Racial inequities in homicide rates and homicide methods among Black and White women aged 25–44 years in the USA, 1999–2020: a cross-sectional time series study

Abstract

Introduction:

US Black women aged 25–44 are disproportionately murdered. Despite ongoing efforts to reduce racial and structural inequities, the result of these efforts remains unclear particularly in light of the COVID-19 pandemic.

Methods:

The present study examined a cross-sectional time series of homicide death rates from CDC WONDER for women, aged 25–44, by race between 1999 and 2020 among the 30 states in the US with enough events for analysis. Homicide death was classified using underlying cause and multiple cause of death codes; mortality rates were calculated per 100,000 based on US Census Bureau population sizes. Homicide methods were classified as firearm, cutting/piercing, and other.

Results:

In 2020, the homicide rates among Black women was 11·6 per 100,000; compared with 3 per 100,000 among white women. This inequity has persisted over time and is virtually unchanged since 1999. Homicide inequities vary across US states; in 11 states, racial inequities increased since 1999. The racial inequity was greatest in Wisconsin, where in 2019–2020, Black women aged 25–44 were 20 times more likely to die by homicide than white women. Death by firearm is increasing; women in the US had 2·44 [95% CI 2·14, 2·78] times the odds of homicide involving firearms in 2019–2020 compared to 1999–2003. Firearm homicide deaths are disproportionately concentrated among Black women in every region in the US.

Conclusion:

Findings suggest that more progress is needed in reducing racial and structural inequities that underpin the disproportionate homicide rate among Black women.

Introduction

Homicide rates continue to surge among women <44 years old in the United States (US) and Black women are overwhelmingly impacted.¹ Racial inequities evidenced in the rates of homicide is a long-standing public health priority.¹ Previous epidemiological studies examined these disparate rates, especially among women, through lenses that include intimate partner violence (IPV);² US state differences;¹ correlations with firearm legislation;³ gun prevalence and homicide rates;⁴ structural inequities, namely education, unemployment, economic status/wealth distribution, and home ownership;⁵ and its association with pregancy.⁶ Considering the COVID-19 pandemic and on-going efforts to reduce structural inequities, continued surveillance of homicide inequities is needed. A comprehensive analysis is a critical step toward targeting the development of effective, culturally salient interventions that reduce the likelihood of murder among Black women.

IPV is a key contributor of homicide among Black women.¹ Importantly, more than half of the women murdered in the US experienced IPV before they were killed.¹ Black women are murdered at younger ages and higher rates than other women of color, including Native American and Alaska Native women.⁷ As of 2020, Black women are not only nearly three times more likely murdered than white women, they are killed six years younger than the national average.⁷ Black women also account for 44·6% of pregnancy-related violent deaths resulting from IPV.⁶ The underlying culprit for these disproportionate rates may be racism, and clear descriptive epidemiology that documents differences in rates across US states using updated data represents a critical need to uncover the scope of the problem.

A dearth of research provides clear descriptive epidemiology illuminating the disproportionate rates of homicide among Black women, and it is not without limitations. Epidemiological literature examines homicide rates among Native American women,⁸ its relationship with firearm legislation³ and IPV homicides in North Carolina.⁹ We found one article that analyzes racial and ethnic differences in homicides. Petrosky and colleagues¹ interrogated homicide deaths between 2003 and 2014, and found that Black women experience the highest rates of IPV homicide. Still, there are noteworthy gaps. Findings only account for 18 states, spans 12 years, and are calculated through 2014. There is an urgent need to identify trends thereafter, especially since there have been limited, yet coordinated and multifaceted efforts to address structural barriers to reduce inequities associated with elevated rates of homicide among US Blacks.^10,11

We advance this research by providing state- and regional analyses of homicide rates among women, aged 25–44, between 1999 and 2020. We focus upon this age range, because this is when women are most likely murdered.¹ This epidemiological analysis is a next step to developing a comprehensive agenda to help ameliorate inequities among homicide victims.

Methods

Data source.

Data for this cross-sectional time series were drawn from the National Vital Statistics System (NVSS) complete census on causes of death in the US Death records contain year, of death, sex, age, and racialized group membership.^12,13 This is a publicly available dataset. We restricted analyses to those who were identified to be female, aged 25–44, and deceased between 1999 and 2020. We also restricted to Black or white race, and did not use ethnicity as an inclusion or exclusion criteria; thus, women recorded as Black race may also have Hispanic ethnicity, and as well as those recorded as white; hereafter we referred to these groups as Black and white, respectively. Race and ethnicity data on the death certificate is reported by the funeral director as provided by an informant (often the surviving next of kin) or observation in the absence of an informant whereas race and ethnicity information from the census is by self-report. The bridged race variable was used. We also restricted the analysis to 30 states where the event rate (i.e. number of homicide deaths) was above the minimum for reporting per NVSS guidelines (>9 deaths in any year) in all analysis periods. STROBE reporting guidelines were followed (see Online Table 1).

Death codes.

Homicides included ICD-10 codes *U01-*U02, X85-Y09, Y87·1. Crude mortality rates were calculated per 100,000 based on US Census Bureau population sizes. Death records contained information on cause, which we categorized based on frequency into firearms, cutting/piercing, and other (suffocation, fire/flame, struck by or against, poisoning, unspecified, or other).

Statistical analysis.

Locally estimated scatterplot smoothing (LOESS) regression curves were estimated to visualize overall trends of homicide rates by race annually from 1999–2020. To statistically estimate the number of linear slopes across time that best fit the data, we used Joinpoint regression software. We also ran a parallel pairwise comparison to statistically test for slope differences between Black and white women.

