A central question in the fields of criminology and the economics of crime concerns the extent to which incarcerating the criminally active reduces national crime rates. The short-run causal pathways connecting incarceration to crime are two-fold. First, those who are incarcerated cannot commit crimes. Second, individuals may be deterred by the threat of incarceration. Despite the simplicity of these ideas, the results from empirical research vary considerably. Much of this research suffers from a key problem: any societal forces that increase criminal activity will increase the incarceration rate. For example, if the economy were to sour and more people commit crime as a result, more people will be arrested and sent to prison. In the aggregate, we observe crime increasing as incarceration increases, creating the false impressions that incarceration increases crime.

This project will measure the effect of incarceration on crime by making use of an unusual policy event. In August 2006, the Italian government released one-third of the nation's prison inmates in an effort to relieve prison overcrowding. Any inmate with less than three years remaining was immediately released. As a result, in August of 2006 the Italian prison population declined abruptly, discretely, and substantially. Since this change in incarceration was clearly driven by a policy choice rather than changes in external factors that cause crime, any observable changes in crime surrounding the pardon can be attributed to the change in the prison population. This project will evaluate the impact of this pardon on Italian crime rates. The resulting empirical results will provide estimates of the amount of crime caused by prisoner releases and will be used to produce estimates of the annual crimes prevented through incarceration. Most importantly, the project will provide estimates that are free from bias associated with a reverse impact of crime on incarceration.

Project Report

A growing body of studies finds significant and in some instances quantitatively substantial causal impacts of incarceration on crime. While results are sensitive to estimation methodology and tend to be context specific, most careful research finds that policy-induced increases in incarceration rates generally lead to decreases in crime. However, the exact mechanisms driving this inverse relationship have proved difficult to pin down. Whether this relationship is driven primarily by deterrence of the potentially criminally active or the incapacitation of the already criminally active is an open empirical question. Moreover, there is growing interest in how the marginal impact of incarceration on crime changes as the incarceration rate increases, especially in the United States where incarceration rates are currently at historic highs. These particulars of the prison-crime relationship are more than mere academic debates, as they bear directly on crime control policy as well as general theoretical research on criminal participation. With regards to policy, to the extent that potential criminals are deterred by severe punishment, optimal sentencing structures should emphasize stiff penalties over apprehension since the latter policy tool is resource-intensive while the former may decrease crime at low cost. On the other hand, if prison reduces crime primarily through incapacitation, greater resources should be devoted to identifying the most criminally active, apprehending them, and incapacitating them. Moreover, heterogeneity in the propensity to reoffend among the incarcerated implies that the crime preventing benefits of additional prison years served will vary from inmate to inmate. Lengthy sentences that incarcerate people beyond the age of desistence as well a liberal application of incarceration as punishment may incarcerate some offenders whose behavior may be effectively controlled through less restrictive means. Being able to distinguish the relative importance of deterrence and incapacitation in explaining the prison-crime effect would also inform theoretical reasoning regarding the decision to participate in criminal activity. The economic model of crime postulates the existence of a rational offender who weighs the expected costs and benefits and makes decisions accordingly. Alternative criminological and sociological theories emphasize human capital endowment, socialization towards anti-social norms, peer-influence, biology, and other determinants of crime that do not fit neatly within the rational choice framework. A quantitative assessment of the relative importance of incentives as opposed to pre-determined characteristics among those deterred or constrained by prison would provide information regarding which set of theories best describes the criminal behavior of the most serious offenders in society. This project presents what we argue to be lower-bound estimates of pure incapacitation effects based on variation in incarceration caused by an unusual policy event in the Italian corrections system. In August 2006, the Italian government released more than one-third of the nation’s prison inmates following the passage of national legislation aimed at relieving prison overcrowding. The collective pardon did not impact sentencing for future offenders who were not incarcerated at the time of the pardon and enhanced sentences for those pardoned offenders who reoffend within the five years following their early release. On net, these two factors likely induced a modest deterrent effect on criminal activity. Hence, any observed increase in crime associated with the mass pardon arguably reflects a lower-bound incapacitation effect estimate. We make use of three sources of variation in incarceration rates. First, using national level monthly crime data, we test for a discontinuous break in national crime rate time series associated with the August 2006 mass release. Second, a simple theoretical model of incapacitation suggests that a massive one-time increase in the prison release rate should induce a dynamic adjustment process back towards steady-state values for both the crime rate as well as the incarceration rate. We use variation along this adjustment path (ignoring the discontinuity) to provide a second estimate of the incapacitation effect. Finally, we exploit provincial variation in the impact of the pardon based on the province of residence of pardoned inmates. In addition to providing an alternative source of variation for an overall average incapacitation effect estimate, variation across provinces in pre-pardon incarceration rates permits a test of whether reverse incapacitation effects are smaller in provinces with higher pre-pardon incarceration rates -- i.e., whether the crime prevention returns to incarceration diminish as the incarceration rate increases. All three sources of variation yield estimates of sizable incapacitation effects, ranging from 13 to 46 crimes prevented per prison-year served. Nearly all of this impact is attributable to theft and robbery, with mixed evidence regarding other offense categories. Our cross-province analysis finds average incapacitation effects that are generally consistent with the estimates based on the breaks in the national level time series caused by the pardon. We find strong evidence of diminishing crime-prevention returns to scale. Specifically, the reverse incapacitation effect of the pardon is considerably smaller among those provinces with higher pre-pardon incarceration rates, holding pre-pardon crime rates constant.

National Science Foundation (NSF)
Division of Social and Economic Sciences (SES)
Standard Grant (Standard)
Application #
Program Officer
Marjorie Zatz
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of California Berkeley
United States
Zip Code