Identification and stochastic specification in Rosen's hedonic price model

The use of an explicitly specified utility function to derive the inverse demand functions in S. Rosen's hedonic price model provides considerable insight into the correct stochastic specification of the model. It turns out that except in special cases, the inverse demand equations are nonlinear in parameters and cannot be formulated conveniently as regression models. Moreover, the inverse demand functions and the hedonic price function must be estimated simultaneously to obtain consistent estimates of the parameters of these functions. A tractable estimation technique is described. It is desirable to derive the inverse demand functions from a utility specification that is not a strongly separable function of houses' attributes since strong separability implies the existence of deterministic relations among incomes, prices, and observed housing attributes that may fail to hold in applications. Finally, it is shown that the use of an explicitly specified utility function does not guarantee identification of the parameters of Rosen's model.