 Good afternoon. I'm Vincent Resch and I'm a professor of the Graduate School and the chair of the Hitchcock Professorship Committee. This afternoon, we're very pleased to present Professor Raj Chetty, our full speaker in the Charles and Martha Hitchcock lecture series. This lecture is being co-sponsored by the Department of Economics and also the College of Data Science. Computing, Data Science, and Society and also the Graduate Council. As a condition of the original request, we're obligated and pleased to tell you about how this endowment came to UC Berkeley. It's a story that exemplifies many ways that this campus is linked to the history of California and the San Francisco area. Dr. Charles Hitchcock was a physician with the US Army. He came to San Francisco during the Gold Rush and established a very thriving practice and in 1885 he established his professorship. It was an expression of his long-held interest in education, especially higher education, and his daughter, Lily Hitchcock Coyt, also a very well-known name in San Francisco, greatly expanded her father's original gift to establish this professorship and making it possible to present a series of lectures. The Hitchcock Fund is really one of the most cherished and oldest funds and endowments on the Berkeley campus in terms of bringing scholars to Berkeley as we will have a talk this afternoon. We have brought, if you look at the program, dozens of speakers over the years and really it was this original gift of Charles Martha and Lily Hitchcock Coyt that gave us this this chance to do it. So we want to thank them again. Now our speaker Raj Chetty is the William A. Ackerman Professor of Public Economics at Harvard University and he is the Director of Opportunity Insights, which uses big data to understand how we can give children from disadvantaged backgrounds better chances of succeeding. Chetty's research combined empirical evidence with economic theory to help design more effective government policies. His work on topics ranging from tax policy and unemployment insurance to education and affordable housing has been widely cited in academic, media outlets and congressional testimony. Professor Chetty received his PhD from Harvard in 2003 and is one of the youngest tenured professors in Harvard's history. Before joining the faculty at Harvard, he was a professor at UC Berkeley and Stanford. He's received numerous awards for his research, including MacArthur Fellowship and the John Bates Clarks model, middle, given for the economist under 40, whose work is judged to have the most significant contributions to the field. You will also be presenting a lecture tomorrow and that is announced in your program. So without further delay, we're pleased to welcome Professor Chetty and the presentation this afternoon will cover the science of economic opportunity, new insights from big data. Thank you very much for joining us today. Thanks so much Vince for the very warm welcome. It's really a pleasure to be here with all of you today. Thank you all for coming and I just want to say before I start that it's a particularly a joy for me to be back here at UC Berkeley. I vividly remember back in 2003 getting a call from Rich Gilbert, who's here in the audience today saying that I've gotten an offer from the economics department at UC Berkeley when I was completing my PhD at Harvard. And you know, that was really a transformative experience for me. I came to Berkeley when I was 23 years old and, you know, being influenced by people like David Card and Emmanuel Saez, who's also here, who've been tremendous mentors to me over the years has really, I think, shaped my own career and had a great impact on many of those around me as well. And so much as I'm grateful for UC Berkeley as a great public institution, I'll show you some data showing the tremendous impacts of public institutions like UC Berkeley on economic mobility in the United States, as I've seen that in my own research. Going back one generation, my family, my parents came to the United States having grown up in low income families in India to another great state institution, the University of Wisconsin-Madison, where my dad did his PhD and my mom did her medical training. And like many immigrants, my parents came to the United States in search of the American Dream, which is of course a multifaceted concept that has different meanings. But I want to start this talk by distilling that notion of the American Dream to a single statistic that we can measure systematically in the data, which I think captures a cornerstone aspect of the American Dream, which is the idea that we at least aspire to be a country where through hard work, any child has the chance to move up in the income distribution relative to their parents. Certainly it was that sort of notion that drew my own parents to come to this country and countless other immigrants. And so in a paper, some colleagues and I, including Jimmy Narang, who's a graduate student here at Berkeley, wrote a few years ago, we set about to assess the extent to which America actually lives up to that aspiration by simply measuring the fraction of kids who go on to earn more than their parents did, measuring both kids and parents incomes in their mid 30s and adjusting for inflation. Now there's some further technical details in the background. In particular, you don't have longitudinal historical data going this far back in time. So it's not that easy to compute the statistic. And so we develop some methods in the paper that we think allow us to get reliable estimates. I'll skip those details here, but let me just note more generally. I'm going to present this focusing on the key