Are housing returns predictable? If so, do households take them into account when making their housing consumption and portfolio decisions? We document the existence of housing return predictability in the U.S. at the aggregate, census region, and state level. We study a portfolio choice model in which housing returns are predictable and adjustment costs must be paid when a house is purchased or sold. We show that two state variables aect the agent's decisions: (i) her wealth-to-housing ratio; and (ii) the time-varying expected growth rate of house prices. The agent buys (sells) her housing assets only when the wealth-to-housing ratio reaches an optimal upper (lower) bound. These bounds are time-varying and depend on the expected growth rate of house prices. Finally, we use household level data from the PSID and SIPP surveys to test and support the model's main implications.
This paper was revised in December 2011.