Mary Pattillo & Jordan Conwell on 'Race, College Quality, and Intergenerational Mobility'

Event

Wednesday, March 7, 2018

Mary Pattillo of Northwestern University and Jordan Conwell of the University of Wisconsin-Madison, present their research at UC Berkeley at this Feb. 9 event titled 'Race, College Quality, and Intergenerational Mobility.'

This talk was part of the Research to Impact series. Learn more about the series here: https://haasinstitute.berkeley.edu/researchtoimpactseries

Transcript

Janelle Scott: Thank you all for coming. We're very excited to welcome our guests today. I asked and was told to give very short introductions. I think the power of their ideas and scholarship are going to come through in their talk, so I won't tell you all their amazing and impressive accolades. But it's my pleasure to introduce Mary Pattillo, who's the Harold Washington Professor of Sociology and African-American Studies, is an Institute for Policy Research Faculty Affiliate, and an African-American Studies Director of Graduate Studies. So welcome, and thank you for being here. We also have Jordan Conwell, who is a postdoctoral fellow at the University of Wisconsin, Madison but is also starting as an assistant professor this year at the very same university. So we're very excited, and we want to offer you a very warm Berkeley welcome. So welcome.

Mary Pattillo: Thank you so much, Janelle, for the introduction and for the invitation. It's really a joy to be here. You all might not know because I'm sure it's not on your literal radar screen, but we're getting 10 inches of snow back where we're from, so I know. So we're very happy to be here, happy that Jordan made it because soon after his flight they started canceling all the flights. I might not get back, but oh well. So it's really, really great to be here.

We decided to take this opportunity, given that this is the Haas Institute, to talk about our process because we think it's relevant for the kinds of work that you guys do in particular. Also, this is very much a work in progress. We are looking for your feedback, for your thoughts about our questions and about our methods, so we look forward to that. The backstory of how this paper came to be ... So you all don't know us, so I'll tell you a little bit. I am an ethnographer, a qualitative interviewer. I study the black middle class and racial inequality in the realms of education, criminal justice, and housing and other kinds of urban issues. Jordan got his PhD at Northwestern and is an education scholar and studies racial inequality and class inequality in educational outcomes, and he uses quantitative methods.

So we got together because of this MacArthur Summer Research Grant at Northwestern. I'm talking about it because if there are folks in the room who could do such a thing here at Berkeley or perhaps you have such a thing, this is a summer grant award that incentivizes faculty and graduate students to work together by giving a pot of money. Usually, that pot of money just goes straight to the student. You write a budget for it, and it really just kind of is like the student's going to work this many hours and the money goes to the student. But it could go for other research costs. You could buy data with it. You could pay research subjects with it. You can pay other things. But oftentimes it goes to the student, and it's meant to be a shared research project that is not either the faculty member or the student's existing project, so it's not just meant to fund what you have going on together. It's about coming up with new ideas.

So I knew Jordan was working on education, and I had some questions that I couldn't answer with interviews, and so we got together that way. I want to emphasize that while the grant is meant to be a collaboration that is supposed to assist in graduate student training, I put in here that when you really participate in it, you realize it's also about faculty training. Especially as a qualitative researcher who has a lot to learn about quantitative methods, it's very much been enlightening to me in both positive and negative ways. On the one hand, I'm like, "Oh, look what these numbers can do," and on the other hand, I'm like, "Y'all call this real empirical approach to science?" So it's fascinating.

The motivation was, in my own research, I read a lot about intergenerational mobility and racial gaps in intergenerational mobility. The research is pretty clear that there is greater downward mobility and less upward mobility among African-Americans than among whites. I was wondering how might higher education moderate that effect, the fact that blacks are more downwardly mobile than whites and less upwardly mobile? How does higher education fit into that? Then the second motivation is we actually started this right after the University of Texas case was going through the Supreme Court, and some of you all might remember this quote: "There are those who contend that it does not benefit African-Americans to get them into the University of Texas where they do not do well, as opposed to having them go to a less-advanced school, a slower-track school where they do well." So this is the late Supreme Court Justice Antonin Scalia arguing that basically affirmative action harms its supposed beneficiaries by sending them somewhere where they may not do well. We thought perhaps this logic might and perhaps Supreme Court decisions down the line might put high quality colleges more out of reach for blacks and Hispanics.

So these two questions are the motivation for the research. I'll talk a little bit about the literature in both of these areas. First, this point about upward mobility and downward mobility. It's a clear established fact that there are these racial gaps, that blacks are more downwardly mobile from ... So we're talking intergenerationally, so if a middle class black family is more likely, their children are more likely to be poor than a similar middle class white family and their children to be poor and the opposite, that a poor black family and a poor white family, the second generation of the white family is more likely to be upwardly mobile than the second generation of the black family.

There is very little but just starting to be published research on what explains that gap. Some of the gap could be due to intergenerational transfers and other kinds of things, but a paper by Mazumder published in 2014 finds that the gap between black and white upward mobility is essentially closed among adult children who graduated from college. So in the second generation among blacks and whites who graduated from college, that upward mobility gap is almost closed, and the downward mobility gap is significantly reduced. It's not closed, but at least it's reduced among college graduates. This gives us some sense that the answer to the first question is that definitely higher education moderates that effect. We then ask the next question, which is will any kind of college do, meaning does it matter if you go to a high quality college? Might that even more close the gap? Or, on the downward mobility side, where the gap was significantly reduced, might bringing in college quality actually make that gap disappear? So that's the first contribution to the literature, is thinking about college quality to build on those findings.

A second contribution is that we include Hispanics. Mazumder's research and much of the research on intergenerational mobility only looks at blacks and whites because to do research in intergenerational mobility, you need long longitudinal studies because you need information on the parents' generation and information on the children's generation once the children are grown, right? You can't really study intergenerational mobility on income if the kids are still 16 years old because they don't have any income. The best data set for that is the Panel Study of Income Dynamics, and it did not have a large enough sample of Hispanics when it started. So I'll talk a little bit about the data we use, the NLSY, or actually I think Jordan's going to talk about those data, but it is a better-suited data set to look at Hispanics.

