The epidemiological transition

Our period of analysis coincides with the late stages of the epidemiological transition in the United States. This transition was characterized by two broad changes in people’s life chances: a decline in deaths from infectious disease and a shift in the most common ages of death from young to old ages (Deaton Reference Deaton2013: 31).

The causes of this transition remain a subject of debate. One school of thought emphasizes the importance of improved nutrition (Fogel Reference Fogel2004; McKeown and Record Reference McKeown and Record1962), especially the fact that increasing incomes enabled people to buy more and better food, which translated into better health. However, much of the early work on the transition was based less on precisely estimated effects of nutrition than on a failure to find strong evidence for other causes (Cutler et al. Reference Cutler, Deaton and Lleras-Muney2006).

Other scholars emphasize public health initiatives, particularly clean water and sewerage (Cutler and Miller Reference Cutler and Miller2005; Troesken Reference Troesken2004; Alsan and Goldin Reference Alsan and Goldin2019) and local public health regulations driven by class struggle (Szreter Reference Szreter1988). Cutler and Miller (Reference Cutler and Miller2005) found that water filtration drastically reduced typhoid, infant, and total mortality. Anderson et al. (Reference Anderson, Charles and Rees2022) replicated this analysis after adding several cities, correcting transcription errors, and making other changes to its measurement and methods. They estimate nearly identical effects of filtration on typhoid mortality but revise the estimated effects on total mortality downward to the point of statistical insignificance. Precisely estimating the effects of public health interventions is challenging. Beach (Reference Beach2022: 7) notes that attempting to recover an average treatment effect of each clean-water intervention, as both Cutler and Miller (Reference Cutler and Miller2005) and Anderson et al. (Reference Anderson, Charles and Rees2022) do, can be misleading because “there was often one intervention that was important for improving water quality, but the intervention that mattered varied from city to city.” If a city began filtering its water before it began chlorinating it, for example, the estimated effect of chlorination could be muted because most of the mortality gains would have already taken place. Ultimately, despite disagreements about the magnitude of the urban mortality decline caused by clean-water technologies, all these authors agree about the importance of public health investments generally and clean-water technology specifically.

The epidemiological transition in the United States was concentrated in cities. In 1900, mortality rates were higher in American cities than in rural areas, but by mid-century, this gap had vanished (Haines Reference Haines2001). Between 1900 and 1948, the share of urban deaths from infectious causes fell from a median of 37% to 6% (Feigenbaum et al. Reference Feigenbaum, Muller and Wrigley-Field2019).Footnote ²

Racial inequality in mortality in the early twentieth century

At the beginning of the 20th century, the Black child mortality rate far exceeded the white child mortality rate (Preston and Haines Reference Preston and Haines1991: 96). This large total disparity, moreover, understated even larger disparities within urban and rural areas because most Black Americans lived in rural areas, where the risk of death from infectious disease was comparatively low. Over the next four decades, the evolution of racial inequality in infectious disease mortality was shaped by two countervailing trends: the disappearance of the urban mortality penalty and the large-scale migration of Black Americans to cities, where they were often pushed into segregated and crowded housing.

As discussed above, the introduction of clean-water technologies in US cities was an important contributor to the urban mortality decline. Because it was difficult to segregate urban water systems, these technologies also reduced racial inequality in mortality from waterborne diseases (Troesken Reference Troesken2004). Using data from 33 US cities from 1888 to 1920, Troesken (Reference Troesken2004: 124) shows that water filtration had a large negative effect on the Black typhoid mortality rate, but a much smaller effect on the white typhoid mortality rate. Anderson et al. (Reference Anderson, Charles, Rees and Wang2021) find similar effects of clean-water technologies on racial inequality in infant mortality. Filtration reduced both Black and white infant deaths, but chlorination reduced only Black infant mortality, thereby narrowing the Black-white infant mortality gap (Anderson et al. Reference Anderson, Charles, Rees and Wang2021). Although Black migrants began moving to cities when infectious death rates were higher in urban than rural areas, over time, this urban penalty weakened and ultimately reversed. Consequently, most of the gains in Black life expectancy between 1900 and 1940 were concentrated in cities (Troesken Reference Troesken2004: 4).

