Global Policy Forum

Distributional Aspects of an Environmental Tax Shift: The Case of Motor Vehicle Emissions Taxes


By Margaret Walls & Jean Hanson

National Tax Journal
March 1, 1999

Policymakers are beginning to appreciate the efficiency properties of pollution taxes, but concerns over equity remain. In this paper, we assess the distributional aspects of three types of emissions fees for vehicles--a fee based on total annual emissions; a fee based on emissions rates, in grams per mile; and a fee based on annual vehicle mileage--and we compare the impacts of fees to existing vehicle registration fees. We find that all three of the emissions-related fees look quite a bit more regressive, on the basis of annual household income, than do existing registration fees. However, when we construct a lifetime income variable for our households, we find that the differential impacts of the three emissions-related fees over existing fees are quite small.

Policymakers are currently viewing with favor economic incentive approaches to reducing pollution, yet enthusiasm for such approaches on efficiency grounds is usually tempered by concerns over equity. This is especially the case for motor vehicle emissions policies. We explore this issue in this paper and address, in particular, the question of whether a judicious return of the revenues raised from emissions fees might alleviate the equity concerns. We examine a policy whereby current vehicle registration fees, which are based on the value of a vehicle, are replaced by one of three emissions-related fees--a fee based on annual mileage, a fee based on total annual emissions, and a fee based on emissions rates, in grams per mile.

We use a microlevel data set on households in California that contains information on make, model, model year, and annual mileage for each vehicle in each of 978 households. We combine these data with information on emissions from a remote sensing experiment. We use both annual household income and a measure of lifetime income and compare the incidence of the current California registration fee system to our three alternative emissions-related fees.

We find that all three of the alternative fees look quite a bit more regressive, on the basis of annual household income, than do existing registration fees. However, when we construct a lifetime income variable for our households, we find that the differential impacts of the three emissions-related fees over existing fees are quite small. Households in the lowest lifetime income decile pay only $ 14 more per year, on average, with the fee based on total emissions compared with the current value-based fee.


Regulation of emissions from motor vehicles has relied for the most part on traditional "command-and-control" approaches, with new car emissions standards and inspection and maintenance (I&M) programs being the dominant strategies. These approaches have been criticized by many observers as inflexible, poorly targeted to actual in-use emissions, not compatible with motorists' incentives, and ultimately very costly ways to reduce pollution (Harrington, Walls, and McConnell, 1995; Kessler and Schroeer, 1993). Several economic incentive approaches have been suggested as alternatives. One such approach that looks particularly promising is charging vehicle owners an annual fee based on emissions. Recent research suggests that such an approach could yield net welfare benefits several times those from an I&M program in which all vehicles must be maintained to the same standards (Harrington, McConnell, and Alberini, 1998).

Vehicle emissions fees have another advantage: the revenues they generate could be used to reduce other taxes. This advantage of a tax approach over command-and-control--or even over economic incentive approaches that do not raise revenues, such as (non-auctioned) marketable pollution permits--is becoming widely recognized. Any environmental policy will interact with pre-existing taxes in the economy and increase the distortions from those taxes (Bovenberg and DeMooij, 1994; Goulder, 1995; Parry, 1995; Bovenberg and Goulder, 1996), but, unique among environmental policies, a tax raises revenues that can be used to offset some of the distortions. Its revenue-raising potential might also make the environmental tax more politically palatable. There are several potential taxes that could be reduced, but reducing a vehicle-related tax, such as a registration fee, might be the best option for gaining public acceptance. This type of "linked compensation" for the losers from corrective tax policies has been suggested as a way of both gaining approval of the tax and promoting equity (Burtraw, 1991; Small, 1992).

Of course, any tax substitution will produce winners and losers--some vehicle owners would pay more to register their vehicles under an emissions-based system than under existing systems and some would pay less. If the poor end up paying more, it might doom the policy--or at least lead policymakers to search for ways to redistribute tax revenues. One of the most common criticisms of pollution taxes in general is that they are often believed to be inequitable; i.e., low-income households are thought to be disproportionately harmed. This seems to be especially important for vehicle-related taxes. In every debate over increasing the gasoline tax, numerous arguments are made about the detrimental impact on the poor.

In this study, we assess the distributional impacts of vehicle emissions fees. We do this using household data on the make, model, model year, and annual mileage of each vehicle in each of 1,018 California households; we combine this household information with model year average emissions data from a California remote sensing experiment. We look at the incidence of substituting revenue-neutral emissions fees for California's existing vehicle value-based registration fee.

To assess the distributional impacts, we use two measures of economic well-being: annual household income and lifetime income, a variable that we construct from information on households' education levels and other variables. We calculate current registration fees as a fraction of annual income and lifetime income; we then look at three alternative revenue-neutral fees--a fee based on annual vehicle-miles-traveled (VMT), a fee based on emissions rates in grams per mile (g/mi), and a fee based on total emissions (g/mi multiplied by VMT). Although only the last of these is a true emissions fee, the other two could have other features that make them worth considering, such as lower administrative costs. We discuss these points below.

