Self-destructing prophecies: Long-term forecasting of municipal correctional bed need


Although municipal jails consume a significant amount of resources and the number of inmates housed in such facilities exploded in the 1990s, the literature on forecasting jail populations is sparse. Jail administrators have available discussions on jail crowding and its causes, but do not have ready access to applications of forecasting techniques or practical demonstrations of a jail inmate population forecast. This article argues that the underlying reason for this deficiency is the inherent unpredictability of local long-term correctional population levels. The driving forces behind correctional bed need render local jail population forecasts empirically valid only for a brief time frame. These inherent difficulties include the volatile nature of jail populations and their greater sensitivity when compared with prison populations to local conditions; the gap between the data needed for local correctional population forecasting and what is realistically available to forecasters; the lack of reliable lead variables for long-term local correctional population forecasts; the clash of the mathematics of forecasting and the substantive issues involved in the interpretation of forecast models; and the significant political and policy impacts of forecasts on local criminal justice systems and subsequent correctional population trends.

The differences between the accuracy of short-term versus long-term jail bed need forecasts means that forecasting local correctional bed need is empirically valid for, at best, one to two years. As the temporal cast is extended, longer-term forecasts quickly become error prone. Except for unique situations where jails exist in highly stable local political, social, and criminal justice environments, long-term forecasts of two years or greater are fatally flawed and have little empirical accuracy. Long-term forecasts of local jail bed needs are useful, though, as policy catalysts to encourage policymakers to consider possible long-term impacts of current decisions, but forecasts should be thought of and presented as one possible future scenario rather than a likely reality. Utilizing a demonstration of a local jail forecast based upon two common empirical forecasting approaches, ARIMA and autoregression, this article presents a case study of the inherent difficulties in the long-term forecasting of local jail bed need.


Criminology and Criminal Justice

Document Type




Publication Date


Journal Title

Journal of Criminal Justice