Racialized Police Use of Force and Birth Outcomes

Abstract

Objective:

To estimate associations between police use of force (PUOF) in local jurisdictions and birth outcomes of Black women compared to White women.

Methods:

Using birth records linked to municipal police department data on PUOF incidents, we estimated associations between overall and racialized PUOF and birth outcomes of 75,461 Black women and 278,372 White women across 430 municipalities in New Jersey (2012-2016).

Results:

Overall PUOF was not associated with birth outcomes of Black or White women. A 1% increase in racialized PUOF was associated with a .06% increase in the odds of low birth weight (β: 0.06; 95% CI: 0.03-0.09) and preterm birth (β: 0.06; 95% CI: 0.03-0.10) among Black women but had no associations with birth outcomes of White women.

Conclusions:

Lack of associations between overall PUOF and birth outcomes coupled with significant associations between racialized PUOF and birth outcomes among Black women only is consistent with mounting evidence that structural racism adversely affects the health of Black people in the U.S.

Policy Implications:

The findings point to the need to address health inequalities at the structural level.

INTRODUCTION

Rates of low birth weight and preterm birth among infants born to Black women have been twice those of infants born to White women in the U.S. for decades (Artiga, et al., 2020; MacDorman & Mathews, 2011; Mathews & Driscoll, 2017). These large and persistent disparities in infant health have implications for disparities in cognitive development, educational achievement, earnings, and other long-term outcomes (Case, et al., 2004; Conley & Bennett, 2000; Figlio, et al., 2014; Lambiris, et al., 2022). Differences in maternal education, income, occupation, health insurance, and prenatal health behaviors explain little of the observed racial disparities in birth outcomes (Reichman & Teitler, 2013).

Few studies have explored the roles of racism or discrimination—particularly institutional racism—in explaining the much poorer birth outcomes of Black women compared to White women, despite an increasing recognition that entrenched social disadvantages associated with racial minority status in the U.S. take a toll on health (Braveman, et al., 2022; Noonan, et al., 2016). Some of the obstacles to studying the effects of institutional racism on health have been difficulties in constructing measures that convincingly reflect institutional racism and finding appropriate population-level micro-level data on health that can be linked to those measures.

Emerging research has found adverse associations between violent policing and birth outcomes of Black women (Curtis, et al., 2022; Freedman, et al., 2022; Goin, et al., 2021; Hardeman, et al., 2020; Jahn, et al., 2021). Potential mechanisms underlying the associations include direct effects through harm to pregnant women (which would be very rare) or, more plausibly, indirect effects. One person’s experiences can vicariously affect acquaintances or other community members (Heard-Garris, et al., 2018; Moody, 2022), and policing practices that result in high rates of incarceration could lead to changes in the social fabric of communities (e.g., by removing significant numbers of young men and constraining their opportunities when they are released) (Roberts, 2003). Finally, when violent policing is racialized (i.e., disproportionately targeted to Black subjects), it may reflect general levels of racism in communities that permeate local policies and institutions affecting residents’ housing, healthcare, child care, and education, all of which can affect human capital and health.

Existing studies of police violence and birth outcomes have generally focused on indirect, but not race-specific, exposures and differential effects on Black and White women. One study found small and positive associations between fatal police violence (from the Fatal Encounters database and death records, linked to birth records) in the mother’s census tract or within a specified physical distance from her residence and preterm birth (PTB) in California, and it appeared that the associations were stronger for Black women than for White women (Goin, et al., 2021). Another investigated the extent to which fatal police violence in core-based statistical areas of the U.S. census (and counties, which are smaller) was associated with pregnancy loss (Jahn, et al., 2021). Specifically, the authors estimated associations between month-to-month fatal police violence and conceptions resulting in live births and found that exposure to police killings during pregnancy was associated with lower birth rates, especially for Black women; they attributed this finding to increases in pregnancy loss. A third study of fatal police violence and birth outcomes created exposure measures involving 49 high publicity incidents of police lethal force toward Black persons, legal decisions not to indict/convict officers involved, and hate crime murders of Black victims in the U.S. using Google Trends (Curtis, et al., 2022). The authors found some associations between exposure to high publicity incidents and higher odds of PTB, with stronger associations for Black women than for White women, but the findings were not robust to alternative model specifications and their race-specific measures of lethal force involving Black victims were at too broad a geographic level (national) to reflect community-level factors.

