Abstract
Between 2014 and 2018, at least 974 youth were fatally shot by the police. Racial disparities in fatal police shootings (FPS) have been well-established in existing research, but less attention has been paid to patterns in fatal police encounters with youth. This study uses a multisource and externally validated research design to track cases of FPS for individuals aged between 10 and 24. Cases were geocoded to the county-level and linked to multiple variables capturing social, demographic, and policing contexts. The results point to substantial racial disparities in FPS. Black youth are killed at nearly double the rate of white youth. Counties with heightened risks to the physical safety of law enforcement have more FPS. Shootings are also more frequent for Black youth in racially segregated communities.
Introduction
On August 9th, 2014, members of the Ferguson, Missouri police department fatally shot Michael Brown, an unarmed, 18-year-old, Black male. Brown’s death was one of many fatal police shootings (FPS) of minors and early adults, which exhibit sharp racial disparities in their distribution (Edwards et al., 2018; Zimring, 2017). Estimates indicate that the rate of FPS for African Americans is roughly triple that of whites (Shane et al., 2017), consistent with the larger racial disparities in policing and the criminal justice system (Hinton & Cook, 2021). Edwards et al. (2018) suggest that individuals between 20 and 35 have the highest rates of FPS, indicating that the concentration of FPS is also stratified by age.
Research on FPS has grown extensively over the last decade (Sherman, 2018; Zare et al., 2022). Yet, despite incidents capturing national attention such as the death of Michael Brown, less work has directly focused on patterns of fatal encounters between the police and youth. There are several reasons researchers should examine the youth population directly. Police stops are strongly associated with age, and youth are disproportionately stopped by law enforcement (Figures & Legewie, 2019). Police stops also differ by race, with young Black males stopped at nearly double the rates of young white males (Geller, 2017). Youth, and especially minority youth, are disproportionately more likely to have contact with law enforcement, a necessary precursor to a fatal shooting. As well, the consequences of direct or indirect exposure to police violence for youth, particularly minority group members, may result in a variety of negative educational and mental health outcomes at key stages of development (Jindal et al., 2022).
This study expands on existing research examining police shootings by analyzing the social contexts where youth are fatally shot by the police. I emphasize the structural, demographic, and other contextual characteristics that put youth at heightened risk. Identifying the communities most at risk for FPS is an important component of developing a more comprehensive understanding of how racial disparities in police violence take shape. I use a unique database of 974 cases of fatal shootings of youth between 2014 and 2018, which were situated geographically and linked to contextual variables measuring key aspects of policing, racial segregation, social inequality, and demographic composition. The results indicate that there are substantial racial disparities in the overall rates of FPS for youth, with Black youth having notably higher rates of FPS. These results are particularly pronounced in areas with higher levels of death or injuries to law enforcement. Furthermore, racial segregation is positively associated with FPS for Black youth.
Police Violence and Youth: Contexts, Racial Disparities of Risk, and Consequences
Police violence is not equally distributed across the United States, or within different social groups (Nix et al., 2017; Zimring, 2017). Scholars have found that racial biases in FPS are most acute in more urban communities with higher concentrations of racial minorities (Ross, 2015). There is also preliminary evidence that areas with more residents who hold racially biased beliefs might increase the prevalence of FPS (Hehman et al., 2018). While disproportionate police killings of adolescents and young adults has been identified by prior work (e.g., Edwards et al., 2019), less research has focused on contextualizing the rates of police killings. I argue that community characteristics may disproportionately put adolescents and early adults at risk of experiencing police violence and its consequences.
I focus on three areas of research that collectively establish the importance of identifying the local contexts where law enforcement fatally shoots youth. First, I emphasize the differential likelihood of exposure to the police as a joint product of racial segregation, poverty, and age. Second, I turn to work on the criminalization of youth, which may influence how law enforcement assess risk during encounters with the public. Finally, I discuss the consequences of FPS, which have been linked to an increase in mental and physical health problems for youth, even if they do not directly experience violence themselves.
