Many academic studies claim to find widespread racial discrimination in America. However, most of those studies are either based on fabricated data; or omit major variables needed to test whether racial disparities are due to racism, or are instead the result of non-discriminatory causes. (The Supreme Court’s Croson decision warns that it is “completely unrealistic” to expect that minorities will be found in every field in “lockstep proportion to their representation in the local population” in the absence of racism. Yet academics commonly make just that unrealistic assumption).
If a researcher studies a field and claims to find widespread discrimination in it (such as in an industry, or in the criminal justice system), it is easy to get that finding of discrimination published in an academic journal, even if the researcher has omitted major variables needed to actually prove discrimination, or even if the researcher is unwilling to make his data available (many studies contain the claim that the authors’ data will be made available upon request, but this is often a lie — the authors usually decline to turn over the data when it is later requested. 93% of authors who indicated that data was available on request either did not respond or declined to share their data, according to a recent study in an academic journal).
By contrast, if a researcher finds that systemic discrimination is absent in a field or the criminal justice system, the researcher may be unable to get his research published. And if it is published, a finding of non-discrimination may trigger reprisals by left-wing academics, against the study’s author and those who publish or disseminate it (as The College Fix noted in the article “Scholar forced to resign over study that found police shootings not biased against blacks.”).
The College Fix reports on a rare example of a professor facing consequences for fabricating data that allegedly showed racism in the criminal justice system:
A public Florida university fired a professor after completing a months-long investigation into allegations of research misconduct in racial bias studies.
Criminologist Eric Stewart “demonstrated extreme negligence in basic data management, resulting in an unprecedented number of articles retracted” and “numerous other articles now in question,” according to a five-page termination letter Provost James Clark at Florida State University sent to Stewart earlier this month.
“The damage to the standing of the University and, in particular, the College of Criminology and Criminal Justice and its faculty approaches the catastrophic and may be unalterable,” Clark continued.
Reports were provided by FSU’s misconduct inquiry committee in 2019 and 2020 in response to misconduct allegations against Stewart, according to the July 13 letter, which noted Stewart had responded that the reports “indicate that the misconduct claims were rejected by multiple panel experts.”
But the provost’s letter continued, “In the four years since the initial issues arose, you [Stewart] have not taken any meaningful steps to remedy the situation, you have not re-created or attempted to re-create any of the studies, you have not pursued any remedial action, and you have even refused to cooperate with your FSU colleagues and coworkers who requested to work with you on these matters.”….
Stewart left his job in March, years after an investigation into his research began and after six race-related studies coauthored by him had been retracted…“Professor Stewart’s 16-year FSU career appears to have ended, signaled by his abrupt March 2023 absence,” The Florida Standard reported.…“His sudden, unexplained replacement may indicate the looming end of the investigation, with enough evidence of fraud discovered to justify termination.”….
Most recently, the academic journal Criminology retracted on December 12, 2019, a study co-authored by Stewart called “Ethnic Threat and Social Control: Examining Public Support for Judicial Use of Ethnicity in Punishment.”
In the discredited article, Stewart had argued “that as black and Hispanic populations grew, the surrounding white populations wanted more racially discriminatory sentencing,” The Fix reported.
Stewart himself “identified a mistake” in the data, and a co-author, Justin Pickett, “has publicly stated his view that the identified discrepancies are not attributable to researcher error,” according to the publisher’s retraction statement.
“Scientific fraud occurs all too frequently….and I believe it is the most likely explanation for the data irregularities in the five retracted articles,” Pickett wrote March 2020 in an Econ Journal Watch article, “The Stewart Retractions: A Quantitative and Qualitative Analysis.”
“The retraction notices say honest error, not fraud, is the explanation.” Pickett wrote. “Fortunately, if that is true, Dr. Stewart could easily prove it: recreate the original sample that produces the findings in Johnson et al. (2011) and then publicly explain how he did it.”
Pickett teaches criminology at the University of Albany, according to his university bio. The Fix reached out to Pickett, who declined to comment.
“There’s a huge monetary incentive to falsify data and there’s no accountability. If you do this, the probability you’ll get caught is so, so low,” criminologist Pickett told the Florida Standard. “There’s too much incentive to fake data and too little oversight.”….Stewart’s work garnered nearly $3.7 million in funding….He worked on grants from the Florida Department of Juvenile Justice, the National Science Foundation, the National Institute of Mental Health, and the National Institute of Justice.
Stewart made $190,000 per year at the university, Fox News reported in April.
If Professor Stewart had fabricated the data for fewer studies, he likely would never have been detected. But he fabricated the data for so many that he eventually got caught. Even so, the university only calls him out for “extreme negligence,” even though he obviously committed extensive fraud. False data does not come into being over and over again on a vast scale without fraud.
It is easy to get away with fabricating one or two studies, if they have a progressive slant that is shared by most readers and staff of academic journals.
