INTRODUCTION
Over the years, extensive data overwhelmingly confirm a disturbing trend: Blacks are killed by the police at significantly higher rates than any other racial group in the country (Demby, 2017). These killings are disproportionately carried out by White[1] officers. A study conducted by ProPublica between 2012 and 2014 found that White officers were responsible for 68 percent of all people of color killed (Gabrielson et al., 2014). While racism is often cited as a major factor in the shooting of unarmed Blacks (Mesic et al., 2018), little is known about such incidents involving officers of color. After all, 32 percent of all police killings of people of color are perpetrated by officers of color (Gabrielson et al., 2014). From 2012-2014, Black officers accounted for only 10 percent of all police killings, but among that 10 percent, 78 percent of those killed were Black (Gabrielson et al., 2014).
The topic of Black officers’ involvement in police killings of unarmed Black civilians changed dramatically on January 28, 2023, when the Memphis police department released a video showing the brutal beating of Tyre Nichols. After being stopped for reckless driving, five Black police officers pulled him from the car and subjected him to a severe beating, resulting in his tragic death. This incident, unique compared to other high-profile killings of unarmed Blacks, has prompted questions about the potential impact of policies aimed at increasing the number of Black police officers to reduce police violence against Black civilians.
The prevailing narrative suggests that Black officers, due to their strong racial identities, can better relate to Blacks and are less likely to shoot unarmed members of their own race when compared to their White counterparts. However, we argue that this narrative is incomplete. We argue that Black officers undergo similar socialization as their White counterparts and can also develop anti-Black attitudes. The key difference lies in the fact that Black officers may also possess robust levels of racial identity that act as a constraint on their anti-Black attitudes, reducing the likelihood of them resorting to shooting unarmed Blacks.
In this article we probe into the question of why Black police officers choose to shoot, or not shoot unarmed Black civilians. We borrow methods that have traditionally been used to examine bad behavior on the part of White police officers to understand such behavior by Black officers. Specifically, we employ the use of a simulation-based reaction-time game administered on the computer to investigate whether officers are more likely to shoot or not to shoot unarmed Black civilians when compared to armed Black suspects. One of the major shortcomings of the existing literature on police decision making is that there is a dearth of research that includes African-American police officers as participants. For example, Correll et al. (2007) recruited 10 African Americans out of 113 in one study and 18 out of 33 in another. Additionally, Ma & Correll (2011) used only 6 blacks out of 31 subjects in their study, and Plant and Peruche (2005) research had only one African-American police officer. To our knowledge, Correll, et al. (2002) has included the largest number of African-American participants, with 25. The research in this paper seeks to build on the extant literature by focusing solely on African-American police officers and some important factors that may influence their behavior towards Black citizens.
Here, we contend that Black officers who shoot unarmed Black civilians identify less with members of their race on the implicit level and possess anti-Black attitudes that are comparable to those of White racists. In contrast, we advance the argument that Blacks who possess racial solidarity with other Blacks will be less likely to shoot unarmed Blacks in error. In making this argument, we pull from the seminal work of W.E.B. DuBois and apply the double consciousness concept. According to DuBois:
It is a peculiar sensation, this double-consciousness, this sense of always looking at one’s self through the eyes of others … One ever feels this twoness, an American, a Negro; two souls, two thoughts, two unreconciled strivings.
DuBois (1903) concept of ‘double consciousness’ speaks to the internal tension that exists within the Black psyche. Many Blacks are socialized to live in two Americas. On the one hand, they share the anti-Black sentiments of American mainstream culture as is manifested through negative stereotypes of Blacks, and on the other hand, they live in communities whereby issues such as economic subordination and police violence experienced by Blacks, drive Blacks to collectively align themselves with other Blacks. We argue that Black police officers, like other Blacks, possess both sets of attitudes. For these officers, they see the world through both a ‘Blue’ lens as police officers and a Black lens as African Americans. And while we argue that Black officers’ anti-Black attitudes will lead to them making more errors in shooting unarmed Black civilians, we also argue that attitudes associated with racial solidarity will constrain such decisions.
In examining Black police officers’ decisions to shoot unarmed Black civilians as a function of ‘double consciousness’ (Du Bois, 1903), the current research utilizes both pro-Black and anti-Black attitudes measured on the explicit and implicit levels. Specifically, we apply an implicit measure of Black identification, and an explicit measure of disillusionment with inequality in America, both of which we expect to inhibit erroneous shootings of unarmed Black civilians by Black police officers. In contrast, we expect that anti-Black racist stereotyping (internalized racism), will encourage erroneous shootings of unarmed Black civilians. The model also controls for age, which we argue serves as a proxy for socialization, that is, either exposure to ‘Blue’ police subculture or exposure to racism in American mainstream culture more generally (or both). We expect that age will be associated with a higher error rate in shooting unarmed Black civilians, although the opposite may be possible based on research by Paoline and Terrill (2007).
