How Much Are Electric Vehicles Driven?

APPLIED ECONOMICS LETTERS 2019, VOL. 26, NO. 18, 1497–1502 https://doi.org/10.1080/13504851.2019.1582847

ARTICLE How much are electric vehicles driven?

Lucas W. Davis Haas School of Business, University of California, Berkeley, CA, USA; Energy Institute at Haas, Berkeley, CA, USA; National Bureau of Economic Research

ABSTRACT KEYWORDS The prospect for electric vehicles as a climate change solution hinges on their ability to reduce Electric vehicles; plug-in gasoline consumption. But this depends on how many miles electric vehicles are driven and on hybrids; vehicle miles how many miles would have otherwise been driven in gasoline-powered vehicles. Using newly- traveled; rebound effect available U.S. nationally representative data, this paper finds that electric vehicles are driven JEL CLASSIFICATION considerably fewer miles per year on average than gasoline-powered vehicles. The difference is D12; L62; Q41; Q54; Q55 highly statistically significant and holds for both all-electric and plug-in hybrid vehicles, for both single- and multiple-vehicle households, and both inside and outside California. The paper discusses potential explanations and policy implications. Overall, the evidence suggests that today’s electric vehicles imply smaller environmental benefits than previously believed.

I. Introduction how many miles the driver would have otherwise driven in a gasoline vehicle. Many studies have pointed out that where in the This paper uses newly-available nationally repre- country you drive an electric vehicle has impor- sentative data from the U.S. Department of tant implications for the environment. This point Transportation’s National Household Travel has been made repeatedly – both in the economic Survey (NHTS) to provide some of the first evi- (Zivin, Kotchen, and Mansur 2014; Holland et al. dence on how much electric vehicles are driven. 2016) and engineering literatures (Tessum, Hill, Prior to the 2017 NHTS, the most recent NHTS and Marshall 2014; Tamayao et al. 2015). was conducted back in 2009, when there were Another important factor is how much you virtually no electric vehicles on the road, so this drive. This has received far less attention, but has represents one of the first opportunities to measure major implications for the environmental impact vehicle miles traveled (VMT) for electric vehicles. of electric vehicles. These data show that electric vehicles are driven After all, it isn’t the manufacturing of electric considerably less on average than gasoline- and die- vehicles that gives them their environmental edge. sel-powered vehicles. In the complete sample, electric If anything, the copper, aluminum, lithium, cobalt, vehicles are driven an average of 7,000 miles per year, and other materials used to build batteries actually compared to 10,200 for gasoline and diesel-powered make electric vehicles more resource intensive vehicles. The difference is highly statistically signifi- (Notter et al. 2010). Altogether, the negative extern- cant and holds for both all-electric and plug-in hybrid alities from manufacturing have been calculated to vehicles, for both single- and multiple-vehicle house- be $1,500+ higher for electric vehicles than for gaso- holds, and both inside and outside California. line-powered vehicles (Michalek et al. 2011). This pattern is surprising because electric vehi- Instead, the prospect for electric vehicles as cles tend to cost less to operate per mile, so should a climate change solution hinges on their ability be attractive to high-mileage drivers (Sivak and to reduce gasoline consumption. But how much Schoettle 2018). Instead, it appears that the exact gasoline is actually saved when a driver buys an opposite has happened, with low-mileage drivers electric vehicle? This depends, crucially, on how being more likely to buy electric vehicles. Limited many miles the electric vehicle is driven and on range, particularly for the first generation of

