Intro. [Recording date: July 26, 2023.]
Russ Roberts: Today is July 26, 2023, and my guest is economist Roland Fryer of Harvard University. He was last here in October of 2022 discussing educational reform. Roland, welcome back to EconTalk.
Roland Fryer: Thanks for having me, Russ. It's great to see you.
Russ Roberts: Our topic for today is diversity, affirmative action, racial disparities. I want to start with a very general question which we could spend the whole time on--we'll try not to--but, what have we learned going back to, say, Gary Becker's book, which was in 1957, if I remember correctly, on The Economics of Discrimination? What have we learned about the source of racial disparities in American education and income, for example? How important is racism, and how can we measure that, if at all? How do economists approach that challenge?
Roland Fryer: Oof, well, you just jumped right in there, Russ. That's a long question, and I'll try to be brief. You're right. The study of discrimination in economics was really started by Gary Becker in his doctoral dissertation at the University of Chicago [Fryer mistakenly said Columbia University--Econlib Ed.]. And, that was in the 1950s. And, Gary's basic view was that discrimination happened because people had a taste for it. The reason that I don't hire you, Russ, is because I just don't like hanging out with you. My utility goes down when I hire Russ. That means I have a stomachache--whatever it is--my utility goes down, so I don't hire those types of people as often. And that's called taste-based discrimination.
There's been another theory of discrimination put out by Kenneth Arrow, and his view was it's not really about taste: It's more about information. It's not that I don't like you, it's that I believe people who are similar to you just are not that qualified or that productive. And, given I have imperfect information about who you are, I don't know if you're good or not until I hire you. Then I will use stereotypes--those stereotypes I have about your group--and I will assign them to you. That's an information-based type discrimination.
A third one that the Sociologists really have come up with and are more keen on is called a structural-based discrimination. It's not that I don't like you. It's not that I have stereotypes. In fact, I have policies and procedures that unwittingly have a disparate impact on you. So, a referral program might be something like that.
Those are the basic tenet--I mean, that's a very broad overview. There are lots of papers in economics and sociology within those categories. I've been fortunate enough to write some of those over the years, but at a broad generalization, social scientists break up discrimination into one of those three buckets.
Now, the question is: How much is there? How do you measure it? Wow, this is a really, really hard question, right? I used to teach a course here at Harvard on the economics of race, and on the first day of class I'd say: Understanding discrimination is equivalent to understanding the causal effect of race on an outcome. Right? Just imagine how hard that is.
Economists have gotten really, really clever over the years at determining causality out of other things because you have random variation here. Sometimes you set up your own experiment. Sometimes you use natural experiments, experiments that happen in natural life that you can exploit. But, that doesn't really happen with race very often, right? It's really hard to imagine randomly changing race.
Now, some people have gotten, again, clever with that. People will randomize names on resumes that are indicative of a certain race. So, it is possible, but is very, very difficult.
In terms of how much discrimination is there in the world? Well, newsflash: It does exist. Okay? Let me be the first one to tell you. It does exist in the data. It exists anecdotally in my own life. Right? I grew up in Texas in the 1980s. There's no way you cannot imagine that there's discrimination in the world in Texas in the 1980s. It was very blunt then. Right?
There's a weird story I'll tell here. When I first moved to the Northeast--to Cambridge--in 2003, I was amazed that someone could send you preferences about their eating habits when you invited them over for dinner. It's just not something that happens--I don't know if it's--Russ, you're laughing out-- but it's not something that happens in the South. Or maybe it's income. We grew up with not a lot, and we certainly--first of all, no one invited you over for dinner. There were no dinner parties in my neighborhood. But, if someone had you over Sunday dinner, you couldn't, like, send them a note being like, 'I don't like coconut. I'm gluten-free.' I mean, we didn't have those things. Okay. I'm not saying--please don't write Russ and say I love gluten-free people, whatever. But you just couldn't do it.
So, when I got here, I invited one of my colleagues over for dinner and they sent me an email with a long list. 'I don't have bell peppers, coconuts.' I just thought, this is the craziest thing in the world. But, that is what it was like to grow up in the South in the 1980s. We would be out playing--we played street football every day; because there weren't a lot of parks, so we played in the street. And it was relatively racially mixed. Okay? At the end of playing football, we would decide maybe we'll go over Russ's house for some snacks. We're hungry. We've been playing football all afternoon. And, every now and then, literally once a week or so, someone like a Russ would turn to me and say, 'Hey, my bad man. Our parents don't allow Black people in the living room.'
And, then there was a decision to make. Right? And, it was so clear. There wasn't a philosophical thing. They were like, 'Okay, well, what kind of snacks do you have? If they're really good snacks, Roland, we're out. If they're not-so-good snacks, we'll choose somebody else's house.' It was that matter of fact. Right?
I don't even know what theory of discrimination that is. Maybe [for--Econlib Ed.] their parents [it--Econlib Ed.] was taste, and they were just maximizing. But, you get the idea. It was a very explicit way; and yes, discrimination exists.
The question really is whether or not it is as limiting a factor in labor markets, and loans, and education as people think it is. And I think that's the big question, right?
My worry is that the whole concept of disparity has turned into discrimination. Not every disparity is discrimination. This is an obvious thing to say, statistically; but it is something that in the ethos--in the culture--over the last two years, I believe, has been purposefully confused.
And even really, really smart people that I admire and respect, when a disparity gets too big, they'll say, 'Well, that's got to be discrimination.' I'll say, 'Well, how do you know that?' 'Well, because it's just too big.'
Well, I'm not familiar with the Too-Big Statistical Test, okay? So, that doesn't work for me.
