Police officers face off with protesters during demonstrations in Charlotte, North Carolina, on September 21, 2016, following the fatal shooting of 43-year-old Keith Lamont Scott by a police officer. | Sean Rayford/Getty Images

Phillip Atiba Goff explains what we know — and don’t know — about racial bias in policing.

Our conversation over race and policing — like our conversations over virtually everything in America — is shot through with a crude individualism. Talking in terms of systems and contexts comes less naturally to us, but that means we often miss the true story.

Phillip Atiba Goff is the co-founder and CEO of the Center for Policing Equity, as well as a professor of African American studies and psychology at Yale University. At CPE, Goff sits atop the world’s largest collection of police behavioral data. So he has the evidence, and he knows what it tells us — and, just as importantly, what it doesn’t even attempt to measure. He knows what we can say with confidence about race and policing, and what we wish we knew but simply don’t. He thinks in systems, in contexts, in uncertainty — in the bigger, harder picture.

That’s what this conversation on The Ezra Klein Show is about. What do we know about racial bias in policing? At what levels does it operate? Where has it been measured, and what haven’t we even tried to measure? How much of policing is driven by crime rates? How do we think about the conditions that create crime in this analysis, and what do we miss when we ignore them? What do we know about the investments that actually make people safe? How do we balance the reality that police do help reduce violent crime with the fury communities have at being overpoliced or victimized by police? How do we experiment with other models of safety carefully and systematically?

There’s a lot in this one. This conversation could’ve gone for hours longer. But these are tough issues, and they deserve someone who understands both the micro-level data and the macro-level context. Goff does, and he shares that knowledge generously and clearly here.

An edited excerpt from our conversation follows. The full conversation can be heard on The Ezra Klein Show.

Subscribe to The Ezra Klein Show wherever you listen to podcasts, including Apple Podcasts, Google Podcasts, Spotify, and Stitcher.

Ezra Klein

You recently said that “the way we set up law enforcement is almost the exact way we’d set up an experiment to make you engage in more discriminatory behavior.” Can you talk about why that is?

Phillip Atiba Goff

Social psychologists, at least for the last quarter-century, have been making our research smaller and smaller and smaller to make it more and more precise. What that allows us to do is to get the exact situations that are most likely to get everyday people to engage in discriminatory behavior. Things like: I’m tired. If you’re tired, your brain gets lazy and you overrely on overlearned associations. So you end up behaving in line with stereotypes.

If you’re multitasking, if you’re cognitively depleted, if your adrenaline is off, if you’re in a new situation, if you’re being negatively stereotyped because of your membership within a visible group that matters to you — all of those things are really robust predictors [of discriminatory behavior]. You can manipulate them in a laboratory and you will get people to discriminate against Black folks and not white folks across a whole host of different dimensions.

But also all of those things are literally the job description for law enforcement. So it is not surprising that we see some pretty significant racial disparities on the other side. It’s exactly what we would predict. In fact, most social psychologists would consider that to be cheating. You just added all the things you knew were going to produce bias and you put them all together.

Except it’s not an experiment — it’s with people’s lives.

Ezra Klein

There is an argument on the right, from people like Heather Mac Donald, that there’s no racial bias in policing, despite what the headline statistics say. And often they cite work like the Roland Fryer study from a few years back. Can you give me your gloss on that? How do we know if there is racial bias in policing? And what can we say about where it does or doesn’t show up?

Phillip Atiba Goff

There is nobody — and I mean nobody — I’ve met who is remotely serious, who has the skills to evaluate research, who’s looked at actual data, who says there’s no racial bias in policing. That’s just not a serious position. That’s like saying there’s no oxygen on Earth and there’s no such thing as the color green. Of course there’s bias in policing. Where there could be a reasonable debate is at what level and at what magnitude. Those are serious questions to which we have the beginnings of some answers.

Roland Fryer, for all of the things that I think are irresponsible in the way he has written up some of his stuff on policing, does not contend that there is no bias in policing. In fact, in the first article that he put out — the one that got all the unnecessary, undue attention — it was really clear there are loads of racial disparities which probably constitute bias in police use of force. For whatever reason, he decided to emphasize that he didn’t see it in use of deadly force. But in every other form of use of force, he finds it.

