Hidden Evidence in Racial Bias Research by Cesario and Johnson

In a couple of articles, Cesario and Johnson have claimed that police officers have a racial bias in the use of force with deadly consequences (Cesario, Johnson, & Terrill, 2019; Johnson, Tress, Burkel, Taylor, & Cesario, 2019). Surprisingly, they claim that police officers in the United States are MORE likely to shoot White civilians than Black civilians. And the differences are not small either. According to their PNAS article, “a person fatally shot by police was 6.67 times less likely (OR = 0.15 [0.09, 0.27]) to be Black than White” (p. 15880). In their SPPS article, they write “The odds were 2.7 times higher for Whites to be killed by police gunfire relative to Blacks given each group’s SRS homicide reports, 2.6 times higher for Whites given each group’s SRS homicide arrests, 2.9 times higher for Whites given each group’s NIBRS homicide reports, 3.9 times higher for Whites given each group’s NIBRS homicide arrests, and 2.5 times higher for Whites given each group’s CDC death by assault data. Thus, the authors claim that for every Black civilian killed by police, there are 2 to 6 White civilians killed by police under similar circumstances.

The main problem with Cesario and Johnson’s conclusion is that they rest entirely on the assumption that violent crime statistics are a reasonable estimate for the frequency of encounters with police that may result in the fatal use of force.

One cannot experience a policing outcome without exposure to police, and if exposure rates differ across groups, then the correct benchmark is on those exposure rates.” (Cesario, Johnson, & Terrill, 2019, p. 587).

In the context of police shootings, exposure would be reasonably approximated by rates of criminal involvement for Blacks and Whites; the more group members are involved in criminal activity, the more exposure they have to situations in which police shootings would be likely to occur” (p. 587).

The quotes make it clear that Cesario and Johnson use crime statistics as a proxy for encounters with police that sometimes result in the fatal use of force.

What Cesario and Johnson are not telling their readers is that there are much better statistics to estimate how frequently civilians encounter police. I don’t know why Cesario and Johnson did not use this information or share it with their readers. I only know that they are aware that this information exists because they cite an article that made use of this information in their PNAS article (Tregle, Nix, Alpert, 2019). Although Tregle et al. (2019) use exactly the same benchmarking approach as Cesario and Johnson, the results are not mentioned in the SPPS article.

The Police-Public-Contact Survey

The Bureau of Justice Statistics has collected data from over 100,000 US citizens about encounters with police. The Police-Public Contact Survey has been conducted in 2002, 2005, 2008, 2011, and 2015. Tregle et al. (2019) used the freely available data to create three benchmarks for fatal police shootings.

First, they estimated that there are 2.5 million police-initiated contacts a year with Black civilians and 16.6 million police initiated contacts a year with White civilians. This is a ratio of 1:6.5, which is slightly bigger than the ratio for Black and White citizens (39.9 million vs. 232.9 million), 1:5.8. Thus, there is no evidence that Black civilians have disproportionally more encounters with police than White civilans. Using either one of these benchmarks, still suggests that Black civilians are more likely to be shot than White civilians by a ratio of 3:1.

One reason for the proportionally higher rate of police encounters for White civilians is that they drive more than Blacks, which leads to more traffic stops for Whites. Here the ratio is 2.0 million to 14.0 million or 1:7. The picture changes for street stops, with a ratio of 0.5 million to 2.6 million, 1:4.9. But even this ratio still implies that Black civilians are at a greater risk to be fatally shot during a street stop with an odds-ratio of 2.55:1.

It is telling that Cesario and Johnson are aware of an article that came to opposite conclusions based on a different approach to estimate police encounters and do not mention this finding in their article. Apparently it was more convenient to ignore this inconsistent evidence to tell their readers that data consistently show no anti-Black bias. While readers who are not scientists may be shocked by this omission of inconvenient evidence, scientists are all to familiar with this deceptive practice of cherry picking that is eroding trust in science.

Encounters with Treats and Use of Force

Cesario and Johnson are likely to argue that it is wrong to use police encounters as a benchmark and that violent crime statistics are more appropriate because police officers mostly use force in encounters with violent criminals. However, this is simply an assumption that is not supported by evidence. For example, it is questionable to use homicide statistics because homicide arrests account for a small portion of incidences of fatal use of force.

