The Scope of Racial Bias in Policing: Behavioral Science’s Role in a Systemic Problem

Fatal Shootings and Violence by Police

In the United States, police officers are more likely to use lethal force by firearms against Black people than against White people. An analysis of officer-involved shooting data from 2010 to 2017 indicates that at least 30 percent of Black people shot by the police would not have been shot had they been White (Clark et al. 2023). The rate of fatal shootings of Black people by the police is about two and a half times the rate of fatal shootings of White people (Statista 2024). One in 1,000 Black men and boys will be killed by police over the life course (Edwards et al. 2019). Evidence of these disparities persists when researchers control for a range of relevant factors other than the race of the victim (Goff et al. 2016; Ross 2015; Ross et al. 2021). Among victims of fatal police shootings, Black, compared to White, victims are also less likely to have been armed (Charbonneau et al. 2017; Nix et al. 2017).

Racial disparities in police use of force against unarmed individuals at a nonlethal level also occur (Goff et al. 2016). About a third of Black men report that they have been the victim of police violence (Pickett et al. 2022). A US survey of Black people found that their level of fear about being the target of police violence is five times the level among White people (Graham et al. 2020).

Use of Nonlethal Force

With rare exception, studies investigating racial disparities reveal that police use of force, nonlethal as well as lethal, is significantly more likely in encounters with Black people (and especially Black men) than with White people (Goff et al. 2016; Hollis and Jennings 2018). Police officers are generally more lenient and less likely to use force with White than with Black people (Kovera 2019), even when controlling for the severity of the suspected illegal behavior. An analysis of body-worn camera footage reveals that Black people are 63 percent more likely to experience force in a police encounter than are White people (Willits and Makin 2018). In addition, as both the analysis of the camera footage and a separate study of case files (Kahn et al. 2017) reveal, Black people are more likely to experience police use of force sooner in the encounter than are White people.

Overall, the effects of the target’s race on police use of force typically occur over and above the influence of a range of other factors, including more reported resistance to police officers (Crow and Adrion 2011; Terrill et al. 2008; compare with Paoline et al. 2018) and are observed across the country. Latine people are also targets of police use of force at higher rates than non-Latine White people (Hoekstra and Sloan 2022), particularly with respect to high levels of use of force (such as pointing or using a firearm or taser, or canine; Smith et al. 2023).

The likelihood of police use of force relates not only to the race of the target but also to the neighborhood and cities in which the police encounter occurs. Police officers are more likely to use force in neighborhoods in which non-White people represent a greater percentage of residents (Lee 2016). These effects overlap with the higher use-of-force rates observed in cases of low socioeconomic resources and relatively higher rates of violent crime. In addition, police officers are more likely to engage in use of force in cities with larger percentages of Black and Latine populations, especially when the cities are more residentially segregated (Smith and Holmes 2014).

“Stop and Frisk” Pedestrian Stops and Vehicle Stops

Constitutionally, police officers in the United States can only detain a person if they have reason to believe the person has or is about to commit a crime. However, all fifty states have had some version of stop and frisk laws that permit a police officer to stop and frisk a person without stating a reason. For many of the laws, all that is needed is “reasonable suspicion,” which can mean the person appears suspicious to a police officer.

Stop and frisk laws, as originally proposed, are typically portrayed and frequently seen as a race-neutral way to reduce crime. However, the racial disparities in the actual enforcement of these laws are glaring. Between 2004 and 2012 in New York City, 4.4 million people were detained under this law. Although the Black and Latine residents of New York City represented slightly over half the city’s population, they constituted 83 percent of the people stopped (New York Times 2013). Moreover, the likelihood that an investigatory stop by a police officer would escalate to use of force was significantly greater for Black and Latine targets than for White targets (Levchak 2017). Because of successful legal challenges to the stop and frisk policy in New York City, the number of these actions by the police declined dramatically. By 2017, stop and frisk incidents were only a fraction of what they had been earlier, but 90 percent of the people stopped were still Black residents.

With respect to traffic stops, Black and Latine drivers are much more likely to be stopped than are White or Asian drivers (Baumgartner et al. 2017; Lim et al. 2024). The disparity in traffic stops between Black and White drivers was greater in areas in which White residents were more racially biased (Stelter et al. 2022). In addition to where a traffic stop occurs, when a stop occurs is a significant factor relating to this disparity. There is a large racial disparity in traffic stops during the daytime when a driver’s race can be easily identified (Pierson et al. 2020); this difference is often reduced at night when it is more difficult to identify the race of a driver (Racial and Identity Profiling Advisory Board 2021). Black drivers are about five times more likely than White drivers (44 percent versus 9 percent) to report that they have been unfairly stopped by the police, and the percentage is almost twice as high for Black men than for Black women (59 percent versus 31 percent; Desilver et al. 2020). Black and Latine drivers are also twice as likely to be searched following a vehicle stop (Pierson et al. 2020). This disparity persists even though police are more likely to find drugs on White than Black targets. Nationwide, while Black drivers are stopped by police and, after they are stopped, searched at a disproportionately higher rate than White drivers, the searches of the vehicles of Black drivers are less likely to discover contraband than those of White drivers (that is, they have a higher “false-alarm rate”; Meyer and Gonzalez 2024).

Impact of Fatal Shootings and Police Violence

Exposure to police violence through the media can have extended negative consequences, particularly for Black people. A USA Today/Ipsos poll (2021) found that in the month after George Floyd’s death, trust in law enforcement declined significantly among US residents in general, to a low of 56 percent. One year later, 69 percent—a significant increase—said they trusted law enforcement. However, the gap between White and Black people was particularly large, 77 percent versus 42 percent. Also, whereas in the month after George Floyd’s death 60 percent of US residents reported that they personally believed that George Floyd was murdered, by the time Derek Chauvin (the Minneapolis police officer who appeared in the viral video of George Floyd’s death) was preparing to stand for trial, that percentage significantly declined to 36 percent. At that later time, the racial divide was substantial: only 28 percent of White people agreed at that time that George Floyd was murdered, compared to 64 percent of Black people. Derek Chauvin was subsequently found guilty of murder.

Police incidents of violence against Black people have an enduring impact on Black communities (Haile et al. 2023). For example, the levels of anxiety and depression among Black people increased substantially following George Floyd’s murder (Fowers and Wan 2020). One study looked at “poor mental health days” among Black people over five years and found that these days were much more likely to occur during weeks when such acts of racial violence received considerable attention in the media. These incidents did not have similar effects on the mental health of White people (Curtis et al. 2021). With respect to physical health, community residents who perceive a greater use of force in their area by police experience higher levels of stress, and this relationship is particularly pronounced for Black residents (Stansfield 2022). Additionally, greater exposure to police killings of Black people adversely affects the cardiovascular health of Black people, increasing the likelihood of stroke for Black men and hypertension for Black women (Talbert 2023).

In addition, exposure to cases of police violence against unarmed Black men was found to suppress citizen crime reporting among citizens of Cincinnati generally, but particularly among those living in Black and Latine neighborhoods (Pearson and Timberlake 2023), and exposure to police use of fatal force generally suppressed people’s willingness to place 911 calls for assistance in Los Angeles, but that impact was greater in Black and even greater in Latine neighborhoods; the effect was significant in White neighborhoods as well, but the impact occurred to a lesser degree (Sheppard and Stowell 2022).

Impact of Pedestrian and Vehicle and Stops on Community Health and Action

Common encounters between police officers and community members can also have significant negative effects on the health and actions of the residents of areas in which police officers are “protecting and serving.” For instance, adolescents who were personally stopped by police officers (on the street, in a car, at school) or who observed another person stopped by police subsequently experienced more depressive symptoms, and this effect was particularly pronounced for Black adolescents (Turney 2021; see also Burke et al. 2023).

The character of police contacts is a critical element of how people respond to police officers. While the number of stops in an area does not directly relate to negative health among residents, the incidence of stops culminating in invasive action by police (for example, frisking) does relate to poorer health (for example, ratings of overall health and specific conditions such as high blood pressure; Sewell and Jefferson 2016). These effects, particularly those associated with high levels of stress and psychological distress, occur not only for individuals who have personal contact with police but also for those who reside in highly and inequitably policed areas. The negative effects are particularly strong for Black and Latine individuals. In addition, residents living in minority communities with higher incidences of use of force against pedestrians have a higher risk of diabetes and obesity (Sewell 2017).

The greater stress aroused by encounters with law enforcement, such as police stops and arrests, has an additional consequence: adolescents who experience police contact show an increased likelihood of subsequently engaging in delinquent behavior. In a four-wave longitudinal study, contact with law enforcement in the form of pedestrian stops by police predicted increases in Black and Latine adolescents self-reported criminal behaviors across an eighteen-month period, in part because of the psychological distress (for example, anxiety and depression) these adolescents experienced (del Toro et al. 2019).

In terms of the impact on the community, while a greater number of stop and frisk encounters in disadvantaged communities was generally positively related to more calls for service (311 calls) in those neighborhoods, more stop and frisk encounters that involved searches or the use of force, particularly when they did not result in an arrest, reduced the likelihood that community members would seek these services (Lerman and Weaver 2020). In general, involuntary encounters with police reduce willingness to engage with or seek support from police or local government (Brayne 2014). In areas around Philadelphia, greater incidence of pedestrian stops in which police frisked residents predicted lower use of emergency health services (Kerrison and Sewell 2020). As Vesla Weaver, Gwen Prowse, and Spencer Piston (2020, 3) explain, “This is because people tend to engage the state only when they have a basic trust that it will not dominate them, humiliate them, or physically assault them.”

Furthermore, this response is not necessarily one of passive withdrawal; it also often involves “collective autonomy,” in which communities distance themselves strategically from public institutions (for example, police and government) while building community resources to promote racial collective consciousness and self-determination. An analysis by Weaver and colleagues (2020) of conversations between Black residents from highly policed areas in five US cities reveals consistent themes of perceptions of police oppression and consequent efforts to withdraw from engagement with police and government while developing community cohesiveness, autonomy, and power.

In summary, there is robust evidence of racial and ethnic disparities in policing in the United States. This disparate treatment of Black residents relative to White residents has significant lasting impact on Black people, both individuals who experience direct police contact and those who learn about incidents involving others. Beyond the impact of these events in terms of psychological distress and its consequences for mental and physical health, these occurrences undermine trust in the police, which reduces the likelihood that residents of affected communities will seek police assistance or use other government services, such as emergency medical services.

As noted earlier, racial disparities in policing-related outcomes do not automatically imply racial discrimination. In subsequent sections, we explore the potential role of racial discrimination in policing and how it contributes to racial disparities. The challenge of distinguishing the role of discrimination relative to other potential influences in these disparities is a complex task because of methodological challenges in determining causality, the fact that discriminatory influences operate in different sociopolitical levels, and the very different nature of the activities in which disparities are observed. Bias has many facets, and the circumstances and motivations involved in different police activities (such as use of force and traffic stops) vary widely. The actions of police officers are also shaped by influences at multiple levels of analysis including, for example, officer biases, situational officer discretion, and organizational or institutional guidance (Camp 2024). While acknowledging these complexities, we attempt to triangulate from the diverse evidence to better understand the potential role of racism in racial disparities in policing.