There is some research already using the NLSY and looking at Hispanics, and it finds that looking at downward income mobility among blacks, whites, and Hispanics with middle class parents ... He finds that blacks are more downwardly mobile than whites, as I've already mentioned, but he doesn't find any significant difference between Hispanics and whites in downward mobility. When you look at upward mobility, there is more upward mobility among Hispanics and especially educational upward mobility, but much of that stems from the much lower educational origins of Hispanics than whites, and so it's easier to be upwardly mobile, especially when you're talking about immigrant generations, when the first generation has much lower educational attainment. So there's not much research. There's this one study by [Axe 00:09:26] and another study by [Wenn 00:09:29] and all that look at the intergenerational mobility among immigrants, so we're adding to that research as well.

Then, finally, the research on college quality. We know from research by Sean Reardon and others that blacks and Hispanics are significantly less likely to attend highly selective universities when compared to whites, even when you net out family income. But much of the research outside of Reardon that looks at college quality is done by economists, and they actually skip the question of predicting college quality and looking at racial differences in college quality and instead move to the next question, which is does college quality affect outcomes, outcomes like getting a bachelor's degree or income down the line. The early research found that college quality mattered for how much you make.

We haven't given this talk much, but giving a talk to folks at Berkeley, this is comforting, right? You go to a high quality university, and so I'm sure part of the reason you come to a high quality university is you think it's going to pay off in your income. The early studies found that, although there is some more recent research by Dale and Krueger that challenges that relationship and suggests instead it's more selection effect, that those of you all who get into Berkeley are going to make good money no matter where you go. That's what a selection effect means. So it's not really Berkeley that's given you the high income after you've finished here, it's what you came with when you came into Berkeley. We are not looking specifically to predict the outcomes, although we are looking at intergenerational mobility and some income outcomes. We're going to focus more on intergenerational mobility. But we also want to predict or look at are there differences in college quality among blacks, whites, and Hispanics.

So a little bit about the logistics just, again, as the backstory part, and Jordan will go more into depth. We had to figure out when we first came together to start this research what data sets we wanted to use. I already mentioned the Panel Study of Income Dynamics, which is the gold star. To be just quite honest because, again, I'm kind of taking this both as a behind the scenes approach to a collaborative research endeavor, really, I had just heard that the PSID is a bear to navigate. From everybody I knew, the PSID is more difficult to navigate than NLSY, so I was like, "Let's start with the NLSY." NLSY97 could have been an option, but there's not a second generation to study. The National Survey of Adolescent Health is starting to get a second generation income, but for intergenerational mobility, it's much better to study income when people are in their 30s and 40s than studying income when people are in their 20s, so we went with NLSY97 on that.

Nonetheless, you should keep in mind the particular cohort that we're studying. I mean, you guys are a nice diversity of folks in the room. For the younger people in the room, these folks are really old to you, and to the older people in the room, these are us, and so commentary on affirmative action today is a little bit different than thinking about folks who were going to college in the '80s, for example.

What question should we ask? We started off thinking about college match because that's really what Judge Scalia was talking about, meaning you have an ACT score, for example, and do you go to a school where the average ACT score is above yours or below yours. Above yours would be that you're over-matched. Below yours would be you're under-matched. We started with that, but that started to get a little too complicated in how do you define match. So college quality is much easier, and Jordan will talk about how we measure college quality. We are including Hispanics. Initially we thought most of the research is on blacks and whites, let's follow that, but we're happy to contribute to the literature on Hispanics. As you'll see, we are focusing on income attainment and mobility rather than educational attainment and mobility. We try to explain the race gaps in intergenerational mobility and also explain some outcomes on income, which Jordan will elaborate on. So, with that, I'll turn it over to Jordan.

Jordan Conwell: Yeah, perfect, thank you. So, just put more formally, the one question we're going to talk about today out of this broader project is here. So among this cohort in the NLSY79 followed from young adulthood to middle age, and they're still being followed, do black/white and Hispanic/white differences in the quality of colleges that folks attend account for racial gaps in income around age 40? I'll talk a little more, as Mary started to note, about the importance of taking income at that particular time in the life course. And then intergenerational income mobility, so where your parents ranked in the income distribution at that time to where you end up in the income distribution among your contemporaries.

So we've talked a little bit about the NLSY79. This survey started with about 13,000 folks who were aged 14 to 22 in 1979. They were interviewed every year until 1994, and they've been interviewed in even-numbered years since then. One of the things about NLSY that makes it useful for this kind of research is that although they've been following people for a long time, their response rates are still fairly high for a survey of this size.

Our data and part of the logistical challenge of this project was figuring out college quality. In the restricted use version of the NLSY that you apply at the Bureau of Labor Statistics to get, in the 1984 to 2012 survey waves, respondents were asked to provide the name and location of the three most recent colleges they had attended. What we do is we link that to another restricted use data source, which is the Barron's Administrative Index of College Quality. I think folks are probably familiar with the Barron's rankings. These rankings are most competitive, highly competitive, very competitive, competitive, less competitive, and non-competitive. They're based on the previous year median SAT or ACT of the incoming class, that class' GPA and class rank, and the acceptance rate of the colleges. Barron's doesn't rank two-year colleges, and so we include those in a separate category, as we'll discuss. A large share of college enrollments, of course, are at community colleges, and we want to assess the affects of community college as well, especially given racial differences in who ends up at community colleges, a very important part of the story for racial differences in these outcomes, as we'll talk about.

So here's an example of these ratings. This is from California in 1999. We have Barron's rankings for many years, so we know what year the respondent to the NLSY indicated they went to that college, so we're able to use the quality rankings at the time the respondent was enrolled at that college, and some of these rankings do shift a little bit over time.

Mary Pattillo: I'll just add a joke. When we first put this slide together, Jordan didn't want to do California because Berkeley comes up in the highly competitive rather than the most competitive, and he was like, "They're not going want to see that."

Jordan Conwell: That's on Mary. That's not on me.

Mary Pattillo: Right, right. I said, "Well, this is California."

Jordan Conwell: Well, I just wanted to be a nice guest. I don't want to say that they're not-

Janelle Scott: That was 18 years ago.

Jordan Conwell: Yeah.

Mary Pattillo: Yes, exactly.

Jordan Conwell: Yeah, and that's why I made sure that these rankings do change over time, and also, as we will see, students who went to highly competitive colleges do have very nice outcomes. Everybody can breathe easily. It's not going to ruin your life if you did not go to Stanford or Pomona.

