Life in prison is meant to be difficult. But it doesn’t always get better once you’re out. Re-entering offenders often have a tough time finding employment, even when they are motivated and able to work. But “ban the box” - a popular policy aimed at helping ex-offenders find jobs - doesn’t help many ex-offenders, and actually decreases employment for black and Hispanic men who don’t have criminal records. This is a classic case of unintended consequences. We should repeal “ban the box” and focus on better alternatives.

How did we get here? Helping ex-offenders find jobs is a top policy priority for liberals and conservatives alike, for good reason. Without a stable job, it’s tough to build a life free of criminal behaviors and influences. When fewer jobs are available, re-entering offenders are more likely to reoffend. This increases crime and incarceration rates, which is costly to all of us.

A major obstacle to changing this status quo is that employers are reluctant to hire individuals with criminal records, often discarding applications from anyone who has a record. You might think this is unfair to individuals with criminal records who would make good employees. Perhaps it would be better to prevent employers from asking about criminal backgrounds up front, so that everyone gets a fair chance at an interview. This is the motivation behind “ban the box” policies.

“Ban the box” forbids public and often private employers from inquiring about an applicant’s criminal history until late in the hiring process. Such policies have been adopted in cities and states across the country. Late last year, President Obama even banned the box on applications for many federal government jobs. Proponents of second chances for ex-offenders have rejoiced at this apparent progress.

Here’s the problem: employers still don’t want to hire ex-offenders. Many ex-offenders would make good employees, and some were convicted of relatively minor crimes. But on average, ex-offenders are more likely than non-offenders to have engaged in violent, dishonest or otherwise antisocial behavior, and are more likely to engage in similar behavior in the future. (About two-thirds of released prisoners are rearrested within three years.) Employers, understandably, want to hire peaceful, honest, agreeable employees who won’t be taken off the job by an arrest or conviction. But these characteristics are largely unobservable when reading through job applications. You might be able to divine some of them from an interview, but first you have to decide which applicants to consider.

To do this, an employer will “statistically discriminate,” using the observable information that is most correlated with the unobservable information of interest to sort applications into “probably job-ready” and “probably not job-ready” piles. This type of discrimination is common and, in most cases, perfectly legal: For instance, employers might prefer applicants with a college degree not because of what one learns in college but because earning that degree is correlated with greater motivation, intelligence, and diligence - unobservable characteristics that make you a more productive employee. Sorting people based on whether or not they have a criminal record is far from perfect - some ex-offenders are more job-ready than others, just as some college grads are more job-ready than others. But the employer can’t see which ex-offender is more job ready, only which applicants are ex-offenders. They know that the average ex-offender is less job-ready than the average non-offender, and so base their hiring decisions on that.

If you take information about criminal records away, what happens? Employers are forced to use other information that is even less perfect to guess who has a criminal record. The likelihood of having a criminal record varies substantially with demographic characteristics like race and gender. Specifically, black and Hispanic men are more likely than others to have been convicted of a crime: the most recent data suggest that a black man born in 2001 has a 32% chance of serving time in prison at some point during his lifetime, compared with 17% for Hispanic men and just 6% for white men. Employers will guess that black and Hispanic men are more likely to have been in prison, and therefore less likely to be job-ready.

Statistically discriminating based on race and gender is, of course, unfair to the many black and Hispanic men who don’t have criminal records - just as statistically discriminating based on criminal history was unfair to the ex-offenders who were more job-ready than the average. (Discriminating based on race and gender is also illegal, but there’s plenty of evidence that such discrimination occurs anyway.) Taking information away in this context hurts more people than it helps. Just because employers can’t see an applicant’s criminal history doesn’t mean they don’t care about it. Under “ban the box”, they will avoid ex-offenders by avoiding groups that are more likely to contain ex-offenders, like black and Hispanic men.

What effects has “ban the box” had so far? Two new working papers suggest that, as economic theory predicts, “ban the box” policies increase racial disparities in employment outcomes.

Amanda Agan and Sonja Starr submitted thousands of fictitious job applications before and after “ban the box” went into effect in New Jersey and New York City, randomly assigning race and criminal history to each “applicant.” They then tracked the number of callbacks received. When employers asked about criminal records on the job application, they called white applicants slightly more often than identical black applicants - but that small gap became more than four times larger, and statistically significant, after “ban the box” went into effect. (White applicants with criminal records benefited the most from the policy change - they’re the ones who got a chance to prove themselves in an interview, though it’s unclear if they would have gotten a job offer. Employers are still allowed to check criminal records before making a final offer, so applicants could be turned away at that point.) Because of the randomization, they can attribute this effect to the removal of criminal history information from job applications.

In a separate paper, Benjamin Hansen and I exploit the variation in adoption and timing of “ban the box” policies across the country to measure the policy’s net effects on the employment outcomes of young, low-skilled men. We find that black and Hispanic men without college degrees are significantly less likely to be employed after “ban the box” than before. This result is not explained by pre-existing trends in employment, and persists for several years.

Overall, the unintended consequences of “ban the box” are large, and run counter to one of its goals: reducing racial disparities in employment. For this reason, I hope jurisdictions repeal their “ban the box” laws. But I also hope this doesn’t stop efforts to improve the lives of people coming out of prison. This is a group that our country has long neglected, and we should be doing much more to help them succeed. Advocates could push for policies that would provide more information to employers about ex-offenders’ job-readiness, rather than taking information away. Better yet, they could help disadvantaged ex-offenders improve their job-readiness. The more employable the average ex-offender, the less cautious employers will be about hiring one.