Closing Arguments in the Great Gatsby Curve Wonk Fight of 2012

Scott Winship
Scott Winship Director - Center on Opportunity and Social Mobility, AEI

January 20, 2012

Editor’s Note: This opinion piece first appeared on January 17, 2012 as a guest post on National Review Online’s domestic policy blog, The Agenda.

Several researchers have misinterpreted my point in criticizing Krueger’s mobility claims from last week. Here it is, stated as clearly as I can muster: Krueger was deceitful in his claim that the middle class is shrinking, and the Great Gatsby Curve is remarkably weak evidence that mobility is declining.

Let me start with the second objection to the Krueger claims, related to the now-famous Great Gatsby Curve. My assessment has been the subject of quite a bit of criticism from Miles Corak, my Brookings colleague Justin Wolfers and others. Miles and Justin basically accuse me of not knowing my stuff, making arguments that weaken my own casse, and trying to muddy up Krueger’s evidence indiscriminately to cast doubt on it. Harsh, Heather.

I’ve outlined my detailed objections to the evidence at length, but there is more to say.

To review, Krueger plotted inequality against immobility across ten countries, computed the best-fitting line through the points, and then used this Great Gatsby Curve to estimate that today’s children will have less mobility than their predecessors. I want to emphasize for the record that I am agnostic on the questions of whether economic inequality robustly correlates with economic imobility across nations and whether economic inequality causes economic immobility. What I did assert in my critique of Krueger’s figures is that for purposes of projecting whether mobility is on the decline or not, extrapolations using the Great Gatsby Curve are “uninformative”. The approach is an exceedingly weak one, which replaced the even-more-flawed approach from the president’s Osawatomie speech. The specific way that Krueger implemented the approach also has specific weaknesses (such as using inequality measured in adulthood rather than in childhood).

Justin made the colorful charge that my criticisms of the Gatsby Chart’s imprecision actually strengthen Krueger’s and Corak’s case. Wolfers apparently thinks I am arguing that all or some of the ten points in the Gatsby Chart are measured imprecisely, such that they are equally likely to be too high or too low, too far left or too far right. He also appears to think that I am arguing that the perfectly estimated best-fitting line would be flat, indicating no relationship between inequality and immobility.

If I were arguing both of these points simultaneously, I would indeed be inadvertently strengthening the other side’s case. This would be embarrassing. Since I am actually making neither argument, it is just frustrating, since most of Justin’s readers will assume he is right and will not end up reading this post. So to be clear, I am arguing that the Gatsby Curve is too imprecise for worthwhile estimates of the trend in American mobility. It may be too flat (if most of the error in the points is of the equally-likely-to-be-high-low-left-right variety). But it may be too steep. Many of the issues I have raised do not necessarily imply that the Gatsby Curve is too steep—Justin is just wrong about this.

Let me illustrate dramatically with two examples. The economic theory that Miles cites regarding why inequality might affect mobility is most logically applied using wealth inequality data rather than income inequality data. For five countries where there are comparable wealth Gini coefficients (see Table 3), the Gatsby Curve (using Miles’s immobility figures and the wealth Ginis from roughly 2000) really is flat. The correlation is 0.01, while the correlation using income Ginis from roughly 2000 is 0.64. (Ask me if you want the spreadsheet.)

For another illustration of how imprecise Gatsby Curves are for this sort of exercise, take a look at the last three rows of Jo Blanden’s Table 5. The rows provide 18 estimates of mobility-inequality correlations, using 6 different sets of mobility estimates and 3 different income Ginis. If Miles and Justin are right, these 18 estimates should be similar. But they instead range from -0.15 to 0.87. That is to say, the Gatsby Curves are all over the place. (And, yes, I realize that the Ginis are measured at different times, but as I noted, Krueger is using Ginis for the wrong time (adulthood instead of childhood). Miles and others have not found this objection compelling up to this point.)

If the Administration wants to say, “There is a relationship across countries between inequality and mobility, so that would lead us to expect that with rising inequality there will be less mobility,” that’s less objectionable, in a sense, than trying to nail down a point using the Great Gatsby Curve, giving the illusion of precision. However, the claim would still be weak, for a number of the reasons I have laid out. Better yet would be to just say, we have too little mobility and to support it with the comparatively rock-solid evidence on levels of American mobility. That is, one does not have to tie insufficient mobility to inequality, and one does not have to show that mobility is declining to argue that it is insufficiently high.

Noah Smith is the only person who has attempted to defend Krueger’s claim that the middle class has shrunk, which obscured the point that it has shrunk by his definition because more people are now richer than middle class. Noah claims that this claim isn’t deceptive at all, that it is narrowly true and illustrates a point about rising inequality. Brad DeLong gleefully approves.

Give Noah the point that in a narrow sense, the middle class as Krueger defines it has shrunk. Was Krueger being deceitful? Summarizing the chart that showed a declining middle class, he said, “we have more families falling into either extreme end of the distribution, and fewer in the middle.” No—we have more families falling into one extreme end, and it is not the low end, it is the high end. For that matter, the upward mobility estimates from the Osawatomie speeech used the same constricted definition of middle class: upward mobility from poverty to richer-than-middle class did not count as upward mobility in those estimates. There is absolutely no reason to be substantively concerned about whether the poor have less oportunity to make it to the middle class but no further.

I hope readers have found this debate to be informative, and I am grateful to Reihan for letting me crash here for a few days. I have spent too much time on it this week to continue, barring some response that I feel compelled to defend myself against. I share Justin’s hopes that the Gatsby Curve will spark a lot of additional research on this question—I’ll certainly be pursuing some of it.