Really, really smart people have thought about how to detect how much discrimination there is in the market. And, I would say that the basic idea is somewhere between 10% and 25% can be attributed to discrimination. And, the flip side of that is 75-90% is something else.
In other words--let me try to sharpen a point on this. If you're an employer, the fundamental question is: If you see disparities in hiring, is it because people are coming to your firm with the same skills and you were choosing different processes or functions to hire them or criterion to hire them? Well, that's discrimination. If you have the exact same skills and the likelihood of being hired is determined by the race of the person or the gender of the person, that's discrimination.
Or, is it that people come into your candidate pool and they have different characteristics, and to really truly say a disparity is discrimination, you have to disentangle those two things. You have to understand is it the same people being paid different prices or different likelihoods of being hired, or is it different people? Right?
And that is a fundamental question and one that I don't believe gets enough attention. And so, whenever I've seen people disentangle that--well, yes, there is typically, not in every study, but typically seem to be discrimination. But, it is not the major component of the disparity. Okay?
Last thing I'm going to say on this, and please interrupt me or ask me to go deeper on any one of these things. I'm going to try to communicate this well, I'm not sure I can. What makes the--the other thing that makes the study of discrimination so hard, and just observational data: If you have an experiment, great, that's simple. But, in observational data, it's in some sense, Russ, a one-sided test. Meaning, if you find that there's no discrimination, then, sure, there's no discrimination.
But, if you find that, after accounting for education, and job experience, and a measure of skill, there's still a big disparity in hiring: well, that residual, is that discrimination or is that because we didn't have enough data to make better comparisons? Right?
In statistical parlance, we'd say, is that residual discrimination or is it mismeasured x's? Right? In layman's terms, what we're saying is: did you measure education in the appropriate way? Do you have a really great measure of skills, or is it noisy? And that matters. Again, because if we're in a situation where we're innocent until proven guilty in some sense, or we assume the person is not discriminating unless we can find evidence they are, then it's just really hard to pinpoint discrimination without really good data and really great techniques.
Russ Roberts: And we know, if you've ever spent any time thinking about managing human beings and what productivity really is, there are many, many skills that are intangible. One of my favorites is integrity, seriousness of a person: can't measure it. You can observe it in casual interactions over a period of time. If you're a really good interviewer, you might be able to get at that in a job interview. But, the skills that we have data on are never precisely mapped into the productivity of an employee in the workspace.
Of course, that always leaves leeway to be a racist or other--or sexist or antisemite--but it's a fact, as you say, that these things are very hard to measure. I think culturally the challenge for us in general is thinking about what the default is, whether it's innocent until proven guilty or guilty until proven innocent.
But certainly, to take the first step--which you've written about in your own exploration of this--and saying, 'Well, how could I begin?'--forget final--'How could I begin to get at what parts of this are racial versus differences between human beings?'
Then the next question, of course, should be: If it is differences in human beings, what could we do to change that? And we'll talk about that later, I expect. But I think that's--inevitably impossible question to answer precisely--but we can get some information on it, is how I would describe it.
Roland Fryer: Well, Russ, your point about innocent till proven guilty or guilty until proven innocent, that is a fundamental point.
And, the reason--this is a point I used to debate with my grandmother--and the reason that it is, in America, innocent until proven guilty is because discrimination has typically been handled in the courts. And that's the way the courts operate. But it's hard as a human to imagine that in 1963--before the Civil Rights Act, when it was okay, or when it was more tolerated to discriminate--when we flip that and go 1964, 1965, that we also flip the null hypothesis. That's a strange thing to imagine, just historically.
So, I get it. I had a statistics professor in college--when I first--like, my love of economics had just, like, taken hold. Like, I've had a love affair with economics since my first class in 1995. And, the professor said something like, 'Life is all about who gets to claim the null hypothesis.' And I thought: That was one of the most beautiful statements I've ever heard, whether it's relationships or the study of discrimination. Really, who gets to get the null hypothesis? Who gets to claim the null, has a weighted advantage.
Russ Roberts: And that's--for the non-statisticians in the crowd--the null hypothesis is your default that you're comparing other things against. It's sort of your baseline; and that can change everything.
Russ Roberts: Talk a little bit about your grandmother and her attitudes on this question. How you--you've written about it very poignantly about how your attitudes changed over time because you tried to measure them--what we're talking about.
Roland Fryer: Well, my grandmother partly, but at least emotionally, fully raised me as a kid. She passed away in 2016--actually 2014, sorry. She liked to tell people that she brought me home from the hospital, which is when I was born, which is true. We had a phenomenal relationship, a phenomenal relationship; and she was an amazing human being. She played basketball for Bethune-Cookman University in the 1940s, late 1940s. When I was a kid, we always thought we had plenty because she had two or three jobs--always. And she just had this--life was simple, to my grandmother. There was right and there was wrong. And [?fought? thought?]: she integrated schools in Florida in 1969. And, in going to school, she lived in an all-Black neighborhood till the day she passed away. I could never get her to move to an integrated--didn't want any of it. She had lived in an all-Black neighborhood. And she was asked to go integrate a school that was 25 minutes away in an affluent suburb, and she didn't.
And, to hear her tell her stories of being spit on and rocks thrown at her walking from her car to get to a class, it's incredible. She was just a phenomenal human, Russ.
Her first day of school when she integrated--imagine this, first day of school when she integrated--she's lining up the kids to go to--I'm going to get teary-eyed thinking about my grandmother. She lines up the kids to go to lunch and one kid is a little unruly in line, so she grabs the kid's hand and she walks with him as the class walks to the lunchroom. First day of integration. And obviously the kid is white. And he takes my grandmother's hand, and he puts it in his mouth, and he bites her.
