Interview with Economics Professor John Yinger on Mortgage Discrimination
I recently had the opportunity to talk with Dr. John M. Yinger, a professor of economics at the Maxwell School at Syracuse University who researches discrimination within credit markets. His bio reads:
“John Yinger is a Trustee Professor of Economics and Public Administration and International Affairs, director of the Education Finance and Accountability Program, and associate director of the Center for Policy Research. He has published dozens of articles in professional journals on the topics of education finance, discrimination in housing and mortgage markets, and urban economics. His most recent book, “Housing and Commuting: The Theory of Urban Residential Structure,” is forthcoming. His edited volume, “Helping Children Left Behind: State Aid and the Pursuit of Educational Equity,” published in 2004, and his co-authored book, “The Color of Credit: Mortgage Discrimination, Research Methodology, and Fair Lending Enforcement,” appeared in 2002. Yinger has also taught at Harvard University, the University of Michigan, and the University of Wisconsin; served as a senior staff economist at the President’s Council of Economic Advisers and co-directed several state-level tax or aid studies (from Syracuse Directory).”
In this article, I want to share the transcript of my interview. We have a lot to learn from Professor Yinger and his work with discrimination:
Could you talk about how you personally got involved with the topic of discrimination within credit markets and how you explored your interests?
That is a little bit of a long story, but the starting point is that my father was a Sociologist who wrote a textbook on race relations in the 1950s. That textbook went through five editions and he wrote many other articles on race relations in his career. Then I grew up in the 60s, which was a time when we all thought a lot about [discrimination] issues, and my personal interest was stimulated in 1965, the year I graduated from high school. My father invited Dr. Martin Luther King Jr to give the commencement address at Oberlin College, where he was a professor. I was fortunate to ride in a car with MLK from his hotel to the place he was speaking at Oberlin College. It was a very short ride, but a very memorable one: he turned around and shook my hand, and asked me how I was doing. This made a great impression on me. Then, when I started graduate school, I stumbled across a topic that seemed to be understudied and fit my interests. That was some essays on prejudice, discrimination, and segregation, and that was my dissertation and I kept going from there.
I noted that you pursued a Ph.D. from Princeton University that was titled “Three Essays on the Economics of Discrimination in Housing,” defended in 1974. Could you talk about how our collective knowledge of mortgage lending discrimination evolved as new academic studies came out, Housing and Urban Development (HUD) regulations became amended, and new data (such as HMDA) were released?
My dissertation and my early research did not have much to do with discrimination. The scholarly literature was pretty limited at that time. In the early 1980s, a book came out by people who ended up becoming my colleagues at the Harvard City Planning Department, where I had a teaching position. They (Bob Schafer and Helen Ladd) wrote a book on mortgage discrimination that was a really fine book given the data limitations of the time. Then, I got involved with people at the Urban Institute to study housing discrimination, and after working with them, I was asked to help write a survey on mortgage discrimination, where I was joined by Steve Ross, who had been a student of mine at Syracuse. He and I wrote a couple of chapters in a book that the Urban Institute put out, and eventually, we turned that into our book “The Color of Credit.” It was kind of a progression of 10 years as I got into this topic but the culmination of it was the book with Steve.
I noted that you and Stephen Ross have various publications together, from “The Color of Credit” to literature reviews, among others. Could you talk about the importance of collaboration in these types of fields?
Economics, like many fields, has gotten very complicated. It takes a lot of knowledge and time to assemble a data set, figure out the conceptual issues, and figure out the appropriate statistical methodology. It is very natural that people want to do that in teams. Like any other relationship, you have to negotiate a relationship and it has to work. Some scholars work alone, others in teams, and some like me do both. I think it’s great to work in teams. You can be a little more ambitious and get the advantage of different points of view. A lot of the work in economics is done in teams, and I have had some wonderful teams on discrimination and other topics. Steve is a wonderful scholar at the University of Connecticut, and we had several different publications that we did together, including the mortgage book, and it was a delight to work on the book with him. I am very proud of that book, and probably I would say that the main lessons of the book still apply. It is not as if we have solved the problem of mortgage lending discrimination. It’s in a different world though, as you wrote your paper on methods for predicting default, a lot of complicated statistical issues come up. Our book did not address those directly but the principles in our analysis apply. Maybe we can come back and talk about that in relation to your work.
What is your personal favorite work?
That’s a tough one. I have a pretty long list. The thing I am most excited about now is some work on what is called hedonic regressions, which are regressions where the house value is the dependent variable and the characteristics of houses in neighborhoods are the explanatory variable. This is an interesting topic to me because it helps us understand what people care about, including the racial composition of their neighborhood. It also helps understand how people allocate themselves across neighborhoods — why some types of people live in one place and some types in another. Of course, racial segregation is a part of that but is not the only thing. I have a paper in the journal of urban economics which is my main paper on this but it leads to other papers as well. But I am very proud of my work with Steve.
