Can we fix discrimination in markets?

Ian Ayres: One of the areas of discrimination in markets that I've written about is disparate treatment in retail stores. I got a call from the attorney general's civil rights division in a major office that you guys would all know. They're interested in having studies done on testing for discrimination in big-box stores. They were calling to find out what tests they should do, and if I would help.

They were worried about disproportionate surveillance — and investigation — of minority shoppers. But I was trying to have them broaden it to include idiosyncratic accommodation. I wanted them to go in places that don't have a restroom and say, "Can I use the employees' restroom?" Or go in with a kid, and say, "My kid needs to use the restroom." Or "I'd like to take more than three items into the dressing room." My sense is that you're not going to find discrimination on something that other people can see the action and compare. You won't see it on price.

Fiona Scott Morton: No, because prices are posted.

Ayres: Exactly. So what should be tested?

Rick Brooks: I've run surveys where I asked people the question, "How often have you been racially profiled while driving?" And more relevant for what you're talking about, Ian, "How often have you been racially profiled or followed while you're in a store?" When looking at the correlation between their answers to these questions and their skin color I observed an interesting result. I had two measures of their skin color: One self-determined, where I ask them to report their skin color on a Likert scale from 1 to 5. And a second report from a research assistant coding their skin color on the same scale.

When I used the RA's assessment of the subject's skin color to predict how likely people are to report being racially profiled while shopping — controlling for gender and occupation and a few other things — the RA's assessment didn't predict much. But when I used the subject's self-assessment, it was one of the strongest predictors in the equation.

Scott Morton: So you're mapping a cultural thing, not a visual color. Like hairstyle, plus color, plus the way they're dressed, plus their cultural orientation?

Brooks: Right. There's something there, but I'm not sure what the causation is. I don't know, for example, if your perception of your stereotypicality affects your behavior and drives your interactions. But this is very intriguing.

Scott Morton: I feel sorry for the people working in the shops, in the following sense. Of course, I don't think they should be using race to discriminate against people. But on the other hand, they're being given incentives to, say, reduce shoplifting: they get a bonus, for example. And they're being given incentives to provide good-quality service. So when somebody wants to take four items instead of three into the dressing room, they have to make a decision about whether this is the kind of person who, if they'll let them take four in, there will be a positive return or a negative return. Were I in their circumstances, it's not the skin color you care about — it's the respect they treat you with, and the quality of the clothes they're wearing, and whether they speak in complete sentences that is what you want to use to decide if they will steal the items or not.

James Baron: But this is still a statistical discrimination problem, right? The clerk is still making a prediction about individual behavior based on group characteristics. He's just looking at complete sentences versus no complete sentences rather than white versus black.

Scott Morton: Isn't the point that we're supposed to train these guys to not use factors we're worried about, namely skin color, and to use things that they should legitimately use, like whether the person looks like they're going to steal something? Is the store supposed to not take that into account at all, and just say, "I'm sorry, there's a rule that you can only take three items into the dressing room, and even though you look terribly trustworthy and you have two toddlers, so you obviously don't want to be running back and forth between the shelves and the dressing room, we're going to drive you away to a different store, because we refuse to be helpful?"

Brooks: Part of it is about efficiency, and part of it is about distribution. The costs are going to be spread. Let's say we're going to force the store to adopt an inefficient monitoring technology, where they keep an eye on everyone. It's going to raise prices, and the cost's going to be shared, broadly, by consumers. Alternatively, we can let them work with a more efficient system that will lower costs, but those costs are going to be concentrated on the actual offenders and the innocent people who are also part of that same subcategory.

Scott Morton: In terms of what group you fall into, skin is something you can't choose. Speaking in complete sentences, sounding polite, and not having tattoos are choices you make. And you as a consumer can decide to make those choices. You know how you're going to be treated in the labor market or in the market for shoppers. And maybe that's something you have to deal with.