For each study period (1999–2003; 2004–2008; 2009–2013; 2014–2018; 2019–2020) and US states with sufficient homicide deaths to meet the minimum for reporting per NVSS, ^12,13 we estimated the rate of recorded homicide per 100,000 among Black and white women. We examined the absolute and relative differences in homicide within the study period (1999–2003 vs 2019–2002; 1999–2003 vs. 2009–2013; and 2009–2013 vs. 2019–2020), reporting rate ratios and rate differences and 95% confidence intervals (CI). Rate ratios were estimated in SAS PROC GENMOD in a log-link Poisson model with population offset terms, with confidence intervals estimated from the “lsmeans” option, with the following formula:

$$\log \left({\mu }_i\right)=\log \left({\eta }_i\right)+{\chi }_i^{\prime }\mathcal{B}$$

where ${\mu }_i$ is the number of deaths, ${\eta }_i$ is the population denominator, and ${\chi }_i^{\prime }\mathcal{B}$ is the set of independent variables (race, time period). Rate differences were estimated using the SAS NLMEANS macro within the log-link Poisson model. We estimated the relative change in the rate ratio in homicide rates comparing Black to white women for each time period comparison, by state, by testing the interaction between year and race for the first and last study period. We examined the absolute change in homicide rates for each time period comparison, stratified by racialized group membership, and statistically tested whether the absolute disparity in the homicide rate changed using chi-square.

Finally, we examined the method of homicide death by region in the US, for four US regions (South [Alabama (AL), South Carolina (SC), Arkansas (AR), Tennessee (TN), Georgia (GA), Oklahoma (OK), Kentucky (KY), Florida (FL), North Carolina (NC), Maryland (MD), Texas (TX), Delaware (DE), Virginia (VA)]; Midwest [Illinois (IL), Indiana (IN), Michigan (MI), Ohio (OH), Wisconsin (WI), Kansas (KS), Minnesota (MN), Missouri (MO)]; West [Arizona (AZ), Colorado (CO), Nevada (NV), Washington (WA)]; Northeast [Connecticut (CT), Massachusetts (MA), New Jersey (NJ), New York (NY), Pennsylvania (PA)]). Logistic regression models in SAS were estimated with firearm death (vs. other) as the outcome, and predictors of race and year category, stratified by region, with the general formula for the logistic model given by:

$$logit(\pi )=\log \left(\frac{\pi }{1-\pi }\right)=\alpha +{\beta }^{\prime }\chi$$

Where $(\pi )$ is the probability of firearm death vs. other, $\alpha$ is the intercept parameter and ${\chi }_i^{\prime }\mathcal{B}$ is the vector of covariates (race, year group). Race by year interactions were also estimated and reported.

States were included in each region for each year that there was a sufficient number of homicides (>9 deaths) by race for analysis. All analyses were conducted using R version 4·2·1 and SAS version 9·4.

Ethics.

This project did not involve human data or participants; therefore, per the guidelines of the Columbia University Institutional Review Board, IRB assessment was not necessary. The funder had no role in data collection, analysis, interpretation, manuscript writing, and decision to submit.

Results

Trends in homicide rates among Black and white women aged 25–44 in the US

This study included 31,627 homicide deaths (12,243 among Black women and 19,384 among white women). Figure 1 shows the rate of homicide per 100,000 among women aged 25–44 stratified by racialized group membership (Black, white). Black women have persistent and large inequities in homicide compared with white women. In 1999, the homicide rate among Black women was 11·6 per 100,000 compared to 2·9 per 100,000 among white women. Although homicide rates remained significantly higher among Black women, rates declined from 1999–2013 and then shifted course, increasing from 2013–2020 whereas homicide rates remained static among white women. Inequities persisted through 2020; the homicide rates among Black women was 11·6 per 100,000; compared with 3·0 per 100,000 among white women. Joinpoint regression analysis indicated that the best fitting model included 2 linear slopes (1 joinpoint) for the Black trend line with a significant change in the magnitude of the slope in 2014 (1999 to 2013, slope=−0·35, SE=0·04, p<0·010; 2014 to 2020, slope=1·05, SE=0·15, p<0·010). The best fitting model included 1 linear slope (0 joinpoints, slope=−0·02, SE=0·01, p=0·080) for the white trend line. Additionally, we ran a parallel pairwise comparison to statistically test for parallelism between the two groups. This test rejected parallelism with a p-value of <0·0010, indicating that the slopes were significantly different between Black and white women.

Figure 1. Homicide rate per 100,000 among Black and white women aged 25–44 in the United States from 1999–2020.

*Figure 1. Includes LOESS regression estimated homicide rates and 95% confidence intervals.

*Joinpoint regression analysis indicated that the best fitting model included 2 linear slopes (1 joinpoint) for the Black trend line with a significant change in the magnitude of the slope in 2014 (1999 to 2013, slope=−0·35, SE=0·04, p<0·010; 2014 to 2020, slope=1·05, SE=0·15, p<0·010). The best fitting model included 1 linear slope (0 joinpoints, slope=−0·02, SE=0·01, p=0·080) for the white trend line. Additionally, we ran a parallel pairwise comparison to statistically test for parallelism between the two groups. This test rejected parallelism with a p-value of <0·0010, indicating that the slopes were significantly different between Black and white women. Joinpoint estimates are visualized with black dashed line.

Racial Inequities in Homicide Rates by State and Region

Table 1 presents the homicide rate ratio and rate difference among Black and white women, aged 25–44 for 30 US states with sufficient (>9) deaths for analysis. Overall, the homicide rate for women in the US ranged from 6·6 per 100,000 in AL to 1·6 per 100,00 in MA, representing a 4-fold difference (see Online Table 2). State-by-state analysis revealed substantial heterogeneity in the magnitude and trend in relative as well as absolute inequities.

Table 1.

