In their influential 2010 book, The Race between Education and Technology, Harvard University economists Claudia Goldin and Lawrence Katz offer an explanation for the United States’ decades-long rise in wage inequality. In their view, the main reason that inequality has increased so much is because the supply of educated workers has not kept pace with an ever-growing demand–especially for workers with a college degree. The short supply of college-educated workers has driven up their price relative to the rest of the workforce, accounting for most of growing gap between workers at the top and the bottom of the earnings ladder. The research implies that the most direct and effective way to reduce the wage gap is to expand the share of the workforce with a college degree.

Goldin and Katz’s diagnosis and prescription represent the predominance of rising wage inequality within academic and Washington policy circles. But, this spring, first in public remarks and later in an interview with the Washington Post, Harvard economist Lawrence Summers declared that focusing on education and training as a way to reduce inequality is “whistling past the graveyard” and “fundamentally an evasion.”

After making these informal comments, Summers–together with Melissa Kearney and Brad Hershbein, both of the Hamilton Project at the Brookings Institution–produced a more formal analysis of how much increasing the share of college-educated workers could aid in reducing inequality. Their more formal analysis concluded “Increasing educational attainment will not significantly change overall earnings inequality” but would “reduce inequality in the bottom half of the earnings distribution, largely by pulling up the earnings of those near the 25th percentile.”

We argue that Hershbein, Kearney, and Summers get it right when they conclude that even a large jump in college attainment would have little impact on overall earnings inequality. But we also believe that they are overly optimistic in their assertion that increasing college attainment will reduce inequality at the bottom.

As Hershbein, Kearney, and Summers correctly argue, expanding the college-educated workforce would do little to lower inequality because “a large share of earnings inequality is at the top of the earnings distribution, and changing college shares will not shrink those differences.” The reason that the top one percent earn so much more than the rest of the workforce is not fundamentally because they have a college or advanced degree. About one third of workers already have a college degree or more, and inequality has increased substantially within that group between 1979 and 2014. As Hershbein, Kearney, and Summers maintain, even a sharp increase in the share of the college-educated population is not likely to put meaningful downward pressure on the earnings of those at the very top.

Their analysis is too sanguine, however, with respect to non-college-educated workers. The authors’ conclusions about workers at the bottom and the middle rest on two assumptions: First, that arbitrarily giving some non-college-educated workers a college degree will automatically give them access to earnings equal to those of existing graduates, and second, that reducing the supply of non-college-educated workers (by turning some of them into college graduates) will boost the earnings of the remaining non-college workers substantially. Both assumptions are unlikely to be true. As a result, the hypothetical plan to bestow 10 percent of non-college-educated men with a diploma would do nowhere near as much for inequality between the middle and the bottom as Hershbein, Kearney, and Summers suggest.

We note that Hershbein, Kearney, and Summers limit their analysis to men, because the period over which they estimate the effect of education attainment on wages is characterized by a large increase in the share of women in paid work, which complicates the analysis. Since the default for working-age men has been market labor throughout the period they analyze, it’s sensible to consider the effect of attainment on the wage distribution of men only, while understanding there are implications for women as well.

To help understand the Hershbein, Kearney, and Summers’ thought experiment, imagine that there are two bowls: one filled with non-college-educated men and one filled with college-educated men. The hypothetical exercise takes 10 percent of the people in the first bowl (of non-college graduates) and puts them into the second one (for college-graduates). This has three effects: It changes earned income for the people in the first bowl (those without degrees) by reducing the supply of non-graduates, bidding up their earnings. It changes earned income for the people in the second bowl (graduates) by increasing the supply, pulling down their wages. And it changes earned income for the people who were moved from the first bowl to the second bowl (from non-graduates to graduates) by giving them access to the higher earnings received by graduates. (See Figure 1.) In each of the three cases, however, the effects assumed by Hershbein, Kearney, and Summers are likely overstated.

