Forecasting Prison Populations - Where to Begin? (From National Workshop on Prison Population Forecasting, P 23-43, 1982, Charles M Friel, ed. - See NCJ-85289)

• NCJ Number: 85291
• Author(s): C M Friel
• Date Published: 1982
• Length: 21 pages
• Sale Source: SEARCH Group Inc.; National Institute of Justice/
• Language: English
• Country: United States of America
• Note: Available on microfiche as NCJ-85289

Annotation

Guidelines for correctional forecasting divide the process into five major steps: getting things organized, predicting admissions, predicting time served, sensitivity analysis, and ongoing forecasting.

Abstract

The first phase is gathering historical data on persons admitted to prison, the total population at different points in time, and the number of releases. The forecaster must define these categories, be aware of changes in the definitions over time, and analyze fluctuations in the historical record. The accessibility and quality of current and historical inmate records should be examined, along with criminal justice system, demographic, and economic trends in the outside world. Future admissions can be predicted in two ways: (1) linear or curvilinear extrapolation of historical admissions data and (2) using a lead indicator, such as unemployment or changes in the crime rate. Because the search for a single lead indicator is often unsuccessful, forecasters commonly use multiple linear regression including two or three indicators. Time served can be conceptualized as the likely time that future prisoners will remain in prison or the number of prisoners likely to be released at some time in the future. Forecasters can reasonably estimate the total population if they know the number of people to be admitted in the future and the likely time served. This interrelationship can be used to validate a forecasting model. Several States have developed interesting and relatively transferable models for calculating time served, including California, Colorado, Minnesota, and Florida. A forecasting model can be validated on the historical data, although this may not be wise in a dynamic correctional environment. Another strategy is developing multiple forecasts based on different sets of assumptions. Forecasting is an ongoing activity, since the likelihood of error increases as time passes and the prison population turns over. Errors in the model, however, provide the forecaster with the unique opportunity to analyze the dynamics of the prison population.