VoxEU Column COVID-19 Health Economics Poverty and Income Inequality

Race and the COVID-19 pandemic

COVID-19 pandemic is having a disproportionate impact on African Americans, who are dying at a rate two to three times higher than their population share. This column uses a detailed individual-level dataset from Cook County, Illinois, to explore the relationship between COVID-19 mortality and race. Not only are Black Americans disproportionally affected by COVID-19, but they also started to succumb to it earlier than other groups. Such asymmetric effects can be traced back to racial segregation introduced by discriminatory lending practices in the 1930s.

Since the COVID-19 pandemic hit the US, discussion on its disproportionate impact on African Americans has been at centre stage. The Atlantic (Kendi 2020) was the first to launch a call, on 4 March 2020, for adequate race-disaggregated data to assess the extent of this phenomenon. On 7 April, The New York Times (Eligon et al. 2020) reported that in Illinois, 43% of people who had died from the disease were African Americans, a group that represents only 15% of the state population. In Michigan, African Americans represented 40% of deaths against 14% of the population; in Louisiana, 70% against 33%. On the same day, The Chicago Tribune (Reys et al. 2020) reported that 68% of the dead in Chicago were African Americans, who represent 30% of the city’s population. These figures combined uncovered that Black Americans were dying at a rate two to three times higher than their population share.

The urgency of the racial issue has been widely acknowledged within the medical literature. While the high risk of COVID-19 death for minorities tends to correlate with pre-existing health conditions, possibly because of genetic and biological factors, the consensus is that race differentials are also associated with socioeconomic factors reflecting living and working conditions (Yancy 2020).1

Due to the unavailability of race-disaggregated individual data, economists have so far evaluated the effect of race on COVID-19 outcomes using aggregated data, either at the ZIP-code level (Borjas 2020, Schmitt-Grohe et al. 2020, Almagro and Orane-Hutchinson 2020), or across counties (Desmet and Wacziarg 2020, McLaren 2020), or based on survey data (Wiemers et al. 2020).

In a new paper (Bertocchi and Dimico, 2020), we take advantage of an unexplored and extraordinarily detailed individual-level dataset that covers daily deaths from COVID-19 and other causes, and includes race among a wide array of individual characteristics, such as age, gender, pre-existing conditions, and georeferenced home address. The data are collected by the Medical Examiner’s Officer of Cook County, Illinois, the county that hosts the City of Chicago.2 Data collection started on 16 March, when the first COVID-19-related death was recorded in the county. The present analysis is based on data up to 16 June.

The Cook County data

In the three months from 16 March to 16 June, the Medical Examiner reported 4,325 COVID-19 deaths, of which 35% were of Blacks, against a Black population share of 27%. Thus, Black Americans have been dying at a rate 1.3-times higher than their population share. While these figures do confirm that Black people are overrepresented in COVID-19 deaths, they also paint a somewhat more moderate picture compared to the one reported by media.3

Figure 1 COVID-19 deaths, by race (Cook County, 16 March – 16 June 2020)

Note: The figure reports the number of COVID-19 related deaths by day, for Blacks and for all other races combined.

This seeming inconsistency is explained in Figure 1, which plots the number of COVID-19 deaths in each day separately for Black people and all other races combined. By 9 April, the cumulative share of Black Americans who had died from COVID-19 represented 58% of the total; by then, Black people were dying at a rate more than twice their population share. Past its peak in mid-April, deaths among Black people started to decrease so that, by 16 May, their cumulative share was down to 39%, to reach 35% by 16 June.

In the meantime, deaths for other races started to increase, peaked, and then decreased later than those for Black people. Thus, not only are Black people disproportionally affected by COVID-19, but they also started to succumb to it earlier than other groups, which explains the consequent decline in the share of cumulative Black deaths as the epidemic followed its course. What the epidemiological curve reveals is an extraordinary degree of racial segregation, with different groups displaying distinct patterns even in the timing of their exposure to the epidemic.

Historical segregation

To search for the roots of the higher vulnerability of Black Americans to the epidemic, we dig into its potential historical determinants by exploiting information on the georeferenced home address of the deceased. In Figure 2, we plot the Medical Examiner’s COVID-19 death data on a map of the Chicago area, where we also superimpose the Residential Security Maps produced in the 1930s by the Home Owners’ Loan Corporation and georeferenced by the University of Richmond (Nelson et al. 2020).4

Figure 2 Cook County map and COVID-19 deaths (16 March – 16 June 2020)

Note: The map reports census block group boundaries and Home Owners’ Loan Corporation areas, with green, blue, yellow, and red denoting respectively grade A, B, C, and D neighbourhoods.

