Air quality related equity implications of U.S. decarbonization policy

Greenhouse gas (GHG) emissions are often co-emitted with air pollutants that form fine particulate matter (PM2.5), which causes substantial premature mortality in the U.S. and disproportionately harms racial/ethnic minorities and low-income populations. While climate policy can yield air quality co-benefits, the distribution of these benefits and their implications for exposure disparities remain unclear. The U.S. federal government has articulated goals to direct 40% of the benefits of certain federal investments to disadvantaged communities, raising questions about how national decarbonization strategies might affect pollution inequities. Despite overall air quality improvements, disparities by race/ethnicity persist across income groups and reflect systemic environmental racism and historical discriminatory practices. This study asks whether and to what extent U.S. national CO2 policy of near-term ambition (40–60% reductions by 2030 relative to 2005) can mitigate racial/ethnic disparities in PM2.5 exposure at national, state, and urban scales, and whether alternative distributions of CO2 reductions across sectors and regions could improve equity outcomes.

Prior work shows that climate and clean energy policies can deliver substantial air quality and health co-benefits, sometimes exceeding direct climate benefits and costs. However, evidence on equity impacts is mixed. Retrospective analyses of California’s cap-and-trade have found increased emissions near some facilities located in communities with higher shares of racial/ethnic minorities and low-income residents, while other studies report limited or mitigating effects on disparities when accounting for pollutant transport and secondary PM2.5 formation. Studies of wind power deployment show heterogeneous effects on disparities across states. Prospective studies in California, Texas, and nationally indicate that deep decarbonization scenarios can reduce exposures but may not uniformly reduce racial/ethnic disparities; sectoral focus matters, with transportation emissions reductions offering higher potential to narrow racial inequities in some analyses. Reports have also proposed scenarios that target greater air pollution damage reductions among Black, Hispanic, and low-income populations. Overall, literature suggests that least-cost CO2 reductions may not align with maximal equity gains, and that sector- and location-specific strategies are important for addressing disparities.

Study design: The analysis evaluates national-scale decarbonization scenarios for 2030 achieving 40–60% CO2 reductions relative to 2005 and quantifies PM2.5 exposure and racial/ethnic disparities. A primary policy case represents a national CO2 cap-and-trade achieving −50% by 2030; a 2030 baseline without the policy and a 2017 historical year provide references. Additional optimization and sensitivity scenarios explore alternative spatial/sectoral distributions of CO2 reductions and associated equity outcomes. Energy-economic modeling: The USREP-REEDS framework combines MIT’s U.S. Regional Energy Policy (USREP) computable general equilibrium model (30 regions) with NREL’s Regional Energy Deployment System (ReEDS) electricity capacity expansion model (134 balancing areas). The Baseline aligns with EIA AEO 2020 reference, NREL ATB 2019 mid-range technology assumptions, updated state policies, and COVID-19 adjustments. The Cap 50% scenario imposes a national CO2 cap-and-trade (allowance price rising to $99 in 2030), covers energy and industry CO2, allows national trading, and allocates allowances to states per capita with per-capita household rebates. Emissions inventory and projections: Base emissions draw from US EPA NEI 2017 (point and area sources) with spatial allocation to InMAP grid cells and effective stack height (ESH) layers. Area source spatial allocations use NEI 2014 surrogates when 2017 spatial distributions are unavailable. Biogenic and wildfire emissions are held constant (2005). For 2030 scenarios, 2017 emissions are scaled by USREP outputs across 29 regions using 20 mapping variables (545 combinations) while assuming fixed non-CO2 emission factors at 2017 levels. Power sector CO2, SO2, and NOx emissions for coal and gas are scaled to ReEDS balancing areas; total CO2 is reconciled to USREP by sector (electricity, transportation, industry, residential/commercial). Air quality modeling and exposure: Annual average PM2.5 (primary and secondary components) is simulated using the InMAP Source-Receptor Matrix (ISRM; ~52,000 grid cells; variable 1–48 km resolution; three ESH layers). ISRM translates emissions changes in each grid cell and height layer into concentration changes for primary PM2.5, sulfate, nitrate, ammonium, and SOA. Population exposures are computed using ACS block-group racial/ethnic counts (2012 5-year) allocated to grid cells, scaled to 2030 with state-level total population growth rates from UVA projections. Racial/ethnic groups include non-Hispanic white; racial/ethnic minorities (all except non-Hispanic white); and non-Hispanic Black, Asian, and Hispanic (all races) groups. Disparity metric is the percentage difference between group-average exposure and total population-average exposure. Uncertainty analysis: A sensitivity scenario varies the spatial allocation of point-source reductions within each USREP-ReEDS region-sector while holding total regional-sector reductions fixed, yielding upper/lower bounds on exposure and disparity changes relative to Baseline. Optimization of CO2 reduction distributions: To test equity-improving allocations under fixed CO2 targets, linear programs minimize PM2.5-attributable mortality for racial/ethnic minorities by scaling emissions (0≤Si≤1) at eligible sources (those with PM2.5-related mortality and nonzero CO2 in 2030 Baseline), subject to meeting national CO2 reduction targets and scenario-specific constraints: State-sector (fixed state-by-sector totals), State-total (fixed state totals, sectoral reallocation allowed), Nation-sector (fixed national sector totals, geographic reallocation allowed), Nation-total (national cap only). Additional Nation-total runs target −40% and −60% CO2. Marginal mortalities per ton by pollutant and grid cell are derived from ISRM with Krewski et al. concentration-response and all-cause mortality rates. A set of 5000 randomly sampled non-optimal −50% scenarios illustrates the feasible outcome space; a worst-case scenario maximizes minority mortality under −50%.

