2. Theoretical basis and hypothesis

Our objective is to determine how the financial benefits and costs of the CVRP vary by income class. During the study, we make the following assumptions: (i) the only funding source for the CVRP program in California is from the cap-and-trade program; (ii) the government revenue from the California cap-and-trade program solely depends on energy consumption; and (iii) the zip code mean income represents the mean income for rebate receivers. We also focus on the California CVRP program and do not consider other electric vehicle policies, such as charging station distribution and charging rates. With these assumptions, our results focus on the CVRP program’s redistribution effects, but not the general electric vehicle equity problem.

We evaluate the distributional impact of policies by exploring the ratio of benefits to costs in different income groups. Progressiveness occurs when the ratio of benefit to income declines with income or when the ratio of cost to income rises with income. Regressiveness implies the opposite patterns. In contrast to a complete benefit-cost analysis, we introduce a public finance perspective and focus on the financial redistribution of the CVRP program. Thus, we do not consider the effectiveness of the CVRP program in stimulating electric vehicle sales or its environmental benefits and risks. By focusing on the financial redistribution effect, this study provides a straightforward overview of how the CVRP program reallocates income in California.

The main index of interest, therefore, is the net benefit-cost to income ratio at the zip code level. In our study, we use the following equation to calculate the net benefit to income ratio:

(1) $$ rati{o}_i=\left( benefi{t}_i-\mathit{\cos}{t}_i\right)/ incom{e}_i $$

In Equation (1), the ratio is the net benefit to income ratio in zip code $ i $ . The benefit is zip code $ i $ ’s benefits from the CVRP program. The cost is zip code $ i $ ’s costs from the CVRP program. The income is zip code $ i $ ’s median income. If the ratio is greater than zero, then zip code $ i $ ’s benefits are more than the costs from the CVRP program. If the ratio is less than zero, then zip code $ i $ ’s benefits are less than the costs from the CVRP program.

The funding source and mechanism of the CVRP program also suggest how its benefits and costs distribute among income groups. The funding source of the CVRP program distributes among all residents, because its funding source, the California cap-and-trade program, charges all residents who consume electricity and gasoline. Meanwhile, only residents buying electric vehicles receive rebates. The benefits should be expected to concentrate on high-income consumers, given that purchasers of new PEVs are predominantly high-income consumers. For example, a recent trade survey found that the average Tesla buyer’s income was $503,000 for the Model X, $267,000 for the Model S, and $160,000 for the new Model 3 (Prenzler, Reference Robinson2016, Reference Rubin and St-Louis2017). Therefore, we expect that the benefits of the CVRP program will be regressive but the distributions of costs and net benefits are not obvious without careful study.

3. Data and methods

We then consider the data and methods concerning financial benefits and then the data and methods concerning financial costs.

4. Results of unconstrained program

Table 1 summarizes the California data on rebates received and mean income, reporting the mean rebate at the individual level and the mean income at the zip code level. Among those receiving a rebate, the average rebate was about $2200. The mean income was about $95,500. The average ratio of rebate to population was about $3.9 per 100 persons.

Table 1.

Summary statistics for California rebate program.

Note: Reported is the mean of each variable with standard deviations in parentheses.

Figure 1 displays on a map of California the distributions of zip codes benefiting and losing due to the CVRP. The map provides a visual indication of how the two quantities are related. In the map, those zip codes receiving more rebates than costs are marked as red, and those zip codes contributing fewer rebates than costs are marked as blue. Those zip codes without rebate data are left in white. Although Figure 1 does not include the income category, it reveals that only the high-income regions of San Francisco and Los Angeles, and a few other regions tend to receive more rebates than costs. The rest of the areas in the state, which tend to have lower median incomes, tend to have smaller rates of rebates compared to costs.

Figure 1.

Net financial impact of the CVRP program by zip code.

The income distribution of the cap-and-trade program’s mean cost is compared to the income distribution of the mean rebate in Figure 2. In this figure, income, rebate received and rebate costs are averaged at the zip code level. The right side of Figure 2 reports the distribution of rebate percentage received by each income category while the left side reports the percentage distribution of the cap-and-trade cost by income (Jones & Kammen, Reference Langbroek, Franklin and Susilo2014). The income intervals are chosen to ensure that each category has the same population size.

Figure 2.

Distribution (%) of financial burden and financial rebate as a share of income, by zip code.

Note that the distribution of financial costs among income classes is different from the distribution of financial benefits. The financial benefits of the rebates are concentrated in the high-income categories while the low-income categories receive virtually no benefit from the rebates. To be more specific, the group with the highest 10 % of income accounts for 11 % of the total costs but receives 33 % of the total rebates.

