Systemic Racial Bias in the Criminal Justice System Is Not a Myth

Many on both the left and the right tend to speak of systemic racism simply as a 0/1 state: either the system is fundamentally and inextricably racist or it is not racist at all. But recognizing distinct mechanisms at play in a racialized system should help us see systemic racial bias as a matter of degrees—as something that can improve or worsen over time. Indeed, research suggests that racial disparities have been declining over time, though there is no guarantee of inevitable progress, and our present situation makes it clear that we still have a long way to go.

One response of some conservatives to the nationwide protests against police brutality and systemic racism has been to reject the existence of systemic racism altogether. A widely cited Wall Street Journal article by Heather Mac Donald, for instance, claims that “A solid body of evidence finds no structural bias in the criminal-justice system with regard to arrests, prosecution or sentencing.”

Such claims, as I show in this essay, rest on cherry-picked studies that do not withstand scrutiny. They also misunderstand what systemic racism means.


It’s easy to refute the assertion that there is no systemic racial bias at all. Considerable evidence points to some degree of racial bias across nearly every aspect of the criminal justice system. Consider just a few examples: compared to whites, blacks are significantly more likely to be searched after traffic stops but less likely to turn up contraband or weapons; not to be told why they were pulled over; to be ticketed for jaywalking; to be treated with less respect by officers; to be arrested for low-level offenses; to be wrongfully convicted of murder, sexual assault, and drug crimes; to be offered plea bargains that include jail time; to serve life without parole for nonviolent offenses; to receive longer sentences for the same crime; to be incarcerated while awaiting trial; to have probation revoked; to be assigned higher bail amounts for the same crimes; and so on. Consider, in addition, that blacks are significantly underrepresented among prosecutors (95 percent of elected prosecutors and 85 percent of lawyers are white); more than a third of states have never seated a black state supreme court justice; and black judges have a higher likelihood of being overruled than their white counterparts. This is just a fraction of the research on the topic (see more than 300 research studies that are listed here and here, for example).

Given the history of racism in America, these shouldn’t be seen as mere disparities. Disparity by itself doesn’t imply bias, but these differences aren’t the outgrowth of a system that originated on some level playing field. Rather, our criminal justice system from the beginning has produced and sustained disadvantages for African Americans. This bias is systemic in the sense that it is perpetuated primarily by structures such as policies, codes, norms, laws, and organizations across the criminal justice system. Systemic racism means you can have “racism without racists.”

In what follows, I first clarify some confusions surrounding the term “systemic racism.” Next, returning to the Wall Street Journal article by Heather Mac Donald that I mentioned at the beginning, which focuses on fatal police officer-involved shootings, I show that the key claims on which it rests (besides botched calculations) are not empirically supported: (1) the studies it cites as evidence against systemic racial bias do not withstand scrutiny; and (2) the claim that blacks’ higher crime rates explain racial disparities is not supported either. I conclude by considering how systemic biases in police shootings can be produced by other mechanisms besides racist intent.

Three Mechanisms That Create Systemic Racial Disparities

In spite of its rhetorical power, “systemic racism” may not be the most helpful term. This isn’t simply because of the defensiveness provoked by the pejorative connotations of “racist.” It’s rather because the term conflates at least three distinct types of mechanisms, all of which work together to contribute to racial disparities as a system-level output. Before turning to the data, then, we should explore these mechanisms.


Type 1: First there’s racism as we generally understand it—prejudice and discrimination based on ideas of racial superiority and inferiority. Racist ideologies, originating as justifications for the dehumanization and exploitation of blacks, shaped laws and decision-makers in our criminal justice system for centuries. While the Civil Rights Act struck down most (but not all) overtly racist laws, the racists in the system didn’t disappear overnight. But racial attitudes have certainly changed over time, and explicit racism may play less of a role today in sustaining systemic racial bias. Subtle or implicit racial bias, in the form of heuristics, tastes, preferences, and negative stereotypes, may be more relevant, though evidence on whether implicit biases can be validly measured or “treated” is mixed (e.g., see criticisms here; but see here for a more adequate understanding of implicit biases, and here, here, and here on their relevance to law enforcement). When nationally representative data show that simply the darker one’s skin tone, the higher one’s likelihood of arrest and incarceration (controlling for crime rate, socioeconomic status, and so on), Type-1 mechanisms are likely at work.

