I've long been "meaning" to read and critique more Larry Summers papers...
Finally, I was assigned to read one in a course I'm auditing, entitled "Tax Incidence" with Larry Kotlikoff.
On some level, the paper is basically fine. However, it falls into the category of what I like to call "pre-Modern" Economics, as the premise of the paper is that "we" know nothing about human behavior except that humans are uber-rational machines. It is a paper chalk full of obviously erroneous assumptions such as "all agents are the same" and "Production exhibits Constant Returns to scale" while maintaining innocence that the conclusions are a product of the assumptions...
Take the first (seemingly harmless) insight that tax incidence need bear no correlation with who bears the legal incidence of a tax. And that who bears the legal incidence will make no difference to the equilibrium price or quantity. It's a fascinating insight! Except it's not true in the real world case.
Theoretically, fine. But as Larry Summers himself would be wont to say "There are idiots. Look around..." If the government taxes each carton of milk, and charges the company directly, so that a consumer sees only the post-tax price when they walk into the grocery store, there will be a different equilibrium outcome than if the tax is not levied until the consumer gets to the cash register... Should there be? No. But, people are, well, people. $1.99 with 5% tax later looks like a better deal than $2.09. The milk-producing folks might well be better off with the tax incidence falling on the consumer...
But, whatever, you say, you can't expect someone to theorize about everything at once. Fine. The next section in the paper deals with factor taxes. Summers & Kotlikoff just consider the homogenous-worker tax incidence case. Some things, such as "a tax on capital doesn't affect the real wage in the short run" are obviously conclusions that will not extend to the real world case (although I suspect, in this case, the authors realize it).
More strikingly though, the basic conclusion that, when taxing factors, the more elastically demanded and inelastically supplied factor will bear the burden of the tax, does not hold if we start relaxing some of our assumptions (which nobody believes anyway). Let's assume everyone, and every job is not the same. Let's look at NFL teams -- clearly the quarterback position is the most important on the team. The Colts demand for Peyton Manning is clearly more inelastic than Peyton Manning's supply of labor to the Colts. If the Colts unilaterally announced they would cut his salary to $5 million, Peyton could threaten to sign w/ Tennessee and could easily get way more (in which case, how many seconds would it take the Colts to come crawling back?). On the other hand, for lineman and/or wide receivers, especially the second-stringers, its quite plausible that the Colts demand for these guys is quite elastic (second-stringers are easy to replace), while at the same time, the Colts have an inelastic demand for some second string lineman.
Summers' theory suggests that, when we tax Peyton Manning, the incidence should not primarily fall on Peyton Manning. But if Peyton were taxed by declaring all income over $5 million taxed at a 99% rate, Peyton would likely just settle for $5 million since it's not the Colts organization which is insulting him. Now, with more room under the salary cap, the lineman, receivers, and second stringers would be getting more. The incidence falls completely on Peyton Manning, even though he's the guy, according to Summers' analysis, who can't be touched by taxes.
Am I cheating here by relaxing so many assumptions? And, isn't the NFL a special case? Well, in a sense yes, and in a sense no. Summers' didn't mention that there isn't just one labor supply curve, that if the Colts offer Peyton less or more he'll react differently than if the government taxes Peyton less or more (even if it's the city of Indianapolis which taxes him to pay for inner-city education). Or that Peyton Manning will have a different supply curve if the Colts offer him $5 million, but promise to sign a pro-bowl lineman with the proceeds. And Peyton's labor supply curve for playing for the Colts is different than his supply for playing football at all.
In the last section, he considers an OLG model, which implies that a higher tax on capital reduces the capital stock and the real wage.
How might we tweak this model? Instead of assuming everyone is the same, we could merely assume two types of Agents. The first type is the Walton family, and the second type are workers who work at Wal-mart. The ultra-rich tend to do the lions share of saving, so let's assume the workers all have tight budget constraints, while the Walton family all goes to grad school until they are 40 and then soothe their soles working on humanitarian projects in between ski trips to Aspen. In other words, let's assume the Waltons are rich enough they are at or near a satiation point.
What happens now when we increase the tax on savings? Well, nothing really changes, especially if the taxes get refunded, lump-sum. The workers weren't saving anyway, and the Waltons are basically satiated enough to the point where they are just giving money away... Now, probably the real world is somewhere in between my story and their story, but, nevertheless, the conclusion of Summers and Kotlikoff "Thus, an interest income tax compensated in the second period unambiguously reduces capital intensity. This means that the pre-tax return to capital rises and the wage falls. The tax is thus at least partially shifted to labor." only follows because of the authors assumptions, assumptions which do not happen to hold in the real world case...
Of course, there are dozens of ways we could relax the assumptions and reverse tweedle-dee and tweedle-dum's results. We could assume, for the poor group, that their consumption largely consists of education, food, and health care expenditures, and that productivity in the second period depends on consumption in the first. In this case, the larger the tax on capital, the higher productivity is, and the higher the steady-state capital stock. But, of course, if the capital stock mostly consists of the Waltons' vacation homes, this might not even be a good thing!
19 minutes ago