Evaluating Treatment Efficacy with Searual Offenders: The Insensitivity of Recidivism Studies to Treatment Effects

Abstract

Treatment programs for sexual offenders have been implemented in prison settings with the objective of reducing recidivism among released offenders. Reviews of the literature evaluating the efficacy of treatment have not found convincing evidence that institutional treatment reduces recidivism. However, these reviews have been most concerned with the possibility of a Type I error in hypothesis testing: specifically, that we might reject the null hypothesis when it is true, concluding that recidivism among treated offenders has been reduced compared with that of untreated offenders and concluding that treatment is effective when it is not. The present paper explores the risks of Type II error by examining the sensitivity of recidivism studies to treatment effects and the power of statistical tests of treatment hypotheses in recidivism studies. Using a series of "what if" analyses and power calculations, the sensitivity of statistical hypothesis testing was explored in recidivism studies under a variety of N's, base rates, and treatment effects. The size of treatment effects required for significance at the p < .05 level at various N's and base rates was calculated and the N required to obtain significance at the p < .05 level in the "average" recidivism study was estimated. This paper examines the sensitivity of statistical hypothesis testing in three of the most oft-cited recidivism studies of institutional sexual offender treatment. Recidivism studies were found to be quite insensitive to the effects of treatment and these findings are discussed with respect to the likelihood of a Type II error. Alternative methods of assessing sexual offender treatment efficacy are described and recommended