The Structural Bases of Exposure to Police
A long line of research indicates that police priorities and behaviors are heavily informed by community features (e.g., Bittner, 1967). Communities with extensive racial segregation and poverty receive disproportionate attention by police (Gordon, 2022) and are sites of a disproportionate amount of violent crime (Krivo et al., 2009). Such areas also experience more fatal police shootings (Feldman et al., 2019). While the existing literature on fatal police shootings focuses primarily on how adults are policed, local contexts play an important role in explaining racial disparities in arrests for adolescents and early adults too (Gase et al., 2016). In the current context, community characteristics are particularly important because they increase the overall exposure rates of when, why, and how often youth encounter the police.
Youth experience disproportionate contact with law enforcement due to their structural location and routinized participation in institutions where the police are present. For example, the presence of police officers in American schools has grown rapidly since 2000 (Na & Gottfredson, 2013). This has been linked to an increase in arrest rates for Black students and boys of all races (Homer & Fisher, 2020) and the larger criminalization of school discipline (Hirschfield, 2008). Similarly, police hyper-surveillance in poor, minority communities creates stress and uncertainty for youth as they complete routine activities such as walking down the street (Campos-Manzo et al., 2020). Such sentiments increase levels of distrust and cynicism for youth, who are subject to gendered and racial discrimination during police encounters (Brunson & Miller, 2006; Hitchens et al., 2018).
Age and Discretion During Encounters With Law Enforcement
Police discretion plays an essential role during encounters with the public (Brown, 1981). Law enforcement officers take a variety of factors into account in the regular course of their duties that are partially shaped by the context where the encounter occurs (Klinger, 1997). Officer reactions are also strongly linked to the race of those they interact with (Hinton & Cook, 2021) and an officer’s own demographic characteristics (Brandl & Stroshine, 2013). Implicit racial bias in policing may further exacerbate racial disparities in police encounters and fatal shootings by heightening a police officer’s assessment of risk (Price & Payton, 2017; Spencer et al., 2016).
Researchers have increasingly focused on how suspect age influences police officer perceptions of risk during encounters. Criminologists have long emphasized the age-crime curve, which suggests that criminal activity skews to the young and declines as individuals age (e.g., Hirschi & Gottfredson, 1983; Sweeten et al., 2013). For such reasons, youth, and especially minority youth, are often criminalized by law enforcement (Rios, 2006). These dynamics may place youth at disproportionate risk of an encounter ending in a physical assault, violent arrest, or shooting, particularly in community contexts where police officers face higher rates of violence from suspects.
The Consequences of Police Violence on Youth
Another reason to examine the geography of where fatal shootings occur is the growing evidence of spillover effects. By this, researchers refer to the mental health strain, lowered community cohesion, and other deleterious effects of FPS on those who heard about the shooting, but did not necessarily have direct ties to the individual that was killed (Bor et al., 2018). Evidence indicates that residence in a community where a fatal police encounter occurred or with any police violence may be associated with higher blood pressure, risk of diabetes, and obesity (Sewell, 2017; Sewell et al., 2021). Though there are concerns about the extent and validity of spillover effects due to data quality issues about FPS (Nix & Lozada, 2021), the general point that fatal shootings have implications that reverberate beyond the incident itself is well-documented in the literature.
Youth are particularly vulnerable to spillover effects from police encounters, violence, and killings, which may impede vital aspects of social and emotional development. Legewie and Fagan (2019) examine the impact of surges in surveillance and policing on New York City youth. They conclude that aggressive policing decreases the educational test scores for Black boys, though not for all other groups, such as Black girls or Latinx youth. Browning et al. (2021) point to increased cortisol levels—a marker of physiological stress—among Black boys who were exposed to a police-related death relative to Black girls and whites. A recent systematic review of research on the spillover effects of policing on youth by Jindal et al. (2022) identified 29 published studies establishing several negative consequences of aggressive, racial disparities in policing, including maladaptive coping, psychological distress, and risk taking behaviors.
