NCJ Number
202799
Date Published
April 2002
Length
46 pages
Annotation
This document examines prison population forecasting techniques to estimate future trends on New Mexico's female prison population.
Abstract
Population projections are an important tool in correctional planning and management. While the New Mexico Corrections Department has used standard forecasting techniques with some success for the male prison population, the female prison population has proved to be much more difficult to model. The reasons for this are its relatively small size, its erratic trend over the last decade, and the lack of information on determinants of the prison population. Prison population forecasting techniques can be subdivided into time series analyses that focus on prior time-trends and their extrapolation for future dates, and simulation methods that model the relationship between variables in the criminal justice system. A variety of time series projection methods are described and applied to historical data on the female prison population in order to produce estimates of future trends. Comparison of univariate and multivariate models in terms of projections through December 2003 indicates that the autoregressive integrated moving average (ARIMA) Component Transfer Model provides the best fit, although the projected population is considerably higher than in other models. Comparison of the projections for the period February 1999-August 2000 reveals that all models overestimated the growth in female prison population. The total in August 2000 was 491, but the models estimated a population of between 511 and 611 prisoners. Once again, the ARIMA Component Transfer Model performed well in terms of variance explained, but the full regression ARIMA transfer model performed better on other measures. 2 figures, 12 tables, 22 references