- Under the 50% CO2 reduction policy (Cap 50%) relative to the 2030 Baseline, national population-weighted PM2.5 decreases by 0.37 µg/m3 on average (range across counties −1.97 to +0.44 µg/m3). Relative to 2017, average exposure increases by 0.20 µg/m3 due to higher activity levels in 2030. - Sectoral drivers: Electricity sector accounts for 77% of CO2 reductions, transportation 10%, industry 7%, residential/commercial 6%. Near-elimination of coal power and reductions in other combustion reduce SO2 and NOx by 49% and 16% relative to Baseline; primary PM2.5 −7%, NH3 −1%, VOC −5% relative to Baseline (some pollutants exceed 2017 levels). - Contribution to exposure reductions by sector (relative to Baseline): coal electricity −0.16 µg/m3 (about half of total reduction), transportation −0.06, residential −0.06, industrial −0.05, non-coal electricity −0.02, food/agriculture −0.01. - Group-specific exposure changes (2030 Cap 50% vs Baseline): Black −0.44 µg/m3; white −0.37; Asian −0.33; Hispanic −0.32; racial/ethnic minorities overall −0.36; total population −0.37. - Disparities: Relative disparity for racial/ethnic minorities increases from 12.4% to 13.1% (vs total population). White population’s relative advantage increases (from −7.3% to −7.7%). Black disparity slightly decreases (from 17.9% to 17.8%). Asian and Hispanic disparities increase (Asian 9.1% to 10.1%; Hispanic 12.1% to 13.3%). Thus, while exposures decline for all groups, disparities persist or widen slightly at national scale. - Spatial heterogeneity: State- and urban-level impacts vary widely; policy narrows disparities in some states and widens in others. Within the 20 largest urban areas, within-city disparities generally worsen slightly, but regional-scale effects can offset these in aggregate. - Uncertainty in point-source spatial distribution (policy case): Results are robust; for most groups changes are limited. For Black populations, relative disparity may increase or decrease depending on allocation, but changes are small (~0.5 percentage points relative to ~18.4%). - Optimization results (−50% CO2 target): Alternative distributions can further reduce racial/ethnic minority exposures and modestly reduce disparities. Allowing reallocation within states across sectors (State-total) can reduce disparity by about 0.34 percentage points; allowing nation-wide reallocation (Nation-total) achieves an additional ~1.67 percentage points. The largest additional gains arise from reallocating reductions in the transportation sector. Even the best-case among all feasible scenarios reduces the disparity gap by only about 2.7 percentage points (baseline gap ~12%). - Stringency sensitivity: A −60% target further reduces exposures for minorities and the total population but yields a smaller change in disparity; a −40% target produces a greater percentage-point change in disparity but smaller overall PM2.5 exposure benefits.

