One way to test for an anti-black bias in police killings is to test if officer race is associated with lethal force after controlling for possible covariates. Johnson et al. (2019) [archived] studied officer characteristics and racial disparities in fatal officer-involved shootings. Their analysis differs from many other measures of racial disparity in fatal police shootings because, instead of benchmarking fatal police shootings against a chosen reference (e.g., crime rates), researchers instead examined whether certain factors (officer, civilian, and county characteristics) predicted the race of a citizen fatally shot by the police. The significance statement of the study reports that the data shows “no evidence of anti-black or anti-Hispanic disparities across shootings, and white officers are not more likely to shoot minority civilians than non-white officers” (page 15877–15882). However, as Knox and Mummolo (2020) have correctly explained, this sentence is misleading. To address this, Johnson et al. released the following correction to correct this sentence:

Goff et al. (2016) [archived] investigated racial disparities in use of force across 12 departments participating in the National Justice Database. Use of force was benchmarked against demographics of local arrest rates. All data were collected between 2010 and 2015. While the study investigate various categories of use of force by police (i.e. lethal, less lethal or Taser, canine, OC spray, weapon, and hands or body), lethal force by police is the only relevant measure for my post. When controlling for arrests for all offenses, the results demonstrated that black citizens suffered lower rates of lethal force compared to white citizens. In particular, black citizens had lower mean rates of lethal force when benchmarked against total arrests (0.37 vs 0.64 instances of lethal force per 1,000 arrests, Table 6) and when benchmarked against arrests for violent offenses (6.92 vs 16.7 instances of lethal force per 1,000 arrests, Table 7).

Miller et al. (2017) [archived] reviewed injuries resulting from law enforcement during arrests or police stops in 2012. Researchers found that, while blacks are more likely to be stopped or arrested by the police, they are no more likely to be injured or die conditional upon such an incident. In other words, blacks have higher rates of death at the hands of the police, this is due to greater exposure to police whether than anti-black bias in police decision to initiate lethal force:

On average, an estimated 34 people were killed or medically treated for injury by law enforcement per 10 000 stops/arrests. That ratio is surprisingly consistent by race/ethnicity. Blacks have high arrest and stop rates, and per capita are much more likely than whites to die at the hands of police. However, when blacks are stopped or arrested, they are no more likely than whites to be injured or die during that incident.

Note that this study did not directly investigate the cause of racial disparities in arrests or police stops. Keep in mind that this study provides evidence against anti-black police bias at just one of the many stages of police decision-making that can produce anti-black disparities. For example, there may be anti-black bias at the stage of deciding whether to initiate an arrest or stop despite there being no anti-black bias at the stage of deciding whether to initiate lethal force. It is possible that an anti-black bias at the first stage combined with no anti-black bias at the second stage produces an overall anti-black disparity in police killings. While this is a possible explanation, one plausible explanation for the racial disparities at the first stage (i.e. racial disparities in arrests/stops) is that racial disparities in arrests/stops are largely driven by racial disparities in criminal activity. Regardless, this study is still valuable in showing no anti-black bias at the stage of initiating lethal force.

Tregle, Nix and Alpert (2018) [archived] analyzed fatal shootings by police of black and white citizens from 2015 to 2017. The authors examine the odds of being fatally shot by the police for black versus white citizens against 7 different benchmarks: population sizes, average number of traffic stops, average number of street stops, average number of police-initiated contacts (combination of traffic stops and street stops), total arrests, arrests for violent crime, and arrests for weapons offenses. The results did not vary considerably by year (Table 1). As expected, when benchmarked against population size, blacks are roughly 3 times as likely to be killed by the police as whites (Table 1). The problem with this benchmark, the study notes, is that “most of the population (white or black) was not exposed to the risk of being fatally shot during this time; therefore, these people should be excluded from any analysis comparing racial differences in fatal OIS” (page 14).

