Journal of Transport Economics and Policy, Volume 39, Part 1, January 2005, pp.

124

Equity Implications of Vehicle Emissions

Taxes
Sarah E. West

Address for correspondence: Sarah E. West, Department of Economics, Macalester

College, 1600 Grand Avenue, St Paul, MN 55105 USA. The author thanks Mark
Delucchi, Don Fullerton, Raymond Robertson, Dan Slesnick, Rob Williams, and Ann
Wolverton for their helpful comments and suggestions. For helping her to understand
fuel chemistry, she thanks Keith Kuwata. For providing data, she thanks the California Air Resources Board, Satya Devesh, Mary Hostak, and Raphael Susnowitz.

Abstract
This paper considers the equity implications of vehicle emissions taxes by examining the
incidence of a tax on local pollutants. It uses emissions data from the California Air
Resources Board and household vehicle and income data from the US Consumer
Expenditure Survey. It incorporates household price responsiveness that diers across
income groups into a consumer surplus measure of tax burden. Since poor vehicle owners
spend more on miles as a proportion of their income and drive vehicles that pollute more
per mile than those owned by the wealthy, the incidence of tax on vehicle emissions falls
relatively heavily on them. This burden, however, is mitigated to some extent by low
vehicle ownership rates and high price responsiveness in the lower half of the income
distribution. A uniform tax on miles that does not distinguish between dirty and clean
vehicles is less regressive than the emissions tax.

Date of receipt of nal manuscript: April 2004

1

Journal of Transport Economics and Policy

Volume 39, Part 1

1.0 Introduction
Between 1980 and 2000, vehicle miles travelled (VMT ) in the United States
increased by over 76 per cent (FHWA, 2003). This dramatic increase in
vehicle use frustrates attempts to reduce local air pollution. As of January
2004, for example, fourteen metropolitan areas in the United States are
classied as extreme or severe ozone nonattainment areas.1 Faced with
these challenges, policy makers seek cost-eective means to attain abatement goals. While command-and-control standards still dominate vehicle
pollution policy in the United States, many studies have shown that taxes
and other incentives can attain the same amount of pollution reduction
at lower cost (see for example Bohm and Russell (1985) and Harrington
et al. (1994)). Pollution taxes also generate revenue that can be used to
fund government projects or to reduce taxes on labour or investment.
One argument against taxes on vehicle pollution and gasoline is that
they disproportionately burden poor households. Several studies conrm
fears that a tax on gasoline is regressive. Only Walls and Hanson (1999)
and Sevigny (1998), however, explicitly consider the incidence of a vehicle
emissions tax.2 Both those studies consider local pollutants: Walls and
Hanson examines a tax on hydrocarbons (HC) and Sevigny examines a
tax on HC, carbon monoxide (CO), and oxides of nitrogen (NOx ).
Five main contributions distinguish this paper. First, it considers a tax
on two particularly damaging pollutants, particulate matter (PM) and
sulphates (SO4 ) along with CO, HC, and NOx . Second, it uses data from
the California Air Resources Board to estimate vehicle emissions per mile
as a function of vehicle vintage, engine size, import status, and vehicle
type. Walls and Hanson (1999) and Sevigny (1998) assign emissions per
mile to vehicles in households according to vehicle vintage. Other papers
estimate vehicle emissions functions as complicated as those estimated
here but do not conduct incidence analysis (see for example Harrington
(1997) and Kahn (1996a)).
Third, my analysis examines the eect of including households that do
not own vehicles. Walls and Hanson (1999) includes households without
1

See the US Environmental Protection Agencys Green Book Nonattainment listings at [Link]
[Link]/oar/oaqps/greenbk.
2
Kahn (1996b) determines the eect of households neighbourhoods median incomes on the likelihood
of passing vehicle emissions tests. Dill et al. (1999) consider the incidence of existing California vehicle
licence fees that are not based on emissions. Other papers analyse the incidence of other pollution
taxes. Brannlund and Nordstrom (2004) and Cornwell and Creedy (1997), for example, examine
the incidence of carbon taxes. Poterba (1991) and West and Williams (2004) consider the incidence
of a gasoline tax, West (2004) examines taxes on miles and subsidies to new vehicles, and Metcalf
(1999) calculates incidence for a variety of environmental taxes but not a tax on vehicle emissions.

Equity Implications of Vehicle Emissions Taxes

West

vehicles but does not compare results with simulations that omit these
households. Sevigny (1998) considers only vehicle owners. Measures that
ignore households that do not own vehicles will overstate the incidence
on income groups with fewer vehicle owners, and understate the incidence
on groups with more vehicle owners.
Fourth, I incorporate households responses to an emissions tax and a
tax on vehicle miles travelled into incidence calculations. I use elasticities
of demand for VMT that dier across income groups, obtained from
estimation using household data from the 1997 US Consumer Expenditure
Survey. Other papers calculate incidence with no price responsiveness or
with no income-varying price responsiveness.3 They assume away the
possibilities that, for example, poorer households are more price responsive
because their spending on miles occupies a larger fraction of their budget,
or that wealthier households are more price-responsive because they have
more transport options. To the extent that demand elasticities vary with
income, measures that ignore this will overstate the incidence on income
groups with relatively elastic demand, and understate the incidence on
groups with relatively inelastic demand.
Fifth, incorporation of price responsiveness allows me to use the change
in consumer surplus to calculate the reduction in welfare due to the tax.4
This measure adds the Harberger triangle (Harberger (1962)), to the
rectangle of tax paid.5 Ignoring this triangle biases incidence estimates.
For example, if poor households are more price responsive than wealthy
households, they will escape more of the tax than wealthy households,
mitigating regressivity, but their Harberger triangles will be larger,
exacerbating regressivity.
The incidence of vehicle emissions taxes depends on the relationships
between emissions per mile, VMT, and income. Thus the income measure
used is a critical determinant of incidence. For the purposes of incidence
analysis, an ideal income measure places a money value on material well
being with which the value of welfare loss from the policy change can be
scaled. Consumption is a better indicator of material well being than

Metcalf (1999), Poterba (1991), and Walls and Hanson (1999) calculate incidence assuming no price
responsiveness. Sevigny (1998) allows price responsiveness to dier according to how many vehicles a
household owns but not according to income.
4
West (2004) uses consumer surplus when examining incidence of a miles tax. Brannlund and Nordstrom (2004) and Cornwell and Creedy (1997) use equivalent and compensating variation to measure
the welfare eects of carbon taxes.
5
The Harberger triangle is usually associated with deadweight loss. A distorting tax, such as a tax on
labour or capital, creates a deadweight loss by raising price above marginal cost. A tax on emissions,
on the other hand, is corrective rather than distorting it raises price to marginal cost. Rather than
creating a deadweight loss, emission taxes eliminate one.