So we have to set the table stakes here. No serious person is claiming there’s no racial bias. The people who are claiming that are politically motivated — they’re not really set up to evaluate the science, and they’re not interested evaluating the science. There’s not controversy there.

If we’re gonna go to where there’s some unanswered questions, we should be asking: What are the outcomes where we see bias? Where do we have good evidence and not good evidence? And what are the levels at which bias might exist?

Ezra Klein

You say the way to think about this is that it operates at three levels. What are those three levels?

Phillip Atiba Goff

First, let me talk about the overall outcomes. When people are talking about use of deadly force, that’s a really small percentage of an already small percentage of incidents, and small numbers work differently statistically. So I could look at use of force generally and we see incontrovertible evidence of not just disparities but bias. If I look at vehicle stops and pedestrian stops, if I look at ticketing and arrests, those are biased even when benchmarked against the most conservative benchmarks of crime. Other outcomes might not look biased depending on how you analyze them.

But the levels are what we really should be talking about. Police bias works on three levels. First, an individual officer may or may not be biased, may or may not exhibit bias. If an officer treats Black people one way and white people another way — if they treat Latinos, white people, and Asian people differently — that’s at the level of the encounter. If you send Black people and white people out under the same circumstances to encounter the police, how do they come back? It’s definitely the case that we see bias there, but if we restrict our conversations to just biased officers, we’re missing the other two levels.

The second level is probably the most important one: the level of community. Let’s say I’m a robot officer, so I treat everybody I come in contact with exactly the same, but I am deployed differently to Black neighborhoods than white neighborhoods. You’re still gonna see racial disparities. That is a form of bias that afflicts Black communities and poor communities tremendously, but it’s not going to look like the individual officers are biased. The decision-making and the deployment patterns are biased. And we must be able to hold that to account because policing is fundamentally a neighborhood issue even more than it’s an encounter issue.

Ezra Klein

When you say deploy differently, what do you mean by deployment?

Phillip Atiba Goff

It could mean a number of different things. So one thing is, in these neighborhoods over here, you’re going to deploy stop, question, and frisk. That’s one way. But it also could just be: I want 20 officers per square block right per block over in this neighborhood and one officer per 20 blocks in this other neighborhood. Just the concentration of officers is going to lead to different outcomes.

It could also be something like: Everywhere across the city where you see an open-air drug market, I want you arresting folks. Now, I don’t know when the last time was that you were on a college campus, but most college drug use happens in dormitories. It’s an indoor drug market. And even if it weren’t, we don’t treat colleges in the same kinds of ways that we treat low-income Black and brown communities. So the same instruction for an officer to bust open-air drug markets is going to afflict some communities different than others just because of the geographic-spatial elements of those communities. That’s the community level at which bias can operate.

The last one, which I think is harder to conceptualize, is the city level. Let’s say everybody in Philadelphia is getting treated the same at the encounter level, and every neighborhood in Philadelphia is treated exactly the same. But let’s say across the board, Philadelphia cops treat folks more harshly than cops in a city like Bridgeport, Connecticut. It’s actually not that hard to imagine. But because percentage-wise, there’s more Black people in Philadelphia than Bridgeport, you see racial disparities.

That’s also potentially a source of bias: We have a different function — literally a different mission — for law enforcement in cities that are a higher percentage of Black and brown than we have in cities that are a higher percentage white. We’ve got to be able to understand each of those levels.

Ezra Klein

I want to go through some of the objections here. The thing you hear next in this discussion is that this is all generated by crime rates: Officers police more aggressively where there’s more crime, and there’s more crime in Black communities. So everything you see here is simply a function of police going where the crime is. What is the relationship between the crime rate of a local community and police violence?

