CEPR - Center for Economic and Policy Research


En Español

Em Português

Other Languages

Home Publications Blogs CEPR Blog Unequal Access to Education and the Gatsby Curve

Unequal Access to Education and the Gatsby Curve

Written by John Schmitt and Shawn Fremstad   
Thursday, 01 March 2012 09:45

The central issue in the furor that erupted a few weeks back over the "Gatsby Curve" was whether or not the sharp increase in inequality over the last three decades has depressed economic mobility. President Obama, the chair of his Council of Economic Advisers, and most of the leading experts in the field argue that today's extreme inequality is not only bad in itself, but it also lowers economic mobility. Brookings Institution fellow Scott Winship, however, says mobility is no worse today than it was earlier in recent economic history.

Winship's critique of the standard view on declining mobility centers on what he sees as limitations in the available data and problems with the statistical techniques used by economists analyzing the recent trends. At this point, quite a few people have offered responses to Winship's technical concerns (including one of us (pdf) at a recent event sponsored by the New America Foundation). But, most of this recent debate has been confined to a fairly narrow (and generally technical) discussion of Winship's criticisms of the way surveys capture, and the way economists model, income.

A recent paper (pdf) by Martha Bailey and Susan Dynarski, however, allows us to completely sidestep Winship's concerns about measuring income across generations. Bailey and Dynarski analyze college completion rates over time and find a disturbing pattern that provides strong support for the view that economic mobility has declined substantially over the last three decades.

Here is the key figure in their analysis (click to enlarge):


Source: Author's calculation based on data from the National Longitudinal Survey of Youth. 1979 and 1997 (U.S. Bureau of Labor Statistics, 2010a. 2010b)

Bailey and Dynarski used two data sets, one that followed people born in the early 1960s (1961-64) and the other that followed people born in the late 1970s and early 1980s (1979-82). For each of the two initial periods, the economists divided families into four equal-sized groups by income, from lowest to highest. Then, they asked how many children in each of these income fourths had completed a four-year college degree 25 years later. For the group born in the early 1960s, 5 percent of the children born into the poorest fourth of families had finished college by age 25, compared with 36 percent of those in the richest fourth of families. For the group born about 20 years later, in the late 1970s and early 1980s, 9 percent of those born into the bottom forth had finished college by age 25, compared with 54 percent for those children born into the top quartile.

It is hard to interpret these numbers as anything other than making a strong case that mobility is a lot lower for the children of the 1980s than it was for the children of the 1960s, especially given the large increase between the 1980s and the 2000s in the earnings of college graduates.

To see this point most easily, it helps to invert the numbers in the Bailey and Dynarski chart. The figure below shows the share of each quartile that does not have a four-year college degree by age 25. For those born into the lowest income families in the 1960s and trying to make their way into the top quartile later in life, 95 (out of 100) did not have a college degree. These non-college-educated young people born to low-income families in the 1960s were competing for slots in the top quartile with the 64 (out of 100) non-college-educated young people and 36 (out of 100) college-educated young people born into families in the top fourth of families.


Source: Bailey and Dynarski (2011), Figure 3.

For those born into the bottom quartile almost 20 years later, the odds were much worse. Now, there were 91 (out of 100) without a college degree --a slight improvement over the situation for the earlier generation. But, these 91 non-college-educated young people born into low-income families were competing for top slots against 54 (out of 100) college graduates --not 36 (out of 100) barely two decades earlier.None of this evidence depends on how we measure income (except to assign children into the initial family income distribution) or on any modeling needed to overcome limitations in the data. And the education data leave little doubt that it is much harder for low-income children born in the 1980s to get ahead than it was for low-income kids born in the 1960s (when, of course, it wasn't very easy either).