Okay. So, again, we have lots of reports for each respondent of different colleges they attended, and people who graduated in later survey waves, the most recent college you attended is whatever college you graduated from, so you report that over and over again. So we need one ranking of college quality for each person who was in this survey, and we're going to consider the college that you attended to be the highest quality college that you attended as opposed to the first college or the last college. Measuring college quality in this way is going to capture any upward transfer behavior, so if you were to transfer to Pomona, your highest quality college would be most competitive, not highly competitive. We're going to miss downward transfer behavior in this case, so if you transferred to San Jose State, we would still capture your college quality as highly competitive because you reported enrollment at Berkeley.

We're also going to use data on those who we do not have a valid college enrollment for at any time. So this is going to allow us to assess outcomes for the whole cohort and gauge effects of not going to college as well as racial differences in those effects. Just like the community college question, this is important to keep this in mind and include this in our analyses because there are obviously racial differences in who doesn't go to college and we want to look at the penalties for not going to college potentially and any potential racial differences in that penalty. So we have 5772 verified college-goers, 6914 non-college-goers, and that adds up to the full sample of the NLSY. So we're going to talk about this whole cohort.

This is the distribution of college attendance and college quality among those who attended. More than half of the individuals in this cohort did not attend college or at least we do not capture a reported college enrollment for them in the data. Among the 45% who did attend college, these enrollments are concentrated at community colleges and competitive colleges, so in the middle of the quality distribution. I'm going to return to some of this later, these effects of college quality on outcomes.

While we're introducing this measure of college quality, we want to show that college quality does have the expected effects on our outcomes of interest. So these are the results for an individual's income on the average from 2006 to 2014, so when respondents are between the ages of 41 and 57, by college attendance and quality. For example, compared to those who went to competitive colleges who are the category not shown, the excluded or comparison category, those who went to most competitive colleges reported 60% higher average incomes, and those who did not go to any kind of college reported 50% lower average incomes. Of course, as Mary noted, one of the reasons these effects are so large is that people who go to colleges of different quality are systematically different in many other ways that also affect their income, so this is not just the effect of college quality. This is the effect of college quality and everything else that's different about people who went to different kinds of colleges. So, later, we're going to talk about controlling or adjusting for some of these things. But this just goes to show that we are getting some important variation in what's happening with these outcomes with this measure of college quality.

Here is the distribution of [inaudible 00:20:53] and quality by race. Later in the talk, we're going to show these as percentages and discuss racial differences in the probabilities of ending up at certain quality levels. Here, I'll just talk about the numbers of observations because they're relevant to our methods. In particular, the numbers of individuals in the higher quality categories for blacks and Hispanics are relatively low, and so we're sometimes going to talk about these college quality categories in larger groups just to get more precise estimates of the effect. So, at times, we're going to talk about non-competitive and less competitive colleges together and very competitive, highly competitive, and most competitive colleges together.

So if starting with racial differences in college attendance and quality, these are the odds of attending college versus not attending college from a ... Sorry ... from a logistic regression. Here, we see that we're not controlling for any other aspects of individuals' background. Blacks were about three-quarters as likely as whites to attend college, and Hispanics were about 60% as likely to attend college.

But, again, we know there are lots of other factors that differ by race that affect people's probability of going to college, so what we want to do is control for those. Those controls are shown on the bottom, so gender, mother's age at student's birth, whether a student grew up in a two-parent household, whether they resided in an urban area, mother's education, number of siblings, parental income, and then we have in the NLSY students' armed forces qualifying test score, which is used in a lot of different literature in sociology, economics as a measure of cognitive ability.Once we control for all these things, blacks and Hispanics are both significantly more likely than whites to go to college. So what we're comparing here is blacks and Hispanics who are theoretically equal on all these things, so grew up in the same types of neighborhoods, had a mother who had the same education, had parents who had the same household income. In that theoretical comparison from this model, blacks were three times more likely than whites to go to college and Hispanics were two times as likely, and those differences are statistically significant.

This is an example of what in the literature is called a net non-white advantage, so this is not totally uncommon in research on educational outcomes. Research on degree receipt, for example, has found that when you're controlling for all these things, blacks are more likely to earn degrees than whites, and we're finding this pattern here for college attendance and expanding it to see this difference for Hispanics as well. So not entirely surprising but worthwhile to note.

Now, we're just talking about the folks who went to college, a community college or a four-year college, and we're going to look at what are the racial differences in the odds of ending up in higher quality categories. So, compared to whites, blacks ... The way you read this, this is called an ordered logistic regression. Blacks went to significantly lower quality colleges than whites who also went to college, and Hispanics also went to significantly lower quality colleges than whites who went to college. But, again, we observe net black and net Hispanic advantages in college quality among college-goers once we're controlling for these background factors.

So moving to individual income, I showed you a slide on this before. This is individual income from 2006 to 2014, so these respondents were aged ... The youngest respondents were age 41. The oldest respondents were age 57 by the end. Income measured around age 40 minimizes life cycle bias in income, so income fluctuates over the life course. It goes up, it stabilizes sometimes at a peak, and then it starts to go down again towards retirement. This is one of the reasons that we don't want to use data from NLYS97, because those respondents are just finishing college, some of them are still enrolled in graduate school, so you don't get good estimates of lifetime income for that cohort. Actually, you get systematically bad estimates because people who are going to earn the most are still in school and the ones who are out are going to end up probably earning lower incomes.

So the nice thing about this is we have people at what is considered their peak lifetime earnings point and we're going to take the average of any reports from 2006, 2008, 2010, 2012, and 2014, which is going to minimize noise in the individual income reports. We also include years where the respondent reported they did not ... They had zero income in the average, and zero is different from missing. So if it's missing, it's not counted in the average. But if it's non-missing and you reported zero, that is in the average.

So without considering college quality, without considering any controls, we observed significant black/white income gaps and significant Hispanic/white income gaps. On average, blacks' incomes were less than 60% of whites' incomes, so if on average whites earn, say, $50,000 a year, we would observe blacks earning about $30,000 a year on average. We also observed a significant Hispanic/white income gap smaller than the black/white income gap. Hispanics' incomes were on average about 80% of whites' incomes, so, again, if whites are averaging about 50K, we observe Hispanics earning on average 40K.

Our question is does college quality explain these racial income gaps net of other factors that we also know differ by race and influence income? Before getting to college quality, we want to control for the variables for a model of this cohort's income attainment from previous research on the cohorts. One of the other nice things about using data that's been used by a lot of different social scientists is we already have some accepted models in the literature of the things that affect these cohorts' socioeconomic outcomes. We're going to see if after we control for that kind of model, we put college quality on top of it, do we get additional explanatory power?

