Wednesday, July 2, 2014

Glenn Loury on Discrimination

Douglas Clement has yet another in his fine series of interviews with economists, this one with Glenn Loury, published in the June 2014 issue of The Region, a publication of the Federal Reserve Bank of Minneapolis. Here are a few insights from Loury's work on discrimination, but the interview also touches on crime and incarceration, inequality, and the evolution of economic theory. 

A standard approach to studying discrimination in labor markets is to collect data on what people earn and their race/ethnicity or gender, along with a number of other variables like years of education, family structure, region where they live, occupation, years of job experience, and so on. This data lets you answer the question: can we account for differences in income across groups by looking at these kinds of observable traits other than race/ethnicity and gender? If so, a common implication is that the problem in our society may be that certain groups aren't getting enough education, or that children from single-parent families  need more support--but that a pay gap which can be explained by observable factors other than race/ethnicity and gender isn't properly described as "discrimination." Loury challenges this approach, arguing that many of the observable factors are themselves the outcome of a history of discriminatory practices. He says:

"By that I mean, suppose I have a regression equation with wages on the left-hand side and a number of explanatory variables—like schooling, work experience, mental ability, family structure, region, occupation and so forth—on the right-hand side. These variables might account for variation among individuals in wages, and thus one should control for them if the earnings of different racial or ethnic groups are to be compared. One could put many different variables on the right-hand side of such a wage regression.
Well, many of those right-hand-side variables are determined within the very system of social interactions that one wants to understand if one is to effectively explain large and persistent earnings differences between groups. That is, on the average, schooling, work experience, family structure or ability (as measured by paper and pencil tests) may differ between racial groups, and those differences may help to explain a group disparity in earnings. But those differences may to some extent be a consequence of the same structure of social relations that led to employers having the discriminatory attitudes they may have in the work place toward the members of different groups.
So, the question arises: Should an analyst who is trying to measure the extent of “economic discrimination” hold the group accountable for the fact that they have bad family structure? Is a failure to complete high school, or a history of involvement in a drug-selling gang that led to a criminal record, part of what the analyst should control for when explaining the racial wage gap—so that the uncontrolled gap is no longer taken as an indication of the extent of unfair treatment of the group?
Well, one answer for this question is, “Yes, that was their decision.” They could have invested in human capital and they didn’t. Employer tastes don’t explain that individual decision. So as far as that analyst is concerned, the observed racial disparity would not be a reflection of social exclusion and mistreatment based on race. ...  But another way to look at it is that the racially segregated social networks in which they were located reflected a history of deprivation of opportunity and access for people belonging to their racial group. And that history fostered a pattern of behavior, attitudes, values and practices, extending across generations, which are now being reflected in what we see on the supply side of the present day labor market, but which should still be thought of as a legacy of historical racial discrimination, if properly understood.
Or at least in terms of policy, it should be a part of what society understands to be the consequences of unfair treatment, not what society understands to be the result of the fact that these people don’t know how to get themselves ready for the labor market.
When I'm giving a talk on these issues, I point out that there are a variety of kinds of discrimination. One kind of discrimination is when an employer treats two people with the same qualifications differently because of race or gender. Another kind of discrimination can cause social conditions that lead to people being more likely to have different qualifications in the first place. I have argued that this pattern means that suing employers for discrimination should thus have a smaller place, and trying to equalize qualifications should have a bigger place. But Loury has a thought-provoking response to this approach.

In one of his papers, Loury considers various kinds of interventions that have a goal of reducing the effects of discriminatory behavior. He draws two distinctions. You can intervene early--say, trying to assure better grade-school performance, a better high school graduation rate, and a higher level of college attendance. Or you can intervene later, when people are actually applying for jobs. You can also intervene in a :"blind" way, which favors a broad group that will be disproportionately of a certain race, or in a "sighted" way that favors the group specifically. Thus, expanding government funding for preschool programs for low-income families is in these terms an early "blind" intervention. A quota for hiring a certain percentage of African Americans or other ethnic groups to certain jobs is a late "sighted" intervention.

I tend to favor the first kind of intervention, to which Loury offers a couple of counterarguments. One is that a later "sighted" policy is also an incentive for skills acquisition at an earlier age. As Loury puts it:
One of our key insights is that under sightedness (again, overt discrimination in favor of a particular group), the very act of boosting people’s access to slots—that is, putting a thumb on the scale in their favor at the point where they compete for positions—implies a subsidy to their acquisition of skills. ... [I]f a later intervention is properly anticipated, then an earlier intervention may not be necessary; it may be redundant. ... Now, this result—that we find quite interesting—requires the assumption I just referred to: that when making their decisions about how to invest in the development of their skills, people be farsighted enough to anticipate the consequences of their being favored at the point of slots allocation. That assumption will not be plausible in every case (youngsters can be unnervingly short-sighted...)."

However, if one is willing to grant the possibility that knowing certain jobs are likely to be available will tend to encourage skills acquisition earlier, Loury then offers another point. There is a classic question in the economics of taxation that considers whether a country should impose taxes on a variety of inputs to the production process, or instead a tax on outputs. The general finding is that the negative effects of the tax are smaller if you impose them at the end of the production process, because then the taxes don't also have a distortionary effect on the process of production itself. There's a related classic question in the economics of monopolies, which asks whether it's worse for an economy to have a monopolist that raises the price on an input used by many producers, or for a monopolist to raise the prices to consumers. Again, the negative effects of monopoly are smaller if they result in higher prices at the end of the production process.

Loury and a co-author create a model that applies similar reasoning to the question of when to intervene to stop discrimination, with the implication being that intervention at the later stage of being hired may be less burdensome to an economy than intervention at the earlier stage. Loury says:

"The distortion (in our case, preferences for a disadvantaged group) should take place “downstream,” at the point of competition for final positions, rather than “upstream,” at the point where people are investing in their own productivity. ... Now, you’d think that for affirmative action it might be different, that, well, it’s always better to go early. ... Pre-K is something people are advocating these days. And, indeed, there may be other reasons, not in our model, having to do with cycles of development and so forth, which would explain why early intervention of a different kind is warranted. But if it’s purely in the framework of our model, I think our finding is explicable in terms of intuitions that you find in other areas of economics."
I'm not sure I'm persuaded! But Loury's tight analysis and probing insights are always worth reading. If you would like to read more of Loury laying out these ideas, a possible starting point is an article he wrote for the Spring 1998 of the Journal of Economic Perspectives, called "Discrimination in the Post-Civil Rights Era: Beyond Market Interactions." Like all JEP articles, it is freely available on-line courtesy of the American Economic Association. (Full disclosure: I've been Managing Editor of the JEP since the inception of the journal in 1987.)