Crime-Control Effect of Incarceration: Reconsidering the Evidence, Final Report

This study critiqued two previous studies that measured the empirical relationship between incarceration and crime. The first study was by Levitt in 1996 and the second study was by Marvell and Moody in 1994. These studies have been used to support the expanded use of incarceration. It was suggested the methodologies used were limited in discussing punishment policy. It was also noted that estimating the effect of prison on crime was difficult due to the causation potentially flowing in both directions; the increase in imprisonment may reduce crime, but increased crime will put more people in prison (causing overcrowding litigation). In attempting to solve this problem, Levitt used instrumental variables regression, using regressed crime on prison. Three problems were identified with Levitt’s choice of instruments. First, it was just as susceptible to the simultaneity problem (rising crime rates cause increased prison population and increased crime determined overcrowding litigation). Second, for the time period under study only 12 States experienced system-wide litigation leaving a generalization for the other 38 States. And, third, seven of the 12 States were from the South and have had historically high rates of incarceration. The Marvell and Moody regressed crime rates on prison population over a 19-year period, finding that the size of the State prison population had a significant, short-term negative impact on crime. They determined that the regression model determined a causal relationship for at least short-run movements. Several technical problems were identified with this approach to prison-crime relationship. First, they used a narrow statistical test to justify the model and implemented the test incorrectly. The attention became the short-run effects of State prison populations (SPP) and not the long-run effects. Second, they used a narrow set of control variables. Lastly, they included a lagged dependent variable on the right-hand side of their model, controlling for dynamic effects and autocorrelation. This study extended the analyses to the use of flow measurement of prison admissions constructed from a data source using sentencing as a control. Sentencing practices were viewed as potentially the driving force explaining variation in crime rates. Prison admissions were separated into non-drug and drug offender admissions. A different picture seemed to emerge when using prison admission flows rather than prison inmate population. The impact on the crime rate due to the imprisonment of a large number of drug offenders has clear policy implications. Evidence showed that the effect of non-drug admissions was not the same as that of drug offender admissions. This study analyzed the effects of these admissions on violent and property crime rates. This study’s intent was to create speculation regarding existing estimates of the relationship between crime and incarceration and suggested future work pay close attention to the choice of model specification and find justification for chosen specifications in theoretical considerations. Tables