The Missing Mortality Advantage for European Immigrants to the United States in the Early Twentieth Century

Abstract

Immigrant populations typically have lower mortality rates and longer life expectancies than their nonimmigrant counterparts. This immigrant mortality advantage has been a recurrent finding in demographic and population health research focused on contemporary waves of immigration. However, historical data suggest that European immigrants to the United States in the early twentieth century had worse health and higher rates of mortality, yet it remains unclear why a mortality advantage was absent for immigrants during this period. This article combines Vital Statistics records and Lee–Carter mortality models to analyze mortality by nativity status for the U.S. White population from 1900 to 1960, examining variation by age, sex, time, and place. Contrary to contemporary expectations of a foreign-born mortality advantage, White immigrants had higher mortality rates in the early 1900s, with the largest foreign-born disadvantage among the youngest and oldest populations. Although foreign-born and U.S.-born White mortality rates trended toward convergence over time, the foreign-born mortality penalty remained into the 1950s. A decomposition analysis finds that immigrants' concentration in cities, which had higher rates of infectious disease mortality, accounted for nearly half of the nativity difference in 1900, and this place effect declined in subsequent decades. Additional evidence, such as a spike in mortality inequalities during the 1918 influenza pandemic, suggests that common explanations for the immigrant mortality advantage may be less influential in a context of high risk from infectious disease.

Introduction

Immigrant populations in the United States often have better health outcomes and longer life expectancies than their nonimmigrant counterparts. An immigrant health and mortality advantage has been a recurrent finding in demographic and population health research and has been documented among migrants from all regions of the world (Riosmena et al. 2017), in numerous destination contexts (Anson 2004; Boulogne et al. 2012; Guendelman et al. 1999; Kennedy et al. 2006; Malmusi et al. 2010), and even among internal migrants within a country (Chen 2011; León-Pérez 2019; Lu 2008). This pattern has been dubbed a “paradox” in part because the mortality patterns of immigrant populations often do not correspond to their social position in a country's ethnoracial hierarchy (Bakhtiari 2021; Markides and Rote 2015). Immigrants often live longer despite having other risk factors for early mortality, such as relatively low socioeconomic status and experiences of discrimination.

However, the bulk of research on immigrant population health has focused on contemporary waves of migration that began in the latter half of the twentieth century. At the turn of the century, the United States was experiencing one of its historically largest waves of immigration. Immigrant populations arriving from Central, Eastern, and Southern Europe by many accounts occupied a “middle tier” status in the U.S. ethnoracial hierarchy (Barrett and Roediger 1997; Brodkin 1998; Ignatiev 2009; Jacobson 1999; Roediger 2006). Although formally classified as White in the U.S. Census, many new immigrants and their descendants experienced interpersonal discrimination, residential segregation, occupational exclusion, and other conditions that today are considered important social determinants of health and mortality.

Available data suggest that the mortality outcomes of European immigrants in this era corresponded to their middle-tier social position. Estimates based on census records from 1900 and 1910 have found higher rates of childhood mortality among White immigrants relative to their U.S.-born White counterparts, although still substantially lower than rates of the Black population (Bakhtiari 2018; Dribe et al. 2020; Haines and Preston 1997; Preston and Haines 1991). Rates of infant and adult mortality were also higher for foreign-born Whites in early 1900s Vital Statistics records (Lieberson 1980). This foreign-born mortality penalty eventually disappeared and reversed, as more recent studies of older-age mortality have found higher relative life expectancy for White immigrants by the end of the twentieth century (Dupre et al. 2012; Mehta et al. 2016; Singh and Hiatt 2006).

It is unclear why European immigrants did not exhibit a similar foreign-born mortality advantage in the early 1900s. Researchers often attribute immigrant health and mortality advantages to selection factors that shape the demographic and health profiles of immigrant populations (Jasso et al. 2004; Kennedy et al. 2006; Riosmena et al. 2017), as well as to differences in health behaviors that may offset other risks tied to socioeconomic status and living conditions (Blue and Fenelon 2011; Fenelon 2013; Lariscy et al. 2015). Most studies of immigrant selection during the early 1900s focus on economic outcomes, rather than health, and evidence is mixed as to the direction of selection (Abramitzky et al. 2012; Spitzer and Zimran 2018). However, even if initial selection led to better health for new arrivals, the epidemiological context of the era may have rendered initial differences short-lived. Infectious diseases represented a substantially larger proportion of overall mortality in the early 1900s (Armstrong et al. 1999; Feigenbaum et al. 2019), particularly in urban areas, where immigrants tended to disproportionately settle (Haines 2001).

