Abstract: Numerous studies have shown that several characteristics of offenders are related to their likelihood of recidivism after release from prison. Nearly all of these studies, however, have focused on offenders from just one state. Few studies have examined recidivism rates controlling for the characteristics of offenders from multiple states, and virtually none have examined recidivism rates controlling for characteristics of offenders from multiple states during different periods of time. Additionally, few studies have explored different types of recidivism across multiple jurisdictions. To address these shortcomings, this dissertation applied logistic regression models to data from the publicly available Prisoners Released in 1994 dataset to investigate the extent to which nine individual level factors explain variation in recidivism rates within three years of release from prison across 15 states. The nine factors are: 1) gender, 2) age at first arrest, 3) race, 4) age at release, 5) number of prior arrests, 6) type of current offense, 7) time served, 8) admission type and 9) release type. Eight forms of recidivism were examined: 1) rearrest for any offense, 2) rearrest for a new violent offense, 3) rearrest for a new property offense, 4) rearrest for a new drug offense, 5) rearrest for a new public order offense, 6) reconviction probability if rearrested, 7) reimprisonment probability if reconvicted, and 8) parole violations. The dissertation investigated differences in the effects of the individual level factors on each form of recidivism. To investigate the effects of criminal justice policies and practices on state differences in recidivism rates, multilevel models were estimated that include three contextual variables, in addition to the nine individual factors. The state-level contextual variables are: 1) drug arrests per 100,000 residents, 2) police officers per 1,000 residents and 3) the arrest-offense ratio. In a final analysis, regression analyses were conducted to determine the extent to which the nine individual factors explain the increase in the three-year rearrest rates among persons released from prison in 1983 and 1994. The findings reveal that differences in individual level characteristics help to explain the variation across states for some, but not all, forms of recidivism. The findings related to rearrest for a new violent offense, reconviction probability, and parole violations were not conclusive. Results from the multilevel models indicate that the contextual factor of police officers per 1,000 has a significant impact on property rearrests and a marginal impact on drug rearrests and reconviction probability. The analysis of rearrests during two separate time periods revealed that changes in contextual factors, as opposed to individual level characteristics, were responsible for the increase in rearrest rates which occurred between 1983 and 1994. This study provides evidence that both individual level and contextual factors play a role in recidivism and need to be taken into consideration in implementing policy and designing programming. Two conclusions consistent with the findings are that treatment services need to be based on offender need and risk level and that states should consider reinstating discretionary parole. It would be beneficial for future research to examine the effect of additional individual and contextual variables on recidivism rates, particularly if a multistate dataset, similar to the one used in this study, becomes available in which county of release is specified.