National Patterns in Environmental Injustice and Inequality: Outdoor NO2 Air Pollution in the United States

We describe spatial patterns in environmental injustice and inequality for residential outdoor nitrogen dioxide (NO2) concentrations in the contiguous United States. Our approach employs Census demographic data and a recently published high-resolution dataset of outdoor NO2 concentrations. Nationally, population-weighted mean NO2 concentrations are 4.6 ppb (38%, p<0.01) higher for nonwhites than for whites. The environmental health implications of that concentration disparity are compelling. For example, we estimate that reducing nonwhites’ NO2 concentrations to levels experienced by whites would reduce Ischemic Heart Disease (IHD) mortality by ∼7,000 deaths per year, which is equivalent to 16 million people increasing their physical activity level from inactive (0 hours/week of physical activity) to sufficiently active (>2.5 hours/week of physical activity). Inequality for NO2 concentration is greater than inequality for income (Atkinson Index: 0.11 versus 0.08). Low-income nonwhite young children and elderly people are disproportionately exposed to residential outdoor NO2. Our findings establish a national context for previous work that has documented air pollution environmental injustice and inequality within individual US metropolitan areas and regions. Results given here can aid policy-makers in identifying locations with high environmental injustice and inequality. For example, states with both high injustice and high inequality (top quintile) for outdoor residential NO2 include New York, Michigan, and Wisconsin.


Supporting information for environmental injustice and inequality metrics
Equation S1 presents the calculation of population-weighted NO 2 concentration (C), where i indexes the Block Groups, c i is the mean concentration for each Block Group i; p i is the population of Block Group i; and n is the number of Block Groups. As an example, for calculating the population-weighted NO 2 concentration for urban whites, c i is the mean concentration for each urban Block Group i; p i is the white population of urban Block Group i; and n is the number of urban Block Groups presents a sensitivity analysis on the selection of the Atkinson Index (with inequality aversion parameter, ε = 0.75) as the core environmental inequality metric presented in the main text. This core environmental inequality metric is highly correlated (Pearson's correlation coefficients > |0.96| and Spearman's rank coefficients > |0.98|) with the alternate environmental inequality metrics we considered (Atkinson Indices with ε = {0.25, 0.5, 1, 1.25, 1.5, 2), Gini coefficient, and Gini coefficents on modified and inverse NO 2 datasets) among the 448 urban areas. Thus, the conclusions presented in the main text are not highly sensitive to the core metric selection for environmental inequality.
As a supplement to Figure 2 and Table 3 in the main text, Figure S2 and Table S1 present alternate metrics for environmental injustice (relative percent difference between lowerincome nonwhites and higher-income whites) and inequality (Gini coefficient) for US regions, states, counties and urban areas. Table S2 provides details for the public health impacts (reductions in Ischemic Heart

Supporting information for health impact estimates
Disease mortality) associated with disparities in NO 2 concentration differences observed between nonwhites and whites.

Supporting information for regression models
Tables S3-S18 present linear regression model details for Figure 1 in the main text. The dependent variable in each model is the population-weighted NO 2 concentration for Census householders. The independent variables are income, income-squared, and, for urban models, a dummy variable to control for specific urban area. We developed separate regression models for each of the 4 largest race-ethnicity categories (white, black, hispanic, asian) in 4 location categories (large urban areas, medium urban areas, small urban areas, rural areas), yielding 16 total regression models.
As an alternative analysis to Figure 1 in the main text, Tables S19-S30 present NO 2 regression models for which each observation is a Block Group concentration rather than population-weighted concentration. The dependent variable for each model is the Block Group mean NO 2 concentration. The independent variables are Block Group average income, Block Group average income-squared, and Block Group percent white population. We developed separate regression models for each of the 3 Block Group percent white population tertiles and for each of 4 location categories (large urban areas, medium urban areas, small urban areas, and rural areas), yielding 12 total regression models. Compared to the population-weighted concentration analyses (Figure 1; Tables S3-S18), Block Group analyses indicate a more varied relationship with race and with income, but in general suggest that NO 2 concentration disparities are greater by race (percent white tertile) than by income. Figure S1. Correlations among environmental injustice and inequality metrics (Pearson's correlation coefficient, r; Spearman's rank correlation coefficient, s) for urban areas (n=448). "Atkinson (0.75)" indicates Atkinson Index calculated with the inequality aversion parameter (ε) = 0.75. "Gini (mod.)" indicates the Gini Coefficient calculated on a modified NO 2 dataset in which the BGs with the lowest 10% of NO 2 concentrations in each UA are clipped to the 10th percentile concentration in the UA. "Gini (inverse)" indicates the Gini Coefficient calculated using the inverse of concentration (ppb -1 ) for all BGs. Figure S2. Supplemental environmental injustice and inequality in residential outdoor NO 2 concentrations for US regions, states, counties and urban areas. The left column shows relative difference in population-weighted mean NO 2 concentration between low-income nonwhites and high-income whites, with larger positive differences (red colors) indicating higher injustice (larger relative percent difference between lower-income nonwhites and higher-income whites). The right column shows the Gini Coefficient, with higher values indicating greater inequality.  Counties 3 (n = 3,109) 11% (-52% to 67%) 0.14 (0.0008 to 0.38)

Environmental Justice
Urban Areas (n = 448) 12% (11% to 47%) 0.08 (0.008 to 0.18) 1 Larger Gini Coefficients indicate greater inequality. 2 Larger positive percent differences indicate greater injustice (low-income nonwhites more exposed relative to high-income whites). Negative differences indicate that high-income whites are more exposed relative to low-income nonwhites. 3 This analysis excludes counties that consist of 1 Block Group (n=29; total population = 21,500 people) or contain 0 low-income nonwhites and/or 0 high-income whites (n=16; total population = 65,800 people).      Dallas--Fort Worth--Arlington, TX Urbanized Area  Dallas--Fort Worth--Arlington, TX Urbanized Area