This epidemiological environment could have shaped immigrant mortality inequalities in two ways. First, overall nativity differences in mortality may have reflected the population composition and the disproportionate urban location of immigrants. In 1900, nearly 64% of the foreign-born White population lived in cities, compared with only about 37% the U.S.-born White population (Ruggles et al. 2021). Until the mid-1900s, mortality rates were higher in U.S. cities relative to rural areas, in part owing to the impact of infectious diseases in urban environments (Haines 2001). A portion of the foreign-born mortality disadvantage may have been due to higher risk exposure in cities, given the place distribution of the foreign-born population.

However, there are also reasons to expect higher rates of mortality for immigrants even after accounting for location. Although contemporary immigrants often have “paradoxically” low rates of mortality, this may be partially explained by the unique behavioral risk factors and relatively prolonged disease etiology tied to noncommunicable conditions such as heart disease, stroke, and cancer (Blue and Fenelon 2011; Fenelon 2013; Lariscy et al. 2015). Infectious diseases have a shorter time between exposure and disease manifestation and a different set of risk factors, such as crowding and sanitation, that may be more influential than initial selection patterns or health behavior differences. Even in contemporary contexts, infectious diseases appear to be the one disease category in which immigrants do not have a health advantage (Aldridge et al. 2018), and they in fact may be particularly disadvantaged during periods of epidemic transmission (Horner et al. 2022; Labberton et al. 2022; Tuckel et al. 2006).

Despite the size and significance of the White immigrant population in the early 1900s, there are still gaps in knowledge about its mortality outcomes and how they varied. This study examines variation in mortality patterns by nativity status in the early to mid twentieth century using data on deaths of the White population collected from multiple Vital Statistics sources. I examine trends in national-level mortality between foreign-born and U.S.-born White populations from 1900 to 1960, using Lee–Carter mortality models to smooth noisy data and bridge gaps in coverage. In addition to documenting variation in patterns of mortality by age and sex over time, I decompose rate differences across urban and rural locations to assess how variation in population age structure and urban residence patterns contributed to the immigrant mortality disadvantage during this era.

Background

Explanations for the Immigrant Mortality Advantage

Recent estimates of life expectancy at birth suggest that the foreign-born population in the United States will live approximately four years longer than the U.S.-born population (Medina et al. 2020). Evidence of an immigrant health and mortality advantage has been found among immigrants to the United States from every region of the world (Riosmena et al. 2017), and studies have found comparable advantages for foreign-born populations in other destination contexts, including Canada, Australia, and countries of Europe (Anson 2004; Boulogne et al. 2012; Guendelman et al. 1999; Kennedy et al. 2006; Malmusi et al. 2010). Similar advantages in health and mortality outcomes have been documented for internal migrants moving within a country (Chen 2011; León-Pérez 2019; Lu 2008).

Researchers typically attribute the initial immigrant health advantage to a combination of selection processes that shape the health, demographic, and behavioral profiles of immigrant populations. Migrant populations are not representative samples of their origin-country populations but are instead “positively selected” on a number of characteristics, including age, education levels, health status, and even psychosocial traits such as motivation or risk aversion (Akresh 2007; Feliciano 2020; Kennedy et al. 2006). At a basic level, migration across long distances can be taxing and challenging, so individuals who are disabled, dealing with chronic health problems, or critically ill may be less able to migrate, resulting in a migration flow that selects out the less healthy tail of a population distribution.

In addition to demographic selection prior to and during migration, there are related factors that may explain the immigrant mortality advantage. Migration is not always a one-way journey, and some migrants return to their origin countries later in life, particularly in cases of economic migration. Research on Hispanic migrants in the United States, for instance, has examined how selection during return migration may skew mortality statistics, as immigrants who are older and at greater risk of mortality may be more likely to return to their origin countries (Bostean 2012; Turra and Elo 2008). This phenomenon, known as the “salmon bias,” may lead to an undercount of immigrant mortality as return migrant deaths are recorded in a different country (Palloni and Arias 2004).

Aside from health and demographic selection, a substantial portion of the mortality advantage in the United States is also related to differing health behavior profiles, although it is unclear whether this is driven by selection on behavioral patterns relative to origin-country nonmigrants. Hispanic immigrants in the United States are less likely to smoke and drink than their U.S.-born counterparts, have healthier diets, and exhibit other behaviors that may explain a portion of the healthy immigrant effect (Blue and Fenelon 2011; Fenelon 2013; Lariscy et al. 2015). These differences contribute to lower rates of mortality from heart disease, stroke, cancer, and related noncommunicable diseases, which make up the major causes of death in recent decades.

Initial health and mortality advantages, attributable to either selection or behavioral patterns, are not permanent. Numerous studies have found that health and mortality outcomes of immigrants become worse with duration of residence (Cho et al. 2004; Hamilton et al. 2011; Landale et al. 2000; Rumbaut 1997). The children of immigrants often have worse health and mortality outcomes than their parents, and even among the first generation of immigrants, indicators of health often become worse with greater duration of residence in the destination context, sometimes within as few as 10 years. Although immigrant health patterns are often considered paradoxical because many contemporary immigrant populations are ethnoracial minorities in their destination countries and experience disproportionately low socioeconomic status, over time their outcomes often converge with those of nonimmigrant coethnics.