Type 2: Systemic disadvantages to blacks also result from laws, policies, norms, and institutions within the criminal justice system that may not be intrinsically racist. Type-1 racist intentions motivated the design and implementation of some of these laws and policies. For instance, historical research (e.g., see here and here) shows how felon disenfranchisement laws, poll taxes, changing classifications of crimes, and other measures were introduced strategically to suppress blacks. Policies such as the “War on Crime” and “War on Drugs” also shaped racial disparities through their effects on policing and mass incarceration. Other organizational and institutional factors, such as cultures of police departments, arrest quotas, and biases inadvertently built into policing algorithms also play a role in sustaining systemic racial disparities.

Type 3: Structural and cultural factors outside the criminal justice system may contribute to racial disparities. Racist policies of the past have enduring and measurable effects on many of today’s most disadvantaged neighborhoods, shaping the degree of social cohesion and violent crime in these communities. Potentially active mechanisms here include “cultural responses to chronic economic and racial subordination;” the “culture of poverty” thesis; effects of systemic injustice on the black family; the stigmatization of “blackness;” a range of institutional deficits and assets in urban black communities; indifference and apathy toward racial minorities; racial disparities in healthcare, education, employment, and so on.

All three types of causal mechanisms operate simultaneously. While the relative influence of these different mechanisms may change over time, African Americans experience these effects as the legacy of the same racist system that has endured over generations.

A weakness of the term “systemic racism,” however, is that it conflates the three types of mechanisms, making “racism” both the explanans and explanandum: it becomes both the explanation and the thing being explained. Such conflation also leads to distortions on both sides of the political spectrum, generating more heat than light.

For these reasons, it might help to use a more analytical (rather than pejorative) concept, such as a “racialized system” to refer to a complex of institutions that produces systemic racial biases and disadvantages. Doing so can avoid defensiveness and confusion around the term “racism,” while preserving the core reality of enduring patterns of racial disadvantage as the systemic outcome. Recognizing the range of mechanisms at work might also help us avoid the pernicious problem of confirmation bias—cherry-picking only those pieces of evidence that reinforce our ideological priors.

No Racial Bias in Police Shootings?

Turning now to the data: Mac Donald cites two key studies as the most damning evidence against systemic bias in policing. The first is a study by Harvard economist Roland G. Fryer Jr. She claims that the authors found no racial disparities in police shootings. But she blatantly neglects the primary finding of the article: blacks are more than fifty percent more likely than whites to experience non-lethal force in interactions with police—effects that persist in spite of statistical controls. “Even when officers report civilians have been compliant and no arrest was made,” Fryer notes, “blacks are 21.2 percent more likely to endure some form of force in an interaction.”

Fryer’s claims about fatal shootings are couched in a number of caveats. First, he clearly states that these data are not nationally representative. Instead, they come from ten police departments that were willing to provide him with data. As he puts it, “In essence, this is equivalent to analyzing labor market discrimination on a set of firms willing to supply researcher with their Human Resources data!” Fryer cautions that these data may also contain biases in reporting, since officers “may have an incentive to distort the truth.” While Fryer insists that his data show no racial bias when it comes to fatal shootings, scholars have pointed out problems with Fryer’s analysis (see here and here), which render his approach incapable of demonstrating the absence of racial bias.

More problems plague the second major study on police shootings that Mac Donald cites. This article by Johnson et al. claims to have found “no significant evidence of anti-black disparity in the likelihood of being fatally shot by police.” But this isn’t true. The article only examines victims who are fatally shot, not non-fatalities. It can’t make any claims about if and when an officer will engage in a fatal shooting in the first place.

Two response papers to this article in the same journal identified serious flaws, and an investigation by the journal’s editorial board affirmed the article’s critics. One response re-ran its analysis and found that “Young unarmed nonsuicidal male victims of fatal use of force are 13 times more likely to be black than white.” Another found that the original article’s approach was “mathematically incapable of supporting its central claims.” In response to critics, the authors of the original study issued a correction, admitting that they could no longer justify their initial significance statement.