Figure 2 also shows that the rebate program’s costs are distributed relatively evenly across income categories, with a slight pattern of progressiveness. The highest average financial cost of the program is reported for people with incomes between 70 and 80% percentile. Overall, the financial costs appear somewhat larger in the high-income categories than in the low-income categories.

Table 2 shows that the ratio of net benefit to income increases from negative to positive as income increases. The sign of net benefit shows whether a specific income group receives more benefits than the costs. If the sign is positive, then the income group receives more rebates than costs. Both mean benefits and costs increase as income increases, but the mean benefits only surpass the cost for people with incomes higher than $66,670, and the benefits continuously increase with income. The increasing ratio indicates that the net benefits of the rebate program are concentrated in the highest income groups.

Table 2.

Net benefit to income ratio.

Table 3 reports the correlations, derived from simple linear regression, between mean income, and mean financial cost and mean financial benefit of the CVRP. Previous studies also suggest household demographics (i.e., the number of children and number of household vehicles) and consumer demographic information (i.e., race, household age, education, housing type, and political attitudes) (DeShazo et al., 2017) are correlated with buying electric vehicles. However, due to the data limitations, we include only income and race in the analysis. A more complete dataset might need a survey specific to electric vehicle buyers.

Table 3.

Statistical predictions of mean cost and mean rebate.

Note: t statistics in parentheses.

* p < 0.05;

^** p < 0.01;

^*** p < 0.001.

The result shows the magnitude of the associations between the variables, but we do not aim to test causality or make predictions. We also include population density and race into the analysis, as previous literature suggests (Jones & Kammen, Reference Langbroek, Franklin and Susilo2014; Muehlegger & Rapson, Reference Neubauer and Wood2019). Structured similarly to models 1 and 2, models 3 and model 4 include the population density and race as control variables.

The results of models 1 and 2 show that zip codes with higher incomes both contribute more to and receive more from the rebate program. The mean income coefficient of model 3 is a little bit lower than that of model 1, as Jones and Kammen (Reference Langbroek, Franklin and Susilo2014) suggested, but the conclusion about the significance of the coefficient does not change. Therefore, the previous implication is maintained. The difference between models 2 and 4 is minor, showing that including the population density variable does not change the basic finding.

In models 1 and 3, the mean income coefficient is significant and positive, even though the coefficients are small. The coefficient magnitudes mean little because income in the original surveys was categorical. Instead, the scale matters: the significant p-values show a high correlation, implying that the higher income zip codes contribute significantly more to the rebate program through the California cap-and-trade program. Meanwhile, in models 2 and 4, the income coefficients are also small and are also significant and positive. The significant p-values here mean that, on average, the higher-income people receive more rebates.

By comparing the results of the two estimates, we see that the positive average income coefficients for receiving rebates are 90–100 times that for contributing to the program. The ratio shows that, although higher income groups contribute somewhat more to the rebate program as income increases, the benefits received from the program increase far more rapidly with income. One reason for this result is that the cap-and-trade program’s financial costs are shared by all residents in California (through higher prices for products), but the benefits of receiving the rebates are concentrated among wealthy consumers to buy new PEVs, a highly expensive consumer product. Therefore, the results of the estimates and coefficients suggest the same conclusion as the visual inspection of Figure 2: the rebate program’s distributional impacts are regressive.

4.1 Comparing before and after income restrictions

In 2016, there was a major change in the CVRP that sought to address concerns about income inequity. Before the change, there was no income restriction on receiving the clean vehicle rebate. After 29 March 2016, two stages of income ceilings were imposed on the rebate, as well as two stages of additional rebates for low-income consumers. From that date to 31 October 2016, the income ceiling was $250,000 for a single filer and $1500 was added to the rebate for lower-income buyers. After 31 October 2016, the income ceiling was lowered to $150,000 and the additional rebate was raised to $2500.

The previous analysis did not consider the change of policy, and therefore it is reasonable to consider how much the new policy changed the impacts on income distribution. To examine whether our previous conclusion is robust to the change, the data used in Figure 1 are split into two parts: before 29 March 2016, and after 29 March 2016. After splitting, the dataset for observations before the policy changes contains 9929 observations while the dataset after the policy changes contains 2763 observations.

Figures 3 and 4 display the income distribution of both the rebate contribution ratios and the rebate received ratios, before (Figure 3) and after (Figure 4) the policy changed. The distribution of the contribution to the rebate program before the policy change is more like that of Figure 2, with 39 % of rebates flowing to the highest income groups. This is not surprising because most of the observations in the main dataset were collected before the policy changes. Since the income interval has been chosen to even the population within each category, the scales are somewhat different. After the income constraint was imposed, the ratio decreased to 25 %.