Here, first, before getting into college quality is that model. So we're controlling for gender, the AFQT test score, how much schooling the respondent attained, the average weeks they worked per year, their mother's education, their age, their share of time in rural area during 2006-2014, their share of time in the South, their share of time they reported their work was limited by their health, the shame of time they were married, whether they resided in the South when they were younger, the average number of children they had during this time period, and the average number of preschool children they had during this time period. So, again, that's a model from a previous study of this cohort's income.

Controlling for all those things, so comparing people who are theoretically equal on that background and those labor market and family experiences, blacks still earned 7% lower incomes than whites. So controlling for those factors brings blacks and whites incomes' closer together, but we still observe a significant gap. So, again, if we observe whites earning about $50,000 on average, we would observe blacks earning about 46.5. These same controls do account for the entire Hispanic/white income gap, so the Hispanic/white income gap is no longer significant net of those controls.

What we want to do now is determine whether racial differences in college quality further account for these gaps net of this model, so we're going to add the controls for college quality. What we want to see for the black/white gap is this reduced perhaps to insignificance, and then for this non-significant net Hispanic advantage, does this get bigger and become significant?

So net of college quality and other controls, blacks still earned approximately 7% lower incomes than whites, so net of this model of income attainment for this cohort, the racial differences in college quality, which we showed you, don't give us additional explanatory power for the black/white income gap. They don't explain any more of that gap. This finding is actually consistent with some recent audit studies that finds that blacks don't receive the same labor market returns to elite college credentials as whites, so this is kind of the same thing. Net of all these other things, there's no additional return to college quality. We should note that we're also controlling for educational attainment in the model, so, again, we're looking at people who had the same attainment who went college of different quality, for example. Sorry, this is skipping ahead. And, also, in that slide, you see the Hispanics' net advantage does not get any larger. So Hispanic/white differences in college quality also are not doing more to explain Hispanic/white differences in income.

Even though racial disparities in college quality don't explain more of racial differences in incomes, net of all those controls, we still do observe some strong effects of college quality. So, compared to those who went to competitive colleges, which is again the excluded or omitted category here, those who went to most competitive colleges earned significantly higher incomes, while those that did not go to college at all earned significantly lower incomes. So still seeing some strong expected effects of college quality in the tails of the quality distribution, so really high quality colleges versus those who didn't go, net of those controls.

Turning now to what we feel is one of our main contributions to the literature is do racial differences in college quality account for some of these persistent racial disparities in intergenerational mobility from the parent generation to the child generation? So we looked at individual income before ... Okay, there we go ... and we're now going to turn to family income in order to be consistent with previous research in mobility literature, the paper by Mazumder that Mary mentioned as well as work by Deirdre Bloome and Bruce Western.

So we're looking at parents' family income from ... Or you can think of this as the household income, from 1979 to 1982, the average in any years where the NLSY respondent was living in the parental home. So some of the respondents were out of the home by the time the NLSY started, some of the older ones, so they would not be included in this, and then some people were in the home in certain years and out of the home in other years, so we're taking any years where you were living with your parents and they reported an income, the average of that. Then we're taking respondents' own family income from the time period we just talked about, so 2006 to 2014. So moving from individual income to family income does subsume a couple important issues regarding assortative mating and family formation and racial differences in those, and there are some interesting race/education differences in those kind of processes, and I'm happy to talk about those in the Q&A.

But, for now, we're going to be consistent with this literature and talk about the family income of the respondent. Today, we're going to give just an illustration of what we think we're finding here. So following Mazumder, we're going to focus on a few illustrative key intergenerational mobility transitions. For upward mobility, we're going to focus on individuals who were born to the bottom income quintile, so individuals whose parents were in the bottom fifth, the bottom 20%, of the income distribution, what were their chances of exceeding that, so moving to the 21st percentile or higher, as an adult? Then, for downward mobility, we're going to look at those who were born to the top half of the income distribution, what were your chances of falling out of the top half as adult? Again, these are some of the key transitions that Mazumder and other have focused on, so we're going to use those today for illustrative purposes, and then we'll talk a little bit about future directions we want to go given some of the intricacies of our question and how they engage with this.

I'm going to begin with upward mobility. I'm going to fill this plane with a bunch of different stuff. But before doing that, in the legend of the graph, on the bottom of the slide, we see racial differences in the chances of being born in the bottom 20% of income in the first place. So only 11% of whites in this cohort grew up in those circumstances compared to more than half of blacks and 40% of Hispanics. So before we start filling this with chances of getting out of that position, it's important to note that there are huge racial disparities in being born into those circumstances in the first place.

Beyond those compositional differences, there are also racial disparities in the chances of moving up from that position, so not surprisingly, we see some double disadvantage here for blacks and Hispanics, more likely to be there, and as I'll show you in a second, less likely to get out. So, among whites, 68% of those born into the bottom quintile exceeded that quintile as adults. Again, exceeding here just means getting to the 21st percentile or higher, but even given that small difference that we would accept as upward mobility in this example, we see big racial differences even in that. So, for blacks, only 47% of the larger group of blacks who were born into the bottom quintile make it out as adults, and this results in what is called a black/white upward mobility gap, in this case of 21%. And, then, for Hispanics, as Mary talked about, we see slightly smaller gaps compared to the black/white gaps but still a difference. 61% of Hispanics who were born into the bottom income quintile exceeded that position as adults for a Hispanic/white mobility gap of 7%.

So what we would like to know is how did these patterns differ by college attendance and quality? To do this, we predict the chances of upward mobility by race in college attendance using separate models for each race to allow the effects of college quality to differ by race. So beginning here with the black/white mobility gap, we see that this gap is largest among those who did not attend college. In that group, 63% of whites moved up compared to only 36% of blacks, for a gap of 27%, which is larger than that overall gap of 21%. The gap was much smaller among college attendees. For example, it was only 10% among those who attended high quality four-year colleges, so that together group of very competitive, highly competitive, or most competitive. In that group, 83% of blacks were upwardly mobile, greatly exceeding the racial average, compared to 93% of whites.

It's also illustrative here to compare across races, so comparing some of the points on the red line for blacks, the blue lines for whites. For example, blacks who attended very highly or most competitive colleges had upward mobility probabilities that were roughly equivalent to those of whites who went to competitive four-year colleges. But blacks who went to non and less competitive four-year colleges had upward mobility chances that were not much better than those of whites who didn't attend college at all. Also, for blacks, we see that getting to community colleges as opposed to not going at all appears to have provided a big boost for moving out of the bottom quintile. So we see a big slope there between not going to college at all and just getting to community college, and we see another big move from the competitive college to the very highly or most competitive margin.