And, my grandmother, as she tells the story, didn't really flinch. She just allowed him to finish. And, she asked him, 'Are you done?' And, when he was done, she looked at her hand and it had teeth marks all over it. And she flipped his hand over and she bit him back--heh, heh, heh, heh, heh [laughing]. Raw. I know people listening are going, 'This is the most insane story ever--.' I guess that's how it was in the 1960s.
My point is: She suffered no fools, and she refused to be treated poorly by anyone. And, she had a very large sense, and clear sense of what her values were. And we used to go to church every Sunday--yada, yada, yada. I love my grandmother.
Anyway, we--our relationship was phenomenal and it taught me how cool it can to have creative friction, I'd call it now. We had creative friction, me and my grandmother.
We disagreed a lot. I was a little precocious kid who, I think, if she had known a lot about economics, she would have realized very early on I was an economist. My grandmother used to drive around to every gas station trying to find cheaper gas. And, I told her at some point, like at age eight, 'Don't you--why are we--doesn't it make sense to think about the gas we're spending trying to find lower gas?' Right? My grandmother said to me, 'Just stop talking.' Right? It would frustrate her.
But these are the kinds of questions I would ask, and we would debate them all day long. I was the only grandchild; and so I grew up hanging around with older, mainly black women, who--they would sit and talk about race in America, continuously. And I had the great fortune of sitting on plastic covered couches, drinking sweet tea and listening to this wisdom.
And so, when you listen to that kind of wisdom, inevitably as I grew up, I viewed the world as being rampantly racist. The view in my grandmother's living room was that you could spot a racist on sight. You could just look at them. They clearly are racist. And so, I subscribed to that view. I didn't have any other--I still think my grandmother walks on water or she did. And so, that was the view I had.
On the other hand, I was really, really--because of those conversations, I was really, really interested in these areas.
And so, when I got to graduate school, I had zero intention, Russ, of studying discrimination. I don't know why I was there. I think I was there not to get a real job. I did pretty well an undergraduate. I graduated in two and a half years. I went to graduate school at age 20. And so, I didn't know what else to do at age 20. I seemed to be pretty good at economics, and economics was fantastic to me. So, I decided to keep going.
And I remember going into my first Price Theory class in graduate school and being amazed.
But where my mind was really blown was my first Game Theory class. And they set up an extensive-form game on the board and he wrote, you know, 'You have choice A or B, and once you choose A, then the next choice is B or C,' or whatever. And so, you go down this extensive form; and at the end of the extensive form, there were numbers: the utilities, the payoffs. And so I, was so confused--the power of this, it was like it had glowed off the chalkboard. It was three-dimensional, in my mind. I asked the Game Theory professor, 'Wait, wait, wait. Why are[?] you accounting for the following?' I gave a bunch of complicated things. 'It's in the number six.' Wow--like all of that, all the clutter that was in my grandmother's living room on those plastic slip covers went away.
The world was clear to me. And I was pretty damn certain that anything could be modeled in an extensive-form game.
And, I went at it. I'll never forget it. I left that class and I went home and I started writing what would soon to become a paper called "A Dynamic Theory of Statistical Discrimination." I wrote that in my first year of graduate school, the spring semester, after learning about extensive-form games. And that paper was, 'Hey, I had the following intuition. If you have multi stages in a game and you are discriminated against in the first stage of the game, actually you're going to be discriminated for, potentially, in the second stage of the game.'
In other words, if you understand that the labor market has, without any regulation--let's take Affirmative Action, all that, out of here--without any regulation. If you see a woman or minority pilot in the 1950s, that must be the best pilot you could imagine, because of the hurdles they had to cross to actually get to be in that position. And so, therefore, conditional on being a pilot, the people I want to promote are the actual minorities and women. Right? It is just a simple conditional distribution argument, but you could see it very beautifully in this extensive-form game. I wrote that out, etc. Actually, ended up publishing that paper.
There's some evidence of that. If you look in data for large enterprises, you see exactly that phenomenon. Yes, there seems to be discrimination in hiring, conditional on hiring. You'll see a lot of times--not always, depends on the industry--you'll see that conditional on hiring, actually the people who are most likely to be promoted are sometimes women and minorities. I showed that data was true in a few enterprises in the United Kingdom as part of that paper. But you get the idea.
I was fascinated. And so, these two worlds came colliding because there was a world of pure beauty of these extensive-form games that was clear. The end of an hour you could circle something. You could solve for an equilibrium and say, 'This is it.'
And, there was another world which was this intuitive, is there, is there not? I can feel it, obviously. That kind of thing. And, I respected both.
And honestly, they'd never felt rivalrous in me until I had a phenomenal study partner one night, and at 2:00 in the morning--we're in the Sparks Building, Penn State, that was my first year of graduate school--and he was brewing coffee. I'll never forget it: he was brewing coffee. And I hated his coffee because we were so cheap, he would never empty the coffee and put in new coffee grounds. He'd just sprinkle some on top and re-brew it, and it was disgusting. But, it was two o'clock in the morning. It's funny, the things you remember, right? It was two o'clock in the morning and I needed something to keep up, and so it was me and my buddy. We had just met a few months ago. We had done all this math together, but we really didn't know each other, so we started talking.
So, I told him a little bit about where I grew up, and he told me that he grew up on a farm in Illinois. And we had had a shared comradery of cow tipping--I grew up in Texas--and we talked about all the things that rural people do. So, I felt relaxed around him, and so I decided to tell him I knew one of our professors was racist. And, he says, 'Well, how do you know?' I said, 'Well, you can just tell.' And so I gave the kind of logic that my grandmother would give on those slip covers, drinking that sweet tea that always seemed to pass muster to me.