Although evidence for both initial health selection and postmigration declines is widespread, it is not entirely consistent nor universal (Shor and Roelfs 2021). There is variation in the degree of health advantage and postmigration decline based on origin country (Akresh and Frank 2008; Cho et al. 2004; Frisbie et al. 2001; Hamilton and Kawachi 2013; Van Hook and Stamper Balistreri 2007), reasons for migration (DesMeules et al. 2005; Giuntella et al. 2018; Lu 2008), age–period–cohort characteristics (Fenelon 2017; Hamilton et al. 2015), and the context of reception in the destination country (Akresh et al. 2016; Bakhtiari et al. 2018; Bollini et al. 2009). However, research in this area has almost exclusively focused on contemporary waves of migration, which in the United States began following legislation in 1965 that eliminated previous quota-based immigration restrictions. There has been less research on variation across historical contexts and how the different social and epidemiological environments of the early 1900s may have affected selection processes and risk factors for previous immigrant populations.

Immigration to the United States in the Early Twentieth Century

During the “age of mass migration” that spanned the late nineteenth and early twentieth centuries, nearly 30 million immigrants arrived in the United States. New arrivals originated primarily from various regions of Europe, initially Northern and Western Europe, and in the early 1900s primarily countries of Southern, Central, and Eastern Europe. By 1910, the foreign-born share of the U.S. population peaked at 14.7% (Ruggles et al. 2021).

Although selection processes in theory can affect migrant population composition across historical contexts, research on the direction and magnitude of migrant selectivity during the early twentieth century contains mixed results. There is evidence from linked Norwegian and U.S. censuses that male immigrants from Norway to the United States were negatively selected on the basis of economic prospects, with migrants coming from poorer families than nonmigrants and having lower earnings than brothers who remained in Europe (Abramitzky et al. 2012, 2013). However, research relying on arrival records from Ellis Island suggests that Italian immigrants were positively selected at a local level on the basis of height—a proxy associated with various social and economic outcomes—despite evidence of negative selection at a national level (Spitzer and Zimran 2018). Other studies found mixed selection patterns (Connor 2016; Kosack and Ward 2014), but it is worth noting that most research on selection during this period examined economic characteristics, rather than health or mortality.

Return migration also contributed to the demographic contours of immigrant populations, although there is again mixed evidence about whether return migrants were positively or negatively selected (Abramitzky et al. 2012; Bandiera et al. 2013). In the cases of both initial inflows and out-migration, patterns of selectivity appear to have changed in relation to immigration restrictions that created new barriers to movement. As the number of immigrants from Central, Southern, and Eastern Europe increased in the early 1900s, so did anti-immigrant discourse and nativist politics, culminating in the passage of the 1924 Johnson–Reed Act, which instituted a quota-based system that effectively curtailed new immigration for several decades. Policy restrictions on immigration appear to have increased positive selectivity effects (Abramitzky and Boustan 2017; Massey 2016; Spitzer and Zimran 2018), prolonged the duration of stay of migrants, and reduced the amount of return migration (Greenwood and Ward 2015). Such policies also drastically altered the demographic profile of the remaining foreign-born population, as there were fewer young arrivals to balance the existing immigrant cohorts as they aged.

In addition to shaping migrant flows, the nativist climate and policy restrictions impacted immigrant populations who remained. Although European immigrants were officially classified as White in census records, historians and sociologists have documented degrees of difference and exclusion within the White population that created a hierarchy between Anglo-Saxon and Nordic Europeans and more recent arrivals from Southern, Central, and Eastern Europe (Brodkin 1998; Jacobson 1999; Roediger 2006). Such distinctions within the White population were not just a matter of classification but shaped the social experiences of immigrant populations and their descendants. Immigrants in the early 1900s were subject to stereotyping and interpersonal discrimination that at times manifested in violence or organized lynching (Jacobson 1999; Kenny 2006). They were also excluded from certain jobs, occupations, and other opportunities for mobility (Roediger 2006). Immigrants in the early 1900s were also highly segregated from the U.S.-born White population (Eriksson and Ward 2019; Logan and Zhang 2012). Although immigrants often self-select into ethnic enclaves upon arrival, many Southern, Central, and Eastern European populations were denied housing opportunities or subject to predatory practices that often congregated them in closely packed tenements in urban areas with poor sanitation and limited resources (Ager et al. 2020; Riis 1901).

