Modeling the suppression effect of correctional programs on juvenile delinquency

Studies of the effects of correctional programs on juvenile delinquency have observed that delinquents exhibit a sharp rise in their arrest rates up to the time of intervention. The drop to a lower rate following intervention has been labeled a suppression effect. A controversy has arisen regarding the nature of the suppression effect; some scholars attribute it to the effectiveness of the correctional programs, while others claim that it is due to a selection artifact. In this study, we examine attempts to model such phenomena and point out that the general terms in the model are not identifiable. Without identifiability, one can construct models that attribute the suppression effect either to the correctional program or to the selection artifact. Some identifiable models are proposed and their associate likelihood functions are used to present a process of model-based analysis to analyze data collected originally by the American Institutes for Research. Discussion of the feasibility of this type of probabilistic modeling approach to criminal justice phenomena is also given.