CONTACT Lucas W. Davis [email protected] Haas School of Business, University of California, Berkeley, CA 94720, USA © 2019 Informa UK Limited, trading as Taylor & Francis Group 1498 L. W. DAVIS electric vehicles, has likely played a key role, so Thus the basic economics of electric vehicles implies this pattern may change over time as more long- the exact opposite of what is observed in Figure 1, range electric vehicles are introduced. with electric vehicles being driven more miles The evidence suggests that today’s electric vehi- per year than other vehicles. cles imply smaller environmental benefits than pre- The pattern in Figure 1 also goes against what viously believed. The less electric vehicles are would be implied by the ‘rebound effect’ (see, e.g. driven, the smaller the implied reductions in gaso- Borenstein 2015). Electric vehicles tend to cost less line consumption, and thus the smaller the implied to operate per mile (Sivak and Schoettle 2018), so reductions in carbon dioxide emissions. Current drivers should use them more.3 Thus the rebound policies like the U.S. federal $7,500 electric vehicle effect provides another reason why to expect, tax credit treat low- and high-mileage drivers uni- ceteris paribus, the exact opposite of what is formly, so don’t provide the right incentive to observed in Figure 1, with electric vehicles being induce adoption by high-mileage drivers. driven more than other vehicles. Figure 2 examines whether this pattern is different for vehicles owned by California house- II. Preliminary evidence holds. About half of all U.S. electric vehicles are Figure 1 shows average miles driven per year for in California, so this breakdown is of significant four different categories of vehicles. This figure intrinsic interest. Moreover, electricity rates in was constructed using data from the 2017 California are higher-than-average (Borenstein National Household Travel Survey (U.S. and Bushnell 2018), so one might have expected Department of Transportation 2018). A valuable to see lower mileage for California electric vehi- feature of the NHTS is that respondents fill out an cles. As it turns out, however, the pattern is ‘Odometer Mileage Record Form’ which requires roughly similar for California- and non- them to write down the current odometer reading California electric vehicles. In both cases, these for all vehicles in the household. For this analysis, data show that all-electric and plug-in hybrid these odometer readings were divided by the age vehicles are driven considerably less than other of each vehicle to calculate the average number of types of vehicles. miles driven per year. Thus according to these data, electric vehicles III. Additional analyses are driven considerably less than other types of vehicles. All-electric and plug-in hybrid vehicles Table 2 compares average annual miles driven for are driven 6,300 and 7,800 miles annually, respec- electric- and non-electric vehicles. Across the tively, compared to 10,200 for gasoline and diesel entire sample, electric vehicles are driven an aver- vehicles, and 11,800 for conventional hybrids.1 age of 7,000 miles per year, compared to Table 1 provides descriptive statistics. 10,200 miles per year for gasoline and diesel vehi- This pattern is somewhat surprising. Purchasing cles. Thus, in the 2017 NHTS electric vehicles are any vehicle requires a buyer to make an intertem- driven 30% less than other vehicles. poral tradeoff between purchase price and operating This pattern holds for several different subsam- costs.2 Relative to gasoline-powered vehicles, electric ples. First, the pattern holds both for all-electric vehicles tend to cost more to purchase but less to and plug-in hybrid vehicles. Second, the pattern operate, so consumer guides recommend electric holds both for single-vehicle and multiple-vehicle vehicles for high-mileage drivers (McDonald 2016). households, suggesting that the pattern cannot be

1The overall average of about 10,000 miles driven per year is consistent with previous studies using data from the 2009 NHTS (see, e.g. Archsmith, Kendall, and Rapson 2015). Similarly, Levinson (forthcoming), shows using data from the 2009 NHTS that conventional hybrids tends to be driven more than gasoline- and diesel-powered vehicles, consistent with the two right-most bars in Figure 1. 2A related literature on gasoline-powered vehicles finds that vehicle buyers are relatively attentive to future operating costs (Busse, Knittel, and Zettelmeyer 2013; Allcott and Wozny 2014; Sallee, West, and Fan 2016). These studies are closely related to an older literature on a broader class of energy-related investments (Hausman 1979; Dubin and McFadden 1984). 3Early electric vehicle adopters were even able in some cases to charge their vehicles for free. During its early years, Tesla famously offered buyers unlimited free charging for life on its fast-charging network, though this practice was ended for buyers placing an order after 1 January 2017. See, e.g., Tim Higgins, ‘Tesla Motors Plans to Charge for Its Quick-Charge Access’ Wall Street Journal, 6 November 2016. APPLIED ECONOMICS LETTERS 1499

Average Miles Driven Per Year 12,000 9,000 6,000 3,000 0 All-Electric Plug-In Hybrid Gasoline/Diesel Conventional Hybrid Note: Constructed using data from the 2017 NHTS. Bars indicate 95% confidence intervals. All estimates calculated using sampling weights.