This paper does not assess the efficiency advantages of emissions fees. We briefly discuss other research on this issue, but focus our attention here on equity. Since concerns about equity are often a major stumbling block for pollution taxes, we feel that our work sheds light on an important issue.

In the following section, we discuss alternative approaches to assessing tax incidence based on annual income, annual total consumption expenditures, and lifetime income. We briefly discuss some findings in the literature and end showing our lifetime income variable and how it compares with annual income. In the third section, we describe our household data and the emissions data that we merge with it. In the fourth section, we discuss the different types of environmental registration fees that we analyze, calculate our revenue-neutral fees, and briefly summarize other research on the efficiency of emissions fees and the optimal level of such fees. In the fifth section, we show our distributional findings for both annual and lifetime income. The final section draws some overall conclusions.

Tax Incidence and Lifetime Income

Economists have long argued that using annual income as a basis for determining tax incidence is problematic because of the tendency for individuals to consume based on permanent income, or earnings over their life cycle (Friedman, 1957; Modigliani and Brumberg, 1954). Most people tend to earn their highest incomes around middle age and their lowest incomes when they are young or old. Grouping people by annual income using cross-sectional data will lead to some young and old people in the lower income groups who may not belong there on the basis of lifetime income. Likewise, higher income groups will include some middle-aged people who belong in a lower lifetime income category. Moreover, differences in annual income are often the result of transitory components that should have smaller effects on consumption than differences in permanent income.

Several studies have compared the incidence of various taxes on a lifetime versus annual income basis. Poterba (1989, 1991) and Metcalf (1993) use annual total consumption expenditure as a proxy for lifetime income. Poterba shows that taxes on gasoline, alcohol, and tobacco appear to be much less regressive when viewed as a percentage of total consumption expenditures rather than as a percentage of annual income. Metcalf finds that sales taxes appear to be equally as progressive as income taxes and property taxes appear to be approximately proportional to consumption expenditures.

Fullerton and Rogers (1993) calculate actual lifetime incomes for a sample of households using an 18-year span of data from the Panel Study of Income Dynamics (PSID). They categorize households based on this lifetime income variable and use a computable general equilibrium model to assess the lifetime tax burden of several different taxes. Lyon and Schwab (1995) use the same data to look at the incidence of alcohol and cigarette taxes. Fullerton and Rogers find that all of their taxes look less regressive on a lifetime income basis. They note that in fact, "all distributional effects of taxes are likely to be muted in the lifetime context" (p. 19). Lyon and Schwab find very little difference in their annual income and lifetime income results. The approach taken in these two studies is only feasible with panel data, which we do not have. We also do not have data on annual total consumption expenditures to carry out the Poterba approach.

Rogers (1993) and Casperson and Metcalf (1994) present interesting alternatives for cross-sectional data sets. Both studies use the PSID to estimate relationships between various demographic variables and lifetime income. They then apply those estimated relationships to a separate cross-sectional data set. Both studies compute incidence by assessing annual tax payments as a fraction of lifetime income. In other words, they use their calculated lifetime incomes to better categorize each household's ability to pay, but they still calculate an annual rather than a lifetime tax burden. Both studies find that the taxes they look at are less regressive on a lifetime income basis than on an annual income basis but more regressive than when viewed as a fraction of annual consumption.(1)

In our study, we adopt Rogers' approach and apply her regression results to our cross-sectional data set. Rogers relies on her earlier work with Fullerton, regressing the calculated lifetime incomes for their PSID households--i.e., the present value of annual income over the lifetime of the household--against education level, education squared, and interactions between education and dummy variables for whether the household is married, white or female-headed.(2) We use her regression results to predict lifetime income for each household in our data set. We then annualize that lifetime income by computing a 60-year constant annuity using a four percent real interest rate.

Our approach has some limitations. First, unlike the PSID, our data set does not identify a household "head," only a "reference person." We assume the reference person is the head unless it is a married household and the reference person is female; in this case, we use information on the male spouse (to be consistent with the PSID, which always assumes that the male is the head of the household). Second, we have a number of unmarried group households in our sample. As we stated in the previous footnote, Rogers uses the average income of the husband and wife in a married household; thus, we must double our calculated lifetime incomes for married households to get a more accurate measure of household income. Although it is less than ideal, we do the same for group households. Third, Rogers' lifetime income variable is a potential lifetime income, calculated assuming everyone works the total number of hours available in each year (see Fullerton and Rogers (1993) for an explanation). Because we want to directly compare our results with results using annual income, we adjust the predictions we get from her regression by multiplying each predicted value by the ratio of the sum of annual incomes for our households to the sum of the predicted lifetime incomes from the Rogers regression. This ensures that the mean adjusted annualized lifetime income is equal to the mean annual income for our sample. The adjustment does not affect the relative position of the households, but it is a rather ad hoc adjustment that is necessary because we do not know the number of hours worked by our households. Fourth, there is a general problem faced by anyone calculating lifetime incomes for a cross section of households: one has virtually no choice but to assume that the status of the household in the cross section has held and will continue to hold throughout its lifetime. In other words, a recently widowed elderly person will mistakenly be counted as single and a sophomore in college who eventually earns a degree will mistakenly be assigned 13 years of education. Without panel data, there is no good solution to this problem. Finally, individual fixed effects can be an important determinant of annual and lifetime income (Fullerton and Rogers, 1993). When creating a lifetime income variable, ideally, one would like to smooth out the transitory components in annual income but leave in the fixed effects. However, without panel data, this is not possible. This means that our measure of lifetime income probably overstates true lifetime income for some of our households and understates it for others.