Aren’t Blacks Killed by the Police at Higher Rates Because They Commit More Crimes?

Mac Donald claims that the higher police fatality rate of blacks is explained by the higher rate at which they commit crimes—particularly homicides and robberies. But most black fatalities in police shootings are not homicide or robbery suspects. Rather, the data show that “a substantial number of fatal shootings each year result from interactions that frequently do not end in arrests, such as traffic stops, domestic disturbances, suspicious person stops, and officers responding to 911 calls about mentally ill persons.”

Indeed, research on fatalities of police shootings from 2009-2012 in seventeen states finds that in 22 percent of all cases, the contact with law enforcement was initiated due to mental health problems, and an additional 17 percent of all cases were “suicides by cop” (i.e., the victims were suicidal and were trying to provoke the police to kill them). Mental illness features significantly among victims of police shootings. And no firearm was present in 44 percent of fatal shootings.

If crime rates were indeed the key explanatory factor, statistical controls for them should eliminate racial disparities in the data. But that’s not the case. Research analyzing data from the US Police Shooting Database found “no relationship between county-level racial bias in police shootings and crime rates (even race-specific crime rates), meaning that the racial bias observed in police shootings in this data set is not explainable as a response to local-level crime rates.” Still other data from the fifty largest US cities finds no correlation between city-level violent crime rates and police killings. Another study, examining police fatal shootings in 2015, found that net of controls, blacks were more likely than whites to have not been attacking officers or civilians, and more than twice as likely as whites to have been unarmed (see also here and here). By contrast, another study, using different crime rate estimates, found that their measures of crime rates could statistically explain racial disparities in all but one category: blacks were significantly more likely than whites to be killed when “holding/reaching for a harmless object.”

What the Data Can and Cannot Tell Us

Contrary to what Mac Donald claims, “crime rates and civilian behavior before and during interactions with police” do not sufficiently explain the racial disparities in shootings. But there are limitations in what existing research on police shootings can tell us about racial bias. First, the data simply aren’t good enough. There is no comprehensive national database on police shootings, and existing data are either not nationally representative or don’t record all the relevant variables.

Second, as Lyman Stone argued last week here at Public Discourse, while it’s clear that the data on police killings clearly demonstrate the existence of racial disparities, we don’t have a good way of quantifying the extent of bias. It is often cited that blacks are three times more likely to be killed by the police than whites even though they make up 13 percent of the population. But the national population isn’t the best benchmark, since not everyone in the country is equally at risk of encountering the police. Yet there is no consensus as to what benchmark would be most appropriate (e.g., see here, here, and here).

Perhaps the most challenging methodological issue is causal inference. None of the data on police shootings can tell us whether racial bias is a motivating factor. Some argue that even better datasets won’t be able to ascertain whether racial disparities reflect Type-1 racial biases on the part of officers.

Finally, police shootings themselves may be driven by prior systemic biases that make African Americans more likely to encounter the police. As researchers have argued: “If even a small subset of police more frequently encounter and use non-lethal force against black individuals than white individuals, then analyses of pooled encounter-conditional data can fail to correctly detect racial disparities in the use of lethal force.”

How Police-Citizen Encounters May Go Awry

Let’s consider a different aspect of policing to examine the likelihood of police encounters in the first place—traffic stops. Here too, there are systemic disparities. A 2020 article in Nature analyzed nearly 100 million traffic stops and found that blacks were more likely to be stopped in the daytime than at night, suggesting bias in stop decisions, and that the bar for searching blacks was significantly lower than for whites. Similarly, a Nashville study found that black drivers were stopped 44 percent more often per driving age resident than white drivers, mostly for high-discretion non-moving violations related to minor infractions (e.g., broken taillights), and only 1.6 percent of stops resulted in custodial arrest. A study of 20 million traffic stops in North Carolina yielded similar findings. Such tactics are ineffective for reducing crime, and subject blacks to a disproportionately higher likelihood of police encounters.

