In the public debate over incarceration policy, there is considerable disagreement about what value of individual offending frequency (lambda) is appropriate to use in estimating incapacitative effects. This article provides an approach for estimating the mean values of lambda for diverse subsets of the total offender population, with particular emphasis on subsets generated by filtering through various stages of the criminal justice system. Sharp differences in offending frequency are displayed between robbery and burglary inmates, across three states, and particularly between resident inmates and free, active offenders. Free offenders average I to 3 robberies and 2 to 4 burglaries per year, while resident inmates have lambda values 10 to 50 times higher. Differences result front the underlying levels of criminal activity and the sanction levels that offenders face. A highly heterogeneous distribution of offending frequency in the total population of offenders combines with relatively low imprisonment levels to lead to substantial selectivity of high-lambda offenders among resident inmates and a correspondingly low mean value of lambda among those offenders who remain free. These results have important implications for estimating incapacitative effects of an increase in incarceration since the additional inmates will be drawn from free offenders whose mean lambda is at least an order of magnitude lower than that of the current inmate population.