Available evidence suggests that health and mortality patterns of European immigrant populations corresponded to this relative “middle tier” social position. European immigrants to the United States in the early 1900s had higher mortality when compared to U.S.-born non-Hispanic Whites, according to Vital Statistics data. Estimates of child mortality based on census records have found higher rates among foreign-born Whites, relative to their U.S.-born non-Hispanic White counterparts, in 1900 and 1910 (Bakhtiari 2018; Dribe et al. 2020; Haines and Preston 1997; Preston and Haines 1991). By the latter decades of the twentieth century, this mortality penalty appears to have reversed. Studies of older-age life expectancy and mortality have found higher relative life expectancy for foreign-born Whites older than 65 from the 1990s on, although the difference appears smaller than the foreign-born advantage among other ethnic and racial groups (Dupre et al. 2012; Mehta et al. 2016; Singh and Hiatt 2006).

The Epidemiological Context of the Early Twentieth Century

Whereas contemporary patterns of immigrant mortality are considered surprising or paradoxical because outcomes are often better than predicted by their socioeconomic and ethnoracial statuses, the mortality disadvantage of immigrants in the early twentieth century is in many ways more in line with their social position and corresponding determinants and risk factors. Two important differences in the epidemiological contexts across eras may offer partial explanations for the contrasting links between social position and mortality outcomes. First, a larger portion of overall mortality was due to infectious disease (Armstrong et al. 1999; Feigenbaum et al. 2019). In 1900, influenza and tuberculosis were the leading causes of death in the United States. Infectious diseases remained a significant source of mortality in subsequent decades, declining as the United States completed its “epidemiological transition,” in which overall life expectancy increased and infectious disease deaths drastically declined. Research on the immigrant mortality advantage has predominantly focused on contemporary destination countries where mortality is dominated by noncommunicable conditions, yet there is some evidence that even in contemporary societies immigrants experience equivalent or disadvantaged outcomes when it comes to infectious diseases, particularly during times of epidemic transmission (Aldridge et al. 2018; Horner et al. 2022; Labberton et al. 2022; Tuckel et al. 2006).

Two of the common explanations for the immigrant advantage—initial health selection and health behavior differences—may have been less protective in an epidemiological context in which the link between risk exposure, disease, and mortality is quicker, as is the case with infectious disease. For instance, although there is mixed evidence of positive selectivity among the wave of Southern, Central, and Eastern European immigrants (Spitzer and Zimran 2018), it is unclear how long initial health differences would remain postmigration. Contemporary research suggests that initial health and mortality advantages diminish with duration of residence, as the outcomes of immigrant populations come to resemble those of the native-born population. It is likely that postmigration changes would occur more rapidly in a context of high exposure to infectious disease.

Similarly, health behavior differences related to diet or substance use may account for better outcomes among chronic noninfectious conditions for contemporary immigrant populations, but the major risk factors for mortality are different in a context of high infectious disease risk. Sanitation, population density, immediate living conditions, and related risk factors play a larger role than smoking, diet, exercise, and other health behaviors that contribute to the contemporary immigrant mortality advantage. Importantly, the lag time between risk exposure and disease outcomes is typically much shorter. Whereas differential patterns of health behaviors linked to noncommunicable conditions may be protective for years, even in the face of other social and economic risk factors, they may offer little benefit during periods of epidemic pathogen transmission (Aldridge et al. 2018; Horner et al. 2022; Labberton et al. 2022; Tuckel et al. 2006).

A second difference, related to the disease profile of the era, was the relative rate of mortality in urban and rural areas. An urban mortality penalty, in which urban areas had higher rates of mortality than rural areas, persisted through the first decades of the twentieth century, before eventually reversing (Feigenbaum et al. 2019; Haines 2001). This penalty was driven by the higher risk of infectious disease transmission in crowded cities. New immigrant arrivals disproportionately settled in urban areas, raising the possibility of mortality inequalities due to the distribution of the foreign-born and U.S.-born populations.

By the middle of the twentieth century, both epidemiological and social conditions had changed. Infectious disease represented a much smaller portion of overall mortality, and the urban mortality penalty disappeared and reversed (Haines 2001). The social conditions of many European immigrant populations and their descendants also improved, coinciding with a realignment of racial boundaries along a stricter Black–White binary (Lieberson 1980). Distinctions within the White population blurred across generations and over time (Brodkin 1998; Roediger 2006), and this temporal and generational pattern of mobility formed the basis for many early models of linear assimilation and immigrant integration (Alba and Nee 2003; Waters and Jiménez 2005).

It is likely that such changes corresponded with mortality improvements, but there is a gap in knowledge about White immigrant mortality patterns, as the Vital Statistics reports only sporadically included nativity information after 1932. It is unclear when the foreign-born mortality penalty crossed over to a foreign-born mortality advantage for White immigrants in the United States or how various demographic and epidemiological factors contributed to the unique patterns of the period. This study aims to fill gaps in coverage by combining multiple sources of data to examine trends in mortality among the White population by nativity status from 1900 to 1960.

