Exposure Disparities by Income, Race and Ethnicity, and Historic Redlining Grade in the Greater Seattle Area for Ultrafine Particles and Other Air Pollutants

Long-term exposure to air pollution is associated with adverse health effects, such as cardiovascular and respiratory diseases and cognitive decline.^1–3 Historically, racialized groups and low-income communities have been more likely to be exposed to higher concentrations of air pollution, contributing to health disparities.^4–13 Many different factors can affect health outcomes, and it has been well documented that life expectancy, morbidity, and mortality in the United States varies based on income level, race and ethnicity, and residential location.^14–19

Disparities in environmental health in the United States are a result of structural racism, including ongoing and historic segregation and urban planning decisions. For example, starting in the 1930s, the Home Owners’ Loan Corporation (HOLC) graded residential areas by “mortgage security” on an A-to-D scale: “A” indicated the lowest financial risk to lenders that borrowers would default on mortgage loans and “D” indicated highest risk to lenders. Several factors, such as closeness to transportation, amenities, or industry presence, also contributed to grading.²⁰ Racial, ethnic, and religious composition were significant contributing factors to a neighborhood’s grade, allowing lenders to reduce the accessibility of mortgage financing for people of color and immigrants.²¹ The HOLC’s maps depicted Grade D in a red color, and this grading practice became known as redlining. HOLC grades were delineated decades ago, yet they correlate with present-day social and environmental attributes of neighborhoods.^22–25 In addition, “racial covenants” on homes and neighborhoods established and enforced “White only” communities, resulting in further racial residential segregation.²⁶ Furthermore, zoning regulation, the barriers to homeownership faced by Black people as a result of discriminatory lending practices, the continual lack of enforcement by government officials in the face of discriminatory practices, and placement of polluting activities such as highways and industry, into communities of color all reinforced residential segregation, which resulted in higher exposure to environmental hazards among these communities.²⁶ Historically, and still today, racialized and low-income communities generally have fewer health-promoting resources to adequately address the health effects of air pollution, including less access to amenities such as greenspaces and health care facilities, maintaining the health disparities created by structural racism.²⁷ When racism or classism impede the right to a healthy living environment, including the air we breathe, this is an environmental injustice.

Ultrafine particles (UFPs) are particles suspended in air with aerodynamic diameters of $<100\text{\hspace{0.17em}nm}$. As a result of their small size, UFPs can pass through the blood–air barrier in the lungs that transfers oxygen to blood.²⁸ Growing evidence suggests that UFPs can also cross the blood–brain barrier, which serves to protect the brain from toxins and pathogens.²⁹ UFPs are composed of toxic metals and compounds that have been associated with adverse heart, lung, and brain health effects.²⁸ UFPs are emitted by transportation and other sources of fossil fuel combustion.^28,30 In most ambient urban environments, UFPs dominate particle numbers but weigh very little and thus contribute nearly zero to the regulatory mass-based measurements for fine particulate matter [PM $\le 2.5\mathrm{\mu }\mathrm{m}$ in aerodynamic diameter (${\mathrm{PM}}_{2.5}$)].³⁰ This is why UFPs are often measured as particle number count (PNC).

Limited existing research has explored UFP exposure disparities by race and ethnicity, income, or redlining grade. Chambliss et al. compared nitric oxide (NO), nitrogen dioxide (${\mathrm{NO}}_2$), black carbon (BC), and UFP exposure disparities between different races and ethnicities in the San Francisco Bay area.⁵ The authors measured pollutants by mobile monitoring, and they found median concentrations of NO, ${\mathrm{NO}}_2$, and UFPs to be higher for Hispanic and Black populations and lower for White populations in their study region. Saha et al. studied UFP disparities by race and ethnicity and income across the United States, and found 35% higher than average concentrations for Asian, Black, and Hispanic populations.⁶ Other observational studies on air pollution exposure disparities by race, ethnicity, and socioeconomic status have focused on criteria pollutants, such as ${\mathrm{NO}}_2$, ${\mathrm{PM}}_{2.5}$, and ozone (${\mathrm{O}}_3$) across the United States; most have found higher pollution levels in low-income and racialized communities.^7–13 Previous studies have also examined the association of redlining grade with BC, ${\mathrm{NO}}_2$, and ${\mathrm{PM}}_{2.5}$, and found that pollutant concentrations increased with less desirable HOLC grades.^31–34 Other studies examining the health or environmental effects of redlining have found an association with asthma, lack of health insurance, and reduced present-day greenspace with more hazardous HOLC grades.^24,35–37

