Race, *class, and human ecological factors in the spatial distribution of manufacturing emissions

The environmental justice movement has prompted social scientists to investigate whether disadvantaged residents experience disproportionate exposure to environmental hazards in the United States. Evidence is mounting that disparities do exist, though there is not yet a clear picture of how various environmental risks are distributed. The research presented here tests for the existence of environmental inequity using national-level data on toxic emissions from manufacturing facilities from the EPA's Toxics Release Inventory dataset for 1993. Several methodological improvements are made over previous research designs. These include an improved estimation of residents' exposure to air toxins through modeling of atmospheric dispersion; elimination of the problem of sample truncation; investigation of aggregation bias through the use of parallel analyses at two geographic levels of analysis, the county and the census tract; analysis of cross-level effects; and estimation of spatial autocorrelation and spatial dependence.;Overall, these results confirm environmental justice activists' claims that Blacks, Asians, Hispanics, and communities with lower average income, are all exposed to higher levels of industrial air pollution than are middle and upper class White Americans. Regression analyses indicate that Blacks, Asians, and Hispanics are exposed to higher concentrations of airborne toxins from manufacturing facilities, as are communities with lower median household income. The census tract level data do appear to more precisely estimate the levels of industrial air pollution as well as residential characteristics. However, with a few important exceptions, substantive findings from the county and census tract data are comparable and this research finds no evidence that the county is an inappropriate unit of analysis for environmental justice research. These results challenge the suggestion that previous findings of environmental inequity are the result of an ecological fallacy. Investigation into cross-level effects reveals that the use of larger aggregates (counties) does lead to inflation of bivariate regression coefficients. Measures of spatial autocorrelation reveal moderate levels of spatial dependence with regard to industrial air pollution. Spatial regression analysis reveals that the spatial grouping of counties into larger regions has an effect on the county-level estimates of some regression coefficients.