Journal of Transport Economics and Policy

Volume 39, Part 1

annual income.6 University students and retired people, for example, may
have very low incomes but high levels of consumption and thus high
material well being and working households can maintain levels of well
being in the face of temporary reductions in income by taking money
from savings or by borrowing. In good economic times, households
smooth consumption by saving.7 For these reasons, and since the
Consumer Expenditure Survey contains detailed and accurate data on
expenditures, I use total consumption expenditures as my measure of
income.
This paper focuses on incidence of the costs of vehicle emissions taxes,
ignoring the distribution of the external benets of reduced vehicle
pollution and miles driven. Incorporating such benets would reduce tax
burdens for all income groups and might have important distributional
eects if the benets are unevenly distributed across income groups.8 In
addition, I do not consider the distributional eects of rebating tax revenue
or using it to reduce other taxes.9 As pointed out in Small (1983), the
concept of regressivity is not appropriate for a corrective tax where there
are net welfare increases. For the purposes of this paper, therefore, I use
the terms regressive and regressivity when comparing one scenario
to another, not to characterise the incidence of one particular policy.
Among vehicle owners, emissions per mile of local pollutants decrease
as income increases; poor vehicle owning households drive vehicles that
pollute more than those owned by wealthy households. But a signicant
proportion of the population owns no vehicles and therefore emits no
vehicle pollution.10 This population is concentrated in the lower third of
the income distribution. In addition, poor households are more price
responsive than wealthy ones; elasticities of VMT demand in the poorer
deciles of the income distribution are about twice as large as those in the
wealthier deciles. While the incidence of tax on local vehicle pollutants
falls relatively heavily on poor households, this burden is mitigated to
6

Sevigny (1998) uses annual income. Walls and Hanson (1999) use a measure of lifetime income, which
usually produces incidence estimates more similar to those that use consumption. Poterba (1991),
West (2004), and West and Williams (2004) use consumption to analyse environmental policy
incidence but do not consider a vehicle emissions tax.
7
See Slesnick (2001) for further discussion of dierences among income measures. Consumption may
be a better approximation to welfare than income, but it is still an approximation. It is problematic,
for example, for households that cannot smooth consumption by saving or borrowing.
8
See Baumol and Oates (1988) and Brooks and Sethi (1997) for general discussion of the distribution of
benets.
9
I assume that the government discards all tax revenue. For incidence analysis of a gasoline tax whose
revenues are used to reduce a labour tax, see West and Williams (2004). For analysis of a carbon tax
whose revenues are used to reduce a variety of taxes, see Brannlund and Nordstrom (2004).
10
Of course, households that do not own vehicles may pollute indirectly by riding diesel-fuelled buses or
trains fuelled by electricity generated by coal-burning power plants.

Equity Implications of Vehicle Emissions Taxes

West

some extent by low vehicle ownership rates and high price responsiveness in
the lower half of the income distribution.
Since poor households drive dirtier vehicles than wealthy households, a
uniform tax on miles that does not distinguish among vehicles is less
regressive than the emissions tax. In the case of vehicle emissions of local
pollutants, the pollution control policy that is easier to implement is also
less regressive.
Section 2 provides details on the incidence measures employed in the
paper. Section 3 discusses the data, emissions per mile estimation, and
elasticity estimation. Section 4 presents the incidence results and Section
5 concludes.

2.0 Measuring the Incidence of Emissions Taxes

A households total emissions equals the sum of emissions per mile of each
pollutant i (EPMih ) times vehicle miles travelled (VMT h ). An ecient tax
on emissions equals the money value of marginal external damages
(MED) per unit of emissions.11 A tax per gram of emissions could be
assessed by measuring the emissions per mile and multiplying by miles
travelled, where miles travelled is obtained by comparing current odometer
readings with previous readings. The total tax paid by a household (T h ) on
local pollutants would equal:
X
MEDi EPMih VMT h :
1
Th
i

The tax on emissions is thus equivalent to a vehicle-specic tax on miles.

An ideal measure of the incidence of a tax on vehicle emissions would
calculate the general-equilibrium changes in prices that would occur
throughout the economy in response to the tax, and then calculate the
eects of those price changes on households welfare. A tax on vehicle
emissions, even if it was assessed only on households, would presumably
aect prices of fuel and automobiles and thereby aect many sectors of
the economy. Calculating such eects, however, requires a great deal of
information, most notably the demand and supply elasticities for all
aected industries, and the distribution of ownership of rms in those
industries. Thus, for simplicity, I focus on the short-run partial equilibrium
incidence of the tax. I assume that in the short run, households respond to
the emissions tax by reducing the number of miles they drive, but not by
11

This is the familiar Pigouvian tax, introduced in Pigou (1932).

Volume 39, Part 1

Journal of Transport Economics and Policy

switching to less polluting vehicles or other modes of transport.12 I also

assume that the supply of consumer goods is perfectly elastic. This implies
that the imposition of the emissions tax does not aect the producer prices
of related goods such as fuel and vehicles, and thus the entire burden of the
tax falls on consumers.13
Each household faces its own initial price per mile, determined by fuel
costs and other variable costs per mile. Each household also faces its
own
per mile, which is the initial price per mile plus
P with-tax price
h
i MEDi EPMi . While I assume that MED are the same for all
households, emissions per mile are vehicle specic.14 In addition, I allow
VMT demand elasticities to vary across income deciles.
I use the change in household consumer surplus to measure the change
in household welfare due to the tax. Because I consider short-run incidence,
where households respond to the emissions tax only by reducing miles,
consumer surplus is dened as the change in the area under the households
VMT demand curve over the quantity purchased, which reects the eect
of changing prices on utility. Assuming a linear VMT demand curve, the
change in consumer surplus for household h can be expressed:
X
MEDi EPMih VMT1

CSh
i

1
2

X

MEDi EPMih

VMT1h
VMT0h ;

where the rst term is T h , the rectangle of tax paid dened in equation (1),
after the household has adjusted miles in response to the tax. The second
12

Studies conrm that in the long run, households respond to higher gasoline prices by switching to
more fuel ecient vehicles (see Dahl and Sterner (1991)), but none examine the dierences in this
adjustment across income groups. We can therefore only speculate that since wealthier households
can more easily aord to replace a dirty car with a less polluting one and thus avoid more of the
tax, the emissions tax may be more regressive in the long run.
13
While no study examines the degree to which a vehicle emissions tax would be borne by gas and vehicle
producers versus consumers, we might expect the gas tax to have a similar incidence. CBO (2003) nds
that gasoline consumers would bear nearly 85 per cent of the total long run costs of an increase in the
gas tax. Vehicle manufacturers would bear part of the burden by lowering vehicle prices; Goldberg
(1998) nds that carmakers would respond to gas tax increases by lowering the relative prices of
less fuel ecient luxury and standard sized cars. These studies, however, do not address the possible
dierences in the producer incidence of a gas tax across dierent income groups. To the extent that
gasoline and vehicle suppliers bear part of the tax burden, my estimates will overstate the incidence
on households that consume gasoline and vehicles, and will understate the incidence on households
that own rms that supply gasoline and vehicles. If rm owners are concentrated in the top deciles,
this would mean that my estimates would overstate the taxs regressivity.
14
A referee points out that to the extent that poorer people live and drive in relatively densely populated
areas, the true MED for them will be higher because MED is nearly proportional to population
density.