Two studies from U.S. cities focused on measures of police contact more broadly. The first found that higher levels of community-level police contact were associated with higher rates of PTB for both Black and White women in Minneapolis (Hardeman, et al., 2020). The authors also found that individuals in predominantly Black neighborhoods had more police contact than those in predominantly White neighborhoods but they did not specifically measure racially-disproportionate rates of police contact within neighborhoods. The second investigated associations between formal complaints filed for excessive use of force against police in Chicago and cardiovascular disease and birth outcomes of Black women from one hospital (Freedman, et al., 2022). The authors found that Black women had a 19% higher risk of PTB, 16% higher risk of delivering a small-for-gestational age infant, and a 42% higher risk of cardiovascular disease (which is associated with adverse pregnancy outcomes) in census block groups where residents frequently complained about the police using excessive force, even after accounting for neighborhood socioeconomic conditions and homicide frequency.

In this study, we expand upon this literature by using unique linked data from the state of New Jersey (NJ) to investigate associations between measures of police violence at the community level and individual-level birth outcomes of Black women compared to White women. The state of NJ has one of the largest numbers of municipal police departments in the country (Reaves, 2011), one of the highest numbers of police officers per resident (55% higher than the national rate) (Reaves, 2011), one of the largest racial disparities in incarceration in the nation (Mauer & King, 2007), and has been collecting data on police use of force. NJ is therefore a highly suitable state for investigating associations between police practices and disparities in birth outcomes.

We consider measures of overall and racially-disproportionate police use of force in municipalities, which generally have their own police departments that implement policies and provide oversight. We hypothesized that: (1) Racially-disproportionate police use of force (PUOF) against Black people is adversely associated with birth outcomes of Black women but not White women. (2) Overall PUOF, which is not necessarily racialized, is adversely associated with birth outcomes of Black women, but less so than racially-disproportionate PUOF is.

METHODS

Data

We pooled records on all births in the state of NJ from 2012-2016, which included maternal residential addresses under a restricted data use contract with the NJ Department of Health. Using the mother’s municipality and county of residence, we linked the birth records to information on PUOF incidents that involved local or NJ state police officers in each of 468 municipal police departments in NJ during the same period.

The information on PUOF incidents was from The Force Report (FR) dataset, which contains information from >70,000 use-of-force forms completed by NJ police officers and compiled by NJ Advance Media. The data include reports of all violent encounters between officers and civilians in every police department in NJ from 2012-2016 (NJ Advance Media, 2019). The recording of these encounters was mandated by the NJ Attorney General’s Office for all incidents that involved compliance holds, takedowns (leg sweeps or tackles), strikes and punches, leg strikes, baton strikes, pepper spray, or deadly force (e.g., firing a service weapon). The most common types of force in the data were compliance holds (81%) and punches and hand strikes (28%). Discharging of weapons was rare (0.4%). Our measures of use of force were computed using the total number of incidents involving any type of force within each police department. There were 8441 reported incidents of PUOF in NJ in 2012, 8455 in 2013, 8381 in 2014, 8054 in 2015, and 8121 in 2016.

We used police department-level information on racial-ethnic proportion of PUOF subjects, average number of PUOF incidents per officer, size of police force, and an age-adjusted measure of racial-ethnic distribution of residents in their jurisdiction that the Force Report team constructed using 2012-2016 population data from the American Community Survey (ACS). We also linked municipal-level information published by the NJ Department of Labor on resident population from the 2010 decennial U.S Census (DEC Redistricting Data - PL 94-171), median household income from the 2010 ACS 5-year estimates, and 2012-2016 city and town violent crime data published by the Criminal Justice Information Services Division of the U.S Federal Bureau of Investigation (FBI). Violent crimes in the FBI’s Uniform Crime Reporting Program are offenses that involve force or threat of force, and include murder and nonnegligent manslaughter, rape, robbery, and aggravated assault. We merged all of these data elements to the individual-level birth records using identification variables for municipalities or county subdivisons (boroughs, cities, and towns).

A total of 501,689 singleton births took place in the state of NJ in 2012-2016. Multiple births were excluded because they are more likely than singletons to be low birth weight and preterm and it would have been necessary to address that additional source of clustering. Of the 501,689 births, 500,904 (99.8%) had non-missing data on maternal county and city of residence. Of those, we excluded 30,417 that were to non-residents of NJ or to mothers whose residence was not in one of the municipalities in NJ that had a police department or was served exclusively by State police. Eight municipalities were under the jurisdiction of police departments that served multiple municipalities; in those cases, the consolidated police department information was used.