When benchmarked against police-initiated contacts, traffic stops, or street stops, blacks are 2-3 times as likely as whites to be fatally shot by police officers (Table 1). The authors note that one problem with this benchmark is that “it is a conservative estimate of potentially lethal interactions, since so many interactions are for minor offenses that typically end peacefully” (page 14). When benchmarked against total arrests, the black-white disparity in fatal police killings was reduced, but still statistically significant: depending on the year, black citizens were between 1.23 and 1.37 times as likely as white citizens to be fatally shot by the police. While arrests is a better benchmark than previous benchmarks, the authors state that “many arrests each year are for minor offenses (e.g., liquor laws, curfew), wherein the probability of the encounter turning deadly is low” (page 16). Finally, when benchmarked against arrests for violent criminal crime and weapons offenses, “black citizens were slightly less likely to be fatally shot by police from 2015 to 2017” (page 17). Black citizens were between 0.6 and 0.86 times as likely to be fatally shot by the police, depending on the year and depending on the type of crime used in the benchmark (Table 1).

Cesario et al. (2019) [archived] analyzed officer-involved fatal shootings in 2015–2016 to determine if there was any evidence of anti-black disparities. The authors analyzed whether blacks or whites are more likely to be fatally shot by the police when benchmarked against 16 different measures of criminal activity, for 3 different types of police shootings, and for 3 different time periods. In total, the authors performed 144 (16 x 3 x 3) tests in their analysis. The 16 measures of criminal activity varied depending on the specific crime measured (i.e. murder/non-negligent manslaughter, violent crime, or weapons violations) and the source of data (data was pulled from SRS, the NIBRS, the NCVS, and the CDC). They 3 types of police shootings were (1) all shootings, (2) shootings of unarmed and non-aggressing citizens, and (3) shootings of unarmed citizens reaching for or holding an object. The 3 time periods were (1) 2015, (2) 2016, and (3) 2015-2016 (combined). The authors note the following conclusions from analyzing these tests:

Three conclusions are apparent from these analyses. First and most important, across all three types of shootings, only 1 of the 144 possible tests (0.7%) showed statistically significant anti-Black disparity. Second, the analysis shows the robustness of the conclusions to sampling time frame, as the year chosen makes almost no difference. Third, the small sample sizes in the unarmed and misidentification shootings confirm that there is not enough data to make definitive statements about subsets of police shootings beyond all fatal shootings.

The 1 test (of 144 total tests) that found a statistically significant anti-black disparity was the test of all shootings benchmarked against the NCVS’s measure of “less severe” violent crime for the year 2016. The other 143 tests found no statistically significant anti-black disparity. In fact, 61 of the 144 tests (42%) of the tests found statistically significant anti-white disparities (see Figure 4).

One limitation of some of the earlier studies is that they use national level data, which means they aggregate data across lower levels of analysis (e.g., neighborhood-, county-, agency-level, etc.), which can introduce bias and mask critical information. Mentch (2020) overcomes these limitations by analyzing racial disparities in fatal police shootings using county-level data racial demographics, arrest records, and law enforcement density. Mentch found that the race of those fatally shot by the police “are more or less in line with what would be expected when such victims are considered as a random sample from local arrestee populations.” Interestingly, the study also found that “the racial distribution of shooting victims appeared to be the same regardless of whether the police officers involved were wearing body cameras.”

Shjarbacka and Nix (2020) analyzed national racial disparities in fatal police killings benchmarked against violence against police. Violence against police was measured using the FBI’s Law Enforcement Officers Killed and Assaulted (LEOKA) reports. Violence against police included both felonious homicides and nonfatal assaults with injury in which a firearm or a knife/other cutting instrument was used as a weapon. So authors used two benchmarks, one for each measure of violence against police. The authors find that “From 2015 to 2017, black citizens were less likely to be fatally shot by US police (odds ratios < 1.0) than white citizens based on both benchmarks used” (page 6). Benchmarked against felonious homicides, blacks were about 0.55 times as likely as whites to be fatally shot by a police officer (Table 2). Benchmarked against non-fatal assaults, blacks were about 0.6 times as likely as whites to be fatally shot. Keep in mind that the authors do note that “caution should be exercised while interpreting the odds ratios due to the small base rates of fatal and serious transactional violence between citizens and the police” (page 6).