Equity Implications of Vehicle Emissions Taxes

West

term is the Harberger triangle, which, with a linear demand for VMT, is
half of the tax per VMT times the change in VMT that occurs in response
to the tax.15
For ease of comparison with previous studies, I also calculate the change
in household welfare as the tax paid by that household, using equation (2)
but omitting the second term. I divide both welfare loss measures by total
expenditures to obtain tax burden as a percentage of income. Comparison
of these ratios across deciles allows one to determine which income groups
bear more of the burden of a particular tax relative to their incomes.
To compare the overall incidence of two dierent tax policies, however,
one needs to measure the distribution of the tax burden across all deciles for
each policy. Suits (1977) derives just such a measure. The Suits index is
traditionally calculated using tax paid as the measure of welfare change.
For a given tax policy Tx , the Suits index calculated over ten income deciles
is:
X

10
1
Sx 1

3
2Tx yi Tx yi
1 yi
yi
1 =K;
i1

where yi is the accumulated percentage of income through decile i, Tx yi is

the accumulated percentage of total tax paid through decile i, and K is a
constant. Scaling the term in brackets by K and subtracting it from one
gives an index similar to the Gini coecient. The Suits index ranges from

1 for a perfectly regressive tax, to zero for a proportional tax, and to
1 for a perfectly progressive tax.
For ease of comparison with other studies, I present the traditional Suits
index using tax paid. I also calculate a Suits index using consumer surplus
as the measure of welfare loss. The Suits index equivalent for consumer
surplus (taxes paid plus the Harberger triangle) is obtained by dening
Tx yi as the accumulated percentage of consumer surplus lost through
decile i.

3.0 Data, Emissions per Mile Estimation, and

Elasticity Estimation
Determining the incidence of a tax on vehicle emissions requires data on
household income, emissions per mile, vehicle miles travelled, and a
measure of households responsiveness to the tax. This section describes
15

For detailed discussion of this and related measures of welfare loss, see Willig (1976).

Journal of Transport Economics and Policy

Volume 39, Part 1

the three main sources of data used here: the US Consumer Expenditure
Survey (CEX), the California Air Resources Board Light Duty Surveillance
Program (CARB), and the American Chambers of Commerce Researchers
Association (ACCRA) cost of living index. It also explains estimation of
emissions per mile and estimation of elasticities of VMT with respect to
operating costs per mile.
3.1 Household data, operating costs, and VMT derivation
The household data consists of 7,073 households from the 1997 US
Consumer Expenditure Survey that own zero, one, or two vehicles.16
Households that own no cars make up 24 per cent of the sample, 45 per
cent own one vehicle, and 31 per cent own two vehicles. The CEX includes
total expenditures, the amount spent on gasoline, and detailed information
on each households vehicles. Variables in the vehicle le include year,
make, model, number of cylinders, odometer reading, the amount paid
for the vehicle, and other characteristics.
I dene the operating cost per mile for each vehicle as the price of
gasoline divided by fuel eciency, plus maintenance and tyre costs per
mile. The ACCRA cost of living index lists for each quarter the average
prices of regular, unleaded, national-brand gasoline for over 300 US
cities. Since the CEX reports state of residence of each household, but
not city, I average the city prices within each state to obtain a state gasoline
price for each calendar quarter. Then I assign a gas price to each CEX
household based on state of residence and CEX quarter.
Unfortunately, the CEX does not record vehicles fuel eciencies. I
therefore use data from the California Air Resources Board (CARB) to
estimate a regression of miles per gallon (MPG) on engine size and vehicle
vintage.17 Smaller or newer vehicles are more fuel ecient. For one-vehicle
households, fuel eciency is calculated directly from the regression results.
For two-vehicle households, I calculate the fuel eciency of each two car
pair by averaging the two cars estimated eciencies.
I then calculate the fuel cost per mile as the price of gasoline divided by
fuel eciency. The ORNL (1998) provides maintenance and tyre costs per
mile, by vehicle vintage. I add these per mile costs to the fuel cost per mile to
obtain operating costs per mile.
16

Eighty-two per cent of 1997 CEX households own zero, one, or two vehicles (for comparison, the 1995
US Nationwide Personal Transportation Survey lists this number as 81 per cent (ORNL (2000)). The
CEX data for households with more than two vehicles is very spotty; 70 per cent of these households
have missing data for engine size or vintage of at least one vehicle. Households with three or more
vehicles have higher average expenditures than the households included here; by ignoring them this
study focuses on a less wealthy portion of the income distribution.
17
For more detail on these data and results from the MPG regression, see West (2004).

Equity Implications of Vehicle Emissions Taxes

West

The CEX reports one number per household for gasoline expenditures.
To obtain VMT driven per household, I divide the households gas
expenditure by its gas price to get gallons of gas consumed. Then, I multiply
gallons by fuel eciency to obtain VMT for the household.18;19
3.2 Emissions per mile estimation
The CEX does not include information on vehicle emissions. The CARB
contains complete information on 671 vehicles emissions of carbon
monoxide (CO), hydrocarbons (HC), and oxides of nitrogen (NOx ) in
grams per mile.20 Several studies nd that important determinants of
vehicle emissions include vintage, cumulative miles (obtained from an
odometer reading), engine size, vehicle class (light-duty truck versus car),
and import status (that is, manufactured by a US company versus a nonUS company) (see for example Bin (2003), Harrington (1997), and Kahn
(1996a)). All of these variables are included in both the CEX and the
CARB data. Odometer readings, however, are missing for more than 10
per cent of the CEX vehicles in my sample and, since CEX interviewers
do not check the odometers of the vehicles but instead rely on the
household to report accurately their vehicles mileage, readings are subject

18

Using total VMT versus the VMT in each vehicle ignores the possibility that households respond to
changes in the operating cost per mile by driving more in one vehicle and less in another. Green and
Hu (1985) nd that substitution among vehicles within a household in response to changes to the price
per mile is negligible, and so the bias is not likely to be large.
19
Miles data by vehicle is available in two data sets that were conducted close in time to the 1997 CEX: the
1994 Residential Transportation Energy Consumption Survey (RTECS) and the 1995 Nationwide
Personal Transportation Survey (NPTS). Since both surveys, however, report income only as
categorical indicators for annual income ranges, I do not use them in this paper. I did conduct a
simple comparison of the distribution of VMT across deciles from these surveys. I assigned each
RTECS and NPTS household that owns fewer than three vehicles a random income from a uniform
distribution within the households income range category and divided the households into deciles.
Both the RTECS and NPTS show more miles driven in lower deciles and fewer in upper deciles than
shown in the CEX. This is due in part to the fact that both surveys contain fewer households with no
vehicles than the CEX. Using those surveys miles and income data (but CEX-derived price elasticities
and assuming that NPTS and RTECS vehicle characteristics distributions are similar to those in the
CEX) would result in incidence estimates that are more regressive than those found here.
20
The CARBs Light-Duty Vehicle Surveillance Program, Series 13 and 14, was part of ongoing eorts
by the CARB to accumulate vehicle emissions data, to investigate vehicle maintenance practices, and
to determine the frequency and eect of tampering with pollution control equipment. The programme
performed dynamometer tests used in Californias Smog Check II Program in Enhanced Areas. Such
tests do not capture all increased emissions that occur in real-world driving conditions. They miss
those emitted, for example, during heavy acceleration or high-load operating conditions. To the
extent that driving behaviour and loads dier across income groups, my estimates will understate
incidence on households that are aggressive drivers or carry heavy loads and overstate incidence on
households that drive less aggressively or with light loads. For more details on the benets and
disadvantages of dierent vehicle emissions testing methods, see BEST (2001).