Phillip Atiba Goff

Before I respond, let me just point out that this is the more reasoned version of the question. What you usually hear is, “What about Black-on-Black crime?” And the other version of it that is more popular recently is, “But what about Chicago?” It’s important to distinguish between the two forms of that question, which are good faith and bad faith. The bad-faith version of that question is: Come on, you know Black people just deserve it. It’s almost expressly bigoted, and it should be treated as such. The bad-faith version so proliferates that people should understand why folks have such a strong and emotional reaction to that form of the question — because it’s meant to indicate that Black people deserve what’s happening to them.

But there is another form of it, the good-faith version, which is: What cops do is go where crime goes, and if there is more crime in Black communities, which some statistics show, shouldn’t we expect that that’s a portion of it? That you have to deal with that seriously. If you don’t deal with that seriously, you haven’t been careful in your methodologies.

But that’s a solvable empirical question. You can answer that by looking and seeing every piece of the best research from the work done by Andrew Gelman and Jeff Fagan to the work that we did with the Urban Institute to the theoretical models of Cody Ross and John Mummolo and Dean Knox. Every one of them says crime is a big predictor of police deployment and police contact. So is poverty, by the way — not that that makes it okay.

So poverty and crime are big predictors, but in all the cities we’ve ever looked at, they aren’t sufficient to explain the racial disparities. Crime and poverty matter, but there are still disparities after that. There’s evidence that there’s still bias after that. In some cities, crime and poverty predict about 80 percent of the disparity; in other cities, crime and poverty rates are about 20 percent. And that means there’s a real difference in how much police behavior and policy is a driver of inequality in policing and therefore in criminal justice outcomes. That’s where the educated conversation in the country should be, but we’re nowhere near that.

Ezra Klein

I’m glad you brought in poverty here, because I want to add in what potentially seems to me to be a fourth level, though it might be embedded in one of the ones you offered. All the methodology operates once a crime has occurred, more or less. But there’s this whole other question of what creates crime itself and what creates criminogenic conditions. We know a bunch about this: Poverty creates crime. Not having health insurance creates crime — you cut crime if you expand Medicaid in different areas.

One of the things that has happened in America is that we have used policy — redlining and segregation and disinvestment and job discrimination and more — to create much more criminogenic conditions in Black communities. The outcome of that has been more crime, and then we aggressively police those communities in response.

We tend to see crime as an individual decision, and to some degree it is, but we also have good evidence that crime is a social and policy outcome. That seems like another way structural racism shows up in this conversation that can be very hard to track and does not show up in a lot of the studies.

Phillip Atiba Goff

That’s exactly right. What we choose to make a crime is deeply influenced by the racial legacies of the United States and the ways in which we create conditions for crime are deeply racially biased. So even if you are benchmarking crime, what you’re benchmarking is an endogenous variable. So even the things that we find are pretty aggressively conservative estimates of the level of bias that we see out there exactly for that reason: Racial bias is already baked into what we’ve defined as how we’re policing various crimes.

The crack versus powder cocaine disparities is the quintessential example. But there’s literally tens of thousands of smaller and bigger examples throughout the codes of different states and different cities.

One of the big problems that we had when we first started doing this work at the Center for Policing Equity was just this — they called it the problem of the benchmark or the denominator problem. If a police department beats up Black people four times more often than it beats up white people, that’s clearly a problem. But nobody knows how much of that problem is the fault of police as opposed to how much is inequality or discrimination in housing and health care and employment and education.

It’s not erasing racism to acknowledge that there are upstream factors that afflict Black communities that are also racist. But our conversations about policing have been so narrow — because our understandings around public safety have been so narrow — that we have erased all the upstream factors that end up showing up as outcomes when we’re talking about policing.

Help keep Vox free for all

Millions turn to Vox each month to understand what’s happening in the news, from the coronavirus crisis to a racial reckoning to what is, quite possibly, the most consequential presidential election of our lifetimes. Our mission has never been more vital than it is in this moment: to empower you through understanding. But our distinctive brand of explanatory journalism takes resources. Even when the economy and the news advertising market recovers, your support will be a critical part of sustaining our resource-intensive work. If you have already contributed, thank you. If you haven’t, please consider helping everyone make sense of an increasingly chaotic world: Contribute today from as little as $3.