RT Article
T1 Parolee Recidivism and Successful Treatment Completion: Comparing Hazard Models Across Propensity Methods
JF Journal of quantitative criminology
VO 31
IS 1
SP 149
OP 181
A1 Peters, David J.
A2 Hochstetler, Andy
A2 DeLisi, Matt
A2 Kuo, Hui-Ju
LA English
YR 2015
UL https://krimdok.uni-tuebingen.de/Record/1764279344
AB Objectives Ascertaining the effect of treatment on recidivism is a core area of investigation in criminology and corrections research. The two objectives of the current analysis are: (1) to determine the true effect of treatment regimen completion on time to recidivism controlling for propensity to complete treatment; and (2) to examine the sensitivity of results under various propensity score methods. Methods Drawing on the population (n = 1,270) of parolees residing in a Midwestern state, we examine the effect of completing a treatment regimen on days to recidivism (using two failure outcomes) over a 2-year period using proportional hazard models. We adjust for the propensity to complete a treatment regimen using the covariate adjustment, inverse weighting, case matching, and strata methods. Results Completing a treatment regimen has a sizable effect at reducing recidivism risk, which grows stronger the longer offenders are on parole. This effect is consistent across treatment propensity methods. It is driven mainly by completion of alcohol and drug treatment regimens. Treatment effects are stable across two measures of recidivism (arrest/prison-return and prison-return only). Conclusions Discussion centers on the implications for assessing treatment success in the parole population as well as on methodological implications for researchers conducting similar research. In the current analysis propensity scores produce stable results regardless of propensity method. Guidance is provided on selecting propensity methods based on data distortion, technical expertise, and presentation of results. We conclude that the covariate adjustment method is best suited for novice researchers, and the case matching method for expert researchers. The strata method is recommended for supplemental analyses. Future research should examine treatment effects reporting at least two propensity methods.
K1 Hazard models
K1 Parole
K1 Propensity score
K1 Corrections
K1 Treatment
K1 Recidivism
DO 10.1007/s10940-014-9229-2