Bayesian Inference in a Structural Model of Family Home Prices

Abstract

We review the implications of an intertemporal representative consumer model for the analysis of housing prices, describing the choice between non-housing and housing consumption, and provide an explanation for the excess return of housing over the riskless rate based on weakly separable preferences. Further considerations are presented regarding the role of liquidity constraints. A Bayesian structural vector autoregression predicts relations between real rent growth, interest rates and housing prices consistently with the representative consumer model. The orthogonalized impulse response functions show, that housing prices are relatively unresponsive to shocks to fundamental value. The logarithmic rent/price ratio increases or does not significantly change following shocks to the real rent growth and relative bill rates. The dynamics of housing prices over the business cycle is mainly determined by financial factors. A shock to the natural logarithm of the rent/price ratio does not have significant predictive properties for subsequent real rent growth and relative bill rates. Moreover, the logarithmic rent/price ratio is a highly persistent variable displaying momentum and long term reversal.