In Mankiw's latest post which indoctrinates Dr. Robert Samuelson in Mankiw's gas tax cult, he exaults:
"In his latest column, Robert Samuelson says we should "enact an energy tax equivalent to $2 a gallon on gasoline -- introduced over six years, or about 33 cents annually," making him the latest member of the Pigou Club. I have proposed a more modest $1 increase, in part based on the research of Parry and Small, but of course there is a degree of uncertainty about how high the optimal tax is." [bold mine]
It is refreshing to see that Mankiw has finally acknowledged one of the many complaints regarding a gas tax increase proposal (for a slightly larger list, see my previous post titled: "Cafe or gas tax hike" or any of my posts on Mankiw's blog). Though the use of language "degree of uncertaintly" is laughable and a gross understatment.
Parry and Small use model calibration / simulation technique - which basically draws upon a whole host of varying assumptions to provide the answer that they want to get - namely that a $1 gas tax increase is "optimal."
The biggest assumption of the model is the elasticity of gasoline demand. PandS assume an average price elasticity (in absolute value) to be .55 with .30 being a LOWER bound (which is astronomical - see a previous post where I show research that puts the value close to .04. This obvious inflation biases their result upwards. But of course, being a calibration model, pretty much everything else is based on assumption as well (often using just one article from which to base the assumption on).
All of the following is assumed by Parry and Small: fuel efficiency, pollution damage, congestion cost, accident cost, labor supply elasticity, government spending, PPI of gas, initial tax on gas... the list goes on.
Can "congestion cost" really be quantified so easily? I'm highly skeptical. Also, PandS say:
"the congestion externality is the largest component of the optimal fuel tax. " But congestion reduction does not directly follow from a gas tax hike. In fact, their assumption of such large congestion reduction I think is too high given that most of the effect would be to substitute to a more fuel efficient vehicle as opposed to driving less. Also, the congestion effect is variable depending on where you live. Where I live, in Indianapolis, there would be likely little effect on congestion since there are few transportation alternatives.
Perhaps the biggest problem with this is that some papers I have read say $4 is needed, some say less than $1. Further, researches tend to ignore any of the directly negative consequences of large gas tax hikes (regressivity, gorge the beast...). The point is it is all dependent on your assumptions.
For my opinion on calibration technique, see this excellent paper on the subject entitled "Calibrating the World and the World of Calibration" by Mattias Polborn.
This strongly relates to my recent post regarding post-autistic economics and the strange need for Economists to find "optimal" answers using mathematics based purely on assumption.