COVID-19’s disproportionate impact on racial and ethnic minorities has been widely covered by the news media. My analysis indicates that an important reason why African Americans and Latinos are more likely to become infected with the coronavirus is that these minority groups tend to live in more densely populated urban neighborhoods than white non-Hispanics. My estimates show that higher population densities in the urban neighborhoods in which many African Americans live could account for a 40% higher incidence of COVID-19 and 68% more deaths from COVID-19 relative to white non-Hispanics. Latinos tend to live in even more densely populated urban neighborhoods than African Americans. Consequently, due to the neighborhoods in which they live, Latinos could face a 63% higher incidence of COVID-19 and more than twice as many deaths from COVID-19 compared to the typical white non-Hispanic. It is also important to note that county level data is too highly aggregated to account for differences in the neighborhoods in which whites, African Americans, and Latinos live; even within the same county, African Americans and Latinos live in neighborhoods that are about twice as densely populated as white non-Hispanics.
Contact Rates are Higher in More Densely Populated Areas
Variation across communities in the prevalence of COVID-19 and the number of deaths from COVID-19 are strongly correlated with differences in population densities across neighborhoods. The spread of infectious diseases is more difficult to control in communities where residents live close to one another. The R0, or basic reproduction number of the coronavirus, depends on the contact rate between infected individuals and other residents of the community. The contact rate among residents of more densely populated communities is likely to be higher so that the coronavirus is likely to spread more rapidly and extensively in these neighborhoods.
The goal of social distancing is to lower the contact rate for individuals and slow the spread of the coronavirus. Social distancing is most important in densely populated neighborhoods where contact rates are likely to be higher. The empirical results described below are based on counts of cases and deaths that occurred during periods of shelter-in-place regulations in most densely populated counties. Despite the efforts to practice social distancing I find that more densely populated neighborhoods had significantly more cases of COVID-19 per capita.
Measuring Population Density
The most disaggregated national COVID-19 datasets that are currently widely available are at the county level. There are over 3,000 counties in the US, and counties vary widely in both population and land area. Because areas and neighborhoods within a county are so heterogeneous it is often misleading to rely on a simple measure of a county’s population density—the ratio of the number of residents in a county to the land area of the county.
I calculate population densities using data for over 217,000 neighborhoods (Census Block Groups) from the 2014-2018 American Community Survey. I measure the population density for each county using the number of residents per square mile in the neighborhood in which the median county resident lives. This calculation differs from the ratio of a county’s population to its land area. For example, Fargo is located in Cass County, North Dakota and accounts for more than 68% of the county’s population, but less than 3% of the land area in the county. A conventional measure of population density in the county would be about 100 residents per square mile, but the median resident of the county lives in a neighborhood in Fargo with a population density of over 3,200 residents per square mile. The population density of the neighborhood in which the median resident of the county lives is more relevant to contact rates for county residents and the risk of the spread of the coronavirus than would be a simple conventional measure of population density.
More Densely Populated Urban Areas Face Greater Risks of COVID-19
Figure 1 compares the population densities of New York City and five quintiles of counties in the US (excluding New York City). Quintiles are formed by grouping Americans into five different categories that are equal in population. The top quintile includes the 20% of Americans who live in the most densely populated counties, the second quintile includes the 20% who live in the next most densely populated counties, and so on until reaching the lowest quintile with the 20% of Americans who live in the least densely populated counties. Figure 1 shows that, outside of New York City, 20% of Americans live in counties with an average population density of about 9,874 residents per square mile. At the other extreme, another 20% of Americans live in counties with an average population density of about 244 residents per square mile.
Figure 2 compares the number of COVID-19 cases per million residents in New York City and the five quintiles described above. New York City has over 16,500 cases per million residents, about four times the rate in the top quintile of the most densely populated counties (other than New York City) which have over 3,800 cases per million residents. The lowest quintile, including the least densely populated counties, has over 900 cases per million residents or less than one quarter of the rate in the top quintile and less than one-sixteenth the rate in New York City.