takeaways, but for those interested in discussing the technical details further, I'm more than happy to take questions on that when we get to the Q&A. So just jumping to the punchline of this analysis, you can see it quite clearly for yourself, back in the middle of the last century for kids born, say in 1940, it was a virtual guarantee that you were going to achieve the American dream of moving up 92% of children by our estimation born in 1940 in America went on to earn more than their parents did. But if you look at what has happened over time, you can see that there's been a dramatic fading of the American dream, such that for children born in the middle of the 1980s, who are turning 30 around now when we're measuring their incomes as adults, it's become essentially a 50 50 shot a coin flip as to whether you're going to achieve the American dream of moving up. And so that broad trend is of course of great interest to economists, because it reflects a fundamental change in the US economy that we'd like to understand. But I would argue it's also fundamental social and political interest because I think it's this very trend that underlies a lot of the frustration that people around the United States are expressing, that this is no longer a country where it's easy to get ahead. And that's reflected in a lot of the political outcomes we see recently, and so forth. And so motivated in part by that trend and other similar statistics. I think there's a broad interest in the public and academia and policy circles in understanding what's driving trends like the one that I just showed you and more broadly in creating equality of opportunity, help giving kids better chance of rising up, particularly those growing up in more disadvantaged circumstances, and increasing social mobility or intergenerational income mobility. There's a very large literature, as some of you here likely know, in the social sciences, in economics and sociology and other adjacent fields, analyzing the determinants of intergenerational mobility and equality of opportunity. But and while that literature, you know, I think has made quite a bit of progress over the years. There's recently been a surge of work in this area, which is what I'm going to focus on in this pair of lectures here, where the key technological breakthrough, if you will, is the recent availability of large scale longitudinal administrative data, things like tax records or other kinds of big data like social network data that allow us to follow people over time and understand the determinants of economic mobility with the precision that we've never had in the past. And so in particular, as I'll show you in this lecture, those data allow us to study the determinants of economic opportunity much more sharply for, I think, two key reasons. First, they permit disaggregation of the national picture that I just showed you in very fine ways across different subgroups, so for example, by race and ethnicity, across geographic areas, by income groups, so forth and so on. And second, they, because of their scale and scope, they permit the use of various quasi-experimental and experimental techniques that allow us to understand mechanisms much more precisely than we were able to in the past. So what I'm going to do here is present an overview of a series of papers that we've written in our research team at Harvard Opportunity Insights on these issues with many, many co-authors, John Friedman, Nathan Hendren, Matt Jackson, Teresa Kukler, Pat Klein, here at UC Berkeley, Larry Katz, Emmanuel Saez, who I mentioned earlier, Johanna Strobel, Danny Yegan, also here at Berkeley and many, many others. I'll cite those papers along the way. So let me start in this first map here by showing you one form of disaggregation, which is to disaggregate the picture of intergenerational mobility in America geographically. I'm going to first describe how we construct this map. The original version of it was from a paper in 2014 with Emmanuel, Pat Klein, and Nathan Hendren. And since we've done, since then, we've done further work on it. So what I'm going to do is first just construct, describe how we construct the map, and then tell you what I think we learned from it. So what we're doing here is taking data on 20 million kids, essentially all kids born in America in the early 1980s, and we're using data from anonymized tax returns to link them back to their parents and back to the specific place in which they grew up. We divide the US into 740 different metro and rural areas. And in each of those areas, we compute a very simple statistic, a simple measure of upward mobility. We ask, what is the average household income at age 35 for kids who grew up in low-income families? And for the purposes of this map and much of this talk, I'm going to define low-income families just to pick a specific number as families at the 25th percentile of the national income distribution, which corresponds to a household income of $27,000 a year. Okay, so for example, here in the San Francisco Bay Area, what we see using tax record data is if we take a set of kids who grew up in families making $27,000 a year, on average, if we look at those kids' tax returns when they're age 35, they're making about $37,000 a year. Okay? So similarly, we can compute that statistic for all of the different areas of the United States, and we color the map so that blue-green colors represent areas with higher levels of upward mobility, where kids are more likely to achieve the American dream in some sense, and red-orange colors represent areas with lower levels of upward mobility. So if you start by looking at the scale in the lower right-hand side of this map, you