This report, sponsored by the Pennsylvania Commission on Crime and Delinquency (PCCD), summarizes the results of a reliability and validation study using the LSI-R scores and recidivism data conducted by the Institute on Crime, Justice and Corrections (ICJC) at the George Washington University.
Results indicate that although there are a limited number of LSI-R items for which substantial reliability has been achieved that have a statistical relationship to recidivism, overall, the LSI-R, as tested in institutional settings had problematic reliability. Using a condensed instrument that includes the items with higher rates of reliability would result in greater specificity in the identification of prisoners with the highest and lowest risk to public safety. The LSI-R instrument effectively separated risks on cases into three categories. However, the results do not warrant its use by the parole board for assessing risk at the time of a parole interview. Instead the LSI-SV would be more effective for the board’s risk assessment. Analysis revealed that approximately half of the parolees recidivated within 1 year of their release date. Findings include: the most predominant reasons for recidivism were technical parole violations, followed by new convictions and absconding; only 8 of the 54 LSI-R items were found to be associated with recidivism; while the LSI-R classifies prisoners according to their recidivism rates, the lack of reliability among many of the LSI-R items diminishes the overall validity; the LSI-SV (the Screening Version) can be substituted for the LSI-R; and using a combination of the 8 most reliable and valid items from the LSI-R, plus several other demographic items resulted in the best predictive results. The sample of 1,006 prisoners used for the validation study was not a representative sample of all prison releases. The pilot involved prisoners who were housed at nine Department of Corrections facilities who had an LSI-R completed on them, and were released on parole. 12 tables