The study estimates the upper and lower bounds of the important parameters for selection, based on results of previous work. Using a computer simulation of the criminal justice system, the sensitivity of selectivity measures of the present system to plausible changes of the values of each parameter is demonstrated. Then the simulation is extended to evaluate the predictive accuracy of tests based on the number and rate of prior arrests and on social characteristics such as employment and marital status. Next, the likely crime control effects of selective strategies are assessed. Four strategies are considered: police procedures aimed at ensuring the arrest of repeat offenders, prosecutors' career criminal programs oriented toward more thorough and certain prosecution of them, and practices that alter bail provisions and provide for enhanced sentences for chronic, dangerous offenders. The final section of the study includes tentative conclusions about the relative value of available predictive techniques, the efficacy of each selective strategy, and recommendations for further research and evaluation efforts. Graphic and tabular data are provided. (Author summary modified)