Does automatic conviction relief affect recidivism: evidence from California

• In Event: Automating Justice: State Policy Variations, Discretion, and the Impact of "Clean Slate" Record Relief

Abstract

Criminal records create persistent barriers to employment, housing, and educational opportunities long after a person completes their sentence. To dismantle these barriers, California has enacted the most expansive automatic criminal record relief laws in the country. Under these policies, all arrests and most misdemeanor convictions, as well as most non-serious, non-violent, and non-registrable felony convictions, are eligible for automatic relief. This study provides the first large-scale causal evaluation of the impact of California’s conviction relief policies on subsequent arrest likelihood.

We utilize individual-level administrative records from the California Department of Justice, leveraging a natural experiment: the simultaneous granting of conviction relief to 1.2 million individuals in July 2022. We focus specifically on conviction relief, as convictions carry the most severe collateral consequences in the labor and housing markets. The relief population has a median age of 53 and is primarily male (75%), with the majority having had no criminal justice contact for over two decades. 

Our empirical strategy employs an event study and a two-way fixed effects difference-in-differences design. We estimate impacts on two groups: the full population of individuals with prior convictions, and a "recent conviction" subsample (convictions within the last seven years). This latter group is theoretically most likely to benefit, as California’s Investigative Consumer Reporting Agencies Act (ICRAA) already prohibits commercial background check agencies from reporting convictions older than seven years.

Across multiple specifications, we find statistically significant but substantively small increases in the likelihood of arrest for the group that received conviction relief relative to the comparison group. These effects range from 0.00134 percentage points in the full sample to 0.0026 percentage points in the recent sample, which represent a modest 0.20% to 0.44% change from the pre-period mean. While statistically detectable, both groups have extremely low absolute arrest rates. We hypothesize that a lack of awareness may contribute to these results, as California does not currently notify individuals when their records are cleared.

Finally, we provide descriptive evidence of an "implementation gap" between assessed eligibility and actual relief. Of the 5.5 million people with conviction records remaining as of July 2022, we estimate that fewer than 20% are legally ineligible due to serious, violent, or sex offenses. However, we find millions of potentially eligible individuals, including more than 2.5 million people with misdemeanor convictions alone, who did not receive relief in the initial wave. While some of these records may have had their relief reversed by the courts, we hypothesize that this gap likely reflects conservative eligibility algorithms used by the state to minimize false positives in the face of incomplete data. Our findings suggest that while automating relief is a necessary policy change, it is not alone sufficient to improve outcomes. Its impact is likely constrained by a lack of beneficiary notification and technical implementation challenges that result in the withholding of relief for millions of eligible individuals.

Authors

• Alissa Julia Skog, University of California, Berkeley
  Presenting Author

• Johanna Lacoe, University of California, Berkeley
  Non-Presenting Co-Author