Turning now to Hispanic/white gaps in upward mobility, these are largest among those who didn't go to college and at the top of the quality distribution, among those who attended very highly and most competitive four-year colleges. It's important to note that Hispanics who went to the highest quality four-year colleges had much better mobility outcomes than Hispanics overall, even though they maybe didn't do as well as whites who went to these kinds of colleges. 81% of Hispanics who went to high quality four-year schools were upwardly mobile from the bottom quintile. Yeah, so we want to note that. Among those who attended community colleges, non, and less competitive four-year schools or competitive four-year schools, Hispanics had slightly better upward mobility outcomes than whites did.

Here's the full graph for downward mobility, so I won't fill in the whole plan again. Again, this is those who were born into the top half of the income distribution, chances of falling out of that position as an adult. On the legend of the graph, again, we see big compositional differences in who was born into that circumstance in the first place, 61% of whites born to the top half of the income distribution compared to a fifth of blacks and less than a third of Hispanics. Again, the colored lines indicate the, again, racial differences in chances of moving down, so we've flipped what we were seeing before. Now, blacks not only were the least likely to be born into the top half. Now, they are the most likely to fall out. Hispanics in the middle and whites being the most likely to be born into the top half and the least likely to fall out.

A few other things we want to note here. First, black/white and Hispanic/white gaps in downward mobility are essentially closed among those who went to the highest quality four-year colleges, so that category of very highly and most competitive. A few other things we'd like to note, for blacks, is a big effect, at least for these individuals born into the top half of the income distribution at this margin between community college and the lowest quality four-year schools. So, for chances of moving downward for blacks, it looks like attending a four-year school, even a lower quality one, provides some defense against downward mobility relative to attending a community college, which did have positive effects on moving up from the bottom quintile, perhaps not as big of effects as moving down from the bottom half, but getting to one of those non or less competitive four-year colleges does look to have a big effect. For whites, we see what looks to be a bit of a ceiling effect in the extent to which college quality protects from downward mobility once a student attends a mid-quality or competitive four-year college. But for blacks and Hispanics, the chances of downward mobility continue to decrease ... That's a positive outcome. Your chances of moving down continue to get smaller between mid and high quality four-year schools.

So, just to recap our main findings here and then we're looking forward to your thoughts in the Q&A, in this cohort, blacks and Hispanics were significantly less likely to attend college, and among those who attended, they attended colleges of significantly lower quality. However, we observe net black and net Hispanic advantages once we control for family background and students' academic ability. For individual income at mid-life, the Hispanic/white income gap was accounted for by a standard model of this cohort's income attainment. The black/white income gap was not fully accounted for by such a model, and it is not further accounted for by racial disparities in college quality. Looking at intergenerational income mobility, which is where we feel some of our biggest contributions are in this case, we found that in this illustration racial gaps in some of these key intergenerational mobility transitions were smaller or eliminated at higher levels of college attendance and quality. This builds on previous work finding that college graduation is important for closing mobility gaps. What we're seeing here is that college quality, not even considering graduation in this case, also plays a role.

We want to emphasize that we showed here for these intergenerational mobility transitions was illustrative of some preliminary work and we want to expand on it. In those future analyses of income mobility, we want to, instead of using these people who began at a certain point in the distribution, in the bottom quintile, the top half, use everyone, so a larger sample to get more precision of the estimates, and look at chances of moving up or down by a certain amount, so 10%, 15%, 20%. We also want to know what else, like other aspects of family background or academic ability, is involved in this relationship between college quality and mobility. Just like the income figures we showed you at the very beginning, these big effects for college quality and mobility are also going to be correlated with the other things that push people into colleges of different quality.

And, then, for family income, something that I'm interested in in some of my other work on racial differences in child development is this issue of assortative mating. There's a lot of interesting relationships between gender and college quality and race and who ends up partnering with whom. We know that for blacks rates of educational assortative mating ... So this is where somebody who went to a high quality college or has a high educational attainment marries somebody who has also high attainment ... are much lower than rates of assortative mating for whites because the gender splits in educational attainment are much bigger for blacks. So this is stuff we know in black culture, highly educated black women lamenting the fact that they don't have a good marriage market, assuming a preference for same race partners. That's not just in the movies, those things. I can talk about other projects that I have going on right now. Those things are real, including amongst advantaged families that partner because these women do partner but they partner with lower status men, and that has implications for their household incomes. Also, in some other work that I have on child development and how kids do in school, it has implications for how their kids do in school relative to white kids whose moms have [inaudible 00:43:01] education because those kids' moms married dads that were better educated.

So, again, looking at if we're going to talk about family income here, that's all tied up in this as well, and those are really, really interesting and exciting extensions of these kind of dynamics that we're finding. Then we didn't show you some results for educational attainment today, which we also have, because the income was more than enough. We gave you 45 minutes of stuff. But we also have some stuff on do racial differences in college quality account for racial differences in years of schooling completed and degree attainment. So I'll stop there, and we're looking forward to your questions. Thanks.

Janelle Scott: Thank you. So questions? I'm happy to let you field the questions, or I can field them. However you ...

Mary Pattillo: I think we can field them.

Jordan Conwell: Yeah.

Mary Pattillo: Yeah. Yes?

Speaker 4: That was a fantastic talk. I just came to Berkeley after 10 years at Yale. Much of that time at Yale, I was wondering why Yale was failing black and Native American students especially and Hispanic students to a lesser account, and this was very illuminating. But I wonder coming to Cal, Cal is a much larger school, and I wonder to what extent ... How do we expect that the Yale students felt ... Well, other things that they felt was they felt extremely isolated, and there wasn't a good cohort effect going on.

So I think part of the advantage for highly selective universities is the networking that you do, but if you're one of only two or three or a handful of people, it's very hard to get that networking advantage. At Cal, you can be any tiny minority, and there's going to be at least 20 other people who are in the same boat as you. So I wonder to what extent ... Most of the most competitive universities are the flagship state universities, Wisconsin, Madison, and so what effect does school size play in the most competitive fields, and can you sort of just aggregate the size of the university and use that as a proxy for the size of their cohorts with this data?

