Among COVID-19 fatalities in the US, minorities are overrepresented (Yancy 2020) while women are underrepresented (Peckam et al. 2020). By looking at the intersection between race and gender, we uncover a Black female bias: while Black men are affected as much as White men, Black women are more affected than White women, and this is due to their lower socioeconomic status. The first and most harshly hit by the pandemic were Black women employed as frontline workers who commute on public transport from historically redlined blocks.
In a new paper (Bertocchi and Dimico 2021), we take advantage of extraordinarily detailed individual-level and georeferenced data on US daily deaths from COVID-19 and other causes provided by the Medical Examiner's Officer of Cook County, Illinois, the county that includes the metropolitan area of Chicago. The information includes race and ethnicity among a wide array of other individual characteristics such as gender, age, pre-existing conditions, and georeferenced home address. The present analysis is based on data up to 15 September 2020, covering the first wave of the epidemic in Cook County. Figure 1 shows the spatial distribution of COVID-19 deaths recorded since 16 March 2020, the day the first COVID-19 death was recorded. We superimpose on the map the boundaries of Census block groups.
Figure 1 COVID-19 deaths in Cook County, 16 March to 15 September 2020
We combine death data with US Census data on occupation by sector, public transport use, household crowding, and access to health insurance – down to the block group level of disaggregation. Since the county comprises almost 4,000 block groups, this represents a major advantage compared to other analyses of the racially differentiated impact of the pandemic (Almagro and Orane-Hutchinson 2020, McLaren 2020) that have been conducted at a state, county or, at best, ZIP code level (there are only 164 for Cook County). The resulting unique dataset allows us to jointly investigate the racial and gendered impact of COVID-19, its timing, its determinants, and its geography.
The Black female bias
Our dataset allows us to focus on the potential intersection between race and other demographic characteristics, notably gender. Preliminary correlational evidence suggests that, even after controlling for age and comorbidities, the probability of dying from COVID-19 has been particularly high for Black women, while Black men were not significantly more likely to die from the disease than White men.
To establish our main results, we employ information on all deaths (from COVID-19 and any other cause reported by the Medical Examiner) recorded from 1 January to 15 September in 2020 and 2019 and construct a cell-level panel, with cells aggregated at a race, census block group, week, and year level. The main outcome of interest is a measure of excess deaths for each race in a given block group and week in 2020, relative to the same race, block group, and week in 2019. Using an event study approach, we capture differential trends in deaths between years, pre-and post-COVID-19 weeks, and races. In Figure 2, we compare these differential trends for women and men.
Figure 2 Sex-disaggregated excess deaths for Blacks and Whites and Black-White excess death differential
Note: The plots report coefficients for fixed-effect regressions where the dependent variables are excess deaths for Blacks and Whites, by sex (females in top left panel, males in top right panel) and the Black-White differential in excess deaths, by sex (females in bottom left panel, males in bottom right panel). Vertical lines represent 95 percent confidence intervals. Data refers to deaths from any cause reported between 1 January and 15 September 2020 and 2019. Event time 0 corresponds to the week of 11 March.
The top two panels of Figure 2 show that excess deaths are near zero, as expected, in the weeks that precede the start of the epidemic. They shoot up in the second half of March 2020, when the epidemic starts, and are more numerous for males independently of race. However, we also observe that Black females outnumber White females (top left panel), while among males racial differences are much less pronounced (top right).
The bottom two panels confirm that the racial differential in excess deaths is larger and more prolonged for females (bottom left). This means that the racial disadvantage is largely attributable to Black females, who are hit by the epidemic earlier and more severely. In other words, a male bias is present only within the White population while, strikingly, within the Black population we do not observe any significant sex-related differences. To quantify, in the critical week of 8 April 2020, the Black-White differential in excess deaths was 3 percentage points and was entirely driven by Black women.
What drives the Black female bias?
