The anti-police narrative depends on suppression of facts, and the duplicity of anti-cop forces reached a shameless new low at a congressional hearing last week. Committee members should sanction the false testimony, given under oath, and publicly correct the record.
The House Judiciary Committee, now controlled by Democrats, had called a hearing to address a “series of deaths of unarmed African-American men while in police custody” as well as the “mistrust between police and marginalized communities.” Throughout the four-hour session, a photo array of blacks killed by the police played continuously on video screens around the room, interspersed with statistics allegedly proving that the police harbor lethal racist bias. Committee chairman Jerry Nadler claimed in his opening remarks that the “frequency of these killings and the absence of full accountability for those responsible send a message to members of the African American community that Black Lives Do Not Matter.” Nadler invoked the deaths of Michael Brown in Ferguson, Missouri, and Freddie Gray in Baltimore, as examples of “police misconduct against African-Americans,” though Barack Obama’s Justice Department found no misconduct in the first case, and criminal charges against the Freddie Gray officers were dismissed either before or after trial.
Minutes before the hearing was scheduled to begin, Al Sharpton escorted Gwen Carr, the mother of Eric Garner, to the witness table, surrounded by a mob of photographers. Garner had tragically died of a heart attack after New York Police Department officers tried to arrest him for selling loose cigarettes in July 2014; Garner resisted arrest, and one of the officers used a chokehold to take him down. Garner’s repeated last words—“I can’t breathe”—became an international rallying cry against police brutality. Carr gave impassioned testimony denouncing the lack of consequences for the police “murder” of her son: after five years of federal and local investigation, the officer who used the fatal chokehold was fired, but no prosecutions or other actions in the case have taken place. Carr left the witness table after her remarks to sit in the spectator section, but she remained a frequent reference point for the Democratic argument that the police devalue minority lives.
The eight remaining witnesses then began their testimony; I was one of the two witnesses called by the Republican minority. I asserted that the narrative that we are experiencing an epidemic of racially biased police shootings of black men was false. A study published this August in the Proceedings of the National Academy of Sciences was just the latest piece of research to undercut that narrative, I said. The study, by faculty at Michigan State University and the University of Maryland at College Park, found that it is the rate of violent crime that determines fatal police shootings. The more frequently officers encounter violent suspects from any given racial group, the greater the chance that members of that group will be fatally shot by a police officer. In fact, black civilians are shot less, compared with whites, than their rates of violent crime would predict. There was no evidence that officer race (i.e., whiteness) predicted the race of the shooting victim or that there was anti-black bias in fatal police shootings.
Next up was Phillip Atiba Goff. Goff heads the Center for Policing Equity at the John Jay College of Criminal Justice; the Center purports to document police bias through the use of data. Goff dramatically announced that his “love of country” and his “vocation as a scientist” required him to put aside his prepared remarks to “correct the record” so marred by my testimony. The authors of the PNAS study had recently acknowledged that their “central causal claim” was “unsupported by the data and factually wrong,” Goff asserted. Theirs is a “correlational” study, Goff said accusingly, and the “authors themselves have admitted to others in the scientific community that the central causal claim that they make, which is that there is no bias in [police shootings], is unsupported by the data.” Goff refused to be part of a “laundromat for junk science,” he concluded.
After the final witness—a former civil rights attorney in the Obama Justice Department who sued police departments for civil rights violations— completed her testimony, Chairman Nadler granted the minority request that I be given a minute and a half to respond to Goff. I said that the PNAS study was hardly an outlier; previous analyses had reached the same no-bias conclusion, including a 2017 paper by Harvard economist Roland Fryer that found no evidence of racial discrimination in shootings; a lab study of police shoot-don’t-shoot decisions in the state of Washington that found that officers in a realistic video simulator were three times less likely to shoot unarmed black suspects than unarmed white suspects; and shoot-don’t-shoot experiments by the University of Chicago’s Josh Correll that also found that police don’t shoot unarmed black civilians at any higher rate than unarmed white civilians. (Correll is now at the University of Colorado, Boulder.)
Nadler invited Goff to reply—and gave him nearly twice the time I had been given to respond. “None of that is true,” Goff claimed. Roland Fryer has “confessed to being embarrassed” by his study, which “has been roundly debunked,” Goff said. The Washington simulator study was not timed—so, according to Goff, irrelevant. And it was simply inaccurate to say that Correll’s research revealed no bias. “It absolutely did.” Goff brushed off as of little interest Correll’s “error” rate finding—the fact that officers don’t mistakenly shoot unarmed blacks at a higher rate than unarmed whites. What should matter, Goff said, is that the black armed suspects in Correll’s experiments were “shot more quickly; white armed suspects were shot more slowly.”
It is Goff’s testimony that was inaccurate, however—to say the least. The authors of the PNAS study have not acknowledged that their work was “factually wrong” or “unsupported by the data.” To the contrary, they have reaffirmed their research. Nor has Roland Fryer confessed to being embarrassed by his study. Goff apparently made up these allegations out of whole cloth.
