A Note on Hedonic Valuation of Urban Amenities Using Unbalanced Data

Hedonic valuation of urban amenities often requires estimating housing and labor market regressions. It is difficult to get both types of data for all survey respondents. We show that common practice of handling the unbalanced data by conducting two separate regressions can lead to inconsistent covariance matrix estimation and improper inference regarding confidence intervals for amenity values. We then demonstrate how two easily-implementable yet consistent techniques can be used for hedonic valuation with an application in valuing temperature increases in urban Brazil. All techniques estimate a net positive marginal value of a temperature increase. Unlike the separate equation estimation, however, techniques using a consistent covariance estimator are unable to reject at the 5 percent level the null hypothesis of a zero welfare effect