NCJ Number
205690
Date Published
May 2004
Length
40 pages
Annotation
This report presents a study examining the predictive power of the criminal history measure of the Federal Sentencing Guidelines.
Abstract
Under current Federal sentencing guidelines, judges must refer to a table of two axes involving an offender's "offense level" and "criminal history category" in order to arrive at a minimum and maximum sentence for the offender. The criminal history component is drawn from the validated United States Parole Commission's "Salient Factor Score: and the "Proposed Inslaw Scale." The current study may serve as a performance review of the criminal history component's predictive ability for offender recidivism. Data included recidivism information for a random sample of 6,062 offenders sentenced in fiscal year 1992 under Federal sentencing guidelines. Three types of recidivism were considered for the study: (1) re-conviction for a new offense; (2) re-arrest; or (3) supervision revocation. Three analytical techniques were used to analyze recidivism rates among the sample of offenders: (1) tabular analysis; (2) area measurement under the receiver operating characteristic curve; and (3) survival analysis with hazard modeling. The results indicate that the Federal guidelines' criminal history points and categories statistically predict recidivism. Criminal history points were found to have more predictive power than criminal history categories. Thirteen offender characteristics are described that account for variations of recidivism rates; they include gender, age at sentencing, race or ethnicity, employment status, educational attainment, marital status, illicit drug use, guideline offense level, guideline applied for instant offense, sentence type, length of sentence for instant offense, departure status, and type of recidivating event. The final conclusion indicated that the Federal guidelines' criminal history measurement showed statistically predictive power in predicting offender recidivism and is performing as intended. The statistical difference between criminal history points and categories may not have significance for policy. Other interesting findings that emerged from the research and deserve further consideration include the finding of no correlation between recidivism and the guidelines' offense levels. References, exhibits, appendix