A Reconsideration of Some Parole Prediction Factors

THIS study is predicated on Ohlin's thesis: greatest increases in predictive accuracy will undoubtedly depend on securing better factors rather than on refining the techniques and methods of prediction work. 1 New factors, formulated for the present study, as well as factors used in earlier studies, were tested on 4448 inmates paroled during the decade 1940-49 from the Pontiac Branch of the Illinois State Penitentiary. This is an institution for prisoners classified young and improvable. The mean age at parole was 24.1 years and the average time served before parole was 3.6 years. For two-fifths of these inmates, parole was terminated in the Army. Although comparisons from numerous standpoints revealed little difference between the Army and the non-Army parolees, the violation rate for the Army cases was only 3.8 per cent, while for the non-Army group it was 39.4 per cent. Investigation of this disparity revealed not only that the Army in wartime provides an unusually favorable parole situation, but also that behavior in the Army which would constitute violation in civilian life is frequently not reported. Further, because of special dispensations granted by the Parole Board for military service, the average time on parole was only 15.8 months for men discharged f om parole while in the Army, but for non-Army cases it was 33 months. In view of the differences between the two parole situations, and the similar traits of the two groups of parolees, it was decided to base parole predictions for current cases only on the 2693 non-Army parolees. The Army cases, from World War II, illustrate the point that in parole prediction our primary sampling problem lies in the selection of time periods-not populations. We predict from experience in one period what the outcome will be in another period, although parole violation rates vary with wars, economic trends, Parole Board policies and other time-connected variables. We must, therefore, seek stable factors. For inclusion in this study a factor was required to show relatively consistent predictive utility in each of three periods: 1940-41, when 1451 Pontiac inmates were paroled, of whom 60 per cent were nonArmy cases; 1942-44, when 1970 were paroled, of whom 47 per cent were nonArmy cases; and 1945-49, when only 1027 were paroled, but 88 per cent were non-Army cases. These variations reflect marked changes in parole practice in Illinois during the period studied, which provide a stringent test for predictive stability. Nevertheless, the total number of non-Army cases in each of the three periods ranged only from 869 to 920, and the range of violation rates was only from 37.4 per cent to 40.0 per cent. Unless otherwise indicated, all violation rates and other data cited in this report will be for the non-Army cases for the full 1940-49 time sample.