Clustered into Control: Causal Impacts of Water Infrastructure Failure

Abstract

The causal economic impacts of water infrastructure disruptions in OECD countries are largely unknown. Using details of water main break events in Washington, DC, and hourly traffic speeds for 2,182 road segments in a quasi-experimental difference-in-difference design, we estimate the causal effect of main failure on congestion. We use k-means clustering to match treated road segments to control segments. Although precisely estimated, the magnitude of our treatment effects is economically small even when accounting for temporal traffic heterogeneity. Our results suggest that traffic concerns alone are not a justification for policy makers to alter repair strategy for distributed water infrastructure.