Individual Differences

There is evidence of stable individual differences across time in officers’ orientations in policing. For instance, one robust predictor of police officers’ attitudes toward use of force on the job is their attitudes two years earlier, before they entered the police academy (Oberfield 2012). Building on the idea that police officers differ systematically in the ways they approach policing, one line of research on racial disparities in policing that involves discrimination by police officers has focused on the role of “bad apples,” or a limited number of officers whose bias skews the statistics for unjustified disparate outcomes. Indeed, there is evidence of consistent individual differences among officers in racially biased treatment of residents and others in the community (Terrill and Ingram 2016). Felipe Goncalves and Steven Mello (2021) show that the racial disparity in speeding ticket outcomes among Florida Highway Patrol officers can be accounted for by the actions of less than half (about 40 percent) of patrol officers. Allegations by Chicago residents of police misconduct were significant predictors of future allegations against those same officers (Rozema and Schanzenbach 2019), providing additional evidence of stable individual tendencies. However, a simulation found that replacing the top 10 percent of Chicago officers identified as at risk for engaging in misconduct in their duties with ones not at such risk would reduce the overall number of use-of-force incidents by only 4–6 percent across a ten-year period (Chalfin and Kaplan 2021).

Research identifies a specific individual-difference characteristic that has important consequences for policing with respect to discriminatory treatment of members of particular racial and ethnic groups. Policing, as it has been traditionally conceived in the United States, functions to enforce the existing hierarchical social order. Law enforcement as a profession thus appeals to individuals who value this orientation: policing attracts people who have characteristics (for example, in ideology or personality) that relate to support for maintaining the current social order. Social Dominance Orientation (SDO) is an individual-difference measure that reflects the extent to which a person prefers hierarchical intergroup relations over equality and desires their ingroup to dominate outgroups (Ho et al. 2015; Sidanius et al. 2017). Although the SDO scale does not identify specific other groups in its items (for example, “Some groups are just more worthy than other groups”), it predicts biases toward multiple marginalized groups, including Black people. People who score higher in SDO are generally more prejudiced in their attitudes, hold more negative stereotypes of other groups, and behave in more discriminatory ways toward traditionally marginalized social groups (Duckitt and Sibley 2017).

SDO is particularly relevant to policing. White police officers score higher on SDO compared with members of the general public, college students, and public defenders, even controlling for a range of demographic characteristics (Sidanius et al. 1994; see also data provided by Xu et al. 2014). This orientation is further supported by officer experiences while serving. Junior police officers’ perceptions of a high level of SDO among peer officers reinforces their own orientation (Gatto and Dambrun 2012). Also, specific prejudices toward a range of groups increase with training and experience in policing (Gatto et al. 2010). With respect to policing outcomes, Jillian Swencionis, Enrique Pouget, and Phillip Atiba Goff (2021) demonstrate across three different cities that White officers (but not Black officers) with higher levels of SDO have a greater number of use-of-force incidents during “Terry” stops (forcible stops made when an officer suspects a crime). This set of studies, however, does not include data to test whether the use of force is more frequent with Black than with White targets. Overall, the findings about the role of individual differences indicate that “bad apple” police officers do exist and that their discriminatory actions can contribute disproportionately to racial disparities in policing. However, this factor accounts for only a limited amount of variance in racial disparities in policing.

Prejudice and Stereotypes

There is converging evidence of the role of racial prejudice (an attitude or unfair preference) and stereotyping (overgeneralized and frequently inaccurate beliefs) in policing across a range of research paradigms. Analysis of policing statistics implicates the influences of racial prejudice and stereotyping on a range of police activities and outcomes. Economists Brendan O’Flaherty and Rajiv Sethi (2019, 2024) examine the extent to which racial disparities in police stops and searches and police use of deadly force reflect statistically accurate beliefs that justify these disparities (called statistical or strategic discrimination) or represent unjustified, biased attitudes or beliefs based on race (termed preference-, prejudice-, or belief-based discrimination). This distinction is a critical one in the field of economics, determining the sources of racial disparities in policing (Agan 2024). These researchers conclude that while statistical discrimination does contribute to racial disparities in stops and searches, prejudiced discrimination also exerts a significant effect (for example, Goel et al. 2016; Pierson et al. 2020). In making such determinations of disparities as due to unfair discrimination, appropriate benchmarks are critical. Hit rate analysis, which is a widely accepted benchmark (Agan 2024), compares incidences of stops to the finding of contraband in the searches that follow. Unfair discrimination when the frequency of stops is based on race (for example, Black motorists versus White motorists) significantly exceeds the relative hit rates (for example, finding contraband in searches).

O’Flaherty and Sethi (2024) report that compared with their representation in the population, Black people were significantly overrepresented in fatal police shootings from 2015 to 2022. However, the researchers note that incidences of fatal shootings of Black people by police officers were comparable to the incidence of Black people arrested during that period. While this analysis suggests that racial disparities in fatal shootings may primarily reflect statistical discrimination, O’Flaherty and Sethi further observe that using arrestees as the baseline may obscure the impact of prejudiced discrimination to the extent that arrest rates are themselves a product, at least to some degree, of prejudiced discrimination. The researchers also note that stereotypes Black people hold of police officers may also contribute to their encounters with police, which can escalate into use-of-force incidents and potentially to fatal shootings (see also Agan 2024).

Psychological research has focused on understanding relevant aspects of racial prejudice and stereotypes and how those biases can influence police behavior. That is, this type of research has a strong emphasis on illuminating intrapersonal and interpersonal mechanisms. Goff and colleagues (2008) focus their research on a critical dehumanizing element of stereotypes of Black people: the tendency among White participants in their experiments to automatically associate Black people with gorillas or other kinds of apes. In a series of studies conducted with general samples of participants, these researchers further illuminate the implications of this bias within criminal justice contexts. Specifically, the research demonstrates how this association of Black people with apes alters visual perception and attention and increases endorsement of violence against Black targets. One of their studies, involving analysis of criminal cases, reveals that news articles written about Black people who are convicted of capital crimes are more likely to contain ape-relevant language than news articles written about White convicts. In addition, convicts who are implicitly portrayed as more apelike in these news articles are more likely to be executed by the state than those who are not. These researchers directly demonstrate the causal impact of this form of dehumanization in the domain of policing. They subliminally primed White male participants with words related to ape or big cat and then had participants view a video of a Black or White target being beaten by police officers. Participants exposed to the ape-related words were more likely to believe the police officers’ violent actions were justified when the victim of the beating was a Black person but not when the victim was a White person.

While this set of studies did not examine the stereotypes of police officers directly, the researchers report anecdotal evidence. In Los Angeles, police officers referred to the beating of Black targets as “monkey slapping time” and described a Black family they had to visit as “gorillas in the mist.” It was also common among Los Angeles police to report incidents in which a Black person died as “NHI”—no humans involved (Goff et al. 2008). More recently, an analysis of text messages sent between police officers in Berkeley, California, in 2019 and 2020 finds that these officers “engage in vile dehumanization of (Black people) and express views that are openly racist” (Savidge et al. 2022).

Another core aspect of automatically activated (implicit) racial stereotypes directly relevant to policing is the prevalent stereotypic association of Black people with criminality (Pickett et al. 2014). For instance, one set of experiments (Eberhardt et al. 2004) demonstrates that participants more quickly identify obscured photos of crime-related objects (such as guns) after being subliminally primed with Black than with White faces. The speed with which the crime-related objects are recognized is unrelated to participants’ level of explicit racial prejudice, suggesting the distinct nature of implicit and explicit biases. Another prominent research paradigm examines race-related bias toward identifying guns (compared with non-gun objects) following brief exposure to Black faces compared with White faces (see Payne and Correll 2020 for a review). Research using this paradigm has consistently shown that participants are more likely to mistakenly identify objects as guns when the objects are paired with images of Black men’s faces than when they are paired with images of White men’s faces. This effect does not occur at the earliest stages of perceptual recognition of the objects but develops at later stages in ways that suggest racial stereotypes shape how people construe and react to the objects (Stein et al. 2023).

Researchers studying discrimination in policing have often focused on simulations of decisions to shoot armed or unarmed targets, most commonly male targets. In one of the most widely used paradigms, participants engage in a “shoot/don’t shoot” computer simulation in which they observe images of White and Black men in everyday places like parks or city sidewalks (Payne and Correll 2020). Some of the men are armed with guns, while others are unarmed and carrying mundane objects like wallets or cell phones. Participants are instructed to press a button to “shoot” if the man is holding a gun or another button to “don’t shoot” if the man is not holding a gun. In this simulation, participants unaffiliated with law enforcement tend to mistakenly shoot unarmed Black men at higher rates than unarmed White men. These participants also tend to react faster to unarmed Black men than unarmed White men (see Mekawi and Bresin 2015 for a review; compare with James et al. 2016).

Follow-up research on the intersection between race and gender has shown that Black women are not more likely to be mistakenly shot than White women or White men in these simulations, suggesting that Black men are uniquely associated with violence and danger (Plant et al. 2011). The racial bias does not relate to an explicit measure of racial prejudice (Correll et al. 2002, Study 3), but it is greater among participants sampled generally (Correll et al. 2006) and police officers (Peruche and Plant 2006) who more strongly endorse the stereotype of Black people as “dangerous” or as criminals. In studies employing both White and Black participants unaffiliated with law enforcement (Correll et al. 2002, Study 4) and police officers (Plant et al. 2005), the level of shooter bias against Black targets is comparable across participant race, suggesting that shooter bias is perpetuated by knowledge of cultural stereotypes rather than by an explicit propensity to discriminate against Black people. Experimental interventions that weaken the stereotypic association of race and criminality, in turn, reduced racially discriminatory responses on a shoot/don’t shoot task (Plant et al. 2005).

Compared to the results of studies using participants sampled from general populations, results of research using police officer participants in decisions to shoot in this simulation paradigm are more variable. Some research has shown that, at least under some circumstances, police officers are more hesitant to shoot Black targets than White targets on a shoot/don’t shoot task (James et al. 2017). Also, police officers’ implicit bias, as assessed using the IAT, does not reliably predict racial biases in decisions to shoot in this task (Andersen et al. 2023). However, racial biases in these types of simulated encounters are well documented and depend on the life experiences of the police officers. On average, officers with policing experience do not tend to show a racial bias in decisions to shoot (Correll et al. 2007). However, officers who have less childhood contact with Black people or are assigned to gang and street-crime units that regularly deal with minority gang members tend to show higher levels of racial shooting bias (Sadler et al. 2012; Sim et al. 2013). New police recruits also tend to show evidence of a racial shooting bias, suggesting that certain on-the-job experiences may be important for mitigating racial shooting biases (Ma et al. 2013).

Because the likelihood of activating a stereotype of a group is greater when people are exposed to a more typical member of a group, implicit stereotyping is stronger when people are exposed to a Black individual who is more visually prototypical of their race. Of direct relevance to criminal justice, an analysis by Jennifer Eberhardt and colleagues (2006) demonstrates that when Black defendants found guilty of a capital crime involving a White victim are perceived to have a more stereotypically Black appearance, they are more likely to be sentenced to death (see Kleider-Offutt et al. 2017 for a review). Laboratory research employing the shoot/don’t shoot simulation paradigm (Kahn and Davies 2011) demonstrates that both White and Black participants who are not employed in law enforcement exhibit a greater propensity to shoot more phenotypically stereotypic Black targets (those with darker skin, broader noses, and fuller lips), resulting in a more pronounced racial shooting bias. Phenotypic stereotypicality of White targets also matters. A further analysis of a random sample of booking photos reveals that, controlling for type of arrest, reported level of resistance, and the presence of drugs and alcohol in a target’s system, White targets who are perceived to be more phenotypically White are less likely to be the targets of police use of force and are treated with less force by police (Kahn et al. 2016). However, this study does not find a significant relationship between the phenotypic stereotypicality of Black targets and police use of force.