Methods

The analysis draws on mortality data digitized from Vital Statistics of the United States published reports. These annual reports of deaths in the United States are compiled by the National Center for Health Statistics and can be combined with U.S. Census data to produce mortality rates by nativity for the White population.¹ This study combines deaths from annual reports from 1900–1932 and 1939–1941 with decennial rates for 1950 and 1960 calculated in a Vital Statistics summary report (Grove and Hetzel 1968).²

There are differences in the Vital Statistics reports across years that make comparison over time challenging. States were not required to report mortality information at the beginning of the study period, so the initial death registration area in 1900 represents a subset of 10 states plus the District of Columbia. The death registration area was gradually expanded to include more states, resulting in an unbalanced panel across years. The overall mortality rates discussed in the following are based on mortality in death registration states from 1900 to 1933 and national-level data from 1939 to 1960.³

Nativity data are also incomplete. Some deaths in the annual tables from 1900–1933 had an unknown nativity classification: approximately 1.0–1.4% of all deaths across age brackets lacked such classification. The Vital Statistics reports reference three possibilities for dealing with unknown nativity classifications: omitting the deaths, distributing them evenly between groups, or distributing unknown deaths among groups in proportion to their relative size among all deaths. I follow the last method, and distribute deaths of unknown nativity relative to the proportion of deaths each group represented for a given age, year, and sex combination. I similarly distribute deaths of unknown age across the age-groups in proportion to their relative size among total deaths.⁴

To further account for error in death data that may produce unreliable year-to-year estimates, I use Lee–Carter mortality models to estimate fitted age-specific mortality rates based on temporal and age trends within the data (Lee 2000; Lee and Miller 2001). The Lee–Carter model is a popular method for forecasting trends in mortality based on assumptions that changes in age-specific mortality over time are driven largely by period trends captured by $kt$⁠, with age-specific death rates changing in unison, although to different degrees, represented by $bx$⁠:

Although forecasting is not the aim of this article, the Lee–Carter models are used to detect linear trends and smooth estimates of mortality rates across age-groups and periods. The Lee–Carter model uses singular value decomposition to analyze temporal trends across age-groups, allowing for more systematic assessments of changes in mortality and patterns of measurement error in relation to model residuals.

The age-specific mortality rates used as inputs for the Lee–Carter models were calculated by pairing death data from the Vital Statistics reports with population data from the full-count census files made available by the Integrated Public Use Microdata Series (Ruggles et al. 2021). State-level population data by age, sex, race, and nativity were derived from the full-count 1900, 1910, 1920, 1930, and 1940 census files. For years after 1940, I rely on population estimates by nativity status published in the Vital Statistics reports. Intercensal years were estimated using localized regression interpolation for each state. The interpolated populations of individual registration states for a given year were aggregated to the registration area to produce annual population estimates. Death and population estimates were calculated for the following age brackets: under 1, 1–4, 5–14, 15–24, 25–34, 35–44, 45–54, 55–64, 65–74, 75–84, and 85 and over.

In addition to overall mortality rates, I calculate mortality rates by place using a subset of Vital Statistics tables that includes data on aggregated deaths in cities of a certain size within the registration states, which I use to distinguish urban and rural mortality rates. From 1900 to 1909, the Vital Statistics reports distinguish deaths in cities with 8,000 inhabitants or more according to the 1900 census, and after 1910 the reports distinguish deaths in cities with 10,000 inhabitants or more in the previous decennial census. These deaths are subtracted from the total deaths to produce an estimate of rural mortality.

Given the relatively higher distribution of immigrants in urban areas, these data can be used to decompose nativity differences in the overall mortality rate into three components: a difference due to the age structure of the populations, a difference due to the place distribution of each population (i.e., the population concentration in cities), and a rate difference after accounting for age and place composition. Following Das Gupta (1993), the difference in mortality rates (T / t) between populations stratified by age (i) and place (j) can be expressed as a combination of the composition age effect (I), composition place effect (J), and rate effect (R). In the following equations, A / a and B / b represent the composition I effect and J effect, respectively, which are used to calculate standardized rates:

The age effect can be interpreted as the predicted difference between population rates if only the age structures differed, the place effect as the difference if only the place distribution differed, and the rate effect as a difference in rates standardized with respect to both age and place.

Results

Variation by Age and Sex

The first decades of the twentieth century were a period of substantial declines in overall mortality rates, for both the White population and the United States as a whole. However, this was not the case for the crude mortality rate of the foreign-born White population, which was stagnant and then began to increase by the 1920s (Figure 1). This pattern can largely be attributed to the age structure of the foreign-born population. Immigration restrictions in the 1920s limited the influx of younger populations and resulted in an increasingly older foreign-born age distribution (the extent of this population shift can be seen in Figure A2 in the online appendix). By the middle of the twentieth century, the crude mortality rate of the foreign-born White population was more than three times that of the U.S.-born White population.

