Manhattan Institute for Policy Research.
Subscribe   Subscribe   MI on Facebook Find us on Twitter Find us on Instagram      



The Great Gastby Curve Revisited: Part 1

December 30, 2013

By Scott Winship

Does More Inequality Correspond with Less Economic Mobility Across Local Job Markets?

In policy debates, as in life generally, it is important to be able to admit when you are wrong. We have previously maintained that the evidence that rising inequality has hurt economic mobility is weak. Recent research on the relationship between the two, however, has caused us to reassess that conclusion. There is no shame in reversing positions in the face of new, more persuasive evidence, and we have come to believe that new findings related to the Great Gatsby Curve compel such a reversal.

The Great Gatsby Curve, you may recall, was a chart revealing a strong statistical correlation across countries between household income inequality and men’s earnings mobility. It was developed by Canadian economist Miles Corak and popularized in a speech at the Center for American Progress by economist Alan Krueger while he was the chair of President Obama’s Council of Economic Advisers. Liberals have pointed to the chart throughout the past two years as compelling evidence that more inequality leads to less mobility. Paul Krugman, Krueger’s Princeton colleague, called it “very illuminating—and disturbing.” We, on the other hand, have argued that the Great Gatsby Curve constituted weak evidence that inequality hurts mobility. Essentially, rather than asserting that there is no such relationship, we have maintained that the Great Gatsby Curve is not particularly illuminating on that question and therefore not especially disturbing.

But with the arrival of a remarkable new dataset on mobility in the United States, it is possible to consider whether the same relationship between inequality and mobility shows up across local areas in America. Harvard economist Raj Chetty and his colleagues obtained access to tax records that allowed them to compute the first sub-national intergenerational mobility estimates. They also provided data on a range of potential correlates of mobility for each of the 741 “commuting zones” in the study. CZs are either metropolitan areas or county groupings in rural parts of the country—think of them as local job markets.

In many ways, looking across local job markets at the relationship between inequality and mobility is more informative than looking across countries, as Corak did. Countries differ strongly not only in terms of mobility and inequality but in terms of population size and composition, institutional features of the economy and government, politics and policy, and culture. Of course, there are also strong regional differences within the United States, but to the extent that other factors are driving the relationship between inequality and mobility in the cross-national data, those factors may contaminate the relationship across American CZs less. Of course, cross-CZ differences might also be driving any correlation between inequality and mobility in the Chetty data.

We have now had a chance to analyze the Chetty figures. And the results couldn’t be more convincing.

Below is the cross-CZ Great Gatsby Curve in all its glory. Each of the CZs is represented as a dot, with inequality increasing from left to right, and immobility (not mobility) rising as one moves from the bottom to the top of the chart. The best-fitting line through the dots is also shown—that’s the “curve” in Great Gatsby Curve. The fact that it is upward-sloping means that more inequality leads to more immobility (less mobility). The mobility measure here is Chetty’s “relative” measure and gives the difference in the expected adult rankings of the poorest and richest children. In other words, metropolitan New York’s score of 32 indicates that the richest children had average adult incomes that put them 32 percentiles ahead of the average for the poorest children. The range goes from 9 percentiles in Alpine, Texas to 49 percentiles in Yazoo City, Mississippi.

Wait a second, that’s not right—the chart shows how each CZ’s prevalence of single-motherhood among families with children compares with its relative mobility. It reveals a pretty strong relationship— a 20-percentage-point increase in the share of families with kids headed by a single mother predicts a 15-percentile increase in the gap in average adult income between the poorest and richest child. For that matter, the correlation is 0.61, which means that the variation (the “standard deviation”) across CZs if we simply predicted mobility from that best-fitting line and the CZ single-motherhood figures would be 61 percent as large as the actual variation in mobility. In other words, a CZ’s prevalence of single motherhood predicts its relative mobility quite well all by itself.

Anyway, we were talking about how inequality causes immobility, so pay no attention to that chart.

