Michigan set-asides found to increase wages and reduce recidivism

Preliminary results of an empirical study by two University of Michigan law professors show that setting aside an individual’s record of conviction is associated with “a significant increase in employment and average wages,” and with a low recidivism rate.  We know of only one other similar study, conducted by researchers at the University of California at Berkeley, and it came to essentially the same conclusion.  One relevant difference between the two studies is that in Michigan set-aside results in sealing of the record, while in California it does not.  Such studies are rare because of the difficulty of obtaining data, particularly where relief seals the record, but they are a very important way of advancing a reform agenda.  Thus, Professors Sonja Starr and J.J. Prescott propose that their research “provides important empirical guidance to the broader social policy debates associated with set-aside laws and accessibility of criminal records.”  In the hope that their work will encourage others to undertake similar research, we reprint the entire report below.

Project Outcomes Report: Award No. 1023727

Project Title: Evaluating the Impact of Criminal Record Set-Aside Laws on Recidivism and Socioeconomic Outcomes

Co-PIs: Sonja Starr and J.J. Prescott

Tens of millions of Americans have criminal records, which often carry collateral socioeconomic and legal consequences long after the criminal sentence is completed. To provide relief from these consequences, some states offer procedures by which certain offenders can have their records sealed or “set aside.” In Michigan, during the period of this study, set-asides were available only to offenders with a single criminal conviction, who may apply for them beginning five years after sentencing or release from incarceration. A set-aside removes the record from public view and from background-check databases, eliminates any applicable state occupational restrictions, and legally entitles the recipient to represent herself as having no record in employment proceedings. Accordingly, one might expect that recipients’ employment prospects would be improved. However, it is not obvious that one should expect this effect to be substantial, because people with records often face other employment hurdles and because the employment effects of the type of limited and older records for which set-asides are given are not known. No prior empirical study has quantified the effect of set-asides on employment, and in this study, we sought to do so.

A principal challenge in studying set-asides is that such records, after the fact, are by definition no longer publicly available. However, pursuant to deidentification conditions, we obtained set-aside records from the Michigan State Police and linked wage information on the same individuals from Michigan’s unemployment insurance system. This enabled us to track approximately 4,000 set-aside recipients’ wages and employment status on a quarterly basis for periods of at least three years before and after the set-aside was received. Having received our data later than intended, we continue to complete final analyses and prepare papers for submission for publication; the remaining work entails comparisons between set-aside recipients and comparable non-recipients, for whom we also obtained similar data. Here, we summarize the results of preliminary analyses focused on set-aside recipients alone, before and after receipt of the set-aside. Our analyses are based on regressions that account for pre-set-aside trends for these individuals, as well as for changes and fluctuations in the economy.

We find that receipt of a set-aside is associated with a significant increase in employment and average wages. Most of the gain is observed in the first year after set-aside receipt, during which recipients’ probability of employment rises steadily by a total of about 6.5 percentage points (from about 58.6% to about 65.1%–that is, recipients became about 1.11 times as likely to be employed). Over the same one-year period, recipients’ average quarterly wages rose by about 22%. This increase is too large to be fully attributed to the gain in employment probability—the implication is that set-aside recipients are often able to find higher-paying jobs, in addition to increasing their chances of finding work in the first place.

These trend-change estimates are obtained after filtering out preexisting trends, which are negative for most of the two years before the set-aside and flat in the six months immediately before the set-aside. That is, applicants are somewhat more likely to apply for set-asides after a period of unemployment. It is possible that some of the upturn in employment and wage trends for set-aside recipients could be accounted for not by the set-aside itself, but by mean reversion or by motivated job-hunting (that is, applicants might tend to seek set-asides at a time when they are especially motivated to find work, or to find higher-paying work). Our ongoing analyses seek to disentangle these competing explanations.

Wage and employment gains were similar across racial groups, but were much larger among females (whose wages rose 41% within a year after receiving a set-aside, compared to 15% for males). Females represented 47% of set-aside recipients. About 58% of recipients were white and about 37% were black, with the balance from other groups or unidentified. Approximately 71% had not faced incarceration for their convictions, and only 1% had been incarcerated for more than a year. Approximately 39% of the set-aside convictions were felonies.

Finally, because we had access to applicants’ full criminal records, we also assessed their recidivism probability. Fewer than 4% of set-aside recipients were rearrested within five years of the set-aside, and fewer than 2% were reconvicted. These are extremely low rates; we note for comparison purposes that in a five-year period, Michigan police make about 13 arrests per 100 people in the general population.

In addition to contributing to researchers’ understanding of the employment effects of having a criminal record, this research provides important empirical guidance to the broader social policy debates associated with set-aside laws and accessibility of criminal records. It provides a clearer picture of who set-aside laws affect and what the stakes are, in terms of benefits for recipients and public safety concerns. Moreover, the development of a procedure within Michigan state agencies to match deidentified criminal record and unemployment insurance data could help to benefit future researchers who wish to investigate relationships between socioeconomic variables and criminal offending.