Sunday, April 15, 2007

Bob Frank replies

A few days ago, I expressed here my skeptical view of a recent column by Bob Frank. I offered Bob an opportunity to respond. Below I am reprinting, in its entirety, what he sends along. There is much that one could debate here, and I am sure the commenters will, but I will refrain. Since I picked this fight, and since I have ample opportunity in this forum to express my perspective, in fairness I will let Bob have the last word--at least for now.

First, my thanks to Greg for his gracious invitation to respond to his comments on my recent New York Times Economic Scene column. In that column, I argued against trickle-down theory’s claim that higher taxes on top earners would reduce economic growth. Here I’ll attempt to explain why Greg’s defense of trickle-down theory falls short.

Greg discounts the significance of the negative relationship I cite between wage growth and the average workweek over the last century. This relationship, he argues, is a consequence of the fact that the income effect of rising wages has offset the substitution effect (which is exactly how I described it). But the observed link is irrelevant, he explains, “because the distortionary effect of taxes depends only on the substitution effect. The evidence cited suggests that income effects are larger than substitution effects, not that substitution effects are small.”

Greg is right about what this particular piece of evidence shows. But he is mistaken in claiming that this evidence is irrelevant to my claim. Indeed, the argument I advanced in my column had nothing to do with whether taxes on the rich are distortionary. That’s an interesting question, and I’ll return to it in a moment. My only point in the column, however, was to question a very different claim—namely, that higher taxes on the rich would reduce work effort. That is precisely a claim about the total effect on work effort of lower after-tax wages. In other words, it’s a claim about the combined impact of the income and substitution effects. So the fact that the workweek declined over the last century in the face of substantial growth in real wages is directly supportive of my argument.

A necessary and sufficient condition for trickle-down theory’s argument to the contrary is that the elasticity of supply of labor with respect to real wages be significantly positive. The most comprehensive recent econometric study of labor supply elasticity in the United States will be published in the next issue of The Journal of Labor Economics. The authors, Fran Blau and Larry Kahn, estimate that the labor supply curve for men has been essentially vertical for many decades. The clear implication is that higher taxes on top earners, most of whom are men, will not significantly reduce work effort.

Greg also mentions research suggesting that higher taxes on the rich may reduce the amount of income they report to the IRS. Perhaps so, but that by itself would not imply any reduction in output. And with even the supply-sider Bruce Bartlett now conceding that tax cuts for top earners don’t boost total tax revenues, it’s important not to exaggerate the problem of unreported income. But irrespective of its magnitude, why isn’t the best solution to this problem a simpler and more strictly enforced tax code rather than tax rates that are too low to sustain minimally adequate public services?

Greg and I both have good health insurance, but millions of others have none. Absent the revenue necessary to fund universal health coverage, their ranks will keep growing as our current system of private, employer-provided insurance continues to unravel. President Bush’s recent proposal to make individually purchased health insurance tax deductible is unlikely to help. (As Stephen Colbert put it, “It’s so simple. Most people who can’t afford health insurance also are too poor to owe taxes. But if you give them a deduction from the taxes they don’t owe, they can use the money they’re not getting back from what they haven’t given to buy the health care they can’t afford.”) Without higher taxes on people like Greg and me, this problem will get worse.

Revenue shortfalls have also led to cuts in other important public services. I cannot imagine, for example, that Greg feels any more comfortable than I do about the Bush administration’s cuts in the Energy Department’s program for helping lock down loosely guarded nuclear materials in the former Soviet Union. But such programs cost money. Unless we can raise additional revenue to pay for this program, or unless we want to borrow even more heavily from abroad (loans that will eventually have to be repaid in full, with interest), it will remain underfunded. It is not satisfactory to assert that we can just reduce government waste. The president, who campaigned as an opponent of government waste, is the one who couldn’t find more wasteful or less politically protected programs to cut.

As evidence for his claim that I need to do additional reading, Greg cites a 1988 paper in which Joe Stiglitz argued that the socially optimal marginal tax rate on the most productive person might actually be negative. The reason, Stiglitz explained, is that inducing that person to work more could generate positive spillover effects for less productive workers. In the abstract, this is an interesting claim. (Is it any more than that? Stiglitz, for one, never thought to offer tax policy proposals on the basis of it.) But if we’re going to discuss externalities, then complementarities between skilled and unskilled labor are surely not the most important ones to consider.