There are actually two different inequality measures in the Chetty data. One is the gap between the incomes of the household at the 25th percentile of the CZ (richer than just 25 percent of households) and the household at the 75th percentile (poorer than just 25 percent of households). You can think of this measure as indicating how wide the gap is between the richest and poorest middle-class households; it ranges from $21,400 in Eagle Pass, Texas to $91,500 in San Jose, California (Silicon Valley). Here’s the Great Gatsby Curve comparing relative mobility against “across-the-middle” inequality:

Actually, that’s not that striking, is it? A 50 percent increase in the gap between the 25th and 75th percentiles of parent income predicts just a 3.7-percentile increase in the gap in average adult income between the poorest and richest child. So if, for example, that across-the-middle gap in the parent generation grew from $50,000 to $75,000, the predicted adult income gap between the richest and poorest children would rise from below 32 percentiles to above 35 percentiles. The correlation here is just 0.25, so by itself, knowing the across-the-middle parental gap in a CZ doesn’t tell us nearly as much about the CZ’s likely mobility as knowing its level of single motherhood. Imagine how you would draw the best-fitting line through those dots if the red line wasn’t there.

Let’s try the other inequality measure. That one looks at the share of the CZ’s total parental income that went to parents whose income put them in the top one percent of households nationally. This is the kind of inequality that is rising in America; inequality below the 75th percentile has hardly increased since the 1980s. So if rising inequality has hurt mobility, this is the inequality measure that we’d expect to be important. This “we-are-the-99%” inequality ranges from 0.5 percent in Amery, Wisconsin to 52 percent in Friday Harbor, Washington.

What the what? The line’s sloping down—so more inequality leads to more mobility? That result is actually driven by the two CZs toward the right side of the chart—those would be Friday Harbor, on San Juan Island just off of Seattle, and Jackson, Wyoming, home to the Jackson Hole resort—but take them away and the line is flat, indicating no correlation.

Maybe we are looking at the wrong kind of mobility. Chetty’s data has a slew of mobility measures. Let’s try his measure of absolute mobility—the expected adult income rank of a child whose parents were at the 25th percentile. And let’s also look at the probability that a child starting in the bottom fifth of the income distribution makes it to the top fifth as an adult. We may as well look at both inequality measures, and what the heck, let’s look at single motherhood again. First, the relative mobility results again:

Errr….well, OK, the top-one-percent share isn’t that strongly connected to mobility after all. Looking across states rather than CZs, an increase in the top one percent’s share does consistently (if weakly) predict lower mobility, but the results (not shown) are basically the same as the cross-CZ estimates. In particular, the single motherhood results make inequality look comparatively unimportant for mobility, whether looking across states or CZs.

But, hey, the measure of inequality across the middle of the income distribution clearly is connected to mobility. Maybe the inequality-versus-single motherhood comparison is being distorted in some way. The 741 CZs range from Los Angeles with its 16 million people in the 2000 census to Murdo, South Dakota, population 1,193. Maybe if we focus in on the biggest ones, the importance of inequality will come through more clearly. So let’s hone in on the one hundred largest CZs, going down the list until we get to Madison (population 590,755). That’s 70 percent of the U.S. population.

OK, here are those three charts again, looking at relative mobility:

Uhhh….oops, those charts don’t indicate that either kind of inequality affects mobility. Huh. I guess this weakens the assertion in the widely-cited “Middle-Out Mobility” paper from the Center for American Progress that

"The strong inverse relationship between inequality and mobility is further demonstrated by the Chetty study which shows that the Great Gatsby Curve holds not only across countries but across regions within the United States."

