Housing market segmentation and hedonic prediction accuracy

In an earlier paper, Goodman and Thibodeau [Journal of Housing Economics 7 (1998) 121] examined housing market segmentation within metropolitan Dallas using hierarchical models (Hierarchical Linear Models: Applications and Data Analysis Methods, Sage, Newbury Park, 1992) and single-family property transactions over the 1995:1 - 1997:1 periods. Their preliminary results suggested that hierarchical models provide a useful framework for delineating housing submarket boundaries and that the metropolitan Dallas housing market is segmented by the quality of public education (as measured by student performance on standardized tests). This paper examines whether delineating submarkets in the manner proposed by Goodman and Thibodeau improves hedonic estimates of property value. We include two additional housing submarket constructions in our evaluation: one using census tracts and one using zip code districts. Using data for 28,000 single-family transactions for the 1995:1 - 1997:1 period, we estimate hedonic house price equations for most of Dallas County as well as individually for each submarket. The parameters of the hedonic house price equations are estimated using a 90% random sample of transactions. The remaining observations are used to evaluate the prediction accuracy of the alternative housing submarket constructions. The empirical results indicate spatial disaggregation yields significant gains in hedonic prediction accuracy. (C) 2003 Published by Elsevier Inc.