Study 1
Bird, Nguyen, and Grattet (2020) used a quasi-experimental design to assess the impact of California’s Proposition 47 (Prop 47) on rearrest and reconviction rates for individuals convicted of drug offenses at a 1-year follow-up. Twelve counties representative of the state — Alameda, Contra Costa, Fresno, Humboldt, Kern, Los Angeles, Orange, Sacramento, San Bernardino, San Francisco, Shasta, and Stanislaus — were identified and provided local jail and probation systems data. Taken together, these counties comprise 60 percent of California’s population and represent the state’s geographic diversity by including counties from the northern, southern, coastal, and valley areas. Data from the participating counties were merged with data from the California Department of Justice and the California Department of Corrections and Rehabilitation. Measures from these latter sources include criminal history and recidivism outcomes.
Drawing on a rich set of individual-level characteristics, a genetic matching strategy was used to conduct two separate analyses. Genetic matching is a matching strategy similar to but distinct from propensity score matching. The authors maintain that genetic matching improves on propensity score matching by directly matching on characteristics to construct a control group that is most similar to the treated group across all the observed factors that may drive selection into treatment.
In the first comparison (the main analysis), individuals convicted of drug possession offenses in the pre–Prop 47 period (November 2012 through October 2013) were matched with similar individuals convicted of drug possession offenses in the post–Prop 47 period (November 2014 through October 2015) to estimate the effects of Prop 47 on rearrest and reconviction rates, controlling for any remaining differences in individual characteristics and including county fixed effects. The second comparison (the difference-in-differences analysis) compared all individuals convicted of eligible drug possession offenses both before and after the passage of Prop 47 (the Prop 47 drug possession intervention group) with individuals convicted of other drug offenses such as possession for sale, sales, or manufacturing that are not eligible under Prop 47, both before and after its passage (the non–Prop 47 control group), to assess whether changes in the recidivism rates of the two groups are attributable to Prop 47 or, instead, to broader recidivism trends.
The CrimeSolutions review of this study focuses on the difference-in-difference analysis. This analysis compares individuals convicted of drug possession offenses with individuals convicted of other drug offenses both before and after the passage of Prop 47. The post–Prop 47 drug possession offenses group included 14,002 individuals convicted of drug possession offenses. This group was predominately male (82 percent); on average 24.4 years old at first conviction; and about 40 percent Hispanic, 39 percent White, and 17 percent Black. Further, these individuals had an average of 18.7 total arrests, with an average of 7.5 total arrests for drug crimes. The initial pre–Prop 47 drug possession offenses group included 21,303 individuals from the pre–Prop 47 period. After genetic matching, the pre–Prop 47 drug possession group was reduced to 7,490 individuals who had characteristics similar to those of the post–Prop 47 drug possession group. Similarly, the post–Prop 47 other-drug-offenses group included 20,090 individuals convicted of other drug offenses. This group was predominately male (81 percent); on average, 24.9 years old at first conviction; and 41 percent Hispanic, 36 percent White, and 17 percent Black. Further, these individuals had an average of 17 total arrests, with an average of 6.8 total arrests for drug crimes. The initial pre–Prop 47 other-drug-offenses group included 18,135 individuals from the pre–Prop 47 period. After genetic matching, the pre–Prop 47 group was narrowed to 11,008 persons who closely resembled the post–Prop 47 other-drug-offenses group. The difference-in-differences analysis used regression with controls to compare changes in rearrest and reconviction rates for any crime for all individuals in the Prop 47 drug possession intervention group (n = 21,492) with all individuals in the non–Prop 47 control group (n = 31,098) at the 1-year follow-up.
Finally, there was the possibility that Prop 47 would encourage a shift in charging practices away from Prop 47 offenses, perhaps swapping what previously would have been charged as a Prop 47 offense with a non–Prop 47 offense. To explore this issue of charge-swapping, the authors also report findings for a measure of rearrest that includes supervision violations and a measure of reconviction that includes revocations. All analyses controlled for differences in individual characteristics and included county fixed effects. Subgroup analyses were conducted within specific offense categories.