And I said that to him, and at two o'clock in the morning in the Sparks Building over the worst coffee you've ever had, he spit out his coffee, laughing in my face. 'That is the most ridiculous thing I've ever heard. Who could ever say anything like that? What do you mean? Am I a racist? Do I look like it? What if I turn like this? What if I turn like that?' He just had fun with this. He says, 'Dude, I don't get it. On the problems on the board you are as precise as can be. But, when it comes to this, which is the best application of this problem on the board we've seen all semester, you're using your intuition and not the mathematics.'
That was probably the best lesson I got on economics and game theory in my entire career, man. It was great to have another person say that" 'You just believe these things? You're totally full of crap.' Whatever.
I'm sorry, this is getting too long, but I set out to say, 'All right, let's combine these two worlds. Let's see if we can be really rigorous, but also use our intuition. Let's not put our heads in the sand.'
Then I just was on a quest to understand that and read a phenomenal paper by Derek Neal and Bill Johnson. Derek Neal was at the University of Chicago at the time. Bill Johnson was at the University of Virginia. And similar thing: it felt like an attack on what I knew. They were basically saying there's not a lot of discrimination in the world. They said, 'Look, if you look at the aggregate data, a third of the wage gap--there's a 33% difference in the wages paid to black employees and white employees.' That felt about right to me at that time of life, 33%. Then they said, 'Look, if you account for one measure of skill pre-market--pre-market, nothing to do with the employer--that 33% difference goes down to 7%.' Okay? So, yes, there's still discrimination. Maybe it's 7%, maybe it's 10% depending upon the specification.
But, wow, 7%--that's very different than thinking a third to a half of the disparities are that way.
So I stayed up late at night trying to attack that paper. I was sure that they were racists, themselves. Etc. And, it just made me come to realize that the rigors of economics are phenomenal, and unfortunately most of our colleagues are applying them to problems like the optimal cake-eating problem. Here is a space where a lot of discussion goes on in American and other living rooms around the world about how much discrimination there is. Now that I'm old-er, I have seen this discussion play out in living rooms across the world, right? The Turkish immigrants in Vienna. It's the same discussions we were having in the 1990s about whether or not--is it their effort? Is it institutional discrimination? It's the same stuff, no data. Or whether it's the African immigrants in France: same thing. Or the African immigrants in Spain: same thing.
I see these things replicated over and over again. And I just decided--in whatever it was, 2000, 23 years ago--that I was going to use my grandmother's intuition to root for a result or to have expectations for a result or to have intuitions about what the data may say. But that there was going to be nobody else on the planet more rigorous when it came to studying discrimination than me. And I'm not sure I ever meant that, but that has been my quest for the last 23 years.
Russ Roberts: And God bless you. When you talk about your grandmother, when you mentioned in passing--this is a little digression, but it's related. When you said in passing, 'and she always had two or three jobs,' and I thought, 'Hmm, that explains something about Roland Fryer. I don't know him that well, but he works pretty hard.' I'd say he works very hard. And then I thought, 'Now I know where he got it from.' But, if I tried to try to figure out whether that was nature or nurture, I think we'd have a data challenge.
Roland Fryer: Yeah. It doesn't[?] say your data challenge, right?
Russ Roberts: Is that your innate character or did your grandmother role-model that for you? Who knows? Hard to know. But, whatever it was, I think she had something to do with it.
Roland Fryer: A hundred percent. And that's just the way she was. When she was in her 80s, she was out campaigning for Obama. We would say she has a high motor. She was always [inaudible]. She passed away at 89 years old and was only ill for about six months. But before that, she was out campaigning for the local politics, NAACP [National Association for the Advancement of Colored People], whatever she could do. And, I always found that inspiring. I didn't find her statistical rigor inspiring. For my grandmother when I would share with her my results, my results were always put into two categories: obvious and wrong. There's no one who has pushed me more on these subjects than her, in part, because just trying to understand where our intuition is different.
I even wrote a paper about how one interprets signals. Because, my grandmother and I could have the same interaction: one of us could think it was racism, and one of us think it's something very different.
I wrote about this before, but I'll say it quickly. The one instance that comes to mind is we went into McDonald's because McDonald's had two-for-$2 Big Macs. It was a very exciting time in the life of a young man. We went in and paid for the Big Macs. My grandmother gave the money to the cashier, and the cashier put the change on the counter. My grandmother picked the change up. We walk outside. And, my view--again, as a little precocious nerdy kid that didn't know he was quite a nerd yet--was, like, 'Hmm, that's really sanitary. She didn't want to pass germs back and forth.' My grandmother said, 'She doesn't want to touch me.' And, was irate about it.
Again, if you think of game theory: we don't even allow that in our assumptions--that you and I can see the same signal and interpret it very, very differently. And so, I wanted to understand how two people could come to the same interaction, see the same signal--publicly observable signal--and have two very different interpretations of that.
And so, I wrote this paper with my good friend Matt Jackson out at Stanford where it was a form of information-based bias, but it was you interpret signals as a function of your prior beliefs. Okay?
So, the question was how do you use Bayes' rule when the signals are unclear? Our idea was that most signals in life are unclear, so we need to figure out how to do Bayes' rule. And so we did a simple way of thinking about it, which was a double updating. First, when an unclear signal comes in, you updated it to be a clear signal. Is this an A or B? I don't know. First you use your prior to update that: decide which ones--is it an A or B--and then once it's an A or B, then you update your beliefs accordingly using Bayes' rule, right? So it's a double updating.
