A survival model of pretrial failure

This paper examines the likelihood of rearrest during the pretrial period with a model that depends on both time elapsed since release and on individual and case characteristics. Using data on a sample of male arrestees released on recognizance in the District of Columbia in 1984, we apply a survival or hazard model to the problem of predicting pretrial rearrest. We are particularly interested in whether drug use, as measured by urinalysis at arrest, is predictive of pretrial rearrest and its timing. Results show, for example, that drug use or a charge for larceny is associated with high risk levels in the period immediately following release. In our data, the number of prior convictions exerts a strong effect on rearrest risk throughout the pretrial period, but the initial high risk associated with being on probation or parole or having pending charges decreases rapidly over the course of a year at risk.