Relative and absolute differences in homicide rates, by time period, comparing Black and white women aged 25–44 in the United States

State	Relative or absolute differences	1999–2003	2004–2008	2009–2013	2014–2018	2019–2020	Difference in difference: 1999–2003 vs 2009–2013	Difference in difference: 1999 2009–2013 vs 2019–2020	Difference in difference: 1999 1999–2003 vs 2019–2020)	Overall homicide rate per 100,000 among black and white women 25–44 from highest to lowest
South	Rate ratio (95% CI)	2·88 (2.70, 3·06)	2·56 (2·40, 2·74)	2·45 (2·28, 2·63)	2·46 (2·30, 2·64)	2·78 (2·52, 3·07)	0·86+	1·14+	0·96+	4·80
	Rate difference (95% CI)	7·20 (6·60, 7·70)	5·50 (5·10, 6·10)	4·60 (4·20, 5·10)	4·90 (4·40, 5·30)	6·40 (5·60, 7·10)	−2·60+	1·80+	−0·8+
Alabama	Rate ratio (95% CI)	3·19 (2·47, 4·10)	2·83 (2·14, 3·75)	2·36 (1·74, 3·20)	2·62 (1·98, 3·47)	2·07 (1·35, 3·16)	0·74+	0·88+	0·65+	6·60
	Rate difference (95% CI)	10·35 (7·70, 13·00)	7·65 (5·30, 10·00)	5·30 (3·20, 7·40)	6·98 (4·70, 9·30)	5·48 (1·90, 9·00)	−5·05+	0·19+	−4·87+
South Carolina	Rate ratio (95% CI)	2·09 (1·58, 2·78)	2·72 (1·99, 3·71)	2·08 (1·51, 2·88)	2·16 (1·58, 2·95)	2·56 (1·67, 3·93)	0·99+	1·24+	1·22+	5·70
	Rate difference (95% CI)	5·31 (3·00, 7·60)	6·19 (4·00, 8·40)	4·07 (2·10, 6·10)	4·32 (2.40, 6·30)	6·89 (3·30, 10·40)	−1·24+	2·82+	1·58+
Arkansas	Rate ratio (95% CI)	4·42 (2·98, 6·55)	3·91 (2·62, 5·84)	3·45 (2·29, 5·22)	2·98 (2·04, 4·35)	6·06 (3·59, 10·22)	0·79+	1·76+	1·37+	5·70
	Rate difference (95% CI)	11·62 (7·30, 16·00)	10·18 (6·00, 14·30)	8·68 (4·80, 12·60)	8·82 (4·80, 12·80)	20·33 (12·00, 28·70)	−2·94+	11·65+	8·71+
Tennessee	Rate ratio (95% CI)	3·63 (2·83, 4·66)	2·46 (1·86, 3·25)	2·46 (1·84, 3·29)	3·03 (2·30, 4·01)	3·88 (2·64, 5·69)	0·68+	1·58+	1·07+	5·20
	Rate difference (95% CI)	10·44 (7·70, 13·20)	5·86 (3·60, 8·20)	5·31 (3·10, 7·50)	6·92 (4·70, 9·20)	10·68 (6·60, 14·70)	−5·13+	5·37+	0·24+
Georgia	Rate ratio (95% CI)	3·02 (2·47, 3·71)	2·41 (1·94, 3·01)	2·42 (1·89, 3·11)	2·07 (1·65, 2·59)	2·77 (1·96, 3·90)	0·81+	1·15+	0·92+	4·80
	Rate difference (95% CI)	7·25 (5·70, 8·80)	4·68 (3·40, 6·00)	3·62 (2·50, 4·70)	3·47 (2·30, 4·60)	5·37 (3·50, 7·30)	−3·63+	1·75+	−1·88+
Oklahoma	Rate ratio (95% CI)	2·73 (1·70, 4·39)	2·55 (1·57, 4·14)	3·01 (1·90, 4·78)	2·08 (1·24, 3·48)	5·13 (2·89, 9·11)	1·11+	1·71+	1·88+	4·60
	Rate difference (95% CI)	6·86 (2·23, 11·50)	6·34 (1·80, 10·90)	7·62 (3·00, 12·30)	4·07 (3·41, 7·80)	15·84 (6·90, 24·80)	0·76+	8·22+	8·98+
Kentucky	Rate ratio (95% CI)	2·57 (1·68, 3·93)	1·90 (1·12, 3·23)	3·35 (2·19, 5·13)	1.76 (1.09, 2.83)	3·93 (2·33, 6·63)	1·31+	1·18+	1·53+	4·60
		Rate difference (95% CI)	6·76 (2·40, 11·10)	3·22 (−0·02, 6·60)	8·12 (3·80, 12·50)	3·40 (−0·05, 6·90)	13·88 (5·60, 22·20)	1·36+	5·76+	7·12+
Florida	Rate ratio (95% CI)	2·22 (1·86, 2·66)	2·48 (2·08, 2·97)	2·53 (2·11, 3·04)	2·21 (1·84, 2·66)	2·08 (1·60, 2·70)	1·14+	0·83+	0·94+	4·40
	Rate difference (95% CI)	4·84 (3·50, 6·20)	5·31 (4·00, 6·60)	4·92 (3·70, 6·10)	3·86 (2·80, 4·90)	4·11 (2·40, 5·90)	0·09+	−0·81+	−0·73+
North Carolina	Rate ratio (95% CI)	2·74 (2·20, 3·40)	2·17 (1·72, 2·73)	1·78 (1·35, 2·34)	2·45 (1·90, 3·16)	3·22 (2·22, 4·67)	0·65+	1·81+	1·18+	4·40
