Uncertainty in the Housing Market : A Learning Model and Evidence from Household Survey Data

This paper argues that uncertainty can lead to higher price volatility and persistent estimation errors in the housing market. Empirically, I construct a Household Sentiment Index (HSI) by applying the Case-Shiller repeat-sales estimation method to the households' survey responses in the Panel Studies of Income Dynamics (PSID). I find that compared to the actual housing prices in the same state, the HSI consistently overestimated the growth rate after mid-1990s and exhibited more volatility over the entire period from 1968 to 2011. Theoretically, I show that these facts can result from the households' uncertainty regarding whether a recent change in the market is temporary or permanent. A housing model incorporating this assumption and a Bayesian learning mechanism is able to match the volatility of housing prices. As a proxy for the sentiment index, the one-period forecast from the model shows persistent under-estimation and over-estimation of housing prices, resembling the behavior of the PSID responses