Air pollution concentrations have been rapidly increasing in the major urban areas of Brazil caused mainly by the increasing use of vehicles. Policies to control car emissions in Brazil have relied basically on mandatory emission standards and subsidies for specific cleaner technology resulting in substantial decrease of car emission rates. Nevertheless, taxes on car sales, differentiated by vehicles' size and fuel, have also influenced car emission patterns. This paper analyses the compliance trend of the Brazilian fleet with environmental standards between 1992 and 1997. We find that larger automobiles had the fastest compliance schedule while popular models adjusted very slowly. Also gasoline-fuelled models had a faster adjustment pattern than ethanol cars. Additionally, we analyse the current relationship between pollution emissions and car characteristics in order to
orient policy formulation. We find a positive relationship between emissions rates and horse power, concluding that although the current value-added sale car tax is not environmental harmful, a tax differentiating clean from dirty models, within each tax bracket, could create substantial incentives for emission control in the future.
Urban air pollution is a serious environmental problem in developed as well as in most
developing countries. In the case of Brazil, air pollution concentrations have been rapidly
increasing in the major urban areas over the last decades. As elsewhere, this expansion has
been caused mainly by the increasing use of vehicles. Today, emissions from vehicles are the
major source of air pollution in Brazil's largest cities. In 1997 in São Paulo, for example,
private cars were responsible for approximately 75% of carbon monoxide (CO), 73% of
hydrocarbons (HC), 23% of nitrogen oxides (NOx) and 10% particulate matter (PM)
Costs associated with high air pollution concentrations in large cities are known to be
important. Human health costs predominate, and range from eye irritations to respiratory
problems and increasing cancer rates, all of which induce direct and indirect costs to society2.
They also estimate the health costs associated with concentration levels in
excess of air pollution standards, finding a loss of approximately US$ 700 million per year in
the early 1990s.
Even when consumers can perceive individual emission damage, they are unable to reduce
alone the aggregate social emission costs. Consequently, their preferences will usually not
consider fuel and car cleanliness. In the presence of this negative externality, environmental
regulation is required.
If we were able to measure emissions by individual cars, the first best incentive option for car
emission control would be the imposition of a Pigovian tax on each source according to its
marginal contribution to air pollution damages. This would allow flexibility for car owners in
the choice of emission reduction strategies. However, such first best approaches can incur
high administrative cost. As put by Innes (1996), even if tamper-resistant emissionmeasurement
from tailpipes were available at reasonable costs, such devices do not detect
important non-tailpipe pollution and, therefore, high costly reliable periodic car monitoring
would be required. Consequently, the application of car emission control policies would have
to reckon on regimes which do not require direct emission monitoring.
When emission output measurements are difficult, the economic literature on MBIs proposes
instead that regulators may apply first best taxes on the use of inputs and products which are
related to emissions. For car emissions, fuel and automobile taxes are good candidates for
this option. Fullerton and West (1999 - hereafter FW), have derived a set of fuel and car
optimal taxes which are able to mimic, at least in theory, the unavailable tax on emissions. In
order to derive a closed form solution, FW consider emissions per gallon (EPG) and miles
per gallon (MPG) only to depend on si, the size of the car. Under these specific technological
conditions, FW propose a closed form solution for a fuel tax (tg) differing according to
characteristics of the vehicle at the pump.
The owner of car model i, would pay a tax given by
)( ) ( i i gi s MPG s EPM t : =
where :, represents the marginal social cost of a unit of emissions and EPG and MPG
represent car features. More generally, we could specify such a tax to be a function of other
car characteristics which are likely to affect EPG and MPG, as well as emission features of
different fuel types.
Admitting regulators know the mileage consumption and useful life of each car model i
owned by consumer j, an equivalent car sale tax would consist of the present value of the
above fuel tax. This car tax could be, instead, applied periodically for licensing purposes, and
its value would be set by monitored mileage at that period.