	Rate difference (95% CI)	6·56 (4·80, 8·30)	4·37 (2·80, 5·90)	2·28 (1·00, 3·50)	3·99 (2·70, 5·30)	5·99 (3·70, 8·30)	−4·28+	3·71+	−0·57+
Maryland	Rate ratio (95% CI)	3·56 (2·67, 4·74)	2·59 (1·89, 3·53)	2·87 (2·02, 4·08)	2·80 (1·98, 3·95)	2·15 (1·34, 3·48)	0·81+	0·75+	0·60+	4·30
	Rate difference (95% CI)	6·94 (5·20, 8·70)	4·36 (2·80, 5·90)	3·97 (2·50, 5·40)	4·03 (2·60, 5·50)	3·71 (1.30, 6·20)	−2·97+	−0·26+	−3·23+
Texas	Rate ratio (95% CI)	2·59 (2·18, 3·06)	2·41 (2·04, 2·86)	2·29 (1·92, 2·74)	2·57 (2·19, 3·01)	2·84 (2·24, 3·59)	0·88+	1·25+	1·10+	4·00
	Rate difference (95% CI)	5·54 (4·20, 6·90)	4·92 (3·70, 6·20)	3·93 (2·80, 5·00)	5·01 (3·90, 6·10)	5·72 (4·00, 7·40)	−1·61+	1·79+	0·18+
Delaware	Rate ratio (95% CI)	N/A	2·67 (1·14, 6·30)	N/A	N/A	N/A	NA	NA	NA	3·60
	Rate difference (95% CI)	N/A	4·39 (−0·01, 9·00)	N/A	N/A	N/A	NA	NA	NA
Virginia	Rate ratio (95% CI)	3·38 (2·62, 4·36)	2·96 (2·21, 3·97)	3·27 (2·36, 4·51)	3·32 (2·45, 4·49)	1·77 (1·08, 2·89)	0·97+	0·55+	0·52+	3·50
	Rate difference (95% CI)	7·02 (5·10, 8·90)	4·76 (3·10, 6·40)	4·33 (2·80, 5·80)	4·93 (3·40, 6·50)	2·13 (0·04, 4·20)	−2·69+	−2·20+	−4·89
Midwest	Rate ratio (95% CI)	6·16 (5·61, 6·76)	5·36 (4·83, 5·95)	4·32 (3·85, 4·86)	4·85 (4·37, 5·38)	7·22 (6·24, 8·35)	0·70+	1·67+	1·17+	3·60
	Rate difference (95% CI)	12·70 (11·60, 13·70)	9·70 (8·70, 10·60)	6·90 (6·10, 7·70)	9·30 (8·30, 10·20)	15·00 (13·20, 16·70)	−5·80+	8·1+	2·30+
Missouri	Rate ratio (95% CI)	4·56 (3·43, 6·05)	4·92 (3·56, 6·80)	5·38 (3·92, 7·38)	6·06 (4·67, 7·86)	5·59 (3·88, 8·05)	1·18+	1·04+	1·22+	5·10
	Rate difference (95% CI)	12·25 (8·80, 15·70)	10·12 (7·00, 13·30)	11·46 (8·20, 14·80)	17·79 (13·80, 21·80)	20·97 (14·20, 27·70)	−0·79+	9·51+	8·72+
Indiana	Rate ratio (95% CI)	6·56 (4·98, 8·66)	5·34 (3·93, 7·26)	2·08 (1·39, 3·10)	5·09 (3·79, 6·84)	6·22 (4·09, 9·44)	0·32+	2·99+	0·95+	4·40
	Rate difference (95% CI)	17·49 (13·10, 21·90)	12·52 (8·80, 16·30)	3·47 (1·00, 5·90)	12·47 (8·90, 16·00)	16·74 (10·60, 22·80)	−14·02+	13·27+	−0·75+
Michigan	Rate ratio (95% CI)	6·45 (5·24, 7·95)	5·19 (4·08, 6·60)	5·17 (3·90, 6·87)	4·88 (3·78, 6·30)	6·97 (4·72, 10·28)	0·81+	1·35+	1·08+	4·30
	Rate difference (95% CI)	14·77 (12·30, 17·20)	10·24 (8·10, 12·40)	8·03 (6·00, 10·00)	9·54 (7·40, 11·70)	13·18 (9·40, 17·00)	−6·74+	5·15+	−1·59+
Ohio	Rate ratio (95% CI)	4·47 (3·52, 5·67)	5·05 (3·99, 6·39)	3·73 (2·86, 4·85)	3·46 (2·72, 4·39)	5·26 (3·83, 7·21)	0·84+	1·42+	1·18+	3·80
	Rate difference (95% CI)	8·42 (6·40, 10·50)	9·84 (7·70, 12·00)	6·54 (4·70, 8·40)	7·27 (5·30, 9·20)	13·68 (9·90, 17·50)	−1·88+	7·14+	5·26+
Kansas	Rate ratio (95% CI)	6·15 (3·64, 10·40)	4·83 (2·68, 8·71)	4·30 (2·48, 7·45)	3·68 (2·11, 6·44)	6·57 (3·03, 14·24)	0·70+	1·53+	1·07+	3·80
	Rate difference (95% CI)	14·14 (6·70, 21·60)	10·24 (3·70, 16·80)	10·75 (4·00, 17·50)	9·14 (2·90, 15·40)	15·98 (4·20, 27·70)	−3·39+	5·23+	1·84+
Illinois	Rate ratio (95% CI)	6·81 (5·61, 8·25)	5·34 (4·23, 6·73)	4·84 (3·69, 6·36)	6·02 (4·70, 7·71)	11·21 (7·65, 16·42)	0·72+	2·32+	1·65+	3·60
	Rate difference (95% CI)	13·95 (11·90, 16·00)	8.52 (6.80, 10.20)	5.92 (4.50, 7.40)	8.73 (7.00, 10.50)	15.06 (11.70, 18.50)	−8.03+	9.14+	1.11+
Wisconsin	Rate ratio (95% CI)	6·42 (4·07, 10·12)	5·42 (3·45, 8·52)	5·55 (3·47, 8·87)	5·19 (3·35, 8·06)	20·16 (10·65, 38·16)	0·86+	3·63+	3·14+	2·50
