Structural Time Series Modelling of Crime Rates

Doktorarbeit / Dissertation aus dem Jahr 2012 im Fachbereich Statistik, Hawassa University, Sprache: Deutsch, Abstract: The basic goal of this study was to model monthly crime rates and the impact of sentence reforms on reported crime in Addis Ababa, Ethiopia, over the study period of April 2002 to January 2012 GC by incorporating intervention effects. Monthly data obtained from Addis Ababa Police Commission for a total of 118 months were used. The impact of this legislation on reported crime rates was examined using OLS regression model and time series approaches. In particular, structural time series models were employed for the intervention analysis. The best fitted models were selected based on the model capturing the variations in the data accessed through Rs2, AIC and BIC. Comparisons of the selected model were carried out based on ability of prediction via predictive error variance. Both the local level and local linear trend models with inclusion of seasonal and cyclical effects was found to be the best fitted model for modelling of both property and violent crimes, such as murder, aggravated assault, rape, robbery, burglary and larceny. The result of the analysis shows that the coefficient of intervention was negative to rate of murder, robbery, rape and larceny but it was significant only for larceny. The effect of macroeconomic variables such as food inflation, non-food inflation, general inflation and unemployment were statistically significant for aggravated assault. It is recommended that the government and the concerned body should take some measures to tackle inflation and unemployment issues in the city. The rates of crime reported in the violent and property crimes contained seasonality. Thus, prevention mechanisms based on seasonals need to designed.