Why would two professors want to retract a paper that had no apparent methodological, mathematical, or interpretative errors?  The answer, as described in part one, is “cancel culture.”

In a “Statement on the retraction of ‘Officer characteristics and racial disparities in fatal officer-involved shootings,’” Joseph Cesario of Michigan State University and David Johnson of the University of Maryland, write,

We were careless when describing the inferences that could be made from our data. This led to the misuse of our article to support the position that the probability of being shot by police did not differ between Black and White Americans (MacDonald, 2019). To be clear, our work does not speak to this issue and should not be used to support such statements. We accordingly issued a correction to rectify this statement (Johnson & Cesario, 2020).

Although our data and statistical approach were valid to estimate the question we actually tested (the race of civilians fatally shot by police), given continued misuse of the article (e.g., MacDonald, 2020) we felt the right decision was to retract the article rather than publish further corrections. We take full responsibility for not being careful enough with the inferences made in our original article, as this directly led to the misunderstanding of our research.  [Emphasis added.]

As evidence of such misuse, they cite two articles by Manhattan Institute fellow Heather Mac Donald. (See here and here.)  If you read those articles, however, you’ll see that Mac Donald neither said nor implied that “the probability of being shot by police did not differ between Black and White Americans.” Instead, she simply reported what the study found, which is that once violent crime rates are taken into account, there is no racial disparity. In the first article, Mac Donald provides extensive quotes from the authors explaining their findings and defending them from criticism. In the second, she quotes a passage from the authors’ own summary: “We find 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. Instead, [the rate of] race-specific crime strongly predicts [a] civilian [victim’s] race.”

It turns out, in fact, that the authors’ real reason for wanting to retract their report isn’t that they don’t want their findings to be misused; they want it retracted because they don’t want those findings even to be read. “The mistake we made,” they say, “Was drawing inferences about the broader population of civilians who interact with police rather than restricting our conclusions to the population of civilians who were fatally shot by the police. We are thankful to [Princeton University professor Dean] Knox and [Princeton University professor Jonathan] Mummolo … for highlighting this error.”

To understand the significance of this rather cryptic statement, you have to read Knox and Mummolo’s critique of the PNAS paper and the authors’ reply, but what it boils down to is this. Knox and Mummolo insisted that the only acceptable way to analyze and report police shooting data is by comparing the number of black civilians shot by the police to the number of black civilians in the entire population. When the authors of the PNAS paper originally replied to Knox and Mummolo, they disagreed.  But by making this statement in their retraction request, however, they are indicating that they now accept that Knox and Mummolo are right. When it comes to police shootings, some comparisons must not be made, and some things must not be said.

Comparing one set of data to another in the way recommended by Knox and Mummolo is known as benchmarking. The authors of the PNAS paper didn’t use the benchmarking approach, and, in their response to Knox and Mummolo, they defended their decision not to do so. However, they also pointed out that, even if benchmarking is used, there are a number of different benchmarks to which the number of black civilians shot by the police could be compared. It is far from obvious that the total number of black civilians is the best one to use. For example, using the number of black civilians who commit violent crimes might be a better benchmark because people who commit violent crimes are much more likely to be involved in violent confrontations with the police than people who don’t. If one uses that benchmark, they noted, the results are consistent with the results they obtained by using their alternative approach.

As noted, since the time of their response to Knox and Mummolo, the authors of the PNAS paper have changed their tune, quite possibly because of what had happened to Professor Hsu. In their recent retraction request, they accept the rule that researchers who take anything other than the officially approved approach to police shootings data must refrain from “drawing inferences about the broader population of civilians who interact with police.” Somewhat surprisingly, however, in what may be a parting shot at the cancel mob, they add:

Benchmarking calculations and their derivations generally point to the same conclusion: Relative to the proportion of Black civilians in the U.S., Black Americans are shot more than we would expect. However, relative to various proxies for the proportion of Black civilians who commit violent crime, Black Americans are not shot more than we would expect. This has been consistently shown for the majority of fatal shootings (90-95%) where the citizen shot is an immediate threat to an officer or other citizen (Cesario et al., 2019; Fyfe, 1980; Goff et al., 2016; Inn et al., 1977; Tregle et al., 2019; Worrall et al., 2020), though some evidence has been presented that racial bias may be present in the remaining types of shootings (Ross et al., in press). The lack of racial disparity once violent crime rates are taken into account has also been shown in papers using more complex analytic approaches than proportion comparisons (Fryer, 2016; Mentch, 2020; Ross et al., in press).

I suppose this small act of rebellion may give the authors some sort of satisfaction, but it must be cold comfort, given that, once their retraction request has been granted, the research to which they devoted years of their lives will be consigned to oblivion. As for the rest of us, it’s no comfort at all. Following the review of the benchmarking literature quoted above, the authors say, “One problem with such benchmarking approaches is that debate arises about whether it is more informative to compare the number of civilians shot to overall population proportions or proxies for violent crime proportions.” But debate about this question shouldn’t be regarded as a problem. On the contrary, debate about this and similar questions is the only way to solve the real problem, which is that the police are killing too many black men. If we’re serious about trying to solve that problem, we need to know why so many black men are being killed.

The conventional explanation—which may, of course, be correct—is that the police are racially biased. Because almost everybody accepts that explanation, the solutions that are currently being discussed—which run the gamut from requiring police officers to undergo anti-bias training to defunding the police altogether—all focus on the police. If, however, the soon to be retracted study is correct, it might make more sense to focus on young black men instead. The authors of the PNAS study themselves make this point. After noting that, “One of our clearest results is that violent crime rates strongly predict the race of a person fatally shot,” the authors discuss the policy implications of that finding. “At a high level,” they say, “Reducing race-specific violent crime should be an effective way to reduce fatal shootings of Black and Hispanic adults. Of course, this is no simple task—crime rates are the result of a large and dynamic set of forces. However, the magnitude of [the] disparities speaks to the importance of this idea.”

That, of course, is not something a lot of Americans want to hear right now. When it comes to the question of why the police shoot so many black men, a lot of Americans are passionately committed to the view that racially biased policing is the cause of the problem, and that reforming or disbanding the police is the solution, and, as we have seen, they are all too willing to punish anyone who disagrees. On its own, however, a passionate commitment to a set of ideas tells us nothing about whether those ideas are correct. In the days of Copernicus and Galileo, a lot of people were passionately committed to the idea that the earth stood stationary at the center of the universe. That didn’t make it true. In the 20th century, a lot of people were passionately committed to the ideals of Fascism, National Socialism, and Communism. That didn’t justify the misery and death they inflicted on hundreds of millions of innocent people.

Frightening academics into retracting papers that challenge the conventional wisdom won’t help us reduce the number of black men killed by the police, and canceling everyone who disagrees with us won’t solve any of the other problems we face as a nation either. If we really want to solve our problems—and, more importantly, if we want to make sure we don’t go down the path that Europe went down in the 1920s and 30s—we must do all we can to ensure that everyone is free to criticize the prevailing orthodoxy and propose alternatives. We must, in short, cultivate the rationalist attitude recommended by Popper and live by its motto: “I may be wrong, and you may be right, and by an effort, we may get nearer to the truth.”