We showed this in experiments. It's kind of halfway interesting. Remember Armand[?], and all the old school game theory was essentially that your colleagues there in Israel, the idea was if we have different beliefs and access to the same exact information, those beliefs will converge at a high level. Yes, there've been adaptations, but at a high level, that's what it was.
And what we showed was the exact opposite. We went into the lab and did an experiment on climate change, on Affirmative Action, on the death penalty, and we took people who we screened to have different beliefs, right? You believe climate change is manmade. Some other people believe it's not. And, we gave them the same information. They left the experiment more polarized, because they each interpreted it as being evidence in support of what they initially believed, and then they diverged in the way that Bayes' rule would expect.
Anyway, that's a long-winded way of saying she's been an inspiration for a lot of my work because of how we interacted; and just imagine a kid in graduate school learning all these new techniques. They were like golden tools to me. And then, going back to a neighborhood where those tools, I thought, were desperately needed, but no one knew them and they had a very different manner of coming to conclusions. And trying to fuse those worlds, that's what was so cool to me about my early days as an economist.
Russ Roberts: And that's, I'd say, a microcosm or a template for how your professional career generally in every area. You are a relentless pursuer of truth with data. I think the challenge--and you might reflect on this briefly, because we have a lot of other things to talk about--but I think part of the challenge is: as we've already alluded to, so many things that we care about are not easily measured. So, obviously, we care about trying to assess discrimination: how much of it comes from the characteristics of the applicants, say, versus the characteristics of the managers--as either racist or not--and characteristics of the applicants--as skilled or not, or more or less skilled. And, we can't measure that precisely. And so, we're inherently limited.
And, as you know as well as anyone, the way you get ahead in our profession is by making dramatic statements, and nuance is not always rewarded. It rarely is. It's probably rarely rewarded anywhere, not just in economics.
So I think that's always going to be the challenge. We want to find the truth. We want an answer. Accepting that we can't always find that answer is not comfortable. So, I think sometimes we overstate the value of what we've discovered because it's easier. Sometimes.
Roland Fryer: Yeah, I think that's it. I think I have a simpler version, Russ, which is: We're just lazy. I actually don't subscribe to the view that we can't measure these. I think I subscribed to your view--you said it carefully--you said: you can't measure them easily. That's true, but we can measure them, or at least measure them a whole lot better and make better conclusions.
The problem with this and Affirmative Action, the way it was practiced, and our general view, in my opinion of racism and sexism in America and maybe around the world, is: we're looking for quick wins. Right? We're just lazy as heck. And, this is not the time to be lazy. That's what's so frustrating, is that we could solve these problems if we actually dug in and got serious. But we just want to kind of play around on the periphery.
Russ Roberts: So, let's turn to a controversial claim you made about diversity. The work we've been talking about so far is what in economics we call positive--how does the world actually work and what are its characteristics? As opposed to normative--which is how would we like it to work. I always like to bemoan the fact that those two words are not very helpful, positive versus normative. Only economists know what they are.
But I'll say it one more time. Positive means what is the world like actually. And, normative is how do we wish the world could be or would be.
Social scientists generally look at positive things to try to figure out what's happening, and then they turn to normative things, like, 'How can we make it better if we don't like what's happening?'
So, diversity training, Affirmative Action, these are policies that have been put in place by corporate America and universities to try to address the disparities that we're talking about. And, as you pointed out in your writing, a lot of the times they don't care what the source of it is. They're just going to bluntly do something about it. You wrote that diversity training approximately accomplishes zero, and maybe is negative. The question I'd ask you is: What could we do better, if we wanted to make this a better world in that area? What do you think corporations should be doing? Then we'll turn to universities, because you have some very interesting thoughts there, too.
Roland Fryer: Great. Yeah, I just want to be clear: it's not, this is not a claim I was making. It was research I was describing, that a set of sociologists have looked at--over a thousand corporate trainings, the results of them, even in the best conditions in terms of research design. And, what they found is that the average impact is zero for corporate trainings. And ones that are mandated, the average impact is actually negative on the future hiring and promotion of certain minority populations within those companies. So, that is not a claim, it's just a [inaudible] for researchers.
Russ Roberts: Yeah. Sorry.
Roland Fryer: Okay.
What should they be doing?
Well, what we've talked about so far here has been a roundabout way of understanding the theoretical and underpinnings of what they should do. Right?
So, after George Floyd, when a lot of the companies were signaling how much they loved this or that group, I got really frustrated with that because I thought to myself, 'Why aren't they doing all these techniques that we know about in economics?'
So I called a former graduate Ph.D. student of mine, and I asked her, I said, 'Why are they not doing the Becker outcomes test?' Which is a test to understand whether or not there's discrimination--taste-based discrimination--in hiring practices? Right? Just like when we were kids playing street football, that stuff, same thing, but substantiated into a test. And, she also laughed at me. I'm detecting a pattern. She said, 'Look, Roland, no one reads arcane journals for fun besides you.' Maybe Russ.
And so, what do I think they should do? I think they should use the insights that social scientists have developed over the last 50 years that provides a very, very direct roadmap of how to make progress on these issues.
Now, are they going to go read arcane math journals? No.
But, that's why we developed this software to make it really easy for them. Right? Where we actually substantiate the software--the Becker outcomes test--where we substantiate in the software, Arrow's test of information discrimination that we talked about earlier. Where we substantiate in the software this double updating, where you can see if people are using biased priors when they actually make decisions or when there's imperfect signals. All of that substantiated in the software.
So, what does that mean? It means that companies have a lot of data right now and they keep data on the types of people in their applicant pool, who was hired, how they perform, who gets promoted. When there's attrition, what the demographics of that looks like.
