Finding the European crime drop using a panel data model with stochastic trends

We develop a panel data model with stochastic dynamic processes to empirically verify the possible existence of the European crime drop. This time-varying effect can be captured by the stochastic trend and can be interpreted as the “potential” European crime drop. Due to the flexibility of our modeling framework, it is not needed to make the existence of the crime drop explicit beforehand. We consider three variants of the model, from pooling all parameters over the countries to the possibility of estimating country-specific loadings for the cross-national crime drop. To have an equivocal measure of crime over the countries for the considerable period of interest, we create homicide rates based on the World Health Organization Mortality Database. Our proposed model is able to extract the European crime drop as the underlying time-varying factor of the data. The partially pooled variant of the model is most similar to a two-way fixed effects model. The empirical results from both of these models are aligned. The none pooled variant of the model shows the usefulness of allowing for country-specific crime drop loadings. It is also beneficial to allow for a second stochastic trend for East-European countries. The findings are robust against the inclusion of macroeconomic variables in the model. In an additional explorative analysis, univariate results for the US show that the timing of the European and US crime drops coincide.