Our interest in the statistical modeling of data on the timing of recidivism began in the mid 1970s when we were both junior members of the eco­ nomics department at the University of North Carolina.