A1374
Title: Dynamic panel data models for the potential European crime drop
Authors: Ilka van de Werve - VU Amsterdam (Netherlands) [presenting]
Siem Jan Koopman - VU Amsterdam (The Netherlands)
Abstract: A panel data model is formulated with time-varying trends to empirically verify the possible existence of the European crime drop. The stochastically time-varying component represents the cross-national crime trend, and each country has its own weight on it. By representing the model in state space form, a likelihood-based approach is proposed using Kalman filtering to estimate the parameters and extract the unobserved time-varying component. In the second step, a cluster analysis of the country's fixed effects and weights is used to show (dis)similarities across the European countries. Several representations of the model are compared to the two-way fixed effects model and dynamic factor model. An empirical illustration of the presence of a potential European crime drop in comparison with the US crime drop shows the benefits of the proposed model and estimation methodology.