But look over here! That paper was primarily about how opportunity is fostered by a large middle class (which is, of course, threatened by rising inequality). The authors conclude from their analysis of the Chetty data, “as a region’s middle class expands, so too does mobility.” So, here’s the chart proving the robustness of that finding (top one hundred CZs again):

Oh…that relationship goes in the right direction, but it is statistically no different from a slightly negative correlation, and the best-fitting line itself looks flatter if you ignore Brownsville, Texas, as shown by the black line. In fact, it turns out that across the top one hundred CZs, neither measure of inequality nor the size-of-middle-class measure has a correlation with any of the mobility measures that is statistically different from zero. Meanwhile, the relationship between single motherhood and mobility holds up in all of these analyses.

All right, time to ditch our framing device. You take our point? The evidence across local job markets in the United States throws severe doubt on the idea that income inequality harms opportunity (and it makes a hash of the empirical basis for the left’s “middle-out” sloganeering). As we and others have shown, as evidence goes, the Great Gatsby Curve is paper thin.

While Chetty and his team have been admirably cautious about over-interpreting their correlations, others have been…less so. Krueger used the Curve to project a decline in mobility for American children (after an earlier attempt by the Administration to project a large drop was shown to be ill-founded). Projections using the Great Gatsby Curves above would reassure that rising inequality will not hurt mobility.

Corak has tried to have it both ways, at times warning that the Curve can’t be dismissed as just showing correlation, other times characterizing it as merely a “communication device” to start a conversation around policies that might better promote mobility. That’s a dodge. What do our Great Gatsby Curves above communicate? On the basis of these charts, rather than a new Washington Center on Equitable Growth housed at CAP and devoted to discovering the damages that income inequality inflicts, the left should have started a Washington Center on Single Motherhood.

Corak has also emphasized that there are compelling theoretical reasons for worrying about income inequality’s impact on mobility. So if correlation does not equal causation, then perhaps correlation-plus-theory does. But guess what? The theoretical underpinning of the case for focusing on family structure is at least as solid.

Do not misunderstand us: to conclude from these charts that single motherhood is a primary cause of immobility would be no more justified than concluding the same about inequality from Corak’s chart. (For what it’s worth, one of us suspects that single-parent families are a big problem for upward mobility, while the other is more skeptical.) The point is that finding a statistical correlation between two variables is a starting point for serious analyses of the causes of immobility, not convincing proof you were right all along. Academic economists, in their day jobs, are well aware of this. If liberals want to use evidence like the Great Gatsby Curve to make the case that inequality is the challenge of our time, they cannot complain when conservatives want to focus on family structure instead. More to the point, they have no basis for saying the other side is wrong and their side is right. The case that rising inequality should concern us more than any number of alternative policy issues remains dubious.

At least the Great Gatsby Curve still holds across countries, right? It’s illuminating and disturbing! More on this to come, but here’s a tip for Curve-worshipers: caution should be your watchword, unless you are fully prepared to admit your mistake should it be shown that that cross-national curve has been over-hyped too.

Original Source:



On Obamacare's Second Birthday, Whither The HSA?
Paul Howard, 10-16-14

You Can Repeal Obamacare And Keep Kentucky's Insurance Exchange
Avik Roy, 10-15-14

Are Private Exchanges The Future Of Health Insurance?
Yevgeniy Feyman, 10-15-14

Reclaiming The American Dream IV: Reinventing Summer School
Howard Husock, 10-14-14

Don't Be Fooled, The Internet Is Already Taxed
Diana Furchtgott-Roth, 10-14-14

Bad Pension Math Is Bad News For Taxpayers
Steven Malanga, 10-14-14

Proactive Policing Is Not 'Racial Profiling'
Heather Mac Donald, 10-13-14

Smartphones: The SUVs Of The Information Superhighway
Mark P. Mills, 10-13-14


The Manhattan Institute, a 501(c)(3), is a think tank whose mission is to develop and disseminate new ideas
that foster greater economic choice and individual responsibility.

Copyright © 2014 Manhattan Institute for Policy Research, Inc. All rights reserved.

52 Vanderbilt Avenue, New York, N.Y. 10017
phone (212) 599-7000 / fax (212) 599-3494