Disproportionate Clean Air Act violations occur in communities of color throughout the United States

The United States (U.S.) Clean Air Act seeks to prevent and abate ambient air pollution, while also providing a framework to identify and address violations. Little research has examined where or how frequently violations of the Clean Air Act occur and how marginalized communities may bear a disproportionate share of these violations, despite the fact that marginalized communities experience persistent, disproportionate pollution levels and associated health impacts. Here, we leverage data on Clean Air Act enforcement and compliance together with demographic data to show that the most serious violations of the Act—high priority violations (HPVs)—predominantly occur in communities of color throughout the U.S. Specifically, we find that the number of facilities with an HPV within communities with the largest proportion of people of color is nearly two times greater than in communities with the smallest proportion. Only 6% of facilities with an HPV address their violations within the timeframe mandated by the U.S. Environmental Protection Agency, and a larger share of facilities with an HPV in disadvantaged communities do not address their violations within this timeframe compared to facilities with an HPV in non-disadvantaged communities. Enforcing agencies should improve how violations are communicated and addressed. To this end, we suggest several ways to empower individuals and communities with easy-to-access data related to Clean Air Act violations and that enforcement practices and reporting be standardized across enforcing agencies.


Introduction
The Clean Air Act aims to protect the public from ambient, or outdoor, air pollution in the United States (U.S.).The Act also establishes a permitting system for emission sources.Most air pollution sources across the country are regulated by state and local air pollution control agencies by way of authority delegated by the U.S. Environmental Protection Agency (EPA).Although violations of the Clean Air Act are governed by EPA enforcement response policy, there are considerable differences regarding enforcement and implementation of EPA's policy among these agencies.
The EPA provides inspection, violation, and enforcement data for over 1 000 000 EPA-regulated facilities via Enforcement and Compliance History Online (ECHO) [1].This database has illuminated disproportionate violations of the Clean Air and Water Acts near U.S. carceral facilities as well as nationwide inequities related to the Clean Water Act and Safe Drinking Water Act [2,3].To our knowledge, no study has systematically investigated violations of the Clean Air Act across the U.S. nor explored the distribution of these violations among population subgroups, despite that marginalized populations experience disproportionate air pollution exposure and associated public health damages [4][5][6][7][8].
The root of these air quality and health inequities has been partially traced back to 'redlining,' a discriminatory mortgage lending practice dating back to the 1930s that effectively denied Black and Brown individuals (among other immigrant communities of the day) from home ownership.One lasting impact of redlining was intergenerational devaluation of property and housing stock in redlined communities [9].These communities were and continue to be targeted by city planners, politicians, and developers for the construction of air pollution emission sources such as highways and heavy industry, becoming 'sacrifice zones' [10][11][12].
Here, we leverage enforcement and compliance data from nearly 15 000 major emission sources across the U.S. and coincident demographics to examine a subset of Clean Air Act violations known as 'high priority violations' and explore the following questions: Where are facilities with a high priority violation (HPV)?Which population subgroups live closest to facilities with an HPV, and how do penalties, stack testing practices, and other types of compliance evaluations vary for these subgroups?Finally, we consider how enforcement, compliance, and communication practices could change to increase community awareness of potentially hazardous facilities.

