Would a gas tax be regressive?

Jim over at OIFS is doing his usual good job of making people uncomfortable--in a good way. As the call for an increased gas tax builds to a loud whisper, Jim notes that opponents are quick to play the regressive card. Put simply, a regressive tax takes a larger percentage of income from the poorer than from the richer. A famous example would be a tax on food. Since poorer households spend a larger share of their income on food, a flat tax would necessarily affect poorer households more by the same measure. This is the reason most state sales taxes exempt basic food purchases. The opposite of a regressive tax is a progressive tax, which taxes something richer households spend a larger share of their income on, like a tax on second homes.

Jim's argument is that a lot of things stick it to the poor. If we avoided all of them, we wouldn't get much done. He offers a neat example that 25-35% of Interstate highway construction expenses are mitigating the costs to third-parties. Why not ask what it would take to mitigate the costs of a $1 gas tax to the poor and then get on with the discussion? Well, let's do it. I looked into the 2004 BLS Consumer Expenditure Survey (CE) to get an idea of the costs to the poor. I'll define "poor" arbitrarily as the lowest income quintile (roughly less than $20,000 pre-tax annual income). That should catch most poor and near poor households.

The first thing I found is the same thing Poterba (1991, paywall) and others have noticed--unlike a tax on food or housing, it isn't at all clear whether a gasoline tax would be regressive. Start with the fact that a full 1/3 of households in the lowest income quintile don't even own a car. Effects on those households would be only indirect via rising prices of other goods (food, mass transit). Then continue up the income ladder to find that the second and third quintiles actually spend a larger share of income on gasoloine than the poor. The two highest income quintiles do spend a smaller share on gas, but the dropoff is much smaller and more gradual than for food or housing. (BLS 2004)

Still, one could argue that even if a gas tax doesn't necessarily take a larger share of the poor's income, the share it does take may be more important (the eat or drive argument). Well, that's a separate debate, but if you agree that the burden is unfairly placed on the poor, then push for a gas tax with mitigation! What would it cost?

Again basing my numbers on the 2004 CE, the average poor household purchases about 400 gallons of gas in a year (1), about half the amount of the average household. Given an estimated 100 million total households currently (Census Current Population Report 1996), a gas tax that increased pump prices by $1 per gallon (2) would impose a direct burden on the poor equal to $400 (400 gal*$1/gal) per household. Since about 20 million households meet my definition of poor, complete mitigation would cost $8 billion.

$8 billion per year sounds like a whopping lot of money, but remember what those Interstate mitigation costs were? A gas tax of this size would raise about $125 billion per year (3). Compensating the poor would represent only 6.4% of the total revenue generated.

What about the effects on the other four quintiles? Well, I note that well over half of those households are college educated, and they spend more than the amount of the tax on "television, radios, and sound equipment" (BLS 2004). I'll bet they can figure something out.

(1) $730 spent on gas [BLS 2004]/$1.81 average 2004 price of regular [EIA 2006]
(2) This would represent a total tax of around $1.56/gal using elasticity estimates from Greene and Ahmad (2005) cited in Econbrowser (2006). This is because part of a tax would be absorbed by gas suppliers.
(3) Given an estimated long-run demand elasticity of -0.335, the tax would reduce gas consumption by about 9.5%, reducing average household consumption to about 800 gallons per year. 800 gallons * $1.56 * 100 million households = about $125 billion in total revenue.

Correction: I got so immersed in the household data I sort of forgot it was only part of the picture. Non-household purchases of taxable gasoline might add another 30-40 billion gallons to the quantity side (EIA 2006). Of course, that just makes the mitigation costs to the lowest income look even cheaper!