Using Quantile Regression in Hedonic Analysis to Reveal Submarket Competition

A root concern of hedonic property value models is that some commonly-used estimators aggregate very diverse households into a single regression, which may explain the marked differences in implicit price estimates for the same housing attributes across studies. In this paper, we extend a model that captures household heterogeneity through submarket identification and use quantile regression analysis to explore the role submarket competition plays in setting housing prices in those price ranges where different submarkets occupy homes of similar price. We find evidence of direct competition between submarkets with different preferences for at least some homes in a single neighborhood market. We cluster packages of attributes into three broad indices: dwelling structure variables, location variables, and adjacency variables. In the price ranges of competition between two submarkets, there is a clear premium paid in one of the indexed attribute by the final occupant to ‘outbid’ a member from another submarket. The attributes that realize a premium are those that are expected from prior analysis on what those submarkets prefer; and these premiums introduce variation in housing prices that would not be captured by standard hedonic approaches. By examining hedonic parameter instability at different housing price levels, we uncover not only latent diversity among homeowners but direct competition between them that calls into question policy and market conclusions drawn from standard hedonic price models, especially large sample hedonic studies.