A more reasonable benchmark are incidences of non-fatal use of force. The PPCS data make it possible to do so because respondents also report about the nature of the contact with police, including the use of force. It is not even necessary to download and analyze the data because Hyland et al. (2015) already reported on racial disparities in incidences that involved threats or non-fatal use of force (see Table 2, Table 1 in Hyland et al. (2015).

The crucial statistic is that there are 159,100 encounters with Black civilians and 445,500 encounters with White civilians that involve threat or use of force; a ratio of 1: 2.8. Using non-fatal encounters as a benchmark for fatal encounters still results in a greater probability of a Black civilian to be killed than a White civilian, although the ratio is now down to a more reasonable ratio of 1.4:1.

It is not clear why Cesario and Johnson did not make use of a survey that was designed to measure police encounters when they are trying to estimate racial disparities in police encounters. What is clear is that these data exist and that they lead to a dramatically different conclusion than the surprising results in Cesario and Johnson’s analyses that rely on violent crime statistics to estimate police encounters.


It is important to keep in mind that the racial disparity in the fatal use of force in the population is 3:1 (Tregle et al., 2019, Table 1). The evidence from the PPCS only helps to shed light on the factors that contribute to this disparity. First, Black civilians are not considerably more likely to have contact with police than White civilians. Thus, it is simply wrong to claim that different rates of contact with police explain racial disparities in fatal use of force. There is also no evidence that Black civilians are disproportionally more likely to be stopped by police by driving. The caveat here is that Whites might drive more and that there could be a racial bias in traffic stops after taking amount of driving into account. This simply shows how difficult it is to draw conclusions about racial bias based on these kind of data. However, the data do show that the racial disparity in fatal use of force cannot be attributed to more traffic stops of Black drivers. Even the ratio of street stops is not notably different from the population ratios.

The picture changes when threats and use of force as added to the picture. Black civilians are 2.5 times more likely to have an encounter that involves threats and use of force than White civilans (3.5% vs. 1.4%, in Table 2; Table 1 from Hyland et al., 2015).

These results shed some light on an important social issue, but these numbers also fail to answer important questions. First of all, they do not answer questions about the reasons why officers use threats and force more often with Black civilians. Sometimes the use of force is justified and some respondents of the PPCS even admitted that the use of force was justified. However, at other times the use of force is excessive. The incidence rates in the PPCS are too small to draw firm conclusions about this important question.

Unfortunately, social scientists are under pressure to publish to build their careers, and they are under pressure to present strong conclusions to get their manuscripts accepted. This pressure can lead researchers to make bigger claims than their data justify. This is the case with Cesario and Johnson’s claim that officers have a strong bias to use deadly force more frequently with White civilians than Black civilians. This claim is not supported by strong data. Rather it rests entirely on the use of violent crime statistics to estimate police encounters. Here I show that this approach is questionable and that different results are obtained with other reasonable approaches to estimate racial differences in police encounters.

Unfortunately, Cesario and Johnson are not able to see how offensive their claims are to family members of innocent victims of deadly use of force, when they attribute the use of force to violent crime, which implies that the use of force was justified and that victims are all criminals who threatened police with a weapon. Even if the wast majority of cases are justified and fatal use of force was unavoidable, it is well known that this is not always the case. Research on fatal use of force would be less important if police officers would never make mistakes in the use of force. Cesario and Johnson receive tax-payer money to found their research because fatal use of force is sometimes unnecessary and unjustified. It is those cases that require an explanation and interventions that minimize the unnecessary use of force. To use taxpayer’s money to create the false impression that fatal use of force is always justified and that police officers are more afraid of using force with Black civilians than they are afraid of Black civilians is not helpful and offensive to the families of innocent Black victims that are grieving a loved one. The families of Tamir Rice, Atatiana Jefferson, Eric Garner, Philando Castile, to name a few, deserve better.

4 thoughts on “Hidden Evidence in Racial Bias Research by Cesario and Johnson

  1. Professor Schimmack. I’m just curious. You say homicide arrest account for a relatively small portion of incidents that result in fatal uses of force. What is the exact percentage.

Leave a Reply