitivity to mileage at 3.9% for every additional 10 000 miles and Non-Telsa BEVs having the highest sensitivity at 5%.We also observe lower retention rates for vehicles that remain on dealership lots longer, suggesting that (as would be expected) dealerships reduce the prices of vehicles that take longer to sell.CV prices fall at lower rates (0.08% for every additional 10 d on the lot) relative to non-CVs, which fall at rates between 0.13% (Tesla BEVs) and 0.37% (Non-Tesla BEVs).This suggests dealers are willing to lower prices much more aggressively over time for used alternative fuel vehicles (including HEVs) than used CVs.Finally, we also find that less efficient vehicles (those with higher operating costs) appear to hold their value better than more efficient vehicles (lower operating costs) across all powertrains.While this result may be counterintuitive, it is important to note that more efficient vehicles tend to have higher prices; as a result, higher demand for more affordable (but less efficient) vehicles could lead to less rapid depreciation compared to higher-priced (but more efficient) vehicles.

Driving range and subsidy effects on PEV value retention
Table 4 shows several important effects specific to PEVs.First, we find that BEVs with higher driving ranges have higher retention rates, though this effect is significantly higher for Non-Tesla BEVs, which tend to have lower ranges in general than Tesla BEVs.For every additional 10 miles of range, Non-Tesla BEVs have 5.58% higher retention rates and Tesla BEVs have 1.57% higher retention rates, all else being equal.The range effect is not statistically significant for PHEVs.
Figure 3 shows this relationship, plotting the predicted two-year-old retention rate versus range for multiple model years for select BEVs in our analysis.We only include vehicle models that have at  A4, and the plotted points are the mean twoyear-old retention rates across all listings for a given model year-range pair.Here we see the rapid increase in two-year-old retention rates for newer Non-Tesla BEVs, which have larger electric driving ranges.In contrast, the slope of line for the Tesla Model S is negative because the negative model year effect is larger than the positive range effect, which is smaller in magnitude than that of Non-Tesla BEVs.This suggests that while range is important for value retention, there may be a limit to the higher value retention obtained from increased driving range alone.
We also find an effect from subsidies available in the new vehicle market.A $7500 subsidy for a new BEV (the maximum federal subsidy for which most BEVs in this time period qualified) is associated with lower retention rates in the resale market by 3.3% for Non-Tesla BEVs and 3.4% for Tesla BEVs.This is expected as the retention rates (computed as the listing price divided by the unsubsidized MSRP) should be lower for used vehicles that have an equivalent subsidized model available in the new market compared to those that do not.
To put this subsidy effect in context, we estimate that the total federal subsidies dispensed for new BEVs sold between 2011 and 2019 resulted in $255 million in indirect subsidies to the resale market via lower resale prices (see figure 4).While relatively small in magnitude compared to the $8.7 billion in new subsidies, our estimate is still important to note as it is an additional benefit to consumers that comes at no additional government cost.
Although our data extends up until March of 2022, we chose to censor the data used for all modeling to the end of 2019 due to the supply disruptions in the automotive market that resulted from the outbreak of the COVID-19 pandemic in early 2020.These disruptions led to significant increases in vehicle prices in the new and used markets, with prices for some used vehicles rising higher than their new MSRPs since many new vehicles were unavailable for lengthy periods of time.This rise in pricing made models that rely on assumptions of exponential decline unreliable, hence why we omitted the data    from our models.Nonetheless, descriptions of the post-pandemic listing prices are informative.
Figure 5 shows the mean listing prices for CVs, Tesla BEVs, Non-Tesla BEVs.As of March 2022, mean inflation-adjusted prices are 37%, 39%, and 3% higher compared to in January 2020 for CVs, Non-Tesla BEVs, and Tesla BEVs, respectively.In combination with the overall trend of increased retention rates for Non-Tesla BEVs, this result suggests that, at least in the short term, affordability for all vehicles (including BEVs) is worse than pre-pandemic years, and buyers may have a difficult time finding an affordable BEV in the resale market, adding an additional barrier to more equitable PEV adoption.