This article provides a discussion of bureaucratic representation associated with the racial makeup of police departments. Historically, civil rights advocates have made the case that the remedy to police misconduct toward Blacks is to increase the number of Black police officers. In some instances, however, Black police officers lose sight of their responsibility to the Black community and adopt anti-Black attitudes, which can lead to negative police behavior toward Blacks. Such behavior can be a function of socialization e.g., police training. For example, a police force in Georgia was found to have used Black civilians as targets for police shooting practice (Bellamy, 2023). Hence, we examine the literature on police socialization and anti-Black attitudes. We then delve into the literature related to implicit bias and the use of simulators to study racial bias in the shooting of unarmed Black civilians. The research continues by discussing the extant literatures heavy focus on racial solidarity when examining Black intra-racial attitudes. In this vein, the article introduces an innovative measure of racial identity using methods used to operationalize nonconscious attitudes. We argue that Black police officers develop and retain some of the same attitudes as Black civilians, despite being socialized as a police officer (or American citizen more generally) where they may have been exposed to negative stereotypes of Blacks. One such set of attitudes is the disillusionment that some Blacks may have based on their experience as American citizens. Many Blacks believe that society and the legal system, for example, are unfair to Blacks (Dawson, 2001). After exploring discussions of racial solidarity, we introduce literature on internalized racism as a measure of anti-Black attitudes. We then delineate the hypotheses, present the data and methods section, followed by the findings. Lastly, we close with a discussion and conclusion.
RACE AND POLICE REPRESENTATION
The brutal beating of Tyre Nichols by Black officers runs counter to the expectation of civil rights leaders during the 1960s. For example, given the dearth of Black officers in the 1940s, Martin Luther King, Sr. pushed for the recruitment of Black officers (Foreman, 2017). The impetus behind his efforts was rooted in the notion that Black officers would truly serve and protect Black citizens. This expectation was based on theories of representative bureaucracy where, according to Wilkins and Williams (2008, p. 656), “[e]arly scholars assumed that passive representation would naturally translate into active representation.” The normative origins of representative bureaucracy theory can be traced to Kingsley (1944) who theorized that in a democratic system, bureaucracies should reflect the people they serve (Haider-Markel et al., 2022; Keiser & Haider-Markel, 2022). Passive representation is concerned with “whether the bureaucracy has the same demographic origins—sex, race, income, class, religion—as the population it serves” independently of the bureaucratic agent’s behavior (Wilkins & Williams, 2008, p. 655). Conversely, active representation assumes that “bureaucrats will act purposely on behalf of their counterparts in the general population” (Wilkins & Williams, 2008, p. 656). Passive representation may not result in the fair treatment of Blacks. Indeed, Wilkins and Williams (2008) found that the presence of Black police officers in police departments led to an increase in racial profiling, directly contradicting this assumption. This suggests that the expected correlation between passive and active representative bureaucracy can be neutralized or even reversed by factors that Wilkins and Williams (2008) trace to socialization in ‘Blue’ police culture. They arrive at this conclusion after analyzing “in-depth individual interviews and focus group discussions with Black police executives and officers, [that] seem to suggest that the pressure to conform to the organization or to achieve the goals of the organization weighs heavily on black officers” (Wilkins & Williams, 2008, p. 660).
PRO-‘BLUE’ AND ANTI-BLACK POLICE SOCIALIZATION
Police officers tend to possess occupational identification with their departments that exceeds the occupational identification observed in other organizations (Van Maanen, 1978a, 1978b). This occupational identification is the result of a socialization process “whereby incumbents communicate organizational culture, both formally and informally, to newcomers” (Britz, 1997, p. 129). Research on police socialization goes back to the 1960s when White-male dominated police departments were notoriously lacking in diversity. Skolnick (1966) and Brown (1988) investigated socialization processes and occupational identification within police organizations and found according to Britz (1997, p. 133), “that individual characteristics did not significantly alter the socialization process and occupational solidarity among newcomers.” Instead, evidence suggests that ‘Blue’ police culture “was so salient in police organizations that individuals, regardless of personality differences, adopted the norms and beliefs of departmental members” (Britz, 1997, p. 133). Thus, research by Alex (1969) suggested that the few then employed African-American police officers did not differ significantly from their White peers on working personality, behavior manifestations, or general attitudes" (compare Britz, 1997, p. 131). This suggests that female or non-White officers internalize the ‘Blue’ organizational culture that has historically been characterized by White supremacist bias. Even after the gradual and continuous introduction of women and non-White officers to police organizations, Britz (1997) finds that “background characteristics, including race and gender, did seem to be secondary to occupational culture. … This lends support to earlier research that suggests that the police subculture is so salient to law enforcement personnel that individual differences are quickly overwhelmed” (Britz, 1997, p. 144).