Journal of Transport Economics and Policy

Volume 39, Part 1

to extreme rounding and other measurement error, and so I do not include

odometer readings in emissions per mile estimation.21
I estimate one regression each for CO, HC, and NOx as functions of
indicators for the number of cylinders in the engine (four, six, or eight),
vintage (older than 1980, 1980 through 1989, and 1990 and newer), interactions between number of cylinder and vintage indicators, an indicator
equal to one if the vehicle is a light truck (pick-up truck, van or minivan,
or sport-utility vehicle), and an indicator equal to one if the vehicle is an
import (made by a non-US manufacturer).22
Since the CARB data contain only vehicles from California, the sample
is not representative of the US vehicle eet.23 I therefore use population
weights from the CEX to weight the vehicles in the CARB data so that
the data represent the distribution of all combinations of vintage, engine
size, vehicle type, and import status in the US vehicle eet.24 Examination
of the relationship between emissions per mile of the three pollutants
revealed that a semilog specication best ts the data.25 I estimate the
regressions using weighted ordinary least squares.
Table 1 contains results from this estimation. In the CO regression,
6-cylinder terms and import are not statistically signicant. For HC, the
6-cylinder interaction terms and import are not statistically signicant.
Other coecients in those regressions are signicant at the 5 per cent
level or better. In the NOx regression, the 6-cylinder-1980s-vintage
interaction term, truck, and import are not statistically signicant. The
6-cylinder indicator is signicant at the 10 per cent level and all other
coecients are signicant at the 5 per cent level or better.
Older and larger vehicles pollute with more CO, HC, and NOx per mile
than newer and smaller vehicles. The average 6-cylinder vehicle emits 17 per
21

Inspection of households reported odometer readings across quarters conrms these measurement
error problems; if reported odometer readings were accurate, the data would imply that many
households are driving a negative number of miles per quarter.
22
An alternative specication might replace the vintage ranges with indicator variables for each year.
Some years, however, are very sparsely represented in the CARB.
23
The CARB chose a random sample of all vehicles in California, and then sent requests to owners of
such vehicles within a 25-mile radius of the CARB oce in El Monte, California. The nal sample
includes only those who responded.
24
Even using these weights, because California (CA) vehicles certify to dierent emission standards than
do vehicles in the rest of the US, a relationship between emissions and vehicle characteristics based on
CA vehicles may not be accurate for all vehicles nationally. Since CA standards tend to be more
stringent for earlier vintages than national standards, we might expect my CA-based estimates to
underestimate emissions per mile for older cars nationwide. Since poor households drive older cars
than wealthy households, emissions per mile and therefore the incidence of emissions taxes may be
underestimated for those households.
25
Tests of Box-Cox specications do not reject the null hypothesis of a semilog specication and this
specication results in the highest R2 values.

10

West

Equity Implications of Vehicle Emissions Taxes

Table 1
Emissions per Mile Regressions

1 if cylinders 6
1 if cylinders 8
1 if 1980s vintage
1 if 1990s vintage
(6-cylinders  1980s)
(6-cylinders  1990s)
(8-cylinders  1980s)
(8-cylinders  1990s)
1 if light-duty truck
1 if import
Constant
Number of Observations
R-squared

ln(CO)
(g/mile)

ln(HC)
(g/mile)

ln(NOx )
(g/mile)

0.308
(0.334)
1.486
(0.219)
0.425
(0.177)

0.622
(0.173)

0.199
(0.367)

0.436
(0.358)

0.912
(0.306)

1.337
(0.278)
0.580
(0.148)

0.079
(0.130)
1.775
(0.179)
671
0.28

0.643
(0.321)
2.001
(0.220)
0.490
(0.191)

0.665
(0.185)

0.267
(0.352)

0.387
(0.348)

1.154
(0.299)

1.819
(0.300)
0.385
(0.152)
0.005
(0.136)

0.904
(0.194)
671
0.36

0.449
(0.261)
1.241
(0.195)
0.453
(0.173)

0.487
(0.199)

0.243
(0.284)

0.629
(0.304)

0.729
(0.246)

1.440
(0.309)
0.143
(0.095)
0.083
(0.120)

0.482
(0.177)
671
0.28

Note: HC is hydrocarbons; NOx is oxides of nitrogen. Estimation is weighted OLS. Robust

standard errors are in parentheses. The omitted vehicle is a 4-cylinder, older than 1980-vintage,
domestic car.

cent more CO, 43 per cent more HC, and 7 per cent more NOx per mile than
a 4-cylinder vehicle. The average 8-cylinder vehicle emits 2.3 times more
CO, 4.5 times more HC, and 1.2 times more NOx than the average 4cylinder vehicle. Controlling for other factors, 1980s vintage vehicles emit
42 per cent less CO, 54 per cent less HC, and 25 per cent less NOx than
vehicles older than 1980, while 1990s vehicles emit 83 per cent less CO,
87 per cent less HC, and 77 per cent less NOx than vehicles older than 1980.
These results make sense in the context of US vehicle emissions policies.
Newer vehicles are subject to more stringent emissions per mile standards
than are older vehicles and, since pollution-control equipment deteriorates
with time, even cars that face stringent o-the-assembly-line standards
emit more as they age. This is compounded for larger cars with lower fuel
eciencies; cars that burn more gas per mile and have broken pollution
control equipment emit more per mile than smaller cars with equipment in
the same condition (Harrington (1997)).
11

Journal of Transport Economics and Policy

Volume 39, Part 1

Until 1994, standards for light-duty trucks (including minivans and

sport-utility vehicles) were less stringent than for cars. The regression results
show that they emit 58 per cent more CO and 39 per cent more hydrocarbons
per mile than cars. Light-duty trucks also appear to emit more oxides of
nitrogen than cars, though this is not a statistically signicant eect.
The CARB data do not contain information on emissions of particulate
matter or oxides of sulphur. While the prevailing evidence nds no
signicant health eects of sulphur dioxide (SO2 ), sulphate (SO4 ) has
signicant health eects as a specic form of particulate matter.
Particulates are the most damaging pollutant; it is important to include
them in an emissions tax (see McCubbin and Delucchi (1999)).
Fortunately, data on HC can be used to obtain a rst approximation of
PM. Mulawa et al. (1997) nds that vehicle emissions of PM10 , particulate
matter of aerodynamic diameter of 10 microns or less, and HC emissions
are highly correlated.26 They nd that PM10 averaged 1.3 per cent of the
HC emission rate.27
The sulphur content of gasoline can be used in combination with fuel
eciency to estimate sulphate per mile. The national average sulphur
content of fuel in 1997 was 340 parts per million (EPA (1998)). According
to the EPAs MOBILE vehicle emissions models, two per cent of sulphur in
gasoline is converted to sulphur in SO4 .28 To translate into grams per mile
of sulphate for each vehicle, I multiply the sulphur content of fuel by two
per cent, divide by 100, and divide by the vehicles MPG.
I predict emissions per mile of CO, HC, NOx , and SO4 for each vehicle
in the CEX household data. Households with one vehicle are assigned the
emissions per mile of their one vehicle, while two-vehicle households are
assigned the average CO, HC, NOx , and SO4 of their two vehicles. I
multiply each households predicted HC by 0.013 to obtain PM10 per
mile. Households that own no vehicles are assigned zero values for
emissions per mile.
Table 2a summarises the data on household quarterly total expenditures,
car ownership, quarterly VMT, and emissions per mile, by decile, for the full
sample of households that own zero, one, or two vehicles. Decile 1 contains
the poorest households and decile 10 the richest. Hydrocarbon, oxides
of nitrogen, and PM10 emissions per mile peak in the seventh decile, while
CO and SO4 emissions peak in the eighth decile. Low emissions per mile
26

Mulawa et al. reports an R-squared of 0.84 for a regression of PM10 on HC.