Age-specific rates are more useful for assessing relative mortality inequalities by nativity status, given the drastic differences in population age structures. Figures 2 and 3 present the uncorrected mortality rates and fitted results of Lee–Carter mortality models by nativity status, sex, and age from 1900 to 1960. Table A1 in the online appendix includes fit statistics from the Lee–Carter prediction models by nativity and sex. Across the full set of years, the models for U.S.-born White men and women perform better in terms of explaining variance in unstandardized mortality rates than models for the foreign-born population. However, this was largely due to infant mortality. As depicted in Figure A1 in the online appendix, residuals of the Lee–Carter models were extremely large for the under 1 age category, ranging from −1.3 to 1.9, likely reflecting the small number of infants who were born abroad and moved within the first year of life, as well as the data collection challenges associated with that unique and uncommon scenario.

Throughout the early decades of the twentieth century, mortality rates were higher for the foreign-born White population among most age brackets. The exception to this pattern appears to be among working-age adults in the early 1900s, where we see comparable or lower rates of mortality in some years for the 25–34 age-group. Although there is evidence of a foreign-born mortality penalty among most age brackets, the magnitude of relative inequalities between foreign-born and U.S.-born White populations varied by age, with the largest differences at each tail of the age distribution. Although the under 1 mortality rate appears lower for the foreign-born population, the relative rarity of migration within the first year of life makes those estimates extremely unreliable.

Mortality differences for the 1–4 age category are more in line with age and time trends, and the higher foreign-born mortality is consistent with findings from research that has used census records to estimate childhood mortality (Bakhtiari 2018; Dribe et al. 2020; Haines and Preston 1997; Preston and Haines 1991). Relative gaps in mortality rates were smallest in the middle of the age distribution, between ages 25 and 54. For instance, mortality rates were nearly identical for women 25–34 throughout the period, and in the early 1900s the foreign-born mortality rate was lower for this age-group. A relative foreign-born disadvantage emerged again among populations 55 and over.

Nativity inequalities in mortality also varied by sex. In general, foreign-born White men had larger gaps in age-specific mortality relative to their U.S.-born counterparts than did foreign-born White women. However, this varied by age-group, with a “crossover” in which foreign-born White women had relatively higher mortality penalties than foreign-born White men above age 65.

Decomposition of Place Effects

As noted earlier, the older age distribution of the foreign-born White population likely explained divergent trends in crude mortality rates between foreign-born and U.S.-born populations, although differences remained even after standardizing for age. A second population characteristic, the concentration of White immigrants in urban areas, likely also shaped both crude and age-specific mortality patterns.

A substantially larger portion of the foreign-born White population lived in urban areas, which had higher rates of overall mortality and infectious disease risk for much of the early 1900s. In 1900, for instance, 70.6% of the foreign-born population in Vital Statistics registration states lived in a city of 8,000 or more, compared to only 47% of the U.S.-born population.⁵ Trends across the study period also diverged, as the proportion of the U.S.-born White population in urban areas declined in the 1910s and 1920s. In 1940, 70.9% of the foreign-born White population lived in a city of 10,000 or more, but only 43.5% of the U.S.-born White population lived in a city of that size. It is worth noting that the states included in the Vital Statistics registration area changed almost annually, and variation in the urban concentration of both populations may in part reflect differences across states.

Figure 4 and Table 1 include the results of decomposing mortality differences between the two populations to account for the age structure of each population, the distribution of each across urban and rural areas, and standardized underlying mortality rate differences. In 1900, nearly half of the nativity difference of 2.2 deaths per 1,000 population can be attributed to the place distribution of the populations. With a few notable exceptions, both the place distribution of the foreign-born population and standardized rate differences, controlling for place and age effects, contributed equivalently to overall mortality differences from 1900 to 1920.

There are two exceptions to the foregoing pattern, in which standardized rate differences were particularly influential, regardless of place and age effects. The relative contribution of rate differences surpassed composition factors in 1906–1907 and again in 1918. The spike in 1918, coinciding with the 1918 influenza pandemic, in particular suggests a disproportionate impact of infectious disease, even after accounting for the greater proportion of the foreign-born population living in cities. Across all sets of results there is evidence that the foreign-born White population had substantially higher mortality rates during the 1918 pandemic, in both rural and urban areas. In the age-specific mortality data, this is pronounced among age brackets in the middle of the age distribution, which had smaller nativity differences in prepandemic years. The cause of the spike in 1906–1907 is unclear, although it may be related to a typhoid epidemic in 1906 and 1907 or changes in the composition of the registration area in 1906.

