In 2008, France introduced a new feebate, combining taxes and subsidies for new cars based on their carbon emissions (CO2). The purchase of low-emission vehicles is encouraged through a rebate (‘bonus’) that reaches up to €1,000, while the purchase of high-emission cars is discouraged through a tax (‘malus’) up to €2,600. In advance of the publication of her paper in ‘The Economic Journal’, TSE’s Isis Durrmeyer discusses the effects of the policy on French households, manufacturers, and the environment.
Why is it important to consider a policy’s differentiated impacts on individuals?
Policy evaluation tends to focus on the average or overall impact, but distributional effects can be crucial to public acceptance. For instance, France’s “yellow vests” protests began in October 2018 in response to an increase in the diesel tax that was expected to have little impact on the overall population but severely penalized suburban and rural households that rely heavily on cars. Another example is the strong reaction to a small subway fare price increase in October 2019 in Santiago, Chile. Distributional effects are especially relevant when a policy – such as the 2008 French feebate – favors some individuals at the expense of others.
What was the expected impact of the feebate on consumers?
There are two types of gains and losses for French households: a direct effect through the amount spent by individuals and the characteristics of their car purchases and an indirect effect on air quality. By targeting CO2 emissions, the policy favors diesel cars, associated with higher emissions of the most hazardous air pollutants: nitrogen oxide (NOX) and particulate matter (PM). The increased emissions of these pollutants are a hidden cost of the feebate that my evaluation accounts for. NOX, PM, and other local pollutants downgrade the air quality near their origin, so the question of where they are emitted is relevant. This is not the case for CO2 emissions that generate global pollution.
How do you evaluate the policy’s effects?
I use a structural model to simulate the demand and supply of new automobiles in 2008 in the absence of a feebate, allowing me to predict the prices and car models that would have been purchased. My model represents consumers' preferences for the different car models available. I estimate how their choices relate to the various car attributes (horsepower, fuel costs, vehicle size, etc.) and the prices. By comparing individual car choices with and without the policy, I can identify the causal impact of the feebate. More specifically, I estimate the feebate's effect at the individual level and in monetary terms.
I use granular data on car sales at the local level to build a flexible model of heterogenous consumer preferences for car attributes and price sensitivities. Since I do not observe the direct links between individual purchases and demographic characteristics, I exploit the correlation between the average characteristics of cars purchased and the average demographic characteristics of the local municipality. In contrast to standard models, I identify winners and losers by linking gains or losses from the feebate to households’ demographic characteristics. On the supply side, I model the pricing strategies of car manufacturers and estimate the marginal costs of all cars proposed on the market.
I then measure the potential gains from alternative feebate schemes that achieve the same CO2 emissions reduction at the same budget cost as the 2008 feebate, identifying those that maximize consumer surplus, national manufacturers’ profits, and the reduction of local pollutants and diesel car share. I consider a variety of different weights for each municipality’s outcome in the global objective, depending on population size and income.
What are your main results?
Balancing consumer surplus and manufacturer profits with the budget cost for the government and the impact on emissions, the policy increases annual welfare by €115 to €119 million. The budget cost of €210 million necessitates an average tax of €23.9 for French households, reducing their overall surplus by €39 million. But their loss is more than offset by an increase of €158 million in the profits of French car manufacturers. Under a uniform tax, the feebate favors middle-income individuals. If the tax is proportional to income, the feebate achieves some redistribution from the richest to the poorest households.
Average CO2 emissions from car purchases decrease by 1.6% but annual new car emissions increase overall. In monetary terms, the change in annual CO2 emissions generates a loss of €0.49 million. This cost can turn into a benefit when considering that extra new car sales are old car replacements. Average emissions of NOX and PM increase, while those of carbon monoxide (CO) and hydrocarbon (HC) decrease. Overall, the annual cost of local pollutant emissions increases, reducing welfare by €0.015 to €1.31 million.
The policy may also impact local air quality by pushing the sale of diesel cars. I find that the feebate is responsible for an increase of 0.5% in average car emissions of NOX and 0.3% in PM emissions. These local pollutant emissions increase most in rich and dense municipalities, where they are initially the lowest, achieving another form of redistribution to poor and rural areas. However, richer municipalities are associated with larger decreases in the average emissions of CO and HC.
Could the policy be improved?
The implemented feebate achieves almost all of the maximum potential consumer surplus and French car manufacturer profits. Inequalities across individuals and French car manufacturers could be reduced, but this would come at the cost of global welfare losses.
Alternative feebate schemes could further limit NOX and PM emissions, especially using diesel-penalizing ones. However, such diesel-penalizing feebates would increase emissions of CO and HC and decrease consumer surplus, indicating trade-offs across pollutants and objectives. An optimal feebate scheme design would require specifying the regulator’s objective and the weights associated with each outcome.
I do not model the non-pecuniary side effects of the policy, such as making environmental friendliness a more salient car characteristic. In another paper (D’Haultfœuille et al., 2016), I provide evidence of a shift in preferences towards low-emission cars in 2008, but there is no direct proof that the feebate policy is responsible.