Article ID: | iaor1993149 |
Country: | United States |
Volume: | 40 |
Issue: | 3 |
Start Page Number: | 485 |
End Page Number: | 504 |
Publication Date: | May 1992 |
Journal: | Operations Research |
Authors: | Tien James M., Ellermann Raymond, Sullo Pasquale |
Keywords: | probability, statistics: data envelopment analysis |
Empirical estimates of quantile residual life functions can be employed effectively to obtain properties of recidivism and to help screen parametric mixture models. In this manner, the Burr model is demonstrated to be an appropriate model for characterizing recidivism. When applied to certain data, the model suggests that while the observed declining recidivism rate can be explained by population heterogeneity, individual recidivism rates may in fact be increasing. The quantile residual life function approach to modeling recidivism is applied to two often-referenced data sets, as well as to an extensive data set obtained from the state of New York which is new to the criminal justice literature.