Data and methods
The ECHO database (downloaded 3 July 2023) contains five years of data on facility inspections and formal enforcement actions and three years of data on noncompliance.ECHO also provides data on whether facilities are in non-attainment areas as well as sociodemographics (e.g.percent people of color and percent below low income) within three miles of each facility.We use this three-mile radius definition as it is commonly used by the EPA not just for ECHO but for other studies and literature, such as the Environmental Justice Screening Report for the Clean Power Plan [13].
We additionally incorporate data on whether census tracts containing facilities included in ECHOare considered 'disadvantaged' by the Climate and Economic Justice Screening Tool (CEJST).At the time of this study, U.S. government agencies including the EPA, Department of Energy, and Department of Health and Human Services were actively using CEJST to aid in allocating funding from the Biden-Harris Administration's Bipartisan Infrastructure Act and Inflation Reduction Act [14].Specifically, the agencies were employing the tool to identify disadvantaged census tracts for funding to fulfill the Administration's Justice40 Initiative, which aims to direct 40% of the benefits of federal climate and clean energy investment to disadvantaged communities.For a census tract to be disadvantaged by CEJST it must be both at or above the 90th percentile for environmental, climate, health, or other burdens and at or above the 65th percentile for low income.Currently, the only air pollutants explicitly included in the tool are fine particulate matter and diesel particulate matter [15].CEJST's unit of analysis for sociodemographic data differs from that of ECHO, but we chose to consider both CEJST's unit-hazard (i.e.census tracts) and ECHO's distance-based (i.e., three-mile radius) methods as they provide commonly-used, complementary methodological approaches [16].We note that the potential health and welfare damages associated with emission sources may not be limited to any one three-mile radius or census tract containing the facility.
Our main analysis considers only facilities classified as major (Title V permit) sources; that is, facilities with actual or potential criteria pollutant emissions at or above 100 tons yr −1 with lower thresholds for facilities in non-attainment areas or if facility emissions are hazardous air pollutants (HAPs) [17].Applying this criterion yielded 14 179 facilities, of which most (88% or 12 471) had no violations identified.Approximately six percent, or 844, of facilities had a violation other than an HPV identified, and another 6%, or 864, had HPVs, which are the focus of this study.We focused on facilities with HPVs specifically because, while the EPA considers all violations important, HPVs are a subset of violations authorized by the Clean Air Act that require additional scrutiny, reporting, and oversight.EPA regulates HPVs through their enforcement response policy [18].
Not all HPVs are associated with emissions.Some HPVs may be more administrative in nature, such as inadequate reporting and recordkeeping, and emissions-related HPVs may be associated higher public health risk potential.The specific causes of HPVs, referred to as 'criteria' by the EPA, provided by ECHO are as follows are triggered by different causes or, using EPA terminology, criteria: criterion 1 (failing to obtain necessary permits or install emissions controls), criteria 2, 3, and 4 (violating emission limitations or standards); criterion 5 (violating work practices, testing and monitoring requirements, or recordkeeping); and criterion 6 (other violations) [18].ECHO provides the names of pollutants or pollutant categories responsible for violations.However, these pollutants or pollutant categories do not directly map onto the EPA criteria.We group and distill these pollutants or categories in different categorical groups: administrative (the aforementioned 'criterion 5'), greenhouse gases (GHGs), HAPs, volatile organic compounds (VOCs), criteria pollutants, and other (table S1).It is possible that administrative violations may be associated with specific pollutant emissions.
We report the strength and significance of relationships in a couple different ways and declare relationships as significant if p < 0.05.For two continuous variables, we report the slope of the linear regression, the p-value (p) calculated using the Wald Test with a t-distribution of the test statistic, and the Pearson correlation coefficient (r).For a continuous variable and a binary outcome, we use point biserial correlation and report r and p.