Mary Pattillo: I have one thought before, perhaps, Jordan has some other ones. In some respects, I might challenge how the social networking works. You had mentioned small cohorts of same-race peers as one kind of social networking, but some might argue that the social networking that leads to the higher outcomes is cross-race social networking. It doesn't mean that being isolated is a good thing, and, of course, people would argue that one needs same-race peers to be able to do the cross-race social networking. So I think all of those, I think, would go in the direction that you're suggesting as well. But I might think that a smaller school that might then force more cross-race social networking could have the outcome that you posited for the larger school. So I think it's an empirical question, but I think the hypothesis could go both ways.

Jordan Conwell: Yeah, no, I think you're absolutely right that there are just trade-offs here, and so our work is not qualitative work on student experiences, but we know from contemporary movements ... [inaudible 00:46:33] Harvard, this kind of stuff. I'm sure there was probably something similar here and older work on these cohorts that just because you went to a high quality college and maybe you had very good outcomes, maybe at the end of the day you might decide that it was worth it for you, there is damage in other ways. So it's a trade-off that students are making and that families are making. That's a big part of this ... This under-matched discussion that we started with was this whole idea of what's your frog pond? Is it better to be a big fish in a small pond and blah, blah, blah. I think, in terms of socioeconomic outcomes, like your income, how much schooling you complete, on average, you go to the absolute best school that will admit you. But there are a lot of trade-offs there.

I think something else that you mentioned in this discussion is all the work we know about HBCUs and how even though these may not be schools that are of the same quality as Yale, huge percentages of black students who end up completing advanced degrees, getting MDs, JDs, come to these kinds of schools. All these effects matter in a lot of different really complicated ways.

Yes?

Speaker 5: Well, I noticed that the two most competitive schools were private schools, and so that cut between private and public, I wondered did you think about that? Because, honestly, knowing Pomona, my niece and my sister both work there, the atmosphere of a school like that is totally different than an atmosphere in a big public school. So did you think about those kinds of differences and what that meant for quality and any future attainment?

Mary Pattillo: If we controlled for public versus private in the model ... So the question would be if we controlled for public versus private, would all the ... See, I don't think all the college quality outcomes would be changed because the publics run the gamut of college quality and the privates run the gamut of college quality. So I think, here, we see what you're saying, which is those two happen to be the most competitive, but I don't think that's the case in all states. The publics can still be in the highest quality category. My guess is I don't think it would affect our findings that much because of the stratification of both publics and privates because many of the ones in the lowest competitive category were also privates. Matter of fact, I think they were all privates in California in the lowest category.

Jordan Conwell: Yeah, and while Mary's flipping, I'll just note that one of the difficulties just from a logistical standpoint of-

Mary Pattillo: Yeah, they're both privates.

Jordan Conwell: ... getting results like this is that we're looking at enrollments over a long period of time. So if we were just looking at one cross-section, we could very easily include a lot of these very interesting things about schools. Because we observe enrollments across 30 years and these things also change, that's kind of a bear logistically to do.

Mary Pattillo: Right. Although public and private is something that would be stable across-

Jordan Conwell: Public and private's easier, stable.

Mary Pattillo: Right, the first question would be less easy. The second question-

Jordan Conwell: Still hard. Still hard, yeah.

Mary Pattillo: ... would be easier because the schools maintain their public and private quality. But, yes, so, in California, the least competitive or the non-competitive, not even less competitive, the non-competitive are both ... I assume these are both private schools.

Speaker 5: I don't even know those schools.

Mary Pattillo: Just from names of them, I'm assuming that they are private. Yeah. Uh-huh (affirmative)?

Speaker 6: Just echoing the complimentary remarks, this was really insightful and illuminating. My question is the study period that you're looking at includes a number of regime changes, such as, for example, Prop 209, in other states, the Grutter decision, in Texas, the Hopwood decision. I'm wondering to what extent you're able to control that, especially because the kind of upward mobility channeling function of affirmative action can often get at students who come from much more disadvantaged backgrounds, [inaudible 00:50:54] more superficially in numbers. So I'm wondering to what extent you're able to account for that.

As a follow-up question, I saw another study a couple years ago that shows that ... It doesn't look like you're looking at post-graduate. When you get to the post-graduate, the gap shrinks even further. I'm wondering to what extent you looked into that or are planning to look into that.

Jordan Conwell: That's very interesting. In this, we don't. I think a way to look at that would be to at least find a way to control or adjust or look at college enrollments at certain periods of time because we do know when you reported going to this school. So, to the extent that those regime changes can really be captured based on different time periods or different states or interactions thereof, there are some ways to get at that. One of my colleagues at Wisconsin is Eric Grodsky, who was formerly at UC Davis, and he's done a lot of very important work in this area about trying to actually model these different regime changes. I think it's actually some California data. The different regime changes and how this affects affirmative action and different outcomes.

Speaker 6: So the students you're looking at were born up to 1981 or '82, right? So they'd be entering-

Mary Pattillo: No, they were born much earlier, up to 1965.

Jordan Conwell: No, they were born between 1957 and 1965.

Mary Pattillo: Right, yeah. So I doubt we could look at the regime changes you're talking about, much of which happened after these folks, most of them, had gone to college. But even if it were the same time, these are national data with only a certain number of observations in each state, and so our sample sizes already get small, and that's why we had to combine the categories. So, then, if we start looking at within state, we're likely into very small ... Yeah.

Speaker 6: [crosstalk 00:52:43].

Mary Pattillo: And on the question about Texas and California, Eric Grodsky's work in California and Marta Tienda's work in Texas has very much looked at the post-affirmative action and Ten-Percent Plan kinds of outcomes.

Jordan Conwell: If anyone has access to administrative data for this state and an RA who knows how to work with the data, I have a lot of questions I would like to ... Because this is why state data is great, because you're in ... This is the Grodsky or the Tienda work. Your sample size is huge in states of Texas and California, you have variation in school quality, you have variation in student demographics, but you also know very precisely the different admissions, regimes, and different things that were going on in terms of what the state university system was doing.

Mary Pattillo: But the thing we couldn't answer with that data ... There's always trade-offs, right?

Jordan Conwell: Right, trade-offs.

Mary Pattillo: ... is this intergenerational question is something you could only get with the long-running longitudinal data sets, which are unlikely to be state specific, so the variation you would get in a state data set you couldn't get here. But I guess if you really had a state that wanted to do it ... Oh, no, it would be very hard to get information on the parents. You could get information on the parents through financial aid records of the students, but you wouldn't know what the parents' socioeconomic status was when the students were growing up, for example.