The rate effect, reflecting differences between groups standardized by age and place composition, increased in terms of its relative and absolute effect in the 1920s, as the place composition effect declined. It is worth noting that the composition of the foreign-born population changed throughout this period, and immigrants from Southern, Central, and Eastern Europe began to make up a larger proportion of many foreign-born age-groups (see Figure A3 in the online appendix). The rate effect also appears to decline in the mid-1920s, coinciding with the passage of immigration restrictions.

The age effect came to dominate the proportional composition effects as the older age structure of the foreign-born White population drove the higher crude rate by the 1930s. Although by 1940 the foreign-born White population had a crude mortality rate that was 13.1 deaths per 1,000 higher than that of the U.S.-born White population, 91.6% of that difference was due to population age structure. Differences attributable to place effects declined during this period, although it is worth noting that in 1940 the place distribution still contributed 0.8 deaths per 1,000 to the overall rate difference. Similarly, the remaining rate effect, assuming equivalent age and place compositions across populations, had declined but not disappeared by 1940.

Trends Toward Convergence

Temporal trends in relative inequalities can be assessed using three different measures: the age-specific rates presented in the foregoing, patterns in the decomposition analysis, and life expectancy estimates calculated from the age-specific death and population data. Table 2 depicts various period-based life expectancies, calculated using standard life table methods with the fitted age-specific mortality rates and original observed mortality data as inputs. For both men and women, life expectancy at birth was shorter for foreign-born White populations relative to U.S.-born White populations in most years. Although life expectancy at birth was slightly higher for the foreign-born populations in 1930, this appears to be driven by the lower infant mortality rate for immigrants, which may be due to either measurement challenges or migration selection patterns. Life expectancy calculations at other ages in 1930 are lower for the foreign-born population.

Although mortality rates were relatively similar for middle-aged foreign-born and U.S.-born White populations when looking at age-specific rates, higher death rates at later ages resulted in lower foreign-born life expectancy at 25, 45, and 65 even into the 1950s. In 1900, the nativity gap in life expectancy at birth was 0.7 years for men and 2.5 years for women, based on the fitted Lee–Carter estimates. By 1960 it had fallen to 0.3 years for men and 0.7 for women.

Across the measures, relative inequalities in mortality rates for the foreign-born White population continued for much of the early 1900s, with rates beginning to converge for some age-groups by the 1950s, although it is important to note a gap in data between 1941 and 1950. However, relative inequalities fluctuated, even for the 25–34 age-group, for which there was a foreign-born mortality advantage in some years. For instance, the relative nativity inequality for this group was highest in 1918, during the influenza pandemic, and appears to have increased after the 1930s. Despite trends toward convergence, at the end of the study period in 1960, there was no evidence of a foreign-born mortality advantage commonly seen in more contemporary studies of immigrant health.

Discussion

This article documents trends in mortality by nativity status among the White population in the United States in the twentieth century. By using Lee–Carter mortality models to estimate age-specific mortality rates from noisy annual data, this study provides demographic mortality estimates and life tables by nativity, allowing for a closer examination of variation in mortality patterns across age-groups, sex, time, and place.

Contrary to contemporary research that has frequently found an immigrant mortality advantage, this study found higher mortality among the foreign-born White population in the early and middle parts of the twentieth century. This is consistent with other studies that have found higher rates of child mortality and overall mortality for immigrants in the early 1900s (Bakhtiari 2018; Dribe et al. 2020; Lieberson 1980; Preston and Haines 1991).

Several plausible mechanisms for the contemporary immigrant mortality advantage may have been absent, or operated differently, in the early 1900s. This study was not able to directly test for selection effects, but existing literature returns mixed results about the direction of initial selection and return migration during this period, with some evidence of negative selection for the early waves of migrants from Northern and Western Europe (Bostean 2012; Turra and Elo 2008). However, if negative selectivity was responsible for the foreign-born mortality penalty during this period, it would have likely been most evident among the middle of the age distribution with the largest proportion of new arrivals, but that was not the case. Even if selection shaped initial patterns, the epidemiological context likely accelerated postmigration changes to population health profiles.

Similarly, differences in health behaviors, particularly smoking and diet, play a role in contemporary patterns of immigrant health and mortality (Blue and Fenelon 2011; Fenelon 2013; Lariscy et al. 2015). Although this study was not able to directly test for behavioral influences, the epidemiological context of the early twentieth century suggests that this category of health behaviors was less relevant to overall health and mortality. Infectious disease represented a larger source of mortality, and the risk factors for disease and early mortality differed substantially across eras.