This raises questions about the role of training. There is some evidence that police training at academies transforms new members. Unfortunately, this often goes in the wrong direction whereby training reinforces the role of White supremacist socialization. For example, Moskos (2009) detects a marked increase in feelings of ‘Blue’ group solidarity through training. Simon (2021, p. 25) carefully analyzes police training at academies and finds that fighting crime is conceptualized as a war against a Black male “enemy.” Bell (2018, p. 534) argues that Blackness itself is associated with criminality and that current policing practices establish Black men as “always suspicious and always potentially up to no good.” Simon (2021, p. 25) writes, “This gendered and racialized way of conceptualizing the enemy puts Black officers in a bind. The Black officers I met, both men and women, expressed an investment in this white, male institution, often justifying over-policing in communities of color and defending high-profile police shootings of Black men. At the same time, they also described being subject to racist policing when off duty.” Given the power of ‘Blue’ police socialization, Simon (2021) argues that “that increasing demographic diversity in policing does not necessarily affect how the police use violence” (see also Frydl & Skogan, 2004; Johnson et al., 2019; McElvain & Kposowa, 2008; and B. W. Smith, 2003). Similarly, Schuck and Rabe-Hemp (2021, p. 578) find in a study of 655 recruits from multiple police academies in the United States that ethics training can mitigate the effect of socialization, but that “The results confirm the role of the academy in socializing officers into the negative aspects of the traditional police culture” (Schuck & Rabe-Hemp, 2021, p. 578).
IMPLICIT ANTI-BLACK SHOOTER BIAS
While implicit bias as a concept is used loosely in popular discourse, arguably, very few people understand how implicit bias works. Implicit biases are automatic associations in people’s minds between groups and positive or negative stereotypes. According to the Kirwan Institute (2015), such associations are activated involuntarily and outside of our conscious control. When police officers possess anti-Black implicit biases, they may react in discriminatory ways against Black civilians, even if they do not endorse explicit prejudice.
One of the first experiments using an implicit bias test to examine whether race had an impact on police officers shooting unarmed civilians used the priming paradigm. Law enforcement agents were primed with stimuli that consisted of either a White or an African-American face, followed by a target object consisting of a gun or a tool (Payne, 2001). The respondents were given only 500 ms to identify the target. The results revealed that participants identified the gun more quickly and more accurately when it was preceded by an African-American face as opposed to a White face. In contrast, the respondents identified a tool more quickly if it was preceded by a White face as opposed to an African-American face. Correll et al. (2002) expanded Payne’s research by introducing the First-Person Shooter Task (FPST; also known as the Police Officer’s Dilemma). This task was designed to provide a more realistic experience by including whole-body photographs of armed or unarmed citizens with backdrops that contained images e.g., park benches, street cars, etc. In these experiments, subjects sit in front of a computer screen and are instructed to press a computer key to “shoot” and a different key to “not shoot.” The goal of the task is to shoot armed targets and to avoid shooting unarmed targets. Correll et al. (2002) recruited participants from the general population (non-law-enforcement officers) to participate in the shoot or don’t shoot experiment. The research finds that the participants erred more in shooting unarmed African-American civilians who were holding innocuous objects such as cell phones when compared to White civilians. In addition to shooting the wrong target they also were found to mistakenly shoot an unarmed target faster if the target was African American than if the target was White. The authors enhanced this research by adding police officers as participants (Correll et al., 2007). The police officers were found to exhibit a similar pattern as the general population participants by mistakenly shooting an African-American civilian versus a White civilian. They were also quicker to identify a White civilian as being unarmed when compared to an unarmed Black civilian. Similarly, participants made more errors in shooting targets that were rated by a separate sample as having more Afrocentric features (e.g., skin tone broadness of the nose) when compared to Eurocentric features (Ma & Correll, 2011). Sim, Correll and Sadler (2013), found that police officers from special units who were assigned to minority gangs were more likely to exhibit racially biased behavior when making their decisions. To compare racial differences between participants, Correll et al. (2002), included a sample that possessed both African-American police officers and civilians. They found that African Americans possessed anti-Black racial biases equivalent to those of Whites.
Building on this research our article focuses solely on African-American police officers using both implicit and explicit measures of their attitudes towards African Americans. We expect some African-American officers to possess some of the same anti-Black attitudes of mainstream American culture. Consistent with the ‘double consciousness’ hypothesis (Du Bois, 1903), however, we also expect some to possess more pro-Black attitudes (e.g., implicit Black identification, or disillusionment with legal and social biases in America). We investigate how these potentially conflicting attitudes affect their behavior in the implicit shoot/don’t shoot task when faced with unarmed Black civilians.