Despite these shortcomings, our measure of lifetime income seems reasonable. Most of the differences across households are explained by differences in education.(3) Table 1 shows the resulting lifetime income figures as well as annual income by decile. (It is important to note that the deciles are created differently in the two cases--i.e., households are sorted by annual income to obtain annual income deciles and by lifetime income to obtain lifetime income deciles.) The lifetime income has a mean identical to the mean of annual income (by construction), but the standard deviation is much smaller than the standard deviation of annual income. This is expected and is similar to the findings in virtually all studies of lifetime income (Lillard, 1977; Blomquist, 1981; Davies, St-Hilaire, and Whalley, 1984; Fullerton and Rogers, 1993). Fullerton and Rogers, for example, obtain a coefficient of variation for household lifetime income of 0.456 compared to 1.838 for annual income. Our corresponding numbers, based on the means and standard deviations in Table 1, are 0.384 and 0.936. Also, as in previous studies, the distribution of lifetime income is less unequal than that of annual income. The lowest decile households in the Davies, St-Hilaire, and Whalley study, for example, earn 1 percent of the annual income earned by all households in their sample and 4.2 percent of the lifetime income; the corresponding percentages for households in the highest decile are 30.6 and 18.4. These figures are very close to our results reported in Table 1.

Table 1: Annual and Lifetime Income by Decile(a)

Annual Income (AI)
Decile Mean Percentage of Total AI
1 $ 6,844 1.7%
2 13,681 3.3
3 19,940 4.8
4 26,883 6.4
5 34,506 8.2
6 40,097 9.6
7 49,096 11.8
8 56,889 13.7
9 68,507 16.4
10 100,399 24.1
All $ 40,791 100%
Standard Deviation $ 28,198 --
Lifetime Income (LI)(b)
Decile Mean Percentage of Total LI
1 $ 16,579 4.0%
2 22,499 5.3
3 30,227 7.2
4 35,457 8.5
5 40,869 9.8
6 43,384 10.6
7 46,771 11.3
8 51,325 12.3
9 59,210 14.3
10 69,308 16.7
All $ 40,791 100%
Standard Deviation $ 15,654 --

(a) Each decile contains ten percent of households; the deciles differ for the two income measures; for example, decile 1 includes the poorest ten percent of households on an annual income basis for purposes of computing annual income measures and the poorest ten percent on a lifetime income basis for purposes of computing lifetime income measures.

(b) The figures for lifetime income are annualized adjusted lifetime income (see text).

The California Data on Vehicles, Emissions, and VMT

We use the U.S. Department of Transportation's 1990 Nationwide Personal Transportation Survey (NPTS) as the basis of our analysis. The NPTS is a large survey of randomly selected U.S. households; it contains information on the make, model, and model year of each vehicle in each household, as well as a host of socioeconomic and demographic information about each household.(4) We choose to focus our empirical work on California, because we have a fairly large number of households to work with and because its registration fee system is straightforward and has a feature typical of many states in that fees are a function of vehicle values.(5)

Table 2 shows some summary information about the California households and their vehicles. There are 1,018 households in the sample and they own an average of 1.79 vehicles each. This includes some households--6.2 percent--who own no vehicles. Annual registration fees in California in 1990 were two percent of vehicle value plus a flat annual fee of $ 25. Average fees in 1990 amounted to about 0.68 percent of annual income and a slightly lower percentage of lifetime income, 0.56. This is a small fraction of income but registration fees in California are still a fairly sizable portion of the cost of owning and operating a vehicle. The average vehicle on the road in California in 1990 had a Red Book value of $ 5,063 and cost $ 126 to register. As a comparison, average fuel costs amounted to about $ 700 per vehicle per year and insurance costs in California averaged about $ 870 (Insurance Information Institute, 1991); thus, registration fees amounted to about seven percent of annual vehicle operating costs, excluding maintenance costs. Table 2 also shows that the average vehicle in California is driven 13,400 miles per year, slightly more than the national average of 12,700.