And so, you can do with the types of analysis that has been done time and time again in the best economics journals to understand whether or not there's bias in hiring. You can use those same techniques using actual live employer data.
So, that's what I've been doing over the last couple of years. And, it has been a phenomenal experience because you get to see with large corporations, what's actually going on. And more importantly, how to actually help fix it. Right? So, what do they need to do? Very simple, diagnosed; and then apply solutions that are relevant for that.
Why is that important? Because, remember in the beginning, Russ, we talked about Becker's theory--taste-based--then we had information and then we had structural. Okay. Well, how to solve those--how to solve bias--really depends on which type it is, and what is happening.
A lot--what I see over and over again in corporations--is that they detect a disparity. They don't even know if it's a bias yet. And, they just start applying things: Let's have an employer resource group. Let's have an affinity group where people can get together in the cafeteria. Let's do this, let's do that. There's nothing wrong with that, per se. Let's do training.
And then, they don't see a lot of movement on key performance indicators. Then they get frustrated. Right?
But that's equivalent to how my grandmother used to diagnose and treat. She just gave me NyQuil no matter what was wrong. 'Grandma, I have a stomach ache.' 'Take some NyQuil.' 'My ankle hurts.' 'Take some NyQuil.' It didn't matter, right?
And that's essentially what we're doing in DEI [Diversity, Equity, and Inclusion], to tie all this together.
And, my approach, and the economist approach, is quite simple, but very different, which is: First detect the disparities. We have to answer that first question we talked about at the beginning.
I'm going to try to tie all these last 30 minutes together. We have to answer that first question at the beginning. Is it that people are coming to your company with different skills and you're pricing them accordingly? That's not discrimination.
Or, is it that they're coming with the same skills and you're pricing them differently? That's discrimination.
So you have to figure out: Is that disparity really bias? Step one. Step two, what type of bias is it? Right? Let's actually run the statistical test. You remember--Altonji and Pierret, and all these other statistical tests--I could talk about those in a second if you want. But it helps determine what type of bias it is. And then once you know, then we can curate solutions specific to this type of bias that's in this specific company. And then, you try those solutions. Some of them could be AI-based co-pilots for hiring or promotion or what have you. And, we monitor them, come back in a couple of months, three months--depending upon the frequency of the outcome that we're looking at--and measure our progress. Right? But you could do all of this in a very, very data-centric way.
Russ Roberts: Well, let's take a practical example. Let's say you have a particular division in your firm and it is very disproportionately one race or one gender--and, as are the managers. Okay: so, it's not just the overall number of employees, but it's also the management of that. And so, you find out, let's say--so you could find out either after you've done the analysis you're talking about, that the applicants to your firm are--let's use race--that the black and white applicants are equal quality, or you can find out they're not.
So, what would you do differently based on those two? Because that's a good thing to look at. You don't want to just look at the ratio or proportion of people that are hired. That doesn't tell you anything--anything conclusive. It might tell you to look, it might alert you to an issue; but now you have a decision to make. Let's say you discover you don't get very many black applicants for these positions, and the ones that you do get are not as good as the white applicants.
Russ Roberts: Now what?
Roland Fryer: Yeah. So if that's the case--so, let's take it one by one. If it is the applicants are just as good, but there are issues in the hiring, then we would want to understand what the issues are in the hiring. We'd want to put it into those different buckets and treat it for what it is.
For example, imagine it's information. Then what we would do is work with the company to get more information at the time of hiring. That's what helps information bias, is: providing more and more information, because then the sterotypes get smaller; and because what you're needing to apply your stereotypes to get smaller and smaller.
Now, juxtapose that to this implicit bias stuff, right? Which is going in the opposite direction. I have seen a lot of people who don't diagnose what type of discrimination it is. Just imagine that it's implicit bias, and what you do there is take away information. Well, if it's information bias, you're making the problem worse. Okay? Again, that's an example of why it's important to understand the type of bias that is undergirding the disparities. Okay. But, in that case, we give more information, we monitor. We can even put in a co-pilot with hiring that says, 'Hey, given this person's characteristics, here's the predictive performance score.' And we can make sure that's done--
Russ Roberts: By co-pilot, you mean a non-human--
Russ Roberts: algorithm?
Roland Fryer: A machine learning algorithm that says: Historically, here's been the relationship between characteristics that we know at the time of hiring and performance. Okay? And so, for every applicant that comes in, I'm going to show you what the expected performance is over the next one year, three year, five years. Okay? Then you could just hire based on expected performance. You don't even need to see the race. That would be one way to make it more meritocratic.
Now, if it is your other example--which happens a lot--is 'Look, it's a supply issue. We treat everyone the same, but we don't get enough in the door.' Then we go look at sourcing channels. Right? Like, 'How are you advertising? What does the spec look like? Is it the right spec?' We've worked with a couple dozen employers, Russ: I've never met one who has the right spec. Meaning, if I look--right? because I love talking to you because it's so simple[?]--if I look at the characteristics that they collect of the person and I correlate those characteristics with performance, and I say, 'Okay, which characteristics are the most important?' And then I look at the spec and say, 'Which ones are they actually asking for?' They don't line up. Right?
And so, those are the kinds of things that we can--that this software can help an employer do. Which is, like, figure out what spec I should have for this job, given historical performance in my own company. Right? It sounds simple, right? But it's not often done.
Russ Roberts: I mean, it's a beautiful example because in any organization, there's a huge amount of history. I'll call it prejudices--most of them aren't racial. They're just stupid or blind. They're just blind. 'This is the way we've always done it. It always worked for us.' And the idea of, at some midpoint, saying, 'Well, now let's look and see if it worked.'