One study, Ross, Winterhalder, and McElreath (2020), challenge the findings of Cesario et al. (2019) by performing a re-analysis of their dataset. Ross et al. purport to show some evidence of anti-Black racial disparities among police killings in 2015–2016 with the same dataset. While this study purports to challenge the findings of Cesario et al, it is important to emphasize that Ross et al. purports to find evidence of anti-black disparities only among police killings of unarmed and non-aggressing civilians. Regarding other killings, the study’s findings are consistent with the findings of Cesario et al: “we recover a principle finding of Cesario et al. (2019): Racial disparities in the killing of armed suspects by police are proportional to the relative rates of violent criminality” (page 327). This point is important because, as Cesario (2020) notes, armed or aggressing civilians constitute 90-95% of all fatal police shootings (page 2), leading Cesario to conclude (page 2):

Thus the main difference in results between Cesario et al. (2019) and Ross et al. (in press) is that, for killings of unarmed, non-aggressive citizens (again, roughly 5-10% of shootings), we expressed uncertainty and did not take a strong position whereas Ross et al. conclude anti-Black disparity.

As for the similarities, Ross et al. (in press) confirm that there is no anti-Black disparity in shootings of armed or aggressing citizens (the far majority shootings) once violent crime rates are considered. Indeed, they confirm not just a lack of anti-Black disparity but reveal an anti-White disparity in most of their calculations (see their Fig. 2). That is, the most certain results in all of Ross et al.’s analyses confirm that there is no anti-Black disparity in fatal police shootings of armed or aggressing citizens.

Also, see Cesario (2020) for other critiques of Ross et al. One of the primary criticisms with the Ross et al. study, Cesario explains, is that their analysis “fails to account for the fact that among nonviolent citizens, encountering police is still a prerequisite for being fatally shot” (page 2). Cesario continues:

If encounter rates differ between Blacks and Whites for the nonviolent population, then the estimates produced by Ross et al. will not be unbiased. Indeed, if encounter rates are higher for Black nonviolent citizens, then the analyses will overestimate anti-Black disparity in unarmed shootings by biasing the size of the nonviolent citizen population to appear smaller for the Black population than it actually is.

Even ignoring this criticism, the findings from the Ross et al. study are limited because they are restricted to unarmed and non-aggressing citizens. As Cesario (2020) points out, this only accounts for about 5-10% of total police shootings and about 20 Black citizens and 30 White citizens per year (page 2).

Let’s review the findings of this section. First, we know that black people are extremely disproportionately likely to engage in criminality activity (particularly for the most serious crimes such as violent crimes and homicides), to a degree that is even more disproportionate than their rate of deaths by police. Second, studies show that the number of police killings in an area is influenced by the crime rate of the area. Finally, studies suggest that the disproportionate rate of police killings of black people is the result of disproportionate exposure to dangerous police interactions, which can be explained by disproportionate black violent criminality. Thus, the evidence seems to suggest that the disproportionate rate of police killings of blacks is due to disproportionate black criminality, not due to anti-black bias in the criminal justice system.

Black criminality is not caused by anti-black bias in the criminal justice system

In the remainder of this post, I will argue that disproportionate black criminality is not caused by an anti-black bias within the CJS. The argument is based on the following pieces of evidence. First, I criticize a few of the most common alleged sources of anti-black bias within the CJS. Second, I cite low levels of incarceration among black legal immigrants. Lastly, I cite instances of similar relative levels of misconduct among blacks who have not engaged with the CJS. This includes misconduct among black children that appears as early as preschool, and similar relative levels of criminality among blacks in other countries such as Canada and the United Kingdom. This demonstrates that there are factors independent of the U.S. CJS that may cause the relative levels of black criminality in the United States.

