The seminal economic model of crime (Becker, 1968), claiming that crime can be reduced by increasing either the probability or the severity of punishment, has triggered a huge amount of empirical research. On one hand, the aim was to test the empirical relevance of Becker's model of crime; on the other hand, information was desired to be gained concerning the relative effectiveness of higher sentence probability and harsher sentence severity. Obviously, in view of the growing amount of resources governments are spending on crime deterrence, knowledge about the crime-reducing power of these two policy instruments is crucial to achieve optimal resource allocation. More precisely, a hint as to whether crime should be combated by spending more on police (thereby increasing the probability of a sentence) or by convicting to longer prison sentences (thereby increasing the severity of a sentence) would be of greatest importance. Unfortunately, results published so far allow for no clear-cut policy recommendations.
Studies based on individual data suggest that the number of rearrests for released prisoners decreases with longer prison sentences experienced in the past as well as higher percentages of convictions (given arrest). (1) However, the specific sample selection makes more general recommendations for crime policy difficult. More interesting for this purpose are studies based on aggregate data. Findings from cross-sectional analyses mostly indicate that sentence probability is more effective in reducing crime than sentence severity (i.e. the [absolute] elasticity of the former is larger). (2) However, several flaws inherent in all these cross-sectional analyses (such as the impossibility of controlling for unobserved heterogeneity among the countries) fostered the emergence of panel data estimates with fixed effects. In panel data studies, differences between sentence probability and severity became even more apparent. Whereas higher arrest (and conviction) rates still showed a crime-inhibiting impact, the effect of longer prison sentences turned out to be rather ambiguous. (Sentence severity was insignificant in Cornwell and Trumbull, 1994, and Mustard, forthcoming, negative in Viren, 1994, even positive for certain crimes in Marselli and Vannini, 1997, and not explicitly incorporated in Levitt, 1998a).
A primary goal of this article is to examine whether the seemingly greater effectiveness of sentence probability still holds in a dynamic context. For instance, although the above-mentioned studies suggest that higher sentence severity has a smaller simultaneous effect on crime compared to sentence probability, its impact may be longer-lasting in return. Obviously, a meaningful comparison of these two crime-reducing policies requires information about the overall (short-and long-run) effectiveness of sentence probability and sentence severity.
Second, this article presents a new method that enables the distinction between deterrence and incapacitation effects. (3) Longer prison sentences as well as more frequent convictions (to prison) potentially reduce crime by two different channels: first, by deterring more offenders from committing crimes due to the higher expected penalty (deterrence effect); second, by having (more) convicted offenders (longer) imprisoned, who cannot commit any crimes during imprisonment (incapacitation effect). Because incapacitation is costly, a policy instrument that is able to reduce crime by deterrence rather than imprisonment is ceteris paribus preferable. Therefore, by separating deterrence from incapacitation effects, this article is able to give more precise directions toward a cost-effective crime-reduction policy than prior studies could.
Because the strength of law enforcement is possibly endogenous, the authors consider vector autoregressions (VARs) as a promising method for the stated purpose and apply them to Swiss property crime data (theft and robbery). …