Jordan Conwell: And sometimes you don't observe students long enough in the life course like they do. This is also a national representative. I probably should have noted that this is weighted to be a national representative of this particular age cohort in the US, which is obviously something that you don't have with administrative data.

Mary Pattillo: Mm-hmm (affirmative)?

Speaker 7: Very interesting talk. I'm really intrigued by the intergenerational aspect of your analysis. My question is really looking at the impact of the area of study and specifically funneling into degree programs with a higher return on investment and specifically looking at the most competitive versus highly competitive schools where there might be a secondary admissions process, so you get into Berkeley but you don't get into business school, and how that impacts your income long-term.

Jordan Conwell: I would say beyond the scope of what we can answer here but certainly something that is going on. We know that not only are there racial differences in admissions at these quality of schools, there are big racial differences in the extent to which students are formally accepted into some of these kinds of programs but also informally, especially in a big college. Like you're in letters and science, this is a huge college, right? The difference between getting pushed to African-American studies or Latino/Latinx Studies versus getting into, I assume, stuff like chemistry and other pre-med tracks over here, so even when I think if you're accepted in disciplines. So that stuff, again, it's pretty hard to measure. These are all things that we know are going on from our personal experience as students, as educators.

Mary Pattillo: And I ... Go ahead.

Speaker 7: Oh, I was going to say and it certainly complicates the public/private conversation as well in thinking that if a school is slightly smaller, the secondary admissions pool is also smaller and your likelihood of entering into a degree program with a higher return on investment.

Mary Pattillo: So what you're making me think is wanting to go back to the literature that exists to see are there any data sets that allow you to put all this in the model at the same time? There's a full literature on college major and outcomes that allows you to say exactly what you say, that there are returns to certain college majors, higher returns in terms of income to certain college majors. Those are cross-sectional data sets ... Or they're not, they might have some longitudinal component but not intergenerational the way we have, and large ends that allow you to control for public/private, college quality, college major, all of these things all at once.

So I just am curious to know if any of them can and have put them all in the model to see, really, what's more important because I think there could be a case to be made that an engineering major at San Jose State or Cal State Fullerton could make more than ... And just so I don't run into the trap ... You remember when Barack Obama kind of talked about our history majors and he got ... So I'll just say sociology ... Could make more than a sociology major at Stanford. I think those are the kinds of comparisons that ...

Jordan Conwell: Yeah, this reminds me a lot of [inaudible 00:57:13]'s work on mostly gender disparities but some stuff on racial disparities as well in STEM enrollment, right? So that's work that doesn't look at college quality but I think looks at just differences in selection into majors and returns to STEM majors.

Speaker 7: To my understanding, the intergenerational piece really isn't there in the existing work.

Mary Pattillo: Right, exactly, definitely, yeah. And it's really interesting to think about the intergenerational piece. Again, this might be more qualitative research, and I know there are some of this on that, which is if the young people having grown up in the, say, bottom 20th percentile are more encouraged to go into the majors that are expected to yield higher income returns whereas the ones in the top half or maybe the top 20% are more able to go into a larger range of majors. I know there is qualitative research on that. Especially among immigrant generations is where most of that research is done.

Yes?

Speaker 8: I'm almost hesitant to ask this, but I wonder if the NLSY has any questions on what colleges the respondents got into but didn't go to so that ... It's not part of a question, just kind of an idea I had, but it could look at people who for personal family or financial circumstances didn't get to go to the best possible school they could've gone into and then track their outcomes over time.

Jordan Conwell: Yeah, I don't think NLSY has this. A couple other education ... ELS, I think, Education Longitudinal Study 2002 and some of the other National Center for Education statistics data sets I think have it. This is reminding me also of Caroline Hoxby's work at Stanford because she has this great stuff from is the SAT ... So she has where all the kids took the SAT in the whole country, what schools they set their stuff to, so you get to choose on the SAT, "I want to send it to these five schools," versus where they went and differences in that. But I don't think we had-

Mary Pattillo: I don't remember that being in there either, yeah.

Speaker 9: Well, I've heard that African-Americans are less likely, even when they have scores, even when they have the A to G requirements in California, they're less likely to aspire to higher quality colleges.

Jordan Conwell: I think everyone ... In terms of higher ed, we read a lot of the Hoxby stuff, and we were starting to do this. She's an economist at Stanford on exactly that point, and it is amazing looking at the college application behaviors of ... Again, she has everybody, so she has huge ends, so she can really stratify the sample into "I want to look at kids whose ACT or SAT scores are at the 99th percentile of the country, and I have thousands of those kids." Looking at the college application behaviors of kids who have those scores based on their race and their family background, and the differences are staggering. In a national sample, there's a statistically distinguishable group of students who have scores this high, who are disadvantaged, who only apply to a little community college and Harvard.

Mary Pattillo: But her findings are much clearer on the income than on the race.

Jordan Conwell: Yeah, I think that's true.

Mary Pattillo: Her findings are very clear on that lower income kids with high scores are not applying to the most selective schools and less clear if the race effect is totally driven by the income effect or if, actually, and we found this in some earlier things that we were doing with this paper, if, in fact, black students are more likely to over-match, actually. There is a fair amount of the research that finds that, that not now in where they apply or where they get in but actually where they actually go, when you control for other things, black students are more likely to go to schools where the selectivity, the SAT scores and so on, are higher than what their credentials might predict, which is in alignment with an affirmative action mechanism, and we definitely found that earlier.

I'm just curious in the room how many undergrads are there in the room? Any undergrads? Any grad students? And then faculty members? And then staff? Okay, so we have a great mix. So just about how the sausage is made, which many of you all know from your own experience, but I'm so actually pleased with this paper today that I want to share this. We've been working on this for a little while because, again, this grant is meant to give money for things that is not the focus for either of you. So I have this other big project on monetary sanctions. Jordan was finishing up his dissertation, getting a job, a J-O-B, which he successfully did, all while we were working on this paper. So things move slowly. Oh, and then there's the restricted access data, which at Northwestern means now Jordan has to come back from Wisconsin to work in this little tiny locked room, which is the only place he can work on these data. So it's been slow-going. We finished a first paper on college match, sent it out, we thought it was good, got rejected.