Figure 3.

Distribution (%) of financial burden and financial rebate as a share of income, by zip code, before income restrictions.

Figure 4.

Distribution (%) of financial burden and financial rebate as a share of income, by zip code, after income restrictions.

Despite the policy change, the patterns displayed in Figures 3 and 4 suggest similar conclusions to those drawn from Figure 2: the cost of the rebate program derived from the California cap-and-trade program is relatively evenly distributed among different income groups while the benefits of the rebate program concentrate on the highest income group. Groups other than the highest income group obtain almost no benefits from the rebate program. The cost of the rebate policy is relatively evenly borne among different income groups and actually increases slightly as income increases.

Tables 4 and 5 list the ratios of net benefit to income, before (Table 4) and after (Table 5) income ceiling was included. The income intervals are different in Tables 4 and 5, showing the change in income in California. However, the mean net benefit to income ratio increases with income in both Tables 4 and 5. This means that including the income ceiling does not change the fact that the rebate policy predominantly benefits individuals with higher incomes.

Table 4.

Net benefit to income ratio before income ceiling.

Table 5.

Net benefit to income ratio after income ceiling.

Tables 6 and 7 examine the correlations among mean income, meaningful contribution to the rebate program, and mean rebate received from the program using different emission estimations. Models 1 and model 2 examine the correlation between income, mean rebate received, and mean rebate contribution individually. Models 3 and 4 also include population density as a control variable.

Table 6.

Statistical predictions of mean cost and mean rebate, before income restrictions.

Note: t statistics in parentheses.

* p < 0.05;

^** p < 0.01;

^*** p < 0.001.

Table 7.

Statistical predictions of mean cost and mean rebate, after income restrictions.

Note: t statistics in parentheses.

* p < 0.05;

^** p < 0.01;

^*** p < 0.001.

The results of Tables 6 and 7 show similar results to that of Table 3. In general, people with higher incomes bear more costs and receive more benefits from the rebate program. The mean income coefficients in model 1 in Tables 6 and 7 are lower than those in Table 3. However, the conclusion regarding the significance of the coefficient does not change. In other words, even after the policy change, higher-income people still received more rebates.

In both tables, mean income coefficients are significant and positive. Again, although the magnitude of these coefficients should not be interpreted numerically, their scales are still comparable. The significant p-values also mean that, on average, the higher-income people pay more to, and receive more rebates from, the rebate program.

The ratio of benefits to costs from the rebate program is also similar to that found in Table 3. The positive income coefficients for receiving rebates are approximately 150 times those of the cost from the program between model 3 and model 4 in Table 6, and the number grows to 276 times between model 3 and model 4 in Table 7. The ratio means that, although the high-income groups contribute more to the rebate program, the benefits received from the program rise far faster as income rises. Therefore, the inference that higher-income people receive more from the rebate program is the same and, thus, is robust to the changes in policy.

5. Discussion: potential policy responses

The results of the analysis indicate that the current CVRP has a strongly regressive distribution of financial benefits and a modestly progressive distribution of financial costs. Although the rebate program later included an income ceiling for those receiving the rebate and an increased rebate for lower-income groups, the policy change has not changed the fact that the distribution of CVRP financially benefits the high-income groups. The reason that imposing income restrictions does not change the distribution of CVRP benefits could be that the rebate does not compensate enough for high PEV prices in lower-income groups and a lack of public charging infrastructure discourages low-income consumers more than high-income consumers. Low-income groups are more sensitive to vehicle price (Muehlegger & Rapson, Reference Muehlegger and Rapson2018) and less likely to own their houses (Sierzchula et al., Reference Yang, He, Xia and Chen2014).

In addition to the CVRP’s direct redistribution effect, the program has also helped wealthier people access cheaper transportation. According to the comparison tool developed by the U.S. Department of Energy, the national average price for one gallon of gasoline is $2.84, while the cost for one e-gallon is $1.19 (EGallon, Reference Freire-González2018). (California has higher prices for gasoline and electricity than the national average). Thus, including the income distribution of vehicle operating costs would likely have made the CVRP financial benefits appear even larger than they are without considering energy costs.

The CVRP and its income redistribution could be justified by the social gains from reducing greenhouse gas emissions by encouraging the use of ZEVs. However, studies show that the social gains from reducing emissions are not as large as would be expected by the difference in emissions between the average car and a ZEV (Archsmith et al., Reference Archsmith, Kendall and Rapson2015; Jenn et al., Reference Lambert2016; Muehlegger & Rapson, Reference Parry and Small2020; Xing et al., Reference Xing, Leard and Li2021). The buyers for ZEVs tend to replace fuel-efficient vehicles with ZEVs, so the emission reduction is smaller than average (Muehlegger & Rapson, Reference Parry and Small2020; Xing et al., Reference Xing, Leard and Li2021). Furthermore, the current ZEV policy also encourages manufacturers to produce less-efficient vehicles, due to the compliance averaging in regulatory programs (Jenn et al., Reference Lambert2016).