Both fuel and car taxes would make consumers perceive the emission-increasing cost of extra
mileage consumption and recognise the emission-reducing benefits of fuel cleanness and
economy as well as car pollution abatement devices.
Note that under this approach regulators must know the marginal social cost of a unit of
emissions ( :). Moreover, this parameter will be location-specific since marginal damages are
dependant on total pollution charges and the environment assimilative capacity which, in
turn, varies according to atmospheric variables (e.g., wind speed, temperature, humidity, etc).
Suppose, however, that regulators know : and location-specific taxes can be applied.
Although the fuel tax is simpler than the equivalent car tax, it would still require car features
to be identifiable at the gas station. Again such an approach is likely to generate high
administrative costs in order to be feasible and reliable. Therefore, if we cannot mimic the
first best solution with alternative taxation schemes, we would have to rely on second best
market instruments. The ideal second best mechanism should create price incentives for
consumers to drive fewer miles and, at the same time, buy cleaner cars. While the former
decision is related to fuel use, the latter works through car price differentiation3.
A car tax based on the estimation of a vehicle's annual emission is proposed by Eskeland
(1994) and Sevigny (1998). Emission rates per mile would be estimated based on car
characteristics, and miles travelled could be measured by the change in the vehicle's
odometer in a given year4. What makes this proposal different from the first best alternative
proposed by FW is the lack of knowledge on each specific car's EPM. While FW's model
assumes that it is possible to estimate individual EPM for all car models in order to charge an
emission gas tax at the fuel pump, Sevigny (1998) only expects to be able to derive average
EPM figures. Again, implementation may prove to be costly for the case of odometer
measurement procedures.
An alternative constrained optimal regulation is proposed by Innes (1996). A combination of
taxes on gasoline and automobiles could be combined with a government fuel content
standard. The fuel tax would be independent of individual automobiles, but the car tax would
depend on auto characteristics (eg, power, size, style), fuel economy and abatement features.
Since mileage demand is highly correlated with automobile features, the automobile tax
would also affect miles driven. Moreover, additional incentives could be created. In a
dynamic setting, a car sale tax could be partially returned to consumers as incentive for
scrapping older models according to the abbreviation of the car's useful life5.
For the previous mechanism, a subsidy for pollution control equipment is relatively simple to
define since control equipment such as catalytic converters and filters are directly observable.
The same applies for fuel, insofar as that emissions will rise more or less proportionately with
fuel consumption for a given vehicle and given driving conditions6. On the other hand, a tax
on car characteristics requires a periodic identification of the relationship between car
characteristics and emissions.
In general, market based instruments are difficult to implement and regulators wishing to
apply them would have to combine tax schemes with technological and emission standards.
This approach does not maximise social welfare by setting optimal levels of pollution.
Rather, the aim is to use pricing mechanisms to increase cost-effectiveness in achieving a
certain standard compliance regarded as desirable7. That is, once environmental goals are
defined, economic instruments can reduce the social costs of achieving them. The rationale is
rather simple. Since users face different marginal control cost schedules, pollution taxes
varying directly with users' pollution levels will make users adopt control levels up to the
point where pollution control costs are equal, at the margin, to non-compliance tax costs.
Taxes are set at certain level which will make the maximised individual decisions, in
aggregate, to meet the desirable standards. Society will first start to control from the leastcost
users which will reduce total control cost. In this case, individual emission standards are
3 See Johnstone and Karousakis (1998) for a review.
4 This tax would take the form of Annual tax =, VMT cEPM bEPM aEPM * ) ( 3 2 1 + + where EPM is the emission
per mile, VMT is the vehicle mile traveled and a, b, c are tax rates set to induce the desired level of abatement
for each pollutant.
5 Road fees, varying with air pollution concentration levels and car's characteristics, are also regarded as
possible second best options, although they may induce longer travel distances to avoid charges.
6 See Seroa da Motta and Mendes (1996) for an analysis of fuel taxation in Brazil.
7 Desirable here may mean either one standard politically acceptable or targeted at one specific damage. In the
case of urban pollution, human health damages are usually targeted.
