"Banning the Box" Significantly Increases Racial Discrimination

Employers substitute racial bias for criminal history questions.

"Ban the box" policies forbid employers from asking about a criminal record on a job application. Ban the box policies don’t forbid employers from running criminal background checks — they only forbid employers from asking about criminal history at the application/interview stage. The policies are supposed to give people with a criminal background a better shot at a job. Since blacks are more likely to have a criminal history than whites, the policies are supposed to especially increase black employment.

One potential problem with these laws is that employers may adjust their behavior in response. In particular, since blacks are more likely than whites to have a criminal history, a simple (if imperfect) substitute for not interviewing people who have a criminal history is to not interview blacks. Employers can’t ask about race on a job application, but black and white names are distinctive enough so that based on name alone, employers can guess with a high probability of being correct whether an applicant is black or white.

In an important and impressive new paper, Amanda Agan and Sonja Starr examine how employers respond to ban the box.

Agan and Starr sent out approximately 15,000 fake job applications to employers in New York and New Jersey. Otherwise identical applications were randomized across distinctively black and white (male) names. Half the applications were sent just before and another half sent just after ban the box policies took effect. Not all firms used the box even when legal so Agan and Starr use a powerful triple-difference strategy to estimate causal effects (the black-white difference in callback rates between stores that did and did not use the box before and after the law).

Agan and Starr find that banning the box significantly increases racial discrimination in callbacks.

One can see the basic story in the situation before ban the box went into effect. Employers who asked about criminal history used that information to eliminate some applicants and this necessarily affected blacks more, since they are more likely to have a criminal history. But once the applicants with a criminal history were removed, “box” employers called back blacks and whites for interviews at equal rates. In other words, the box leveled the playing field for applicants without a criminal history.

Employers who didn’t use the box did something simpler but more nefarious — they offered blacks fewer callbacks compared to otherwise identical whites, regardless of criminal history. Together, the results suggest that employers use distinctively black names to statistically discriminate.

A policy to help black job prospects mostly helped white men with criminal histories.When the box is banned, it’s no longer possible to cheaply level the playing field, so more employers begin to statistically discriminate by offering fewer callbacks to blacks. As a result, banning the box may benefit black men with criminal records, but it comes at the expense of black men without records who, when the box is banned, no longer have an easy way of signaling that they don’t have a criminal record. Sadly, a policy that was intended to raise the employment prospects of black men ends up having the biggest positive effect on white men with a criminal record.

Agan and Starr suggest one possible innovation blind employers to names. I think that is the wrong lesson to draw. Agan and Starr look at callbacks, but what we really care about is jobs. You can blind employers to names in initial applications, but employers learn about race eventually. Moreover, there are many other margins for employers to adjust. Employers, for example, could simply start increasing the number of employees they put through (post-interview) criminal background checks.

Policies like ban the box try to get people to do the “right thing” by blinding people to certain types of information. But blinded people tend to use other cues to achieve their interests, and when those other cues are less informative, that often makes things worse.

Rather than ban the box, a plausibly better policy would be to require the box. Requiring all employers to ask about criminal history would tend to hurt anyone with a criminal record, but it could also level racial differences among those without a criminal record. One can, of course, argue either side of that tradeoff, and that is my point.

More generally, instead of blinding employers, a better idea is to change the real constraints. At the same time as governments are forcing employers to ban the box, for example, they are passing occupational licensing laws that often forbid employers from hiring workers with criminal records. Banning the box and simultaneously forbidding employers from hiring workers with criminal records illustrates the incoherence of public policy in an interest-group driven system.

Ban the box is another example of good intentions gone awry because "the man of system" tries to arrange people as if they were pieces on a chessboard, without understanding that:

In the great chess-board of human society, every single piece has a principle of motion of its own, altogether different from that which the legislature might chuse to impress upon it. If those two principles coincide and act in the same direction, the game of human society will go on easily and harmoniously, and is very likely to be happy and successful. If they are opposite or different, the game will go on miserably, and the society must be at all times in the highest degree of disorder. (Adam Smith, Theory of Moral Sentiments.)

PS: The Agan and Starr paper has much more of interest. Agan and Starr, find, for example, evidence of discrimination going beyond that associated with statistical discrimination and crime. In particular, whites are more likely to be hired in white neighborhoods and blacks are more likely to be hired in black neighborhoods.

This post first appeared at Marginal Revolution.

Related Articles


{{relArticle.author}} - {{relArticle.pub_date | date : 'MMMM dd, yyyy'}} {{relArticle.author}} - {{relArticle.pub_date | date : 'MMMM dd, yyyy'}}
{{article.Topic.Topic}} {{article.Topic.Topic}}