Table 2: Vehicles, Annual Mileage, Emissions, and Registration Fees in California in 1990

Number of households in sample 1,018
Average number of vehicles per household 1.79
Average annual miles per vehicle 13,409
Average exhaust HC emissions(a) 2.49 g/mi
Percentage of households with zero vehicles 6.2%
Average annual registration fee per vehicle $ 126
Average annual registration fee per household $ 226
Median household annual income $ 37,500
Average registration fee as percent of annual income 0.68%
Median household annualized lifetime income $ 42,276
Average registration fee as percent of annualized lifetime income 0.56%

(a) This is a weighted average based on miles traveled; the unweighted average emissions rate is 2.80 g/mi.

Our emissions data come from a California data set of over 90,000 vehicles that were subjected to remote sensing in 1991 (Stedman et al., 1994).(6) The primary systematic way in which emissions vary across vehicles is by model year. Older vehicles were subject to less strict standards when they came off the assembly lines. More importantly, emissions systems deteriorate over time and sometimes break down completely; even vehicles that were very clean when new can be high polluters after a few years, particularly if they are not well maintained. We use the remote sensing data to compute average hydrocarbon (HC) emissions rates by vehicle age for cars and light-duty trucks.(7) We assign these averages to our NPTS vehicles. As shown in Table 2, the weighted (by VMT) average of all HC emissions in California in 1990 is 2.49 g/mi, six times the federal standard for new vehicles in that year.

Environmetally Based Vehicle Registration Fees

Various forms of emissions fees have been discussed in the policy arena.(8) A fee based on total emissions would be the most efficient, because it would encourage both reduced driving and repair or scrappage of dirty vehicles. In fact, the most efficient fee would be one that obtained emissions readings during actual driving in areas and at times of the day and year with serious air quality problems. For example, a fee based on HC (and/or NOx) emissions in congested urban areas during the summertime would be ideal to address urban ozone problems.(9)

Although a fee based on total emissions is likely to be the most efficient, we also look at a VMT-based fee and a fee based on emissions rates. The incidence of a VMT fee would look a lot like the incidence of a gasoline tax increase, a policy that would be administratively the easiest to carry out. On the other hand, increasing gasoline taxes in the United States seems to always meet with a firestorm of protest from the driving public. Moreover, it is possible that an emissions fee would be used to substitute for existing I&M programs. In this case, a system focusing on emissions rates as I&M programs do may be more acceptable. For these reasons, it seems important to analyze the distributional aspects of all three types of fees. Knowing their equity, as well as their efficiency, impacts should help identify the best fee.

We need to reiterate that our emissions data vary only by vehicle model year and not by individual vehicle. An ideal data set would have actual in-use emissions (from remote sensing measurements) for specific vehicles matched with income and other information on the owners of those vehicles. Unfortunately, to our knowledge, no such data set exists. Our results are useful for analyzing programs in which fees are based on vehicle age, and they give some sense of the distributional effects of a true emissions fee program, since, as we stated above, vehicle age is the most important determinant of emissions rates. A fee that varies by age would be much easier to implement, especially if substituted for current registration fees.

If we divide total mileage by all passenger vehicles in California into total registration fees paid by the owners of these vehicles under the existing system, we end up with an average VMT-based registration fee of 0.94 cents per mile. This amounts to approximately 16 percent of 1990 fuel costs for an average vehicle in California (equivalent to a gasoline tax of about 19 cents per gallon). If we multiply our HC emissions rates by mileage for each vehicle and divide the resulting total emissions number into total registration fees paid under the existing system, we end up with an emissions fee of 0.46 cents per gram. Finally, performing a similar calculation for emissions rates, we obtain an emissions rate fee of $ 50.72 per g/mi. The emissions rate fee is like an emissions fee calculated with the assumption that all vehicles are driven the same number of miles per year.(10)

Our focus here is on substituting an emissions fee for another type of vehicle tax, but it is interesting that our revenue-neutral emissions fee of 0.46 cents per gram is very close to at least one estimate of a Pigovian fee. Small and Kazimi (1995) summarize evidence on the health benefits of reducing vehicle emissions. They come up with a range of estimates of the marginal damages from HC and NOx emissions, with the upper end of their range for HCs being 0.45 cents per gram.(11)

Harrington, McConnell, and Alberini (1998) use the Small and Kazimi (1995) estimates to assess the efficiency of emissions fees compared to traditional I&M programs. They develop a simulation model in which emissions rates and repair effectiveness are stochastic with parameters estimated from real-world data. In the model, a vehicle owner obtains an initial emissions estimate, and under the fee program, she repairs her vehicle if the sum of the repair costs and the expected postrepair fee is less than the fee without any repairs. Under the I&M program, all vehicles that violate a prespecified standard are repaired. The authors vary the parameters of the model and analyze several different scenarios, but in all cases, the net welfare benefits--i.e., the benefits of reduced emissions less the cost of repairs--are greater with the fee than with the I&M program. Essentially, this result stems from the fact that some highcost repairs that take place under the more command-and-control approach of I&M are avoided with the emissions fee.(12)

Although Harrington, McConnell, and Alberini (1998) estimate positive net benefits from an emissions fee, the average out-of-pocket costs to vehicle owners, including the fee payments, are quite high in many of their scenarios. Thus, despite strong arguments for a fee on efficiency grounds, equity might still be a concern of policymakers. In the next section, we assess the equity implications of our three revenue-neutral fees vis-a-vis the existing registration fee in California.

