Beginning with the seminal work of Becker , an extensive economic literature has analyzed criminal behavior and issues of criminal justice.(1) The linchpin of the economic model of crime is the concept of deterrence: rational agents, faced with higher probabilities of detection or more severe sanctions, will commit fewer criminal acts.
In the real world, however, there are a number of obstacles to effective deterrence. First, criminals may be poorly informed about the likelihood of detection, or may be overly optimistic about their own criminal abilities. Second, whereas the benefits of crime accrue immediately, the costs of crime (e.g. imprisonment) are administered with a substantial lag. To the extent that criminals are myopic (Wilson and Herrnstein ), even large punishments will have little weight in the current decision of whether or not to commit a crime. Finally, among certain groups, serving time in prison is seen as a rite of passage so that being arrested is sometimes viewed as a positive outcome by the criminal (Venkatesh ).
Ultimately, whether deterrence is of practical applicability is an empirical question. Testing for deterrence is difficult because increasing the expected punishment of a criminal act (either through a higher likelihood of detection, or via increased sanctions conditional on being caught), reduces crime in two ways. The first channel is deterrence: larger penalties will lead criminals to commit fewer crimes. The second channel is incapacitation. If criminals commit multiple offenses and punishment takes the form of imprisonment, increasing the expected sanction will also reduce crime by getting criminals off the streets. While a criminal is imprisoned, he is unable to engage in criminal actions that otherwise would have been taken.
The primary source of empirical evidence in favor of deterrence comes from studies that relate measures of the likelihood or intensity of punishment (typically the aggregate arrest rate for a particular crime) to the reported level of that crime.(2) Consistent with the predictions of the deterrence hypothesis, such studies almost always uncover a negative relationship between punishment and crime rates. There are, however, two major shortcomings of those studies. First, they have no power to differentiate between deterrence and incapacitation effects; both deterrence and incapacitation effects are expected to increase with the expected punishment. Without imposing arbitrary functional form restrictions on the utility function of criminals, the two channels cannot be distinguished using this approach. Second, data limitations lead to the possible presence of endogeneity bias in the same direction as predicted by deterrence. The hypothesized relationship in the economic model of crime relates the true arrest rate (arrests/total crimes) to total crime. Because total crimes are not observed by the econometrician, however, the actual measure of both the crime rate and the arrest rate used in empirical studies are computed using reported crimes in place of total crimes. Any measurement error in reported crime will therefore appear both in the left-hand-side variable and in the denominator of the right-hand-side variable. It is straightforward to show that unlike the standard measurement error case, where the estimated coefficients tend to be biased towards zero, measurement error of this type will lead to a negative bias, even if the true coefficients are already negative (Gibbs and Firebaugh ). Given that only 38% of all crimes are reported to the police according to victimization surveys (Bureau of Justice Statistics ), the potential for measurement error in reported crimes would appear substantial.
This paper attempts to discriminate between deterrence, incapacitation, and measurement error as explanations for the empirical relationship between arrest rates and crime.(3) The issue of measurement error is addressed first. …