Projecting crime rates: An age,period, and cohort model using ARIMA techniques

The object of this paper is to generate forecasts of U.S. index property and violent crimes. A three-step procedure is employed. First, the age-specific crime rates are decomposed into age, period, and cohort effects. Then the period and cohort effects are characterized using regression and transfer functions, respectively. Finally, these statistical models are used to generate forecasts of the period and cohort effects and hence forecasted values of the age-specific crime rates. The forecasts are compared with existing data and found to be quite accurate. Issues concerning the forecast efficiency of this approach compared to that of others are also discussed.