Figure 1. How much are electric vehicles driven?

Table 1. Miles driven per year, descriptive statistics. These differences cannot be explained by sam- Number of Standard pling variation. For each row the table reports the Observations Mean Deviation p-value from a test that the means in the two sub- All-Electric Vehicles 436 6,300 4,200 Plug-in Hybrids 426 7,800 4,900 samples are equal. In all cases the differences are Gasoline/Diesel 203,988 10,200 5,900 highly statistically significant. In the latest wave of Vehicles Conventional Hybrids 4,443 11,800 6,300 the NHTS there are 400+ all-electric and 400+ plug- Note: This table reports descriptive statistics for miles driven per year by in hybrid vehicles, so there are enough electric vehicle type in the 2017 National Household Travel Survey. All statistics are calculated using sampling weights. Means and standard deviations vehicles to make these comparisons precisely. have been rounded to the nearest 100. Before proceeding, it is worth noting a couple of features about the NHTS. The sample for the completely explained by within-household substi- NHTS is selected using stratified sampling, so tution across vehicles. Third, the pattern holds sampling weights are used throughout the paper both in California, and outside California, consis- in all calculations. A notable advantage of the tent with the evidence in Figure 2. National Household Travel Survey is the large

Average Miles Driven Per Year

California

12,000 Not California 9,000 6,000 3,000 0 All-Electric Plug-In Hybrid Gasoline/Diesel Conventional Hybrid Note: Constructed using data from the 2017 NHTS. Bars indicate 95% confidence intervals. All estimates calculated using sampling weights.

Figure 2. Is this pattern different in California? 1500 L. W. DAVIS

Table 2. Are electric vehicles driven less than other vehicles? for longer trips. While public charging stations are (1) (2) (3) becoming more common, electric vehicle charging Electric All Other p-value Vehicles Vehicles (1) vs (2) remains nowhere near as convenient as filling up Entire Sample 7,000 10,200 0.00 a gasoline-powered (Li et al. 2017;Li2018). Limited All-Electrics Only 6,300 10,200 0.00 range thus could impact both who buys an electric Plug-in Hybrids Only 7,800 10,200 0.00 Single-Vehicle Households 7,600 10,000 0.00 vehicle and how electric vehicles are used. Multiple-Vehicle Households 7,000 10,200 0.00 It is not clear how much limited range should California Only 7,200 9,800 0.00 Excluding California 6,900 10,200 0.00 matter for plug-in hybrids. With a plug-in hybrid, Note: This table reports average miles driven per year for electric- and non- a driver always has the option to run on gasoline, so electric vehicles in the 2017 National Household Travel Survey. Except where indicated, electric vehicles include both all-electric and plug-in plug-in hybrids are not subject to the same range hybrids. The last column reports p-values from tests that the means in limitations as all-electric vehicles. Still, many plug- the two subsamples are equal. All estimates are calculated using sampling weights and have been rounded to the nearest 100. in hybrid vehicle drivers purchased their vehicles with the intention to use primarily powered by electricity, so this may affect both the type of driver sample size – 129,696 households in the 2017 who buys a plug-in and how these vehicles are used. survey. This large sample size is the reason why A related issue is substitution across vehicles for there are enough electric vehicles in order to make multiple-vehicle households. Table 3 shows that these comparisons. only 10% of U.S. households with an electric vehicle A notable limitation of the NHTS is the low are single-vehicle households. Thus in most cases, response rate. The 2017 NHTS has a lower response electric vehicle owners are able to substitute between rate than previous waves, only 15.6% according to the electric- and non-electric vehicles. Multiple-vehicle survey documentation. The NHTS sampling weights households may prefer to use their electric vehicles attempt to correct for non-response by balancing for short trips, while using their gasoline-powered observable household characteristics, but, of course, vehicles for longer trips.4 This type of within- respondents and non-respondents can also differ household substitution thus could help explain the along other dimensions. The electric vehicle owner- lower average annual mileage for electric vehicles. ship rate in the 2017 NHTS is consistent with aggre- Low VMT for electric vehicles could also simply gate data on electric vehicle sales Davis reflect the type of households who tend to buy an (forthcoming), so at a minimum the data does seem electric vehicle (i.e. selection). For example, urban to provide a reasonable description of the broader vs rural differences could help explain the pattern. pattern of electric vehicle ownership, but it is impos- Urban households may be more likely to buy sible to rule out concerns about non-response bias, so electric vehicles, perhaps because of stronger this is worth highlighting as an important caveat. ‘green’ preferences in urban areas, and may also