To test our hypotheses, we limited the sample to births to the 375,167 Black or White women who lived in municipalities with at least one birth to a Black woman and at least one birth to a White woman. Of those, we excluded 120 births (0.03%) because of missing information on birth weight or gestational age and another 21,214 (5.6%) because of missing data on maternal or municipal-level factors. The final analysis sample consisted of 353,833 births to women living in 430 municipalities; of those, 75,461 (21.3%) were to Black women and 278,372 (78.7%) were to White women.

Measures

Outcomes

Outcomes were low birth weight (LBW) (<2500 grams) and PTB (<37 completed weeks of gestation), which have become standard markers of poor infant health in population-based research studies and are highly predictive of long-term health and economic outcomes (Case, et al., 2004; Conley & Bennett, 2000; Figlio, et al., 2014; Lambiris, et al., 2022). These measures were constructed from indicators of birth weight in grams and clinical gestational age in completed weeks that were available in the birth records. The reporting of birth weight is highly accurate in US birth records, including in NJ (Reichman & Hade, 2001). The clinical assessment of gestational age in the NJ birth records was “judged by the clinician using the best available information (physical examination of the infant and/or ultrasound visualization) (p. 53)” (Bolden & Mammo, 2018).

Exposures

We created two measures of PUOF:

1. PUOF incidents per officer was the number of incidents in a police department jurisdiction, by municipality (m), divided by the number of officers in that department. This measure captures the level of aggressiveness of the average police officer in the department.

   $${\mathit{PUOF\; incidents\; per\; officer}}_m=\frac{{\#\mathit{of\; PUOF\; incidents}}_m}{{\#\mathit{of\; police\; officers}}_m}$$
2. Racially-disproportionate PUOF was the ratio of the number of incidents involving Black subjects to the number of incidents involving White subjects, divided by the ratio of Black to White residents in the municipality.

   $${\mathit{Racially\; disproportionate\; PUOF}}_m=\frac{{\left(\frac{\#\mathit{of\; PUOF\; incidents\; involving\; Black\; subjects}}{\#\mathit{of\; PUOF\; incidents\; involving\; White\; subjects}}\right)}_m}{{\left(\frac{\#\mathit{of\; Black\; residents}}{\#\mathit{of\; White\; residents}}\right)}_m}$$

This relative risk reflects the extent to which PUOF was disproportionately experienced by Black subjects. In supplementary analyses, we used an alternative measure of municipal PUOF—the number of incidents per capita, which is a measure of population exposure to PUOF. The exposure measures were intended to characterize police behavior in communities so they include incidents whether or not the subjects resided in those communities.

Covariates

Maternal characteristics, from the birth records, included parity (first birth), Hispanic ethnicity, age, education, and Medicaid coverage for the birth. Municipality-level variables, from the sources cited earlier, included 2010 median household income, 2010 resident population, and 2012-2016 average rate of violent crimes.

Statistical Analyses

First, we investigated the distributions and stability of the PUOF measures by documenting the ranges of the two measures and calculating Pearson correlation coefficients for the measure of incidents per officer over time within municipalities. We then documented the correlation between the PUOF measures and the racial composition of communities using Spearman’s rank-order correlations. Since our interest is on the effects of community-level policing, we were particularly interested in the extent to which PUOF varied across municipalities, how stable it was over time within municipalities, and how much it varied by the racial composition of municipalities.

Second, we estimated a series of race-stratified logistic regression models that adjusted for the individual and municipal-level covariates listed above and accounted for the data structure by clustering standard errors at the municipality level. Because the distributions of PUOF measures were highly skewed, we used linear logarithmic transformations (in this case, Log(x+0.001)), which de-emphasize outliers, to improve model fit and avoid losing zero-value observations. The density plots of the log-transformed variables approximated a normal distribution (not shown). Logistic regression coefficients of log-transformed measures are elasticities—the percentage change in the odds of the dependent variable associated with a 1% change in the (untransformed) measures.

Finally, we estimated supplementary models that used quartiles of municipal PUOF and PUOF incidents per capita as alternative exposure measures; excluded municipalities with their own police departments but that reported zero incidents of PUOF; limited the sample to municipalities with >= 5000 residents; limited the sample to municipalities with >=20 births to Black women and >=20 births to White women; included year fixed effects; and used alternative functional forms.