By contrast, a truthful academic can lose his job over a single accurate study that angers the progressive readers of an academic journal, or progressive academics. And such a study may well be retracted even if its methodology was reliable and the underlying data was accurate. For example, a study finding that police shootings were generally not racist was retracted because it was cited by a conservative in the Wall Street Journal. Heather Mac Donald, a conservative legal scholar, had discussed the study in a widely-read Wall Street Journal column, “The Myth of Systemic Police Racism.”
That publicity resulted in an enormous backlash against the academics who conducted and discussed the study, by progressive academics angry about its conclusions.
The first casualty of the backlash was Stephen Hsu, the Vice President for Research at Michigan State University. He was forced to resign because he publicly discussed the study, as The College Fix notes in the article “Scholar forced to resign over study that found police shootings not biased against blacks.”
As the Criminal Justice Legal Foundation notes, “Hsu’s crime was publicizing research done at his university, which is exactly what you would expect a VP of research to do. But the particular piece of research reached [an ideologically] Forbidden Conclusion.” The research was valid, but in today’s academia, “the truth shall get you fired.”
After witnessing what happened to Hsu, the researchers who conducted the study, such as Michigan State’s Joseph Cesario, sought their own article’s retraction, citing its “misuse by the media” — that is, the Wall Street Journal, which accurately described the study.
But as a law professor notes, there was no reason to retract this study other than politics. David Bernstein is a law professor at George Mason University, and a legal expert on junk science and the admissibility of scientific evidence. He also is an adjunct scholar at the Cato Institute, which highlights police abuse and wrongful killings by the police.
Professor Bernstein observes that “It’s absurd to ask that a valid study be retracted b/c you think others are ‘misusing’ it. A study says what it says, and so long as it wasn’t actually flawed it shouldn’t be retracted for political reasons except perhaps under truly extreme circumstances, which this isn’t.” In discussing an alleged defect in the study, he noted that “the extrapolation” the studies’ critics “are objecting to seems to me a perfectly reasonable one.” Given the weaknesses of the objections to the study, he agreed with a commenter who concluded that the “request for retraction resulted from a sustained attempt to discredit politically unpopular research,” rather than anything being wrong with the research.
Retracting such studies is harmful to our understanding of police violence. As Stephen Hsu noted, “Cesario’s work (along with similar work by others, such as Roland Fryer at Harvard) is essential to understanding deadly force and how to improve policing.” So these studies finding no racial bias in police shootings should be discussed, not retracted.
As a commenter observed in response to the retraction request:
This retraction is being sought for political reasons, even though the study contained valuable information. The study itself, while not perfect (no study ever is), is of better quality than most research on police shootings.
It is due to the threat of reprisals at odds with academic freedom. It comes after the Vice President for Research at MSU was forced to resign largely because he cited this study, whose conclusions offended political activists. …
So the researcher at MSU had good reason to seek this retraction, to avoid facing a similar fate.
Shootings of black people aren’t any higher than one would expect given the black crime or the black arrest rate, which buttresses research finding that police shootings are not systemically racist. The black crime rate is simply much higher than the general crime rate, as data from the Bureau of Justice Statistics and the FBI’s Uniform Crime Reports shows. Half of all murders are committed by African-Americans, who account for only 13% of the U.S. population.
Despite all this, most unarmed people shot or killed by the police are white.
It is quite obvious that most police shootings are not due to racism.
Many of the people on Twitter angry over this study seem convinced that the police kill only unarmed black people, not unarmed white people, and that police killings of people like George Floyd must be racist as a result.
But that’s wrong. Most unarmed people killed by the police are white. In 2019, 9 unarmed black people were shot and killed by on-duty police officers, compared with 19 whites. Victims remain mostly white if you expand the tally to include off-duty police officers and deaths from causes other than gun shots.
Yet the media often leave the misimpression that the problem of police killings is entirely racial. For example, the New York Times misleadingly wrote on June 5 that “nothing changes with police killings. Gruesome, high-profile cases keep coming — Eric Garner, Michael Brown, Freddie Gray, Breonna Taylor, George Floyd, now Manuel Ellis.” Each of the six people it listed is African-American.
Not only were these killings not proven to be race-based, one of them — the shooting of Michael Brown — wasn’t even unjustified, according to a report by the Justice Department’s Civil Rights Division. As the Obama administration noted on page 82 of that report, “the shots fired” after Michael “Brown turned around were in self-defense.” Indeed, “several of the” mostly black “witnesses stated that they would have … responded” as the police officer did in shooting Brown.
The media leave the false impression that most people killed by the police are black. That’s because it spends weeks talking about the killing of black people like George Floyd, but it only briefly mentions the police killings of unarmed whites, like Tony Timpa or Daniel Shaver. Timpa died after 14 minutes of being restrained and struggling to breathe. Shaver was shot despite pleading not to be killed.