Overall, these studies underscore the importance of not only analyzing the social and racial identities of who is fatally shot by the police, but also where police shootings occur. Police killings have effects that reverberate far beyond the incident itself with unequal impacts on communities, residents, and especially minority group members.
Data and Methods
The core research design for this study involved creating a longitudinal database of FPS in U.S. counties between 2014 and 2018, with the unit of analysis representing county-years. There are 3,143 counties in the United States, producing an N of 15,715 (=3143 × 5). Due to missing values (described below), I analyze 14,901 county-years. Research on FPS has used a variety of areal units of analysis, including states (Delehanty et al., 2017; Mesic et al., 2018) and counties (Ross, 2015). I focus on counties because they provide a middle ground between more coarse units of analysis like states, which may mask underlying heterogeneity linked to FPS, and more granular units such as census tracts or block groups, which decontextualize the broader jurisdictions where police departments engage with the public. To link shootings to counties, I use the latitude and longitude of each case, based on manually coded textual descriptions of the street address or intersection where each individual was shot. I use the Google Geocoding API to create spatial coordinates for the cases of FPS using the location text.
Identifying Fatal Police Shootings
There is not a single, comprehensive listing of FPS. While the federal government collects data on police killings, it significantly undercounts actual rates of police violence (Feldman et al., 2017a, 2017b; Hickman & Poore, 2016). As a result, many researchers have turned to databases of FPS created using a standardized crowdsourced research design, or projects to document FPS established by major newspapers such as the Washington Post (Shane et al., 2017). Here too, researchers have identified certain limitations and miscodes in the crowdsourced accounts of FPS (Nix & Lozada, 2021) suggesting that issues of data quality remain an important point of consideration when understanding patterns of FPS.
The data used in this study are drawn from a larger project aimed at evaluating and improving estimates of FPS nationwide. The core research design has a two-phase process: first, it begins by combining estimates of FPS from four crowdsourced or journalistic databases, which contain essential information on each shooting. I use the data from the Fatal Encounters Project (http://www.fatalencounters.org/), the Washington Post’s Fatal Police Shootings database (https://github.com/washingtonpost/data-police-shootings), the Guardian’s The Counted project (https://www.theguardian.com/us-news/series/counted-us-police-killings), and the now abandoned United States Police-Shooting Database (https://docs.google.com/spreadsheets/d/1cEGQ3eAFKpFBVq1k2mZIy5mBPxC6nBTJHzuSWtZQSVw/edit#gid=1842418396). Individual cases were cross-referenced, then merged and cleaned based on the name, date, and place of each shooting.
Second, the cleaned data were augmented using a semi-supervised machine learning workflow to identify any additional cases of FPS absent from the crowdsourced and journalistic databases. This component of the data collection protocol used the Google Search Application Programming Interface (API) to programmatically identify websites listing cases of FPS between 2014 and 2018. I used up to 100 Google Search API results for the string “[state] [month] [year] fatal police shooting” iterated exhaustively over each state, month, and year for the analytic period. This resulted in 40,206 unique candidate web pages that were passed to a support vector machine classification algorithm (Murphy, 2012) to identify and retain only the web pages discussing FPS. The final classification algorithm predicted the relevance of each web page with 87% accuracy. I found 17,077 web pages classified as relevant, which were manually reviewed and coded by a team of six trained research assistants, along with a random sample of pages coded as not relevant by the classifier. Regular reliability checks during the data collection process consistently yielded 90% agreement across the overlapping items coded by the research team. The Google Search results were combined with the journalist and crowdsourced databases to create a unified database.