Baron: Doesn't this version of discrimination fall into the category of things that the market ought to eliminate? It would seem that between two stores, the one whose gatekeepers are wrongly precluding good customers from going into the dressing room with four items is going to be penalized in the market. It would seem like that behavior ought to be driven out by the market, unless there's something else going on that's sort of counteracting that. If those beliefs are not accurate...

Brooks: Take Denny's. That's a chain that, in some markets, has engaged in proven discriminatory practices. So I don't think that we can trust the market to drive out discrimination, in part because there may not be another option for the customer — the very population that's subject to discrimination doesn't have the advantage of competitive forces. It's a compound. There's a market failure and, potentially, discrimination that may not be corrected by competition.

Baron: Either from an economic point of view or a legal point of view, what is the most effective remedy? If there are two stores, and one of them has a gatekeeper who is statistically discriminating in ways that are actually costing the firm business, but we can't rely on the consumer to go elsewhere...

Ayres: I think the right remedy turns on the underlying cause of the discrimination. So in Fiona's work on car buying, if the race discrimination is caused, in part, by price discrimination, that is, trying to get as much money as possible out of any customer, competition is good — it will help drive out the racial disparities. If it is caused by rational statistical inference, where dealers have determined, accurately, that they can get more money out of certain groups because they lack information, more competition is not going to help. That will force firms to apply those determinations more assiduously so as to not get left behind by the better statistical discriminators. There, the government might want to intervene informationally, to give those groups the tools they need, and sidestep the statistical piece.

If the discrimination is preferential — that is, a bias for one race over another — then competition may be good or bad. If it's preferences of the median consumer that cause the problem, competition may reinforce the effect. If it's preferences of agents — the clerk in the store, for example — then competition might drive it out.

Brooks: When I think about the remedy, the question is the objective. What are we going after? Is it maximizing return? Are we focused more on the economics, or are we also interested in the distribution of rights and protections, even at some economic cost? I think we could probably generate a lot more wealth by taking a lot of rights out of play. But there are some things that we just won't do, or let some firms do.

Baron: It seems to me this discussion speaks to the efficacy of one of the primary remedies that firms have adopted to deal with these problems, which is diversifying the work force. And it's not obvious to me that if there are two Marshalls, next door to each other, and one has 50% of the people who are gatekeepers to the dressing rooms be African American, and the other has 5%, holding constant the ethnic mix of the clientele, that we would observe much difference. It seems to me that the premise that a lot of diversity initiatives have is that somehow if I go in and I see a face that looks like mine, then these processes that would otherwise work against me aren't going to.

Scott Morton: We disagree a little bit. If the African American gatekeeper at Marshalls can then categorize African American shoppers into 50 different buckets, according to their responsibility level, and I can't, because I'm less familiar with that demographic group, then that person knows the top 25 of her demographic buckets are the right ones to allow into the dressing room with four items, and the bottom 25 aren't. I wouldn't make as good decisions because my background limits my knowledge of that group.

Baron: Let me suggest that a firm that thinks it's going to address this problem by hiring a quota of African Americans probably has structured the job in a way so that, even if that discretion existed, it isn't going to be utilized.

Scott Morton: I used to think that affirmative action was a little silly in this way. But I've changed my mind somewhat by watching academics, in particular. I think the gain you get by hiring an academic from a previously underutilized group is their knowledge of topics that everybody else doesn't know anything about. So you get more studies of marriage markets among Muslims, for example; you get Roland Fryer studying "acting white," which, frankly, I knew nothing about until I read his paper.

Baron: That argument gets made all the time: we need a diverse set of perspectives. But is that really true at Marshalls? I'll buy the academics....

Brooks: It sounds like there are a few ways that diversity might contribute value here. One is that if you hire more diverse workers, your workforce may be more effective. Their specialized knowledge, let's say, will improve your bottom line. The second is that diversity creates some value for a targeted population. That is, you might have a more diverse customer base to which you can now better respond. And the third is that diversity will actually change perceptions in the broader society.