Although literature related to the contemporary immigrant mortality advantage often characterizes the pattern as paradoxical because immigrants experience other risk factors and social determinants for earlier mortality, the finding of higher mortality for immigrants in the early twentieth century is in some ways unsurprising. The patterns of mortality outcomes align with their “middle tier” ethnoracial status in a way that is not seen among contemporary immigrants. In a context of high risk of death from infectious disease, initial selectivity effects on population health patterns may be short-lived, and risk factors shaping exposure to pathogens, such as immediate living and economic conditions, may be more influential than individual health behaviors.

Immigrants in this era had greater infectious disease risk exposure in part because they settled overwhelmingly in urban areas with higher overall mortality, and they disproportionately lived in crowded and substandard housing that facilitated the spread of infectious disease (Ager et al. 2020). The results suggest that this place-based distribution of the foreign-born population contributed to nearly half of the overall mortality difference by nativity in 1900, and its effect declined in subsequent decades.

However, the urban concentration of immigrants did not fully account for the nativity differentials in mortality. Patterns in the remaining rate differences suggest that immigrants had a greater mortality risk from infectious disease across contexts, particularly during periods of epidemics. Relative inequalities between the foreign-born and U.S.-born White populations were highest during the 1918 influenza pandemic, with disaggregated data suggesting a foreign-born penalty across urban and rural areas and among midlife age-groups in which typical nativity inequalities were the smallest.

The foreign-born mortality penalty for White immigrants did decline toward the middle of the twentieth century, and by 1960 the foreign-born rate was equivalent or lower for some age-groups. There were three important changes during this period that may have contributed to this shift. First, the epidemiological context changed, as the urban mortality penalty reversed and infectious diseases came to represent a minor portion of overall mortality (Armstrong et al. 1999; Haines 2001). The decomposition analysis suggests this may have contributed to the mortality convergence, as the effect of place declined over time. Second, one of the many consequences of restrictive immigration policies in the 1920s may have been greater selection pressures as barriers to immigration increased. Although not directly measured by this study, inequalities in the standardized mortality rate began to decline in the mid-1920s, when major immigration restrictions were enacted, which could be plausibly linked to increased positive selectivity. Third, social changes may have also played a role. Although European immigrant populations were considered a “middle tier” in the ethnoracial hierarchy for a period, by the middle of the twentieth century they were more explicitly considered part of the White majority and benefited from various economic and social advantages associated with that change (Barrett and Roediger 1997; Brodkin 1998; Ignatiev 2009; Jacobson 1999; Roediger 2006).

This study did not fully account for the change in mortality rates across the period, nor why a foreign-born advantage had not emerged by the 1960s. One possible explanation may be related to an additional demographic consequence of immigration restrictions. The foreign-born population not only was older, but also had a longer average duration of residence in the United States by the middle of the twentieth century. Even among contemporary immigrant populations that exhibit a health and mortality advantage, the pattern tends to diminish with greater duration of residence. There also may be some unique selection or health behavior patterns shaping European immigrant health. Some contemporary analyses have found that European immigrants exhibit smaller life expectancy advantages than other immigrant populations today, although the reasons for that are unclear (Mehta et al. 2016; Singh and Hiatt 2006).

This temporal analysis is limited by two important unavailable variables. First, the Vital Statistics data report deaths for all White immigrants, making it impossible to disaggregate by ethnic or origin-country differences. Relative inequalities in childhood mortality varied by origin country and region (Bakhtiari 2018; Dribe et al. 2020), and there is reason to expect similar variation in overall mortality. A second omitted variable is cause of death. This analysis focused on overall mortality and therefore did not directly test the effects of infectious disease on the relative mortality penalty or its variation across time and place.

Although this study is historical in scope, the results suggest the immigrant health and mortality advantage may vary by type of disease or cause of mortality, and immigrant population risk factors may warrant more attention and resources during periods of infectious disease risk. This “infectious disease exception” to the immigrant mortality advantage may also apply to contemporary contexts in which infectious disease remains prevalent or during periods of epidemic transmission. For instance, there is emerging evidence that some immigrant populations did not exhibit an advantage during the COVID-19 pandemic in the United States and other immigrant destination countries, and in fact foreign-born populations may have had higher rates of infection, hospitalization, and death (Clark et al. 2020; Horner et al. 2022; Labberton et al. 2022). Whereas the life expectancy at birth for the U.S. Hispanic population was previously three years higher than that of the non-Hispanic White population, driven largely by the foreign-born mortality advantage, in 2020 that gap shrank by two years (Andrasfay and Goldman 2021).

This study examined nativity differences in mortality, but there are other important dynamics to immigrant population health patterns, such as changes with duration of residence and across generations. As more historical mortality records and full-count census files become digitized and available for analysis, future research can provide additional answers about how both epidemiological and social characteristics of the early 1900s affected the health and mortality of immigrants and their descendants.

Notes

References