However, these gains in urban Black life expectancy between 1900 and 1940 were partially offset by the crowded, segregated, and poorly ventilated housing that many Black migrants were forced into. Such housing generated the ideal conditions for respiratory infections like tuberculosis to spread (Acevedo-Garcia Reference Acevedo-Garcia2000; Roberts Reference Roberts2009). Segregationists who used high rates of Black mortality to justify their cause (Stein Reference Stein, Cole and Ring2012) thus got the causal relationship exactly backwards. Zelner et al. (Reference Zelner, Muller and Feigenbaum2017) find that after the 1918 influenza pandemic, the estimated per-pulmonary-case risk of TB infection rose among Black Americans but not white Americans in 11 large US cities. These elevated transmission rates slowed the decline of tuberculosis mortality among Black Americans in cities, increasing racial inequality in death from tuberculosis. The nonwhite-to-white ratio in urban tuberculosis mortality jumped sharply from 2.5-to-1 during the influenza pandemic to more than 4-to-1 in the 1930s (Feigenbaum et al. Reference Feigenbaum, Hoehn-Velasco, Muller and Wrigley-Field2022). In the early twentieth century, tuberculosis contributed more to the total racial disparity in infectious disease mortality in US cities than any other disease (Feigenbaum et al. Reference Feigenbaum, Hoehn-Velasco, Muller and Wrigley-Field2022).

Unsegregated urban water systems and segregated, crowded housing should have pulled Black infectious disease mortality in different directions. On the one hand, clean-water technologies reduced racial inequality in typhoid and infant mortality. On the other, residential overcrowding likely increased racial disparity in death from tuberculosis. In the remainder of the paper, we use city-level data on deaths by cause and age to assess these predictions.

Data

To study racial inequality in infectious disease mortality in the early twentieth century, we gather annual data on deaths by age and cause in all reporting US cities. These data come from published volumes of the Vital Statistics of the United States.Footnote ³ We rely on the years 1906 to 1933 because these were the only years in the first half of the twentieth century when deaths were reported separately by age, cause, and racial classification. We combine counts of deaths by cause with age-and-racial-group-specific denominators from the IPUMS Restricted Complete Count Census Microdata (Ruggles et al. Reference Ruggles, Flood, Goeken, Grover, Meyer, Pacas and Sobek2018) to construct cause-specific, age-specific mortality rates.

The U.S. Vital Statistics are the only data enabling us to study racial inequality in mortality by age and cause in the early 20th century, but these data have two notable limitations. First, they include only a small number of unrepresentative cities, listed in Table A.3. Because our primary interest is in comparing Black and white mortality within cities, our sample should accurately capture inequalities inside those particular cities, but may not generalize well to other cities, many of which had very small Black populations. Cities in the DRA reporting age and cause data by racial classification were selected twice over: they were included in the DRA and were further selected to report these specific data. Table A.4 compares the cities that ever appear in our sample to other cities based on their year of entry into the DRA. This comparison confirms that our sample contains substantially larger cities that had larger Black populations and were more likely to be in the South than other cities, either in the United States’ DRA or in the full urban population. The uneven distribution of the Black population across urban areas in this period means that we account for much of the urban Black population and those parts of the urban white population that shared cities with Black Americans, but far less of the urban white population as a whole. Feigenbaum et al. (Reference Feigenbaum, Muller and Wrigley-Field2019) showed that, during most of the period we examine here, high mortality rates in southern cities were driven by extremely high mortality in those cities’ large Black populations. Until the mid-1920s, white mortality was similar across regions, but during the late 1920s and early 1930s, white infectious disease mortality was higher in southern cities than in other cities. Because our sample over-represents cities with high white mortality rates during the period when we document increasing racial inequality, it is likely that they understate racial inequality in urban infectious disease mortality at the national level.

Second, like all mortality series from this period, our data form an unbalanced panel. States were not required to report vital statistics until 1933, and cities entered the DRA in a staggered fashion over the period (Haines Reference Haines, Carter, Gartner, Haines, Olmstead, Sutch and Wright2006), separately from and often well before their respective states. This was also true of DRA cities reporting deaths by age and racial classification (as shown in Table A.3). Because the results reported below document changes in racial inequality during the 1920s, it is reassuring that no cities were added during that decade.Footnote ⁴ Therefore, the changes we document do not simply reflect changes in which cities are included in the sample.

Throughout the paper, we report the median infectious disease mortality rate. In the appendix, we reproduce our results using the median all-cause mortality rate. When calculating the median mortality rate, we weight by the city population to capture the mortality risk experienced by the median person. In the appendix, we show that our results are substantively identical if we use weighted means instead.