The original findings of the PNAS study “largely stand unchanged,” wrote two of the authors in response to a critique of their paper from two Princeton politics professors. That critique “in no way invalidates” the central claim regarding the lack of anti-black racial disparities among those fatally shot by the police, reported David Johnson, a postdoctoral researcher in the Lab for Applied Social Science Research at the University of Maryland, and Joseph Cesario, a professor of psychology at Michigan State University. Even under the study design proposed by the critics, there is again “no significant evidence of anti-black disparity in the likelihood of being fatally shot by police,” Johnson and Cesario conclude.
After the hearing, I asked Fryer if he had repudiated his study. “I literally have absolutely no clue what Goff is talking about,” Fryer wrote back in an email. “My study was published in the top Econ journal [The Journal of Political Economy].” So confident remains Fryer in his findings that he pointed me to another paper of his in the American Economic Review (Papers and Proceedings) that reaffirmed his Journal of Political Economy results. Fryer found that after controlling for whether the suspect was armed, the nature of the suspect’s encounter with the officer, and other relevant factors, blacks in Houston were “27.4 percent less likely to be shot at by police relative to non-black, non-Hispanics. Investigating the intensive margin—who shoots first in an encounter with police or how many bullets were discharged in the endeavor—there are no detectable racial differences.” (Italics in original.) The evidence of police bias in some studies is “mainly a result of misspecified regression equations,” Fryer concluded. The most “granular data”—which takes into account crime rates and civilian behavior before and during a police-civilian interaction—suggest that there is “no bias in police shootings.”
I sent a request to the Center for Policing Equity that Goff supply a citation to the alleged retractions by Johnson, Cesario, and Fryer; there has been no response.
Let’s examine the Princeton critique of the PNAS paper directly. The PNAS researchers had constructed a database of 917 officer-involved fatal shootings from 2015. Fifty-five percent of the victims were white, 27 percent were black, and 19 percent were Hispanic. (Since 2015, the proportion of people killed by the police who are black has dropped; in 2018, it was 23 percent, according to the Washington Post’s database on fatal police shootings.) The PNAS analysis focused on the characteristics of each individual fatal police shooting to test statistically whether officer or civilian race predicted fatal police shootings. As the authors put it in their response paper, they calculated the “probability that a civilian is Black, Hispanic, or White given that a person has been fatally shot.” They found that officer or civilian race does not predict fatal police shootings; posing a threat to officers or civilians does.
The Princeton authors—Dean Knox and Jonathan Mummulo—argued that the PNAS study used the wrong calculation. Instead, the only meaningful test of police bias would look at the probability that a person in some wider population (beyond victims of fatal police shootings) will be shot given his race. Johnson and Cesario readily acknowledge that Knox and Mummulo’s approach is valid, but they reject the implication that their method is not—contrary to Goff’s claims.
The dispute implicates a vexed issue in the analysis of police activity: the benchmark problem. Racial-advocacy groups and the media invariably use population data as the benchmark for evaluating bias in police stops, arrests, or shootings: if blacks have proportionally more encounters with the police than their representation in the local population, police bias must be the reason, conclude those advocacy groups and the press. In New York City, for example, blacks are about 23 percent of the civilian population, but they made up 57 percent of stop, question, and frisk subjects in 2018. Whites are about 34 percent of New Yorkers; they made up 8.4 percent of stop subjects in 2018. Case closed, argue the advocates. The police are singling out blacks based on their race.
Others, myself included, have argued that crime rates should be the relevant benchmark for police activity, since policing today is data-driven: officers are deployed to where civilians are being victimized, and that is overwhelmingly in minority neighborhoods. If crime rates are used to measure police activity, policing no longer looks disproportionate. In New York, to continue the example, blacks were 72.6 percent of known shooting suspects in 2018, according to victim and witness identifications (those victims and witnesses being overwhelmingly black themselves). Whites in 2018 were 2.8 percent of known shootings suspects. Black and white stop rates—57 percent and 8.4 percent—are easily commensurate with rates of violent street crime.
The PNAS analysis was a deliberate attempt to sidestep the benchmark problem by testing for bias within the finite set of police shooting victims. The Princeton approach would revive the benchmark issue by requiring the choice of a benchmark against which to measure the probability of police shootings. In their response to Knox and Mummulo, the PNAS authors note that under a number of reasonable benchmarks, there is still no evidence of police bias in fatal shootings. Measured against Centers for Disease Control data about homicide rates, for example, blacks are five times less likely than whites to be fatally shot by the police. One could also observe that blacks constituted 62 percent of all robbery defendants, 57 percent of all murder defendants, and 45 percent of all assault defendants in America’s 75 largest counties in 2009, the latest year for which such county data is available, though blacks made up only 15 percent of the population in those counties, according to the Bureau of Justice Statistics. A roughly 25 percent share of police shooting fatalities is easily in line with what those county crime rates would predict, since police shootings will occur where officers are most likely to encounter armed and violent suspects.