Distributional Findings

Annual Income Results

Table 3 shows average annual registration fee payments as a fraction of annual income by decile under the current fee system and under the three environmental fees. On the basis of annual income, all of the fees appear regressive. Moreover, the three environmental fees look markedly more regressive than the current value-based registration fee. Under the VMT fee, households in the lowest decile pay, as a percentage of income, over twice what the average household pays--1.62 percent versus an average of 0.76 percent; they pay approximately three times the average under the two emissions fees. With the emissions rate-based fee, not only do households in the poorest decile pay substantially more as a percentage of income than their counterparts in the richest decile, they even pay more in absolute dollar terms on a per-vehicle basis. These numbers are not shown in the table, but the average household in decile 1 pays $ 132 per vehicle under a fee based on emissions rates, while the average household in decile 10 pays $ 107 per vehicle.(13)

Table 3: Annual Fee as a Percentage of Annual Household Income, by Decile

Current Registration Fee
Emissions Fee
Emissions Rate Fee
1 1.37 1.62 2.65 3.04
2 0.70 1.24 1.49 1.24
3 0.74 0.78 0.94 1.22
4 0.64 0.65 0.71 0.75
5 0.57 0.53 0.50 0.60
6 0.64 0.67 0.57 0.60
7 0.56 0.55 0.56 0.46
8 0.53 0.47 0.43 0.46
9 0.48 0.45 0.41 0.38
10 0.45 0.40 0.35 0.30
Average 0.68 0.76 0.90 0.95

(a) Each decile contains ten percent of households; decile is the lowest annual income decile and decile 10 is the highest.

The Suits Indexes for the three environmentally based fees are dramatically different from the Suits Index for existing fees.(14) The existing registration fee has a Suits Index of -0.09; the VMT, emissions, and emissions rate fees have Suits Indexes of -0.15, -0.24, and -0.28, respectively. The two emissions fees thus exhibit quite a bit of regressivity.

Even though these are revenue-neutral fees, their differential impacts over the existing California registration fee are fairly substantial. Under the most efficient fee, the fee based on total emissions, households in the bottom decile would pay, on average, $ 75 more per year in registration fees ($ 177 per household versus only $ 85 under the existing California system), an additional 1.3 percent of their incomes. Households in the top decile, on the other hand, would pay $ 150 per year less in registration fees ($ 291 per household versus $ 441 under the current system). The differential impacts of the emissions rate fee are even greater. These findings seem to confirm the fears of policymakers that emissions fees have serious equity problems--even when they substitute for an existing tax, the net effect appears to be an increase in regressivity.

Lifetime Income Results

The lifetime income results, shown in Table 4 (where deciles are now defined on the basis of lifetime income), are quite different from the annual income results. The VMT fee is not noticeably different from the existing registration fee system and both fees appear only slightly regressive. The Suits Index, on a lifetime income basis, for the VMT fee is -0.06, compared to -0.03 for the existing registration fee.

Table 4: Annual Fee as a Percentage of Annualized Household Lifetime Income, by Decile

Decile(a) Current Registration Fee VMT Fee Emissions Fee Rate
1 0.44 0.42 0.52 0.71
2 0.57 0.56 0.66 0.70
3 0.68 0.60 0.59 0.66
4 0.53 0.67 0.74 0.67
5 0.64 0.69 0.67 0.66
6 0.62 0.60 0.63 0.69
7 0.58 0.60 0.64 0.63
8 0.55 0.62 0.56 0.51
9 0.53 0.44 0.39 0.38
10 0.46 0.41 0.37 0.33
Average 0.56 0.56 0.58 0.60

(a) Each decile contains ten percent of households; decile 1 is the lowest lifetime income decile and decile 10 is the highest.

The two emissions fees are slightly more regressive than the existing registration fee: on average, as a fraction of their annualized lifetime income, households in the bottom decile pay higher fees than do households in the top decile, and the Suits Indexes fall to -0.11 for the fee based on total emissions and -0.14 for the emissions rate-based fee.

On a lifetime income basis, the differential impacts of the three environmental fees over the current system, on a dollar basis, are small. With a fee based on total emissions, households in the bottom decile pay just $ 14 more per year than under the current system ($ 88 versus $ 74). Under even the worst of the fees, the emissions rate-based fee, households in the bottom decile pay only an additional $ 43 per year, 0.3 percent of their annualized lifetime income.