	Rate difference (95% CI)	9·13 (5·00, 13·20)	8·42 (4·50, 12·30)	7·76 (4·10, 11·40)	8·24 (4·60, 11·90)	22·09 (13·60, 30·60)	−1·37+	14·33+	12·96+
Minnesota	Rate ratio (95% CI)	3·87 (1·97, 7·59)	7·36 (4·20, 12·89)	5·55 (2·96, 10·40)	N/A	6·20 (2·88, 13·37)	1·43+	1·12+	1·60+	1·60
	Rate difference (95% CI)	4·65 (0·07, 8·60)	7·84 (3·60, 12·00)	4·91 (1·70, 8·10)	NA	6·99 (2·00, 12·00)	0·26+	2·08+	2·34+
West	Rate ratio (95% CI)	3·36 (2·94, 3·84)	3·67 (3·20, 4·21)	2·36 (1·98, 2·79)	2·73 (2·35, 3·18)	2·78 (2·24, 3·45)	0·71+	1·18+	0·83+	2.90
	Rate difference (95% CI)	6·90 (5·70, 8·10)	6·80 (5·60, 7·90)	3·20 (2·30, 4·00)	4·20 (3·30, 5·10)	4·70 (3·30, 6·20)	−3·70+	1·5+	−2·20+
Nevada	Rate ratio (95% CI)	4·08 (2·49, 6·67)	2·01 (1·15, 3·51)	1·49 (0·76, 2·94)	3·70 (2·32, 5·90)	2·90 (1·35, 6·24)	0·36+	1·95+	0·71+	4·60
	Rate difference (95% CI)	13·65 (6·00, 21·30)	4·73 (−0·02, 9·6)	1·78 (−0·02, 5·30)	8·73 (4·20, 13·20)	5·83 (1·50, 11·50)	−11·87	4·05	−7·82
Arizona	Rate ratio (95% CI)	2·66 (1·63, 4·34)	3·07 (1·91, 4·92)	2·08 (1·21, 3·56)	1·57 (0·88, 2·78)	N/A	0·79	N/A	NA	3·90
	Rate difference (95% CI)	7·97 (2·00, 13·90)	7·43 (2·60, 12·30)	3·54 (0·40, 7·00)	1·76 (−0·09, 4·50)	N/A	−4·43	N/A	NA
Colorado	Rate ratio (95% CI)	4·26 (2·53, 7·17)	3·37 (1·87, 6·08)	2·51 (1·30, 4·85)	1·91 (0·99, 3·66)	N/A	0·59+	N/A	NA	2·90
	Rate difference (95% CI)	8·96 (3·40, 14·60)	6·04 (1·30, 10·70)	3·62 (−0·02, 7·40)	2·44 (−0·08, 5·70)	N/A	−5·34	N/A	NA
Washington	Rate ratio (95% CI)	3·69 (2·05, 6·64)	N/A	N/A	2·19 (1·22, 3·91)	N/A	N/A	N/A	NA	2·40
	Rate difference (95% CI)	5·58 (1·40, 9·80)	N/A	N/A	2·75 (−0·04, 5·50)	N/A	N/A	N/A	NA
Northeast	Rate ratio (95% CI)	3·15 (2·82, 3·51)	3·27 (2·86, 3·73)	2·95 (2·55, 3·42)	2·96 (2·55, 3·43)	3·30 (2·64, 4·12)	0·94+	1·12+	1·05+	2·70
	Rate difference (95% CI)	6·00 (5·20, 6·80)	4·40 (3·70, 5·00)	3·40 (2·80, 4·00)	3·20 (2·60, 3·80)	4·10 (3·00, 5·00)	−2·60+	0·70+	−1·9+
Pennsylvania	Rate ratio (95% CI)	4·86 (3·88, 6·10)	3·76 (2·91, 4·87)	4·17 (3·21, 5·41)	3·65 (2·85, 4·69)	5·45 (3·80, 7·81)	0·86+	1·31+	1·12+	3·50
	Rate difference (95% CI)	9·87 (7·70, 12·10)	6·55 (4·70, 8·40)	6·96 (5·10, 8·80)	6·65 (4·80, 8·50)	10·39 (7·10, 13·70)	−2·91+	3·43+	0·52+
New Jersey	Rate ratio (95% CI)	3·19 (2·45, 4·14)	4·52 (3·31, 6·18)	4·13 (2·82, 6·05)	3·47 (2·43, 4·96)	2·71 (1·52, 4·80)	1·29+	0·66+	0·85+	2·90
	Rate difference (95% CI)	6·10 (4·30, 8·00)	6·05 (4·30, 7·80)	4·03 (2·60, 5·50)	4·03 (2·50, 5·60)	3·09 (0·08, 5·30)	−2·07+	−0·94+	−3·05
New York	Rate ratio (95% CI)	2·30 (1·96, 2·69)	2·77 (2·23, 3·46)	2·25 (1·76, 2·86)	2·58 (1·95, 3·42)	2·65 (1·76, 4·00)	0·98+	1·18+	1·15	2·90
	Rate difference (95% CI)	4·91 (3·80, 6·10)	2·32 (2·40, 4·20)	2·25 (1·40, 3·10)	1·95 (1·20, 2·70)	2·38 (1·10, 3·60)	−2·66+	0·13+	−2·53
Connecticut	Rate ratio (95% CI)	2·10 (1·13, 3·88)	2·85 (1·61, 5·06)	3·98 (2·10, 7·54)	2·42 (1·28, 4·58)	N/A	1·89+	N/A	NA	2·20
	Rate difference (95% CI)	2·31 (−0·02, 4·80)	3·60 (0·09, 6·30)	3·83 (1·30, 6·40)	2·47 (0·02, 4·80)	N/A	1·52+	N/A	NA
Massachusetts	Rate ratio (95% CI)	2·50 (1·46, 4·29)	3·36 (1·94, 5·82)	1·77 (0·90, 3·50)	2·56 (1.48, 4·44)	N/A	0·71	N/A	NA	1·60
	Rate difference (95% CI)	2·60 (0·04, 4·80)	3·06 (0·09, 5·20)	1·05 (−0·05, 2·60)	2·19 (0·04, 3·90)	N/A	−1·55	N/A	NA