The Home Owners’ Loan Corporation ranked neighbourhoods based on default risk. The safest areas, defined as ‘best’, were graded A, followed by the ‘still desirable’, ‘definitely declining’, and ‘hazardous’ ones, with respective grades of B, C, and D. Named after the colour assigned to the lowest grade, the redlining policies introduced with the New Deal are believed to have favoured the development of segregated neighbourhoods plagued by unemployment, low housing quality, and unhealthy living conditions.

The correlates of the racial gaps

Using cross-sectional information about individual deaths from COVID-19, we show that the probability that an individual who died from COVID-19 is Black increases with pre-existing conditions, in particular with hypertension and kidney and respiratory diseases among those with high prevalence. However, even after controlling for pre-existing conditions and other demographic and socioeconomic factors, the probability remains higher in lower-graded (C and D) neighbourhoods.

The influence of the Home Owners’ Loan Corporation policies is eliminated only when the Black population share is also included among regressors, which confirms that the policies did induce segregation along racial lines. Since the cross-sectional analysis is based solely on information about those that died from COVID-19, it is biased by sample selection. This limitation is overcome in the event study approach we present below.

Deep determinants of the racial gaps in the response to the COVID-19 shock

By extending data collection back to 1 January 2020 and to deaths from any cause, we obtain a panel dataset of weekly deaths at the census block group level, over which we can assess the reaction to the epidemic outbreak in neighbourhoods assigned to different Home Owners’ Loan Corporation grades. For each block group-week, we gather information on the reported number of deaths (if any) for a given block group in any of the 24 weeks from 1 January to 16 June.

Figure 3 plots the mortality rate from any cause, for Blacks and for all other races combined, for each week in the sample. From the beginning of the period, mortality is higher for Black people. Starting from mid-March, mortality soars among both groups, but much more steeply so for Black people.

Figure 3 Mortality rate, by race (Cook County, 1 January – 16 June 2020)

Note: The figure reports mortality rates from any cause of death by week, for Blacks and for all other races combined.

Our goal is to capture the impact of the shock introduced by COVID-19 on historically segregated areas, that is, whether in C or D neighbourhoods deaths after the shock deviate more from those recorded in the pre-shock period, compared to A and B neighbourhoods.

Figure 4 shows the dynamic effect of residence in a C- or D-graded neighbourhood. The dots represent ordinary least squares coefficients on residence in a C or D neighbourhood and vertical lines depict 95% confidence intervals.

Figure 4 The effect of residence in lower-graded neighbourhoods on deaths, by race (Cook County, 1 January – 16 June 2020)

Note: The dependent variable is number of deaths, of Blacks (panel a) and of other races (panel b). The dots represent OLS coefficients on residence in C or D neighbourhoods. Block group and week fixed effects are included. Vertical lines represent 95% confidence intervals based on standard errors clustered at block group level.

Before the COVID-19 outbreak, the average number of deaths for Black Americans and for all other races in C and D neighbourhoods is not significantly different from the average number of deaths in A and B neighbourhoods. However, after the epidemic shock, deaths in C and D neighbourhoods increase sharply, especially for Black people. Thus, the influence of the discriminatory lending practices of the 1930s persists to the present day, by way of a diminished resilience of the Black population to the shock of the COVID-19 outbreak.

Channels of transmission

To investigate how redlining could affect COVID-19 outcomes, we exploit neighbourhood heterogeneity and look at how the effect varies with a number of neighbourhood characteristics. We use the Social Vulnerability Index dataset provided by the Centers for Disease Control and Prevention. The index includes 15 characteristics grouped into four sub-indices: socioeconomic status (income, poverty, employment, and education variables), household composition (age and family-structure variables), minority status (race, ethnicity, and English-language proficiency variables), and housing (housing structure and vehicle access variables).

The heterogeneity analysis reveals that the main channels of transmission are socioeconomic status and household composition. In other words, the effect is much stronger for neighbourhoods that score worse on these two dimensions. The components that most drive the effect are income and poverty (for socioeconomic status) and the shares of the elderly and of single parents (for household composition). Strikingly, the influence of all these factors is magnified in combination with a higher Black share.

No Latino paradox

In normal times, the US Latino population exhibits lower-than-average levels of mortality.5 This remains true also in Cook County. However, when we replicate the analysis in Figure 4 for Latinos, we find that – even though the effect kicks in later – they are also affected by residence in lower-graded neighbourhoods. This should not be surprising since their settlement patterns were not that different from those of Black people and they were also subject to redlining.


Overall, the evidence points to the persistent influence of racial segregation introduced by discriminatory lending practices in the 1930s. Such past practices result in the asymmetric effects of the epidemic shock, diminishing the resilience of African Americans. Far from being determined by genetic and biological factors, their vulnerability to COVID-19 is caused by socioeconomic status and household composition, through which the legacy of the past manifests itself.


Aaronson, D, D Hartley and B Mazumder (2017), “The effects of the 1930s HOLC ‘redlining’ maps”, Federal Reserve Bank of Chicago Working Paper No. 2017-12.

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.