Figure 1


Appealing to a 2010 paper by Daron Acemoglu and David Autor, both economists at the Massachusetts Institute of Technology, Hershbein, Kearney, and Summers argue that shifting 10 percent of men from the first bowl to the second would reduce the wage gap between college and non-college workers by 18 percent, which Hershbein, Kearney, and Summers divide half-and-half between an increase in non-college earnings and a reduction in college earnings. The main basis for that 18 percent estimate is the experience of the 1970s, when the share of college-educated workers increased substantially as the Baby Boomers entered the workforce with far more education than their parents’ generation. This large increase in the supply of graduates arguably drove down the earnings of college graduates relative to the rest of the workforce. When the growth in the college-educated share of the workforce slowed in the 1980s, the college wage premium opened up again. That pattern is the principal motivation for the idea that inequality is primarily “the race between education and technology.”

The decision to divide the 18 percent into two equal parts, with a 9 percent increase in earnings increase for non-college-educated men, drives the reduction in inequality in the bottom of half of earners, one of the key findings that the authors highlight. But the labor market now is very different than it was in the period that Acemoglu and Autor analyze. Since around 2000, the labor market has been deteriorating, jobs are scarce, and the share of the adult population that works has declined. The modest expansion of the mid-2000s did not bring workers back to where they’d been in 2000, and the recovery from the Great Recession of 2007-2009 has not (yet) brought workers back to where they were in 2008. There is simply too much slack remaining in the labor market–for both non-college-and college-educated workers—for reassigning workers from one camp to another to make much difference.

That excess supply is fundamentally why reducing the number of non-college-educated workers (removing workers from the first bowl) is unlikely to increase their earnings by 9 percent. All the college-educated workers who can’t find jobs or are in positions for which they’re over-qualified need to find work or better work first. Only then will we see the emergence of a seller’s market for the non-college-educated, one in which employers have to out-bid each other to find workers. That competition among employers, which we last saw at the end of the 1990s, is what’s necessary to trigger rising wages among the supply of non-college-educated workers.

A second empirical problem with the analysis is that the workers who are assumed to receive an instant college degree are, contrary to a core assumption of the analysis, unlikely to command the kinds of earnings received by those who already have a college-degree. Instead, these hypothetical graduates would continue to compete for the same jobs as the non-college-educated, but the degree would give the graduates a leg up. That, in turn, would push some of the remaining non-college-educated workers out of the labor market entirely.

So, yes, a college degree would improve the individual circumstances of the new graduates relative to those who were not granted an instant degree, but an important part of the payoff would be the ability to out-compete non-college graduates for jobs that don’t actually require a college degree. That, more or less, is what a 2015 study titled “Dropouts, Taxes, and Risks: The Economic Return to College under Realistic Assumptions,” by Alan Benson and Frank Levy of the Massachusetts Institute of Technology and independent economist Raimundo Esteva, finds. The economic benefit to obtaining a college degree for the population that is currently dropping out or otherwise on the cusp of getting one is quite modest. Our colleague Elisabeth Jacobs evocatively referred to this phenomenon as a “Cruel Game of Musical Chairs.” (See Figure 2.)

Figure 2


It’s worthwhile to place this analysis within the context of a larger debate about the labor market and why it’s not delivering broad-based wage growth to the people who comprise it. Since 2000, median wages have stagnated and the labor market participation rate has fallen, as have the rates of job-to-job mobility and household and small business formation. Young workers are not climbing the job ladder to the middle class. The share of national income earned by workers declined. Over an even longer timeframe, wages have not kept pace with worker productivity.

All of these phenomena suggest that the labor market isn’t working for most employees—problems that aren’t confined to those without a college education—and that suggests the problem isn’t that too few people have college degrees. Rather than focus on education attainment as the solution to inequality, it’s time for policy-makers to move on from the race between education and technology and focus on our stagnant labor market. As Summers said, “the core problem is that there aren’t enough jobs.” The key to reducing inequality is more jobs and a higher demand for labor. In the absence of more jobs, heroic assumptions about educational improvement are likely to deliver only modest economic benefits.


—Marshall Steinbaum is a Research Economist and John Schmitt is the Research Director at the Washington Center for Equitable Growth.