CREED Working Paper Series No 29 4
dropped out and regulators become only concerned with ambient standards which reflect total
emissions.
Instead of setting prices, regulators may distribute pollution permits to users, as a share of the
desired total emission targets and allow these permits to be traded among users. This
mechanism creates incentives for achieving the same marginal cost equalising outcome given
the competitive structure of the prices emerging from the permit market transactions.
Apart from cost-effectiveness benefits, such economic instruments can generate revenue. In
the case of pricing mechanisms, note that users with non-compliance will face the respective
tax costs and thereby generate a tax revenue. In the case of tradable permits, they can be
distributed through auctioning mechanisms8.
Car emission control policy in Brazil is basically defined on mandatory emission standards.
Since 1988 the Brazilian governmental authorities have implemented a regulatory mechanism
called The Vehicle Air Pollution Control Programme (PROCONVE), establishing maximum
pollution emission standards (in grams per kilometer) for new vehicles entering the market.
The program has been very successful in reducing emissions per kilometer for new car
models. Nevertheless, as with any command and control approach, it is inflexible and
increases the costs of reducing pollution.
This paper analyses the evolution of average emissions in the Brazilian fleet between 1992
and 1997 (the final compliance date). Together with environmental policy, car tax structures
have been differentiated by car characteristics such as size and fuel use, in order to
accommodate sectoral policy aims. Therefore, government initiatives have affected the car
market and consequently the emission pattern of new automobiles. We try to relate these
sectoral policies with the average emission compliance trends comparing the average
emission changes across car sizes and fuel types. Additionally, we analyse the present
relationship between pollution emissions and car characteristics in order to orient current
policy formulation. Our analysis is based on emission data recorded from laboratory tests
undertaken by the São Paulo Environmental Agency (CETESB) which electronically
measures emissions of HC, CO and NOx for each car model along with the model's
characteristics.
The rest of this paper is structured as follows: the next section presents the Brazilian
regulatory framework for car pollution control and the car tax structure. This is followed by a
description of our database and model characteristics. Section 4 presents the econometric
results; concluding remarks and policy recommendations are discussed in the final section.
8 Cost-effectiveness of permits does not depend on permit auctioning. Freely distributed permits have different
equity effects, although they are equally cost-effective if transactions costs are assumed not constrained.
orient policy formulation. We find a positive relationship between emissions rates and horse power, concluding that although the current value-added sale car tax is not environmental harmful, a tax differentiating clean from dirty models, within each tax bracket, could create substantial incentives for emission control in the future.
Urban air pollution is a serious environmental problem in developed as well as in most
developing countries. In the case of Brazil, air pollution concentrations have been rapidly
increasing in the major urban areas over the last decades. As elsewhere, this expansion has
been caused mainly by the increasing use of vehicles. Today, emissions from vehicles are the
major source of air pollution in Brazil's largest cities. In 1997 in São Paulo, for example,
private cars were responsible for approximately 75% of carbon monoxide (CO), 73% of
hydrocarbons (HC), 23% of nitrogen oxides (NOx) and 10% particulate matter (PM)
Costs associated with high air pollution concentrations in large cities are known to be
important. Human health costs predominate, and range from eye irritations to respiratory
problems and increasing cancer rates, all of which induce direct and indirect costs to society2.
They also estimate the health costs associated with concentration levels in
excess of air pollution standards, finding a loss of approximately US$ 700 million per year in
the early 1990s.
Even when consumers can perceive individual emission damage, they are unable to reduce
alone the aggregate social emission costs. Consequently, their preferences will usually not
consider fuel and car cleanliness. In the presence of this negative externality, environmental
regulation is required.