While the data collection process was quite labor intensive, it ultimately produces a more inclusive, exhaustive, and valid approach to sampling FPS. The sampling procedure resulted in a database of over 6,300 fatal shootings between 2014 and 2018. After selecting only those cases where the victim was between 10 and 24 years old, the operational definition “youth” adopted here, there are 974 cases of FPS. This represents 17.1% of all known cases of FPS between 2014 and 2018. I also use information about each victim’s race to create sub-analyses of FPS. The range of 10 to 24 was used to capture early adolescents, teens, and young adults, while excluding younger children. In the complete database, there were 2 children under 10 who were fatally shot, both of aged 6. Due to the rarity of FPS for children I exclude these cases from the analysis, though note that including them does not change the substantive results.
Operationalization
The statistical analysis focuses on patterns of FPS taking place in each county annually between 2014 and 2018. Descriptive statistics for all variables are in Table 1. The regression analysis uses four dependent variables: the number of cases of FPS overall, as well as for Black, Latinx, and white youth. I do not analyze patterns of shootings for other races due to a small number of shootings. Since the risk of FPS is in part shaped by the number of youths in each county, in the analyses that follow I use exposure terms for the number of individuals aged 10 to 24 overall, as well as for Black, Latinx (based on counts of Hispanics), and Whites. As explained below, the primary analytic strategy is multilevel negative binomial models. When exposure terms are included in count models, the raw counts of FPS are converted to the logged rate of FPS. The data on county-level counts of youth aged 10 to 24 overall and by race were drawn from the CDC WONDER database each year for the duration of the analytic period.
Descriptive Statistics for the Dependent and Independent Variables.
Note. Total sample size is 14,901.
I use three independent variables to capture the policing context for each county: first, I create a measure of the police risk in each county using the logged number of police officers assaulted or killed in the line of duty each year. Second, I create a variable for the police capacity of each county, based on the logged number of sworn law enforcement officers per 100,000 population. Last, I include the logged violent crime rate per 100,000 population, which includes cases of murder and nonnegligent manslaughter, rape, robbery, and aggravated assault. The source data for the first two variables comes from the Law Enforcement Officers Killed and Assaulted (LEOKA) series, an annual report collected by the FBI. I restrict the LEOKA estimates to local and county law enforcement agencies since state or federal agencies could not reliably be linked to counties. The measures included the number of assaults and deaths taking place in each county every year, and the total number of law enforcement officers per 100,000 population. The logged violent crime data is based on the annual county-level counts in the Uniform Crime Reports (UCR) series, which is produced by the FBI. In 810 county-years, UCR estimates of violent crime were not available and were excluded from the analysis. These variables are lagged 1 year to avoid potential temporal spuriousness.
The second group of variables emphasizes different aspects of social and racial inequality in each county. I include annually adjusted variables for racial segregation, income inequality, and the percent unemployed. The measure of racial segregation is based on the dissimilarity index. I use variation in the percent of non-Hispanic whites and non-whites in census tracts to determine the level of segregation in each county. The variable is bound between 0 (low racial segregation) and 1 (high racial segregation). Income inequality is based on the Gini Index and is also bound between 0 (low income inequality) and 1 (high income inequality). I use the 5-year American Community Survey (ACS) estimates to create these variables. In all cases, the values are updated annually and are lagged by 1 year.
The final cluster of variables controls for the demographic context of each county based on three measures, all drawn from the annually updated five-year ACS estimates. I include two variables for the percent of residents of each county who are Black (defined as non-Hispanic Black) and Latinx (defined as Hispanic). I also include a variable for population density to control for more urban environments. The distribution of population density was strongly skewed to the right, so I use the natural log in the analysis.
Analytic Strategy
I use negative binomial regression to analyze patterns in FPS overall and by racial group. The variables are overdispersed, making the negative binomial parameterization preferable to other alternatives such as the Poisson model. As noted above, the models include an exposure term, based on the number of youths overall and by race residing in each county. In count models, adding an exposure term converts the raw counts of FPS to logged rates (Hilbe, 2014, pp. 62–66). I specify a two-level hierarchical negative model based on county-years nested in counties. The regression analyses were estimated using the glmmTMB package (Brooks et al., 2017) in R (R Development Core Team, 2023).