^*Notes: Comparisons were made between women categorized as Black (with or without Hispanic ethnicity) and white (with or without Hispanic ethnicity). States in the table are organized from highest to lowest overall homicide rates. “N/A” indicates a state that was excluded due to suppressed data by race on CDC WONDER (numerator or denominator < 10). Included states are those with at least some time periods in which homicide death counts were not suppressed. 21 states were excluded due to suppressed data at all time points.

Columns of difference in difference compare the magnitude of the rate ratio and rate difference between time periods.

⁺* Indicates that the p-value for the interaction test between race/ethnicity and time period was <0·050.

In terms of relative inequities (Table 1), there was no US state where there was not a significantly higher homicide rate among Black women compared with white women in any time period. Inequities were highest in the Midwest; in 2019–2020, Black women had 7·22 (95% CI 6·24–8·35) times the rate of homicide death compared with white women. In terms of the magnitude of relative inequities across time, the relative disparity significantly improved for 8 states (e.g. in MD and VA, the relative disparity in 2019–2020 was 0·60 and 0·52 times that of 1999–2003), and significantly widened for 16 states (e.g. the disparity in OK increased 1·9-fold between 1999–2003 and 2019–2020, and in WI it increased 3·1-fold).

We also examined variation in relative disparities within the time series, comparing the magnitude of the disparity in 1999–2003 to 2009–2013, and in 2009–2013 to 2019–2020 (Table 1). These results indicated variation within the time period across states and regions. In all regions (South, Midwest, West and Northeast), the homicide disparity declined in 1999–2003 compared with 2009–2013 then increased from 2009–2013 compared with 2019–2020 (see Table 1). This pattern was most pronounced in the Midwest (1999–2003 vs. 2009–2013, RR=0·70, p<0·05; 2009–2013 vs. 2019–2020 (RR=1·67, p<0·05).

In terms of absolute inequities (Table 1), there remain strong and significant absolute rate differences in homicide rates in every state. In 2019–2020, the highest rate differences were observed in WI (22·09 per 100,000, 95% CI 13·60–30·60, MO (20·97 per 100,000, 95% CI 14·20–27·70), AR (20·33 per 100,000, 95% CI 12·00–28·70), and OK (15·84 per 100,000, 95% CI 6·90–24·80). In each of these states, more than 15 additional Black women than white women died by homicide for every 100,000 women aged 25 to 44. In terms of the magnitude of absolute disparities across time, descriptive examination indicates that the absolute disparity significantly improved for 7 states, and significantly widened for 13 states. The most pronounced change was in WI, where the absolute disparity increased by 12·96 deaths per 100,000 between 1999–2003 and 2019–2020.

Table 1 also shows variation within the time series in absolute rate disparity. Similar to the relative disparity, the absolute disparity declined in all regions from 1999–2003 to 2009–2013, and the increased from 2009–2013 to 2019–2020. Again, this pattern was most pronounced in the Midwest; from 1999–2003 to 2009–2013, the absolute difference in homicides between Black women and white women declined by 5·8 deaths per 100,000 (p<0·05); then from 2009–2013 to 2019–2020 the absolute difference increased by 8·1 deaths per 100,000 (p<0·05).

Figure 2 shows relative racial disparity in homicide (ratio of the Black homicide rate and white homicide rate) to two periods: 1999–2003 and 2019–2020. For example, in AL, Black women had three times the homicide rate of white women in 1999–2003, and twice the homicide rate of white women in 2019–2020. Of the 30 states included in this analysis 8 states (OH, MI, PA, TN, FL, NJ, NC, TX) had no change in the disparity across the 20 years of inclusion. In 6 states (IN, GA, MD, NV, AL, VA) the disparity decreased, with the largest decrease being two-fold; in MD, the homicide rate among Black women was 4x times higher than among white women in 1999–2003, decreasing to 2x higher in 2019–2020. For 10 states, the disparity increased: AR, KS, IL, KY, MN, OK, MO, NY, and SC, WI. WI was an outlier in the magnitude of disparity, so we did not include it in the same legend as other states in 2019–2020. Homicide inequities in WI increased from 6x between 1999–2003 to 20x between 2019–2020. As an alternative data visualization, we have also included the relative disparity in homicide rate by state in a scatter plot in Online Figure 1, which displays the disparity in homicide in 1999–2003 on the x-axis and the disparity in homicide in 2019–2020 on the y-axis, with each data point a US state.

Figure 2. Relative change in homicide disparities between Black and white women, aged 25–44 from 1999–2003 to 2019–2020.

* States colored in white did not have sufficiently large number of events for analysis. There were 29 states in the first study period and 24 states in the second study period with sufficient numbers of deaths. Data Source: CDC WONDER. The legend indicates the relative homicide disparity between Black and white women. The magnitude of the disparity is indicated by the color (ranging from yellow to orange for the US and light orange to green for Wisconsin). The change in the disparity from 1999–2003 to 2019–2020 can be visualized through the change in color i.e., a change from a lighter yellow color in 1999–2003 to a darker orange color in 2019–2020 indicates an increase in the disparity; a change from a darker color to a lighter color indicates a decrease in the disparity; no change in color indicates no change in the disparity.