Russ Roberts: It's too hard. Part of it's lazy, but a lot of it, I think, is just the human urge to delude yourself into assuming you've been doing a good job.
Roland Fryer: We don't even do it in our economics department. We're a bunch of nerds. For years, I was on the Admissions Committee and I was like, 'Hey, let's collect a bunch of data and figure out what adjectives in a recommendation letter actually correlate with first-year performance.'
Russ Roberts: Interesting.
Roland Fryer: We never could do it. Or, 'You know what? The same types of people recommend students every year. Shouldn't we weight them based on how good their recommendations are, historically?'
All these are obvious things that we don't do. So, I am with you. I don't think it's prejudice in the normal sense. It's just--I'm not going to call it laziness either. It's just it's not the way business has been done, right?
Here's a fun example. Right? I worked with a very large financial institution a couple of years ago. And, they had me to come in just to chat with their executive team, but they were sure they had no issues with their specialization, because they were, 'Look, Roland, there's no way we could be biased because all we ask for is a 3.8 in Applied Math from an Ivy League school. And so, anyone who goes over that, we take a very careful look at.' Okay.
And, I said, 'Huh, that's interesting. Why don't you introduce yourselves and tell me about your backgrounds?' Right? 'Well, I went to University of North Dakota in music.' When you went around, none of them had that spec. Right?
And so, actually, what we ended up doing was a project on leadership development where we wanted to understand, for all the top 100 leaders in this large--or 500 leaders--in this very large financial institution, what were their paths? And so you could put probabilities on the paths to leadership, and then ensure that women and minorities were on the right path to leadership. Right?
I did the same thing for the NFL [National Football League] when it came to black coaches just about a year or so ago, because the idea was: Is there discrimination in the NFL when it comes to hiring black coaches?
To do this, you got to collect a bunch of data on every NFL coach that's ever coached and what their actual career trajectory was before they got there. It's the same thing. It's these leadership paths. How do you actually get there and are minorities, etc., on the paths? Is it that they are on the same paths that lead to leadership but just passed over, or are they on different paths? Right? And the answer--a lot of times--the data tells you, is that they're actually just on different paths. And that is a easier problem to fix than changing the preferences or what have you on the paths. Does that make sense?
Russ Roberts: Yeah. No, I know exactly what you mean.
Russ Roberts: Well, before we end, I want to bring us back to our last conversation where we talked about educational reform. The Supreme Court recently in the United States struck down Affirmative Action at the college level. You wrote a remarkable piece in the New York Times. You wrote, actually, two pieces. You wrote a piece, I think, the first one was on--it might have been in the Washington Post--but the first piece, before, the decision was basically: Universities are worried that they're going to be unable to use race-based criteria. They're getting ready to lower their standards--reduce or what we would traditionally call lowering standards will be neutral--change how they admit people and reduce their reliance on, say, SAT exams, where historically black students have not achieved as high scores on average as white students. So, let's just not use those, because they're obviously as a measure of skill biasing our pool away from this group that we want to help.
You wrote a profound--so indeed, the Supreme Court did strike it down, Affirmative Action. And, indeed, many colleges--well before this, by the way--have been changing and altering what are often thought of as subjective--they're not, of course--but things that result in a number, whether it's an SAT [Standardized Aptitude Test] exam or something else.
And you came up with a rather different approach. You said, I'll quote:
Elite colleges could operate a network of, say, 100 feeder middle and high schools--academies that are open to promising students who otherwise lack access to a high-quality secondary education in cities where such children are common because of high poverty rates and underperforming public schools. These institutions would bring their students up to the sponsoring university's standards so that they're ready for elite schools when they graduate.
Meaning elite high schools--excuse me, elite colleges when they graduate high school. And then you write, this is the part I love:
To undertake such a project, elite institutions would have to believe two things, that they can afford it and that there are enough Ivy League caliber students trapped in poor performing high schools to make it worthwhile. Do they believe that?
you asked. I don't know the answer--that was a rhetorical question--but we learned something about their beliefs in the response to your piece, which I haven't read anything about elite universities starting any high schools. I don't think they're planning it.
Roland Fryer: I don't think it's in the works.
Russ Roberts: Yeah. I mean, I didn't take that as a piece of satire. You could take it as a piece of satire: Of course, they're not going to start a network of feeder schools. It's too expensive. That's not their core business. They're college. This is high school or middle school.
The reason I thought it was profound is--it's exactly what you said. You could respond to this judicial decision by either changing your standards or trying to change the people--your applicant pool--which is what we talked about earlier. I know which one I'd prefer, and I know which one Roland Fryer would prefer. Why are we so lonely?
Roland Fryer: I don't know. I don't know. Maybe it's willingness to do the hard work. I'm not looking for the most efficient solution yet. I'm looking for a solution. It's been going on for so long, and once we find a solution, then we can quibble over whether it's the most efficient one.
But, I can tell you this: it is not a piece of satire. It is a 'put your endowment where your mouth is'-kind of piece. It's very easy to do Affirmative Action in a lazy way. And as I understand it--and I'm probably very naïve--that's essentially what the American public and the courts were reacting to, is the lazy way in which it was implemented. Not that they don't believe that we should be out there finding gems or diamonds in the rough. Affirmative action is super-complicated for me, because I'm pretty sure I was helped along the way by it. I needed help. I love my grandmother and I talk about her all the time. But, the truth is, I grew up in a very difficult circumstance and she was helpful, but not all guardian[?], and I needed help.