Results
Between 2014 and 2018, 13.34% of U.S. counties experienced at least one incident where the police fatally shot an individual aged 10 through 24. Excluding Hawaii, which had no incidents of FPS, all other states had between 2 and 174 shootings over the analytic period. At the state-level, police shootings were most prevalent in California (174 shootings) and Texas (95 shootings). At the county-level, the raw number of FPS was highest in Los Angeles County, CA, which had 56 shootings followed by Maricopa County, AZ and Cook County, IL, with 31 and 29 fatal shootings respectively. Once the data is organized annually, the mean number of shootings per county reduces to 0.06 overall, with respective values of 0.02, 0.01, and 0.02 for Black, Latinx, and white shootings (see Table 1).
Figure 1 shows the distribution of FPS for youth across the United States. The values are the total count of fatal shootings in each county standardized per 100,000 people aged 10 to 24. I truncate the upper bound of the distribution to 20 or more cases of FPS per 100,000 to minimize the visual impact of a small number of outlying values. The population estimates were created by averaging across the annual county-level counts of youths. The map affirms that police shootings are quite widespread. While there are visible clusters in California, Arizona, and Florida, most other states have isolated communities where FPS took place. Overall, Figure 1 corroborates the ubiquity of the police shooting of youth, which are widely geographically dispersed across the country.
The spatial distribution of fatal police shootings for youth, 2014 to 2018.
It is important to contextualize the magnitude of the rates in Figure 1, as they shift the focus from large, populated counties which have high raw counts of FPS, to more rural locations which have low populations and elevated rates. Decatur County, KS, for instance, had a single shooting, but a rate of 260.48 shootings per 100,000 youths. It is important, therefore, to cautiously interpret the relative magnitudes of FPS rates. FPS are rare outcomes of police encounters. While data capturing all police interactions with youth between 2014 and 2018 is not available, it is plausible that over this period several million encounters took place. A single fatal shooting in a rural county will therefore produce a rate that may be disproportionate to the underlying risk that the police will use fatal force.
The discussion above focuses on overall levels of FPS, however, the evidence in Figure 2 indicates that once shootings are examined by race, there are substantial racial disparities in police shootings. Figure 2 shows the annual rate of FPS overall and for Black, Latinx, and white youth. To calculate these values, I took the annual count of FPS divided by the total number of individuals aged 10 to 24 from the CDC’s WONDER database, which was multiplied by 100,000. The overall and race-specific data were the denominator in the rates for Figure 2.
Rates of fatal police shootings of youth overall and by race, 2014 to 2018.
There is a substantial gap in the rates of FPS for Black youth relative to all other racial groups evident in Figure 2. The ratio of Black fatal shootings ranges between a low of 0.64 shootings per 100,000 Black youth 2018 and a high of 0.86 shootings per 100,000 Black youth in 2016. The data also point to more volatility in the rate of Black fatal shootings compared to other racial groups. The temporal trends indicate that the rate of FPS for Black youth may be declining, however, such a decrease was also evident between in 2015 before rising back to its peak value the following year. The rate of Latinx FPS ranges between 0.27 and 0.34 killings per 100,000 Latinx youth, suggesting that the estimates are roughly stable between 2014 and 2018. The rates for white youth are the lowest overall, and range between 0.14 and 0.22 per 100,000 white youth.
To put the results in Figure 2 in perspective, I calculated the ratio of Black to white and Latinx to white rates of FPS per 100,000 over time. The comparison of Black to white rates of FPS indicates that for each white youth killed by the police, between 3.17 and 6.08 Black youth are killed. The comparison between Latinx and white rates is less pronounced, yet still indicative of racial disparities in FPS. Here, the estimates indicate that for each white youth fatally shot between 2014 and 2018, between 1.25 and 2.49 Latinx youth are killed. These results are striking and are strong evidence that the racial disparities in FPS for youth are either comparable or exceeding the rates of police shootings for adults (e.g., Edwards et al., 2019).
