The clinical prediction method relies upon a clinical analysis of the characteristics and circumstances of each person considered for parole. The statistical prediction method relies on quantifiable variables identified from the prior parole outcomes of offenders. Inmates rating high on variables associated with recidivism are not granted parole. From June 1978 through December 1979, the Israel Individual Pardons Boards reviewed the files of 1,897 inmates. This study collected data on the boards' recommendations, inmate sociodemographic characteristics, the current offense, criminal record, and variables concerning the release. In the study's first stage, the link between the dependent variables (the boards' recommendations) and the other four groups of variables was examined. At the second stage, recidivism data on the 622 released inmates were analyzed. When the dependent variable was recidivism after release, the statistical prediction was more reliable than clinical prediction. 3 tables and 29 references.