The Analysis of Prison-Prisoner Data Using Cluster-Sample Econometrics: Prison Conditions and Prisoners' Assessments of the Future

The study investigates whether and how strong prison conditions contribute to the perceived propensity to recidivate after controlling for personal characteristics and criminal background. In order to combine different sources of information on personal characteristics of prison inmates and administrative prison data in an efficient way, we propose the use of matched prison-prisoner data and application of cluster-sample methods such as GEE (generalized estimating equations). Estimated average partial effects based on GEE and random-effects Probit modeling reveal that prison conditions show significant effects on the perceived likelihood of future reincarceration. Particularly, we find that inmates facing prison overcrowding show a reduced likelihood of recidivism.