This paper develops a parsimonious model for predicting utilization of a regional system of psychiatric facilities based on census data. The model combines conceptual simplicity with readily-available data sources. In practical terms, the model removes the need to undertake comprehensive and expensive utilization surveys. In theoretical terms, the model indicates that the fundamental relationship between utilization rates and key social indicators is stable through time. Although further testing is necessary, it would seem that these key indicators can provide reasonable estimates of psychiatric morbidity for up to ten years in the future.