How Does Residential Segregation Affect Homicide Rates Among Non-White Americans?

This study investigates whether residential racial segregation causally increases homicide victimization among non-white Americans. Using a two-stage least squares strategy that exploits the historical layout of railroad tracks as an instrument for segregation, the authors analyze data from 1970 to 2010 across northern U.S. metropolitan areas. They find a strong, statistically significant relationship: a one-standard-deviation decrease in the segregation index would save approximately 0.147 non-white lives per 1,000 residents—or 14 lives in a typical MSA. The study also links segregation to reduced public revenues and lower expenditures on policing and education, which likely exacerbate violent crime and homicide disparities.

Full Citation and Link to Article

Here is the full citation pulled directly from the published journal article:

Cox, R., Cunningham, J. P., Ortega, A., & Whaley, K. D. (Forthcoming 2025). Black Lives: The High Cost of Segregation. American Economic Journal: Economic Policy. https://doi.org/10.1257/pol.20220083 

Extended Summary

Central Research Question

This article investigates the causal impact of residential racial segregation on homicide victimization rates, particularly among non-white populations in U.S. metropolitan areas. The authors ask whether and to what extent the historical and structural configuration of segregated neighborhoods contributes to racial disparities in violent crime outcomes. Focusing specifically on non-white homicides, the paper evaluates if segregation—rather than broader demographic patterns or crime trends—directly increases the risk of death by violence, and whether it does so by reducing local public revenues and expenditures that could otherwise mitigate such outcomes.

Previous Literature

The study builds upon several key strands of scholarship in economics, sociology, and public policy. Foundational work by Massey and Denton (1989) described Black segregation in the U.S. as “hypersegregation,” a condition unique in scale and intensity. Prior research has linked segregation to a host of negative outcomes, including reduced upward mobility, concentrated poverty, and criminal justice disparities (Chetty et al., 2014; Andrews et al., 2017; Bayer et al., 2021). Several descriptive studies (e.g., Peterson and Krivo, 1993; Bjerk, 2006) have reported positive correlations between segregation and violent crime, but few have credibly established causality—particularly in the domain of homicide victimization. Ananat (2011) developed a compelling strategy using the historical configuration of railroad tracks as an instrument for segregation, which this paper builds upon and extends. This work also aligns with literature examining how segregation affects public finance, intergroup political cooperation, and state capacity (Alesina et al., 1999; Tabellini, 2020).

Data

The authors draw on a wide array of datasets to conduct their analysis, focusing on the period from 1970 to 2010. The primary outcome variable—homicide victimization by race—is sourced from the CDC’s National Vital Statistics System (NVSS) and the Vital Statistics Multiple Cause of Death Files. The primary independent variable, residential segregation, is measured using the index of dissimilarity, which captures the proportion of a racial group that would have to move in order for neighborhoods to be evenly distributed by race. This index is calculated for over 100 metropolitan statistical areas (MSAs), using census tract data from decennial censuses.

Additional data sources include:

• U.S. Census data for demographic and economic covariates

• FBI Uniform Crime Reports for violent crime and arrest statistics

• Vera Institute data on incarceration and prison admissions

• Census of Governments and National Center for Education Statistics for local public finance and school spending data

• Fatal Encounters and other datasets on police-related fatalities

The authors also construct historical indices of railroad track arrangements (railroad division index, RDI) based on early 20th-century maps, which serve as instruments in the causal analysis.

Methods

To estimate the causal effect of segregation on homicides and related outcomes, the authors use a two-stage least squares (2SLS) strategy. The first stage uses the railroad division index (RDI)—a measure of how extensively a city’s neighborhoods were divided by railroad tracks—as an instrument for the index of dissimilarity. Because most railroads were constructed prior to 1910, and because they strongly predict the physical separation of communities, the RDI provides a credible source of exogenous variation in segregation.

The second stage estimates the effect of the predicted level of segregation on homicide victimization rates and other outcomes, including public revenues, policing, incarceration, and school spending. The authors estimate models for each decennial census year between 1970 and 2010, reporting both ordinary least squares (OLS) and 2SLS results. Standard errors are robust, and weak instrument tests are reported.

They also examine heterogeneous effects by race, distinguishing between white and non-white homicide victims, and consider whether changes in segregation over time explain shifts in homicide rates. Various robustness checks are performed, including controls for poverty, inequality, education, and labor market conditions.

Findings/Size Effects

The core finding is that residential segregation has a large, statistically significant, and racially uneven effect on homicide victimization. Segregation significantly increases homicide rates for non-white residents but has no consistent effect on white residents.

Key findings include:

1. Effect on Non-White Homicides

   In the most robust specification, a one-standard-deviation increase in the dissimilarity index leads to a 0.97 standard deviation increase in non-white homicides, or about 0.147 additional deaths per 1,000 non-white residents. In a typical MSA of roughly 94,000 non-white residents, this implies 14 additional non-white homicides. Reducing segregation from the top to the bottom quartile could avert up to 23 non-white deaths per MSA.

2. Effect on White Homicides

   There is no statistically significant relationship between segregation and white homicide rates in any census year. In some years, point estimates are even negative, suggesting that whites may experience reduced risk in more segregated cities.

3. Violent Crime and Murder Rates

   Segregation also increases overall violent crime and murder rates, particularly in central cities. For example, a one-standard-deviation increase in segregation raises the violent crime rate by approximately 8 incidents per 1,000 residents in 1990—a 30 percent increase relative to the sample mean. The murder rate rises by 0.047 per 1,000 residents, or 44 percent.

4. Public Finance Effects

   Segregated cities generate lower public revenues, primarily due to depressed property tax collections linked to white flight and reduced property values. In 1990, a one-standard-deviation increase in segregation is associated with a $665 per capita drop in public revenue, with about half due to lost property taxes. Although the absolute amount of spending on police declines, the share of budgets devoted to policing increases.

5. Policing and Incarceration

   While the paper finds limited effects of segregation on police-related killings, it finds strong effects on incarceration. Segregation significantly raises Black imprisonment rates without affecting white imprisonment. In 1990, a standard-deviation increase in segregation led to an additional 390 Black individuals incarcerated per 1,000.

6. Education Spending

   Segregation is associated with lower per-pupil school expenditures. Given existing literature that ties educational investment to reduced crime and improved life outcomes, this channel may be an important mechanism linking segregation to homicides.

7. Temporal Trends

   Both segregation and homicide rates have declined since 1990, but the association between the two remains positive and statistically significant throughout the study period. The declining trend is attributed in part to changes in residential preferences, suburbanization of middle-class Black families, and increased integration in some areas.

Conclusion

This study provides compelling causal evidence that residential racial segregation significantly increases homicide victimization among non-white Americans. By exploiting the historical placement of railroad tracks to isolate exogenous variation in segregation, the authors establish that the effects are not merely correlational but structural and persistent. The burden of segregation is not shared evenly: it primarily results in lost lives, diminished public safety, reduced education investment, and higher imprisonment for non-white communities, particularly Black Americans.

These findings contribute to ongoing discussions about racial inequality, urban development, and structural racism. They emphasize that segregation is not simply a social or cultural artifact—it has lethal consequences. Public policies aimed at desegregation, equitable school funding, and community reinvestment are not just equity measures but potentially life-saving interventions.

Overall, the paper deepens our understanding of how geography, race, and institutional design interact to shape life outcomes. It highlights the “high cost of segregation” in the most literal sense: paid in human lives.