We consider which factors determined the price–rent ratio for the housing market in 18 U.S. metropolitan statistical areas (MSAs) and at the national level over the period of 1975–2014. Based on a present-value framework, our proposed empirical model separates the price–rent ratio for a given market into unobserved components related to the expected real rent growth and the expected housing return, but is modified from standard present-value analysis by also including a residual component that captures non-stationary deviations of the price–rent ratio from its present-value level. Estimates for the modified present-value model suggest that the present-value residual (PVR) component is always important and sometimes very large at the national and MSA levels, especially for MSAs that have experienced frequent booms and busts in the housing market. In further analysis, we find that house prices in MSAs that have larger PVR components are more sensitive to mortgage rate changes. These are also the MSAs with less elastic housing supply. Also, comparing our results with a recent statistical test for periodically-collapsing bubbles, we find that MSAs with large estimated PVR components are the same MSAs that test positively for explosive sub-periods in their price–rent ratios, especially during the 2005–2007 subsample. Our approach allows us to estimate the correlation between shocks to expected rent growth, the expected housing return, and the PVR component. We find that the expected housing return and movements in the PVR component are highly positively correlated implying an impact of the expected housing return on house prices that is amplified from what a standard present-value model would imply. Our results also show that most of the variation in the present-value component of the price–rent ratio arises due to the variation in the expected housing return