**In Wisconsin, the racial disparity in female homicide sufficiently higher in 2019–2020 (20 times) than the disparity between Black and white women for any other state in any time period, that we opted to show Wisconsin with a separate color (green) in 2019–2020.

Racial Differences in Homicide Methods by US Region

Online Figures 2 and 3 show the method of homicide among women aged 25–44 in the US by region. Among Black women in the Northeast, the prevalence of firearms as a cause of death increased, accounting for 62% of homicides (83 out of 135 deaths) by 2019–2020. In the Midwest, firearms are also more common as a cause of death among Black compared to white women, with 84% of deaths (329 out of 394) in 2019–2020 caused by firearms among Black women, compared with 62% (207 out of 337) among white women. Similar trends hold for the South and West. In 2019–2020 in the South, 81% (616 out of 762) of Black women were killed with a firearm, an increase from 54% (947 out of 1739) in 1999–2003, and higher than white women in 2019–2020 (69% (575 out of 839) deaths). In the West, 73% (74 out of 102) of homicides of Black women were by firearm, an increase 1999–2003 (59% (154 out of 259) deaths), and substantially higher than among white women in 2019–2020 (62% (270 out of 436) deaths).

Logistic regression analyses showing the relationship between year and firearm involvement are shown in Online Table 3. Black women had 1·38 (95% CI 1·28, 1·48) times the odds of homicide involving firearms vs other causes compared to white women. Compared to white women, Black women in the Northeast and West had the highest odds of homicide involving firearms vs other causes.

Figure 3 shows the change in the relationship between race and firearm deaths stratified by year and census region for the Northeast, Midwest, and South. We did not have sufficient data to estimate the relationship between race and firearm death by year in the West. By year, firearm homicide deaths among Black women 25–44 in the West ranged from 18 to 43, thus there were not sufficient numbers of deaths for reliable estimation. In 2019–2020, in the Northeast, Black women had 3·02 (95% CI 1·27, 7·14) times the odds of homicide death by firearm vs other methods when compared to white women; in the Midwest Black women had 6·31 (95% CI 3·05, 13·06) times the odds of homicide death by firearm vs other methods when compared to white women. The odds of death by firearm among Black compared with white women is increasing across time in the South. In 1999–2003, Black women had 0·97 (0·85, 1·10) times the odds of firearm homicide death compared to other means. By 2019–2020, Black women had 1·51 (1·06, 2·15) times the odds of firearm homicide death compared to other means.

Figure 3. Homicide death by firearm versus all other methods among Black women compared with white women, aged 25–44 by Year and Census Region.

*Note: The West was excluded from this analysis due to insufficient sample size for interaction models

*p-values for joint test for interaction between year category and race: Northeast (p=0·090); Midwest (p<0·0001); South (p=0·030)

Discussion

Findings underscore pervasive racial inequities in homicide rates among Black and white women, aged 25–44 years old, between 1999 and 2020 across 30 states. IPV is a key causal factor of homicides; however, current mechanisms precluded this interrogation. Nonetheless, Black women continue to experience persistently high rates of homicide, compared to white women. While rates of homicide declined among Black women from 1999–2013, they have changed direction and increased from 2014–2020. Throughout the 22-year inquiry, Black women were generally six times more likely murdered than their white peers. We interrogated regional differences and found that Black women residing in the Midwest and Northeast, respectively, were more likely killed with a firearm than Black women residing in other areas of the country.

To obtain a more nuanced understanding, we examined statewide shifts and found that states which reflected the greatest inequities correlate with areas of the country where concentrated disadvantage is most evident.¹⁴ These areas also correspond with those that have enduring histories of slavery and lynching, along with locations where especially tense Black Lives Matter (BLM) protests occurred during the height of the pandemic.^15,16 MO and WI reflect pronounced inequities in homicide rates. These two states have similar histories of structural inequities that were further exposed when high-profile police misconduct triggered national public outrage and a series of protests, along with federal charges against local police.

The greatest inequity in homicide rates is in WI. Black women were 20 times more likely murdered in 2019–2020 compared to compared to six times in 1999–2003. This increase is noteworthy. This midwestern state ranks last in racial equity in the US,¹⁴ based on a score tabulated using US Census data that accounts for inequities in wealth distribution.¹⁴ Inequities in WI may be driven by large cities such as Milwaukee where there are higher concentrations of Black individuals than other areas across the state.¹⁷ The racial equity index score contextualizes our understanding of the structural inequities that may further isolate Black women during help-seeking.

Racism may provide insight as to why Black women, regardless of their ethnicity, disproportionately experience high rates of homicide. Racism is endemic and its effects are insidious.¹⁸ Structural racism offers a comprehensive societal understanding of the pernicious, mutually reinforcing [inequitable] systems that reinforce discriminatory beliefs, values, and distribution of resources.¹⁹ Structural racism is institutionalized and deeply entrenched throughout the US and associated with higher levels of homicide among Black women.⁵ Specifically, educational attainment, employment, poverty, residential segregation and home ownership are well-known indicators influencing disproportionately high rates of homicide in areas were Black women largely reside.⁵ Structural racism also underscores many of the psychosocial framings for homicide inequities.^20–22 It is critical that we understand the ways that overlapping forms of oppression precludes Black women from securing the nature and level of assistance necessary to reduce the likelihood that they will be murdered.^23,24