Ayres: How important do you think the internet's going to be in disabling racial disparity?

Scott Morton: I think in ordinary markets, like cars, it's hugely important. You go online, you get quotes, you become armed. So anybody who can do that can beat the statistical discrimination, or any other kind, that's out there.

Baron: Is that true, though? One could have said, a kind of test that Ian should have talked to the attorney general about is the capacity to come in with a kind of print-out from Vehix.com and leverage that into an actual purchase. So do you think it necessarily follows that that information at the point of purchase leads to the ability of different consumers to get the same treatment?

Scott Morton: Yes, I do. But the question is whether or not consumers who need the information the most will use it. The evidence that we have is no, they do not. I mean, this is what the digital divide is. If you don't have educa-tion, you are less likely to have or to use an internet connection. You don't know that you should go to the library. But a guy who knows just enough to know that he should go online actually ends up on a pretty level footing with everybody else because the online page spits out the same number to him as it does to you, and he prints it out and he takes it down to the dealership. And the dealership can make up a story about why they can't match that price, but that's a harder job than just quoting him the MSRP as they were able to do before.

Brooks: What do you all think of the role of information? The internet has two possible ways of leveraging information here. One is that it produces more information that everybody can see. This is what you and Jim were talking about. And there's another way, though, in that it conceals information. Like closing off the identity of the buyer.

You can potentially purchase things on the internet without ever revealing that you're a woman or a minority. Perhaps we should be talking about discrimination in markets — the extent to which we want to reduce it, or to take advantage of some of its beneficial aspects — as really an information management problem?

Scott Morton: The situation that Florian Zettelmeyer and I studied, in our research on cars, has the interesting feature that, when you go online for this price quote, you give them your name and address and you talk to a dealer on the phone. So the dealer pretty well knows what your race is, whether you can speak English, what your gender is. And yet we found that those people were getting the same prices as the white men when they came through the internet.

So it seemed to us that what the dealers were doing is they're ready to take money from anybody who looks like they're uninformed. But if you come through the internet, I think they figure out, correctly, that you know how to use the internet, and it takes you about five seconds to find another dealer who will offer a better price. And it's not worth their time messing with you, because it's too easy for you to walk.

Baron: Why don't we see more consumer unions with the mandate of ensuring the purchasing rights of people who might meet discrimination?

Scott Morton: I think Costco's done a good job in this regard. You can get insurance there. You can get eyeglasses. You can get caskets and funerals. Those are all markets in which there's a lot of price dispersion and a lot of stores that have high markups, and where you're not purchasing frequently enough to be that well informed.

Ayres: To me, the crucial question is the extent to which the internet is going to reduce price dispersion. And it's not a completely done deal. The first decade has been very good for us. But unlike McDonalds, where you can see if you and I got the same price for a hamburger, Amazon could try charging different prices for the same book.

But the big place where I think disparities are going to go away actually relates to some of your work, Rick, on lending. I think LendingTree can be extended to home mortgages more generally, to credit cards, to payday loans. The digital divide is going to go down, and there's no reason you couldn't jump on the internet and put a payday loan up to a thousand lenders. Even if this reinforces statistical discrimination, the competition is going to simultaneously be driving down the margins. It's got to be reducing disparity.

Baron: I think you're right that the internet has the capability of doing this. But I've been really struck by the variation across sectors in the speed with which that opportunity has actually been realized.

I bought a used car on the internet a couple of years ago. And it was a very specific kind of car. The manufacturer lists all of the secondary market cars that it knows of. So you can log on and find out this particular year, this color... I was still living in California, and I found a version of this car in Oakland, 40 minutes away from where I was. And I found exactly the same car — and there were only 800 of these particular cars ever made — I found the same car at a place in Philadelphia, with considerably less mileage on it, but everything else the same, for $10,000 less. When I called the guy in Oakland, he said, "You can't compare a car that you buy here against a car that you buy in Philadelphia."