IV. Potential explanations Table 3. Number of vehicles per household. U.S. Households with at U.S. Households with at least Thus the evidence shows that in the United States, least one vehicle one electric vehicle electric vehicles tend to be driven considerably Single 37% 10% Vehicle fewer miles per year than other vehicles. This sec- Two Vehicles 36% 46% tion discusses possible explanations and then the Three 16% 26% Vehicles following section considers policy implications. Four or More 10% 18% The most obvious explanation is limited range. Vehicles Note: This table was constructed using data from the 2017 National The first generation Nissan Leaf, for example, has Household Travel Survey. Electric vehicles include both all-electric and a range of less than 80 miles, making it impractical plug-in hybrids. All percentages are calculated using sampling weights.

4In related work, Archsmith et al. (2017) shows that households substitute between vehicle attributes when deciding which vehicles to purchase. For example, a household with one fuel-efficient vehicle may be more likely to purchase a second vehicle that is less fuel-efficient. APPLIED ECONOMICS LETTERS 1501 tend to drive fewer miles per year.5 In addition to unusual in the literature. Another well-known urban vs rural, the pattern could be influenced by model in this space is the GREET model, which other differences between ‘green’ and less ‘green’ assumes that electric vehicle batteries are used for communities. Previous research has shown that a lifetime of 160,000 miles (Michalek et al. 2011). environmental ideology is a major determinant Similarly, Archsmith, Kendall, and Rapson (2015) of adoption of energy-efficient vehicles (Kahn assumes that all vehicles have a total lifetime of 2007), and ‘green’ communities may tend to be 257,000 kilometers (159,700 miles). Moreover, places where people drive less. Federal CAFE standards use an assumed lifetime for cars and trucks of 195,000 and 225,000 miles, respectively (Leard and McConnell 2017). All of V. Policy implications these measures are high relative to the newly- available evidence from the 2017 NHTS. For exam- “It seems a foregone conclusion, both in policy and media representations, that electric vehicles are a climate change ple, dividing 195,000 by 7,500 mile per year, implies solution. However determining the potential greenhouse an implausibly long average lifetime of 26 years. gas benefit from electric vehicles is complicated.”– Calculating exactly how much gasoline is saved Archsmith, Kendall, and Rapson (2015). when a household buys an electric vehicle goes Americans have now purchased more than beyond the scope of this study. It could be that 800,000 electric vehicles, counting both plug-in households adopting electric vehicles also change hybrids and all-electric models. Although this is other transportation choices, for example, being still less than one percent of all U.S. registered more likely to use intercity rail for long trips. In vehicles, policymakers see electric vehicles as hav- future work it would be interesting to use long- ing great potential to reduce carbon dioxide emis- itudinal data to assess how household transporta- sions and other forms of pollution, and are tion choices evolve over time. supporting tax credits and other policies to encou- The broader point is that mileage matters for rage people to buy electric vehicles. California, for electric vehicle policy. From a climate change per- example, aims to have 1.5 million electric vehicles spective, the goal is to reduce gasoline consumption. on the road by 2025, and 5 million electric vehicles But current policies like the $7,500 federal tax credit on the roads by 2030 (California Office of the treat low-mileage and high-mileage drivers uni- Governor 2018). formly. A better approach would be a carbon tax How much electric vehicles are driven has or, equivalently, increasing the gasoline tax. Making major implications for the effectiveness of these gasoline more expensive would incentivize electric policies. In particular, the less electric vehicles are vehicles for all drivers, but the biggest incentive driven, the smaller the environmental benefits would be for people who drive a lot of miles. from electric vehicle adoption. Take the following example. Holland et al. (2016) finds that the environmental benefits from Acknowledgments replacing a gasoline vehicle with an electric vehicle I am grateful to Severin Borenstein, Chris Knittel, Jing Li, in California are worth $2,800 over the lifetime of Andrew Yates, and seminar participants at UC Berkeley for the vehicle. The authors make this calculation helpful comments. assuming that all vehicles are driven 15,000 miles per year. If, instead, each electric vehicle is avoiding only 7,500 miles per year driven in a gasoline vehi- Disclosure statement cle, then the benefits are half as large, only $1,400. I have not received any financial compensation for this The assumption by Holland et al. (2016)that project nor do I have any financial relationships that relate vehicles are driven 15,000 miles per year is not to this research.