If we were able to measure emissions by individual cars, the first best incentive option for car
emission control would be the imposition of a Pigovian tax on each source according to its
marginal contribution to air pollution damages. This would allow flexibility for car owners in
the choice of emission reduction strategies. However, such first best approaches can incur
high administrative cost. As put by Innes (1996), even if tamper-resistant emissionmeasurement
from tailpipes were available at reasonable costs, such devices do not detect
important non-tailpipe pollution and, therefore, high costly reliable periodic car monitoring
would be required. Consequently, the application of car emission control policies would have
to reckon on regimes which do not require direct emission monitoring.
When emission output measurements are difficult, the economic literature on MBIs proposes
instead that regulators may apply first best taxes on the use of inputs and products which are
related to emissions. For car emissions, fuel and automobile taxes are good candidates for
this option. Fullerton and West (1999 - hereafter FW), have derived a set of fuel and car
optimal taxes which are able to mimic, at least in theory, the unavailable tax on emissions. In
order to derive a closed form solution, FW consider emissions per gallon (EPG) and miles
per gallon (MPG) only to depend on si, the size of the car. Under these specific technological
conditions, FW propose a closed form solution for a fuel tax (tg) differing according to
characteristics of the vehicle at the pump.
The owner of car model i, would pay a tax given by
)( ) ( i i gi s MPG s EPM t : =
where :, represents the marginal social cost of a unit of emissions and EPG and MPG
represent car features. More generally, we could specify such a tax to be a function of other
car characteristics which are likely to affect EPG and MPG, as well as emission features of
different fuel types.
Admitting regulators know the mileage consumption and useful life of each car model i
owned by consumer j, an equivalent car sale tax would consist of the present value of the
above fuel tax. This car tax could be, instead, applied periodically for licensing purposes, and
its value would be set by monitored mileage at that period.
Both fuel and car taxes would make consumers perceive the emission-increasing cost of extra
mileage consumption and recognise the emission-reducing benefits of fuel cleanness and
economy as well as car pollution abatement devices.
Note that under this approach regulators must know the marginal social cost of a unit of
emissions ( :). Moreover, this parameter will be location-specific since marginal damages are
dependant on total pollution charges and the environment assimilative capacity which, in
turn, varies according to atmospheric variables (e.g., wind speed, temperature, humidity, etc).
Suppose, however, that regulators know : and location-specific taxes can be applied.
Although the fuel tax is simpler than the equivalent car tax, it would still require car features
to be identifiable at the gas station. Again such an approach is likely to generate high
administrative costs in order to be feasible and reliable. Therefore, if we cannot mimic the
first best solution with alternative taxation schemes, we would have to rely on second best
market instruments. The ideal second best mechanism should create price incentives for
consumers to drive fewer miles and, at the same time, buy cleaner cars. While the former
decision is related to fuel use, the latter works through car price differentiation3.
A car tax based on the estimation of a vehicle's annual emission is proposed by Eskeland
(1994) and Sevigny (1998). Emission rates per mile would be estimated based on car
characteristics, and miles travelled could be measured by the change in the vehicle's
odometer in a given year4. What makes this proposal different from the first best alternative
proposed by FW is the lack of knowledge on each specific car's EPM. While FW's model
assumes that it is possible to estimate individual EPM for all car models in order to charge an
emission gas tax at the fuel pump, Sevigny (1998) only expects to be able to derive average
EPM figures. Again, implementation may prove to be costly for the case of odometer
measurement procedures.
An alternative constrained optimal regulation is proposed by Innes (1996). A combination of
taxes on gasoline and automobiles could be combined with a government fuel content
standard. The fuel tax would be independent of individual automobiles, but the car tax would
depend on auto characteristics (eg, power, size, style), fuel economy and abatement features.
Since mileage demand is highly correlated with automobile features, the automobile tax
would also affect miles driven. Moreover, additional incentives could be created. In a
dynamic setting, a car sale tax could be partially returned to consumers as incentive for
scrapping older models according to the abbreviation of the car's useful life5.
For the previous mechanism, a subsidy for pollution control equipment is relatively simple to
define since control equipment such as catalytic converters and filters are directly observable.