High priority violations in the U.S
Of the facilities classified as major emission sources, those with an HPV occur in all states besides North Dakota, Rhode Island, and Vermont; however, their geographic distribution is uneven (figure 1).The states of Texas, California, Louisiana, and Indiana have the largest number of facilities identified by ECHO (2763, 1095, 657, and 574, respectively).Given a greater number of facilities in these states and the fact that Texas and California are large in area, we expect they would contain a greater number of facilities with an HPV, and we indeed find this to be the case for California, Texas, and Indiana (200, 132, and 71, respectively).While Michigan ranks eighth among U.S. states for its 359 facilities identified by ECHO, it is in the top four states alongside California, Texas, and Indiana when ranked by the number of facilities with an HPV (51 facilities).Despite Louisiana ranking third among U.S. states for the number of ECHO facilities, we found only 23 facilities with an HPV in this state (ranking seventh).Examining the proportion of facilities with an HPV to the total number of facilities varies from 0%-22% across states, with the largest proportions in the District of Columbia (22%), California (18%), and Hawaii (18%).
According to the EPA, HPVs should be addressed within 6 months, and failure to do so is generally deemed untimely unless a resolution timeline is demonstrated [18].A 2009 report from the EPA Office of Inspector General detailed how the EPA and states often did not address HPVs in a timely manner and that 30% of state-led and 46% of EPA-led HPVs were unaddressed after 270 d [19].Our analysis, using more up-to-date data, reveals that the proportion of HPVs unaddressed after 270 d has increased to 90%.Moreover, only 6% of facilities with an HPV addressed their violations within 6 months, and the median length of time facilities are in HPV is 36 months (the length of ECHO's data record on noncompliance).
Nearly half (44%) of facilities with an HPV occur in the 37% of census tracts designated as disadvantaged by CEJST.Within individual counties containing the greatest number of facilities with an HPV, the distribution of these facilities is heterogeneous (figure 1), and clusters of facilities with an HPV are often located in industrial areas likely overburdened by other pollution sources.The proportion of facilities with an HPV in disadvantaged tracts can be even greater within individual counties compared with the national scale.As examples, 76%, 64%, and 82% of facilities with an HPV in Kern County, CA; Harris County, TX; and Fresno County, CA, respectively, are in disadvantaged tracts.The tracts containing facilities with an HPV represent a different subset of tracts than those represented by other pollution-related indicators within CEJST (e.g. at the national level only 20% of tracts containing facilities with an HPV are disadvantaged for PM 2.5 exposure, 13% for superfund site proximity, and 12% for hazardous waste site proximity).
If we consider the median elapsed time since facilities' most recent compliance evaluation as proxy for oversight at facilities, we find that there are not substantial differences across facilities with an HPV in disadvantaged versus non-disadvantaged tracts (158 and 152 d, respectively).However, the time facilities with an HPV remain in violation varies as a function of CEJST-designated disadvantaged status.Specifically, only 2% of census tracts containing facilities with an HPV for ⩽6 months are disadvantaged.This proportion increases to 6% for facilities with an HPV for ⩽1 year, 13% for ⩽2 years, and 43% for ⩽3 years.