Through the Clean Air Act,³⁸ the U.S. Environmental Protection Agency (EPA) sets national standards for six criteria air pollutants, including ${\mathrm{O}}_3$, ${\mathrm{NO}}_2$, and PM, all of which have been shown to harm health. These pollutants have been monitored continuously at $>\mathrm{4,000}$ monitoring stations operated by states and local agencies throughout the country.³⁹ Although some states choose to monitor UFPs, there is a lack of long-term and widespread UFP monitoring.⁴⁰ At this time, there is no standard metric of measurement for UFPs. PNC is the most commonly used metric, but concentrations vary owing to differences across instruments in the range of the particle sizes they measure and in their efficiency in counting particles.⁴⁰ Few studies have examined disparities of UFPs and other noncriteria air pollutants, such as BC, despite increasing evidence that these pollutants are also associated with adverse health effects.^41–43 Similarly, to the best of our knowledge, no studies have considered disparities associated with redlining grade and UFPs.

The primary aim of this paper was to examine how contemporary income, race and ethnicity, and historic redlining grade impact present-day UFP exposure disparities in the Seattle, Washington, area. We compared UFP exposures between income groups, racial and ethnic groups, and redlining grade to see how much pollutant concentrations must decrease to make groups with the highest exposures comparable to those with the lowest. The disparities by redlining grade show the lasting effects of discriminatory urban planning practices, whereas the disparities by present-day income and race and ethnicity reflect both the ongoing impacts of racist policies and the modern populations that are disproportionately affected. Further, in descriptive analyses, we estimated magnitudes of difference in pollutants between HOLC Grade A and more hazardous grades and examined how accounting for income and race and ethnicity impacts these estimated differences. We additionally considered how these disparities relate to BC, ${\mathrm{NO}}_2$, and ${\mathrm{PM}}_{2.5}$ exposure levels and compared relative disparities across pollutants to put our study results into context with previous studies and identify the potential for cumulative health effects for groups disproportionately exposed to multiple pollutants. With the current lack of research on UFP disparities, this paper will help inform future research and regulation to prevent detrimental health effects.

The present study is one of only a few to examine UFP exposure disparities by income, race and ethnicity, and HOLC grade and to compare these with other pollutant disparities. Our findings suggest that in the Greater Seattle Area there are disparities by income, race and ethnicity, and HOLC grade for UFP, BC, ${\mathrm{NO}}_2$, and ${\mathrm{PM}}_{2.5}$ and that those disparities are generally greatest for UFPs, regardless of how the disparities are characterized. These disparities were present even though pollutant concentrations are generally low in the Greater Seattle Area. Concentrations of all four pollutants were higher in lower-income communities than higher-income communities. In comparison with the population-weighted study region average concentration, exposure to both UFP and ${\mathrm{NO}}_2$ was at least 40% higher on average in the lowest ($<\$\mathrm{20,000}$) income group. For incomes between $20,000 and $60,000, UFP concentrations were 9–18% higher than the average concentration; this was the largest relative disparity across the four pollutants. In addition, historically racialized groups were exposed to higher concentrations of UFPs than non-Hispanic White residents on average. Black residents had the highest exposures across all pollutants.

Our findings are consistent with and extend previous studies that found ${\mathrm{NO}}_2$, ${\mathrm{PM}}_{2.5}$, ${\mathrm{O}}_3$, and carbon monoxide concentrations to be higher for racialized people and PM to be higher in areas with lower socioeconomic status.^5–13 We extend those findings by showing similar patterns with UFP and BC concentrations. Although the ${\mathrm{NO}}_2$ and ${\mathrm{PM}}_{2.5}$ concentrations in our study region were lower than U.S. EPA standards, we show that disparities in these pollutants, especially in ${\mathrm{NO}}_2$, are still present. Further, the associations for income and for each racial and ethnic group revealed larger disparities for UFPs and BC than for ${\mathrm{NO}}_2$ and ${\mathrm{PM}}_{2.5}$, which may lead to greater health disparities. As described above, adverse health effects due to UFP exposure may be magnified owing to the potential for UFPs to enter the bloodstream.²⁸

