The impact of state prison populations on crime is typically estimated
by applying the lambda, the individual crime rate, of prisoners or ar
restees. We outline the problems with this approach, attempt to reanal
yze the widely divergent lambdas derived in past research, and make ad
justments necessary to use lambdas for estimating the incapacitation i
mpact. The result is an uncertain estimate of 16 to 25 index crimes av
erted per year per each additional prisoner. We argue that regression
analysis can provide a better estimate of the impact of prison populat
ion growth. Applying the Granger test to pooled state data over 19 yea
rs, we found that prison population growth leads to lower crime rates
but that crime mte changes have little or no short-term impact on pris
on population growth. Next we regressed crime rates on prison populati
on and conclude that, on average, at least 17 index crimes are averted
per additional prisoner. The impact is limited mainly to property cri
me.