IMPLICIT BLACK IDENTIFICATION
The internal conflict depicted by DuBois’s double consciousness thesis of African Americans identifying separately as both Americans and Blacks has gone largely unacknowledged in scholarly research. Traditionally, the political science literature has measured African-American group identification using items asking about feelings of closeness toward Blacks (Berelson et al., 1954; Converse & Campbell, 1968; Herring et al., 1999; Tate, 1993). Racial identification, as a construct, has taken on almost as many labels as there are researchers on the topic. These labels include linked or common fate (Tate, 1993 and Dawson, 1994), Black consciousness (Gurin et al., 1980, 1989; Gurin & Epps, 1975; Miller et al., 1981; Reese & Brown, 1995; Shingles, 1981), and Black nationalism/Black autonomy/racial solidarity (R. A. Brown & Shaw, 2002; Davis & Brown, 2002; Dawson, 2001; Welch et al., 2001). The common denominator between all of these labels is the shared experiences and unique worldview possessed by Blacks.
The focus on racial identity from the lens of racial collectivism paints an incomplete picture of DuBois’ double consciousness philosophy. While we are very familiar with the racial solidarity possessed by many Blacks, we know very little about intra-group differences within the Black community, and by extension, their strong allegiance to traditional American values prescribed by the dominant group (in this case Whites).
Prior research has primarily utilized explicit measures of racial identity by employing Likert-items to measure racial identity. Social desirability pressures for ingroup solidarity, however, may lead some Blacks to overstate how close they feel towards members of their own group. To circumvent this problem, we apply a reaction-time based implicit measure of identification with Blacks as a group. The reaction-time task is based on a social-psychological measure of identification strength with romantic partners in long-term romantic relationships (Aron et al., 1991). The same measurement procedure can also measure implicit identification with social ingroups like sororities and fraternities (Coats et al., 2000; E. R. Smith & Henry, 1996).
Craemer and his colleagues applied the reaction-time measure to racial in- and out-groups and found it to predict political opinions regarding race-related policies like aid to minorities and affirmative action (Craemer, 2008), opinions regarding Black reparations (Craemer, 2009), and vote intentions for President Barack Obama (Craemer et al., 2013). Blacks tend to score higher on the implicit Black identification measure than Whites, although some Whites identify with Blacks in politically meaningful ways. However, in both groups, there is politically meaningful variance whereby those Whites or Blacks who score higher on implicit Black identification are also more likely to support aid to minorities, affirmative action, Black reparations, and support for President Barack Obama.
The implicit Black identification measure is believed to measure the cognitive overlap between an individual’s self-concept and that individual’s mental representation of the other, be the other a romantic partner (Aron et al., 1991), a social group (Coats et al., 2000; E. R. Smith & Henry, 1996), or a racial group (Craemer, 2008, 2009; Craemer et al., 2013). Higher levels of implicit Black identification among Black law enforcement officers should be associated with lower error rates in timed decisions to shoot unarmed Black civilians.
BLACK DISILLUSIONMENT WITH INEQUALITY IN AMERICA
During the final weeks of the 2008 Democratic presidential-primaries, then presidential candidate Barack Obama endured heavy criticism from the media, based on comments made by his former pastor, Reverend-Dr. Jeremiah Wright. This flap emerged shortly after incendiary comments were released by the media concerning Dr. Wright’s criticism of the American treatment of Blacks. Specifically, he stated, “[t]he government gives them the drugs, builds bigger prisons, passes a three strikes law, and then wants us (Blacks) to sing ‘God Bless America.’ No, no, no; not ‘God Bless America,’ God damn America… for killing innocent people; God damn America for treating its citizens as less than human…” (Wright cited in Ross & El-Buri, 2008). Such sentiments capture what Dawson (2001) has conceptualized as a disillusioned-liberal ideology or the belief that America has failed to fulfill its promise of equality. This disillusionment with inequality in America is rooted in the many observations of racial inequalities witnessed by Blacks daily. Empirically, the work of Dawson (2001; see also Block, 2011) proves to be the most rigorous effort at measuring and testing disillusionment with inequality among Blacks. Using the concept disillusioned liberalism, Dawson (2001, p. 280) reports that over 75 percent of Blacks believe that they live in a country that is “racially unjust.” Disillusionment with inequality in America should render Black police officers more reluctant to shoot unarmed Black civilians in a timed shooter-bias task. Here, we argue that Blacks with disillusioned attitudes may be likely to empathize more with other Blacks and therefore be less likely to shoot unarmed Blacks.
INTERNALIZED ANTI-BLACK RACISM
A surprising number of studies have found that some Blacks are just as likely, and in some cases more likely, to provide negative stereotypes of other Blacks. Parent (1985), for example, finds that more than half of all Whites, 59.1 percent, and almost half of all Blacks, 46.5 percent, believe that many of the problems faced by Blacks in this country are a function of their lack of will power and motivation. Parent’s data also reveal that approximately 53.7 percent of Blacks, compared to roughly 78.1 percent of Whites, agree that many of the problems that confront Blacks are brought on by Blacks themselves. Additionally, Sniderman and Piazza (1993) report that more African American respondents (40 percent) stated that Blacks are irresponsible than White respondents (21 percent). These authors also find that roughly 59 percent of Blacks describe Blacks as aggressive or violent, compared to 52 percent of Whites. Pride (1995) reports that 56 percent of Blacks compared to 60 percent of Whites responded that Blacks do not get the better things in life because they simply do not try hard enough.
