Crime and Arrest Rate Predictions for the States of Alaska and Oregon - Box-Jenkins Monthly Forecasts

• NCJ Number: 92959
• Author(s): K Johnson; R McCleary; J Angell; J Eidson
• Date Published: Unknown
• Length: 215 pages
• Grant Number(s): 82-BJ-CX-K420
• Sponsoring Agency: Bureau of Justice Statistics (BJS)
• Corporate Author: University of Alaska
• Sale Source: National Institute of Justice/; NCJRS Photocopy Services
• Language: English
• Country: United States of America

Annotation

This technical report details the statistical method used in a crime and arrest forecasting study for Alaska and Oregon, along with monthly forecasts and other relevant statistical data for readers interested in replicating the Box-Jenkins procedures used in the analysis.

Abstract

The document first explains the three integrative analytic measures involved in a Box-Jenkins or ARIMA (Autoregressive Integrated Moving Average) time series: identification, estimation, and diagnosis. If a tentative model satisfies all diagnostic criteria, it is accepted as the final forecasting model for a time series. However, metadiagnosis procedures must be performed. In this study, this consisted of checking each series for a 'best' Box-Cox transformation and adding extra parameters to the tentative model. The final ARIMA forecasting models for the Alaska and Oregon time series consisted of three conceptually distinct components: seasonality, trend, and error or noise. The report presents models and monthly forecasts for violent and nonviolent or property crime for 96 Alaska and Oregon crime and arrest time series. Crime forecasts for Alaska cover statewide totals, Anchorage, Fairbanks, Juneau, and the rest of the State. Forecasts for Oregon are given for the entire State and seven regions. Forecasts also are broken down by three categories: over 18, under 18, and nonwhite.