So all that to say now, from the comments from that paper, I think it put us in the right direction. I think college quality is a cleaner analysis for us and helps us to more crisply show this intergenerational thing, which is really what motivated the research to begin with. We'll send this paper back out, and we'll see how it goes, but it's a story about persistence for the grad students and commiseration for the faculty [inaudible 01:02:47] and hopefully a good outcome.

Speaker 9: Also, I'm wondering the implications of your work seem really, really important, and somebody that has experienced that whole thing, now as a parent ... I have a 16-year-old, and so we're doing the SAT, we're doing all that. But this seems to be enormously important as a message for-

Mary Pattillo: Yeah, I think we want to make sure our findings are precise and strong, so, for example, I was taking down some notes about how we might make sure that's the case, when we might combine the competitiveness categories to get a stronger estimate, for example. I think we want to make sure those findings are tight and strong, and I agree, then, that closing the intergenerational mobility gap is huge. If college quality, doing that tiered by college quality, that has real implications.

Speaker 9: Also, I think, just in terms of I experienced it, I talked to a lot of students that experienced it, and they don't get after-college counseling. So we're not steered. Working with students here who've had disastrous experiences, especially when they first go to college, so I think it has a lot to ... There's a lot of implications for how, especially kids of color, need to be counseled about choosing a college and about going to college, period.

Mary Pattillo: Yeah.

Speaker 10: Pardon me. I had the same comment. I was curious about how you saw the research possibly being applied. College admissions is one of the things that came to my mind. But since that's already been mentioned, I'll just say I am not social scientist, so if you are planning to take this on the road, I'll just share a struggle I had with the data visualization was there was one chart that went this way, and I kept reading it as falling down when it's-

Mary Pattillo: Yeah, yeah.

Jordan Conwell: Yeah, yeah, yeah.

Mary Pattillo: That's a good point.

Speaker 10: So it was cognitively hard to discern what the meaning ... Actually, I think you want to fall down, or I don't know.

Jordan Conwell: You want to not fall down, yeah.

Mary Pattillo: But you orally made the point in that particular ... I remember you saying it, that, here, a lower chance of falling down is a good thing.

Speaker 10: It's a double negative.

Jordan Conwell: Right, it is a double negative.

Mary Pattillo: You're right. We have to visually I think-

Speaker 10: [inaudible 01:05:13]. The other thing I had a hard time absorbing was the controls, the ones that had all the little controls at the bottom. I couldn't figure out what being in the South was seen a positive or what [crosstalk 01:05:27]-

Mary Pattillo: That's meant to not draw your attention to those things. So, in some crowds, it's for their information who want to know what we controlled and what those controls were doing, but the attempt was to try to appease both crowds and maybe you appease neither, which is for those people who don't want to think about the controls, just look at the bigger things up there, and for those who want to know what we controlled and what were the effects of those controls, that's why those were down there.

Speaker 11: I had one question.

Mary Pattillo: Is this the last ... I see Janelle. Is that a hand or a-

Janelle Scott: Oh, no, we have until 1:30.

Mary Pattillo: Okay.

Speaker 11: This goes back to the public issue, is that the public flagship schools take a huge number of transfers whereas most of the privates don't take transfers, it's very rare. Since you have data on the number of colleges attended and you do this sort of peaking, where you're taking the most competitive one of the three that they give, can you look at the transfer effect and whether or not the transfer effect is positive?

Mary Pattillo: I think you see us both writing this down because I think that is definitely something we have to think about. In the first version of the paper, we got the comment about is it best to take the first college ever attended, highest college ever attended, or last college ever attended. We still landed on highest college, but we might have to think about transfer.

Jordan Conwell: Yeah, I think to transfer ... So whether you want to look at the first, the last, the highest depends on what the question is. One of the reviewers on that paper said we, and I'm assuming this person was an economist of education ... He said we are interested in first college because we care about transfer behavior, right, so we want to see if there are differences in starting at a community college or a low quality four-year school and moving up or starting at a very high quality four-year school and moving down. There's an argument to look at last for the kind of outcomes we're looking at because what we want to figure out ... We want the college that the labor market observes, which I think technically is the last. And, again, the result that we showed that ... If I could grab this.

This result, this kind of preliminary result, shows that it seems like by looking at the highest quality college, we are picking up for most people the college that the labor market observes. If there was a ton of transfer behavior and a kind of noise here, we wouldn't be seeing these strong of effects on income. But, yeah, we could look at did you move down in quality, did you move up, yeah, yeah.

Mary Pattillo: We could look at transfer behavior, and we might want to run the models looking at last college attended. I think first college attended is probably not the best.

Jordan Conwell: Right, for this question, it's not.

Mary Pattillo: But last or highest could be better, and so we might run it all doing last and see if we get the same thing, and then we can just put that in a footnote, and that would appease.

Janelle Scott: I wanted to ask. You started off very strikingly with a quote from Scalia and wanting to hear you talk about if Scalia were here, how the work speaks to his critique. How does it challenge what he seems to be asserting about harm? Because I'm imagining his harm is that they can't cut it, so they don't graduate.

Mary Pattillo: Yeah, so I think that has been ... I think Jordan alluded to this in his comments as well. There are lots of outcomes one could measure based on where you go to college. Those outcomes don't always go in the same direction. In fact, there's a recent paper by Lutz et al. that asks the college match question and finds that over-matching is bad for your grades, but you're more likely to graduate. Then it becomes a policy question. Well, what's more important? You have a college degree, or did you get a 4.0? I think we would all agree, and I think the labor market economists would say, having a college degree is more important than your GPA because 10 years after college, nobody ever looks at your GPA. Maybe even right out of college, nobody looks at your GPA, but sometimes in your first internship or something, they ask for your GPA. But it becomes less and less important, whereas the credential itself remains important.

So I think the literature out there ... Oh, so then adding our findings obviously shows if we take high quality colleges off the table for blacks and Hispanics who might be over-matching into them, then we are maintaining the racial gaps generation after generation. That's, I think, what our findings suggest.

Janelle Scott: Any more questions or comments? All right, well, thank you again for coming.

Mary Pattillo: Thank you.

Jordan Conwell: Thank you.

 

Mary Pattillo is a Professor of Sociology and African American Studies, and Faculty Affiliate Institute for Policy Research at Northwestern University.

Jordan Conwell is an Assistant Professor of Sociology and Educational Policy Studies at the University of Wisconsin-Madison.

Resource Type: