The Long View of Housing Wealth Effects

The response of consumption to house prices, sometimes called a “housing wealth effect,” has become a central object in the narrative of the great recession. Recent research has generally come to the conclusion that the wealth effect in the Great Recession was large: between four and ten cents per dollar of increased housing wealth was consumed within a year on average, and this figure is even higher for high LTV and low credit score borrowers who are more likely to be constrained (Mian, Sufi, and Rao, 2013; Case et al., 2013;, Caroll et al., 2011; Campbell and Cocco, 2007; Attanasio et al., 2009; Berger et al., 2016). One popular narrative is that financial innovation and an expansions of mortgage credit in the early 2000s allowed households to to use their houses as “ATMs,” leading to a much more powerful housing wealth effect than in the 20th Century. This paper evaluates whether a substantial housing wealth effect is a new phenomenon using 40 years of data going back to 1975. We propose a new instrument to evaluate these effects. We show that the large housing wealth effect experienced in the 2000s is not a new phenomenon. Indeed, the housing wealth effect has been large since at least the 1980s. We find our approach yields precise estimates of the elasticity of consumption and employment to house prices over time that are similar to estimates using the Saiz instrument in the 2000s, but that it has distinct advantages, particularly in the earlier period. We use our approach with rolling 5- and 10-year estimation windows to show that the housing wealth effect was low in the 2000s relative to its historic average. Furthermore, we find some speculative evidence of a nonlinearity: the housing wealth effect is stronger in busts. We develop a partial equilibrium heterogenous agents model with long-term mortgage debt and in which housing serves as collateral to evaluate these findings. Our new facts help discipline the model mechanisms that account for the housing wealth effect. In particular, these results cast doubt on the view that financial innovations in the 2000's led to a large increase in the housing wealth effect. Indeed, our model shows why the credit expansion in the 2000s may have reduced the strength of the housing wealth effect. We use our model together with our estimates of the consumption response to housing wealth to infer changes in credit supply throughout the 2000s boom and bust.