presented an investigation of selectivity bias in hedonic price functions estimated from sold subsamples. The empirical findings were sensitive to the degree of skewness allowed by the choice of offer distribution. With the generalized model presented in this paper, the degree of skewness present in the offer distribution is determined empirically, rather than imposed by assumption. This allows a test of the hypothesis advanced in Zuehlke and Rasmussen, that by failing to allow for an upward skew in offers, the normal distribution incorrectly attributes skewness in observed selling prices to censoring.
An important source of price dispersion in housing markets is the heterogeneous nature of housing. The hedonic price function summarizes the price variation that results from variation in housing attributes. Another source of price dispersion in housing markets is informational uncertainty. An interesting implication of the decision rules that arise under informational uncertainty is that selling prices are not unique even among a homogeneous subset of houses. Given the attributes of a house, the expected selling price will vary with variation in seller search costs and risk preferences. Homes owned by sellers who have chosen a relatively low reservation price will have lower expected selling prices and shorter expected market durations. When sampling the housing market, these houses will be overrepresented in the flow of houses exiting the market (the sold subsample) relative to the stock of houses offered for sale.1 Consequently, there is potential self-selection bias in the estimation of hedonic price functions because selling prices are observed only for houses in the sold-subsample. 2 presented an investigation of"sold-subsample" selectivity bias in the estimation of hedonic price functions. The empirical findings were sensitive to the degree of skewness assumed to be present in the underlying offer distribution. The coefficients of the sold subsample hedonic price function exhibited a downward bias in absolute value of approximately 40% when