As the media picked over the carcass of Donald Trump’s disastrous debate performance last week, Hillary Clinton’s critics continued to eye one statement by the Democratic nominee. “I think implicit bias is a problem for everyone, not just police,” she said in response to moderator Lester Holt’s question about recent police shootings of black men. “I think, unfortunately, too many of us in our great country jump to conclusions about each other.” She promised to fund implicit-bias training for police forces if she is elected.

By citing “implicit bias”—defined as unconsciously held “negative associations” about a group—the ever-studious Clinton introduced into the political conversation an academic term that has migrated from social-justice circles to the mainstream. A new paper by the Yale Child Study Center on implicit bias among preschool teachers gives us a chance to get a closer look at the concept in practice. The results are dismaying. Though endorsed by the United States Department of Health and Human Services and gullibly covered in major media outlets, the study is a mess of contradictions and spin. If it proves anything other than the bias of both social scientists and the media on racial issues, it’s that implicit bias—assuming there is such a thing, and that we know how to measure it—has no clear real-life consequences.

The Yale researchers’ motivation was uncontroversial. They wanted to examine the “underlying causes” for the disproportionate number of black boys expelled from preschool. “Black preschoolers are 3.6 times as likely to receive one or more suspensions relative to White preschoolers,” they note correctly. “Black children make up only 19 percent of preschool enrollment, but comprise 47 percent of preschoolers suspended one or more times. Similarly, boys are three times as likely as girls to be suspended one or more times.”

Their hypothesis about the racial discrepancy, too, was reasonable enough: perhaps teacher bias could explain some of the suspension gap. To test the theory, they showed a group of 132 black and white preschool teachers a series of thirty-second videos of four preschool children (two black and two white) “engaging in typical activities in a classroom” without any obvious misbehavior. An eye scanner revealed that the teachers watched the black children more than they watched the white children. The teachers were also given a follow-up questionnaire asking which of the children in the video needed the most attention. The results gave more nuance to the eye scanner findings: 10 percent of the teachers chose the black girl, 13 percent chose the white girl, 34 percent pointed to the white boy, and 42 percent pointed to the black boy.

On first glance, the findings could be chalked up as a win for implicit-bias theory. But there were several puzzles. For one thing, black teachers watched the black boys more than the white teachers did. This is consistent with the report’s later finding that black teachers rated black children’s behavior more severely than white teachers did. In an unexpected finding, white teachers were more critical of white children. How do the researchers explain this? By assuming that the white teachers were “holding [black children] to a lower behavioral standard.” If the white teachers had judged the black children more severely than the white children, would the authors have concluded that they were simply holding black children to higher standards? The question answers itself.

Also worth noting: the questionnaire given after the video revealed more gender bias than racial bias. In fact, the teachers were less likely to think a black girl needed to be watched than a white girl. The researchers weren’t interested in the implicit bias suggested by this gender gap—perhaps because they share it. They have good reason. Boys of any race are more likely than girls to be put in special education classes, to be diagnosed with attention deficit disorder, and to be suspended. Is this evidence of bias against boys? If so, parents are also guilty. In surveys, they consistently report more behavior problems with their boys than with their girls. Boys misbehave more than girls. This is not bias; it’s fact.

Now, keeping an eye on black boys as potential troublemakers does suggest that teachers (white and black) have more negative expectations about them. Is this a matter of racial bias, or is it something more on the order of the above-mentioned gender bias? That is, could it reflect something in the real world? There’s abundant research that—regardless of color—boys who grow up in father-absent homes have more behavioral problems than those who grow up in two parent homes. Black children are far likelier to grow up without their fathers. If fatherlessness leads to acting out then, yes, black boys would have proportionately more behavior problems than white boys.

Any teacher worth her salt taps into her accumulated experience. What her eyes watch in a six-minute, fictionalized video shouldn’t worry anyone. It could be a problem if a teacher lets her past experience determine the way she treats her students. Not only does the Yale paper find no sign of that, it finds evidence for the opposite. “No main effects were found for the single indicator item assessing participant’s recommendations regarding suspension and expulsion or the number of days to suspend or expel the child.” Let this sink in for a moment. A paper hypothesizing that implicit bias is the reason why black children are suspended at disproportionately high numbers finds no sign that a child’s race affects whether a teacher will recommend him for suspension or expulsion. 

How did a media fascinated by putative proof of racism in preschool cover this unexpected—and for the researchers, unwelcome—finding? They didn’t. The headline on CNN’s story suggested that racial bias might start as early as preschool. NPR’s Morning Edition noted that bias isn’t just a police problem, it’s a preschool problem. Dozens of other articles on the study ignored the fact that it failed on its own terms.

Maybe before recommending implicit-bias training, Clinton’s policy staff will read the Yale study more carefully. Call me biased, but I’m not counting on it.

Kay Hymowitz is a contributing editor of City Journal, the William E. Simon Fellow at the Manhattan Institute, and the author of books including Manning Up: How the Rise of Women Has Turned Men Into Boysand The New Brooklyn: What it Takes to Bring a City Back (forthcoming in 2017).

Editor’s note: This article originally appeared in City Journal.