The evidence of anti-black bias in the criminal justice system is weak

Here I would like to refute what I believe are the two most commonly alleged sources of anti-black bias in the CJS. The first alleged source is the fact that black suspects have higher arrest, conviction, and incarceration rates and longer sentences than whites suspects of similar backgrounds. The second alleged source is the fact that blacks are incarcerated for drug-related crimes more often than whites despite the fact that the two groups self-report that they use and sell drugs at comparable rates.

First, I would like to address the claim that black suspects, compared to white suspects suspected of the same crime, are arrested, convicted, and incarcerated at higher rates, and are given longer sentences. These disparities do exist. However, an anti-black bias is not necessarily the best explanation for the disparities. Consider the following:

• Beaver et al. 2013 [archived] found that these disparities were “completely accounted for after including covariates for self-reported lifetime violence and IQ.” The abstract does not mention the disparities in sentencing, but the full study notes that “race was not significantly associated with sentence length in the baseline model. After controlling for the effects of self-reported lifetime violence and verbal IQ, the effect of race on sentence length remained non-significant” (page 31).
• Moreover, even if there is an anti-black bias in the criminal justice system, it is unlikely that this could explain much of the crime disparity between blacks and whites. For example, consider sentence length. Rehavi and Starr 2014 [archived] estimate that blacks face a 9% premium compared to whites with regard to the lengths of federal criminal sentencing (page 1349). The study notes that it cannot rule out the possibility that this gap is due to unobserved characteristics of the defendant that influence prosecutorial choices (page 1348) rather than discrimination. However, even if this 9% premium is completely the result of discrimination, this does not seem large enough to play a substantive role in explaining the high black crime rates. That is, it doesn’t seem that the 600-700% higher black murder rate (compared to whites) is significantly caused by a 9% higher sentence length imposed on blacks. In general, the black-white disparities in treatment by the CJS are often negligible compared to the black-white disparities in crime.

Now, I will address the claim that blacks are disproportionately arrested for drug-related crimes despite using and selling drugs at the same rates as whites. Before doing so, I must note that the vast majority of blacks are not imprisoned for drug-related offenses. A report on prisoners in 2018 from the Bureau of Justice Statistics [archived] shows the following:

• The vast majority of black state prisoners are not imprisoned for drug-related crimes. The majority (61%) are imprisoned for violent crimes. Only 13% of black prisoners are imprisoned for drug-related offenses (including e.g. trafficking), and only 3.2% are imprisoned for drug possession. For comparison, 4.5% of white prisoners are imprisoned for drug possession (Table 13). Note that state prisons contain the vast majority (88%) of all prisoners (page 3).
• About half (45%) of black federal prisoners are imprisoned for drug-related crimes (Table 15) (federal prisoners constitute only 12% of all prisoners). But almost none of them were imprisoned for mere drug possession; 99% of federal drug offenders were sentenced for trafficking (see note g of Table 15).

Now, back to the claim that blacks are disproportionately imprisoned for drug-related offenses given their self-reported rates of drug usage and drug sales. There are explanations for this disparity that do not involve appealing to an anti-black bias. For example, blacks are more likely to sell and do drugs in manners that increase the probability of an arrest:

• A 1995 report [archived] by the Bureau of Justice Statistics showed that “Although blacks are 13% of drug users, they should comprise over 13% of drug possession arrests since the types of drugs they use, the frequency with which they use them, and the places where they use them, put blacks at greater risk of arrest.”
• Ramchand et al. 2006 [archived] showed that black people are more likely to engage in risky purchasing behaviors that increase their likelihood of arrest. The study revealed that “African Americans are nearly twice as likely to buy outdoors (0.31 versus 0.14), three times more likely to buy from a stranger (0.30 versus 0.09), and significantly more likely to buy away from their homes (0.61 versus 0.48). These results provide an additional explanation for the differential in arrest rates between African Americans and whites.”
• Tonry (2008) [archived] reported that “The reason why so many more blacks than whites are arrested and imprisoned for drug crimes is well known and long recognized. They are much easier to arrest. Much white drug dealing occurs behind closed doors and in private. Much black drug dealing occurs in public or semipublic, on the streets and in open-air drug markets. And much black drug dealing occurs between strangers” (page 27).