Threat

Threat is highly salient in police work (Jamieson et al. 2000) and takes many forms. It often involves a general concern about personal safety, which leads police officers to be highly suspicious and preoccupied with maintaining an edge in police-resident exchanges, particularly in those involving Black targets. For instance, Joscha Legewie (2016) finds that police use of force during stops of Black (but not White) pedestrians significantly increased during the two weeks immediately after two officers were killed by Black targets.

Police may thus develop an “us versus them” orientation toward residents. In general, an orientation of this type makes people concerned about the motives of outgroup members, wary of their actions, and more defensive in interactions with them. In policing, it can often lead to an animosity toward community members generally, but this antipathy may be particularly pronounced toward Black people. Exposure to Black people spontaneously arouses threat, even among Black participants who do not harbor prejudice—either explicit or implicit—against Black people (March 2023).

This threat may be generated through elements of the cultural racial stereotype of Black people. For instance, White people, particularly those higher in implicit racial bias, tend to perceive even neutral facial expressions of Black persons as expressing anger (Hugenberg and Bodenhausen 2003). Other research, using a representative sample of White participants, shows that priming people with the race of a seventeen-year-old male juvenile offender who committed a violent assault affects the level of punishment that they recommend: the juvenile offender is viewed as more blameworthy and more deserving of a sentence of life in prison when he is Black than when he is White (Rattan et al. 2012). These effects are likely attributable to a greater cultural association of Black people with violence. Consistent with this reasoning, when individuals are primed with words associated with Black people, they are more likely to perceive another person’s actions as hostile, even where the person’s race is not known (Devine 1989). Additional research that extends this work to policing shows that when police officers and juvenile probation officers are subliminally exposed to words related to the Black cultural stereotype (compared to words that were unrelated to racial stereotypes), the officers view a juvenile offender who either shoplifted or assaulted a peer as more hostile and culpable, even though the race of the offender is not directly identified (Graham and Lowery 2004).

The activation of racial stereotypes can additionally affect perceptions of Black juveniles in a particular way that arouses threat, also with direct relevance to policing. Goff and colleagues (2014) conducted a series of studies in which both undergraduate students and police officers judged the age and culpability of youths between ten and seventeen years of age in a criminal justice context. While both undergraduates and police officers estimated the ages of White youths accurately, they perceived Black individuals suspected of committing a misdemeanor as over two years older than they were. Black individuals suspected of committing a felony were seen as more than four years older. Perceptions of the culpability of the youths for the crime showed a similar pattern. These findings suggest that the force typically used by police for adults may be viewed as appropriate for Black youths. Other research reveals that people who see young Black men as threatening are more likely to also perceive their bodies as taller and heavier than comparable young White men of the same height and weight (Wilson et al. 2017). Overall, these results pertaining to race and threat may partially explain why police are more likely to use force with Black than with White targets.

Another form of threat—stereotype (or identity) threat—also contributes to racial bias in policing (Pryor et al. 2020; also see Swencionis and Goff 2017 and Burke 2023). Stereotype threat occurs when a person is in a situation that arouses concern about being judged in accordance with unfavorable stereotypes associated with a social identity that is important and relevant to the context (Steele 2011). Marie Pryor and colleagues (2020) explain how a White officer may experience stereotype threat with respect to being judged in accordance with a stereotype that White people are racist. Also, a police officer of any race or gender may be concerned that members of the community will stereotype police officers as racist. Research conducted in the southwestern United States finds that both officers and participants unaffiliated with law enforcement feel less positively toward law enforcement when faced with the prospect of enforcing what they feel would be seen as racist, and officers fear for their safety as a result (Epstein and Goff 2011).

In encounters with police, citizens also experience threat (Olivett and March 2023). Because of this experience of threat, they display both greater physiological reactivity and defensive behavioral responses (such as a defensive freeze reaction) in encounters with police officers than with other people. Black people, because of the high level of police surveillance they experience, the threat of physical and psychological abuse by police officers, and the quantity of negative police encounters they have had, are especially apprehensive and defensive (Algrim et al. 2022; see also Henning 2021).

Moreover, like police officers in their interactions with the community, community members experience stereotype threat in these encounters. One study shows that Black people, but not White people, report concern that police officers stereotype them as criminals because of their race (Najdowski et al. 2015). Black men but not White men report they would experience stereotype threat in encounters with police and anticipated anxiety, self-regulatory efforts, and behaviors that are commonly viewed as suspicious by police officers.

Contextual Moderators

Situational factors that heighten the risk of discriminatory police behavior include cognitive demand, salience of crime, officer inexperience, and officer discretion (Swencionis and Goff 2017; Pryor et al. 2020; and Glaser 2024). These risk factors for discriminatory policing reflect the hypothesized influences of implicit prejudice and stereotyping on behavior that have been identified in basic psychological research.

The literature on cognitive demand and racial bias reveals that when the demands of a situation exceed the processing resources available, the resources that are typically devoted to controlling the expression of racial bias (which tend to be quite effortful) are otherwise occupied and fail to suppress racial biases that are automatically activated (such as implicit biases). Faced with cognitive overload, most people, including police officers, will resort to cognitive shortcuts, such as using the racial stereotypes that associate Black people with crime (Swencionis and Goff 2017).

Police officers have multiple, simultaneous demands on their attention and cognitive processing, and “they have to make life-altering decisions under severe time constraints, often under stressful or dangerous conditions” (Pryor et al. 2020, 350). Moreover, interracial contact can itself act as an additional stressor, especially for individuals, particularly those who score high on measures of implicit bias (Trawalter et al. 2009). This stress can have detrimental effects on subsequent performance (Richeson and Shelton 2003): following interracial interactions, White participants—particularly those who scored higher in implicit racial bias—display impaired performance on subsequent tasks requiring cognitive effort to inhibit inappropriate responses.

Aligning with basic research on prejudice and stereotypes, situational influences systematically moderate the impact of implicit biases on actions relevant to policing. For instance, conditions that interfere with conscious control of behavior both increase the level of police officers’ level of implicit bias and exacerbate the impact of implicit racial biases on behavior. Police officers show higher implicit prejudice, as assessed by the IAT, following nights in which they have less sleep (James 2018), and officers who feel tired (rather than alert) exhibit stronger racial bias in decisions to shoot (Ma et al. 2013).

Time pressure similarly places demands on people’s cognitive resources, increasing the likelihood that racial biases, particularly implicit ones, will be expressed and critically shape behavior. As discussed earlier, White participants typically identify guns faster when they accompany Black (as opposed to White) men’s faces, but time pressure intensifies these effects. When given a short response deadline, participants are significantly more likely to mistake a tool for a gun when it is paired with a Black man’s face. In other words, when participants are faced with a split-second decision, race influences their tendency to perceive a gun that is not actually there (Payne and Correll 2020). The racial biases observed in the shoot/don’t shoot paradigm are even greater when participants must make their decision more quickly (Correll et al. 2002).

As discussed earlier, cultural stereotypes associate Black people, and particularly Black men, with criminal activity. A criminal focus, in turn, exacerbates racial biases and their impact. Crime is chronically salient in much of police work and, in addition to concerns about personal safety, other aspects of policing emphasize criminality as a core issue in policing. Police performance is often assessed in terms of its effects on crime rates, which tends to incentivize a focus on crime to the exclusion of other aspects of policing, such as protecting the public safety, providing service to the public, having positive interactions with community members, or being accurate (Swencionis and Goff 2017). This focus on crime control represents a risk factor for discrimination because it heightens the accessibility of pervasive stereotypes associating Black people (in general) and Black men (in particular) with crime-related concepts (Eberhardt et al. 2004). Chronic and transitory factors that make crime salient are likely to increase racially disparate behavior by police officers that harms Black people. For example, when police officers are placed in situations where crime is emphasized, they attend more quickly to faces of Black men than of White men and are more likely to falsely identify Black men in lineups (Eberhardt et al. 2004).

The context of gangs further illustrates this point. Between 11 percent and 27 percent of self-identified gang members are estimated to be White, and White gang members’ level of criminal activity is equal to those of their non-White counterparts. However, there is a strong stereotypic association of gang membership with Latine and especially Black youths and adolescents. One likely consequence of this stereotype is that police officers will attend more to Black and Latine people in their activities relevant to gangs. For example, in Chicago, police officers recorded more than 128,000 adults and 33,000 children and youths in a database of gang members and “gang-affiliated” persons (Sweeney and Fry 2018). Non-White individuals were vastly overrepresented in these databases, with over 95 percent of the entries for adults identified as Black or Latine. These racial and ethnic disparities in the “gang” listings, which likely were rooted in racial biases, had disproportionate negative consequences for Black and Latine people: the database was used for criminal investigations, background checks, immigration enforcement, and criminal sentencing (Dumke 2018).

Inexperience is associated with uncertainty and has a more limited set of policing skills that may normally be acquired over time. Perhaps as a consequence, more inexperienced police officers are generally more inclined to use force in their policing. Younger officers and officers with fewer than five years’ experience use more force and are 2.5 times more likely to be investigated for use of force (McElvain and Kposowa 2004; Terrill and Mastrofski 2002). Officers who are twenty-five years old or younger are over three times more likely to be investigated for use of force compared with officers over the age of forty (McElvain and Kposowa 2004).

While the greater tendency to use force, often with insufficient justification, in itself may increase the likelihood of racial disparities in policing, racial stereotypes further increase the likelihood through discriminatory actions. Stereotypes influence behavior more strongly under conditions of uncertainty. Thus, officers who are new to police service or who receive a new assignment are likely to be particularly susceptible to the racialized errors that can arise from uncertainties and consequent reliance on limited cues—such as community members’ race, gender, or presence in a high-crime neighborhood—to determine whether to suspect a person of breaking the law (Willis et al. 2010). Also, because of the role of seniority in police assignments, newer officers are more likely to be assigned to rotating or night shifts, which produce sleep disturbances that (as discussed earlier) independently increase the likelihood of racially discriminatory behavior.

The literature on skill acquisition and expertise concludes that individuals with more practice are more likely to perform tasks accurately and are less likely to engage in racial discrimination (see, for example, Kawakami et al. 2005; Plant and Peruche 2005). Thus, people with less training and experience are not only less successful in performing tasks but also more susceptible to racial biases affecting their responses. With respect to policing, practice can increase accuracy in weapon identification tasks and reduce racial biases in shoot/don’t shoot tasks. Simulations that more closely mirror real-world deadly force encounters, which provide more detailed information about the situation and the target, generally show less racial bias in decisions to shoot (Andersen et al. 2023; Cox et al. 2014; James et al. 2013). In general, likely because of their greater training in firearm use and procedures for use of force, police officers are still less likely than community members to make these racialized errors on shooting tasks (Correll et al. 2007). Also, in support of the proposed roles of experience, training, and expertise, community members who receive training and practice in the shooter task demonstrate improved accuracy and diminished behavioral effects of racial bias (Correll et al. 2007; Plant et al. 2005; Sim et al. 2013). Even among police officers, practice with the simulated shooting task reduces racialized errors against Black targets (Plant and Peruche 2005).

As basic research has shown, ambiguity is an important factor in the extent to which personal biases are manifested in discriminatory behavior. When decisions or actions are governed by clear-cut rules or strong norms, a person is more likely to behave in a racially unbiased manner. However, when rules are nonexistent, appropriate behavior is not well defined, or norms are ambiguous, White people, including White police officers, are more likely to act in ways that favor White persons and disadvantage non-White individuals. Ambiguity and personal discretion are particularly relevant factors in policing. In their day-to-day work, police officers must contend with high degrees of ambiguity (such as determining who fits a suspect description or who poses a potential threat) and have high discretion (for example, deciding whether to search during a traffic stop). Police officers necessarily exercise a great deal of discretion in the course of their work: they must continually make judgments about the meaning of others’ behavior and about how to weigh facts, circumstances, and evidence. In the absence of binding directives or established protocols, officers’ reliance on discretion or personal judgment can allow their prejudices to influence their decisions.