Structural racism compounded by the Ferguson effect that increased tensions between police and Black communities could also help us to understand why there is a noticeable U-shape in homicides over the 22-year inquiry.²⁵ The Black community has a tenuous relationship with the criminal legal system, evidenced in highly publicized abuses ranging from over-criminalization and harsher penalties to unrestrained police misconduct that largely goes unpunished.⁵ Vicarious experiences of police violence result in a general refusal to engage with law enforcement. When Black women decide to rely upon the police, they are often met with delays, disbelief and are sometimes denied basic services and interventions critical for their wellbeing.²⁴ A confluence of these high-profile and well-known inequities have ruptured Black women’s trust in systems that their white peers readily employ. As a result, Black women typically only engage law enforcement when they believe their lives are in imminent danger.²⁴

Homicide inequities largely persisted in 2019–2020, which included the first year of the COVID-19 pandemic. Long-documented societal inequities were further exacerbated during the pandemic.²⁶ State-imposed shelter-in-place measures resulted in closed businesses and churches, widespread unemployment, increased gun and ammunition purchases, reduced social supports, as well as increased stress and isolation.^26,27 Further, Black women residing in areas where there is concentrated disadvantage also experienced higher rates of homicide compared to their white counterparts. Communities that were already bereft of basic services and supports resulting from wealth and income inequities were overwhelmingly burdened with job losses and employment furloughs. This compounded the long-standing effects of disproportionately high rates of unemployment.⁵ A confluence of these factors contributed to elevated incidence of IPV²⁸ and may have had a subsequent impact on homicide rates among Black women.

The findings in this report are subject to at least four limitations. First, sparse and incomplete data prevented us from including all 50 states in analysis and precluded a county-level examination; thus, data are generalizable only to the states included. Second, available data indicate that homicides increased in the US substantially from 2019 to 2020,²⁹ particularly among Black populations; we combined 2019 and 2020 data to increase the number of events for reporting purposes, thus estimates of current inequities are likely underestimated.²⁹ Third, the data do not provide insight regarding victims’ gender expression. As such, we are unable to provide insight regarding the experiences specific to Black or white women who are members of the LGBTQIA community.

Finally, Black women are not a monolith. They are often categorized as a homogenous group despite the diversity of their sociohistorical experiences. African American women are the largest subgroup of US Black women. Black women additionally includes African, Afro Caribbean, Black Hispanic, and Black European women.³⁰ Current reporting mechanisms hinder our ability to specify subpopulations of Black women.

Additional funding is needed for violence research. Funding priorities are inconsistent with community needs. Current data surveillance systems capture insufficient and incomplete data. This limits our ability to determine the number of homicides resulting from IPV. Investing in national surveillance systems would improve our ability to track these trends.

Notwithstanding, this analysis is an important contribution to the literature. This is the first examination that includes a 22-year analysis of homicide trends among women, aged 25–44. Findings provide comprehensive insight that includes the onset of the COVID-19 pandemic when many states enforced shutdown rules. This analysis is timely. Many states are developing strategic plans and allocating funding to support underserved communities of color more fully. This interrogation highlights the need to make continued progress toward reducing structural inequities that result in the disproportionate murders of Black women. Systems that have long disserviced communities of color have enduring and nocuous consequences. Dismantling these systems will take sustained, concerted and multitiered efforts.

RESEARCH IN CONTEXT.

Evidence before this study

Black women are more likely than women of any other racial group in the United States (US) to experience homicide. This is one of the most robust, longstanding, and widely reported findings about women in homicide epidemiology, yet the data is limited. Emerging evidence suggests a strong association between elevated homicide rates and the effects of long-standing structural inequities – educational attainment, unemployment, economic status/wealth distribution, poverty/concentrated disadvantage, and home ownership – across the US. This implies that elevated rates of homicide among women may be preventable through social and structural change; understanding the extent to which disparities change across time provides critical surveillance on the nature and scope of the ongoing problem, and state-level analyses allow for concentrated focus on areas in the most need of intervention. Indeed, there is also some evidence to suggest that racial inequities in homicide among Black and white women are larger, on average, in states that perform worse on indices of structural equity, thus state-level assessment is critical. Reasons for this trend are unclear and there are important limitations to the existing evidence that preclude robust inferences about the contexts and conditions that give rise to, and those that mitigate and prevent, inequities in homicide rates among Black and white women, including potential disparities in the lethal means through which homicide is perpetuated. The evidence is largely cross sectional, inclusive of a just over one third of states (only accounting for 18 out of 50 states) and, crucially, fails to account for trends homicide trends during the height of COVID-19 pandemic.

Added value of this study

We used time-series data from the largest and most comprehensive dataset, CDC WONDER data from 1999–2020, reporting deaths in the US among women aged 25–44, and examined (a) the extent to which homicide rates varied between Black and white women across states, and (b) how inequities in homicides between Black and white women shifted across 22 years, including during the height of the COVID-19 pandemic. We found strong evidence that indicates inequities in homicide rates between Black women and white women. Notably, Black women are murdered six times more often, on average, that of their white peers. Further, Black women residing in the Midwest and Northeast, respectively, were more likely killed with a firearm than Black women residing in any other area of the country. Importantly, we found the greatest inequities are in areas of the country where concentrated disadvantage is pronounced. This is critical initial step toward developing targeted solutions to reduce inequitable homicide rates among Black women.

Implications of all the available evidence

There is strong evidence that the size of racial inequalities in homicide rates are driven by social/structural context. Disparities in deaths by firearms underscores the ongoing role of gun violence in the US. Against a backdrop of high and rising rates of homicide among Black women, there is an urgent need to understand the contexts and conditions that reduce the likelihood of their premature death.

Data sharing:

All data used for the present study are publicly available through the US National Vital Statistics Sytem (https://www.cdc.gov/nchs/nvss/index.htm)