A small hornet's nest was stirred when Greg Mankiw posited in the WSJ that it was time for a gas tax of about $1 a gallon, citing seven reasons for it.

1. It would be a way to internalize the externality of carbon emissions;
2. It would be a way to internalize the externality of road congestion;
3. It would serve as a substitute for regulations that have perverse consequences, like fuel efficiency standards and the SUVs they spawned;
4. It would help balance the budget;
5. It would be borne at least in part by the Saudis, the Venezuelans and other oil producers, whose incomes we do not particularly want to expand;
6. It would be pro-growth to the extent that consumption taxes are better than income taxes in their drag on economic activity;
7. It is a national security issue.

I am less persuaded that Professor Mankiw, in particular by the plan for phasing in the tax in ten cent increments over ten years. Some of it is figuring out what the right size of the tax is; John Palmer has done a nice job laying out those arguments.

But how does he arrive at this precise amount for the tax? The simple answer is we don't know for sure. We have to guess. One would hope the guess is well-informed and documented by people who know what they are doing. And this is the heart of the criticism of Mankiw's Pigou club: it is easy to draw these things on the chalkboard, but measuring and identifying the externalities (not to mention the general equilibrium effects) precisely is probably not possible with today's knowledge and technology.

Let's just say it's tricky.

We know that the short-run elasticity of fuel use is quite small; probably 0.25 is a median estimate. Let's guess the long-run elasticity is about 0.75. (I've not looked at every study, but the ones I've seen would have those as the midpoints of the ranges offered.)

Key to this as well is the elasticity of supply -- how responsive are suppliers to a change in the price they receive for gas brought to market? See Professor James Hamilton's elucidation. In order for points 5 and 7 to be true, you have to assume supply is highly inelastic. In the short-run, perhaps this is true. But how long does it take for refiners to stop producing gasoline out of crude and instead produce something that is not taxed? Given the rapidity of adjustment between gasoline and diesel I observe driving by gas stations, I'd say not long at all. Crude oil is still going to flow to alternative uses.

Thus the burden of the gas tax in the short run is borne by both an inelastic demander and an inelastic supplier, but if supply can adjust quicker than demand, it will be the American consumer who bears the brunt of it. In the short run, there isn't much change in the quantity of gas sold -- it will swell the government's coffers, but if that was your plan along with getting more growth in tax policy, would you argue for Fair Tax or a VAT over the income tax rather than using an ad valorem tax on gasoline?

Last, the phasing part of this assumes, I think, that you can borrow the long-run demand elasticity to encourage the reduction of gasoline to come faster. If we know prices will rise a dollar in the long run, wouldn't we start finding ways to conserve now? Yes, if and only if you expect the phase in to continue for ten years. How many people will believe that government can commit credibly to a ten-year tax plan? The caterwauling over the estate tax should be instructive: a tax rise or fall expected some years from now must be discounted by the rational investor or household for the possibility that the government will renege.

I agree with Palmer -- Pigou taxes are a chalkboard exercise; most of tax policy is blunt force applied to shove things in one direction or the other at the desire of those who have power and think they know better. The fact that you would phase in the increase, however, indicates to me some uncertainty about whether or not the tax is the right amount. As a matter of policy advising, I follow the Hippocratic Oath of economics: First do no harm to market signals.

UPDATE: Aplia Blog has a nice analysis of Prop 87 in CA, which taxes oil production there to pay for alternative fuel research and development. It's not Pigovian; the blog asks "Would a Pigovian tax on gasoline consumption be a better way to fund research on and development of alternative fuels?" You'd first need to answer the question, What are the social costs of not using alt fuels?