EPA (1993) summarises a number of studies that nd a similar relationship.
28
Delucchi (2000) says that the MOBILE models might overstate sulphate emissions. Changes in the
conversion rate of sulphur in gasoline into sulphate, however, do not signicantly aect my incidence
estimates.
27

12

1,383
2,384
3,148
3,872
4,662
5,582
6,726
8,309
10,832
19,223

675
1,212
1,713
2,106
2,506
2,936
3,388
3,907
4,009
4,618

Quarterly
Vehicle Miles
Travelled
(VMT)
4.63
6.64
7.31
7.63
8.22
8.43
8.38
8.45
7.96
7.27

CO
(g/mile)
0.51
0.70
0.81
0.80
0.86
0.89
0.90
0.86
0.84
0.75

HC
(g/mile)
0.49
0.66
0.75
0.78
0.82
0.84
0.84
0.84
0.81
0.74

NOx
(g/mile)
6.64
9.07
10.55
10.45
11.18
11.51
11.72
11.19
10.92
9.77

PM10
(mg/mile)

1.41
2.12
2.30
2.65
2.85
2.88
2.96
3.12
3.11
3.09

SO4
(mg/mile)

39.4
57.4
65.7
74.5
81.2
82.6
86.0
89.2
89.6
89.4

Percentage of
households that
own at least one
vehicle

Note: CO is carbon monoxide; HC is hydrocarbons; NOx is oxides of nitrogen; PM10 is particulate matter of aerodynamic diameter of 10 microns or
less; SO4 is sulphate. Decile 1 is the poorest, decile 10 is the richest. The sample includes households that own zero, one, or two vehicles. The total
number of households is 7,073. Numbers are means unless otherwise noted.

1
2
3
4
5
6
7
8
9
10

Decile

Income (Total
Quarterly
Expenditures
$ 1997)

Table 2a
Mean Income, VMT, Emissions per Mile and Vehicle Ownership Percentages by Decile (Full Sample)

Equity Implications of Vehicle Emissions Taxes

West

13

14

1,955
3,118
3,944
4,726
5,573
6,534
7,733
9,395
11,906
20,515

1,825
2,527
2,794
3,087
3,453
3,855
4,109
4,292
4,644
5,250

Quarterly
Vehicle Miles
Travelled
(VMT)
11.57
11.46
10.15
10.38
9.97
9.81
9.66
9.40
8.91
7.87

CO
(g/mile)
1.25
1.26
1.07
1.09
1.05
1.04
1.00
0.96
0.95
0.81

HC
(g/mile)
1.19
1.17
1.04
1.03
1.00
0.98
0.96
0.93
0.90
0.80

NOx
(g/mile)

16.19
16.38
13.93
14.20
13.65
13.53
13.04
12.52
12.40
10.49

PM10
(mg/mile)

3.61
3.57
3.54
3.54
3.48
3.47
3.49
3.48
3.44
3.47

SO4
(mg/mile)

Note: CO is carbon monoxide; HC is hydrocarbons; NOx is oxides of nitrogen; PM10 is particulate matter of aerodynamic diameter of 10 microns or
less; SO4 is sulphate. Decile 1 is the poorest, decile 10 is the richest. The sample includes households that own one or two vehicles (5,343 total). Numbers
are means.

1
2
3
4
5
6
7
8
9
10

Decile

Income (Total
Quarterly
Expenditures
$ 1997)

Table 2b
Income, VMT, and Emissions per Mile by Decile (Vehicle Owners Only)

Journal of Transport Economics and Policy

Volume 39, Part 1

Equity Implications of Vehicle Emissions Taxes

West

gures in the lower deciles of the full sample are due to low vehicle
ownership rates in those deciles. Less than 40 per cent of households in
the lowest expenditure deciles own at least one vehicle; that number is
about 57 per cent for the second decile.
Table 2b shows the summary statistics for vehicle owners only. With the
exception of SO4 , emissions per mile are signicantly higher in the lower
deciles of vehicle owners than in the higher deciles. Sulphate is only
somewhat higher for lower deciles than in higher deciles: variation in
sulphate across deciles is due to variation in fuel eciency, which does
not vary as much across deciles as much as other determinants of emissions
per mile. In sum, poor households that own vehicles own dirtier vehicles
than wealthy vehicle owners.

3.3 VMT elasticity estimation

A tax on vehicle emissions is equivalent to a vehicle-specic tax per mile. In
the short run, households respond to such an increase in vehicle operating
costs by reducing the number of miles they drive. If households in dierent
income groups respond dierently to the tax, this dierence in elasticities
will aect incidence calculations.
The joint nature of the demands for vehicles and miles complicates
estimation of VMT demand elasticities. An unobserved household characteristic that aects the utility of miles driven in a particular vehicle is likely
to aect both its probability of selection and its intensity of use. For
example, a large household may gain more enjoyment from driving in a
spacious vehicle. The household may also have to drive children to more
activities, and so they may drive more miles. Moreover, factors that
aect the intensity of use will aect the probability of choosing particular
vehicle bundles. A person that lives in a region with long commutes
drives more miles and may be more likely to choose a vehicle bundle that
has low operating costs. Cases such as these imply that the residuals in a
miles regression are correlated with vehicle choice indicators. Unless one
controls for the endogeneity of vehicle choice in the determination of
miles travelled, operating cost elasticity estimates will be biased.
West (2004) provides such unbiased elasticity estimates. That study
estimates a model of the joint determination of miles driven and vehicle
attributes in two stages. The rst stage estimates a nested logit on the
choice between owning zero, one, or two vehicles, and within these vehicle
number categories, the vintage and engine size of each vehicle. This discrete
choice model includes variables that might aect a households vehicle
choice, including total expenditures, gender of the household head,
family size, whether the household owns a home, or lives in a large
15

Journal of Transport Economics and Policy

Volume 39, Part 1

metropolitan area, and the number of earners and potential drivers in the
household.
The second stage estimates a regression of vehicle miles travelled as a
function of vehicle operating cost per mile, indicators for vehicle choice,
total expenditures, interactions between total expenditures and vehicle
operating costs, and demographic characteristics. To correct for the
endogeneity of the vehicle choice indicators, West (2004) employs the
conditional expectation correction approach introduced by Dubin and
McFadden (1984). This method corrects for the bias due to the fact that
the vehicle choice indicators are correlated with the error term. This is a
sample-selection correction along the lines of that presented in Heckman
(1979); the VMT regression is therefore estimated on vehicle owners only.
Results from this second stage can be used to calculate many dierent
measures of the elasticity of demand for VMT with respect to operating
costs, all conditional on vehicle choice. The regression results, for example,
can be used to calculate this elasticity evaluated at sample means of miles,
operating costs per mile, and total expenditures. This elasticity equals