We additionally extended findings from previous studies on historical HOLC grade and air pollution to disparities for UFPs and BC. Lane et al., Schuyler and Wenzel, and Kane previously found disparities by HOLC grade for ${\mathrm{PM}}_{2.5}$, with Lane et al. finding additional disparities for ${\mathrm{NO}}_2$, and Schuyler and Wenzel for nitrogen oxide, sulfur dioxide, and volatile organic compounds.^31–33 We found that, in the parts of Seattle subjected to historical redlining, neighborhoods that were once characterized as hazardous were exposed to higher concentrations of UFP, BC, and ${\mathrm{NO}}_2$ in 2019, with the largest differences apparent for HOLC Grade D. Kane additionally compared ${\mathrm{PM}}_{2.5}$ concentrations in non–HOLC-graded regions to HOLC-graded regions within county boundaries, finding higher ${\mathrm{PM}}_{2.5}$ concentrations in HOLC-graded regions.³³ We also found that concentrations of each pollutant were higher in HOLC-graded regions than non–HOLC-graded regions (Table 1; study region section). We further defined a Grade X for the interior uncategorized areas within the HOLC-graded region, which were blocks with residential populations in the 2010 U.S. Census that were deemed to be business or industrial areas in the 1930s. The elevated pollutant concentrations were the highest in the Grade X area. One reason may be that the industrial infrastructure remains and is a potential source of air pollution, despite present-day residential population. These disparities were greater than those separately observed by racial and ethnic groups, even when the race and ethnicity analysis was restricted to the HOLC-graded region (Table S6). Similar to the other studies, we first calculated average concentrations by HOLC grade, but because of the complex relationship between redlining, income, and race and ethnicity, we then additionally evaluated these associations in a multivariable analysis (Table 5). We found, after adjusting for income and non-Hispanic White population percentage, that although estimated concentrations for less desirable HOLC grades were lower, there remained statistically significant elevated pollutant concentrations in areas previously assigned Grade D and Grade X.

Our goal was to describe disparities, rather than ask causal questions. Thus, the multivariable analysis in Table 5 used regression adjustment to make comparisons of the areas assigned more hazardous grades with those assigned Grade A. Although this is not a realistic counterfactual comparison across all grades, owing to the low probability of any areas graded as hazardous receiving an A grade, it did allow us to describe how much exposures would need to decline in the groups with the highest exposures to be comparable to those with the lowest. Further, in our multivariable analyses, we were able to assess whether the observed disparities differ when we consider HOLC grade alone as compared to models that also consider modern-day income and race and ethnicity percentage. The results show that the grade-specific disparities indeed changed after adjusting for modern-day census characteristics. After adjustment for income and race and ethnicity, the disparities in grades D and X remained statistically significant for UFPs, BC, and ${\mathrm{NO}}_2$. Although racial and economic status were significant factors in assigning HOLC grades, other factors, such as industrial presence, were higher in areas graded D and ungraded areas (Grade X), which may have contributed to our findings.²⁰ In the 1950s and 1960s, local governments across the country zoned Black neighborhoods to permit industrial activity but did not allow industry in White communities.²⁶ Furthermore, decades of disinvestment in Black neighborhoods likely contributed to the exposure disparities we observed. We continue to see the legacy of redlining in Seattle today, with lower average incomes, higher proportions of racialized groups, and overall higher pollutant concentrations in formerly redlined areas.

Schuyler and Wenzel and Kane both additionally compared pollution exposures to asthma prevalence in their study regions, finding a higher prevalence of asthma associated with lower HOLC grades in Pittsburgh, Chicago, and Dallas-Fort Worth.^32,33 Given that our study region was confined to the Greater Seattle Area, it is noteworthy that our overall results aligned with previous research on air pollution exposure disparities. However, there may be specific features of these data and this region, such as lower variation of ${\mathrm{PM}}_{2.5}$ and lower levels of pollution overall, that differ from those observed in other locations and result in different spatial patterns. Despite the lower levels of air pollutants, disparities were still apparent in the Greater Seattle study region. Prior studies on ${\mathrm{PM}}_{2.5}$ were also able to leverage consistent long-term monitoring data, which is not yet readily available for UFPs. It will be important to study changes in UFP exposure and disparities over time, and the associated health effects.

This study offers important strengths. The air pollution data we used are a unique resource with multiple pollutants collected simultaneously and modeled by the same study, thus strengthening the cross-pollutant comparisons. They were collected by a mobile monitoring campaign that was designed to produce unbiased estimates of annual average air pollutants, without needing to adjust for short-term temporal trends as is commonly incorporated into mobile monitoring study estimates, and to target residential rather than on-road exposures. Although some studies use fixed site locations, many mobile monitoring studies collect data only while the vehicle is moving (nonstationary sampling), and thus the measurements and predictions represent on-road exposures.^49–51 Because the Seattle mobile monitoring study used stationary sampling, where the vehicle stopped by pulling off or to the side of the road to sample for 2 min, its measurements more closely approximate off-road concentrations. Furthermore, the 309 stop locations were selected to represent residences of the target cohort, so that the predictions obtained from this study are likely to represent residential exposures and to be relevant for epidemiology, whereas the number of sites allowed us to maximize spatial coverage.

