Environmental Justice: Evidence from Superfund cleanup durations

•We investigate demographic biases in Superfund cleanups.•We develop an empirical identification strategy based on the cleanup durations.•We estimate a semi-parametric Bayesian duration model with unobserved heterogeneity.•We find evidence of discrimination in the early stages of the Superfund program, which disappears after 1994.•Community involvement plays an important role in reducing the cleanup duration. This paper investigates the extent to which cleanup durations at Superfund sites reflect demographic biases incongruent with the principles of Environmental Justice. We argue that the duration of cleanup, conditional on a large number of site characteristics, should be independent of the race and income profile of the neighborhood in which the site is located. Since the demographic composition of a neighborhood changes during the cleanup process, we explore whether cleanup durations are related to neighborhood demographics recorded at the time when the cleanup is initiated. We estimate a semiparametric Bayesian Proportional Hazard model, which also allows for unobserved site specific heterogeneity, and find that sites located in black, urban and lower educated neighborhoods were discriminated against at the beginning of the program but that the degree of bias diminished over time. Executive Order 12898 of 1994 appears to have re-prioritized resources for the faster cleanup of sites located in less wealthy neighborhoods. We do not find that the litigation process is an impediment in the cleanup process, and support the notion that community involvement plays an important role.