As you noted in the paper you had sent me yesterday, Regulation B does not say anything about the steps a regulator needs to take to determine whether a lender is practicing disparate-impact discrimination and also that the business may have an incentive to discriminate against customers in that class, that is, to practice statistical discrimination. As a result, what are some proposed solutions to mitigate disparate impact discrimination?
This is a little bit of a complicated question. Statistical discrimination arises when a lender or some other economic actor uses the characteristics of a group to make a decision about an individual, not knowing what the characteristics of the individual are. This is a violation of the principles of civil rights that we have in this country. It is not always directly linked to legal issues, though that is a complicated story. Clearly, we (the broad consensus) want people to be treated in this country based on their own traits, not the traits of groups to which they belong. This is complicated in mortgage discrimination because there are statistical issues for sorting out when people are practicing statistical discrimination. One of the key issues is that the data that you use in your study (the kind of data you were talking about) has issues of race and ethnicity built into it. You have to take it out of the data if you want a fair procedure. Taking it out of the data means you have to account for it — you have to put it in your analysis and then take out its effect. A lot of people find that very troubling, so they want a neutral policy for not accounting for race or ethnicity at all. But, what Steve and I wrote in that book and what I discussed in that article I sent you, is that if you do that, you automatically build disparate impact discrimination in your analysis — you have to take race out in order to be fair. You can’t take it out without figuring out what its contribution is. My impression is that the federal regulatory agencies in mortgage lending — and there are a lot of them — have not really caught up to this. I don't know of any cases — there may have been some background settlements — where the statistical analysis has specifically tried to see if there has been disparate impact discrimination. There are lots of cases where the analysis looks for disparate treatment disclination, which is just treating someone differently using different rules for people in different groups. Disparate impact is more subtle because it uses the same rules for each group but the rules are designed to reflect information about what group you belong to. Unless you take that out, you will have disparate impact discrimination.
In my paper, I talked about algorithmic lending and the transition into the Fintech era. Machine learning algorithms have become increasingly proficient at predicting race even if race is not included. What are your thoughts on this and how is our knowledge impacted as statistical technology becomes increasingly advanced?
That is exactly the problem. If you are able to predict race based on other characteristics, then you can practice discrimination at will. Again, that is what Steve and I show in our book and is still true today. My opinion, which is expressed in that article I sent you, is that federal regulatory agencies have to invest in statistical knowledge and data sets so they can keep up. One thing that is important here is that there is an established procedure for determining whether someone discriminates that gives the lender the right to explain what they are doing. Let’s suppose — referring back to your paper — that there is a federal regulatory agency that uses a simple procedure to predict default. Then they do an analysis and say that the procedure implicitly builds in race and involves disparate impact discrimination. Then you can turn to the lender, and the lender can say “we use a much fancier method and we don’t predict what race they belong to and our technology does not have disparate impact discrimination in it.” They should be allowed to do that. So, you want to make sure that regulatory agencies keep up. Up to now, these agencies, at least to my knowledge, have not made major investments in the statistical methodology that would allow them to keep up with what’s going on. Maybe they have — I can’t say I am terribly well-informed — but we have had a major transition in consumer finance protection when we went from the Trump administration to the Biden administration. It is possible that there is more work on this than I know.
What research, if any, are you currently working on?
I told you that I was excited about hedonics stuff and I am excited about the ability of the hedonic method to shed light on the way people sort into different locations. For example, you might ask the extent to which the distribution of neighborhood amenities influences the way people with different incomes allocate themselves across urban areas. That is the question I am looking at right now.
What do you wish you knew when you first started out in the field and what advice would you have for someone like me or other young people who are interested in similar work?
The key thing if you want to be an economist and do this kind of work is that you have to enjoy working with the tools of economics and you have to be self-motivated because nobody will tell you what to study. You have to figure that out for yourself. Then you have to find what gets you excited and pursue it. Figure out what you need to do to shed light on it. The tools of economics are getting more and more technical over time. My training is a little old, so I am not up on all the fancy econometric techniques. But that is why you put together teams — you find some young person who is up on all the latest techniques and you team up with them. Economics is a nice way to look at a lot of different questions. If you like the way economists think, it opens up a lot of doors. But there are plenty of other things to do that are worthwhile.
There is plenty of work on mortgage discrimination going on now. One of the challenges is that the data are very hard to assemble. When loans get issued, they are often sold to another institution. A lender provides a mortgage and then the right to receive the mortgage payments is an asset, which is sometimes marketed. From a researcher's point of view, information about mortgage loans moves around depending on who owns them. There are not generally ways to link the data across institutions. Some people have found ways to meet this challenge because there are some lenders out there who don't send very many loans and they are big enough for statistical analysis. Maybe someday we will figure out how to make the links and so on. There is still room for an enormous amount of research on mortgage lending. One recent study used the details of Automated Underwriting Systems and found quite a lot of discrimination in that. It required a sophisticated understanding of how that system works. So you have to learn a lot about institutions and the statistical capabilities of lending organizations, in order to make progress here.