Disparate compliance, enforcement, and impacts
While CEJST does not use race to identify disadvantaged communities, environmental injustice is often the result of racist policies and practices.We therefore explore associations between facilities with an HPV and the race and ethnicity of the nearby population, characterized by the percent people of color of the nearby community.If we bin all facilities into deciles based on the nearby percent people of color to ensure an equal sample size for each bin, the number of facilities with an HPV significantly increases as the percent people of color increases (r = 0.05, p < 0.001; figure 2), and the number of facilities with an HPV within communities with the largest proportion of people of color (10th decile, >78%) is 1.8 times greater than those within communities with the smallest proportion (1st decile, <5%).Put another way, only 4.8% of facilities within communities with the smallest proportion of people of color incur an HPV, while 8.9% of facilities in communities with the largest proportion have such violations.
The time since the last compliance evaluation at facilities with an HPV increases as the proportion of people of color increases, with an additional 4.4 d since last evaluation per 1% increase of people of color (r = 0.17, p < 0.001; figure 2).This result stands in contrast to our previous analysis based on CEJSTdesignated disadvantaged status, which found negligible differences in the time since the last compliance evaluation between facilities with an HPV in disadvantaged versus not disadvantaged census tracts.
While a greater proportion of people of color near facilities with an HPV is associated with increasing time since facilities' last compliance evaluation, the length of time facilities remain in violation does not exhibit the same relationship with community demographics.Facilities with an HPV that address their violations within 6 months have a similar median proportion of people of color (37%) as facilities that remain in violation for up to 1 year (35%), 2 years (34%), or 3 years (38%).
Investigations, partial compliance evaluations, stack tests, and financial penalties are additional means of compliance and enforcement reported by ECHO.For facilities with an HPV there are positive, although generally not significant, relationships between these measures and the percent people of color (figure S1).When taken together, these results present a paradoxical finding that facilities with an HPV in communities of color have fewer compliance evaluations but more overall investigations, stack tests, and partial compliance evaluations and greater penalties.One potential explanation is that a compliance evaluation at a facility in a community of color-although less frequent on average than in majority non-Hispanic, white communities-is more likely to uncover practices or infractions that necessitate further actions such as stack tests or penalties.Also, it should be noted that each state, local, tribal agency, or EPA Region may report and address HPVs differently.Some states do not report a violation until after the case is prosecuted or settled.In some instances, the state attorney general is the lead prosecutor for the agency, not the agency itself, which may result in delays and inconsistencies in reporting times and enforcement actions.In other words, HPVs in communities of color may not be reported in a timely manner until investigations and compliance processes are fully resolved.
The cause of HPVs varies with the ethnoracial composition of the nearby community, and communities with a higher percent people of color tend to have more or administrative-related, or criterion 5, violations; fewer violations related to HAPs and criteria pollutants; and a similar, small number of violations related to VOCs, GHGs, and other causes (figure 2).While HPVs for administrative reasons might seem frivolous, operating without a permit, violating testing requirements, or failing to meet safety standards are serious, and the potential health effects involved with these criterion 5 violations therefore vary from facility to facility, depending on their cause.
Within nonattainment areas where air quality is worse than National Ambient Air Quality Standards, the increase of facilities with an HPV with increasing percentage of people of color is even more stark than for the U.S. as a whole (figure 2).There are over five times more facilities with an HPV in communities with the largest percentage of people of color compared with facilities located in communities with the smallest percentage (10th versus 1st deciles, respectively).Similar to our results for the U.S. as a whole, facilities with an HPV in non-attainment areas are inspected less frequently when they are located in communities of color and tend to receive their HPVs for administrative-related or criterion 5 violations.
The association between communities' ethnoracial composition and HPVs somewhat reflects national-level HPV and demographic patterns and Facilities with high priority violations (HPVs), their causes, and the time since the last facility inspections versus the proportion of people of color in communities in which facilities are located.The number of HPVs (black), the median number of days since the most recent compliance evaluation for facilities with HPVs (orange bars), and the pollutant or pollutant categories associated with the HPVs (pie charts) are presented for deciles based on the percent people of color near (⩽3 miles) facilities.The percent people of color to which the deciles correspond are provided below the pie charts using interval notation.Top, all HPVs in the United States; bottom, only HPVs within the boundaries of at least one defined Nonattainment Pollutant Standard geographic area.
is most pronounced in urban areas, while facilities in individual states, EPA regions, or rural areas do not always display the same significant, positive associations (figure S2).Still, within many limited geographical regions, we generally found more facilities with HPVs in communities of color than in communities with fewer people of color.A few states (e.g.California) and EPA Regions (e.g.Region 9, which is dominated by California given the large number of facilities in the state) are exceptions to this general finding, and in these areas we uncovered a non-statistically significant relationship between the number of facilities with an HPV and communities of color (figure S2).We also find that, on a national basis, communities of color experience fewer of the less-severe violations (as opposed to high priority violations) than communities with a large proportion of white and non-Hispanic persons (figure S3).Facilities with an HPV do not increase for communities with the largest proportion of low-income residents (figure S3), indicating that race, not income, is linked to whether a facility is more likely to have an HPV.

