Researchers and policymakers have directed increasing attention and resources to addressing the longstanding Black-White disparities in infant health. Non-Hispanic (NH) Black infants experience the highest rates of low birthweight, preterm birth, and infant mortality of any racial/ethnic subgroup in the United States (Underwood et al., 2021, Culhane and Goldenberg, 2011, Green and Hamilton, 2019). These disparities persist even after accounting for birthing parents’ individual-level characteristics and health behaviors, signaling the need to better understand how structural factors drive infant health. Identifying the most salient structural factors and assessing the magnitude of their impact is essential to reducing racial/ethnic disparities in early-life health and thus addressing the long-term consequences of these disparities on adult economic status and well-being.

In the present study, we focus on one such structural factor: Black-White racial residential segregation (i.e., segregation). Segregation refers to the degree of spatial separation between racial/ethnic groups. NH White and NH Black Americans comprise 60 percent and 12 percent of the U.S. population, respectively (2020 U.S. Census Bureau). Yet the typical White American lives in an area that is 71 percent White and the typical Black American lives in an area that is 45 percent Black (2014–2018 American Community Survey). These patterns did not occur by happenstance: both private and governmental discriminatory processes served to segregate Black and White Americans for more than a century. While average segregation levels have declined over the past several decades, this overall trend obscures extreme racial segregation levels in certain U.S. metropolitan areas (e.g., Detroit), as well as the increasing spatial isolation between poor and affluent communities. Due to segregation, Black Americans are more likely than White and Asian Americans to live in impoverished neighborhoods with reduced access to high-quality schooling and other important resources. In sum, Massey (2020) described segregation as “a kind of spatial glue that holds a [racial] stratification system together and intensifies its effects” (p. 1).

Segregation’s negative impacts on a neighborhood’s socioeconomic status and resources likely partially explain the strong association between segregation and health inequalities throughout the life course. For example, racial segregation is correlated with racial gaps in childhood asthma and early-life health outcomes such low birth weight (Alexander and Currie, 2017, Mehra et al., 2017). Further, one study found that Black children who consistently lived in segregated neighborhoods were more likely to experience worse adult health and more likely to smoke or drink than those who moved in and out of highly segregated neighborhoods (Schwartz et al., 2022). In contrast, studies have generally found no effects, or sometimes even positive effects, of segregation on health outcomes among White populations, including White infants (Kramer and Hogue, 2009).

Despite the large body of evidence linking residential segregation and infant and child health, very few studies have assessed whether these links are causal. Such assessments are difficult because unobserved area-level characteristics simultaneously affect both segregation and health outcomes for Black infant populations. For example, structural racism or racial animus may simultaneously influence the level of segregation within a region and the health outcomes of its infant population. Failing to account for these omitted variables produces biased estimates from OLS or even fixed effects models.

To address these potential biases, we use an instrumental variables (IV) identification strategy (Ananat, 2011), leveraging plausibly exogenous historical railroad divisions to estimate the causal impacts of residential segregation on the health of an area’s NH Black and NH White infant populations. Specifically, we leverage the fact that, in the early 19th century, train companies laid thousands of miles of railroad tracks in metropolitan areas to transport goods and passengers between cities. The arrangement of these tracks created ready-made neighborhood subdivisions, which local policymakers used to establish and enforce racially segregated neighborhoods in response to large influxes of Black migrants from the U.S. South (Ananat, 2011). These patterns have persisted to the present day, and thus railroad lines often define the boundaries between Black and White communities (see Fig. 1). In other words, although train companies did not choose track configurations to facilitate racial segregation, they were a key technology that enabled segregation (Ananat, 2011). Importantly, we argue that the historical placement of railroad tracks is unrelated to contemporary birth outcomes and can serve as an exogenous instrument for estimating the causal effects of segregation on health.

The main IV results show that plausibly exogenous exposure to residential segregation has significant negative impacts on the birth outcomes of NH Black infant populations in the area. Specifically, a 1 SD increase in racial segregation (equivalent to a 13.6% change) leads to a 49 g decrease in birth weight, a 1.2 percentage point increase in the likelihood of low birth weight, a 0.3 percentage point increase in the likelihood of very low birth weight, a 1.7 percentage point increase in being categorized as small for gestational age, a 0.16 week decrease in gestational length, a 1.9 percentage point increase in the likelihood of preterm birth, and a 0.95 gram per week decrease in fetal growth among Black infant populations. The OLS results were generally smaller in magnitude than the IV results. This pattern suggests that analyses that fail to account for the endogeneity of segregation likely underestimate its consequences for infant health. Importantly, we find few to no effects on White infant populations. Finally, we explore structural pathways through which segregation could potentially affect the health of infant populations in a given area. The analyses suggest that segregation is linked to lower access to prenatal care during the first trimester, higher levels of anti-Black prejudice, greater transportation barriers, and increased area food insecurity (particularly among Black populations), findings consistent with those of prior studies linking each of these mechanisms to poor birth outcomes (Simonovich et al., 2020, Grilo et al., 2022, Orchard and Price, 2017). Taken together, the findings contribute to a growing literature documenting the causal impacts of racial segregation on the health and economic outcomes of Black American populations.