A Robustness Check of the Spatial Hedonic Difference-in-Differences Methodology

This paper builds on a previous paper that evaluates a spatial hedonic difference-in-differences model to estimate the effects on home prices of a public housing project in North Dallas. Here, we perform a spatial falsification exercise by simulating the spatial GMM estimation on over 5,000 randomly drawn locations satisfying the court's criteria for where public housing projects can be built. The goal of the paper is to test whether using "rings" of different sizes around a location is an appropriate method for defining treatment and control groups. The previously estimated Announcement Effects for the 0.5 mile and 1.75 mile treatment groups are in the 13 percent and two percent tails, respectively, of the simulated distributions. Both effects seem to reinforce the peculiarity of the site in the previous study, thereby suggesting that the methodology and its "ringed" treatment/control groups are appropriate for understanding price dynamics around a location. In this spatially unbalanced sample, however, the presence of a nearby exogenous force (i.e., Frankford Townhomes) may cause an under-estimate because its effect cannot be averaged away. The simulation also reveals that hedonic price models should include variables for living area, land area, baths, pool, and good/bad condition. By contrast, the age of a home and race are not relevant in explaining price changes, on average, for the City of Dallas