Incomplete and Partial Thoughts on Greg Mankiw's Updated "Lazear vs Krugman"

Greg Mankiw quotes my claim that the Piketty-Saez data are hard to interpret as the result of a general rise in the economy's skill and education premium driven by skill-favoring technological change:

Greg Mankiw's Blog: Lazear vs Krugman: Update 2: Brad DeLong tries to explain what Paul might have been thinking:

DeLong: The big rise in inequality in the U.S. since 1980 has been overwhelmingly concentrated among the top 1% of income earners.... It's hard to attribute this pattern to a rise in the premium salary earned by the well-educated by virtue of the skills their formal education taught them. Such a rise in the education premium would produce a much smoother rise in relative incomes among the whole top tenth of the income distribution...

And he comments:

I am not convinced that the conclusion follows from the facts presented. I would guess that the top 1 percent of income earners (those [household tax units] earning more than $276,945) are disproportionately very well educated--doctors, lawyers, MBAs, etc. So the rise in the income of the top 1 percent could well represent in large part a higher education premium.

What might well be true is that the returns to education have become increasingly non-linear: The most educated are now getting a bigger return from a marginal year of education than those with moderate amounts of education. In other word, two years getting an MBA from Harvard Business School may increase a person's income more in percentage terms than does two years getting an Associate Degree from Mass Bay Community College. My understanding from my labor economist friends is that some evidence favors this hypothesis of increasing nonlinearity...

I don't think this works particularly well. Yes, if you confront a computer with a strongly nonlinear increase in inequality and ask it to explain it by increases in skills and the values of skills of those at the top, it will spit back that there is evidence of nonlinearity. But so much? The top 0.1% in the United States has gone from 2.3% of income in 1980--23 times average--to 7.6% today--76 times average. The next 0.9% has gone from 6.3 times average to 9.2 times average. And the next 4% has gone from 3.2 to 3.7 times average. Just what have been the changes in technology over the past twenty-five years that have made the skills of the 130,000 households in the top 0.1% so much more highly-valued vis-a-vis the skills of Mr. and Ms. 95th percentile? We are awarding 550,000 advanced degrees a year in this country. The overwhelming majority of them must be gaining little or nothing in relative-income terms vis-a-vis their predecessors of 1980--and those 15,000 a year or so who will someday join the top 1% have seen their relative incomes triple. Continuity: just what is it that the top 13,000 have learned that the other 537,000 have not that is so valuable?

Mankiw also comments, I think more promisingly:

To some extent, the returns to human capital are random (as is true of physical capital). Getting an MBA gives you a shot at being CEO, but it is not a guarantee. This may be part of the Lemieux finding that higher levels of education are associated with higher residual variance. And perhaps it can reconcile the differing perspectives of Krugman and Lazear...

The implicit model, I think, is that when you get an advanced degree--or perhaps when you get an advanced degree from a good school--you not only get skills, but you also get a lottery ticket. Either because of dumb luck or because of the interaction of talent with formal education and technology or because of the interaction of the willingness to work like a dog beyond all reasonable measure with formal education and technology, the lucky or talented or workaholic today can, thanks to revolutions in computer and communications technology, leverage their symbolic-analyst skills over a much larger base of routine manufacturing, marketing, and distribution workers than they could have a generation ago. In this model, we have become much more of a "winner take all" economy than we used to be. Much more income is distributed in the form of winner-take-all tournaments than used to be the case.

My first reaction is that this is possible, but unproven. My second reaction depends on whether victory in the winner-take-all tournaments is due to luck, talent, or industriousness. If it is luck or talent, the 60% of me that is a social democrat thinks that this is grossly unfair, and that we should think very seriously about powerful public policies that will level the distribution of income.

Then the 20% of me that is a libertarian breaks its chains, comes running out of the cave, hits the social democrat on the head with a brick, grabs the microphone and rants: "Do you really trust the American government to manage a major downward redistribution of wealth without doing immense harm? And who would you rather have deciding how Bill Gates's $100 billion and Warren Buffett's $40 billion will be spent--Karl Rove or Bill Gates and the staff of the Bill and Melinda Gates Foundation?"

Now let's return the libertarian to its cave. There, there, everything will be all right. See? Just inside? Catalaxy. Zero government. Private order. A free society of associated autonomous producers...

Where was I?

Oh yes, in the event that it is neither luck nor talent, but instead industriousness, things become harder to evaluate. Organizations adopt winner-take-all compensation policies because they see a benefit, and the benefit is that they get a lot more people than there are big prizes to work very hard. These non-tournament winners are probably irrational--overestimate their chances of winning the tournament--and are thus exploited by the rest of society: putting in MorganChase hours and doing MorganChase-quality work without getting MorganChase pay scales. Jim Mirrlees has argued that such a society--in which those with a high marginal product were forced to work like dogs and had a relatively low utilitarian happiness--could well be optimal on utilitarian grounds.

We should think, and think hard, about all these issues. But I don't think that it's useful to characterize this mechanism for increasing inequality as "a rise in the premium paid to the skills acquired through education." I'm not sure what to call it, but it is something very different.

---

UPDATE: And Paul Krugman emails:  

First, the question of which dates to look at depends on the question you're trying to answer. If you're asking why the public doesn't feel good about the economic growth since 2003, pointing out that inequality fell between 2000 and 2003 is irrelevant; everyone felt lousy about the economy during those years. The point is to explain why most people don't feel better about performance since 2003 - and rising inequality since then is the explanation.   Second, the data aren't encouraging about the long-term trend. The slump in top income shares after 2000 gave us reason to hope that the extreme income concentration at the end of the 90s was an artifact of the bubble, and would not return. But 2004 data already show a return almost to 2000 levels of inequality, and other indicators suggest that the trend has continued since then. Gilded Age II, here we come.    Third, both Greg and Eddie Lazear have asserted not just that rising inequality is partly due to an increased skill premium, which is true, but that it's mainly due to [increases in the amount of] skill[s possessed by those relatively high up in the income distribution], which is false. I don't see why this distinction is so hard to understand. The median income of college grads is up since 1980, but only modestly; the big gains are for people at the 99th percentile and beyond.