0.87.29
This estimate, however, conceals two critical characteristics of miles
demand that vary across the income distribution. First, as shown in
Table 2a, a large number of lower income households do not own vehicles
and therefore drive no miles. In the lower half of the income distribution, as
expenditures increase, spending on miles as a proportion of total expenditures also increases.
Second, lower income households are more responsive to price
changes than are high income households. For the full sample, VMT
elasticities range from
1.51 in the poorest decile to
0.75 in decile 8.
Elasticities for vehicle owners tend to be slightly smaller than full sample
elasticities. Because of how income and operating costs are dened, these
elasticities are not strictly comparable to estimates from previous studies.
However, they are generally larger in absolute value than others.30

29

The expenditure elasticity of demand for VMT calculated at sample means is 0.02. This estimate is
smaller than estimates from other studies and implies that a tax on miles, ignoring the distribution
of benets and possible uses of revenue, would be quite regressive. For other income elasticity
estimates, see Archibald and Gillingham (1981), Hensher et al. (1992), and Mannering and Winston
(1985).
30
For example, Walls et al. (1994) has VMT price elasticity estimates that range from
0.120 to
0.583.
Berkowitz et al. (1990) estimate a VMT price elasticity of
0.21. Similarly, Mannering and Winston
(1985) nd a VMT price elasticity of
0.228, and Hensher et al.s (1992) results range from
0.28 to

0.39. Sevignys (1998) VMT estimates are the only ones that are in the same neighbourhood as those
used here; they range from
0.85 to
0.94.

16

Equity Implications of Vehicle Emissions Taxes

West

4.0 Incidence Calculations

This section discusses results of incidence calculations for two taxes
designed to reduce vehicle pollution. The rst is the vehicle emissions tax,
equivalent to a vehicle-specic tax on miles. While this tax may be the
most ecient way to reduce vehicle pollution, its measurement requirements may render it prohibitively costly it would require the taxing
authority to measure both emissions per mile and vehicle miles travelled.
In addition to facing the technical diculties involved in accurately
measuring emissions per mile, if taxing authorities used odometer readings
to measure VMT, they would face possible cheating on the part of
households, who might roll back their odometers before taking their
vehicles in for inspection. The second policy is a uniform tax on miles,
set so that the revenue it generates equals that generated by the emissions
tax. It would still require the taxing authority to measure VMT but
would not require measurement of emissions per mile. Such a tax would
not dierentiate between dirtier and cleaner vehicles.
4.1 Vehicle emissions tax
As mentioned in Section 2, the ecient tax on emissions equals the
marginal external damages per unit of emissions. Because I consider ve
pollutants, I need estimates of the MED of each. McCubbin and Delucchi
(1999) calculates the dollar value of health costs of CO, NOx , PM10 , oxides
of sulphur in the form of sulphate PM10 , and volatile organic compounds
(VOC). Hydrocarbons are the main component of VOC. I ignore the
distinctions between VOC and HC because these dierences are small
for petroleum based motor vehicle emissions (Small and Kazimi (1995):
p. 8). McCubbin and Delucchi (1999) provides low and high estimates of
health costs per kilogram in 1991 dollars. A meta-hedonic price analysis
in Delucchi et al. (2002) suggests that the true values of MED lie
closer to the low estimates provided in McCubbin and Delucchi (1999). I
therefore use the geometric mean of McCubbin and Delucchis low and
high estimates to obtain one MED per pollutant. Translating these
estimates into dollars per gram and inating to year 1997 dollars using
the Consumer Price Index yields $0.00004 per gram of CO, $0.0004 per
gram of HC, $0.005 per gram of NOx , $0.04 per gram of PM10 , and
$0.025 per gram of SO4 . Multiplying these MED per gram by emissions
per mile yields an average vehicle-specic VMT tax of about one-half
cent ($0.006, to be precise) per mile.31
31

These measures of MED do not include the value of lost crops or damaged ecosystems.

17

18

0.87

0.87

0.87

0.87

0.87

0.87

0.87

0.87

0.87

0.87
0.36
0.34
0.37
0.32
0.32
0.32
0.30
0.28
0.20
0.14

Suits Index
0:200

Suits Index Equivalent for
Consumer Surplus
0:203

0.34
0.33
0.35
0.30
0.30
0.31
0.28
0.27
0.19
0.13

1.51

1.31

1.06

1.01

0.95

0.84

0.78

0.75

0.78

0.83
0.31
0.31
0.34
0.30
0.30
0.31
0.29
0.27
0.20
0.13

0.34
0.34
0.36
0.32
0.32
0.32
0.30
0.28
0.20
0.14

1.46

1.09

1.02

0.94

0.86

0.79

0.78

0.77

0.77

0.84

0.65
0.55
0.42
0.39
0.35
0.36
0.30
0.27
0.21
0.14

0.71
0.58
0.44
0.41
0.37
0.38
0.32
0.28
0.22
0.14

Tax as % of
CS as % of
Total
Total
Expenditures Expenditures

Suits Index
0:253

Suits Index Equivalent for
Consumer Surplus
0:288

Tax as % of
CS as % of
Total
Total
Decile-Specic
Expenditures Expenditures
Elasticities

Suits Index
0:193

Suits Index Equivalent for
Consumer Surplus
0:199

Tax as % of
CS as % of
Total
Total
Decile-Specic
Expenditures Expenditures
Elasticities

Vehicle Owners Only

Note:
CS is the change in consumer surplus dened in equation (2). Elasticities are taken from West (2004). Tax as a percentage of Total Expenditures
and
CS as a percentage of Total Expenditures are means for each decile.

1
2
3
4
5
6
7
8
9
10

Decile

Elasticity at
Sample Means

Full Sample

Table 3
Vehicle Emissions Tax Incidence

Journal of Transport Economics and Policy

Volume 39, Part 1

Equity Implications of Vehicle Emissions Taxes

West

Each household faces its own vehicle-specic VMT tax and responds by
reducing VMT. The rst two panels of Table 3 present incidence estimates
for the full sample of both vehicle owners and households that do not own
vehicles. The left-hand panel shows results assuming that all households
have the same degree of price responsiveness, while the middle panel shows
results allowing price responsiveness to dier by income group. Tax burden,
for both the tax paid and the consumer surplus measures of welfare loss as
a percentage of total expenditures, peaks in decile three of the full sample.
Suits indexes indicate that if we were to ignore the distribution of
benets and possible uses of tax revenue and focus only on costs, the
emissions tax would be quite regressive. The traditional Suits index using
elasticities calculated at sample means is
0.200. Allowing elasticities to
vary across deciles results in a less regressive Suits index of
0.193; poor
households are more price responsive than wealthy households and
therefore avoid more of the tax by reducing VMT by a greater proportion.
Because of this greater price responsiveness, however, poor households
Harberger triangles are larger than those of wealthy households. This
results in Suits index equivalents for consumer surplus that are more
regressive than the traditional Suits indexes based on tax paid.
As a comparison, the Suits index for the taxes considered by Suits (1977)
ranged from
0.17 to 0.36, so the emissions tax appears more regressive than
the most regressive tax (the payroll tax) considered in that study. My
estimates of the traditional Suits index, however, are less regressive than
those found in the two studies that consider the incidence of a tax on vehicle
emissions. Sevigny (1998) reports a traditional Suits index of
0.226 and that
in Walls and Hanson (1999) is
0.24. Their results are more regressive for
three reasons. First, they do not incorporate price responsiveness that
varies across deciles. Second, they do not consider SO4 , which is more
evenly distributed across households than other local pollutants. Third,
Walls and Hansons (1999) data contain fewer households that do not
own vehicles and Sevignys (1998) data contain vehicle owners only.
Indeed, consider the results for an emissions tax on vehicle owners only,
presented in the right-hand panel of Table 3. Ratios of tax paid and
consumer surplus losses to total expenditures are highest in the lowest
decile and decrease substantially as total expenditures increase. Resulting
Suits indexes for emissions taxes are thus more regressive among vehicle
owners only.

