What determines crime rates? An empirical test of integrated economic and sociological theories of criminal behavior

Abstract

Research on crime has by no means reached a definitive conclusion on which factors are related to crime rates. We contribute to the crime literature by providing an integrated empirical model of economic and sociological theories of criminal behavior and by using a very comprehensive set of economic, social as well as demographic explanatory variables. We use panel data techniques to estimate this integrated crime model for property and violent crime using the entire population of all 100 counties in North Carolina for the years 2001–2005. Both fields contribute to the explanatory power of the integrated model. Our results support the economic explanation of crime with respect to the deterrent effect of the probabilities of arrest and imprisonment concerns, as well as the time allocation model of criminal activities. In contrast, the integrated model seems to reject the impact of the severity of punishment on crime levels. With respect to the sociological theories of crime, we find most support for the social disorganization theory and for the routine activity theory. Finally, we find differences between property and violent crimes, mostly explained by the sociological models.

Keywords:

• Crime
• Property crime
• Violent crime
• Deterrence
• Integrated model

Notes

1 Strictly speaking Cesare Beccaria already introduced this idea in 1764.

2 See for instance Cornwell and Trumbull (1994) for an overview of these early empirical studies.

3 For a review on economic and legal and illegal income opportunities theories of crime, see Ehrlich (1996).

4 Although we focus on those prime categories of serious crimes, our study does not include certain types of crimes such as tax evasion (Cebula, 1997) or securities fraud (Engelen, 2004).

5 We only present the regression equation for the integrated model as this is our final objective.

6 There is a possible endogeneity problem with including this variable. We will discuss this issue in detail in Section 6.4.

7 We only introduce this theory in the robustness section as including the rehabilitation programs reduces the sample size to about 260.

8 We formally tested the difference between the OLS and the 2SLS results and the coefficients were not significantly different from each other.

9 We thank an anonymous reviewer for this valuable suggestion.

10 We thank an anonymous reviewer for this valuable suggestion.