On the other hand, both the Random Utility Model estimation (Langbroek et al., Reference Muehlegger and Rapson2016) and the recent experience from the state of Georgia show the rebate program’s importance to the commercial viability of the electric vehicle industry. For several years Georgia stimulated PEV sales with a generous PEV rebate of $5000. When Georgia canceled the financial incentive for PEV sales, the volume of PEV sales fell significantly (Lambert, Reference Ming, Ximei, Yulong and Lilin2017). Thus, insofar as the rebates’ main policy goal is to increase PEV sales, efforts to enhance the equity of rebate design need to be analyzed carefully.

The most obvious modification to increase equity would be to increase the rebate that lower-income people receive. Since low-income consumers are more price-sensitive (Muehlegger & Rapson, Reference Muehlegger and Rapson2018) and more likely to purchase used cars than new cars (Muehlegger & Rapson, Reference Neubauer and Wood2019), it may be worthwhile to consider rebates for purchases of used PEVs. California’s efforts to help low-income people purchase new PEVs may be misguided because new PEVs are quite expensive, well above the $35,000 average transaction price for a new light-duty vehicle in 2017. On the other hand, the current federal policy of restricting federal tax credits to new PEVs has had the indirect effect of reducing used PEVs’ resale value, which in turn has made used PEVs more affordable than they otherwise would be. Thus, efforts to make PEVs more affordable for low-income consumers need to be analyzed with care.

Although the average prices of new PEVs have decreased in recent years, the savings on fuel and rebates for the owner still may not be large enough to cover the initial price premium. This factor is especially significant for the lower and middle-income groups, as they access the used car market where consumers are more sensitive to prices, rebates, and operating expenses (DeShazo et al., 2017). Affordable leasing arrangements for used or new PEVs might be an attractive option for some low-income and middle-income households.

Another policy alternative is to focus rebates on the types of vehicles that lower-income consumers purchase. Rebating on luxury BEVs such as the Tesla Model X does not make the BEV more accessible to lower and middle-income groups, yet Tesla buyers have been a major beneficiary of federal and state incentives (Hardman et al., Reference Hassett, Mathur and Metcalf2017). Since the lower and middle-income consumers are usually in the market for nonluxury vehicles, and since the average ratio of rebate to vehicle price is higher for nonluxury vehicles, policymakers should consider focusing the rebates on nonluxury vehicles. As this article went to press, California was considering more equity-related changes to the CVRP (Carpenter, Reference Carpenter2021).

Another aspect worth considering is whether PEVs can meet the daily needs of lower-income drivers. Although PEVs have much lower energy costs than gasoline vehicles, only high-income households are able to utilize the strength of PEVs while minimizing the difficulties. First, high-income households tend to have two or more vehicles while low-income households may have only one vehicle. The high-income household can buy a PEV that will be utilized only for short-range or inner-city driving and that requires 2–4 hours to charge and receive a rebate. Low-income households are less likely to have more than one vehicle and, thus, that vehicle must have the capability to meet all of the household’s needs. In this respect, the short-range BEV is a less attractive option than a PHEV or even an HEV that does not require lengthy charging (Van Haaren, Reference Sandler and Pezzullo2011; Graham-Rowe et al., Reference Grainger and Kolstad2012; Neubauer & Wood, Reference Prenzler2014; Xing et al., Reference Xing, Leard and Li2021). Secondly, despite the mass investment in public charging infrastructure, house renters still have difficulties in charging their PEVs. Therefore, increasing the charging infrastructure in low-income areas (Sierzchula et al., Reference Yang, He, Xia and Chen2014), targeting the rebate on vehicles meeting a certain driving range, and enforcing recharging standards might also attract lower and middle-income buyers. China is rapidly gaining experience in commercializing affordable PEVs by deploying public infrastructure in advance and, thus, the Chinese experience should be examined for its relevance to other settings.

In addition to lowering the price of new PEVs, changing the funding source for the rebate program also helps to increase equity. The resource redistribution among different income groups raises equity concerns for PEV rebates. Our results show that the middle-income groups pay for most of the rebates, but the highest 10 % income group receives most of the rebates. If the tax base for the CVRP program changes to a capital-gains tax, the cost of the CVRP program would be distributed more heavily in higher-income groups that experience the lion’s share of capital gains.