can see that even in the current generation, kids born in the 1980s who were entering the labor market in recent years, there's a tremendous amount of variation in children's chances of rising up across different parts of America. There are some places, like much of the rural Midwest, take places like Iowa, for example, where kids growing up in families making $27,000 a year, on average, one generation later, are making more than $45,000 or $50,000 a year, so a substantial amount of upward mobility in a single generation. Yet you have other places, like Atlanta, Georgia, or Charlotte, North Carolina, much of the southeastern United States, where kids growing up in families at that exact same income level, $27,000 a year, one generation later, are actually making less than their parents were. And that's kind of remarkable, given the amount of economic growth that has occurred in America over the past 30 years. So motivated by this map, you know, you can see the broad geographic patterns for yourself, right? Much of the rural Midwest has high levels of upward mobility, parts of the coast have high levels of upward mobility, much of the southeast and the industrial Midwest, cities like Cleveland and Cincinnati and Detroit have very low rates of upward mobility. You know, I think this map has motivated a lot of research and interest in our field and around two sorts of questions. So first, this sort of data gives us a new granular lens, sort of a microscope, if you will, to understand the determinants of economic opportunity, the science of upward mobility with a precision we never had before. If you just look at the initial graph that I showed you, you might have various explanations for what changed in the United States over the past 50 years, but it's going to be very difficult to test between those explanations because many things have changed in the U.S. over that period. In contrast here, you can start to ask, you know, what is it that's different about Salt Lake City versus Atlanta versus Dubuque, Iowa versus Boston, et cetera. And you can start to look, for instance, at people who move between these places to uncover what the mechanisms might be through which you're seeing these differences in outcomes across places. So this gives you a powerful way to understand the determinants of mobility just from a scientific perspective. But a second, I think equally important reason to be interested in these data, is from a policy perspective, which I'll turn to in the second of these lectures, where the motivation there is if I can figure out what is different in a place like Salt Lake City relative to Atlanta. Maybe I can start to replicate some of those lessons in these cities that are in the redder colors and increase upward mobility in the U.S. as a whole. So motivated by that logic, the way I'm going to structure this first lecture today is basically to walk through a series of potential explanations. You might already have some in your mind for what is driving the variation you're seeing in this map and systematically test a bunch of hypotheses to understand the science of what's going on. And then in the second lecture, I'll turn to think about what kinds of policies might actually make a difference on the ground in light of what we've learned about the determinants of opportunity. So let's start with the first hypothesis that I think is very natural, particularly in coming from economics, which is that maybe this is about differences in the types of jobs that are available in different places. Take the Bay Area, for example, with the tech sector booming in recent years. In the past few decades, maybe that's why you have relatively high rates of upward mobility here. So to evaluate that explanation, let's turn to this scatter plot here where we're going to take the data on upward mobility from the map that I just showed you and put that on the y-axis for the 30 largest cities here in the United States. And we're going to plot those measures of upward mobility against a simple measure of job growth rates from 1990 to 2010, the period over which these kids were growing up in those same cities. So if you look at this graph, you can see that this basically looks like a cloud. There's basically no relationship between these two variables. And in particular, you have cities like Charlotte and Atlanta, which you might know are some of the most rapidly growing cities in America, kind of viewed as the engine of jobs in the southeastern United States. If you look at any measure, the number of jobs, the number of high paying jobs, the average incomes, Charlotte and Atlanta will rank at the top of the list in terms of cities that are doing really well. Yet if you look at these longitudinal measures that we're now able to construct with these tax records, looking at upward mobility for the kids growing up in low and middle income families in Charlotte, you can see that Charlotte actually ranks the lowest among large cities in America in terms of rates of upward mobility. So you might first ask, how is that even possible arithmetically? If you look at repeated cross sections or snapshots of data, it looks like Charlotte is systematically getting richer over time, but if you follow the set of kids growing up in Charlotte, they don't seem to be getting richer over time. Like how does that add up? So the way I think about it is that Charlotte imports talent. Lots of people move to Charlotte to get high paying jobs at firms like Bank of America, for example, which is headquartered in Charlotte. But what we're seeing in the longitudinal data is that that doesn't necessarily translate to the kids in Charlotte getting those jobs, and in