Also, several studies have shown that blacks tend to under-report their rates of drug usage and other misconduct relative to whites:

• Kleck 1982 [archived] found that “Correlations of black juvenile’s SR delinquency scores with other external measures of delinquency were consistently smaller than was the case for white juveniles, suggesting less honest self-reports of black juveniles. The external measures included not only police records and juvenile court records, but also a measure of school misconduct and teacher ratings of the boys as discipline problems and of the likelihood they would get in trouble in the future.” (page 430).
• Fendrich and Johnson 2005 [archived] reports “This study provides evidence that compared with other groups, African Americans may provide less valid information on drug-use surveys. The findings suggest that African American respondents had significantly lower concordance rates…the results replicate and extend a growing body of research suggesting that African Americans underreport substance use on surveys” (page ii77).
• Kirk 2006 [archived] cited a study which found that blacks were more likely than whites to fail to report being arrested: “Hindelang and colleagues (1981, p. 172) similarly find that a large number respondents in their sample failed to report being picked up by the police, and also that the failure to report varies by race and gender. white males failed to report 24% of the occasions when they were picked up by police; the figure for black males is 50%; for white females, 48%; and for black females, 70%.” (page 6)
• Ledgerwood et al. 2008 [archived] found that race was the best predictor for whether a person would underreport their drug usage: “The present study also identified predictors of discrepancies between self-report and hair testing. Race was the most salient predictor of cocaine disagreement. Even when other factors were controlled for, the self-report and hair test results for African Americans were more discrepant than for non-African Americans, a finding consistent with past studies (Fendrich, et al., 1999; Feucht, Stephens, & Walker, 1994). In a large study of youth (9−20), underreporting of cocaine was documented with urine testing validation as well (Fendrich & Yanchun, 1994) where African Americans in comparison to Caucasians who were urine positive were about 6 times less likely to report cocaine use when other factors are controlled for.” (see section 4.2).

This only covers a few of the alleged sources of anti-black bias within the U.S. criminal justice system (differences in sentence lengths, general differences in rates of arrests/convictions/incarcerations, and differences in punishment for drug crimes). For the sake of brevity, I will not consider all of the alleged instances of anti-black bias. The data that I have presented does not prove that there is no anti-black bias in the CJS. However, I believe the data should be sufficient to motivate a healthy level of skepticism for any future allegations of anti-black bias. We should be skeptical that such allegations are based on studies that have made the proper controls for characteristics that may correlate with race which could explain disparate treatment (e.g. behavior, IQ, etc.). Moreover, insofar as we do find evidence of an anti-black bias in the CJS, the bias is usually too small to account for a significant portion of disproportionate black criminality.

Black immigrants in the United States

According to a 2018 brief [archived] by the Cato Institute, black legal immigrants have far lower crime rates than black natives in the United States. The incarceration rate for black legal immigrants, while high relative to other legal immigrants, is low relative to natives. In fact, the incarceration rate for black legal immigrants is lower than the rate for natives of all racial groups, including white and Asian natives (Table 1).

This would be an unexpected outcome if the criminal justice system was a significant cause of black criminality and incarceration. If an anti-black bias in the criminal justice system explains black criminality and incarceration, then shouldn’t this anti-black bias also apply to black immigrants? Does the anti-black bias only apply to black residents who are citizens? This is implausible. A better explanation is that any anti-black bias in the CJS (assuming such a bias exists) plays a minor or negligible role in the elevated levels of black criminality and incarceration.