These findings are consistent with the previous research on sales, property, value-added, and excise taxes that we discussed briefly above: taxes appear less regressive on the basis of lifetime income than annual income. The reasoning is straightforward. Annual income essentially miscategorizes some households--some households with low annual incomes have high lifetime incomes and vice versa. When we divide tax payments by lifetime income rather than annual income, we are using a larger denominator for some of the low-income households and using a smaller denominator for some of the higher income households. This has the tendency to make the taxes look less regressive. This same logic applies to all four of the fees we analyzed.(15)


Policymakers are currently looking more favorably on economic incentive approaches to reducing pollution, yet the enthusiasm for such approaches on efficiency grounds is usually tempered by concerns over equity. This is particularly the case with policies concerning motor vehicles. We explore this issue in this paper and address, in particular, the question of whether a judicious return of the revenues raised from emissions fees might alleviate the equity concerns. We examine a policy whereby current vehicle registration fees, which are based on the value of a vehicle, are replaced by one of three emissions-related fees--a fee based on annual mileage, a fee based on total annual emissions, and a fee based on emissions rates, in g/mi.

Our results are heavily dependent on the measure of income we use. If we use annual household income, we find that our three fees appear regressive, and returning the revenues by reducing motor vehicle registration fees does not completely offset the problem. The fee based on mileage is less regressive than the two emissions fees--poorer households tend to drive fewer miles than wealthier households. The emissions rate-based fee is the most regressive; thus, equity advantages of a fee based on total emissions rather than emissions rates reinforce the efficiency advantages of such fees.

If we use a constructed measure of lifetime income, our results are quite different. All of the fees appear regressive, but much less so than on the basis of annual income. Moreover, none of the fees looks markedly different from current vehicle registration fees. We find that a VMT fee would look about the same as current registration fees, on a lifetime income basis, while the two emissions-based fees look only slightly more regressive than current fees.

Whether annual income or lifetime income is the more appropriate measure to use in an evaluation of emissions fees is an open question. While economists have long espoused the notion that consumption is based on lifetime income, policymakers have yet to embrace the concept for purposes of evaluating tax incidence. Barthold (1993) argues that it is politically impractical to talk about lifetime income because of the inherent uncertainty in measuring it and because of the shorter time horizons of elected officials and the voting public. And empirical evidence on the lifetime income hypothesis is mixed. There is some evidence to suggest that consumers respond to changes in the pure timing of income, which the lifetime income hypothesis would say should not happen (Shapiro and Slemrod, 1994). Some other empirical evidence suggests that increases in households' consumption expenditures match increases in income (Carroll and Summers, 1991).

We do not pass judgment on which income measure is more appropriate; both results should be of interest to policymakers. A more important issue might be the way in which fee revenues are returned to the public. We examine substitution of emissions fees for vehicle registration fees for two reasons. First, as discussed in the Introduction, some authors have espoused the idea of linked compensation for the losers from corrective taxes. Reducing an existing vehicle-related tax might be one way of carrying this out and might be viewed favorably by motorists. Second, replacing an existing registration fee with one calculated a slightly different way might have lower administrative costs than simply introducing a new tax or reducing the level of a completely unrelated tax. Nonetheless, there may be other tax reductions or subsidy schemes that would do a better job of offsetting the regressivity exhibited by emissions fees in an annual income context. This is an interesting topic for future research.


This research was funded in part by the U.S. Environmental Protection Agency (EPA) and the U.S. Department of Transportation, Federal Highway Administration (DOT/FHWA), through an EPA cooperative agreement (CR 815934-03) with Resources for the Future. On an earlier version of this paper, we received very helpful comments from Don Fullerton, Gilbert Metcalf, Winston Harrington, Alan Krupnick, Molly Macauley, Virginia McConnell, Paul Portney, and participants in NBER's 1995 Summer Institute Workshop on Public Policy and the Environment. We also appreciate the helpful comments of three anonymous referees.

(1) Rogers (1993) analyzes gasoline, alcohol, tobacco, and utilities taxes; Casperson and Metcalf (1994), a value-added tax.

(2) She uses the education and race of the head of the household only, stating that these are very highly correlated with the education and race of the spouse. The PSID defines the head of the household as male; thus, the female-headed household dummy will pick up single, female-headed households. For married households, her measure of household income is the average of the income of the husband and wife.

(3) Nearly half of our households are married; 31 percent are single, female-headed; and 73 percent are white. The average years of education of our household heads is 13.5, slightly more than a high school degree. There is a very low correlation between years of education and age of the household head; thus, we can be reasonably sure that education is truly picking up the effects of education and not age.

(4) See U.S. DOT/FHWA (1993) for a general description of the NPTS and a summary of the information from it.

(5) We have performed a similar analysis for Ohio and Illinois with distributional results very similar to those for California.

(6) Remote sensing is a technology that combines roadside monitors that send infrared beams from one side of the road to a detector on the other side, measuring a vehicle's emissions, with a video camera that obtains a photograph or electronic identification of the license plate. We have also used emissions rates from the EPA's MOBILE5 emissions simulation model. Although average emissions rates tend to be lower from the EPA model than from the remote sensing data, our distributional results are the same using either source of emissions data.