For present purposes, by far the most important externalities are those stemming from the link between context and evaluation. As decades of behavioral evidence clearly demonstrates, virtually every evaluation is heavily shaped by local context. As Richard Layard put it, “In a poor society a man proves to his wife that he loves her by giving her a rose, but in a rich society he must give a dozen roses.” Because evaluation drives consumer choice, context is an important determinant of consumer demand. The upshot is that almost every consumer choice generates significant context externalities.
Consider, for example, a job applicant’s decision about how much to spend on an interview suit. His goal is to make a favorable impression. But his ability to do so depends far less on the absolute quality of his suit than on how it compares with those worn by other applicants. And when he spends more on a suit, he shifts the context within which other candidates will be evaluated.

Context externalities are pervasive. A good school, for instance, is one that compares favorably with other schools in the same local environment. The amount parents must spend to ensure that their children attend such a school is thus an increasing function of the amounts spent by other parents. The evaluations that guide an employer’s promotion decisions are similarly dependent on context. A worker’s odds of promotion depend less on his absolute performance than on how well he performs relative to his coworkers.

The dependence of evaluation on context lays waste to any presumption that individual decisions about how many hours to work or how much to spend on interview suits will be socially optimal. The general result predicted by theory is that if context shapes evaluation more heavily in some domains than others, too many resources will flow to the most context-sensitive domains and too few to the least context-sensitive domains. In my forthcoming book, Falling Behind, I summarize what available evidence says about the extent to which context differs across domains. For the discussion at hand, the relevant finding is that evaluations of leisure tend to be far less context-sensitive than evaluations of income. The implication is that individual valuations of leisure tend to understate social valuations. Thus people work longer hours in the hope of moving higher on the income ladder, only to discover that when others do likewise, their position remains unchanged.

It would be unfair to single out Greg for ignoring context externalities. After all, most of the standard economic models that serve as the basis for policy analysis make no mention of these externalities. At some point, the economics profession will look back in embarrassment about that fact. But even absent explicit mentions of context externalities, most practical policy analysts already seem to recognize that trickle-down theory’s portrait bears little relation to the behavior of people in the real world.

My point is not that people don’t care about money. On the contrary, when the pay in one occupation goes down relative to others, fewer people enter that occupation. Public school teachers, whose starting salaries were more than 20 percent higher than those of the average college graduate in the 1960s, now earn below-average starting salaries. So we are not surprised that fewer qualified people now enter teaching.

But trickle-down theory is about what happens when after-tax pay falls not just in some occupations but for top earners generally. In a largely meritocratic society like the United States, most top earners are extremely driven people. And as recent studies have shown, most of them will never spend more than a small fraction of their earnings. The trickle-down theorist’s insistence that they will begin slacking off in response to a small increase in their marginal tax rates strains credulity.

While serving as chairman of the Council of Economic Advisers, Greg actively supported the Bush tax cuts targeted at top earners by arguing that the cuts would spur them to work harder. Greg would have been astonished to observe such a response from his colleagues at Harvard. Does he have a behavioral model that leads him to expect different behavior from high achievers in other occupations? Or does he have one that explains why any such differences consistently fail to reveal themselves in the data? In the absence of a plausible behavioral model backed by persuasive empirical evidence to the contrary, I stand by my conclusion that trickle-down theory is supported neither by economic theory nor by empirical evidence.

The tax cuts that were sold by invoking this theory did little to promote the well-being of even the well-to-do Americans who were their ostensible beneficiaries. Money that could have been spent rounding up loose nuclear materials in the former Soviet union was spent instead on larger houses and more expensive cars. In light of what we know about the empirical magnitude of context externalities, the principal effect of such spending was simply to redefine what counts as adequate. As in the familiar stadium metaphor, all stand to get a better view, yet none sees better than if all had remained seated.

Greg titled his response to my column “Frank Needs To Read More Widely.” On that point, he is surely right. I don’t know Greg well enough to presume to know what he needs. But he would almost surely offer better policy advice if equipped with an economic model that better fits current scientific knowledge about human behavior.

Again, my thanks to Greg for inviting me to respond to his critique of my column.