Furthermore, officers’ discretionary exercise of their powers to stop, search, and use force against community members is not meaningfully constrained by the courts. The discretion allowed under stop and frisk laws and in traffic stops is thus a primary reason Black and Latine people are disproportionately targeted by police in these activities (Desilver et al. 2020). Consistent with this proposition, policing records show evidence of biased discretion during traffic stops and police interactions in California. Officers choose to search individuals perceived to be Black at over twice the rates of individuals perceived to be White. However, the probability of finding contraband during the search (the discovery rate) is higher for individuals perceived to be White (Racial and Identity Profiling Advisory Board 2021), indicating evidence of racial discrimination. Although the introduction of body-worn cameras (as we will discuss in more detail) and the ubiquity of cell phone recordings of police encounters may expose officers’ behavior to increased public scrutiny, it is not clear that the availability of video recordings has systematically increased accountability for police misconduct.

Screening and Selection

One approach for reducing racial discrimination in policing is to reduce the likelihood that individuals who have biases enter the policing profession. Reflecting this direction, California Assembly Bill 846, which became effective in 2021, requires that candidates for a position as a “peace officer” be screened in the hiring processes in ways to help ensure that they are “free from any physical, emotional, or mental condition, including bias against race or ethnicity, gender, nationality, religion, disability, or sexual orientation, that might adversely affect the exercise of the powers of a peace officer.” These screening assessments could include measures of candidate racial or ethnic bias. As noted earlier, individuals higher in SDO, which relates to greater prejudice against a range of minoritized groups, are particularly attracted to policing as a profession; police officers, on average, tend to score high in SDO compared to the general public (Sidanius et al. 1994; Xu et al. 2014); and police officers higher in SDO are more likely to use force against targets (Swencionis et al. 2021). However, current measures of explicit prejudice and stereotyping are not well suited to screening for bias because individuals tend to adjust their responses to appear less biased than they are when that is socially desirable. Implicit measures of bias are less amenable to such strategic control and relate significantly to discriminatory behaviors on average (Kurdi et al. 2019), but they do not have sufficient predictive strength for any specific individual to justify use of such measures for employment screening (Greenwald, Dasgupta et al. 2022; Lai and Wilson 2021). An alternative approach is to screen for other individual differences that promote positive relations with members of minoritized groups or that tend to mitigate the potential effects of automatically activated (implicit) bias. These can include high levels of empathy and perspective taking, which attenuate the impact of implicit bias (Todd et al. 2011), or general personality traits associated with more positive behaviors toward people from diverse groups (for example, Agreeableness; see Crawford and Brandt 2019).

Background investigations in screening procedures can also be used for prior candidate activities that would be useful in predicting, in combination with other materials in the review process, the likelihood that an individual would perform their policing duties without bias. Engaging more frequently in positive contact and having close relationships with members of another racial or ethnic group are among the most robust predictors of less bias against minoritized groups, which is supportive of research on contact theory (Pettigrew and Tropp 2013). People who have had more positive intergroup contact with members of another racial or ethnic group or who have a friend who is a member of that group have lower levels of prejudice against the group and other members of the group, and they experience less threat when they encounter members of the group. Conversely, those who have more negative experiences with members of another group are more biased against the group and feel greater anxiety and threat (Paolini et al. 2021). These latter findings suggest a possible reason why, given the generally negative experience of interracial police-resident encounters, police officers with greater prior contact with Black people displayed greater racial bias against Black people in a shoot/don’t shoot task (Sadler et al. 2012).

Because little is known about what exact indicators most reliably predict discriminatory behavior by a police officer, further research is needed to maximize the promise of using selection criteria to reduce police bias. However, the largest and best apples-to-apples comparison of officers by race and gender recently revealed that White male officers engaged in far more use-of-force incidents in Chicago than female or non-White officers and that this behavior was driven disproportionately by White male officers’ use of force toward Black residents (Ba et al. 2021). In other words, there is recent, strong evidence that who is doing the policing matters in terms of equitable policing outcomes.

Antibias Training

While assessment for bias and background reviews can be helpful for identifying individuals at major risk of behaving in a discriminatory manner, limiting their entry into the profession, other interventions focus on those already serving as police officers. Antibias training is one such approach that has been commonly used across a range of professional domains. However, antibias training programs are likely to be substantially curtailed in many settings, including policing, because of actions by President Donald Trump in his first days in office in 2025. His Executive Order requires all agencies and departments in the federal government to “terminate, to the maximum extent allowed by law, all DEI, DEIA, and ‘environmental justice’ offices and positions (including but not limited to ‘Chief Diversity Officer’ positions); all ‘equity action plans,’ ‘equity’ actions, initiatives, or programs, ‘equity-related’ grants or contracts; and all DEI or DEIA performance requirements for employees, contractors, or grantees.”²

Nevertheless, many forms of antibias training are likely to continue to be available and accessible to police departments. Antibias training is a family of programs “aimed at facilitating positive intergroup interactions, reducing prejudice and discrimination, and enhancing the skills, knowledge, and motivation of participants to interact with diverse others” (Bezrukova et al. 2016, 1228). With respect to policing specifically, the Final Report of the President’s Task Force on 21st Century Policing (US Department of Justice 2015) encourages the implementation of implicit bias training for police officers (see also Fridell 2016; Swencionis and Goff 2017). A survey of over a hundred police departments in large metropolitan areas revealed that 96 percent had implemented some form of antibias training program, typically with a focus on implicit bias (CBS News 2019).

While antibias training in recent years has been a booming business that has earned billions of dollars annually, assessments of its effectiveness have yielded mixed results (Devine and Ash 2022; Dobbin and Kalev 2022). However, findings that appear to indicate inconsistent evidence of effectiveness can be reconciled by considering the different ways effectiveness has been assessed. Determining whether antibias training is effective depends on the criteria applied. The evidence of the impact of antibias training is consistent for specific types of outcomes.

One of the most common outcome measures of the effectiveness of antibias training is evidence of prejudice reduction toward traditionally underrepresented groups. However, there is no existing evidence that raising consciousness about implicit bias can produce significant changes in discriminatory behavior (Swencionis and Goff 2017), and there are not sufficient results to make the case that even more comprehensive antibias training can reduce explicit or implicit bias in an enduring way (Greenwald, Dasgupta et al. 2022; Paluck et al. 2021). Both explicit and implicit racial biases are deeply entrenched, often supported—intentionally and sometimes unintentionally—by images in the media and interconnected within a web of social relationships and political orientations and grounded in long-term socialization experiences. Prejudice and stereotypes appear to be highly resistant to change on the basis of antibias training, which tends to be limited in duration.

Nonetheless, when other outcomes are considered, there is consistent evidence of benefits of antibias training. Antibias training does increase people’s knowledge about concepts such as implicit bias and topics relevant to improving intergroup relations. It often piques interest in learning more about ways to improve relations between members of different social groups and motivates people to behave more equitably toward others (Bezrukova et al. 2016).

Regarding antibias training in the context of policing, there is evidence that learning about the science underlying work on implicit bias can increase police officers’ recognition of the potential negative impact of implicit biases on racially equitable policing and create more positive attitudes toward antibias training (Vitriol et al. 2024). Participation in antibias training can, in turn, further expand police officers’ knowledge of implicit biases and increase their intentions to apply antibias strategies to reduce the impact of implicit bias on their professional actions (Kochel and Nouri 2024).

However, the evidence of long-term change in behavior directly caused by antibias training is weak: there is currently not strong or consistent evidence that antibias training can produce an enduring influence on police behaviors (Machado and Lugo 2022; Worden et al. 2024). For example, Calvin Lai and Jaclyn Lisnek (2023) assess the results of a large-scale antibias training—the Managing Bias program, which was developed by the Anti-Defamation League—involving over 3,500 police officers. The main goals of this daylong training were to reduce the influence of bias in interactions and decision-making by police officers, improve police-community relations, and increase officer safety. Participants engaged in activities to increase their awareness and understanding of bias and to build skills (such as perspective taking and individuation) to help reduce bias in policing. The researchers find that the training increased knowledge of bias immediately after the session and one month later. It also increased, relative to the pretraining baseline levels, concerns about holding biases and personally behaving in a discriminatory way, as well as intentions to use the strategies they learned to address bias. However, these effects were fleeting: they were not significantly different from baseline one month later. The researchers therefore conclude that “diversity training[s] as they are currently practiced are unlikely to change police behavior” (Lai and Lisnek 2023, 424). Similarly, a one-day antibias training with over fourteen thousand New York City police officers failed to produce any significant changes in racial disparities in stops, frisks and searches, summonses, arrests, and use of force (Kochel and Nouri 2024).

Rather than dismissing entirely the potential benefits that antibias training (particularly with an emphasis on implicit bias) can have for reducing racial discrimination in policing, Jack Glaser (2024) proposes tailoring such interventions with police officers to focus on actions most vulnerable to the influence of implicit bias. Specifically, he suggests that, because implicit bias tends to influence highly discretionary behaviors (such as stop-and-search decisions) more strongly, both the training and its assessment might concentrate on these types of policing activities. However, antibias training that emphasizes implicit bias may inadvertently undermine people’s motivation to exert the effort needed to overcome the influence of implicit bias on their behavior. Because implicit bias reflects a high degree of automaticity and thus limited controllability, when discrimination is attributable to implicit bias people are perceived to be less accountable for their actions and view their discriminatory behavior as less deserving of punishment (Daumeyer et al. 2019).

When trainings are viewed as a cure-all, therefore, the science is clear that any form of training is likely to be a weak lever for change at best. When viewed as a signal about cultural norms, however, trainings may yet prove valuable. While poorly integrated diversity, equity, and inclusion efforts can send the wrong signals about institutional norms and values (Dover et al. 2020), this same research literature holds out hope that, framed correctly, trainings can be part of an organizational signal that combating discrimination is important. Trainings, therefore, need not be dismissed if they do not provide immediate changes in behaviors. Rather, they must be understood in the context of their broader organizational context.

Focused Training

While attempts to reduce levels of explicit and implicit prejudice through antibias training have had limited success, training that is focused primarily on developing and practicing skills related to specific aspects of policing promises greater success, at least in the laboratory. For instance, research has shown that with practice on the shoot/don’t shoot simulation paradigm, police officers, in general, are able to reduce their discrimination against Black (relative to White) targets on this task (Peruche and Plant 2006). This reduction in racially discriminatory responses on this task is particularly strong for officers who have previously had relatively high amounts of positive personal contact with Black people. Other research (Sim et al. 2013) also has demonstrated that training on a shoot/don’t shoot task reduces racially biased errors by police officers when race is unrelated to the presence or absence of a weapon. However, when Black people are experimentally associated or chronically associated with danger (for police officers who routinely deal with minority gang members), training does not reduce racial bias on the task but may exacerbate it. Taken together (while recognizing the existence of boundary conditions), these findings reveal the potential for narrowly tailored training involving activities in which officers have opportunities to practice their skills to reduce police use of lethal force. However, such training requires the use of appropriate scenarios and simulations to facilitate their application to the conditions under which they may be enacted within the duties of police officers (Fridell 2016).