RESULTS

Compared to White women, Black women had significantly higher proportions of LBW and preterm infants, lower levels of education, and higher rates of Medicaid-financed deliveries (Table 1). They also resided in municipalities with higher crime rates, lower median household incomes, more incidents of PUOF per capita, and more aggressive officers as characterized by PUOF incidents per officer. However, Black women lived in municipalities with lower rates of racially-disproportionate PUOF, which was higher in municipalities with higher proportions of White residents.

Table 1:

Individual and municipal-level characteristics of sample by mother’s race, 2012-2016

	Black Women	White Women
Birth Outcomes
Low Birth Weight	9.7	4.7
Preterm Birth	10.4	6.7
Maternal Characteristics
First Birth	38.8	39.2
Hispanic	15.5	31.8
< 20 years	8.2	3.1
20 to 35 years	74.8	73.3
35+ years	16.3	22.6
High School or less	51.7	35.5
Some College	26.6	20.2
College Graduate	21.7	44.3
Medicaid	54.9	27.8
Municipal Characteristics
Number of PUOF Incidents per Officer	2.6	2.2
Racially-disproportionate PUOF	3.1	9.6
PUOF (per 1000 residents)	7.7	5.2
Violent Crimes (per 100,000 residents)	690	276
Median Household Income ($1000s)	52.5	69.7
N	75,461	278,372

Notes: Figures are column percentages unless indicated otherwise. Municipal characteristics are weighted by number of individuals in the sample; e.g., on average, Black women resided in municipalities with 2.6 PUOF incidents per police officer. Among municipalities with at least 20 births per year, on average, the rate of low birth weight ranged from 0.8% to 9.6% and the rate of preterm birth ranged from 2.5% to 11.9%.

Average annual municipality-level number of incidents of PUOF per officer ranged from close to 0 to 12 (Figure 1). The annual municipality-level racially-disproportionate PUOF measure (the relative risk of PUOF involving Black subjects compared to White subjects) ranged from nearly 0 to 2,943; i.e., municipal police departments ranged from no observed racially-disproportionate PUOF to extreme racial disproportionality in PUOF.

Figure 1: Rank-ordered municipality-level police use of force (PUOF) measures, NJ, 2012-2016.

*6 Municipalities with racially-disproportionate PUOF values of 160,172, 223, 269, 523, and 2943 are ommited to improve the readability of the figure.

Within-municipality correlations of the number of incidents of PUOF per officer across the 5 years ranged from .59 to .80 and were highest for adjacent years, indicating that the measure was fairly stable from year-to-year (Table 2).

Table 2:

Correlation of within-municipality police use of force per officer over time

	2012	2013	2014	2015	2016
2012	1
2013	0.72*	1
2014	0.70*	0.79*	1
2015	0.65*	0.70*	0.80*	1
2016	0.59*	0.68*	0.69*	0.74*	1

^*p < 0.01. Pearson Correlations

Observations: 2,300

PUOF was also correlated with racial composition of municipalities. The percentage of residents that were Black was positively correlated with higher PUOF per officer (r_s = 0.43) but negatively correlated with racially-disproportionate PUOF (r_s = −0.26).

Logistic regression models of associations between PUOF and birth outcomes, adjusted for individual-level characteristics as well as municipality-level median household income, rate of violent crimes, and population size, indicated no significant associations between incidents per officer and birth outcomes for Black women (LBW: β: 0.04; 95% CI: −0.01 to 0.09; PTB: β: −0.00; 95% CI: −0.06 to 0.05 ) or White women (LBW: β: 0.01; 95% CI: −0.04 to 0.06; PTB: β: 0.00; 95% CI: −0.05 to 0.05) (Figure 2). However, in corresponding models using racially-disproportionate PUOF instead of incidents per officer, PUOF was significantly associated with worse outcomes for infants born to Black women; specifically, a 1% increase in the racially-disproportionate PUOF was associated with a .06% increase in the odds of LBW (β: 0.06; 95% CI: 0.03 to 0.09) and a .06% increase in the odds of PTB (β: 0.06; 95% CI: 0.03 to 0.10). There were no significant associations between racially-disproportionate PUOF and birth outcomes of infants born to White women.

Figure 2: Adjusted logistic regression estimates of associations between municipality-level police use of force (PUOF) measures and birth outcomes by maternal race.

