Local Factors Explaining the Incidence of Criminal Offences in Romania. A Geographically Weighted Regression Model

Zizi Goschin (zizi.goschin@csie.ase.ro)
Dept of Statistics and Econometrics, Bucharest University of Economic Studies 
Institute of National Economy, Bucharest, Romania


Criminality has long became a matter of increasing concern for national and local decision makers alike, given the high disturbances, hardships and adversities it brings to the social and economic environment. In the last decades the number of criminal offences displayed growing territorial inequalities across Romania. The incidence of criminality in a region depends not only on local socioeconomic conditions, but also on the ones in nearby regions, due to significant population mobility. Spatial econometric techniques account for such territorial correlations, including the use of spatial weights that capture the influence of each region upon its neighbours. Among the spatial methods, the geographically weighted regression (GWR) is a valuable instrument that allows estimating local coefficients, specific to each location, thus providing useful information for appropriate policy design at regional level. In this context we employed a criminality GWR model in an attempt to find the local determinants, both economic and demographic, that explain the spatial distribution of criminal offences in Romania. The results indicated that the incidence of this phenomenon in Romania is linked to factors largely acknowledged in the literature, such as local development, incomes, unemployment, and population density. The novelty brought about by the GWR model compared to previous research is that it also revealed important spatial variations in the impacts of the variables and indicated which counties are more vulnerable to specific factors. From an econometric perspective the GWR model represents a better fit than the classic OLS model, in addition to capturing the spatial variation in coefficients’ estimation.

Key Words: criminality rate, geographically weighted regression, regions, Romania
JEL Classification: R19, C54.

[Full Text]

Romanian Statistical Review 2/2019