An alternative approach involves training police officers in new skills for effectively interacting with community members to prevent conflict escalation, thereby reducing the perceived need to use force. Currently, police academies in the United States devote over three times more hours to firearms training than to de-escalation techniques (Horton 2021). While there is very limited evidence concerning whether de-escalation training works and reduces racial disparities in policing (Agan 2024; Engel et al. 2020), there are techniques that may help de-escalate conflicts between police officers and community members and reduce racial disparities in policing. George Wood, Tom Tyler, and Andrew Papachristos (2020) tested the impact of a one-day procedural justice training program with the Chicago Police Department. The intervention “emphasized the importance of voice, neutrality, respect, and trustworthiness in policing actions.” Before making a decision, “officers were encouraged to provide opportunities for civilians to state and explain their case.” The officers were also encouraged to “apply consistent and explicable rules-based decision-making, treat civilians with dignity and respect their status as community members, and demonstrate willingness to act in the interests of the community and with responsiveness to civilians’ concerns” (Wood et al. 2020, 9815). This training provides officers with a broader range of strategies, grounded in a body of empirical evidence, for handling challenging encounters more effectively. It also communicates priorities and norms of the department, which can influence behavior in additional ways beyond the specific knowledge and skills acquired in the training. This training significantly decreased officers’ use of force in the community. Before training, in a police force made up of 7,785 police officers, there were 363 use-of-force incidents per month. After training, according to a conservative estimate, there were 8 fewer use-of-force incidents per month; according to a more aggressive estimate, there were 156 fewer use-of-force incidents per month.

While any officer trainings that yield positive results in the field should be lauded, they would be incapable of correcting for policies or cultural norms that produce unequal outcomes. Consequently, we turn our attention next to micro-level processes.

Psychology

The traditional approach in psychology has been laboratory research using designs with experimental manipulations and appropriate controls to form cause-and-effect inferences. Psychological analysis is thus complementary to the types of approaches frequently used in other behavioral science disciplines. With respect to understanding situational-level dynamics of racial bias in policing, the relevant research builds on foundational interpersonal processes identified in psychological research and the application of these dynamics to activities related to policing studied with samples unaffiliated with law enforcement and, less often, directly with police samples.

One example of psychological scholarship that leaves the laboratory to answer questions about micro-level discrimination leverages newly available police body-worn camera footage to examine police behaviors during interactions with residents. Early results provide evidence of racial bias in the ways police speak to citizens in their encounters. During routine traffic stops, an analysis of interactions recorded on police body-worn cameras reveals that police officers speak with less respect to Black than to White community members (Voigt et al. 2017) and display a more negative tone in their communications (Camp et al. 2021). While these studies do not allow us to determine to what degree these officer behaviors are driven by intrapersonal, nano-level and interpersonal, micro-level factors or shaped by meso- (organizational), macro- (pan-organizational, institutional), or mega- (cultural) level factors, this line of research at least casts doubt that mere “suspect resistance and/or non-compliance” is the only factor that escalates police-resident interactions, a common argument made in both popular media and research to account for racial disparities in police violence (Terrill 2005).

The availability of body-worn camera footage also potentially allows for quantitative and time-sequenced analyses of encounter escalations—the process of making interactions more tense and likely to produce violence (Ariel et al. 2016). However, without the help of artificial intelligence, aggregating and coding this footage is labor intensive. Additionally, the footage is not available at scale, making it difficult to know what results will and will not generalize.

Finally, some research integrates the literature on stereotype threat with concerns of being stereotyped as criminal (Najdowski et al. 2015) or as racist (Trinkner et al. 2019). As noted in the section on nano-level processes, stereotype threat represents people’s concern about being judged in accordance with unfavorable stereotypes associated with their social identity. Stereotype threat has interpersonal consequences (micro-level effects) relevant to policing, as well as intrapersonal impact (nano-level effects). As research shows, Black participants report higher levels of concern with being stereotyped as criminals than do White participants. That elevated concern leads to higher regulatory efforts, something previous research has identified as producing racial disparities in behaviors that are judged to be indicators of guilt (Najdowski 2011). In other words, as with other stereotype threat domains, concerns with confirming or being evaluated in terms of a racial stereotype ironically produce behaviors consistent with the stereotype that can have direct consequences for how Black people and police officers interact. This is also an area in which nano-level (intrapersonal) and micro-level (interpersonal) processes operate in tandem. There are individual differences (a nano-level influence) in the concern with being stereotyped that reliably predict differences in behaviors (McCarthy et al. 2021; Trinkner et al. 2019), and these behaviors influence interactions between police officers and community residents. It is further possible that processes at the meso-, macro-, and mega-levels, such as discriminatory policies or regional reputations, can moderate the impact of these nano-level and micro-level processes in influencing racial inequities in policing.

The dynamics of stereotype threat in encounters between police officers and Black residents can produce a reciprocal set of influences mutually increasing threat and escalating the potential for conflict and excessive use of force (Najdowski 2023). Indeed, footage from police body-worn cameras offers supportive evidence for this process (Rho et al. 2023). A linguistic analysis of footage from 577 stops of Black drivers finds that stops with escalated outcomes (that is, those ending in arrest, handcuffing, or a search) differ systematically from other stops from the initial moments of the interaction—even in the first forty-five words spoken by the officer. For stops that escalated, officers were more likely to begin the interaction with commands, without explaining why they made the stop. Black men who viewed the footage perceived more negative emotion, viewed officers more negatively, worried more about force being used, and predicted worse outcomes after hearing only the officer’s initial words in escalated versus non-escalated stops. This escalation was likely the outcome of the reciprocal responses, possibly stimulated by stereotype threat experienced by both parties.

Ethnography

The bulk of research on police-resident interactions belongs to ethnographers. For instance, Nikki Jones’s (2014) work on interactions between police and youth in California’s Bay Area reveals how Black youth learn the informal sequencing of a police stop, as well as when and how they can voice their objections to abuses. Similarly, Rod Brunson’s (2007) work reveals which direct and vicarious interactions lead Black youth to believe that “police don’t like Black people” (from the title of Brunson 2007), as well as how those observations create both ideologies about policing and specific tactics to navigate communities where those same police are omnipresent. These findings build on a generation of ethnographic observations about police-resident interactions. However, while work such as Elijah Anderson’s (2013) Streetwise paints incredibly detailed pictures of how aggregated interactions with police shape neighborhood mindset and praxis, the quantitative tests of how these observations scale beyond the communities in which they are observed have lagged behind considerably.

Criminology, Economics, and Political Science

Other quantitative fields have developed literatures at the micro-level, particularly economics and political science. The (mostly) economics literature around “hit rates” (the rate at which officer searches yield contraband) is a notable example. Much like the “cover of darkness” methodology, the literature on hit rates (or yield rates) attempts to bypass a common methodological problem with analyzing and determining the causes of racial disparities in police behavior—namely, that the universe of possible outcomes is rarely known. In other words, it is difficult to tell whether there is racial discrimination in police use of force without being able to account for all incidents where force might have been used but was not. Analyses of the rate at which police searches yield contraband (“hits”) sidestep this methodological limitation (at least in part; see Knox et al. 2020) and by testing the outcomes of a conditional action—that is, a search. The result of such tests (Knowles et al. 2001) reveals that, even if we set aside whether a decision to search someone is racially biased, officers are robustly not using the same criteria for searching Black residents as White residents—and the result is less effective policing generally, even beyond the potentially racist consequences. This tendency to “over-search” Black residents could stem from departmental policies or a broader culture in policing (see Ratcliffe 2004), yet there is some evidence of individual differences.

Along these same lines, economists and political scientists have recently provided the best test yet of the importance of micro-level phenomenon. Specifically, an unparalleled matching study in Chicago concluded that police demographics play a major role in police use of force, with White men accounting for significantly higher rates proportional to the number of Black men they target (Ba et al. 2021). While it is not clear if or how these results generalize, this first true apples-to-apples comparison of officers by location, shift time, and department suggests that individual differences may play an underappreciated role in how interactions with police produce community violence—a merging of nano- (intrapersonal), micro- (interpersonal), and meso- (organizational) level processes.

One theme across all these micro-level literatures is that it is difficult to disaggregate micro-level processes from processes at other levels. Another way to frame that methodological hurdle is this: racial discrimination exists at so many levels and so far before a particular police-resident interaction that it is often difficult to tell when the police are the problem or when the outcome of the interaction is merely a symptom of a larger issue. Dean Knox, Will Lowe, and Jonathan Mummolo (2020) make this point in a highly technical article by demonstrating the many steps one must take to demonstrate racial bias at the level of an individual stop—often preventing scientists, litigators, and policymakers from making appropriate inferences about racism in policing. Importantly, they admit that getting access to all the data necessary to conduct the ideal analysis is impossible in many cases. (The article, notably, does not feature any analyses of police data.)

This theme is worth noting because, ironically, while the first two levels of sociopolitical organization (nano, involving intrapersonal processes, and micro, focusing on interpersonal processes) are the levels at which it is most difficult to tie processes to outcomes, they are also the most commonly invoked in national conversations about racism. In other words, the hardest levels to prove are the most common to be highlighted in discourse, making broad consensus concerning when and where racism is functioning that much harder.

Warrior Prototype

As Amie Schuck (2023, 1) explains, a warrior police prototype prioritizes “the crime-fighting mission, stresses control-focused communication approaches, and characterizes the relationship between the police and the community as detached and guarded” (see also Balko 2021). The warrior prototype valorizes gun-centric, aggressive masculinity. Seth Stoughton (2014; see also Simon 2023) has observed that police culture and training within the United States encourage the development of a warrior mindset and mentality in which police officers “are taught they live in an intensely hostile world, a world that is quite literally gunning for them” (Stoughton 2014, 227). Beginning from their time of training in police academies, officers are encouraged to conceptualize their relationship with the public as a war, identifying their enemy in gendered and racialized ways, primarily as Black and male (Simon 2023). This orientation contributes substantially to a tendency for police officers to perceive a threat from the initiation of an interaction with a resident, particularly a male member of a minoritized groups, and therefore to respond with excessive force, leading to negative outcomes that are more violent than the situation warrants. Indeed, greater officer endorsement of traditional police culture—which includes distrust of the public, distrust of community members, a crime-fighting orientation, and concerns about danger and bravery (Crank 2014; Paoline 2003)—relates to greater support for use of force and less support for policing that involves treating community members with respect and helping residents in ways that go beyond what is necessary to resolve a situation (that is, procedural justice; Silver et al. 2017). Adopting a warrior mentality can contribute to racial disparities in policing through processes of discrimination. For instance, one study (Mummolo 2018) finds that across departments in Maryland, SWAT teams are disproportionately deployed to Black, compared to White, communities—even when researchers control for crime rates.

Guardian Prototype

In contrast to the warrior prototype, “the guardian orientation prioritizes the service mission of the police, values creating authentic connections with residents and building meaningful partnerships with representatives of the community, and emphasizes using communication approaches grounded in dignity and respect” (Schuck 2023, 1). While there have been frequent calls for law enforcement to move away from the warrior mindset to a guardian mindset to reduce excessive use of force (for example, McLean et al. 2020), this may not be sufficient to achieve racial equity in policing. One reason is that the warrior and guardian orientations are not as antithetical as they might initially appear. They can, in some forms, represent aggressive and violent policing. Indeed, officers who are more supportive of aggressively enforcing the law are also more supportive of the guardian orientation (Clifton et al. 2021). Moreover, both the guardian orientation and the warrior orientation frequently reflect a racialized image that also promotes discriminatory use of force against Black people and members of other minoritized groups. Jennifer Carlson (2020, 399) reports, on the basis of seventy-nine in-depth interviews with police chiefs across the country, “Whereas the ‘warrior’ brand of police masculinity emphasizes aggressive enforcement against (black and brown) perpetrators, the ‘guardian’ brand of police masculinity emphasizes assertive protection on behalf of (white) victims.”