Early black misconduct

Rampant levels of misconduct appear in black children very early in life, as early as preschool. I give more details in a separate post, but here are some important facts:

• Losen and Skiba 2010 [archived] reported that black males are nearly 3 times as likely as white males to be suspended (28.3% vs 10%) in middle school [Figure 2]. Black females are over 4 times as likely as white females to be suspended (18% vs 4%) [Figure 2].
• A 2014 report [archived] by the U.S. Department of Education Office for Civil Rights shows that “Black students represent 18% of preschool enrollment, but 42% of preschool students suspended once, and 48% of students suspended more than once.”
• A 2018 study [archived] by the United States Government Accountability Office shows black preschoolers are over 3 times as likely as white preschoolers to be suspended (1.1% vs 0.3%) [Table 17].

Black criminality around the world

Now, consider the fact that we find similar relative levels of black criminality in other countries. Again, I give more detailed facts in a separate post, but here are a few important points:

• United Kingdom: according to data from the UK Ministry of Justice [archived], black people are arrested at 3 times the rate for white people (66.9 vs 22.8 per 1,000 population) [Table 3.12] and represented 13.1% of the prison population [Table 5.03] despite representing 3.1% of the general population [Table A]. Crime statistics [archived] by the UK government show that for the three-year period year ending March 2017 to the year ending March 2019, blacks accounted for 18% of principal suspects for homicide. This implies that the black homicide rate is over 6 times the white homicide rate, a similar pattern that is found in the United States.
• Canada: a fact sheet [archived] by the Colour of Poverty reported “In 2016, black people comprise 3.5% of the general Canadian population, but made up 10% of the federally incarcerated population.” A 2014 commentary [archived] revealed that “homicide rates among blacks in Montreal were estimated to be as high as 24 per 100,000, strikingly higher than their white counterparts at 3 per 100,000” (page 193). This implies that the black homicide rate is 8 times the white homicide rate, a similar pattern found in the United States.
• Brazil: a 2019 article [archived] published in Think Brazil reported that “In 2017, the number of blacks killed was almost three times higher than among non-blacks (a group encompassing white, Asian, and indigenous people).” A 2013 study [archived] published in Aggression and Violent Behavior grouped the homicide rates by race and found that the homicide rate for blacks (34.6 per 100,000) was over 2 times the rate for whites (16.3), and over 5 times the rate for Asians (6.8) (see section 4.1). Trends in homicide rates can be found here [archived].
• Consider The Handbook of Crime Correlates, a huge book published in 2009 that collects the summary analysis of crime research around the world. Table 2.3.2a summarized the results of a large number of studies that compared the rates of violent crime by blacks and whites in North America, the Middle East, Europe, and Oceania. Every study found that being black was significantly associated with criminality. Also, every study that compared the rates of non-violent criminal and delinquent behavior found that being black was significantly associated with property offenses, drug offenses, and delinquency (Table 2.3.2b).

It seems to me that we have very solid evidence that U.S. black criminality is not caused by anti-black bias in the CJS. First, two of the most common alleged sources of anti-black bias seem to be unsupported after reviewing other studies. Secondly, we find relatively low levels of incarceration among black legal immigrants in the United States. Finally, we find comparable levels of relative criminality among blacks who have never engaged with the U.S. CJS. This includes black children who display behavioral problems as early as preschool age, and it includes black criminality in other countries around the world. The best explanation of this widespread misconduct and criminality does not involve any appeal to an anti-black bias in the CJS.

Conclusion

My goal here was to provide reason to refute the seemingly mainstream assumption that the disproportionate rate of police killings of black people is caused by anti-black bias in the U.S. criminal justice system. I argued that this assumption is mistaken because there is a better explanation of the disproportionate police killings. The better explanation consists in the fact that black people disproportionately engage in criminal activity and the fact that the police disproportionately kill criminals. Further, I argued that disproportionate black criminality is not the result of an anti-black bias in the U.S. criminal justice system.

This post is agnostic with respect to the normative prescriptions that we ought to derive from this data. Also, this post is agnostic with respect to the cause of widespread black criminality (I explore possible causes of black criminality here and here). My only hope is that discussions involving black criminality and the criminal justice system can be more fruitful when all parties are aware of the relevant facts.