I said, "I called a car company and they said it's $900 to ship it. I believe $10,000 minus $900 is $9,100. Is there something about the comparison that I'm missing?" You know, it is a bit of a hassle to buy a car in Philadelphia and ship it to California, but it's not $9,100 worth of a hassle.

The question is, is there some entity that could speed up the extent to which those kinds of transactions can be made by more people?

Ayres: So what happened?

Baron: I bought the car from the guy in Philadelphia.

Scott Morton: So what's your complaint?

Baron: My complaint is that to the guy in Oakland, it was inconceivable that somebody would do that. So I'm saying, is there some entity that would make it plausible to both sides of that transaction that that can, in fact, happen? That would seem to me it could democratize it more.

Brooks: I don't think what you want is an entity so much as access. That's what all of this is about. When the seller knows that you can just go to another website very quickly, and access another seller, then you can get competition. When you have access to the labor unions. When you have a credit score that gets you access. The problem, generally, I think, is a failure of access.

Scott Morton: Right, because then markets are working full steam ahead, if you have access.

Brooks: Right. So maybe the discussion isn't so much about market failures per se, as about certain people being closed off from otherwise well-functioning
markets. Costco is great; there is just the inability of some to drive to Costco. The inability to get to the internet. Or to get a credit score rating. An inability to get to the credit union or the consumer union. Discrimination remains, not because markets are problematic, but because there are some populations that can't get to them.

Scott Morton: If you can't get to the good market, you're stuck with the payday loan shop that's next door to your house, because you don't have the credit scores.

Other problems are more difficult than just having access to a market or to a price from Consumer Reports. A lot of what we're doing in the world these days, like buying houses and getting credit cards, requires financial literacy, which many consumers do not have.

In the New York Times this morning, it said that to prevent bad sub-prime mortgages, Illinois wanted to make the borrowers go to two hours of counseling. This is supposed to provide financial literacy because nobody understands the pieces of paper they give you, the disclosures where you're supposed to sign, "Yes, I understand that my interest rate is going to do this..."

Baron: In the auto example, I can log on to Edmonds and it will tell me that people in my area are paying $47 for an AM/FM stereo option on a Subaru. Do we have any evidence that using these kinds of tools changes outcomes in the market — that if more African Americans had access to those kinds of tools, we would see different outcomes?

Scott Morton: The papers I've written suggest that, yes, using the internet lowers the price that you pay.

Brooks: Why isn't there a brokerage market more like that for minorities? "I will be your agent and figure out all the sources..."

Scott Morton: There are. You can go to a car broker, and the car broker will tell you "I'll get you a car for $500 over invoice."

Brooks: But does the population that needs it most even know about it?

Scott Morton: That's the problem. That the people you want to help don't know that they need help.

Baron: That brings me back to the consumers union question. Why doesn't, say, the NAACP have a huge part of its identity be the creation of a part of its website that says, here's where you go to figure out what's the market price for a casket with green velvet
and six feet long in your market?

Brooks: Why would it have to be a nonprofit? Why not some savvy MBA from SOM who cares about minority markets or the minority community?

Scott Morton: You know, it might be the fear is that it is taking business away from the minority-owned businesses. So it may be that those guys are paying high prices at the local Chevrolet dealer and the local Chevrolet dealer is African American. Or the local funeral home owner is African American. And steering business to the ethnic Chinese funeral home across town is perceived as harming the community.

Ayres: I think there's also a market failure in unwillingness to embrace for-profit services to help out minorities. There's only one term for somebody who rents a house to a poor person, and that's "slumlord." If you do it on a nonprofit basis, you're fine, but anybody who, on a for-profit basis, rents a house to a poor person, that's what they call them.

And there's an analogy to that for anybody who provides a TV set to a poor person. That's actually a famous law case. It's where the court says, "How dare you sell a TV set on credit to somebody you knew was on welfare?" As if this is something that's a bad thing to do, to help poor people get chattel in their house.