Research by Cohen (1999) proves to be instructive here. Her work is important because it moves beyond the dominant paradigm of studying race relations as a function of the dominant group’s regulation of the marginal group, to a discussion of the marginal group’s regulation of their own group members. Cohen (1999, p. 43) states that White stereotypes of Blacks “have great staying power.” Hence, it is argued that because Blacks are exposed to the same agents of socialization that depict Blacks in a negative light, they are bound to internalize negative stereotypes and then apply them when judging members of their own race. According to Cross (1991, p. 191), “they look on Black people with a perspective that comes very close to that of white racists.” Here, we employ internalized racism to capture the internalization of anti-Black stereotypes by Blacks. Carter (2007, p. 36) defines internalized racism as “self-blame and feeling responsible in the context of racism.” In addition, Williams and Williams-Morris (2000) indicate that internalized racism is the acceptance by marginalized racial groups of negative racial stereotypes of themselves. Speight (2007) argues that the dominant group in society normalizes negative racial stereotypes. Watts-Jones (2002, p. 592) notes when Blacks internalize racist stereotypes they experience “self-degradation and self-alienation: one that promotes the assumptive base of … inferiority.”
Given that Blacks are not monolithic with respect to their attitudes, it is plausible that not all Black police officers implicitly identify with Blacks or perceive a sense of disillusionment with inequality in America. As stated here, some Black officers may internalize negative racial stereotypes of members of their own group. Similar to the ‘double consciousness’ described by DuBois (1903), we argue that Black law enforcement officers may possess both, pro-Black implicit identification, and/or disillusionment with inequality in America, as well as internalized anti-Black racism. Depending on what attitude is salient, they may be more or less trigger happy when making decisions about whether to shoot unarmed Black targets in the timed shooter Bias task.
DURATION OF (POLICE) SOCIALIZATION
According to Reiner (1985, p. 162), some scholars argue racism may be an unfortunate byproduct of occupational solidarity in police departments, while other scholars suggest the level of racism inherent in the law enforcement community is actually “a reflection of the racism prevalent in the … society and the social groups from which the police are drawn” (compare Britz, 1997, p. 130). This may suggest that we can use age as a proxy for immersion in anti-Black White supremacist culture, be it a ‘Blue’ police subculture specifically, or mainstream U.S. culture more generally. However, using age as a proxy for socialization does not allow us to distinguish between the specific and the more general explanation.
HYPOTHESES
We hypothesize that Black police officers will be less likely to shoot unarmed Black targets in error (referred to as ‘false alarms’) as the scores on the implicit Black identification measure increase, and as their disillusionment with inequality in America increases. Blacks who possess high levels of identification with other Blacks will feel closer to Blacks and be less likely to shoot unarmed Blacks in error. Similarly, Blacks who feel that society and the legal system are unfair to Blacks will empathize with other Blacks and be less likely to shoot Blacks in error. On the other hand, false alarms will increase as internalized anti-Black racism increases. Like Whites who possess anti-Black attitudes toward Blacks, Blacks with internalized racist attitudes are expected to behave like racist Whites. Based on the police socialization literature, we further expect that false alarms will increase with age. We argue that older officers have served longer in an atmosphere of ‘Blue’ police subculture and have been exposed to the anti-Black connotations of American racial mainstream culture for a longer time than younger officers. This socialization could lead such officers to erroneously shoot Black civilians.
Formally:
H1: An increase in implicit Black identification scores will result in decreasing shootings of unarmed Black civilians.
H2: An increase in scores on the disillusionment with inequality scale will result in decreasing shootings of unarmed Black civilians.
H3: An increase in scores on the internalized racism scale will result in increasing shootings of unarmed Black civilians.
H4: An increase in officer age will result in increasing shootings of unarmed Black civilians.
DATA AND METHODS
This study included a total of 43 Black law enforcement officers recruited from a metropolitan area in the deep-South. These officers consisted of city police, college-campus police, hospital police, and members from sheriff departments.
Participants were recruited using a snowball sampling method. Each of them was informed that they would receive $20 for participating in the study. Prior to arriving at the lab, the participants were sent a survey that consisted of questions related to their demographics and racial and political attitudes. They were then invited to participate in the experiment.
