Objective. This article gleans important lessons for environmental justice (EJ) policy from the empirical literature and demonstrates them using EJ analyses of Superfund sites. It constructively critiques the EJ literature and suggests some future directions for designing policy and framing the discourse.

Methods. The research methods of many quantitative EJ studies are reviewed. To demonstrate sensitivity to researchers' choices, logit models are estimated to explain the nationwide distribution of Superfund sites (and deletions from the NPL) using various spatial scales, control variables, and definitions of proximity.

Results. Most empirical EJ studies examine cross-sectional distributions for aggregated demographic measures, dwelling on the implied counterfactual or causal mechanism briefly, if ever. The EJ analysis for Superfund sites exhibits much sensitivity to modeling choices and to whether environmental conditions or policy impacts are modeled.

Conclusions. The findings highlight the limitations of conventional EJ research designs. Several implications for the design of EJ policy are drawn, emphasizing the importance of clarity about counterfactuals and which sorts of impacts are relevant. Principles for a revised EJ policy framework are proposed.