I show that when house prices are high relative to rents (that is, when the rent-price ratio is low) changes in real rents tend to be larger than usual and changes in real prices tend to be smaller than usual. Standard error-correction models provide inconclusive results about the predictive power of the rent-price ratio at a quarterly frequency. I use a long-horizon regression approach to show that the rent-price ratio helps predict changes in real rents and real prices over three-year periods. This result withstands the inclusion of a measure of the user cost of capital. I show that a long- horizon regression approach can yield biased estimates of the degree of error correction if prices have a unit root but do not follow a random walk. I construct bootstrap distributions to conduct appropriate inference in the presence of this bias. The results lend empirical support to the view that the rent-price ratio is an indicator of valuation in the housing market.