5Income differences, in contrast, are unlikely to explain the pattern. Previous research has shown that electric vehicle ownership is strongly correlated with income. Borenstein and Davis (2015), for example, shows using data from U.S. income tax returns, that the top income quintile has received 90% of all electric vehicle tax credits. But high-income drivers tend to drive more than other households (Bento et al. 2005; Levinson forthcoming), so if anything income differences tend to lead electric vehicles to be driven more. 1502 L. W. DAVIS

ORCID mer Choice.” Journal of Environmental Economics and Management 54 (2): 129–145. doi:10.1016/j.jeem.2007.05.001. Lucas W. Davis http://orcid.org/0000-0002-6801-9387 Leard, B., and V. McConnell. 2017. “New Markets for Credit Trading under US Automobile Greenhouse Gas and Fuel Economy Standards.” Review of Environmental Economics References and Policy 11 (2): 207–226. doi:10.1093/reep/rex010. “ Allcott, H., and N. Wozny. 2014. “Gasoline Prices, Fuel Levinson, A. forthcoming. Energy Efficiency Standards are Economy, and the Energy Paradox.” Review of Economics More Regressive than Energy Taxes: Theory and ” and Statistics 96 (5): 779–795. doi:10.1162/REST_a_00419. Evidence. Journal of the Association of Environmental Archsmith, J., K. Gillingham, C. R. Knittel, and D. S. Rapson. and Resource Economists. doi:10.1023/A:1008324605045. “ 2017. “Attribute Substitution in Household Vehicle Li, J. 2018. Compatibility and Investment in the US Electric ” Portfolios.” National Bureau of Economic Research Vehicle Market. MIT Working Paper. “ Working Paper #23856,September. Li, S., L. Tong, J. Xing, and Y. Zhou. 2017. The Market for Archsmith, J., A. Kendall, and D. Rapson. 2015. “From Cradle Electric Vehicles: Indirect Network Effects and Policy ” to Junkyard: Assessing the Life Cycle Greenhouse Gas Design. Journal of the Association of Environmental and – Benefits of Electric Vehicles.” Research in Transportation Resource Economists 4 (1): 89 133. doi:10.1086/689702. “ Economics 52: 72–90. doi:10.1016/j.retrec.2015.10.007. McDonald, Z. 2016. How Long Does It Take To Recoup ” Bento, A. M., L. H. Goulder, E. Henry, M. R. Jacobsen, and The Extra Cost Of An Electric Car? Fleetcarma. June. R. H. Von Haefen. 2005. “Distributional and Efficiency Michalek,J.J.,M.Chester,P.Jaramillo,C.Samaras,C.- “ Impacts of Gasoline Taxes: An Econometrically Based S. N. Shiau, and L. B. Lave. 2011. Valuation of Plug-In Multi-Market Study.” American Economic Review 95 (2): Vehicle Life-Cycle Air Emissions and Oil Displacement ” 282–287. doi:10.1257/000282805774670536. Benefits. Proceedings of the National Academy – Borenstein, S. 2015. “A Microeconomic Framework for of Sciences 108 (40): 16554 16558. doi:10.1073/ Evaluating Energy Efficiency Rebound and Some pnas.1104473108. Implications.” Energy Journal 36 (1): 1–21. Notter, D. A., M. Gauch, R. Widmer, P. Wager, A. Stamp, “ Borenstein, S., and J. B. Bushnell. 