Research Briefs in Economic Policy No. 65

Ban the Box, Criminal Records, and Statistical Discrimination

By Amanda Agan and Sonja Starr
December 7, 2016

In recent years, many cities and states have passed “Ban-the-Box” (BTB) laws, which seek to expand access to employment for people with criminal records. These laws ban employers from including questions about criminal records on job applications or in interviews. Employers are still permitted to conduct background checks, but not until the end of the hiring process. The theory is that once hiring managers have met the applicants in person, they will be more likely to consider carefully whether a conviction is job-relevant, rather than categorically dismissing applicants with records. Many advocates of BTB have framed it as a tool for reducing racial disparity in employment, and especially for reducing high rates of unemployment among black men. The rationale is straightforward: black men have higher felony conviction rates and thus should benefit disproportionately from policies that open doors to people with such convictions.

However, there is a theoretical reason to worry that this approach could backfire. Economists have long predicted that when employers are deprived of individualized information about job applicants they will engage in “statistical discrimination”: that is, they will rely on other observable characteristics to make (accurate or inaccurate) group-based generalizations about an applicant. In the BTB context, employers might use the race of the applicant to guess at the likelihood the applicant has a criminal record: they may simply assume that black male applicants are likely to have them, while white applicants are not. In short, employers who cannot discriminate directly based on criminal records might, instead, discriminate based on race. Doing so would be unlawful, of course, but laws against racial discrimination in hiring have proven difficult to enforce.

Continue to full version

To investigate the potential for statistical discrimination after BTB, we submitted nearly 15,000 applications on behalf of fictitious applicants before and after BTB laws went into effect. Our study focused on New Jersey and New York City, both of which implemented BTB laws for private employers in 2015. We randomly varied the race of the applicant (black or white) and whether the applicant had a felony conviction; otherwise, applicants’ characteristics were similar and randomized. Because our pools of white and black applicants thus had identical sets of other traits, we can confidently attribute systematic differences in white and black callback rates to racial discrimination. Likewise, because our applicants with and without criminal records were otherwise identical, we can confidently conclude that any pre-BTB differences in their callback rates were due to those records. Our field-experimental design thus has an advantage over real-world observational approaches, in which causal inferences can be confounded by other differences in applicant pools.

Our results pose a potential dilemma for policymakers: they support BTB’s basic premise that criminal records are a major barrier to employment access, but they also support the concern about statistical discrimination. When employers asked about criminal convictions, applicants without a felony conviction were 63 percent more likely to be called back than those with a conviction (5.2 percentage points over a baseline of 8.2 percent). This effect essentially disappeared after the enactment of the BTB law because almost all employers complied by removing the criminal-record question from their applications. Our study tested only initial callbacks, so we could not evaluate BTB’s additional premise that getting a foot in the door would help applicants with records get jobs even though employers eventually conduct background checks. However, our results do suggest that BTB is effective at helping applicants with records to obtain job interviews.

However, we also found that BTB substantially increases racial discrimination in employer callbacks. At companies that asked about criminal records before BTB, white applicants received 7 percent more callbacks than similar black applicants; after BTB, this gap grew to 45 percent. (In percentage point terms, we found that BTB expanded the black-white gap by about 4 percentage points: a large increase, given that the overall callback rate for the sample was just under 12 percent.) In contrast, we did not see any significant change at companies that did not ask about criminal records before BTB went into effect and were thus unaffected by BTB, and if anything, the black-white gap shrank at those companies over the same period. This provides reason to believe that the large expansion in the black-white gap at affected companies was a casual effect of BTB, rather than an unrelated change that happened to occur over the same time period.

The growth in the black-white gap appears to come from a combination of losses to black applicants and gains to white applicants. In particular, black applicants without criminal records see a substantial drop in callback rates after BTB, while white applicants with criminal records see large gains. This pattern supports the statistical discrimination theory: when employers lack criminal record information, they tend to assume that black applicants (and not white applicants) are likely to have records. Thus, the gains that BTB offers to people with records (in particular, white people with records) may come at the cost of black applicants without records, who lose their ability to neutralize employers’ negative assumptions by conveying their clean records. Further analysis suggests that these assumptions are exaggerated relative to the actual distribution of felony convictions in the population: that is, employers may be relying on assumptions or stereotypes about black criminality that are statistically ill-founded.

We believe our findings suggest a complicated challenge for policymaking. Of course, BTB’s implications for racial discrimination are not the only relevant consideration. Policymakers might decide that because of the especially serious employment barriers people with criminal records face, expanding job access for them is important enough to be worth pursuing despite the unintended consequence of reduced opportunities for black men without records. Alternatively, they might seek to pair BTB with strategies that attempt either to improve enforcement of racial discrimination prohibitions (a difficult task historically) or to change employers’ underlying incentives vis-à-vis people with records (for example, by expanding tax credits for hiring them). What we think is clear is that BTB, at least taken alone, should not be seen as a strategy for reducing racial disparity in employment: with respect to that goal, our study finds that it is counterproductive.

Note:

This research brief is based on Amanda Agan and Sonja Starr, “Ban the Box, Criminal Records, and Statistical Discrimination: A Field Experiment,” University of Michigan Law and Economics Research Paper no. 16-012, June 14, 2016,

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2795795##.

Amanda Agan, Rutgers University, and Sonja Starr, University of Michigan.