The States involved in the multisite evaluation were Florida, Georgia, Louisiana, New York, Oklahoma, South Carolina, and Texas. Community supervision performance was analyzed with survival time models. These models analyze the length of time until an event occurs (e.g., recidivism), rather than only whether or not that event occurred. Louisiana and New York were the only States in which any evidence that the boot camp program reduces recidivism was found. In New York, the reduction in recidivism was limited to technical violations. In both Louisiana and New York, the possibility that these differences were the result of the intensive community supervision following release cannot be excluded. Overall, the study concludes that if success is measured in terms of recidivism, there is no evidence that the in-prison phase of boot camp programs has been effective. Future studies of recidivism should use random assignment to either shock incarceration or a control group. In addition, evaluation efforts would be improved if treatment and control groups received equal levels of supervision upon release to the community. 25 references