(7) Cars and trucks must be treated separately, because until the 1994 model year, the federal standards for lightduty trucks were less stringent than those for cars. Minivans and sport-utility vehicles are classified as trucks. Carbon monoxide emissions are also available, but because ozone is a more serious air quality problem, we focus on HCs (HCs and NOx combine in the atmosphere to form ozone). NOx emissions are not available from the remote sensing data set.

(8) The EPA, in its proposed Federal Implementation Plan (FIP) for California in 1996 included a recommendation for some type of emissions fee. In response to the FIP, California revised its State Implementation Plan to include a VMT fee (Wallerstein, 1995). No fees have yet been implemented. Maricopa County in Arizona has considered fees based on emissions rates and vehicle age (see Energy and Environmental Analysis (1993) for an analysis). The President's Federal Advisory Committee on Reducing Greenhouse Gas Emissions from Motor Vehicles seriously considered promoting VMT fees as a way of reducing carbon dioxide emissions (Policy Dialogue Advisory Committee, 1995).

(9) See Harrington, Walls, and McConnell (1995) for a discussion.

(10) One can verify using the numbers in Table 2 that the fees are revenue-neutral. The average household continues to pay approximately $ 226 per year in registration fees. Of course, this assumes no change in motorists' behavior. If the fees do what we want them to do, VMT and emissions will drop and tax collections along with them. This means that the rates would have to be adjusted upward to keep revenues constant. Estimating behavioral responses to the fees is outside the scope of this paper.

(11) This particular estimate of Small and Kazimi's (1995) is obtained by attributing all of the ozone morbidity effects to HCs and none to NOx. Because we are looking at a fee assessed only on HCs and not NOx, it seems appropriate to compare it to this estimate from Small and Kazimi.

(12) The fee revenues are not part of the net welfare calculation; they are simply a transfer--i.e., the government is assumed to return those revenues is some way to households. Of course, reducing registration fees is one way in which the revenues could be returned.

(13) Of course, households in decile 10 own more vehicles on average, so they have higher average payments per household.

(14) A Suits Index is the tax analog to the Gini coefficient; it is constructed by measuring at each point in the income distribution the cumulative tax paid relative to cumulative income earned. It is bounded by -1 and 1, with values less than zero connoting regressivity and values greater than zero progressivity; a proportional tax has a Suits Index of zero (Suits, 1977). Interestingly, in Suits' original paper, one of the sets of taxes for which he computed his index was vehicle registration fees and personal property taxes, grouped together. He calculated an index of -0.12 using 1966 tax rates and income and -0.09 using 1970 tax rates and income, identical to our result with 1990 data.

(15) As we noted in the second section, our lifetime income measure does not incorporate individual fixed effects and thus probably has a lower variance than true lifetime income. This means that our results are likely to exhibit less regressivity than results based on true lifetime income.


Barthold, Thomas. "How Should We Measure Distribution?" National Tax Journal 46 No. 3 (September, 1993): 291-9.

Blomquist, N. Soren. "A Comparison of Distributions of Annual and Lifetime Income: Sweden Around 1970." The Review of Income and Wealth 27 No. 3 (September, 1981): 243-64.

Bovenberg, A. Lans, and Lawrence H. Goulder. "Optimal Environmental Taxation in the Presence of Other Taxes: General Equilibrium Analyses." American Economic Review 86 No. 4 (September, 1996): 985-1000.

Bovenberg, A. Lans, and Ruud A. DeMooij. "Environmental Levies and Distortionary Taxation." American Economic Review 94 No. 4 (September, 1994): 1085-9.

Burtraw, Dallas. "Compensating Losers When Cost-Effective Environmental Policies Are Adopted." Resources 104 No. 3 (Summer, 1991): 1-5.

CACI Marketing Systems. The Sourcebook of Zip Code Demographics. Fairfax, 1991.

Carroll, Christopher D., and Lawrence Summers. "Consumption Growth Parallels Income Growth: Some New Evidence." In National Saving and Economic Performance, edited by B. Douglas Bernheim and John Shoven. Chicago: University of Chicago Press, 1991.

Casperson, Erik, and Gilbert Metcalf. "Is a Value Added Tax Regressive? Annual Versus Lifetime Incidence Measures." National Tax Journal 47 No. 4 (December, 1994): 731-47.

Davies, James, France St-Hilaire, and John Whalley. "Some Calculations of Lifetime Tax Incidence." American Economic Review 74 No. 4 (September, 1984): 633-49.

Energy and Environmental Analysis. Draft Working Paper on Emissions-Based Registration Fees. Report prepared for Maricopa County, Arizona, Association of Governments. Arlington, VA: Energy and Environmental Analysis, 1993.

Friedman, Milton. A Theory of the Consumption Function. Princeton: Princeton University Press, 1957.