Goff dismissed the PNAS study as “correlational”—meaning, it looked for causal associations among existing data, rather than setting up an experiment to generate new data and test the effects of deliberately introduced potential causes. But every other study of fatal police shootings is also correlational, since it would be ethically impossible to conduct an experiment regarding fatal police shootings. Goff’s implication that the PNAS study was illegitimate because it was correlational was another misrepresentation.
Ironically, Goff’s Center for Policing Equity has conducted its own study of police use of lethal force that was not only—inevitably—correlational but that reached the same conclusion as the PNAS and Fryer studies: there is no bias in lethal shootings by the police. The average lethal-force rate against white civilians is over twice that against black civilians for every 1,000 violent arrests, and nearly twice that for every 1,000 arrests overall, the study found. That result was buried in the report, however, and went unmentioned in the executive summary or introduction. I didn’t bring it up at the hearing, figuring that Goff’s willingness to distort would reach a hysterical pitch if I referenced his own work.
Goff was correct that the officers in the Washington state simulator study were not under artificial time pressure to make their shoot-don’t-shoot decisions, unlike the Correll studies, which give officers at most 750 milliseconds to respond to potential gun-toting suspects. That absence of a time limit was designed to replicate conditions on the street; even without a time limit, participants’ physiological response showed considerable stress.
Goff’s attempt to minimize the Correll finding that officers do not shoot unarmed blacks at a higher rate than unarmed whites was sheer desperation, since that finding cuts the heart out of the Black Lives Matter narrative. Correll suggests plausible reasons why officers may be quicker—measured by milliseconds—to reach a shoot decision when confronted by an armed black suspect compared with an armed white suspect: their experience with the demographics of street crime. Not only do blacks commit the vast majority of drive-by shootings (and whites almost none), the Department of Justice has found that police officers are five times more likely to die at the hands of a black suspect than a white suspect.
Goff’s conclusion in his second response to my remarks (“The idea that there is not bias is a—it is just not a serious position when you look at the corpus of the science”) was untrue. A robust body of scientific literature challenges the received position that policing—including the use of lethal force—is permeated by racial bias. Goff later said that “there is not a study that takes into account crime rates” that doesn’t find bias. That statement is also untrue. Goff misrepresented specific items in that literature, including the PNAS and Fryer studies, and the general tenor of that literature, as well. The House Judiciary Committee should call him back and demand that he correct the Congressional Record.
As for the rest of the hearing, the Democratic misrepresentations regarding the police were of the more usual nature, reflecting generalized prejudice and ignorance. Ron Davis, who headed the Office of Community Oriented Policing Services (COPS) in the Obama Justice Department, claimed: “The idea that police themselves can reduce crime is false. It is completely false.” This statement, too, was aimed at my testimony, which had also documented the New York policing revolution’s radical effect on the city’s crime rates. University of California, Berkeley, criminologist Franklin Zimring has made the most thorough study of New York’s crime decline to date and concluded that only policing explains it. Davis’s rejection of the efficacy of policing is an odd position for a COPS director to take, since the office distributes federal tax dollars to local police agencies to fight crime. Davis also claimed that white supremacists were infiltrating law enforcement but offered no documentation for that claim.
White supremacy and white nationalism were regular talking points for the Democratic committee members as well. “When we refuse to acknowledge the structural biases that are within our institutions, when we refuse to acknowledge the institutionalized racism that exists within all systems here in government, in law enforcement, in every system, . . . if we don’t acknowledge that white supremacy, if we don't acknowledge the history of institutionalized racism in this country, we cannot move forward,” announced Representative Pramila Jayapal of Washington.
The Republicans on the committee failed to push back against this narrative of systemic police bias, choosing instead to tell feel-good stories about “our brave men and women in uniform.” Such tales do little to rebut Black Lives Matter ideology, since both statements could be true: individual officers display heroism, and policing is infected by “structural racism,” in Davis’s words. The only way to dislodge the “endemic racism” argument is to challenge its factual basis directly. I was the only witness at the hearing with the ability to do so, but the Republicans asked me not one question. This is not a matter of personal ego but rather of the public battle of ideas.
The Democratic committee members and their witnesses clearly laid out their agenda should they retake the White House and Senate: mandatory implicit-bias training for cops, a huge waste of money that could be spent instead on tactical and de-escalation training; mandatory racial-profiling data collection, which will be measured, misleadingly, against a population benchmark; racial quotas for police hiring, which require lowered standards; and more federal consent decrees for police departments, which cripple the ability of cops to engage in proactive policing and divert millions of dollars into the pockets of federal monitors. As inimical as these policy items are to effective policing, the narrative that drives them—that the police are a threat to black communities—is more dangerous still. That narrative rests on duplicity, as amply demonstrated at last week’s hearings. Republicans, who invoke patriotism on a regular basis, are doing the country no favors by ceding the criminal-justice narrative to the activists and race-baiters.