Housing Market Dynamics and the Implications for the Practice of Differences-in-Differences

Differences-in-differences is among the most popular quasi-experimental methods in urban and housing economics. Because houses are immobile and expensive, we expect buyers and sellers to consider seriously and respond to the externalities from which their house cannot escape. As such, house prices are widely used to assess the impact of policy changes, environmental damages, new infrastructure, and other changes to spatial amenities and disamenities. In many applications, the net impact of these significant contributors to house prices are assumed to be fixed. We argue that they are far from fixed, and that the housing market dynamics that are common within metropolitan areas invalidate a general imposition of fixed effects. This is highly problematic for methods that require fixed effects in order to identify the independent effects of a particular policy or externality. Using data from Maricopa County, Arizona, we document the marked asymmetric evolution of the value of location. Far from being stable, we find that the Maricopa housing market is sufficiently dynamic to undermine any of the “standard” approaches to differences-in-differences methods. We run numerous placebo tests, induced treatments, and several policy applications which allow us to quantify the accuracy of typical methods. The results call into question the differences-in-differences method commonly used in urban and housing economics. Is sum, we find that a house’s simple membership in the same metropolitan area is not a sufficient means of judging whether a house is a good control for use in a quasi-experimental method.