Discussion and conclusions
PEVs are a critical component in a low-carbon future, and the resale market will play a critical role in expanding PEV adoption beyond affluent demographics given its greater affordability [30] and sizemore than double the size of the new vehicle market in annual sales [31].In this study, we analyze the retention rates of different cars listed in the resale market, comparing result across different powertrains.
We find in general that PEVs do not hold their value as well as CVs as they age, which is consistent with much of the prior literature on this topic [9,10,12,13,21].However, we also find evidence that this is changing over time, with newer PEV models holding higher retention rates that are approaching those of many CVs.This trend is particularly significant given the rapid advancements in PEV technology, notably in driving range.In our dataset, the mean BEV range grew 76% from 86 to 151 miles from 2012 to 2018.This evolution not only addresses range anxiety but also potentially mitigates resale anxiety [7] as we find that higher-range BEVs have significantly higher retention rates than lower-range BEVs.Mileage continues to be a critical determinant of resale value across all powertrains, with its impact slightly more pronounced in Non-Tesla BEVs.This finding underscores the importance of mileage as a proxy for battery health in BEVs, a key consideration for potential buyers in the used market.
Interestingly, our analysis reveals that subsidies for BEVs in the new market are associated with lower prices in the resale market, effectively passing through an indirect subsidy to used BEVs.While this phenomenon aids in making used BEVs more affordable, it simultaneously could fuel resale anxiety among new BEV purchasers.This dynamic presents a complex scenario for policymakers, especially considering the recent introduction of a used PEV tax credit.The long-term effects of such incentives on PEV resale values warrant further investigation.
The COVID-19 pandemic has significantly impacted the vehicle market by reducing affordability across all powertrains.Recognizing and adapting to these shifts is imperative in maintaining PEV affordability and, consequently, in promoting broader adoption of the technology.The evolving landscape of PEV retention rates, particularly among Non-Tesla models, is noteworthy, but this also introduces a new dynamic where the affordability of subsequent vehicles diminishes due to elevated pricing and retention rates.In the post-pandemic era, the surge in used vehicle prices has somewhat eroded the affordability that once characterized the used PEV market.
Our study is not without limitations.The exclusion of data post-2019 due to the COVID-19 pandemic's impact on the market bounds our findings, with 2018 model years being the most recent models we can analyze.This exclusion means we cannot see how the resale market has evolved with the many newer models that have since been released during the pandemic and post-pandemic years.Additionally, our reliance on listing prices as a proxy for transaction prices may not fully capture the nuances of sale negotiations and If there are systematic differences in listing and transaction prices by powertrain, then our comparative analyses across powertrains may have slight discrepancies with respect to true differences in retained value.Nonetheless, the retention rates presented in this study should accurately reflect what consumers would have observed at dealerships prior to making a purchase decision.Our study also focuses on the U.S. market, which may limit the generalizability of our findings to global markets where different economic conditions, consumer preferences, and policy frameworks prevail.Our analysis also does not extensively delve into the impact of brand perception and consumer loyalty, particularly concerning Tesla and other emerging PEV manufacturers.The strength and reputation of a brand can significantly influence resale values, and as the market continues to evolve, shifts in consumer perception could markedly alter the landscape of PEV value retention.
Finally, the rapidly evolving nature of PEV technology, particularly advancements in battery efficiency and lifespan, could alter the depreciation patterns in ways not captured in our current dataset.As newer models with improved technology enter the resale market, their retention rates could differ significantly from the trends observed in our study.For these reasons, we caution modelers from extrapolating these trends too far into the future.Continuous monitoring and analysis will be essential to fully understand the longer-term trends and implications for PEV adoption and sustainability in the transportation sector. are publicly available on GitHub at https://zenodo.org/doi/10.5281/zenodo.11041585.The vehicle listings data that support the findings of this study are available from www.marketcheck.com,but restrictions apply to the availability of these data, which were used under license for the current study and so are not publicly available.A sample of the data is included in the GitHub repository.The relevant variables in the full original database can be provided on an individual bases for review purposes only to reproduce the study results by contacting the lead contact.All other data used in the study are posted in the repository.

Figure 1 .
Figure 1.Comparison of the median and interquartile ranges of computed retention rates by powertrain and age.The solid line shows the median retention rate and the bands reflect the 25th and 75th percentiles.The same curve for CVs (in grey) is shown for comparison in each sub-figure.

Figure 2 .
Figure 2. Predicted two-year-old retention rates for all modeled vehicle models, comparing model years 2014 and 2018.Model year 2018 HEVs, CVs, and Non-Tesla BEVs are experiencing lower retention rates relative to model year 2014 while PHEVs and Non-Tesla BEVs are experience higher retention rates.Predictions are made using the estimated model coefficients in table A4.

Figure 3 .
Figure 3. Predicted two-year-old retention rate versus range for select BEV models by model year (only vehicle models with at least three model years are included).Predictions are made using the estimated model coefficients in tableA4, and points are the mean two-year-old retention rates across all listings for a given model year-range pair.

Figure 4 .
Figure 4. Annual total indirect subsidy to the resale market from lower prices due to federal subsidies for new PEVs.

Figure 5 .
Figure 5. Summary of resale market prices for CVs (gray), Tesla BEVs (blue), and Non-Tesla BEVs (green).Solid lines are mean prices and bands show the interquartile range of prices.The vertical red line marks March 2020, the start of the COVID19 pandemic.

Table 1 .
Summary of prior studies on PEV resale value.
a Sample sizes estimated based on descriptions of data in papers.Abbreviations: EPA = Environmental Protection Agency (fueleconomy.gov)TMV = True Market Value (private party data) KBB = Kelly Blue Book (private party data) NADA = National Automobile Dealers Association.

Table 2 .
Summary of used vehicle listings.

Table 3 .
Estimated initial sale retention rates and annual depreciation rates for different vehicle powertrains.

Table 4 .
Estimated effects of vehicle model years and vehicle characteristics on retention rates, computed using estimated coefficients in tableA3.Vehicle model and age interaction coefficients are presented in table A4.