particular the kids growing up in low income families and disadvantaged neighborhoods in Charlotte getting those jobs. And so the first, you know, I think very simple lesson from this analysis is, you know, obviously jobs matter at a macroeconomic level and jobs can influence the strength of a local economy as well as shown by Enrico Moretti here at Berkeley and many others, but simply having more jobs in your city, for instance, getting the Amazon headquarters to move to your city is not in and of itself the solution to having more upward mobility for your residents. One has to think, I think, more deliberately about how you equip people with the skills, how you develop the human capital needed to actually get those high paying jobs. So that was the first hypothesis. Maybe this is about jobs. That doesn't seem to be what's going on. So let's come back to the big map and consider a second potential explanation this time coming not from economics but from demography. So anyone familiar with the demographic structure of the United States would recognize that there's a potential link to race here. So in particular, the red and orange colored parts of the map are places with larger African American populations like the Southeast, Cincinnati, Cleveland, and so on. We all know that there's a long history of racial disparities in America. And so you might wonder how much of the differences that we're seeing in this map are really about differences by race rather than differences by place. So to get at that, what we did next is took the tax records and linked them to census data for everyone in the US population, which gives us data on everyone's race and ethnicity in the US. And that allows you to construct this pair of maps here, the same statistics on upward mobility, but separately now for black men on the left and white men on the right. Now if you look at these two maps, initially your reaction might be it looks like they've put these maps on two different color scales, kind of a green blue color scale on the right and then red orange color scale on the left. But in fact, if you look at the bottom of the slide, you can see that we have not done that. It's just that there is such an enormous difference in rates of upward mobility in the United States between black and white men that it's almost like you have two different countries. You basically have non-overlapping distributions here, right? So in particular, if you take a place like Boston where black men growing up in low income families have average incomes of about $25,000 in adulthood, it's one of the best places in terms of upward mobility for black men, they have worse outcomes there than say white men in Atlanta, which offers some of the poorest opportunities for upward mobility for white men. So it's really like two disjoint distributions in terms of economic possibilities for black and white men. So that is to say there's no understating the importance of race in America, even today and importantly, even conditioning on class. You're taking a set of kids, all of whom start out at exactly the same income level and seeing dramatically different prospects for black versus white boys. Now you'll notice when I started to split the data by race, I also began to split it by gender. So I showed you the data specifically for men here. Why did I do that? Turns out if you make the exact same comparison for women, you get a very different picture. If you look at rates of upward mobility for black women on the left versus white women on the right, the spectrum of colors in the two maps looks much more similar. And more broadly, if you look at any measure of economic mobility, you know, rates of rising up, average incomes, rates of college attendance, things like that, once you control for parental income, black and white women have very similar outcomes, black and white men have extremely different outcomes. And so what that's telling us is there's something about the interaction or intersection between race and gender that's extremely important in understanding racial disparities in America. You might think of things related to the criminal justice system. You might think of discrimination that's particularly affecting black men in the labor market. I don't know if we know for sure yet exactly what the answer is, but it's clear that whatever it is has tremendously divergent implications by gender, which I think is useful in narrowing down the possibilities. So what we can see from these maps is that race is undoubtedly important, but even conditional on race, I want to emphasize that there's a lot of difference across places, right, if you look at the map on the right. And so, you know, our takeaway from this and other related analyses is that race is an incredibly important determinant of economic opportunity, but place matters as well. And I will dive into that in further detail in a second. But before I move on from these racial comparisons, I want to make one further point. Throughout this talk, I'm going to focus mainly on upward mobility, because I think most people are interested in that phenomenon. How can you help more kids rise up and achieve the American dream? But I think it's equally important, particularly in understanding the persistence of racial disparities in America, to think about the converse phenomenon of downward mobility that is kids who start out in high-income families and ask where they end up in the distribution. And to show you that data, I'm going to turn to, I think, a powerful visual that The New York Times put together using our data that captures this point very saliently. So the way to think about this is it's basically a depiction of a transition