The research design is based upon a simulation game designed to test decision making processes of law enforcement officers under time pressure. Prior research has incorporated such games as a way to approximate real life situations. Given that it would be unethical to shoot actual humans, this is the closest we can come to mimicking real life decision making about shooting or not shooting human beings. Inquisit by Millisecond.com created a video game patterned after that of Correll et al. (2002). In this experiment, the participant sits at a computer and is given instructions to place his or her fingers on the keyboard using one finger to indicate a ‘shoot-,’ and another finger to indicate a ‘don’t shoot’ decision. The targets are pictures of either Black or White individuals that are either armed (carrying a gun) or unarmed (carrying an innocuous object such as a soda can or cell phone; see Correll et al., 2007, for details).
Participants are provided with 20 practice trials and 80 test trials (i.e., 20 armed Blacks, 20 armed Whites, 20 unarmed Blacks, and 20 unarmed Whites). Each of the targets appears on the screen for 850 milliseconds and timeout afterwards. The goal is for the officer to ignore the race of the target and to shoot whenever the target is armed, and not to shoot when the target is unarmed before the trial is timed out. Once the simulation task is complete, the participant is provided his or her results. The results reveal the number of errors that were made in either shooting an unarmed target or failing to shoot an armed target and how fast the participant responded. At the end of the session, the participant is debriefed and paid $20.00.
Black False Alarms: The dependent variable for our analysis is the number of erroneous decisions to shoot an unarmed Black target (Black false alarms). This variable is a count variable that is increased by 1 each time the officer makes an error in shooting an unarmed Black target. The average number of Black false alarms made by the Black police officers that participated was 2.93 with a standard deviation of 2.772. The distribution of the dependent variable dictates that a model be designed to capture effects on a count-level measure of false alarms. Therefore, we employ negative binomial regression. The independent variables we use are implicit Black identification, disillusionment with inequality in America, internalized anti-Black racism, and officer age.
Implicit Black Identification: The implicit Black identification measure is believed to measure the cognitive overlap between an individual’s self-concept and that individual’s mental representation of Blacks as a group (Aron et al., 1991; Coats et al., 2000; Craemer, 2008; E. R. Smith & Henry, 1996). It measures group identification strength independently of group membership and has proven highly predictive of pro-Black policy opinions among both Black and white respondents in a number of studies (Craemer, 2008, 2009; Craemer et al., 2013). The implicit Black identification measure proceeds in two independent steps, one explicit, and the other one implicit, separated by a distracter task. The former step represents an explicit trait-survey requiring respondents to rate first themselves and later Blacks as a group on a large number of different personality traits (here 45 trait words) using 7-point response scales ranging from 1= “Not at all” to 7 = “Extremely” descriptive (for purposes of analysis ratings of 5 to 7 were considered descriptive, 1 to 3 non-descriptive and rating of 4 neutral). Equal portions of those traits are positive, negative, and relatively neutral. The sole purpose of the explicit trait survey is for the researcher to determine which of the trait words the respondent would use to describe the self as well as the group (matching traits) and which trait words the respondent uses for the self but not the group, or vice versa (mismatching traits). The classification of “matching” and “mismatching” traits is crucial for the interpretation of reaction time patterns during the other step of the procedure.
The other step, the actual reaction time task, requires the respondent to describe him- or herself as quickly as possible using the same 45 trait words (in random order). Three blocks of 45 trials are repeated for a total of 135 trials to allow computation of Cronbach’s reliability coefficient (α = .484; this would be considered low for an explicit measure, but it is remarkably strong for a reaction-time based implicit measure). A trait word appears on the computer screen and the respondent is asked to indicate as quickly as possible whether it is descriptive of the self (1=Yes) or not (0=No). It is important to note that during this task only the self is rated and no reference is made to Blacks as a group or any other racial group.
When reaction times in the self-description task are compared for “matching” and “mismatching” traits separately, individuals who identify with Blacks as a group will display distinct facilitation and inhibition effects, i.e., faster reaction times for “matching” than for “mismatching” traits (Aron et al., 1991; Coats et al., 2000; E. R. Smith & Henry, 1996). These effects do not occur among individuals who do not identify with Blacks as a group on the implicit level. Since these effects only occur among individuals who feel close to Blacks on the implicit level, the reaction time difference between mismatching and matching traits can be converted into a measure of implicit Black identification B according to the formula:
B=¯tMismatch−¯tMatch
In this formula represents the average reaction time a respondent takes to decide whether a mismatching trait is self-descriptive, and the average reaction time the respondent takes to decide whether a matching trait is self-descriptive.
It is also important to note that the traits that are rated as matching, or the number of matching traits during the explicit trait survey, are not used to infer implicit Black identification. Implicit Black identification is inferred only from differential reaction times in response to matching and mismatching traits during the timed trait self-rating task. Aron et al. (1991) interpret this reaction time interference effect as the result of mental “self-other overlap” (see also Coats et al., 2000; E. R. Smith & Henry, 1996). Aron et al. (1991, p. 248) write, “A possible explanation of the . . . effect is that the cognitive structure of the ‘self’ overlaps with the cognitive structure about the other . . . Thus when a trait is descriptive of ‘self’ but not other, there is a bit of confusion in deciding whether it actually represents the self.” This confusion does not occur if the individual does not identify with the other.
