Normative Forecasting - An Application of Stochastic Models to Crime Rates

• NCJ Number: 75947
• Journal: International Journal of Comparative and Applied Criminal Justice Volume: 4 Issue: 2 Dated: (Winter 1980) Pages: 139-145
• Author(s): S J Deutsch; P E Pfeifer
• Date Published: 1980
• Length: 7 pages
• Grant Number(s): 75-NI-99-0091; 78-NI-AX-0040
• Agencies: NIJ-Sponsored

Annotation

In this study, a procedure for normative forecasting is developed which incorporates empirical stochastic models and their forecasts along with a scaling procedure to evaluate the relatively intensity of the crime problem in various cities.

Abstract

Normative forecasting is a procedure designed to measure future intensity within a given region compared to the future level within a larger region - the population as a whole. Such forecasting is needed to choose the location for which a crime program will have the greatest potential impact in view of limited resources. The monthly totals for a given crime over several comparable cities are used here as a norm for comparing the future forecasts of that crime in a given city. Assuming a representative set of cities, this total forms a time series characteristic of the history of that index crime in the country as a whole. A Box-Jenkins model is identified, estimated, and diagnostically checked. From the resulting model, forecasted values are generated to predict the future pattern of the crime's occurrence. These future forecasts represent the expected aggregate number of occurrences of a given crime. However, before a direct comparison can be made between an individual city's future forecast and the forecasts from the model for the total of all cities, some scaling scheme must be implemented. To this end, scaling the entire total crime series, both past history and future forecasts by the current percentage of total crimes observed in the city of interest is done. Thus, the total of a given index crime over all cities is used to represent the average behavior of crime occurrences in major US cities, and comparing an individual city's crime forecast to the scaled norm indicates how one might expect that city's crime picture to change relative not only to crime in the entire population set, but also to that city's current position. Justification of the model, an example using the index crime of robbery in Los Angeles relative to the parallel occurrences of robbery in other metropolitan areas, statistical data, and references are included.