matrix. We're going to take a set of kids who start out in high-income families in the top fifth of the income distribution and ask which quintile of the income distribution they themselves end up in as adults. Do they drop all the way to the bottom fifth? Do they stay in the top fifth and so on? Purple dots are for black men. Green dots are for white men. And what you can see is there's an enormous divergence here for black versus white men. For white men, if you grow up in a high-income family, you can basically expect to float at the top of the distribution. You're likely to be at the top or in the upper middle class. For black men, I think tragically, even if you grow up in the highest income families in America, odds are you will end up in the middle class or even at the bottom of the income distribution in the next generation. So this result, which I think is very disturbing about the United States and certainly was a great surprise to me when we first saw this in the data, I had expected to see when we began to study racial disparities that there would be some divergence by race. But I thought that race might start to matter less as you had sufficient income. If you were from a very high-income family, you could go to the best schools, live in the best neighborhoods, etc., maybe race would become less important. So that hypothesis turns out to be totally wrong. There are enormous differences in outcomes. Even for black and white kids who live in the most affluent neighborhoods, go to the same schools, etc. And I think recognizing that is extremely important to think about how you tackle racial disparities in the U.S. because in some sense, you know, make a visual analogy. For white Americans, achieving the American dream is like climbing a ladder where you kind of start off in the next generation, where you left off in the previous one. Whereas for black Americans, it's more like being on a treadmill, where even after you rise up in one generation, there are these tremendous structural forces that push you back down, only to have to make the rise again. Why is that important to recognize? If you think about a model of income dynamics and how income is going to evolve across generations, if you've got this force of downward mobility, no matter what you do to lift the fortunes of black kids who are currently at lower levels of incomes, that gap between black and white outcomes is just going to reemerge in future generations as people fall back down this treadmill. So you know, the key takeaway here, I think, is if we want to address racial disparities in a sustained way in the long run, we not only need to focus on the most disadvantaged neighborhoods and schools that have less resources and so forth, but also think about why black kids who apparently are growing up in relatively advantaged families, why they are not maintaining those positions, that's really the only way ultimately one will narrow racial disparities in the long run. Okay, so I've shown you a few broad facts about the United States, these broad geographic comparisons, highlighting the importance of race, not so much the importance of differences in job availability, and so forth. So now what I want to do next is dig in deeper to understand what the roots of these differences across places are, conditional on race, and in order to do that, the next thing we did is began to look at these data at even a finer geographic level. So if you look to the sociology literature and the long literature on the potential influence of neighborhoods and environment on people's outcomes, nobody would think of your neighborhood at the level of the Bay Area or Boston or New York City, you'd think about the specific town you live in or the specific school district you're in or specific group of people you're associated with and so on. So to do that, what we did is went back to these data and began to look at the data at a finer geographic disaggregation and the easiest way to show you that is to toggle over to this website, let me just see if this is going to, yep, this comes up here called the Opportunity Atlas, which you yourself can access just a freely accessible website, and the way this works is like a Google map, you can enter in any address you like and literally zoom in to see the data in that area. So given where we are, I'm going to enter an address in Oakland, which I'll tell you about in a second. And so now we're going to zoom in and see these same statistics that I've been telling you about before, but now at a census tract level. So there's 70,000 census tracts in America, each of which has about 4,000 people. There's enough data here because they have data on the entire US population. You can construct quite reliable statistics on these rates of upward mobility for every census tract in America. And so that's what we're doing. It's the same exact statistics I've been showing you, but now just in the Berkeley, Oakland, Bay Area. And so the first observation I want to make before I talk about specific cells of data here is just a simple one. The spectrum of colors that you're seeing on the screen here is the same as the spectrum of colors that you were seeing in the national map. That itself tells you something about the origins of differences in economic opportunity. This is not about state level differences or differences across cities in terms of opportunity and rates of upward mobility. No, it's really about going two miles down the road in Oakland and going over to Alameda or differences between one part of Berkeley and parts of Emeryville in terms of rates of upward mobility. In this area, as in areas, cities across the United States, you can drive two