A spatial multiple treatment/multiple outcome difference-in-differences model with an application to urban rail infrastructure and gentrification

Abstract

We develop a class of difference-in-differences regression models for the case of multiple transportation interventions that may occur sequentially over time and may generate causal spillover effects within a spatial system. We show how these models can be estimated using tools from spatial econometrics, and further extend the models to a system of seemingly unrelated outcomes such that there may be spatial correlation in the error terms. These models facilitate estimation of direct, indirect, and total average causal effects, as well as individual and cumulative effects of transportation interventions that partially overlap in space. Such estimates can assist policymakers in assessing potentially reinforcing effects originating from multiple transportation interventions located in close proximity. We develop an empirical example of our models to evaluate spatiotemporal socioeconomic impacts of the original and expanded light rail system in Denver, CO.