Racialized aggressive policing guided by either warrior or guardian mentality enforces racial dominance through policing practices. Consequently, both warrior and guardian mentalities can perpetuate racial disparities and discrimination in policing through meso-level mechanisms in combination with individual-level processes. “Over-policing” the public generally, and Black members of the public specifically, increases the frequency of contentious police-resident contact. This contact often escalates into police use of forcegenerally, and especially when officers adapt a “warrior-style” approach “that normalizes the expectation of unquestioned compliance with police directives and authorizes police to use physical force in its absence” (Cobbina-Dungy and Jones-Brown 2023, 3). As discussed in the section on interpersonal, micro-level processes, such practices erode trust between police officers and residents, harm members of the community generally, and reduce community member cooperation with law enforcement—which can increase the crime rate in the community. A higher crime rate, in turn, becomes a justification for further over-policing and use of force by police. A higher crime rate in a locality is a strong predictor of greater use of force by police (Shjarback 2018).

Levels of Police Culture

Although much of the research on police culture has focused broadly on the shared institutional culture of policing, which Gary Cordner (2017; see also Ingram et al. 2018) describes as occupational culture, police officers belong to many different groups and thus may be aware of and adhere to the norms of multiple subcultures (see Schein 1990). These subcultures can affect officers’ propensities to engage in excessive force against Black people and to tolerate such acts by other police officers (Micucci and Gomme 2005). In addition to institutional culture, there are subcultures within policing at the organizational level (reflecting the distinctive orientation of each specific agency and the community in which it is located) and within an organization at the levels of rank (involving cultural variation based on an officer’s position in the organization’s structure) and unit (such as patrol groups).

Occupational police culture, which is a culture widely shared across police organizations (a macro-level phenomenon), is extremely resistant to change because it is shaped by traditions and other influences beyond the department or agency and outside the responsibility and influence of police leaders, such as police union influences and community attitudes (Hehman et al. 2018). But reshaping local cultures (which reside in the meso-level), such as those of an agency or department (that is, at an organizational level) or unit, is more possible.

Leadership and Organizational Culture Changes

Within an organizational context, leaders play a fundamental role, both formally and informally, in group dynamics, and ultimately in the culture, of the groups that they lead (Haslam et al. 2020). This influence is particularly pronounced in formally hierarchical organizations with a clear chain of command, such as in police agencies and departments. In organizations structured hierarchically, authority of the leader is explicitly acknowledged, and the leader has the responsibility to make certain decisions and enforce compliance with rules and policies.

Beyond the formal bases of influence within a group, leaders also have broad psychological, and ultimately, behavioral impact. Leaders are seen as embodying the values of a group and thus help shape the meaning that individuals within the group attach to group membership. Thus, in addition to their formal influence in making decisions and enforcing compliance with rules and policies, police leaders have broad psychological impact (Ingram et al. 2023). A police leader is a uniquely powerful role model, and the values and motivations the leader professes and the actions the leader exhibits have enduring influence on how members relate to each other and with the outside community (Dubord and Griffiths 2021). As Fei-Lin Chen and colleagues (Chen et al. 2021, 893) note, “One of the plausible mechanisms of promoting fair and accountable policing on the street is that police agencies and supervisors treat rank and file officers fairly and justly.” Thus, sustained cultural change can occur through the exchanges that police leaders model and foster within an agency or department.

As discussed in the section on micro-level interventions, training programs to help police officers develop procedural justice-related skills in their interactions with community members significantly reduce police use of force incidents (Wood et al. 2020). Procedural justice training for police officers also increases the perceived legitimacy of the police among community members and community members’ trust in and willingness to cooperate with the police (Mazerolle, Bennett, et al. 2013). From the perspective of community members, as the Final Report of the President’s Task Force on 21st Century Policing (US Department of Justice 2015) emphasizes, the perceptions of the police as fair and policing as legitimate are foundational for effective policing.

Beyond providing the opportunity for officers to acquire specific skills, such interventions also communicate the priorities of police leadership, influencing officers’ perceptions of organizational values and norms. In fact, the communication of organizational priorities and norms signaled by new training initiatives may have more impact on organizational change than the acquisition of particular skills through training. For instance, Emily Owens and colleagues (2018) developed an intervention to test the influence of the behavior of supervisors that communicated the importance of procedural justice on officers’ actions in the field. In this experimental test, police officers were randomly assigned to a nonintervention control condition or to an intervention condition in which the officer would meet with a supervisor to discuss an incident involving the officer and for which a report had been filed or an arrest had been made. In this meeting, the supervisor explained that procedural justice was a central part of the department’s mission, and then the supervisor modeled and discussed procedural justice in relation to the incident with the officer. The outcome measure of interest was the subsequent behavior of the police officer. Officers who participated in the intervention made fewer discretionary arrests and were less likely to be involved in incidents in which force was used. These effects were particularly pronounced among officers who worked in areas with a modest level of risk, defined as locations where other officers tended to use force, were injured, or received complaints from citizens.

Moreover, the relationship between police agency internal functioning and engagement with the community is reciprocal. Prioritizing procedural justice practices internally and externally in policing increases trust among police officers and community members, and trust is needed to create productive police-community partnerships (O’Brien and Tyler 2019). Collaborations between police and community partners as part of a public safety system that seeks to address factors that contribute to crime in order to prevent it from occurring reduce mutually experienced threat and improve attitudes. Such conditions of more cooperative and trusting police-community relationships, in turn, can facilitate further changes in police organizational culture. In general, groups are more open to adopting supportive orientations toward members of other groups in contexts that are characterized by cooperation rather than conflict (Arendt et al. 2023).

Policy Changes

While attempts to change police culture at an organizational level can have broad and enduring impact if successful, the most direct way to reduce racial disparities and discrimination in policing is through strategic changes in agency or departmental policies (Durán and Shroulote-Durán 2021). The influence of policies restricting officer discretion in the domains of domestic violence (for example, mandatory arrest), high-speed pursuits (do not engage in them), and treatment of juveniles (for example, when and under what circumstances to ask questions) has been well documented in the literature, producing the commonsense consensus that policy was the tool by which policing could be aligned with community values. During the late 1980s and 1990s, criminologists attempted to extend this understanding to the realm of police violence.

Researchers investigating what mechanisms could reduce unnecessary or particularly brutal police use of force coalesced around the notion that strict regulations through policy were the answer—an institutional or organizational response to aggregated individual problems. As noted earlier, allowing greater discretion in action permits people’s racial biases to exert more influence on their behavior. This principle is also reflected in policing practices and outcomes (Swencionis and Goff 2017). In general, restrictive policies, which limit officer discretion, have reduced a variety of negative consequences related, for example, to high-speed pursuits (Alpert and Madden 1994) and misidentification of people based on eyewitness evidence (Kovera et al. 2022). There is strong evidence that restrictive rules can reduce police use of force and or killings generally (Fyfe 1988; Geller and Scott 1992; White 2001). Nevertheless, while it is likely that restrictive policies that generally reduce officer use of force may also reduce racial disparities in lethal and nonlethal use of force because they limit officer discretion (Swencionis and Goff 2017), there is a relative dearth of evidence about what kinds of policies particularly reduce racially discriminatory behaviors and outcomes.

When police organizations have policies (such as incentives, sanctions, and required in-service trainings) that attempt to control officers’ discretionary decision-making, they reduce the likelihood of misconduct (Eitle et al. 2014). Regular in-service training throughout the year can result in reductions in police misconduct (Eitle et al. 2014). In terms of officer use of force, departments that have adopted more restrictive use-of-force policies demonstrate lower rates of use-of-force incidents (Terrill and Paoline 2017). Even policies that do not directly restrict use of force but mainly monitor it more closely can reduce it. For example, research has shown that having a policy requiring a supervisor to fill out the use-of-force reports is associated with less use of force (Alpert and MacDonald 2001) and that instituting a policy requiring officers to report whether they pulled their firearm and pointed it at someone but did not fire reduces instances of overusing firearms as a control technique (Jennings and Rubado 2017). Given the generally much greater use of lethal force with Black than with White people, the personal and social impact of these policies would likely be greater for Black than for White people. Researchers should, however, be cautious about any benefits observed following discretion-restricting interventions as nearly all police-related outcomes are recorded by police officers. In other words, it is often a methodologically painstaking task to determine if the outcome measure improved or whether officers simply stopped documenting the problem behaviors.

Another intervention that reduces discretion in decision-making and action that has been increasingly implemented across the justice system in the United States is the use of automated decision-making technology to achieve “algorithmic fairness.” While a promising application for policing, to the extent that the data used to inform automated decision-making technology incorporate unrecognized cultural biases or inaccurate historical narratives (Ugwudike 2022) or include inappropriate assumptions in their analyses, these algorithms have the potential to reinforce racial inequities in the justice system. Roland Neil and Michael Zanger-Tisher (2025), for instance, caution that racial biases are likely to result when these procedures treat arrest data as an unbiased measure of criminal offending because evidence shows the role of prejudice-based discrimination in creating racial inequities in arrests.

Officer body-worn cameras are, among other reasons, a frequently used technique for increasing police officer accountability. A study of the Los Angeles Police Department (McClusky et al. 2019), for instance, finds that body-worn cameras increase the extent to which officers exhibit procedural justice (for example, showing interest, behaving respectfully) in their interactions with community members. With respect to use-of-force incidents, a recent review of approximately seventy empirical studies reveals that the use of body-worn cameras significantly decreases police use of force by 9.6 percent (Williams et al. 2021). However, while the overall effect is significant, some researchers have noted that the impact of body-worn cameras on use of force is variable across studies (Lum et al. 2019). One reason is that the policies in some departments allow officers discretion to turn off the cameras (Machado and Lugo 2022). Another reason might be that because the presumed effect of body-worn cameras occurs, at least in part, by increasing officer awareness and thus compliance with rules and norms (Ariel et al. 2017), the effectiveness of body-worn cameras may depend on departmental or agency norms and culture involving the application of use of force. This latter interpretation suggests the importance of considering the impact of culture and policy in combination in examining police use of force generally and particularly toward members of Black people and other minoritized groups. Police culture can significantly blunt the influence of policy (White 2001). For the use of body-worn cameras to have a stronger and more consistent effect for reducing racial inequities in policing, it may thus be valuable to couple the use of the technology with training interventions that help police officers internalize the principles of racially equitable policing (McDonald 2025).

This possibility is supported by results showing that the impact of body-worn cameras on police behavior can be complex and is moderated by factors such as neighborhood characteristics. For example, in a study of the New York Police Department, the implementation of body-worn cameras reduces community complaints against officers by 21 percent but increases police stops of community members by 39 percent (Braga et al. 2022). Other research has found that implementation of body-worn cameras increases police enforcement activity in areas with a high percentage of Black residents and tends to increase these activities in high reported-crime areas—both neighborhood characteristics normatively associated with greater police enforcement activity (Hughes et al. 2020). Taken together, the findings for the effects of body-worn cameras suggest the value of further research to understand how they influence police behavior (particularly with respect to racial disparities) by considering both organizational and structural (for example, neighborhood) factors. Understanding these effects more comprehensively can also help produce a more accurate cost-benefit analysis of body-worn cameras (for example, considering the effects on reducing community complaints versus the cost of the equipment; see Williams et al. 2021).