A number of studies suggest that in segregated black neighborhoods to which the police are frequently deployed, the situation may already be set up for potentially negative interactions in traffic stops—especially if officers are white. Officers are statistically more likely to conduct searches when the suspect is of a different race than the officer. During traffic stops, officers also interact more positively with members of the same race than with members of a different race. Drivers pulled over by officers of the same race are more likely to believe the stop was legitimate. Evidence also suggests that disadvantaged black neighborhoods are both overpoliced and underserved—residents feel harassed for minor offenses while major crimes such as murders go unsolved. Research also finds that police officers expect segregated black neighborhoods to be particularly dangerous, perceive a sense of threat to their personal safety, and fear less punishment for excessive use of force. Fear, or “racial anxiety,” rather than racism as such, creates a tense situation from the get-go. This might explain why white officers are more likely to use guns than are black officers, and in predominantly black neighborhoods. Fatal police shootings also weaken a community’s trust in and perceived legitimacy of the police.

Many who resist the idea of systemic racism point to black “ghetto culture” as the key problem, but neglect how “police culture” may contribute as well. Many police officers have told me that the norm in their departments is to put recruits through paramilitary-style training that keeps them on constant vigil, anticipating that they’ll need to kill or be killed. Officers are trained more for combat situations (a tiny fraction of their job) than for what the majority of their day entails (routine civilian encounters). Plus, nearly 20 percent of officers have a military background (and data suggest this subset is more likely to shoot). Nearly the same proportion may also suffer from PTSD. Studies also indicate high rates of burnoutsubstance abuse, and suicidality among police officers.

Under these conditions, you don’t even need pre-existing Type-1 biases to create the likelihood of violent confrontations with minorities. Greater police deployment to disadvantaged neighborhoods increases the likelihood of regular citizens in those communities encountering police who approach them primarily as potential suspects and criminals. The frequency of such interactions will be experienced by law-abiding citizens there as harassment and will increase mutual mistrust, perceptions of disrespect and de-humanization, and escalation of conflict in police interactions. Even officers with no prior racist inclinations, merely by regular exposure to this environment, can develop heightened anxieties and apprehensions when encountering minority civilians. The regularity with which such conditions obtain is enough to produce encounters that may lead to systemic racial biases in police killings of civilians. Such fatal police shootings might thus be seen as “system crashes.”


Still, a host of other Type-1 and Type-2 mechanisms are likely involved, on which we need much more research. These include the extent of unpunished racist conduct among officers; prevalence rates of white supremacists in police departments; racism experienced by African American police officers; the effects of arrest quotas; criminal behavior of officers; and the role of police unions, qualified immunity, and the “blue wall of silence.” Reforming the structure and culture of policing will certainly require better training and greater accountability and transparency to address such problems, but should also consider measures such as reduced workloads and increased mental health resources to address the challenges noted earlier.

Systemic Racism Isn’t All-or-Nothing

I have proposed a way to understand the complex reality to which the concept of systemic racism refers. Disentangling “racism” as one aspect of a racialized system, and analytically separating the three types of causal mechanisms I suggested (and perhaps more that I’ve missed), can help us grasp the structural and multi-faceted nature of the problem. It can also help us work toward assessing the relative effects of various mechanisms in producing the systemic disparities we observe.

This approach can also help address another aspect of the problem. Many on both the left and the right tend to speak of systemic racism simply as a 0/1 state: either the system is fundamentally and inextricably racist or it is not racist at all. But recognizing distinct mechanisms at play in a racialized system should help us see systemic racial bias as a matter of degrees—as something that can improve or worsen over time, likely unevenly across the various aspects and contexts of the system. Indeed, research suggests that racial disparities have been declining over time, but there is no guarantee of inevitable progress, and our present situation makes it clear that we still have a long way to go.


We need better concepts, measures, and evidence-based strategies (which are still unfortunately scarce) that can help us effectively reduce systemic racial disparities. It does no good to waste our time arguing about whether or not we’re racists.

Keep up with the conversation! Subscribe to Public Discourse today.

Subscribe to Public Discourse!