The study addresses whether near-term U.S. national CO2 policies can also mitigate racial/ethnic PM2.5 exposure disparities. A cap-and-trade policy achieving a 50% reduction by 2030 reduces PM2.5 exposures for all groups but does not substantially reduce relative disparities nationally; some disparities widen slightly. This outcome is driven by the focus of least-cost decarbonization on coal-fired electricity, which, while beneficial, contributes a limited share of overall population exposure and does not primarily target sectors (e.g., industry, heavy-duty diesel transportation) most responsible for current disparities. Optimization analyses show that even with targeted reallocation of the same aggregate CO2 reductions, only modest disparity reductions are achievable, largely by prioritizing transportation emission reductions. Increasing stringency to 60% lowers exposures further but does not close disparity gaps due to parallel benefits for the broader population; less stringent 40% reductions yield larger changes in disparity metrics but smaller overall health gains. These results imply that CO2-focused strategies alone, within the 2030 timeline, are insufficient to resolve entrenched exposure disparities arising from spatially and sectorally specific sources. Complementary, targeted air quality policies are needed to address inequities while pursuing decarbonization.

This work demonstrates that U.S. federal decarbonization policies achieving 40–60% CO2 reductions by 2030 can deliver significant PM2.5 exposure reductions for all racial/ethnic groups but are unlikely to substantially mitigate national-scale exposure disparities. Least-cost carbon reductions target coal electricity, yielding broad benefits but limited progress on equity. Even optimally redistributing reductions under fixed CO2 caps yields only modest disparity improvements (best-case ≈2.7 percentage points reduction of a ~12% gap). Achieving equity goals will require policies beyond CO2 reductions alone, including targeted interventions to reduce direct PM2.5 and precursors from disparity-driving sources (e.g., industrial facilities, heavy-duty transportation), community-focused measures, and potentially more transformative structural changes that phase out fossil fuel use over longer time horizons. Future research should refine sector- and location-specific strategies, integrate dynamic emissions factor changes, evaluate policy bundles (e.g., IRA implementation with equity-focused measures), and assess local-scale impacts with higher-resolution exposure and health data.

- Emissions projections assume fixed 2017 non-CO2 emission factors per unit activity; real-world technological and regulatory changes could alter emission rates by 2030. - Area source spatial allocations rely on 2014 spatial surrogates; biogenic and wildfire emissions are held constant (2005), which may not reflect 2030 conditions. - GHGRP-derived CO2 emissions cover ~85–90% of U.S. GHGs; smaller sources may be underrepresented. - Air quality modeling uses a reduced-complexity model (InMAP/ISRM) with largely linear chemistry and annual-average meteorology; while validated for large-scale analyses, it cannot capture all nonlinearities or temporal dynamics. - Population scaling to 2030 applies state-level total population growth uniformly across racial/ethnic groups and locations, not accounting for differential demographic or migration patterns. - Optimization minimizes minority mortality using national average concentration-response and mortality rates; results may be sensitive to alternative health functions or subgroup-specific baselines. - Policy representation focuses on an idealized national cap-and-trade and broad sectoral reductions; specific program design features (e.g., leakage, offsets, compliance flexibility) and the detailed implementation of the Inflation Reduction Act are not explicitly modeled. - Results emphasize national-scale averages; local co-pollutant mixtures and microenvironment exposures are not resolved.