Identification and Estimation of Hedonic Models

This paper considers the identification and estimation of hedonic models. We establish that in an additive version of the hedonic model, technology and preferences are generically identified up to affine transformations from data on demand and supply in a single hedonic market. For a very general parametric structure, preferences and technology are fully identified. This is true under a strong assumption of statistical independence of the error term. It is also true under the weaker assumption of mean independence of the error term. Much of the confusion in the empirical literature that claims that hedonic models estimated on data from a single market are fundamentally underidentified is based on linearizations that do not use all of the information in the model. The exact economic model that justifies widely used linear approximations has strange properties so the approximation is doubly poor. A semiparametric estimation method is proposed that is valid when a statistical independence assumption is valid. Alternatively, under the weaker condition of mean independence instrumental variables estimators can be applied to identify technology and preference parameters from a single market. They are justified by nonlinearities that are generic features of equilibrium in hedonic models.