Disillusionment with Inequality in America: Disillusionment is based on two explicit survey items in agree-disagree format: “American society [overall] is unfair to Blacks”; and “The American legal system is unfair to Blacks” (Dawson, 2001). These two items load onto one dimension, with a Cronbach’s reliability coefficient α=.81. Higher values represent more disillusionment with racial inequality in America.
Internalized Anti-Black Racism: Internalized racism is measured using a standard item from the National Election Studies (NES) regarding Black and White intelligence. The instruction reads: “On the following scale a score of 1 means that you think almost all of the people in a given group have the characteristic indicated on the scale. A score of 7 means that you think most people in the group have the opposite characteristic. A score of 4 means that you think that most people in the group are not closer to one end or the other, and of course, you may choose any number in between.” The 7-point scale presented labeled the endpoints 1 = “Intelligent” and 7 = “Unintelligent,” and the questions displayed in random order were, “Where would you rate BLACKS in general on this scale?” And, “Where would you rate WHITES in general on this scale?” Black intelligence scores are subtracted from White intelligence scores so that higher numbers reflect the old-fashioned racist stereotype of Whites as supposedly more “intelligent” than Blacks.
Officer Age: The model controls for officer age as a rough proxy for duration of police socialization and/or exposure to the anti-Black connotations of American racial culture. It is based on the self-reported age. Because we failed to ask the officer’s duration of tenure, we cannot analytically distinguish between the effects of police socialization specifically (duration of tenure) from exposure of anti-Black American culture more generally (age). Since both hypotheses point in the same direction, and since older officers are likely to have served longer, we can still interpret the results. However, we recommend distinguishing the two explanations in future iterations of this research.
RESULTS
Descriptive statistics are presented in Table 1 and correlations between dependent and independent variables in Table 2. On average, the sampled Black police officers made 2.93 errors in falsely shooting an unarmed Black target (Black False Alarms) with a minimum of 0 and a maximum of 10. Further, 36 percent of the law enforcement officers were female, and the average age of the officers was 47, ranging from 27 to 65 years. The average reaction-time difference between mismatching and matching traits in the timed self-description task that is part of the implicit Black identification measure, was 31.81 milliseconds, indicating a small but significant (p<.05) pro-Black identification effect. Reaction time differences ranged from -233.95 ms (indicating distance to Blacks) to 316.69 ms (indicating closeness to Blacks).
Average score on the disillusionment with inequality scale is roughly 7 points based on scale from 3 to 10 with a midpoint of 6.5. This average score falls significantly above the midpoint (p<.05) suggesting that the sampled officers tended to be disillusioned with inequality in America. The sample mean of internalized racism (anti-Black stereotypes regarding Black vs. white intelligence) was -0.17, which is indistinguishable from unbiased (0) at conventional levels of statistical significance.
Based on Table 2, Black false alarms are correlated negatively with implicit Black identification and disillusionment with inequality in America (both at the p<.05 level of significance). Internalized racism fails to correlate with the other variables at conventional levels of statistical significance. Officer age is positively correlated with Black false alarms.
Table 3 presents the results from the negative binomial regression with robust standard errors. The dependent variable is the number of Black false alarms; that is, erroneous shootings of Black targets in the timed shooter-task. The frequency distribution of Black false alarms is presented in Figure 1. Negative binomial regression is appropriate for count-data with over-dispersion, that is when the conditional variance (here 7.685) exceeds the conditional mean (here 2.93). The negative binomial model represents a generalization of Poisson regression with an extra parameter to model over-dispersion. We use robust standard errors to prevent heteroscedasticity, that is, unequal error variances across the predicted values, from obscuring statistically significant relationships. Without robust standard errors, the predictor variables approach statistical significance (results not shown), but an inverse funnel shape emerges when the Pearson residuals (i.e., raw residuals divided by the square root of the variance) are plotted against the predicted number of Black false alarms (see Figure 2). Pearson residuals are used to correct for the negative binomial model’s intrinsic tendency to yield heteroscedasticity with increasing counts allowing for increasing variance. Here however, the observed heteroscedasticity is in the reverse direction with increasing counts yielding less error variance. To prevent heteroscedasticity from obscuring the results, Table 3 presents the results model coefficients with robust standard errors.
The coefficients are interpretable as the difference in the logs of expected Black false alarms (erroneous shootings of unarmed Black targets) for each unit increase in the predictor variable, while holding all other variables constant. For example, with each millisecond increase in the difference between mismatching and matching traits in the reaction time task measuring implicit Black identification, the log count of expected Black false alarms decreases by -.002. When multiplying the range of implicit Black identification scores with the coefficient this yields a reduction in log shooting errors of 1.102 going from the most distant to the closest Black identifying officer. In other words, consistent with hypothesis H1, Black police officers who more strongly identify with Blacks as a group on the implicit level are significantly less likely (p=.005) to erroneously shoot unarmed Black civilians in the computer simulation.