Residential Segregation and Policing

While the associations of community socioeconomic standing, composition, and crime with racial disparities in policing are often statistically separable, these factors are correlated at a societal level and can be traced to similar root causes. They are the legacy of formal and informal policies producing racial residential segregation in the United States (see Penner et al. 2023). A range of structural factors over time have produced high levels of racial residential segregation today, particularly between Black and White people. Racial residential segregation was formally enforced through the institution of slavery, and then subsequently through Jim Crow laws, and through laws and statutes (such as those related to “sundown towns,” which were towns in which Black people had to leave at dark) into the 1960s. Even after the landmark civil rights legislation of the 1960s, restrictive zoning and covenants, federal policies for public housing, redlining for housing loans, and the persistent influence of individual-level racial bias continued to promote residential segregation.

Segregated housing has been and continues to be a major cause of racial inequities in all facets of residents’ lives. Policies and individual-level biases that promoted racial segregation in housing also relegated Black neighborhoods to geographic areas that were less desirable, less healthy (because of more exposure to toxins or less access to healthy foods, for example), and more limited in access to essential healthcare services. Clearly, not all majority-Black neighborhoods are underresourced or have residents who are predominantly of low socioeconomic status. Nevertheless, on average, the residents of most predominantly Black neighborhoods are much more likely to be poor than are residents of predominantly White neighborhoods. Poor Black people are more than three times as likely as poor White people to live in neighborhoods where the poverty rate is 40 percent or more. Segregation is a major contributor to “racialized concentration of poverty” (Massey 2016).

This economic disadvantage limits social mobility. In 2020, lenders were 80 percent more likely to deny a mortgage to Black applicants than they were to White applicants (Glantz and Martinez 2018); economic factors were the most commonly cited explanation for this disparity. Also, high school students from families lower in socioeconomic status are less likely to attend college, have fewer choices about where to attend college, and when they do attend college, they are more likely to attend less selective institutions (Walpole 2003). Areas higher in poverty also tend to have higher rates of reported crime (Sharkey et al. 2016). One study reveals, for instance, that men who grew up in low-income families are twenty times more likely to be incarcerated than men who grew up in high-income families (Looney and Turner 2018). Additional research has identified a direct causal link between poverty and criminality, over and above other related causes such as unemployment (Imran et al. 2018).

Such neighborhood contexts also promote racial bias among police officers. Police officers working in neighborhoods with higher rates of poverty, greater crime, and a higher percentage of Black residents have more negative attitudes and express more negative stereotypes of adolescents in the community (Fix et al. 2023). As discussed earlier, negative attitudes and stereotypes of police officers are a significant driver of racial disparities in policing, including stops, searches, and use of nonlethal and lethal force. These biases are particularly likely to produce excessive and racially disparate use of force in these areas because police officers tend to behave more aggressively in places that they consider more dangerous (Gaston and Brunson 2020).

Disinvestment in Systems of Support

Disinvestment in systems and resources previously used to respond to substance abuse (public parks, homelessness, schools, in-patient hospital crises, and mental health) has been accompanied by increased reliance on policing, substituting systems of preventive care with systems of responsive punishment (Hinton 2016; Soss and Weaver 2017). Each produces racial disparities by dint of geography, concentrated disadvantage, and bias across levels. Each is a community crisis that used to have significant dedicated resources devoted to preventing and treating it. Increasingly, policing and the threat of incarceration have been the response of first resort (and sometimes last resort, as well). Much has been written about how police can respond to each of these crises more efficiently, equitably, and humanely. However, given the history of each, the question must be asked: why are police the responders we have decided are best equipped to respond to these crises? The case of mental health may be most instructive, as there is a sizable literature on deinstitutionalization and the corresponding rise in policing (Lamb and Weinberger 2020).

The decline in public spending on mental health has been precipitous since the 1950s and parallels an increase in police contact with individuals navigating mental illness (Lamb and Weinberger 2020). However, this apparent trade-off from care to punishment does not track with public opinion. Rather, the public does not have confidence in police officers’ ability to deliver appropriate care to people navigating mental illness (Milam and McElwee 2021), and the limited research on training officers to improve that care reveals an ineffective standard at best (Compton et al. 2008). Similarly, while officers report believing that individuals with serious mental illness should be treated less harshly than other residents and are less culpable for their wrongdoing than their neighbors, a recent analysis of police use of force toward the same population shows major disparities, with persons with serious mental illness experiencing twelve to thirty times higher rates of police use of force than their representation in the population would predict (Laniyonu and Goff 2021).

Problem-Oriented Policing

Anthony Braga and colleagues (2019; see also Petersen et al. 2023) propose, with supporting evidence, that problem-oriented interventions generally produce larger decreases in crime than traditional police actions in hot spots without producing mass arrests or unnecessary stops that undermine police legitimacy and erode police-community relations. According to these researchers (545–46), problem-oriented policing is “a proactive crime prevention strategy that seeks to identify the underlying causes of crime problems and to frame appropriate responses using a wide variety of innovative approaches” using “strategies can help safeguard against the use of overly harsh and indiscriminate enforcement tactics.” Such interventions are most effective when they involve working with the community to establish bi-directional communication, facilitate joint problem-solving, promote procedural justice, and build resilience and collective efficacy with an enduring structure for productive police-community relations (National Academy of Science, Engineering, and Medicine 2018). Braga and colleagues (2019, 547) explain: “Developing close relationships with community members would help the police to gather information about crime and disorder problems, understand the nature of these problems, and solve specific crimes. Community members can also help with key components of strategies tailored to specific problems by making improvements to the physical environment and through informal social control of high-risk people. In this way, police strategies focusing on particular people and places would cease to be a form of profiling and instead become a generator of community engagement projects.”

Such police-community partnerships are also effective at increasing perceptions of police legitimacy, which contributes to more effective policing. When people see the police as more legitimate, they are more likely to adhere to the law and to cooperate with law enforcement in reporting crime and information that is valuable for addressing crime.

Community Organizations

While police-community partnerships can be valuable for improving public safety, community organizations, made up of neighbors and residents alone, can also be very effective. Patrick Sharkey, Gerard Torrats-Espinoza, and Delaram Takyar (2017) identify numerous nonprofit community organizations, such as Alianza Dominicana and the Community League of West 159th Street in New York City, that have played major roles in reducing crime in their communities. Such nonprofits employ a variety of strategies, including building affordable homes and training and hiring formerly incarcerated residents to maintain neighborhood spaces, but they share a common goal of addressing violence and rebuilding community. Testing the often-overlooked impact of community organizations of this type across a sample of 264 cities and examining effects across a twenty-year period, the researchers conclude that “every 10 additional organizations focusing on crime and community life in a city with 100,000 residents leads to a 9 percent reduction in the murder rate, a 6 percent reduction in the violent crime rate, and a 4 percent reduction in the property crime rate” (Sharkey et al. 2017, 1214). Thus, these organizations can significantly reduce crime in the absence of aggressive policing or expanded police presence.

Neighborhood Investment

To the extent that concentrated poverty relates to high levels of criminal activity in a locality, alternative interventions that do not involve policing may be effective for reducing crime by targeting a root cause. One program intended to do this involved creating “Opportunity Zones” in economically distressed areas by offering special tax incentives to businesses investing in these neighborhoods. Despite the admirable aims of this program, Opportunity Zones have had limited impact on the community while providing more substantial benefits to investors (Theodus et al. 2020). However, the Urban Institute recommended changes that would provide a sound basis for any program designed to infuse communities with jobs and investments, such as prioritizing projects that have the most immediate economic impact on the community (such as job creation) over those that generate the most profit for the investors and seeking more support from Community Development Financial Institutions, which are private financial institutions that are dedicated to providing affordable lending to help low-income and socially disadvantaged people and communities improve their economic standing. In addition to bringing capital into poor and underresourced Black neighborhoods, these investments can also ameliorate the conditions that foster criminal activity. Advocates of “placemaking” argue that unless outside investments are directed at the causes of a neighborhood’s problems, they will not create any sustainable change (Adebowale-Schwarte 2017).

Beyond the macro-level interventions that we have identified in this section are other fundamental structural interventions that represent alternatives to traditional policing. These alternative approaches may not only better address the problems of equitably policing but also create systems of support to better serve the public.

Methodological Refinement

One of the major methodological challenges for understanding the bases of racial disparities in policing involves the issue of directly establishing the causal effect of race, which is essential for determining the impact of racism in policing (National Academies of Sciences, Engineering, and Medicine 2018). In field studies, typically conducted by behavioral science researchers in disciplines other than psychology, techniques such as benchmarking and time-series analysis are often used to test hypotheses about the role of discrimination in racial disparities in policing. However, depending on how a researcher implements these techniques, they may not produce valid inferences about the causal role of discrimination (Neil and Winship 2019).

Traditionally, particularly in psychology, the gold standard for establishing causality in research has been an experiment that tests the impact of a manipulated independent variable on the outcome of interest and controls for other influences in the situation. For both ethical and practical reasons, such direct experimental work in the naturalistic context of policing is relatively rare. Much of the experimental evidence in psychology we have described, such as the shooter bias, has involved simulations of police-relevant behaviors using either participants from the general population or police officer participants. While these experiments help identify underlying causal processes (for example, the role of individual-level biases on perceptions of threat based on the race of another person), the external validity (the generalizability) of such findings is limited. Quasi-experimental tests (for example, the veil of darkness paradigm, which shows that racial disparities in police stops of motor vehicles are significantly smaller at night, when it is more difficult for officers to discern the race of the driver, than during daylight) are thus particularly valuable because they complement the findings of laboratory-controlled research on bias and discrimination in critical ways.

Multilevel analysis can also help, methodologically, to distinguish various types of influences that could account for the same policing outcome. For instance, the finding related to structural-level processes that the greater incidence of police stops in neighborhoods that are predominantly Black than in those that are predominantly White (for example, Fagan et al. 2016) could be a consequence of individual-level police bias, greater police racial bias, or officers’ experiences with more crime in the area (Braga et al. 2019). Time-series analyses considering individual-level factors (for example, increases in personal bias or perceptions of threat) and location-related factors (for example, changes in neighborhood composition or crime rate) could help disentangle potential causal influences for such associations. Determining the nature of causality is essential for designing, implementing, and assessing interventions to reduce racial disparities in policing.

In addition to further refinement of the methodologies used to validly assess the influence of discrimination on racial disparities in policing, the triangulation of findings from less than perfect paradigms across levels of analysis and disciplines with appropriate caution can still contribute positively to theory development and testing, as well as help guide practical interventions.

Theory Development

Considering multiple levels of sociopolitical organization also generates productive theoretical developments. Here we offer one such example. Ryan Hübert and Andrew Little (2023, 2828) describe a model of how police officers “form beliefs using crime statistics but do not properly account for the fact that they will detect crime in more heavily policed communities.” The effect can be understood as an example of the nano-level fundamental attribution error (Ross 1977), in which people tend to overestimate the influence of others’ actions on the personal characteristics of the actor (dispositional attributions) and underestimate the influence of the context in which the action occurs (situational or environmental attributions). These dispositional attributions, in turn, guide the subsequent perceptions of community residents, generating greater suspicion of criminal activity by police officers (a micro-level process) and produce organizational responses, such as intensified policing in specified locations (a meso-level response). This intensified policing can then yield greater evidence of criminal activity, not because of any increase in the individual propensity for such activities by community residence, but because more incidents are detected by greater police surveillance. This process then adds to the body of crime evidence that police officers use to form their beliefs about racial differences in crime. The researchers conclude, “This creates a feedback loop where officers over-police groups that they (incorrectly) believe exhibit high crime rates. This inferential mistake can exacerbate discrimination even among officers with no animus and who sincerely believe that disparities are driven by real differences in crime rates” (Hübert and Little 2023, 2828).

