Testing the Level of Service Inventory–Revised (LSI-R) for Racial/Ethnic Bias

In corrections, the use of objective offender risk classification instruments is increasing. Some observers have raised concerns that these instruments might over classify poor and minority offenders, but there is little research to support these claims. Using contingency tables to compare risk categories and program performance, this study examines the Level of Service Inventory-Revised for racial/ethnic differences in classification errors at a large federal community corrections center in the United States. The study found a tendency toward more classification errors for African Americans than Caucasians or Hispanics, though the types and rates of errors were dependent on the choice of the cutoff score and the performance measure to be predicted. These findings highlight the need for correctional facilities to conduct local tests of their classification instruments. The study provides an easy and practical model for such tests.