4.2 Uniform VMT tax

A uniform tax on VMT would not be ecient because it would penalise
driving in clean vehicles as much as driving in dirty vehicles. This tax,
19

20

0.87

0.87

0.87

0.87

0.87

0.87

0.87

0.87

0.87

0.87
0.26
0.29
0.31
0.31
0.31
0.30
0.29
0.27
0.22
0.15

Suits Index
0:153

Suits Index Equivalent for
Consumer Surplus
0:153

0.26
0.28
0.30
0.30
0.30
0.29
0.28
0.27
0.21
0.15

1.51

1.31

1.06

1.01

0.95

0.84

0.78

0.75

0.78

0.83
0.24
0.27
0.29
0.30
0.30
0.29
0.28
0.27
0.21
0.15

0.26
0.28
0.31
0.31
0.31
0.30
0.29
0.27
0.22
0.15

1.46

1.09

1.02

0.94

0.86

0.79

0.78

0.77

0.77

0.84

0.54
0.46
0.41
0.38
0.36
0.34
0.31
0.27
0.23
0.16

0.57
0.48
0.42
0.39
0.37
0.35
0.32
0.27
0.23
0.17

Tax as % of
CS as % of
Total
Total
Expenditures Expenditures

Suits Index
0:208

Suits Index Equivalent for
Consumer Surplus
0:239

Tax as % of
CS as % of
Total
Total
Decile-Specic
Expenditures Expenditures
Elasticities

Suits Index
0:149

Suits Index Equivalent for
Consumer Surplus
0:151

Tax as % of
CS as % of
Total
Total
Decile-Specic
Expenditures Expenditures
Elasticities

Vehicle Owners Only

Note:
CS is the change in consumer surplus dened in equation (2). Elasticities are taken from West (2004). Tax as a percentage of Total Expenditures
and
CS as a percentage of Total Expenditures are means for each decile.

1
2
3
4
5
6
7
8
9
10

Decile

Elasticity at
Sample Means

Full Sample

Table 4
Uniform VMT Tax Incidence

Journal of Transport Economics and Policy

Volume 39, Part 1

Equity Implications of Vehicle Emissions Taxes

West

however, would require less information than an emissions tax and still
reduce pollution by reducing miles driven.
In order to simulate a uniform VMT tax rate that is similar in magnitude
to the household-specic emissions tax rate, I nd and simulate uniform
VMT tax rates that generate the same amount of revenue as the emissions
taxes. These VMT tax rates dier depending on whether the full sample or
the vehicle-owners only sample is used, and whether the sample-mean
elasticity or decile-specic elasticities are used. They all fall in the
neighbourhood of $0.006 per mile.
Table 4 reports the results from this simulation. Since poor households
drive dirtier vehicles than wealthy households, a uniform tax on VMT that
does not distinguish among vehicles is less regressive than the emissions tax.
Suits index equivalents for consumer surplus for the full sample using the
elasticity evaluated at sample means and decile-specic elasticities are

0.153 and
0.151, respectively. In the case of vehicle emissions of local
pollutants, the pollution control policy that is easier to implement is also
less regressive.

5.0 Conclusion
This paper combines data on emissions per mile from the California Air
Resources Board and household level vehicle and income data from the
US Consumer Expenditure Survey to calculate the incidence of a tax on
vehicle emissions. Incidence calculations allow for household price
responsiveness to dier across income groups and include both households
that own vehicles and those that do not. It compares the incidence of an
emissions tax with that of imposing a uniform miles tax.
While this paper focuses on ve local pollutants, it would be fruitful
to analyse the incidence of a vehicle emissions tax that includes the
marginal external costs of global warming gases such as carbon
dioxide. Since emissions of carbon dioxide are proportional to fuel
use, we might expect the incidence of a carbon tax to resemble that of a
gasoline tax, a tax that is less regressive than the emissions tax considered
here.
The analysis undertaken here is short run and partial equilibrium in
nature. Future research might consider the partial equilibrium incidence
of vehicle pollution taxes in the long run, where households respond to
the emissions tax not only by reducing the number of miles they drive
but also by switching to newer, smaller, better-maintained, or hybrid
vehicles. Other research might conduct incidence analysis in a general
21

Journal of Transport Economics and Policy

Volume 39, Part 1

equilibrium context wherein increases in vehicle operating costs aect

producer prices of fuel, vehicles, and other sectors of the economy.
I assume that the government tax revenue generated by an emissions tax is
discarded. Environmental tax reforms measures that use pollution tax
revenue to reduce taxes on employment or investment are now common
in Europe. Denmark, Finland, Germany, Italy, the Netherlands, Norway,
Sweden, and the United Kingdom have all implemented such reforms
(Hoerner and Bosquet (2001)). Such reforms have also been proposed in
state legislatures in the United States (Hoerner and Erickson (2000)).
Future research could apply methods used to conduct incidence analysis of
environmental tax reforms wherein revenue is used to reduce other taxes
(see for example Brannlund and Nordstrom (2004), DeBorger and Proost
(2001), and Metcalf (1999)), to reforms involving vehicle emissions taxes.
Ignoring possible uses for the tax revenue, the burden of a tax on vehicle
emissions falls relatively more heavily on poor households than on wealthy
ones. This is due to the fact that poor vehicle owners spend more on miles as
a proportion of their income and drive vehicles that pollute more per mile
than vehicles owned by the wealthy. Low vehicle ownership rates and high
price responsiveness in the lower half of the income distribution, however,
mitigates regressivity to some extent. On the other hand, while high price
responsiveness reduces taxes paid by poor households, it increases
Harberger triangles of consumer surplus loss among poor households
and checks the mitigating eect. Overall, however, the results of this
analysis suggest that a tax on vehicle emissions is less regressive than
previously thought.
A uniform tax on miles that does not distinguish between dirty and clean
vehicles is less regressive than the emissions tax. If choosing between an
emissions tax and a miles tax, policy makers should weigh the greater
eciency gains of an emissions tax with the lower implementation costs
and more favourable distributional consequences of a miles tax.

References
Archibald, R. and R. Gillingham (1981): A Decomposition of the Price and Income
Elasticities of the Consumer Demand for Gasoline, Southern Economic Journal, 47,
102131.
Baumol, W. J. and W. E. Oates (1988): The Theory of Environmental Policy, Cambridge
University Press, Cambridge, UK.
Berkowitz, M. K., N. T. Gallini, E. J. Miller, and R. A. Wolfe (1990): Disaggregate
Analysis of the Demand for Gasoline, Canadian Journal of Economics, 23, 25375.
Bin, O. (2003): A Logit Analysis of Vehicle Emissions Using Inspection and Maintenance
Testing Data, Transportation Research Part D, 8, 21527.