* Logistic regression models of associations between PUOF and birth outcomes adjusted for parity (first birth) and maternal Hispanic ethnicity, age, education, and Medicaid coverage, as well as municipality-level median household income, rate of violent crimes, and population size.

Supplementary models using the alternative measures of PUOF (number of incidents per capita) produced results substantively similar to those when using number of incidents per officer (i.e., no significant associations between PUOF and birth outcomes for Black or White women). The estimates were also insensitive to using quartiles of the PUOF measures instead of the logged transformations, excluding municipalities that reported zero incidents of force during the study period, restricting the sample to municipalities with >=5000 residents, restricting the sample to municipalities with >= 20 births to Black women and >=20 births to White women, including year fixed effects, and using alternative functional forms (probit and Ordinary Least Squares models) (results not shown).

DISCUSSION

In this population-level study of New Jersey municipalities, we found significant and robust adverse associations between racially-disproportionate PUOF and birth outcomes of Black women, but no associations between overall PUOF and birth outcomes or between racially-disproportionate PUOF and birth outcomes of White women. Municipal police departments were highly consistent over time in their use of force and there was substantial variation across municipalities in both use of force per police officer and disproportionate PUOF directed at Black subjects. These findings suggest that living in communities with violent policing that is racialized takes a toll on Black women’s health that crosses generations, while living in communities with violent policing that is not racialized does not have more adverse effects on birth outcomes for Black women than for White women.

On average, Black women in our sample resided in municipalities where PUOF involving Black subjects was 8.2 times that of PUOF involving White subjects. Our estimate of the effect of racially-disproportionate PUOF suggests that the elimination of racialized PUOF would reduce the overall rate of LBW among Black women in NJ from 9.7% to 9.3% and the Black-White gap in LBW by 8% (from 5 to 4.6 percentage points) (see Supplemental Material, Appendix 1 for calculations).

The null associations for overall (not specifically racialized) PUOF and birth outcomes, the robust associations between racially-disproportionate PUOF and birth outcomes for Black women, and the lack of associations between racially-disproportionate PUOF and birth outcomes for White women are consistent with racially targeted policing having indirect effects on birth outcomes, but could also reflect broader community values, norms, and tolerance of discrimination and racism and their infiltration into policies, institutions, and organizations. In the latter case, the associations between racialized PUOF and birth outcomes of Black women could reflect broad exposure to structural racism of which discriminatory policing may be a part. Although we cannot identify the precise sources of the associations, the findings point to institutional factors that are associated with racial disparities in health and are consistent with previous theoretical and empirical research suggesting that racialized policing, as a core aspect of structural racism in the U.S., adversely affects the health of Black people (Alang, et al., 2017; DeVylder, et al., 2022; Sewell & Jefferson, 2016; Sewell, 2017). We were also unable to investigate biological pathways linking racial disproportionality in PUOF and birth outcomes in this population-level study. Findings from other studies suggest that stress resulting from racial profiling and real and perceived disproportionate risks of facing police violence can have adverse and cumulative effects on health (Braveman, et al., 2022; Krieger, 2012; Sullivan, 2013; Calvin, et al., 2003).

The findings from our study are generally consistent with, but add to, the small existing literature on police violence and birth outomes by focusing on any PUOF and not just fatal incidents, investigating associations between police violence and birth outcomes at the municipality level within an entire state, and considering both overall PUOF and racial disproportionality in PUOF, the latter of which we argue may reflect structural racism. From a more practical standpoint, the data used and measures created for this study can be used to identify communities with high rates of racialized policing and target interventions.

While fatality at the hands of police is an extreme outcome of interaction with police, only a small fraction of incidents of police violence are fatal and those that are not fatal can induce substantial harm, fear, and stress (Geller, et al., 2014; Hirschtick, et al., 2019). Moreover, because fatal police violence is relatively rare, rates are less stable over time, necessitating measures of exposure with precise timing relative to pregnancy. Finally, studies focusing on fatalities address acute effects of police violence and not overall effects of chronic police or community-level racism. Detailed discussion of the limitations of fatal police violence data is in Klinger (2012). Only two of the previous studies focused on police violence beyond fatal encounters, one in census block groups in a major U.S. city (Freedman, et al., 2022) and the other in census tracts in another large U.S. city (Hardeman, 2020).

