A friend complained to me recently that his son wasn’t getting into Ivy League colleges because it’s so hard for a middle-class white kid to be admitted, even with straight A’s. I asked if the advantages of being a middle-class white kid might be part of the reason his son had become a straight-A student in the first place. It got awkward.

As our politics have fractured increasingly around race, there seems to be more and more confusion about who’s discriminating against whom. For example, a national survey reported that both blacks and whites believed that discrimination against blacks had declined over the past few decades, but whites believed that discrimination against whites was now more common than discrimination against blacks.

The reason, say the study’s authors Michael Norton and Sam Sommers, is that whites see discrimination as a zero-sum game. The more they thought discrimination against blacks was decreasing, the more they felt discrimination against whites was increasing. That’s consistent with other studies showing that if you remind whites that the American population is becoming more diverse and that whites will soon be less that half of the population, their concern about anti-white discrimination increases. Whites tend to view increasing diversity as anti-white bias.

These kinds of data capture an important snapshot of public opinion, but that is the problem—surveys treat the question as a matter of opinion, like which basketball team you like most. But the question of whether discrimination disadvantages whites or blacks is not really a matter of opinion. It is a factual question that can be answered by science. In fact, it has been.

News stories are full of statistical evidence for disparities between black and whites, such as the fact that the average black family earns about half as much as the average white family, or that the unemployment rate for blacks is twice that for whites, or that the wealth of the average white family is ten times the wealth of the average black family. But this kind of evidence is like a political Rorschach test that looks very different to liberals and conservatives. What looks to liberals like evidence of discrimination looks to conservatives like evidence of racial disparities in hard work and responsible behavior.  

This dynamic was captured in a New York Times readers forum about a study showing large racial disparities in economic mobility, especially for black boys. A reader named Michael wrote, “Why is racism the only explanation for this phenomenon? Perhaps something happens to black boys while they are growing up that makes them less capable of succeeding in the U.S. economy… So, why do the authors take the easy way out and blame amorphous racism?” Professor Ibram Kendi responded, “Actually, the easy way out is to say there must be something wrong with these black boys. It is the easy way out that Americans have historically taken in trying to explain racial disparities in our society…Racist ideas of black inferiority is the easy way out.”

This kind of back and forth seems to be everywhere, from intellectuals arguing in books and essays to the general public arguing on social media. When it comes to statistical disparities, this is a rare case in which almost no one is disputing the facts. But the meaning of those facts appears endlessly up for grabs because opponents cannot agree on what is the cause of the disparities—discrimination or differences in merit.

The only kind of evidence that can hope to bridge this divide comes from experiments which directly measure discrimination — and these experiments have been done.

Consider an experiment by sociologist Devah Pager, who sent pairs of experimenters—one black and one white—to apply for 340 job ads in New York City. She gave them resumes doctored to have identical qualifications. She gave them scripts so that the applicants said the same things when handing in their applications. She even dressed them alike. She found that black applicants got half the call backs that white applicants got with the same qualifications.

This study inspired experiments in lots of areas of life. One study, for example, responded to more than 14,000 online apartment rental adds but varied whether the name attached to the email implied a white applicant (e.g., Allison Bauer) or a black applicant (e.g., Ebony Washington). The black applicants were twenty-six percent less likely to be told that the apartment was available.

These kinds of experiments are not ambiguous like statistics on disparities are. There were no differences in merit. Race was the cause. Real employers and landlords discriminated against blacks and in favor of whites, by a large margin.

This kind of direct evidence of discrimination against minorities have been found in other arenas. Professors are more likely to ignore emails from students of color. Airbnb hosts are more likely to tell black renters that the listing has already been taken. Pager and her colleagues published a meta-analysis incorporating every field experiment on hiring since the first ones were carried out in the 1980’s. Across two dozen studies, black applicants were called back 36 percent less than whites with the same qualifications. Not a single study found a reliable anti-white bias. Most sobering of all, the rate of discrimination is the same today as in the 1980’s.

The stakes in this debate are high, as Federal courts and potentially the Supreme Court will soon hear landmark cases on the use of race when making college admissions decisions. Is the present environment one that discriminates against black applicants? Or is it a level playing field made uneven by affirmative action policies? These are empirical questions, and we already have a lot of evidence in hand to answer them.

It may seem naïve, in a way, to think that data like this will change people’s minds. After all, even the overwhelming scientific consensus that humans are causing climate change hasn’t persuaded many whose worldviews conflict with the evidence. But there is a critical difference between short-term arguments and long-term belief change. Confronting an opponent with facts in the middle of a heated argument has probably never changed a single mind. Back-and-forth arguments are social contests in which people are often more motivated to win than to seek truth.

And yet, beliefs in the basic facts of climate change have gradually come more into line with the evidence over the last decade. The key is to keep repeating the facts and their basis in reason and science, until they become part of the background that any conversation takes for granted. It is frustratingly slow work. But to even get started, we need to move the conversation about discrimination beyond evidence of disparities, and focus on the experiments and the stubborn facts they deliver.