How complex a model is needed to predict the demand for health services in the United States? To overcome the high cost of obtaining information for sophisticated models, Newhouse estimates a simplified version of a model for predicting the demand for hospital and physician services to ascertain the model's properties. "Simplification" involved making demand a function of only demographic variables, which can readily be obtained, and omitting measures of health status, price, and insurance coverage, which are unavailable. Unfortunately, this simplification is impractical for today's health planning because changes in insurance influence demand too greatly to be omitted from predictive equations, even if data on labor force participation, income, and education are available as explanatory variables. These variables are not satisfactory proxies. Over time, insurance may cease to be as critical in predicting demand, but for the next several years it is likely to remain important.