One of the challenges of studying effects of structural racism on health is identifying appropriate geographical units of aggregation (Hardeman, et al., 2022). Our study is unique in that it focused on exposure to PUOF at the municipal level, which corresponds more closely than any census-defined areas to police department jurisdictions. Census tracts and block groups, the most widely used geographical levels in the literature on police violence and health, are statistical subdivisions of counties that are designed for census taking and serve no governmental functions, whereas municipalities are governmental jurisdictions within states (and usually within counties) that directly provide services including education, transportation, public health, recreation, and police. As far as we know, no previous studies of police violence and infant health or health more broadly have focused on exposures to police violence at the municipal level. Because municipal police forces are operationally and geographically linked to other administrative and social services that reflect preferences of voters and tolerance or preference for racism at the community level, we argue that municipalities are the appropriate geographic unit for studying consequences of racially-disproportionate police violence and that measures of racialized policing in municipalities are likely to be good indicators of racism in communities more generally.

Two previous studies of police violence and health outcomes considered racial disproportionality in the use of force (Sewell, 2017; Sewell & Jefferson, 2016). Focusing on exposures to not only overall levels of PUOF but also to racial disproportionality in PUOF allowed us to differentiate between estimated effects of policing practices in general and more refined indicators that arguably reflect systemic and structural racism in the community. We found that racialized PUOF, but not overall PUOF, was associated with adverse birth outcomes and then only for Black women. Our measure of racialized policing was somewhat different than that used in the Sewell et al. studies as it captured the relative risk of the Black population in a municipality being subject to a stop involving use of force compared to the risk of the White population being subject to a stop involving use of force, whereas the Sewell et al. studies focused on racial disproportionality in police violence among people who were stopped by police. As such, their measure did not account for the very real possibility that Black and White subjects were differentially exposed to police stops. Despite differences between those studies and ours in how racialized policing was operationalized, the different geographical units considered (they focused on hospital service areas within New York City, whereas we focused on municipalities within a state), and health outcomes examined (they focused on self-rated health, diabetes, blood pressure, and body mass index, while we focused on provider-assessed birth outcomes), their bottom line findings were consistent with ours in that racialized policing was adversely associated with health.

Despite the strengths of our PUOF measures and the focus on municipalities within an entire state, our study has limitations. First, our data are limited to NJ and may not be generalizable to other states. Second, although the FR data were based on reports that officers are required to file after any incident involving force, there is some evidence that officers fail to systematically and accurately document use of force in encounters with civilians and during arrests (Moore, et al., 2018). If PUOF was systematically underreported in our data, the resulting measurement error could have biased our estimates of associations between PUOF and birth outcomes, although the expected direction of bias is not obvious.

Third, some readers might question our interpretation of the study findings, arguing that we should have accounted for the need to use force—e.g., in order to subdue individuals engaged in violent crimes. While PUOF may be justified in some circumstances, it should be limited to preventing immediate harm (to an officer or someone else) by the subject, which is very rare. The Use of Force Policies of the Office of the Attorney General in the State of NJ are clear in stating that PUOF should be used only as a last resort and when absolutely necessary, regardless of the type of crime (Office of the Attorney General, 2022), and most incidents of PUOF in our data involved punches and hand strikes with at least 40%-50% of cases occurring in the course of minor incidents such as traffic stops and violations, suspicion, and disorderly conduct. Additionally, we controlled for violent crimes in the municipality in our models and we also found that racially-disproportionate PUOF tended to be higher in predominantly White communities, which tend to have lower rates of violent crime than predominantly Black communities. For all of these reasons, we argue that it is implausible that the variation in Black/White disproportionality in PUOF across municipalities reflected justifiable and consistently-applied use-of-force protocols.

PUBLIC HEALTH IMPLICATIONS.

Low birth weight and preterm birth are markers of poor infant health that presage lifelong health disadvantages that can severely impact quality of life and are costly for society. Much as physical environments can take a toll on health, so too can social environments that embed structural racism. In this study, we found that racialized police violence was adversely associated with birth outcomes of Black women. We cannot ascertain whether the associations reflect effects of policing or whether police violence is an indicator of other community-level exposures, but either way, the findings point to the need to address health inequalities at the structural level.

Highlights.

• Police use of force (PUOF) varies considerably across municipalities

• Racialized PUOF may reflect structural racism

• Overall PUOF is not associated with birth outcomes of Black or White women

• Racialized PUOF is adversely associated with birth outcomes of Black women

• Findings suggest that structural racism adversely affects health of Black women