Topic: Contributed on Spatial statistics and econometrics Title: Crime forecasting with spatial panel models for counts Authors:  Stephanie Glaser - University of Hohenheim (Germany) [presenting]
Robert Jung - University of Hohenheim (Germany)
Abstract: Crime forecasting has gained increasing interest in the last years but only seldom spatial models are used to exploit the underlying geographic processes in the data for estimation and forecasting. Also, the count nature of the data is often ignored. We apply methods of spatial econometrics to overcome these shortcomings. In particular we propose an additive model with serial and spatial autoregressive terms as well as fixed effects. The model is estimated using system GMM. For forecasting purposes we obtain point forecasts as well as data coherent density forecasts using a Poisson distribution. We apply the model to monthly crime counts at census tract level from Pittsburgh.