Figure 3 provides a similar comparison between the number of COVID-19 deaths per million residents in New York City and the five quintiles based on population densities. New York City has over 1,200 deaths per million residents, more than seven times the rate in the top quintile of the most densely populated counties (other than New York City) which has over 170 deaths per million residents. The lowest quintile including the least densely populated counties has fewer than 40 deaths per million residents, which is less than one quarter of the rate in the top quintile.
New York City is the most densely populated area in the US. The median New York City resident (across all five boroughs) lives in a neighborhood with more than 63,800 residents per square mile. The typical New York City resident lives in a neighborhood with more than double the density of San Francisco, almost three times the density of Philadelphia, and more than five times the density of Chicago or Los Angeles. Because New York City is so much more densely populated than other counties, and because its residents have suffered the most during this pandemic, the inclusion of New York City would skew the results of any statistical analysis of the relationship between population density and the incidence of COVID-19. Consequently, this study of the relationship between COVID-19 and population density excludes New York City from the analysis.
Differences in Population Density Across Neighborhoods Can Help Explain Racial Differences in the Impact of COVID-19
Using the county-level data described above, I estimate a simple regression of the incidence rate of COVID-19 on the population density of counties (excluding New York City) and find an elasticity of .39. This means that a ten percent increase in the population density of a county is associated with 3.9% more cases of COVID-19. I estimate a corresponding simple regression of the rate at which county residents have died from COVID-19 on the population density of counties (excluding New York City) and find an elasticity of .58. This means that a ten percent increase in the population density of a county is associated with 5.8% more deaths attributable to COVID-19.
A closer examination of the more than 217,000 Census Block Groups in the US indicates that the typical African American lives in a neighborhood that is 2.38 times more densely populated than the neighborhood in which the typical white non-Hispanic lives. The typical Latino lives in a neighborhood that is 3.48 times more densely populated than the neighborhood in which the typical white non-Hispanic lives. Moreover, even within the same county, African Americans and Latinos tend to live in neighborhoods that are about twice as densely populated as the neighborhood in which the typical white non-Hispanic lives.
Using the estimated elasticity between COVID-19 incidence and population density, combined with racial differences in the population densities of neighborhoods, the proximity of African Americans to other neighborhood residents could account for a 40% higher incidence of COVID-19 per capita while the proximity of Latinos to other neighborhood residents could account for a 63% higher rate of infection per capita. This relationship is illustrated in Figure 4.
Using the elasticity between COVID-19 fatalities per million residents and population densities, the more densely populated neighborhoods in which minorities live could account for 68% more fatalities per capita for African Americans and 111% more fatalities per capita for Latinos. This relationship is illustrated in Figure 5.
The simple empirical analyses presented here suggest that differences in the neighborhoods where people live can help explain why some demographic groups have been more affected by the coronavirus pandemic than others. People living in more densely populated urban areas are at much greater risk of contracting COVID-19 despite the enormous effort and cost that these communities have devoted to sheltering in place and social distancing.
Another lesson from the simple analyses presented here is that county level data are far too aggregated to understand why some communities face greater risks from the coronavirus pandemic than others. Future studies could be refined by accounting for differences in factors other than population density across neighborhoods that likely have an effect on incidence and fatality rates for COVID-19 such as the age distribution of the population, median income, poverty rates, and access to hospitals and health care professionals. A more detailed study of the factors influencing the risk of COVID-19 should rely on more granular data on the zip codes and/or Census areas within a county where COVID-19 cases have occurred.
1. A conventional county-wide population density measure is akin to repeatedly randomly sampling longitudes and latitudes within a county and measuring the average number of people per square mile within the sampled “neighborhoods.” In contrast, the measure of population density I use is akin to repeatedly randomly sampling county residents and measuring the average number of people per square mile within the sampled “neighborhoods.”
2.These data are from the New York Times and include all cases in this database through Tuesday, April 21, 2020.
3.The relationship I find between the incidence of COVID-19 and population density would be even stronger if New York City was included in the analysis.