How might contextual factors help understand the significant racial disparities in Figure 2? To that end, Table 2 summarizes the coefficients and standard errors from the varying intercept multilevel negative binomial regression model. The table contains the estimates from the full set of independent variables for each of the four dependent variables, including overall logged rates of youth FPS, and specific logged rates for Black, Latinx, and white youth. Since I expect that several of the independent variables are correlated, I reviewed multicollinearity diagnostics for all models in Table 2. None of the variance inflation factors (VIF) were greater than 2.5, indicating that the models did not have problematic levels of multicollinearity.
Multilevel Negative Binomial Regression Predicting Fatal Police Shootings Overall and by Race, 2014 to 2018.
Note. Total sample size is 14,901; Observations are county-years nested in counties. The models include exposure terms for the number of youths overall and by race.
p < .05. **p < .01. ***p < .001 (two-tailed tests)
The variables for policing context have mixed effects on fatal shootings. The covariate for police risk is positive and statistically significant (p < .01), while the variable for police capacity is significant in the models for FPS overall and Latinx shootings (p < .05). Police shootings appear to be related to levels of physical risk faced by police officers, and in counties where more police per capita are assaulted, injured, or killed in the line of duty, fatal force is used more often. The magnitude of police risk varies by race, however, based on a comparison of the coefficients in Table 2. Police risk has the strongest association with Black shootings (b = 0.29) followed by Latinx (b = 0.20) and white (b = .19) shootings. The coefficient for violent crime reaches statistical significance in the model for white shootings (p < .05), but none of the other specifications. Overall, the measures of policing underscore that police risk is a consistent predictor of FPS, while the impact of police capacity or crime is more contingent to the specific group under analysis.
The second group of independent variables measures the levels of social and racial inequality present in each county. Beginning with the variable for racial segregation, there is a positive coefficient (p < .05) for the model of shootings for Black youth. This finding may indicate one mechanism through which the notably higher rate of FPS for Black youth evident in Figure 2 becomes manifest. To the degree that Black youth live in more segregated communities, they are at higher risk for a fatal encounter with the police net of other factors. The other two measures of social inequality—income inequality and unemployment—do not reach statistical significance in any of the model specifications. Overall, racial segregation appears to matter for fatal police shootings, but its effects vary by the race of those shot. Black youth are particularly vulnerable when they live in a racially segregated county.
The final set of variables emphasizes the demographics of each county. There is a positive association between the percent of Black residents in a county and the count of police shootings of youth overall and for Black and white youth (p < .05), and a negative association for Latinx shootings (p < .05). The percent of Latinx residents is positive in the models for shootings overall and for Black youth (p < .01), but not statistically significant in other models. Additional analyses that excluded the exposure term from the model specifications indicated that the coefficients for the percent Black and percent Latinx were positive and statistically significant (p < .05). Accounting for the population at risk via the exposure term changes the underlying risk profile of shootings and helps to explain these findings. Finally, there is a negative association between logged population density and FPS overall and for whites (p < .001), and a positive coefficient for Blacks (p < .001). The effect is not statistically significant for Latinx shootings (p > .05). More densely populated urban environments, therefore, increase the risk of FPS for Black youth, while decreasing or having a null impact for other groups in the analysis. This too may provide a partial explanation for the disproportionately high rate of FPS for Black youths.