The instrumentation and mobile monitoring approach used to obtain the pollutant data in this study has both strengths and weaknesses. Our mobile monitoring study annual averages were derived from $\sim 29$ measurements totaling an hour’s worth of data at each stop, which we modeled to predict annual averages at census block centroids.⁴⁴ This study was not designed to characterize the spatiotemporal nature of the variation in pollution concentrations, and these averages are noisier than those from long-term fixed site monitoring would be. In spite of the short total monitoring duration at each location, mobile monitoring studies are a useful approach for characterizing long-term average exposure to traffic-related air pollutants.^50,52–54 This is particularly true for UFPs owing to the expense and practical challenges of measuring them and their spatial and temporal heterogeneity.⁵⁵ We contrast this with ${\mathrm{PM}}_{2.5}$, a more regional and less spatially heterogeneous pollutant, which is not typically measured in a mobile campaign. Our ${\mathrm{PM}}_{2.5}$ exposure data were challenged additionally by low ambient levels close to the instrument limit of detection. These pollutant and measurement features were reflected in our data set where UFPs had the largest relative spatial variability and ${\mathrm{PM}}_{2.5}$ the smallest.

Other limitations pertain to the demographic data. The years of the ACS and U.S. Census data (2006–2010) do not directly align with the exposure years (2019–2020). At the time of the mobile monitoring study, the 2020 U.S. Census data were not yet available, and predictions were made at block centroids following the 2010 U.S. Census geographies. Although ideally our exposure period and demographic data would align, we wanted to conduct our analysis at the finest scale possible, and even though more recent ACS data (2015–2019) are available, these data are only at the block group level and not the block level. Block-level data permits better small-scale spatial matching to the HOLC areas than block group data and thus more accurate population-weighted exposure averages. Although there was no historical, long-term source of UFP monitoring data in Seattle, Wang et al. documented that modeled spatial surfaces of traffic-related pollutants are stable over a $7\text{-y}$ period, so the temporal misalignment of the demographic and pollutant data should have minimal impact on these results.⁵⁶ Other researchers have found that although ${\mathrm{PM}}_{2.5}$ levels have widely decreased throughout the country over the decades, relative exposure disparities have remained.⁴ We hope that our study will contribute to future research to learn whether and how disparities might change for UFP once more long-term and widespread data are available. In addition, there is the potential for uncertainty in the census data itself, given that low-income neighborhoods have been shown to have higher variation in data than high-income neighborhoods.⁵⁷ Another limitation is that our classification of race and ethnicity was dependent on census classifications. The census limited our analyses to the groups (Asian, Black, Hispanic, Native American, non-Hispanic White, and Pacific Islander) that we studied. We would have liked to disaggregate the data further to, for example, examine Native American and Pacific Islander populations in our regression analyses as well but, unfortunately, the small population sizes did not allow this. Native American and Pacific Islander populations tend to suffer disproportionately from poor health outcomes and have at times been excluded from public health data collection.⁵⁸ Also owing to small population counts, we did not separately divide any groups other than White into Hispanic and non-Hispanic subsets. The degree of overlap of the Hispanic with the non-White groups was small such that the impact on our results should be minimal. Last, we were unable to disentangle the causal effects of the associations of exposure disparities. Structural racism has had broad and enduring effects on human health and exposure. A full understanding of its impacts needs to extend well beyond documentation of exposure disparities.

Overall, this study allows a comparison of disparities in exposure across both criteria air pollutants (${\mathrm{NO}}_2$ and ${\mathrm{PM}}_{2.5}$) and less commonly measured air pollutants (UFPs, BC) that were obtained simultaneously from a mobile monitoring study. Racialized communities, lower-income communities, and historically redlined neighborhoods in the Seattle area were exposed to higher concentrations of UFPs, BC, ${\mathrm{NO}}_2$, and ${\mathrm{PM}}_{2.5}$. Concentrations were highest for the lowest income group ($<\$\mathrm{20,000}$ per household/y) across all pollutants. Further, we observed large disparities by race and ethnicity, especially for Black residents, and by redlining grade for UFP, BC, and ${\mathrm{NO}}_2$. Disparities were largest for UFP and weakest for ${\mathrm{PM}}_{2.5}$. This multipollutant comparison is valuable for putting the results of our study into context with previous studies, most of which have focused on ${\mathrm{NO}}_2$ and ${\mathrm{PM}}_{2.5}$. The presence of higher exposure disparities for UFPs reinforces the need for more long-term monitoring and research on the health effects of UFPs, which will in turn guide regulation of these particles. Promoting cleaner air will help those disproportionately exposed to air pollution and suffering from its resulting health effects.

This research was supported by award numbers R01ES026187, T32ES015459, and 1R25ES025503 from the National Institutes of Health (NIH)/National Institute of Environmental Health Sciences, and CR-83998101 from the Health Effects Institute (HEI) (all to L.S.). This content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or HEI. The corresponding code repository for this manuscript can be found at https://github.com/kayabramble/SeattleAirPollutionDisparities.

The authors declare they have nothing to disclose.