As I look back, I can tell now with the benefit of hindsight that I was a kid who could have benefited from a lot of this type of investment. And so, I'm very grateful for it. And I don't want to sound like a person who walks over a bridge and then burns the bridge and says, 'You guys take a different route.' It gives me pause, and it gives me a little bit of goosebumps when I hear successful Black Americans say, 'There's no discrimination in the world. Just pull yourself up by your bootstraps.' That's not correct. There is discrimination in the world. But, effort matters. Okay.
Now, how does that apply here? I have two daughters who I adore. If they need Affirmative Action, it's not because there's systemic racism. It's because they suck. Because, they have every advantage in the world. Right? We're busting our butts, my wife and I, to make sure they have a life that neither one of us could ever imagine. They go to the best private schools we can find that. Every whim they have is a passion. We spend more money on lacrosse sticks and ballet shoes--whatever we have to do--to fuel their passion. If it turns out they have lower SAT scores, Russ, I just fundamentally don't believe it's because of discrimination. I think it's because they're just not that good; and I don't think Harvard or any institution should admit them if they can't cut it.
On the other hand, they are the anomaly. They'll be fine. But, there's a lot of kids who grew up like I did. Some of which have great grades and not the greatest SAT scores because they didn't go to the best schools. And I think those kids deserve a shot. And, if we're not going to be able to give those kids a shot through the way that I think Affirmative Action should have been practiced--which is hustling to go find the diamonds in the rough, the kids who overcame a lot to get the same SAT score--if we can't do that, which the courts have, I think made it hard, although there was a stipulation in there. I'm not a legal scholar, but it seems like there's a way to do kind of that. Then let's go take supply into our own hands. I've been at Harvard for 20 years, not to exaggerate, but 16 or 17 of those years, I've asked the same question. How are we going to get more diversity? And, by that I mean more Black people among the faculty in our department, be that blunt. And, the answer is always, 'It's a pipeline issue.'
Pipeline. Pipeline. Supply. I get it. I do get it. The New York Times cut this part of my piece because, I guess because it was too long. But, I get why Google doesn't take on computer science education. I'm an economist's economist. I understand public goods.
What I don't understand is why the Ivy League schools can't change supply. I think they can decide: 'Hey, let's intervene in ninth grade, or sixth grade, and let's set up a situation where kids who have that raw talent have a place where they can reach their God-given potential.' And, if you do that with 50,000 kids across America--500 kids in 100 schools--I just don't have any doubt that we can find 5,000 of those--10%--that would need no special treatment to get into Ivy League schools. Yes, Harvard may get a little bit more of their share than Columbia or whatever, or whoever the hot school is at that moment, but essentially they would all benefit from this.
And so for me, having benefited from these things and just watched over time how we treat Affirmative Action, I really believe that Black kids in America who are in tough situations--broken homes, whatever, you know all the things I know--they just want a chance to compete. Give them a chance to compete. It's all we've ever asked for. Look at Du Bois, Philadelphia Negro. Look at what they ask in the back of that--in 1895, they asked Du Bois, or he was in a kind of question-and-answer, 'What should Black people ask from white people?' In 1895. And the answer was, 'A chance to compete.' RightB
I read that now to my students. I teach a class on Black geniuses. I read that passage, but don't tell them it's Du Bois, and they think it's Tim Scott, some other Black Republican. That's another show. Invite me back [inaudible]. That's a whole different thing. We just want a chance to compete. It is insulting who imagine that I need a different standard. I don't. I just need the same opportunity. Or not even same: even close to the same opportunity. I mean, I have no doubt that those kids can compete. Right?
And I wrote in the piece and I meant it--it was not a literary, whatever; it wasn't hyperbole--I wrote that in my grandmother's neighborhood I have seen gallons of talent wasted. Gallons. Right? My best friend when I was just a teenager went to prison for a very long time for shooting the person at the corner store. I've seen gallons of talent wasted.
You fast-forward to the schools my kids--I mean this, I don't know if those kids got talent or not, but I call it teaspoons of talent nurtured. I mean, everything is nurtured. And all I'm saying is if we treated talent the way we do in our elite independent schools in America, we did that for 100 schools in inner cities, we wouldn't have an Affirmative Action problem. We wouldn't have an issue. We'd given them a chance to compete, and they would thrive. And I have no doubt that that's the case.
So, my piece was about the two things they have to believe is affordability--give me a break, right? Just, come on. The cost of this wouldn't even cover 40% of the growth--not level, the growth--in the endowment of the next 10 years. Just the growth. And, if they fundraise, you know what? Ivy League schools want to do it, I promise I'll take the next year off to fundraise for them, like, 'We can do this.' But, they don't want that. But, I haven't offered.
And so we're left with: If you really fundamentally believe there's talent trap there that we have to go get, then let's go get them. We can change supply. For me, this is a phenomenal opportunity and chance to take something the court did that made us all, like, 'Oh my goodness, what's going to go next?' And say, 'You know what? Let's fix the problem. Let's change supply. Let's create opportunity. Let's do all the ideals that we say that we're about.' Right now. Because, the kids aren't going to get ninth grade back.
So, as I wrote in the last sentence of the piece, 'If you're wondering in four years what fraction of minorities we will have--black students we will have--at Harvard or Yale,' that's up to Harvard or Yale. Right? We can actually solve this problem. I fundamentally believe that. Of course, some people have said, 'Great.' Some people have said, 'You're crazy. This will never happen. It's satire.' Yada, yada. It may never happen, but it could. And, if it doesn't, then we have no one to blame but ourselves.
Russ Roberts: My guest today has been Roland Fryer. Roland, thanks for being part of EconTalk.
Roland Fryer: Anytime, Russ. Really great to see you.