Association for Public Policy Analysis & Management

Empirical evidence shows that people of color and low-income communities are more likely to be exposed to actual or potential environmental hazards as well as be subject to inequitable implementation of environmental policy. Although environmental justice issues have aroused a variety of policy and programmatic responses at federal and state levels, studies suggest a limited efficacy. However, research on environmental justice has not yet incorporated the role of public management and dimension of organizational behavior. The primary objective of this study is to explore the links between the representativeness of public institutions, policy outcome, and social justice. Specifically, it examines two questions: (1) Does the representativeness of the EPA administrators promote policy outputs for the environmentally vulnerable communities? (2) Does and to what extent the representativeness of elected officials moderate the effect of bureaucratic representation?

Empirically, we focus on the EPA’s nationwide implementation of the Clean Air Act and Clean Water Act at the state level from 2000 through 2015. We analyze a combined dataset merged from the EPA’s Integrated Compliance Information System – Federal Enforcement and Compliance and Facility Registry Service data, the Census Bureau’s Census 2000, 2005-2009, 2011-2015 American Community Survey 5-year estimates, the U.S. Office of Personnel Management’s Central Personnel Data File, and Congressional Quarterly’s Voting and Elections Collection. The dependent variable is the aggregate informal and formal enforcement actions led by the EPA (block group as level one). In terms of focal explanatory variables, the EPA’s workforce representativeness is measured by the percentage of minority EPA FTEs at the county level (level two). The descriptive and partisan representativeness of elected officials (congressional district as level three) is respectively measured by minority dichotomous variable (1 if minority, 0 if non-minority) and partisanship dichotomous variable (1 if Democrat, 0 if Republican). Client demand (block group level) is measured by the multiplication of potential environmental justice areas (the highest 15th percentile of block groups in terms of percent minority population in a given state) and the number of regulated facilities’ noncompliance incidents. We further include interaction variables between client demand, bureaucratic representativeness, and descriptive and partisan representativeness of elected officials. The control variables at the block group level include the number of regulated facilities, economic class (a factor score based on median household income, education attainment, and poverty rate), unemployment rate, percentage of pollution-intensive industries, percentage of home ownership, percentage of household linguistic isolation, percentage of population under 5, percentage of population at 64 or above, percentage of housing units built prior to 1960, total population, and population density. The control variables at the congressional district level include environmental voting score, committee membership, victory margin, presidential vote, incumbency tenure, ideology, and various district demographics. We use multilevel count data models to estimate the effect of the predictor variables on the outcome variable.