Discussion, Limitations, and Contribution
Fatal police shootings are a significant social problem that disproportionately impacts racial minority groups (Edwards et al., 2019). Scholars have pointed to the importance of community characteristics in understanding patterns of FPS (Feldman et al., 2019; Ross, 2015; Zare et al., 2022), but less work has focused on where the police fatally shoot youth. This study aims to address this gap by examining where 974 youth were fatally shot by the police between 2014 and 2018. The results indicate that there are significant racial disparities in police shootings, with Black youth killed at disproportionately higher rates compared to other races. Community contexts partially explain these findings. Counties where police work is riskier, and counties that are racially segregated, have more police shootings of Black youth.
The results make it clear that there is no single set of community circumstances that consistently predicts all types of FPS. The exception to this is counties with higher risks to police officers. As noted above, young people are hyper-criminalized (e.g., Rios, 2006) which may increase police perceptions of risk during encounters with youth. The strong, positive, associations for police risk across all model specifications indicate that the police may be more willing to use deadly force during encounters with youth in counties where law enforcement work is particularly risky. The correlation coefficient between police risk and violent crime is 0.18, suggesting that the risks in police work are at best moderately associated with patterns of violent criminal activity.
Beyond police risk, the variables predicting FPS are contingent on the specific racial group(s) under investigation. The overall implications for the models of Black youth may point to the larger role of systematic racism in putting these individuals at risk. Specifically, densely populated, majority-minority communities that are racially segregated are part of a larger social legacy of discriminatory housing and lending, the suburban exodus of white residents, and other larger manifestations of racial disparities. These findings are less salient for Latinx youth, as FPS appear to stem more from the risk and capacity of law enforcement agencies. Additional research more directly engaging with the micro-dynamics of police encounters that become fatal may help further explain the differing impacts of the contextual analysis here.
While this study did not examine the consequences of FPS, it is a significant issue that should be examined by future work. Police exposure has been linked to several negative outcomes for youth (Jindal et al., 2022; Legewie & Fagan, 2019) and adults (Sewell et al., 2021). Fatal shootings of youth are geographically dispersed, indicating that the patterns observed in this study may have widespread spillover effects in communities across the United States. An important question for future work, then, is whether fatal encounters between youth and the police exacerbate any spillover effects for adolescents and adults.
This study has some limitations. First, I do not examine the dynamics of the encounters that produce fatal shootings. The steps between the initiation of a police encounter and a fatal shooting are key aspects in understanding overall trends in FPS. My research cannot speak to these important micro-situational sequences of action. However, the contextual approach used here does illustrate where shootings occur, which is necessary to build a comprehensive understanding of FPS. Second, the analytic period ends in 2018, and therefore does not provide insights into FPS between 2019 and the present. I suggest that this limitation is minimized by my use of multiple sources of data sources and external validation to capture FPS more comprehensively. It is a significant undertaking to clean, merge, and standardize the existing databases on FPS, creating a lag in work that attempts to use multiple sources of information.
Overall, this study makes an important contribution to understanding the community contexts that facilitate and impede the police killings of youth. Little work has directly focused on patterns of FPS for early adolescents, teenagers, and young adults, and the results here can be a baseline for future inquiry. The extensive racial disparities in FPS identified here indicate that rates of FPS for Black youths may exceed those of adults. This study underscores the role of local community contexts in the risk of fatal police shootings and identifies potential factors that increase or decrease patterns of shootings overall, and by race.
Footnotes
Acknowledgements
I thank Aalyiah Butler, Megan Faust, Sarah Medina, Emma Sager, Clifford Soloway, Ella Catherine Strahan, Aubry Tedford, and Jarrod Wall for research assistance. I also thank the editor and anonymous reviewers for their helpful comments and feedback.
Data Availability Statement
The data on fatal police shootings is available from the author.
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work has been supported (in part) by Grant # 1902-11743 from the Russell Sage Foundation. Any opinions expressed are those of the principal investigator(s)/Administrator of the Tulane Education Fund AKA Tulane University alone and should not be construed as representing the opinions of the Foundation. This study was also partially funded by a Carol Lavin Bernick Faculty Grant from Tulane University.