We present median mortality rates by age using eight age groups: infants (under 1), ages 1–4, ages 5–19, ages 10–19, ages 20–29, ages in the 30s and 40s, ages in the 50s and 60s, and ages 65/70 and up. The Vital Statistics made a slight change in the age groupings it reported between 1921 and 1922. While age groupings under 29 were unaffected, age groups over 30 do change. For example, whereas we see an aggregation of deaths among 30 to 49-year-olds before 1922, that group changes to 30 to 44-year-olds after 1922. Table A.1 shows the specific ages included in each age group for the periods 1906–1921 and 1922–1933. We construct denominators to match the age groupings used in each year. In all figures involving age groups 30 and over, we indicate the year of the age-grouping change with a vertical gray line.

In addition to reporting infectious disease mortality rates, we also report rates of death from specific causes. Table A.2 lists the causes included in each major grouping we consider – respiratory (pneumonia, influenza, and bronchitis), tuberculosis (all forms), waterborne/foodborne, and childhood illnesses. We study these causes for several reasons. First, diarrheal deaths from waterborne and foodborne diseases were the largest killer of children during this period (Preston and Haines Reference Preston and Haines1991). As discussed above, clean-water technologies also caused infant and childhood deaths to decline (Alsan and Goldin Reference Alsan and Goldin2019; Beach et al. Reference Beach, Ferrie, Saavedra and Troesken2016; Ferrie and Troesken Reference Ferrie and Troesken2008), and these declines affected both Black and white children (Anderson et al. Reference Anderson, Charles, Rees and Wang2021; Troesken Reference Troesken2004). Respiratory causes (influenza, pneumonia, and bronchitis) and tuberculosis, in contrast, should have affected Black and white Americans differently due to their differential exposure to segregated and crowded housing. Finally, we consider the major childhood causes of death (diphtheria, measles, whooping cough, scarlet fever, and smallpox) reported in Preston and Haines Reference Preston and Haines1991 to assess the extent to which these contributed to the total racial disparity in infectious disease mortality.

Implications

The numbers we report here are inadequate to the scale and depth of suffering they imply. Beach et al. (Reference Beach, Clay and Saavedra2022: 48) note that death among young adults and adults in midlife during the 1918 influenza pandemic “was particularly notable and tragic for individuals losing parents, spouses, and breadwinners.” After the pandemic, infectious disease deaths in the prime of life became much less common for white Americans. Although the absolute decline in tuberculosis mortality over time meant that the suffering of bereavement became rarer, Black Americans continued to “die in great numbers without a crisis ever being declared” (Hartman Reference Hartman2020).

With a persistent risk of contracting tuberculosis, Black people wrote to doctors and health advice columnists asking whether they should delay marrying and having children. In 1932, a young man who had contracted tuberculosis asked an advice columnist in the Pittsburgh Courier whether he should marry his partner: “Do you think I ought to tell her that I cannot marry her? You see, we would have puny children and I would not live long enough to educate them” (Strong Reference Strong1932). Responding to a similar question in the Chicago Defender, Dr. A. Wilberforce Williams argued that “Marriage and tuberculosis are not good partners.” He advised the inquirer to postpone his marriage “for one year at least, after you have been cured of all sign and symptoms of tuberculosis” (Williams Reference Williams1916). An obituary for William “Social” Morris published in the Defender in the same year noted that he succumbed to tuberculosis just five months after getting married (“Social” Morris Dead 1916).

Tuberculosis robbed many Black children of parents and economic security. “Tuberculosis is guilty of many a mother’s early death,” noted Elvin Elliott Rawlins, a Black physician who wrote a weekly medical column in the New York Amsterdam News (Rawlins Reference Rawlins1923). “A high death rate in the middle of life,” was also “a serious economic problem for the race,” argued R. Maurice Moss of the Urban League (Moss Reference Moss1926). Even in the absence of deaths from tuberculosis, the fear of contagion may have led some Black workers to lose their jobs. In 1928, a white citizen of Kansas told a newspaper that “Immediately after the newspapers came out with their statement of health conditions among Negroes, as shown by the tuberculosis survey of Dr. W. J. Thompkins, every one of my neighbors discharged every Negro servant in his employ” (False Survey Made Women Lose Jobs 1928). The economic consequences of death and infection in the prime of life thus may have been an important contributor to the Black-white gap in family income (Jácome et al. Reference Jácome, Kuziemko and Naidu2021).