Bertocchi, G, and A Dimico (2020), “COVID-19, race, and redlining”, Covid Economics 38: 129–95.

Borjas, G J (2020), “Demographics determinants of testing incidence and Covid-19 infections in New York City neighbourhoods,” Covid Economics 3: 12–39.

Desmet, K, and R Wacziarg (2020), “Understanding spatial variation in COVID-19 across the US”, CEPR Discussion Paper 14842.

Eligon, J, A D S Burch, D Searcey and R A, Jr, Oppel (2020), “Black Americans face alarming rates of coronavirus infection in some states”, The New York Times, 7 April.

Greer, J L (2014), “Historic home mortgage redlining in Chicago”, Journal of the Illinois State Historical Society 107: 204–33.

Kendi, I X (2020), “Stop looking away from the race of COVID-19 victims”, The Atlantic, 1 April.

Jackson, K (1980), “Race, ethnicity, and real estate appraisal: The Home Owners Loan Corporation and the Federal Housing Administration”, Journal of Urban History 6: 419–52.

Kirby, T (2020), “Evidence mounts on the disproportionate effect of COVID-19 on ethnic minorities”, Lancet Respiratory Medicine 8(6): 547–8.

Krieger, N, E Wright, J T Chen, P D Waterman, E R Huntley and M Arcaya (2020), “Cancer stage at diagnosis, historical redlining, and current neighborhood characteristics: Breast, cervical, lung, and colorectal cancer, Massachusetts, 2001-2015”, American Journal of Epidemiology, in press.

Markides, K S, and J Coreil (1986), “The health of Hispanics in the southwestern US: An epidemiologic paradox”, Public Health Reports 101: 253–65.

McLaren J (2020), “Racial disparity in COVID-19 deaths: Seeking economic roots with census data”, NBER Working Paper 27407.

Nardone, A, J A Casey, R Morello-Frosch, M Mujahid, J R Balmes and N Thakur (2020), “Associations between historical residential redlining and current age-adjusted rates of emergency department visits due to asthma across eight cities in California: An ecological study”, Lancet Planet Health 4: e24–31.

Nelson, R K, L Winling, R Marciano, N Connolly, et al. (2020), “Mapping inequality: Redlining in New Deal America”, in R K Nelson and E L Ayers (eds.), American Panorama: An Atlas of US History, University of Richmond: Digital Scholarship Lab.

Reyes, C, N Husain, C Gutowski, S St. Clair and G Pratt (2020), “Chicago’s coronavirus disparity: Black Chicagoans are dying at nearly six times the rate of white residents, data show”, The Chicago Tribune, 6 April.

Schmitt-Grohe, S, K Teoh and M Uribe (2020), “COVID-19: Testing inequality in New York City”, Covid Economics 8: 27–44.

Wiemers, E E, S Abrahams, M AlFakhri, V J Hotz, R F Schoeni and J A Seltzer (2020), “Disparities in vulnerability to severe complications from Covid-19 in the US”, NBER Working Paper 27294.

Williamson, E, A J Walker, K Bhaskaran, S Bacon, C Bates, C E Morton, H J Curtis, A Mehrkar, D Evans, P Inglesby, J Cockburn, H I McDonald, B MacKenna, L Tomlinson, I J Douglas, C T Rentsch, R Mathur, A Wong, R Grieve, D Harrison, H Forbes, A Schultze, R Croker, J Parry, F Hester, S Harper, R Perera, S Evans, L Smeeth, B Goldacre (2020), “OpenSAFELY: Factors associated with COVID-19-related hospital death in the linked electronic health records of 17 million adult NHS patients”, Nature, forthcoming.

Yancy, C W (2020), “COVID-19 and African Americans”, Journal of the American Medical Association 323(19): 1891–2.

Zenou, Y, and N Boccard (2000), “Racial discrimination and redlining in cities”, Journal of Urban Economics 48: 260–85.


1 The racial dimension of COVID-19 has also been discussed in the UK (Kirby 2020). The so-far largest epidemiological study on the racial impact of COVID-19 looked at NHS England medical records of 17 million individuals (Williamson et al. 2020).

2 The Medical Examiner’s Office reports those deaths that are under its jurisdiction, including those due to diseases constituting a threat to public health.

3 Similarly, as of 16 June, African Americans account for less than 42% of the deaths in Chicago, down from 68% on 7 April as reported by The Chicago Tribune (Reys et al. 2020).

4 On the history of redlining, see Jackson (1980) and, for Chicago, Greer (2014). On the economics of redlining see Zenou and Boccard (2000) and Aaronson et al. (2017). The medical literature on redlining includes Krieger et al. (2020) and Nardone et al. (2020).

5 This poorly understood phenomenon has been referred to as the ‘Latino paradox’, since it occurs in spite of a low socioeconomic status (Markides and Coreil 1986).

2,310 Reads