Similarly, consistent with hypothesis H2, greater disillusionment with inequality in America among the Black police officers leads to a significant reduction in shootings of unarmed Black civilians (p=.011). Multiplying the range of disillusionment with the coefficient, this yields a reduction in log shooting errors of 1.155 going from the least to the most disillusioned officer.
In contrast, consistent with hypothesis H3, greater levels of internalized racism (endorsement of the anti-Black racist stereotype of Blacks being supposedly less intelligent than Whites) is associated with a significantly greater likelihood of shooting an unarmed Black civilian (p=.009). Multiplying the range with the coefficient, this results in an increase in log shooting errors by 1.360 going from the Black officers with the weakest to the strongest anti-Black racist stereotypes. Lastly, consistent with hypothesis H4, older Black police officers are more likely to erroneously shoot unarmed Black targets than younger Black officers (p=.030). Multiplying the range with the coefficient this suggests 1.330 more log erroneous shootings of Black civilian targets going from the youngest to the oldest Black officer.
The Omnibus Test of the model yields a Likelihood ratio χ2 with 4 degrees of freedom (df) of χ2 = 11.252, which is significant at p=.024. It tests the improvement of the model compared to an intercept (constant) only model without predictor variables. The model represents a significant improvement over the intercept only model suggesting that our predictor variables make a significant contribution.
SUMMARY AND CONCLUSION
In this study, we applied Du Bois’ (1903, p. 45) concept of ‘double consciousness’ to Black police officers. We demonstrate that their actions concerning unarmed Black civilians are influenced by their implicit identification with Blacks as a group and their disillusionment with enduring racial inequality in the United States. However, these same Black officers were found to make shooting errors influenced by internalized racism, viewing Black civilians through demeaning racist stereotypes (e.g., assuming lower intelligence among Blacks compared to Whites), and by their socialization, that is their exposure to anti-Black police subculture specifically or American racial culture more generally, which officer age might roughly represent. Our negative binomial regression results, with robust standard errors, support this ‘double consciousness’ interpretation. Black officers are not a homogeneous group; their actions can vary from care to recklessness, or a mix thereof, towards unarmed Black civilians based on prevailing thoughts at the time of a split-second decision to shoot or not to shoot.
If our study’s results are replicated with larger and more representative samples of Black police officers, our findings could have implications for departmental policies to change the culture of police departments and for improving training programs and monitoring progress of cultural-societal efforts at attitude change. To prevent incidents like the one that led to the death of Tyre Nichols, training programs and individual interventions to reduce implicit anti-Black bias may be considered. However, recent research from prominent implicit attitude researchers, including the inventors of the Implicit Association Test (IAT) Greenwald et al. (2022) with Dasgupta, and subliminal racial priming pioneer Fazio, suggests that individual-level strategies to change implicit associations may show little promise as long as the wider culture remains anti-Black. Even diversity training programs in private U.S. companies, according to a large study by Kalev et al. (2006), have shown little effectiveness and sometimes even counter-productive effects (Kalev et al., 2006). This suggests that improved training programs require rigorous testing for effectiveness.
Our findings further suggest that explicit attitude assessments and implicit association measures could be valuable tools in the selection of police officers for community assignments. Officers who strongly identify with Blacks as a group on the implicit level could be prioritized for assignment to communities of color, as they may be more cautious to avoid using lethal force against unarmed Black civilians.
Conversely, high scores on measures of internalized racism might indicate the need for further training for both White and Black police officers. More research is needed to understand the age variable better. While we interpreted it as a proxy for socialization, that is either exposure to ‘Blue’ police subculture specifically, or exposure to anti-Black American racial culture more generally, alternative interpretations, such as generational effects or sensory-motor coordination, may also be worth exploring. Regardless, it is essential to recognize that viewing Black police officers as either solely Black or solely ‘Blue’ is an oversimplification. Our investigation reveals that officers can exhibit both Black and ‘Blue’ attitudes, influencing their behavior towards unarmed Black (and possibly also White) civilians.
In the end, the issue of police shootings involves a combination of factors, including training, use-of-force policies, implicit biases, community relations, and systemic issues within law enforcement agencies. While strengthening Black racial identity may play a role in improving community relations, it is not a singular solution to prevent police shootings.
Reducing police shootings requires comprehensive approaches that address broader systemic issues, promote de-escalation techniques, emphasize community-oriented policing, and provide officers with effective training to minimize the use of lethal force when interacting with civilians.
Acknowledgements
We would like to express our appreciation to Periloux Peay for comments made on earlier drafts.
We capitalize group names to distinguish them from color adjectives (e.g., Black Americans, ‘Blue’ police subculture, or White Americans).