Fullerton, Don, and Diane Lim Rogers. Who Bears the Lifetime Tax Burden? Washington, D.C.: The Brookings Institution, 1993.

Goulder, Lawrence H. "Environmental Taxation and the Double Dividend: A Reader's Guide." International Tax and Public Finance 2 No. 2 (1995): 157-83.

Harrington, Winston, Virginia McConnell, and Anna Alberini. "Economic Incentive Policies Under Uncertainty: The Case of Vehicle Emissions Fees." In Environment and Transport in Economic Modelling, edited by Roberto Roson and Kenneth A. Small. Hingham: Kluwer Academic Publishers, 1998.

Harrington, Winston, Margaret Walls, and Virginia McConnell. "Using Economic Incentives to Reduce Auto Pollution." Issues in Science and Technology 11 No. 2 (Winter, 1995).

Kessler, Jon, and William Schroeer. "Meeting Mobility and Air Quality Goals: Strategies That Work." U.S. Environmental Protection Agency, Office of Policy Analysis. Final draft manuscript, October, 1993.

Lillard, Lee A. "Inequality: Earnings vs. Human Wealth." American Economic Review 67 No. 2 (March, 1977): 42-53.

Lyon, Andrew B. and Robert M. Schwab. "Consumption Taxes in a Life-Cycle Framework: Are Sin Taxes Regressive?" Review of Economics and Statistics LXXVII No. 3 (August, 1995): 389-406.

Metcalf, Gilbert E. "The Lifetime Incidence of State and Local Taxes: Measuring Changes During the 1980s." NBER Working Paper No. 4252. Cambridge, MA: National Bureau of Economic Research, January, 1993.

Modigliani, Franco, and Richard Brumberg. "Utility Analysis and the Consumption Function: An Interpretation of Cross-Section Data." In Post-Keynesian Economics, edited by Kenneth K. Kurihara, 388-436. New Brunswick: Rutgers University Press, 1954.

Parry, Ian W. H. "Pollution Taxes and Revenue Recycling." Journal of Environmental Economics and Management 29 No. 3. Part 2 (November, 1995): S 64-S 77.

Policy Dialogue Advisory Committee to Develop Options for Reducing Greenhouse Gas Emissions from Personal Motor Vehicles. Draft Final Report. Washington, D.C.: U.S. Environmental Protection Agency Air Docket, August 28, 1995.

Poterba, James M. "Lifetime Incidence and the Distributional Burden of Excise Taxes." American Economic Review 79 No. 2 (May, 1989): 325-30.

Poterba, James M. "Is the Gasoline Tax Regressive?" In Tax Policy and the Economy, Volume 5, Cambridge, MA: National Bureau of Economic Research, 1991: 145-64.

Rogers, Diane Lim. "Measuring the Distributional Effects of Corrective Taxation." Paper presented at a National Tax Association session of the Allied Social Science Association meetings, Boston, January 3-5,1993.

Shapiro, Matthew, and Joel Slemrod. "Consumer Response to the Timing of Income: Evidence from a Change in Tax With holding." American Economic Review 85 No. 1 (March, 1994): 274-83.

Small, Kenneth. "Using the Revenues from Congestion Pricing." Transportation 19 No. 4 (1992): 359-81.

Small, Kenneth, and Camilla Kazimi. "On the Costs of Emissions from Motor Vehicles." Journal of Transport Economics and Policy 29 No. 1 (January, 1995): 7-32.

Stedman, Donald H. et al. "On-Road Remote Sensing of CO and HC Emissions in California: Final Report." Prepared for California Resources Board, Sacramento, February, 1994.

Suits, Daniel B. "Measurement of Tax Progressivity." American Economic Review 67 No. 4 (September, 1977): 747-52.

U.S. Department of Transportation, Federal Highway Administration. 1990 NPTS Databook: Volumes I and II. U.S. DOT/FHWA, Office of Highway Information Management, Report Nos. FHWA-PL94-010A and FHWA-PL-94-010B, Washington, D.C., November, 1993.

Wallerstein, Barry. South Coast Air Quality Management District. Personal communication, August, 1995.

More Information on Energy Taxes

FAIR USE NOTICE: This page contains copyrighted material the use of which has not been specifically authorized by the copyright owner. Global Policy Forum distributes this material without profit to those who have expressed a prior interest in receiving the included information for research and educational purposes. We believe this constitutes a fair use of any such copyrighted material as provided for in 17 U.S.C íŸ 107. If you wish to use copyrighted material from this site for purposes of your own that go beyond fair use, you must obtain permission from the copyright owner.


FAIR USE NOTICE: This page contains copyrighted material the use of which has not been specifically authorized by the copyright owner. Global Policy Forum distributes this material without profit to those who have expressed a prior interest in receiving the included information for research and educational purposes. We believe this constitutes a fair use of any such copyrighted material as provided for in 17 U.S.C § 107. If you wish to use copyrighted material from this site for purposes of your own that go beyond fair use, you must obtain permission from the copyright owner.