Discussion
The EPA considers information on the causes of HPV beyond the categories shown in table S1 and figure 2 as confidential, so we were unable to further understand the precise causes of HPVs [20].This lack of detailed information on the causes of HPVs precluded us from understanding specifics about the magnitude of emission releases related to violations, and the public could benefit from additional measures to disseminate information on facilities with an HPV given their gravity.Impacted communities may not even realize their proximity to these facilities or have access to information about what violations have occurred nearby and why.
The EPA should explore what information about violations could be made more publicly accessible and then provide that information in formats that are easy to find and access.An example of related progress is the 2022 release of 'ECHO Notify,' a location-based email digest with information on nearby violations and enforcement activity [21].Including information on facilities with an HPV could also be included as indicators in existing environmental justice screening tools such as EJScreen or CEJST, especially since these facilities exhibit different spatial patterns than other pollution-related indicators included in these tools.Finally, the EPA should also consider resurrecting the Watch List, a program which ran from 2004 to 2013 aimed at facilitating dialogue surrounding enforcement matters related to facilities with alleged violations [22].
Regarding our analysis of facilities with an HPV in CEJST-designated census tracts: some of these facilities may be included in emissions inventories used to model the air pollution inputs included in CEJST.Therefore, the presence of a facility with an HPV in a CEJST-designated disadvantaged tract may not necessarily represent excess pollution beyond what is already accounted for in CEJST.However, since particulate matter and diesel particulate matter are the only air quality-related indicators currently included in CEJST, we expect that this potential for 'double counting' would be minimal since facilities' violations can stem from releases of other criteria pollutants beyond particulate matter as well as HAPs or administrative-related infractions (figure 2, table S1).
Our finding that the number of facilities with an HPV does not relate to median household income (figure S3) and only exhibits small disparities on the basis of CEJST-designated disadvantaged status underscores the importance of explicitly considering race and ethnicity when assessing inequities in violations and designing strategies to remedy these inequities.Research on disparities in ambient air pollution concentrations has similarly found that race and ethnicity, rather than income, better explain pollution disparities, stemming from decades of racist policies, regulatory actions, and land-use planning [5,23,24].
Currently, a given agency enforcing EPA policy might have substantially different policies and procedures compared with the EPA or other agencies (e.g.some states refer enforcement cases to their attorneys general; some states cannot close an HPV until there is a final judgment on the case, even if the issue was quickly resolved).The EPA should work with states and local air pollution control agencies to develop a consistent approach to address violations.To accomplish this goal, the EPA would likely need to develop specific regulations that require consistent and detailed reporting from state, local, and tribal agencies.The National Association of Clean Air Agencies is petitioning the EPA to treat states in a consistent manner and aid state and local agencies in enforcement activities [25].A more standardized approach to enforcement might reduce clustering of HPVs in particular states (figure 1) and variations in compliance activities (figure S3).Beyond simply providing additional, consistent information on all facilities with an HPV, future policies of EPA and state, local, and tribal agencies should be cognizant of the race-and ethnicity-based disparities in facilities with an HPV by, for example, providing additional resources for community monitoring and engagement and oversight to violating facilities in communities of color.To this end, the EPA is beginning to incorporate provisions from Title VI of the Civil Rights Act of 1964 into the air quality permitting process [26].This integration should ensure that federal, state, and local air quality permitting programs that have the potential to affect human health or the environment do not directly discriminate on the basis of race, color, or national origin.

Conclusion
Communities of color shoulder a disproportionate share of facilities with an HPV and less frequent inspections compared with facilities in majority non-Hispanic white communities.This unmistakable increase in the number of facilities with HPVs in communities with the largest proportion of people of color is concerning.While institutionalized and systemic racism has placed emission sources in de facto sacrifice zones, compliance and enforcement measures need not follow these patterns, should deter environmental lawbreaking, and create a level playing field for emitters [19].Relevant enforcement agencies must work to ensure that detailed information on Clean Air Act violations are adequately communicated to nearby communities and that equal protection under the law does not hinge on communities' racial or ethnic composition but that the promise of the Clean Air Act extends to all.

Figure 1 .
Figure 1.Facilities with a high priority violation (HPV) in the United States and the top six counties with the greatest number of facilities with an HPV as of 3 July 2023.Scatter represents the locations of facilities with an HPV, colored by the number of months with an HPV for the (left) U.S. and (right) the top six counties.Census tracts designated as 'disadvantaged' by the Climate and Economic Justice Screening Tool (CEJST) are denoted in green.The principal cities of metropolitan or micropolitan statistical areas are shown in grey, and the primary roadways are indicated in black for the county maps.

Figure 2 .
Figure 2.Facilities with high priority violations (HPVs), their causes, and the time since the last facility inspections versus the proportion of people of color in communities in which facilities are located.The number of HPVs (black), the median number of days since the most recent compliance evaluation for facilities with HPVs (orange bars), and the pollutant or pollutant categories associated with the HPVs (pie charts) are presented for deciles based on the percent people of color near (⩽3 miles) facilities.The percent people of color to which the deciles correspond are provided below the pie charts using interval notation.Top, all HPVs in the United States; bottom, only HPVs within the boundaries of at least one defined Nonattainment Pollutant Standard geographic area.