2018. “Do Two Electricity R. Zah, and H.-J. Althaus. 2010. Contribution of Li-Ion Pricing Wrongs Make a Right? Cost Recovery, Batteries to the Environmental Impact of Electric ” Externalities, and Efficiency.” National Bureau of Vehicles. Environmental Science & Technology 44 (17): – Economic Research Working Paper # 24756, June. 6550 6556. doi:10.1021/es903729a. “ Borenstein, S., and L. W. Davis. 2015. “The Distributional Sallee, J. M., S. E. West, and W. Fan. 2016. Do Consumers Effects of U.S. Clean Energy Tax Credits.” In Tax Policy Recognize the Value of Fuel Economy? Evidence from ” and the Economy. Vol. 30, 191–234. University of Chicago Used Car Prices and Gasoline Price Fluctuations. – Press. Journal of Public Economics 135: 61 73. doi:10.1016/j. Busse, M. R., C. R. Knittel, and F. Zettelmeyer. 2013. “Are jpubeco.2016.01.003. “ Consumers Myopic? Evidence from New and Used Car Sivak, M., and B. Schoettle. 2018. Relative Costs of Driving Purchases.” American Economic Review 103 (1): 220–256. Electric and Gasoline Vehicles in the Individual US ” doi:10.1257/aer.103.1.220. States. Working Paper. California Office of the Governor. 2018. Governor Brown Tamayao, M.-A. M., J. J. Michalek, C. Hendrickson, and “ Takes Action to Increase Zero-Emission Vehicles, Fund I. M. L. Azevedo. 2015. Regional Variability and New Climate Investments, Sacramento, CA. Uncertainty of Electric Vehicle Life Cycle CO2 Emissions ” Davis, L. W. forthcoming. “Evidence of a Homeowner-Renter across the United States. Environmental Science & – Gap for Electric Vehicles.” Applied Economics Letters. Technology 49 (14): 8844 8855. doi:10.1021/acs.est.5b00815. “ Dubin, J. A., and D. L. McFadden. 1984. “An Econometric Tessum, C. W., J. D. Hill, and J. D. Marshall. 2014. Life Analysis of Residential Electric Appliance Holdings and Cycle Air Quality Impacts of Conventional and Consumption.” Econometrica 52 (2): 345–362. doi:10.2307/ Alternative Light-Duty Transportation in the United ” 1911493. States. Proceedings of the National Academy of Sciences – Hausman, J. A. 1979. “Individual Discount Rates and the 111 (52): 18490 18495. doi:10.1073/pnas.1406853111. Purchase and Utilization of Energy-Using Durables.” Bell U.S. Department of Transportation. 2018. National Household Journal of Economics 10 (1): 33–54. doi:10.2307/3003318. Travel Survey. Washington DC: Federal Highway Holland,S.P.,E.T.Mansur,N.Z.Muller,andA.J.Yates.2016. Administration. “Are There Environmental Benefits from Driving Electric Zivin, J. S. G., M. J. Kotchen, and E. T. Mansur. 2014. “ Vehicles? the Importance of Local Factors.” American Spatial and Temporal Heterogeneity of Marginal Economic Review 106 (12): 3700–3729. doi:10.1257/ Emissions: Implications for Electric Cars and Other ” aer.20150897. Electricity-Shifting Policies. Journal of Economic – Kahn,M.E.2007. “Do Greens Drive Hummers or Hybrids? Behavior & Organization 107: 248 268. doi:10.1016/j. Environmental Ideology as a Determinant of Consu- jebo.2014.03.010.