Summary

Search models rationalize housing market dynamics where perfect asset models fail to, by incorporating the time and cost of identifying attractive opportunities to trade. They have been used to consider housing market responses to external shocks, such as widespread foreclosure, transaction taxes, seasonal mobility, improved information on listed homes, and endogenous cyclicality. They provide a framework for time to trade, list prices, and trading volume, which have no role in standard frictionless models. Empirically, time to trade offers an intensive margin as an alternative to studying the extensive margin of trade.

Search models are composed of a number of common components: typically, a meeting function, specifying how many buyers visit a seller as a function of the number of buyers and sellers in the market; a match quality distribution; a sale mechanism that determines when trade occurs and, if so, at what price; and a specification of the number of households entering the market as buyers, sellers, or both. The precise form of each of these components will vary according to the research focus, with those less central to the issue at hand being simplified if necessary.

Usually, the researcher has more information on seller search than buyer search, as the attributes of the home are listed and those of the buyer household are not, because coding property attributes is much easier than coding household attributes. However, the internet now gives researchers access to digital traces left by searching buyers. Common empirical difficulties include incomplete information on home quality and missing information on visits that do not end in trade and rejected offers. Difficulties in separating out state-dependence from seller heterogeneity, whether from impatience, search cost, or some other factor, impede the empirical analysis of stigmatization of long-searching sellers, stock-flow matching, finite horizons, and learning.