We propose that explicitly structuring conceptual analyses around the premise that distinct forms of bias reside in different levels of sociopolitical organization and that these influences combine in particular ways can help provide more theoretical coherence to the existing literature and offer new insights for emerging works. Given that processes at one level may combine with processes at any other level (or multiple levels), one priority for the task of developing this conceptual framework would involve empirically identifying which combinations are most influential in practice in a given context—in this case, the context of policing. Understanding these dynamics more deeply can inform the development of interventions that not only address biases in policing associated with processes at different levels but also have broader impact by targeting unique multilevel dynamics of racial bias in policing.

Intervention Development

Both theoretically and practically, not only does simultaneous consideration of different levels of sociopolitical organization promote a more comprehensive understanding of the dynamics underlying racial disparities in policing, but it can also facilitate more expansive consideration of solutions. Our analysis of conditions that tend to increase the likelihood of racially discriminatory policing through individual-level processes (see also Pryor et al. 2020) suggests changes in practice or policy to reduce racially biased misconduct by police officers at the individual level. As Jillian Swencionis and Phillip Atiba Goff (2017) note, limiting officer discretion—for example, by clarifying conditions under which use of force is permissible—would be an important step for reducing racial discrimination in this area. This can be accomplished through more meso-level changes in policies within a specific police department (meso-level) or at a broader, macro-level. As an example, since the murder of George Floyd in 2020, twenty states have modified their use-of-force standards, typically clarifying that deadly force is justified only as a last resort after exhausting all nonviolent options (Subramanian and Arzy 2021).

As we previously discussed, a large body of research indicates that, at the individual level, police officers’ stereotypes of Black people (and other minoritized groups) and other explicit or implicit biases systematically contribute to racial disparities in police stops and searches and in nonlethal and lethal use of force. Traditionally, the recommended solutions for these effects have also been individual-level, such as improving selection procedures or having officers participate in various forms of antibias training. However, the scholarly evidence indicates that there is not a specific type of training intervention that can reliably reduce implicit bias and possibly even explicit bias in an enduring way (Greenwald et al. 2022; Paluck et al. 2021). A multilevel approach suggests promising alternatives to reducing discriminatory behavior—an ultimate goal.

We considered the role of norms as a meso-level organizational process. Norms are the informal rules and practices that define how individuals should behave within an organization. Whereas stereotypes and attitudes are well-established habits of mind that are difficult to change because they are connected to multiple aspects of social networks and attempts to change them can create backlash, people readily adapt their behaviors to the norms of various situations they encounter. People automatically activate normative guidelines in these contexts, and these norms reliably shape actions both because people are motivated to behave in accordance with these guidelines as members of the group sharing these norms and because other members of the group will enforce these norms when an individual’s behavior deviates from these guidelines.

We acknowledge, though, that there are often conflicting norms in a particular situation. For instance, norms established within particular police agencies or departments for other officers to intervene to prevent unjustified or excessive use of force may be at odds with traditional norms for policing as an institution that emphasizes strong loyalty and mutual support among line officers (Silver et al. 2017). Formal policy changes may therefore be needed to achieve the desired effect (Subramanian and Arzy 2021). For instance, since the murder of George Floyd in 2020 by officer Derek Chauvin while three officers watched or assisted, twelve states and the District of Columbia have instituted laws declaring that police officers are responsible for intervening when they observe cases of excessive or illegal force or misconduct. Penalties for not abiding by these policies range from decertification of officers to criminal liability. Even in the absence of state-level action, agencies and departments are establishing new policies. In the past three years, twenty-one of the nation’s one hundred largest police agencies have adopted “duty to intervene” policies (Subramanian and Arzy 2021).

Intervention Assessment

A multilevel approach offers a more valid cost-benefit analysis for assessing impact of interventions. For example, whereas the impact of increased police stop and frisk activity is frequently assessed in terms of crime rates in a particular location—one goal of such initiatives—the impacts of such actions (for example, on the mental and physical health and perceptions of police legitimacy of community members) are also individual-level outcomes relevant to the longer term and broader effectiveness of policing. The limited impact of stop and frisk interventions for reducing crime may be substantially outweighed by the costs of community health and erosion of perceptions of police integrity (see Petersen et al. 2023). Moreover, if researchers use a broader lens when considering the impact of police interventions, they can stimulate new, important issues to be considered empirically and in application. For instance, Braga and colleagues (2019, 545) observe: “The existing body of scientific evidence on proactive policing programs generally does not consider how these crime reduction and community perception impacts might vary with the racial composition of communities, the police department, and police leadership.” Understanding how various factors that have been typically studied within a particular level of analysis operate jointly to impact intervention effectiveness is essential for developing and implementing new strategies for reducing racial discrimination while improving the effectiveness of policing.

A multilevel approach to understanding and addressing racial discrimination in policing involves increasing equity and improving the effectiveness of policing as presently conceived. The model of policing in the United States today is anchored in policing tradition and in its current structure. Therefore, even with a more comprehensive understanding of the factors contributing to racial disparities in policing, change may essentially be only incremental. Institutions and organizations, particularly when conditions of threat are involved, tend to resist change through both informal and formal means (for example, through system-justifying ideologies; Jost 2020). New models of policing may be needed to address the challenges of policing in society now and in the future.

Tailored Responsibilities

Tom Tyler (2021; see also Kramer and Remster 2022) observes that, traditionally, the focus of policing in the United States has been on controlling reported crime through force and heightened risk of being caught and punished. A major consequence of this approach is that police officers are primarily “trained in one skill set. They learn how to deploy force to compel compliance” (Tyler 2021, 1393). One proposed way to make policing more equitable is to have resources be reallocated to social services and allow police officers to focus more narrowly on controlling serious crime. However, this work acknowledges questions about the feasibility of that approach given the resources needed to create an adequate social service infrastructure to relieve police officers of these responsibilities.

Legitimacy-Based Policing

Tyler (2021, 1398) also proposes an alternative approach: that policing be more fundamentally reimagined around the “question of how people would like the police to deal with social order in their community.” Central to this process would be a shift in policing from a force-based style to gain compliance to a style oriented toward building trust and confidence in the community by establishing legitimacy through procedural justice practices (Chan et al. 2023). When people view the authority of police as legitimate, they more willingly accept and follow police officers’ directives and abide by the laws they enforce. This model of policing promotes the development of police-community partnerships to prevent crime while addressing structural factors that foster crime by establishing “a climate of reassurance within communities that promotes their social, economic, and political vitality” (Tyler 2021, 1388). However, such efforts to improve the legitimacy of policing in the eyes of Black people (and members of other minoritized groups) need to acknowledge that “policing has been seen as racial terrorism in too many communities for too long” (Goff 2021, 678) and that helping communities heal from the harms of policing (see Prowse and Goff 2023) is a key step in achieving the partnerships and trust Tyler envisions.

Asking the Right Questions About Race and Policing

While Tyler’s recommendation for reconceiving policing promises more effective and equitable policing, we offer another perspective. Rather than framing the problem as one of how to change bad police officers into good ones or how to make policing that is limited in effectiveness more effective, it is important to recognize that policing itself is often the problem. Consequently, reforming it is not necessarily the solution. We do not necessarily advocate for the abolition of policing but rather urge a major reimagining of public safety—one in which policing plays a different, less intrusive, and more appropriate role in the lives of citizens.

Gwen Prowse and Phillip Atiba Goff (2023, 528) ask, “Why do police respond to mental health crises rather than healthcare workers? Why does an armed first responder issue traffic violations? What are the consequences of those policy choices?” These questions are basic ones to consider when evaluating policing today. Their answers suggest important ways that policing might be reimagined. For instance, police officers are currently the most common first responders to mental health crises. In some cases, this seems unavoidable, as mental health crises can involve weapons and violence. But in many—perhaps the majority—of cases, behavioral health professionals are more skilled and thus would be more effective at intervening to protect public safety. Also, the argument behind police giving speeding tickets is that unsafe behavior on the roads is a public safety hazard. Why, then, are some jurisdictions able to mail tickets to speeding motorists? And what evidence is there that only police keep our streets safer?

Two traditional elements of policing, crisis intervention (which includes responding to mental health crises) and low-level traffic violations enforcement, have been the focus of interventions for reconceiving policing. Several jurisdictions, including ones in cities like St. Louis and Oakland, have instituted policies that include a range of non-police crisis responders. In addition, the CAHOOTS (Crisis Assistance Helping Out on the Streets) model is growing in popularity across the United States. CAHOOTS is a mobile crisis intervention program that, in coordination with police departments, sends teams of professionals to respond to calls indicating a need for services related to mental health. The program thus offers an alternative to police officers intervening in noncriminal situations by using appropriately trained personnel. Also, an increasing number of jurisdictions are asking these questions and seeking to limit—or eliminate—police from low-level traffic enforcement (for example, for broken taillights or expired registrations). For example, several major cities (for example, Philadelphia and parts of Los Angeles) have recently ended low-level traffic enforcement by police, and several states (including California, Connecticut, and Washington are currently considering such legislation.

The intended impacts of these new programs are broad. They reduce the demand on police officers, relieving stress and allowing more efficient and effective deployment of personnel tailored to their professional training. These programs also reduce exposure to the kinds of complex situations that can lead unnecessarily to inappropriate police behavior (such as excessive use of force) or to racial and ethnic discrimination in policing. The intent of legislation that limits police involvement in low-level traffic enforcement in Connecticut, for example, is to reduce racial and ethnic inequities relating to these policing incidents and the escalation of such stops into these incidents (Chacko 2024). However, we are not yet aware of definitive findings concerning the effectiveness of these changes related to supporting policing with additional mental health services (for example, the CAHOOTS program) or relieving police from pursuing low-level traffic violations.

A major consequence of using “police” as the answer to questions about who should respond to a broad range of public safety situations, including those that police officers are not fully equipped to address, is that an intrusive police presence becomes a problem in itself. The core question guiding the future of policing from this perspective thus becomes, “When is policing the right tool, and when is it the problem?” (Goff 2021, 677). If the rules of policing have always unfairly produced group-based disparities that reinforce social hierarchies, and individual officers are not free of individual risk factors—in fact, they are worse for officers (Pryor et al. 2020; Swencionis and Goff 2017)—and few, if any, trainings are effective in combating individual biases, are there any conditions where police are the best response we can imagine for communities in crises? While this approach inevitably leads to a more limited role for police in society—when it is the right and best tool—the objective is not to punish police as an institution but rather to prepare police officers, who are currently overburdened with multiple demands, to be better equipped to meet the challenges facing officers and society, to serve equitably, and to prioritize the skills that police officers need most. Moreover, considering the multilevel impacts of policing in its current form, the answer to the question of whether policing is the right tool requires full consideration of the question, for whom?

While this may seem a radical framing for some, the question of where to solve problems with policing is, again, as old as the profession, and concerns about the likely problems with using an armed force domestically to regulate public safety date back to the first Western police institutions. In other words, questioning whether to use police is no older nor less fundamental to the project of formal policing than the question of how best to prevent their excesses. And perhaps returning to these formational questions will help provide answers that have been missing for the past two centuries of less-than-stellar results of the experiment of providing organized policing without unjust outcomes for those vulnerable communities whose safety is most precarious. More importantly, it forces us to ask the question: if two hundred years have left us without justice in this first stage of the criminal legal system, was it ever designed to do otherwise? We ask this question in good faith, understanding that a failure to ask it may seem to belittle the injustices shouldered on the backs of Black and Brown communities over the past two centuries.