Tags: inequality | poverty

Comments (7)Add Comment
written by Arjen, March 01, 2012 9:24
Am I missing something here or is the conclusion of this piece wrong? If we just look at the competition between the bottom and top part of the population (and assume this population to be 100), the 5 poor college graduates in the 80's had to compete with 36 rich college graduates, which means they have a chance of 5/41=12% to get a well paying job. Twenty years later the odds are 9/(54+9)=14%, which actually shows a slight improvement for poor people to get a well paying job.
I'm with Arjen
written by tom, March 01, 2012 10:19
I agree with the conclusion, but I don't buy this analysis. First, I think you need to consider how much of an advantage a college degree provides. Second, I think that it would still be best to consider actual mobility.
It's the non-college graduates
written by John Schmitt, March 01, 2012 1:58
The vast majority of the kids born into the bottom quartile are non-college graduates, so the focus in Arjen's comment on what happens to the small share of college graduates in the bottom quartile in the two periods misses the point.

The relevant ratios are 95:36 versus 91:54. For the early 1960s birth cohort, there were about 0.4 (95:36) college grads at the top for every non-college grad at the bottom. For the early 1980s cohorts, there were about 0.6 (91:54) college grads at the top for every non-college grad at the bottom. That is roughly a 50 percent increase in college grads at the top for every non-college grad at the bottom.

Another way to think about this is to ask: if you were a parent, and you knew your kid would be born into the bottom quartile, which world would you rather your kid be born into?

In the later period, your kid has a slightly higher chance of having a college degree (9 versus 5 percent), but is ---either way-- competing against many, many more college graduates at the top (54 versus 36 percent). In this most recent period, 91 out of 100 times, your kid is going to be competing without a college degree against 54 kids at the top who have one. In the earlier period, 95 of 100 times your kid won't have a college degree, but will will be competing against only 36 out of 100 college grads at the top.

Odds ratios
written by BrendanH, March 02, 2012 4:38
You get half way to talking odds ratios: you mention odds, and you invert the outcome. Comparing the odds of getting a degree, for each category relative to the richest quartile, the picture looks like this:

Lowest 25%: 0.094 vs 0.084
2nd quartile: 0.289 vs 0.226
3rd quartile: 0.364 vs 0.401
4th quartile: 1 (reference)

In this view the biggest losers are the second quartile.

Odds ratios are a better measure than differences or ratios of proportions, particularly when the baseline outcome probability changes. That is, they're a better measure of how the underlying structure of inequality changes when there is big growth in college completion.
It's still about the non-college grads
written by John Schmitt, March 02, 2012 6:49
But, BrendanH, the core issue here isn't the odds of getting a college degree, it is the odds of moving from the bottom into the top quartile. It isn't possible with the information here to calculate the odds (or the changes in the odds) of moving from the bottom to the top quartile. We don't know, for example, what share of college grads and what share of non-college grads move to the top in either period.

But, the data are pretty suggestive --which is the point of the post. The bottom quartile is overwhelming non-college educated in both periods, but the top quartile is much more college-educated in the second period. It is hard to tell a story about the world where the benefits of a college degree for 4 extra people at the bottom outweigh the substantial increase in the disadvantage for the 91 people at the bottom who still don't have a college degree (where the increase in relative disadvantage is a function of the big rise in college completion at the top).
Inequality of opportunity vs inequality of outcome
written by BrendanH, March 02, 2012 11:33
As you say, you've got no data here on the outcome you're interested in, but the link between parents' income and education is suggestive.

ORs are good measures of inequality of opportunity, and here it is instructive to see that while the lowest quartile is screwed in both periods, the lower-middle is the one that is losing most in relative terms.
Second quartile
written by John Schmitt, March 02, 2012 11:48
Agreed. The deterioration in relative opportunity that you show for the second quartile is important and interesting --and something we hadn't paid any attention to.

Write comment

(Only one link allowed per comment)

This content has been locked. You can no longer post any comments.


Support this blog, donate
Combined Federal Campaign #79613
budget economy education employment Haiti health care housing inequality jobs labor labor market minimum wage paid family leave poverty recession retirement Social Security taxes unemployment unions wages Wall Street women workers working class

+ All tags