Housing Markets, Models and Indices

Models to predict house prices come in all flavours. In the repeat sales setup, homes that have been sold only once are omitted from the analysis. However, these single sales often constitute a substantial proportion of the data. The proposed repeat sales model eliminates this major weakness. The proposed model is composed of a fixed time effect and a random ZIP (postal) code effect combined with a latent autoregressive component. The latter piece is applied only to homes sold repeatedly while the former two components utilize all of the data. To evaluate the proposed model, single-family home sales for twenty U.S. metropolitan areas from July 1985 through September 2004 are analyzed. The proposed model is shown to have better predictive abilities than the established S&P/Case-Shiller model. Finally, we look at the current housing market crisis in the US and UK compared to a similar meltdown in the late 1990s for Los Angeles, California.