The emergence of a Black female bias exposes an interaction between race and sex that had been so far overlooked. What explains it? A comparison across block groups reveals that it is driven by those with a larger population share in poverty. Differences in poverty rates absorb differences in the shares of people aged 65+ and with pre-existing conditions. This suggests that socioeconomic disparities, rather than demographic and biological differences, lie at the heart of the higher vulnerability of Black women. But what is it, among socioeconomic disparities, that can channel higher viral transmission and mortality?
We look at four potential and not mutually exclusive channels: jobs, use of public transport, housing crowding, and health insurance coverage. The first and second reflect the risk of contracting the virus at the workplace and on the way to work; the third can magnify transmission rates within the household; and the last affects access to medical care once contagion has occurred.
In order to assess whether the higher risk of contracting the virus at the workplace can explain the Black female bias in deaths, we compute the share of women and men employed in 20 industries, at the block group level. Splitting the sample between block groups with above- and below-median shares shows that the Black women’s death differential is explained by female employment in two key frontline, high-exposure sectors: health care and transportation/warehousing. These are sectors where Black women are overrepresented and that pay lower wages (Bertocchi 2020, Ross and Bateman 2019). Other high-exposure, low-pay jobs, for example in restaurants, where again Black women are heavily represented, do not explain death differentials, likely because the shutdown of the food sector protected their health, despite massive layoffs (Albanesi and Kim 2021, Alon et al. 2020).
A second contributing channel is the intensity in using public transport, which we measure with the share of people using it and the length of commute to work (Caselli et al. 2020). By contrast, we find no explanatory power for housing crowding, the diffusion of multigenerational families, and even lack of health insurance. Lastly, using the georeferenced home address of the deceased, we overlay the map of fatalities onto the redlining maps created in the 1930s in order to assess mortgage default risk (Bertocchi and Dimico 2020). We find that the diminished resilience of Black women is geographically concentrated in formerly low-graded blocks, which uncovers a persistent influence of historical racial segregation.
Thanks to a unique source of data, we have established that the COVID-19 death toll in Cook County has been disproportionately imposed on Black women employed in high-exposure, frontline jobs in the health care and transportation sectors, that they reach by public transport from the historically poor neighbourhoods where they reside.
Since we deal with the second most populous US county, which contains the third largest metropolitan area in the country, our findings do carry wider relevance. They also underline the need for granular data combining COVID-19 outcomes by race and sex with socioeconomic information. It is only through such data that scientists can produce evidence capable of guiding effective policy responses, including prioritisation strategies for vaccination campaigns, even after the emergency is over.
Albanesi, S and J Kim (2021), “The gendered impact of the COVID-19 recession on the US labor market”, NBER Working Paper No. 28505.
Almagro M and A Orane-Hutchinson (2020), “The determinants of the differential exposure to COVID-19 in New York City and their evolution over time”, Covid Economics 13: 31–50.
Alon A, M Doepke, J Olmstead-Rumsey and M Tertilt (2020), “The shecession (she-recession) of 2020: Causes and consequences”, VoxEU.org, 22 September.
Bertocchi, G (2020), “COVID-19 susceptibility, women, and work”, VoxEU.org, 23 April.
Bertocchi, G and A Dimico (2021), “COVID-19, race, and gender”, CEPR Discussion Paper No. 16000.
Bertocchi, G and A Dimico (2020), “Race and the COVID-19 pandemic”, VoxEU.org, 29 July.
Caselli F G, F Grigoli, P Rente Lourenço, D Sandri and A Spilimbergo (2020), “The disproportionate impact of lockdowns on women and the young”, VoxEU.org, 15 January.
McLaren, J (2020), “Racial disparity in COVID-19 deaths: Seeking economic roots in census data”, VoxEU.org, 11 August.
Peckham, H, N M de Gruijter, C Raine et al. (2020), “Male sex identified by global COVID-19 meta-analysis as a risk factor for death and ITU admission”, Nature Communications 11: 6317.
Ross, M and N Bateman (2019), “Meet the low-wage workforce”, Brookings.
Yancy, C W (2020), “COVID-19 and African Americans”, Journal of the American Medical Association, Opinion, 15 April.