The effect of imprisonment on recommitment an analysis using exact, coarsened exact, and radius matching with the propensity score
This study examines the effect of prison versus community sanctions on recommitment to prison and compares two levels of community supervision, community control (house arrest) and probation, evaluating whether the findings are contingent on the type of matching methods used in the analysis.
Logistic regression was conducted on unmatched and matched samples. Exact, coarsened exact, and radius-matching procedures were used to create a selection on observables design. Matching variables included current offense, demographics, criminal history, supervision violations, and a rich set of Florida Sentencing Guidelines information culled from an official scoring sheet. Florida judges use this instrument to sentence offenders within the framework of the state determinate sentencing system.
The results show that with exact matching, there is no effect of imprisonment on recommitment, while the other procedures suggest a specific deterrent effect of imprisonment. All four analysis methods showed that offenders under community control are more likely to reoffend than those under normal probation. Analyses between the matched and unmatched prison observations demonstrate that the matched set of prisoners is composed of offenders who have less extensive criminal records and less serious conviction offenses than unmatched offenders regardless of the matching algorithm.
Contrary to a prior analysis of these data, which found a criminogenic effect of prison, a null effect was found using exact matching. Comparing the matching procedures, the more precise the match the less likely there was an effect of prison. However, community control was criminogenic regardless of the matching procedure.
Gaes, G. G., Bales, W. D., & Scaggs, S. (2016). The effect of imprisonment on recommitment: an analysis using exact, coarsened exact, and radius matching with the propensity score. Journal of Experimental Criminology, 12, 143-158. DOI: 10.1007/s11292-015-9251-x