A criminometric study using panel data and latent variables

Abstract

A behavioural model of crime is developed and applied on panel data on the number of crimes and clear-ups for the 53 police districts in Norway for the period 1970-78. The model consists of behavioural relations of the criminals and the police, and of measurement relations allowing for random and systematic errors in the registered crimes and clear-ups. The identification of the model, and the implied testable hypotheses on the covariance structure of the crimes and clearups, depend crucially on the specification of how the distribution of the police district effects of the criminal behaviour and of the police behaviour change over time. The effect of the latent clear-up probability on the crime rate is found to be significantly negative. In our preferred model, a one percent increase in the clearup probability reduces the number of crimes by about one percent. LISREL 7 is applied.