Using physician practice cost functions in payment policy: the problem of endogeneity bias.

Empirical estimates of physician practice cost functions, especially estimates of marginal practice costs and scale economies, could assist in setting physician payment policy. This paper examines the bias that may result in these estimates from the usual approach of treating physician labor input as exogenous. Data come from 207 general surgeons included in the Physician Payment Review Commission's 1988 National Survey of Physicians. The empirical practice cost function is specified as a generalized translog. Results are compared for alternative estimation methods that treat physician labor input as exogenous and as endogenous. The findings suggest that marginal cost estimates which ignore the endogeneity of physician labor are biased downward, while estimates of economies of scale are biased upward. In particular, with physician labor exogenous, statistically significant scale economies are found over a wide range of output levels; with physician labor endogenous, constant returns to scale cannot be rejected. Applying econometric studies of physician practice costs to policy will require attention to important methodological issues and collection of high-quality data.