The same applies for fuel, insofar as that emissions will rise more or less proportionately with
fuel consumption for a given vehicle and given driving conditions6. On the other hand, a tax
on car characteristics requires a periodic identification of the relationship between car
characteristics and emissions.
In general, market based instruments are difficult to implement and regulators wishing to
apply them would have to combine tax schemes with technological and emission standards.
This approach does not maximise social welfare by setting optimal levels of pollution.
Rather, the aim is to use pricing mechanisms to increase cost-effectiveness in achieving a
certain standard compliance regarded as desirable7. That is, once environmental goals are
defined, economic instruments can reduce the social costs of achieving them. The rationale is
rather simple. Since users face different marginal control cost schedules, pollution taxes
varying directly with users' pollution levels will make users adopt control levels up to the
point where pollution control costs are equal, at the margin, to non-compliance tax costs.
Taxes are set at certain level which will make the maximised individual decisions, in
aggregate, to meet the desirable standards. Society will first start to control from the leastcost
users which will reduce total control cost. In this case, individual emission standards are
3 See Johnstone and Karousakis (1998) for a review.
4 This tax would take the form of Annual tax =, VMT cEPM bEPM aEPM * ) ( 3 2 1 + + where EPM is the emission
per mile, VMT is the vehicle mile traveled and a, b, c are tax rates set to induce the desired level of abatement
for each pollutant.
5 Road fees, varying with air pollution concentration levels and car's characteristics, are also regarded as
possible second best options, although they may induce longer travel distances to avoid charges.
6 See Seroa da Motta and Mendes (1996) for an analysis of fuel taxation in Brazil.
7 Desirable here may mean either one standard politically acceptable or targeted at one specific damage. In the
case of urban pollution, human health damages are usually targeted.
CREED Working Paper Series No 29 4
dropped out and regulators become only concerned with ambient standards which reflect total
emissions.
Instead of setting prices, regulators may distribute pollution permits to users, as a share of the
desired total emission targets and allow these permits to be traded among users. This
mechanism creates incentives for achieving the same marginal cost equalising outcome given
the competitive structure of the prices emerging from the permit market transactions.
Apart from cost-effectiveness benefits, such economic instruments can generate revenue. In
the case of pricing mechanisms, note that users with non-compliance will face the respective
tax costs and thereby generate a tax revenue. In the case of tradable permits, they can be
distributed through auctioning mechanisms8.
Car emission control policy in Brazil is basically defined on mandatory emission standards.
Since 1988 the Brazilian governmental authorities have implemented a regulatory mechanism
called The Vehicle Air Pollution Control Programme (PROCONVE), establishing maximum
pollution emission standards (in grams per kilometer) for new vehicles entering the market.
The program has been very successful in reducing emissions per kilometer for new car
models. Nevertheless, as with any command and control approach, it is inflexible and
increases the costs of reducing pollution.
This paper analyses the evolution of average emissions in the Brazilian fleet between 1992
and 1997 (the final compliance date). Together with environmental policy, car tax structures
have been differentiated by car characteristics such as size and fuel use, in order to
accommodate sectoral policy aims. Therefore, government initiatives have affected the car
market and consequently the emission pattern of new automobiles. We try to relate these
sectoral policies with the average emission compliance trends comparing the average
emission changes across car sizes and fuel types. Additionally, we analyse the present
relationship between pollution emissions and car characteristics in order to orient current
policy formulation. Our analysis is based on emission data recorded from laboratory tests
undertaken by the São Paulo Environmental Agency (CETESB) which electronically
measures emissions of HC, CO and NOx for each car model along with the model's
characteristics.
The rest of this paper is structured as follows: the next section presents the Brazilian
regulatory framework for car pollution control and the car tax structure. This is followed by a
description of our database and model characteristics. Section 4 presents the econometric
results; concluding remarks and policy recommendations are discussed in the final section.
8 Cost-effectiveness of permits does not depend on permit auctioning. Freely distributed permits have different
equity effects, although they are equally cost-effective if transactions costs are assumed not constrained.
No comments:
Post a Comment