Panel Paper:
Non-Price Spillover Effects of Foreclosure
*Names in bold indicate Presenter
To test for foreclosure spillovers, we examine whether the likelihood of foreclosure depends on the overall risk of foreclosures in each neighborhood based on the incidence of negative equity in each neighborhood. We employ several empirical strategies to isolate the non-price spillover effects of foreclosure. First, we control for local housing prices by including controls for current loan-to-value (LTV) ratio and the negative equity levels of individual housing units in our sample, which we calculate using changes in housing price indices constructed at the census tract and county levels. Second, we construct our neighborhood foreclosure measure using negative equity exposure at the census tract level—a higher level of geographic aggregation than found in prior research—because previous studies examining the spillover effects of foreclosure find that price effects arise at very localized levels of geography. This census tract proxy measure of foreclosure incidence should be largely independent of price spillovers that operate at narrow levels of geography. Third, we address the endogeneity of the tract-level negative equity measures by exploiting the timing of sales within each census tract during the runup to the crisis, creating an expected fraction of homes in negative equity using marketwide LTV ratios since homes that were purchased earlier in the crisis have a higher risk of negative equity independent of the borrowers’ mortgage choices. Further, our models control for the timing of foreclosure using census tract by foreclosure quarter and purchase quarter by foreclosure quarter fixed effects, and control for trends in purchase behavior during the lead up to the crisis by interacting neighborhood observables, such as tract percent poverty, with purchase quarter fixed effects.
Our study finds evidence of neighborhood spillovers without controlling for neighborhood specific trends, but all effects are attenuated away after allowing neighborhood foreclosure patterns to vary based on the timing of purchase and neighborhood observables measured before the housing crisis. The results are robust at both census tract and block group levels and are estimated without any loss of precision. Our analysis shows that there is no spillover effect of foreclosure after controlling for effects that operate through prices and systematic patterns of trends in foreclosure.