22

Equity Implications of Vehicle Emissions Taxes

West

Board on Environmental Studies and Toxicology (BEST), National Research Council

(2001): Evaluating Vehicle Emissions Inspection and Maintenance Programs, National
Academy Press, Washington DC.
Bohm, P. and C. S. Russell (1985): Comparative Analysis of Alternative Policy
Instruments, in A. Kneese and J. L. Sweeney, eds., Handbook of Natural Resource
and Energy Economics, North-Holland, Amsterdam, 395460.
Brannlund, R. and J. Nordstrom (2004): Carbon Tax Simulations using a Household
Demand Model, European Economic Review, 48, 21133.
Brooks, N. and R. Sethi (1997): The Distribution of Pollution: Community Characteristics
and Exposure to Air Toxics, Journal of Environmental Economics and Management, 32,
23350.
Congressional Budget Oce (CBO) (2003): The Economic Costs of Fuel Economy Standards
versus a Gasoline Tax, The Congress of the United States, Congressional Budget Oce,
Washington DC.
Cornwell, A. and J. Creedy (1997): Measuring the Welfare Eects of Tax Changes using
the LES: An Application to a Carbon Tax, Empirical Economics, 22, 589613.
Dahl, C. and T. Sterner (1991): Analyzing Gasoline Demand Elasticities: A Survey,
Energy Economics, 13, 20310.
DeBorger, Bruno and S. Proost (eds.) (2001): Reforming Transport Pricing in the European
Union: A Modeling Approach, Elgar, Cheltenham UK.
Delucchi, M. A. (2000): Analysis of Particulate Matter Emission Factors in the PART5
Model, ITS-RR-00-21, Institute of Transportation Studies, University of California,
Davis.
Delucchi, M. A., J. J. Murphy, and D. R. McCubbin (2002): The Health and
Visibility Cost of Air Pollution: A Comparison of Estimation Methods, Journal of
Environmental Management, 64, 13952.
Dill, J., T. Goldman, and M. Wachs (1999): California Vehicle License Fees: Incidence
and Equity, Journal of Transportation and Statistics, 2, 13347.
Dubin, J. A. and D. L. McFadden (1984): An Econometric Analysis of Residential
Electric Appliance Holdings and Consumption, Econometrica, 52, 34562.
Environmental Protection Agency (EPA) (1993): Motor Vehicle-Related Air Toxics Study,
Oce of Mobile Sources, Emission Planning and Strategies Division, EPA 420-R-93005.
Environmental Protection Agency (EPA) (1998): EPA Sta Paper on Gasoline Sulphur
Issues, Oce of Mobile Sources, EPA420-R-98-005.
Federal Highway Administration (FHWA): [Link]
[Link].
Goldberg, P. (1998): Eects of the Corporate Average Fuel Eciency Standards in the
U.S., Journal of Industrial Economics, 46, 133.
Greene, D. L. and P. S. Hu (1985): The Inuence of the Price of Gasoline on Vehicle Use
in Multi-vehicle Households, Transportation Research Record, 988, 1924.
Harberger, A. (1962): The Incidence of the Corporation Income Tax, Journal of Political
Economy, 70, 21540.
Harrington, W. (1997): Fuel Economy and Motor Vehicle Emissions, Journal of
Environmental Economics and Management, 33, 24052.
Harrington, W., M. Walls, and V. McConnell (1994): Shifting Gears: New Directions
for Cars and Clean Air, Discussion Paper 94-26-REV, Resources for the Future,
Washington DC.
Heckman, J. (1979): Sample Selection Bias as Specication Error, Econometrica, 47,
15361.

23

Journal of Transport Economics and Policy

Volume 39, Part 1

Hensher, D. A., N. C. Smith, F. W. Milthorpe, and P. Barnard (1992): Dimensions of Automobile Demand: A Longitudinal Study of Household Automobile Ownership and Use,
North-Holland, Amsterdam.
Hoerner, J. and B. Bosquet (2001): Environmental Tax Reform: The European Experience,
Center for a Sustainable Economy, Washington DC.
Hoerner, J. and G. Erickson (2000): Environmental Tax Reform in the States:
A Framework for Assessment, State Tax Notes (31 July), 31119.
Kahn, M. E. (1996a): New Evidence on Trends in Vehicle Emissions, Rand Journal of
Economics, 27, 18396.
Kahn, M. E. (1996b): The Eciency and Equity of Vehicle Emissions Regulation:
Evidence from Californias Random Audits, Eastern Economic Journal, 22, 45765.
Mannering, F. and C. Winston (1985): A Dynamic Empirical Analysis of Household
Vehicle Ownership and Utilization, Rand Journal of Economics, 16, 21336.
McCubbin, D. R. and M. A. Delucchi (1999): The Health Costs of Motor-VehicleRelated Air Pollution, Journal of Transport Economics and Policy, 33, 25386.
Metcalf, G. E. (1999): A Distributional Analysis of Green Tax Reforms, National Tax
Journal, 52, 65582.
Mulawa, P. A., S. H. Cadle, H. Knapp, R. Zweidinger, R. Snow, R. Lucas, and
J. Goldbach (1997): Eect of Ambient Temperature and E-10 Fuel on Primary
Exhaust Particulate Matter Emissions from Light-Duty Vehicles, Environmental
Science and Technology, 31, 130207.
Oak Ridge National Laboratory (ORNL), Davis, S. (1998): Transportation Energy Data
Book, 18th ed., Oak Ridge National Laboratory, Oak Ridge, TN.
Oak Ridge National Laboratory (ORNL), Davis, S. (2000): Transportation Energy Data
Book, 20th ed., Oak Ridge National Laboratory, Oak Ridge, TN.
Pigou, A. (1932): The Economics of Welfare, 4th ed., Macmillan Tax Policy and the
Economy 5, MIT Press, Boston, 14564.
Sevigny, M. (1998): Taxing Automobile Emissions for Pollution Control, Edward Elgar
Publishing Ltd., Cheltenham, UK and Northhampton, MA.
Slesnick, D. T. (2001): Consumption and Social Welfare: Living Standards and Their Distribution in the United States, Cambridge University Press, Cambridge, UK, and New York.
Small, K. A. (1983): The Incidence of Congestion Tolls on Urban Highways, Journal of
Urban Economics, 13, 90111.
Small, K. A. and C. Kazimi (1995): On the Costs of Air Pollution from Motor Vehicles,
Journal of Transport Economics and Policy, 29, 732.
Suits, D. B. (1977): Measurement of Tax Progressivity, American Economic Review, 67,
74752.
Walls, M. and J. Hanson (1999): Distributional Aspects of an Environmental Tax Shift:
The Case of Motor Vehicle Emissions Taxes, National Tax Journal, 52, 5365.
Walls, M. A., A. J. Krupnick, and H. C. Hood (1994): Estimating the Demand for VehicleMiles Traveled Using Household Survey Data: Results from the 1990 Nationwide Transportation Survey, Discussion Paper 93-25 Rev, Resources for the Future, Washington, DC.
West, S. (2004): Distributional Eects of Alternative Vehicle Pollution Control Policies,
Journal of Public Economics, 88, 73557.
West, S. and R. C. Williams, III (2004): Estimates from a Consumer Demand System:
Implications for the Incidence of Environmental Taxes, Journal of Environmental Economics and Management, 47, 53558.
Willig, R. (1976): Consumers Surplus Without Apology, American Economic Review,
66, 58997.

24