The academic backlash against the police-shooting study helps explain why there are fewer new studies finding an absence of bias in the criminal justice system — even as societal racism continues to diminish, according to surveys like the General Social Survey. Researchers used to regularly find that the criminal justice system was fair to racial minorities, in arrests and sentencing, as noted at this link. But researchers now have an incentive to conduct misleading studies in order to reach the opposite conclusion. Specific examples are given in this article.
Researchers used to regularly find that the criminal justice system was fair to racial minorities, in arrests and sentencing. In 1994, federal statistician Patrick Langan looked at the nation’s 75 largest counties and found “no evidence that, in the places where blacks in the United States have most of their contacts with the justice system, that system treats them more harshly than whites.” As he noted in “No Racism in the Justice System,” “Many studies have been conducted that show no bias in the arrest, prosecution, adjudication, and sentencing of blacks.”
Similarly, statistical expert Stephen P. Klein of the RAND Corporation studied California’s state criminal justice system and found that criminal sentencing in California was racially fair and non-discriminatory. (See Stephen P. Klein, et al., “Race and Imprisonment Decisions in California,” 247 Science 812 (1990)). That was the opposite of what Dr. Klein expected to find. He had served as an expert witness for civil-rights groups in landmark cases such as Serrano v. Priest — and studied criminal justice on the recommendation of the liberal California ACLU, which sees racism everywhere. But his statistical analysis debunked claims that the criminal justice system was systematically racist.
These studies involved painstaking statistical analysis that sought to take all relevant factors into account. The more factors a researcher takes into account, the more accurate a statistical analysis becomes.
But the more factors a study takes into account, the more time and money it takes to do the study. So researchers are tempted to cut corners by omitting factors or variables that are hard to measure, or that the researcher suspects may not be all that important.
There is another, even bigger reason for researchers to wrongly omit relevant variables or rely on incomplete data: Taking into account more data or variables can end up debunking claims of discrimination, rather than providing the “proof” of discrimination that progressive officials and journalists want. Studies frequently allege discrimination precisely by ignoring key variables. Their authors are rewarded by being given tons of favorable publicity; or having their studies lead to social change.
A classic example is the gender-bias study used to give female faculty pay raises in Smith v. Virginia Commonwealth University (1996).
The study claimed female faculty were being paid less than men by VCU due to sex discrimination. But it turned out that the study ignored relevant factors actually used by the university to set pay — such as scholarly productivity, and whether a faculty member had previously served as an administrator. If these important variables had been included in the statistical analysis, there would almost certainly have been no finding of discrimination.
But these important variables were excluded, leading to the university giving its female faculty pay raises to compensate for the non-existent discrimination. Male faculty then sued, alleging that because there was no discrimination against women to remedy, the gender-based pay raises discriminated against men.
A federal appeals court ruled that the male faculty could sue the university over the gender-based pay raises. It concluded that the omission of these “major” variables (such as productivity and administrative experience) meant that the gender-bias study was flawed. After its ruling, the university paid off the male faculty to settle the lawsuit, because it was fairly obvious the university would lose.
The appeals court was interpreting the Supreme Court’s murky and vague decision in Bazemore v. Friday, which says a study should include the “major” factors and variables to be admissible in a race or sex discrimination case.
But what makes a variable major versus minor? The Supreme Court’s vague decision itself gave little guidance as to what is a “major” variable that must be included, versus a minor one that can be excluded.
That vagueness has been the source of endless mischief, and incentivized countless bad studies. A researcher who wants to engineer a false finding of discrimination can just omit variables that are supposedly “minor,” but which, if included, would show that a racial or sexual disparity isn’t due to discrimination, but rather, something else (like fewer minorities than whites having the qualifications needed for a job, or more blacks than whites having a prior criminal record).
The researcher’s false finding of discrimination can then be used to justify doing things that progressive officials are eager to do, like creating an affirmative action plan or awarding gender-based pay raises.
For every flawed study alleging discrimination that is successfully challenged in court because it omitted major variables, there are countless others that are never even challenged, because such a challenge is just too costly. To challenge such a study, it is often necessary to pay an expert witness to explain to the court why the omitted variables are major rather than minor, and thus should have been excluded. Such experts usually charge at least $750 per hour for their work, and take many hours to complete and write up their analysis.
If you are a researcher, why conduct a painstaking statistical analysis that takes all relevant factors into account, and thus finds no racism or sexism, when you can make your job easier and reduce your workload by deliberately omitting relevant factors, and thus reach the politically less risky conclusion of “discrimination”?
The challenge to the flawed gender-discrimination study in Smith v. VCU was only successful because the challengers lucked out, and received hundreds of thousands of dollars worth of free legal assistance from the Center for Individual Rights.
