Fun with statistics: the regressivity of gasoline taxes
Posted by Rich Sweeney on April 17, 2008
Yesterday on Grist, Joseph Romm responded to a reader who argued that he should care more about incidence when talking about gas taxes. His response was a litany of citations suggesting that gas taxes were not in fact regressive, partially because the poor don’t own cars/ drive as much. Romm’s supporting evidence is pretty compelling. He cites a recent study by Don Fullerton (who used to be at RFF) which finds that a gas tax would only be regressive across the top half of the income distribution (BTW, Fullerton’s paper is one of the best I’ve seen on the public finance options for curbing vehicle emissions if you’re interested in that issue). Yet despite all of the evidence Romm put forth, I was still skeptical of his claim. I’m working on an incidence paper myself right now, and do see regressivity in the data. It just depends on how you define it. Below is a table with gas expenditure by income quintile from the 2007 BLS Consumer Expenditure Survey (which, btw, is the same source Fullerton uses):
Notice how different the story looks depending on whether you put income or consumption in the denominator. Both measures have their faults, and, in the policy world, people tend to flip back and forth between them depending what point is being made. Ideally economists would like to put lifetime income in the denominator, assuming that people incorporate their future welfare/ wealth into their current decision making (ie college student go into debt with the expectation that they’ll make much more money later on). For a more detailed discussion on defining incidence see this working paper by Gilbert Metcalf.
For those who don’t have time to check out the Fullerton paper, I’d also just like to point out something about the table Romm presented in his post.
However, he omitted the third column. To be fair, his point was about average effects across all people. Yet one can imagine how the second statistic might be relevant to policymakers. While poorer people are less likely to own cars/ drive, the ones that do are likely to be significantly impacted by an increase in the price of gasoline. If a household making under $20,000 a year does own a car, it’s probably because they live in an area where there aren’t any other options for getting to school, work, etc. This means that their demand response in the face of a price increase would probably be negligible, and a gas tax would simply crowd out other consumption. Thus you could imagine framing incidence in terms of rural vs. urban households, as well as simply rich vs. poor.
None of this is to say that one way is best. And I definitely support a gas tax, although I think we should be smart about implementing it. I just thought I’d point out that on this issue, like just about every other issue in DC, different people can use the same data to make different points.