miles down the road and it's like you're going from Alabama to Iowa in terms of rates of upward mobility. So what that tells us, I think, is that the roots of differences in economic opportunity are hyper-local. It's not about we shouldn't just look to broad differences in state level policies, or I was talking with Erin earlier, you know, is it about federal policies? I actually think these data suggest that there's a lot that that can be done at a local level. If we can just figure out what is different about these environments that seem to have much better outcomes relative to other nearby areas that have much worse outcomes even in the current era in the same time period. So this particular address that I entered in is, it's in a census tract in the center of Oakland and that happens to be a place called the Adeline Lofts. It's an apartment complex in Oakland, which was built with low-income housing tax credit, light-tech funding. So it's affordable housing. And so the kind of problem I want to think about and I'll get more into this in the second lecture when we think about policy solutions is does it make sense to have tax dollars going to build affordable housing here as opposed to other places where you might be able to also cite affordable housing? For instance, you know, just to pick, you know, an example of another census tract, which I'll come back to in a second, you know, take this place, you know, over here in Alameda where you see much higher rates of upward mobility. So now one thing I just want to note when you look at this map, you might wonder how much of these differences are due to differences by race, right? I just pointed out earlier that there are sharp differences in upward mobility by race. You all know that the Bay Area, like many other places, is quite racially segregated. So one thing you can do with this tool, which is helpful in that regard, is we can subset the data to different subgroups. So if I click black here, first of all, a lot of the map is going to disappear because there are no black people in a lot of the Bay Area. But where you do have a black population, you can see that even conditional on race, we see very different outcomes for kids growing up in the center of Oakland versus those tracks in Alameda, which turns out are also the site of low-income housing tax credit developments. And so, you know, one wonders like would outcomes be very different if more low-income kids were to grow up here rather than here? And so this level of granularity can, you know, give us a sharper sense of what the determinants of opportunity are. So motivated by that, I'm going to now come back to the slides and dig in to understand in a little bit more detail what is going on at that census tract level of granularity. And in order to do that, you know, so I mentioned the Adeline lofts before, and here's another light tech development called the Playa del Alameda apartment complex. In order to understand what is different about these two places, the next thing I want to do is look at families that move across these different neighborhoods. So why is that interesting to do? So before we get into what that analysis reveals, you know, just from a social science perspective, there's been a long-standing interest in understanding how much of the difference in outcomes we see across people is due to nature versus nurture, environment versus other factors that may be harder to manipulate. And so you may wonder, you know, how much of the outcomes we're seeing, difference in outcomes we're seeing across these places are just that different types of people live in these different places, and maybe that's where we're seeing differences in outcomes, and how much of it is actually about the causal effect of growing up here versus there. So if you think about it conceptually, what you'd like to be able to do if we were able to run experiments is basically randomly assign kids to grow up in these blue-colored areas versus the red-colored areas and compare their outcomes over time. That is obviously a very difficult experiment to run in practice. But the power of large-scale big data administrative data is that we can start to approximate these experiments with the data we have in hand, the observational data that we have. And so to show you that, I'm going to talk about a study where we look at five million families that move across neighborhoods. But rather than getting into the statistical details of the study here, I'm going to summarize what we find in the context of a simple example, that example of those two places in Oakland and Alameda that I was showing you on the map. So imagine a set of kids who move from Oakland to Alameda at different ages, starting with kids who move when they're exactly two years old. So what we're doing in this first dot here is think of it as take the set of kids who move from Oakland to Alameda when they're two, track them forward 30 years in the tax data, and measure their own incomes when they're adults. And what we can see, consistent with the kind of average levels of earnings we see for kids who grow up in birth from Alameda, is that those kids earn about $25,700 per year in adulthood. Okay, so that's for the kids who move when they're exactly two years old. Now let's repeat that analysis for kids who move when they're three, four, five, and so on. Make the exact same move. You get this very clear downward-soping pattern. The later you make the move from Oakland to Alameda, the less of the gain you get. And if you move after you're in your early 20s, the relationship becomes completely flat, and you get no further gain at all. So what do you see from this chart? I think there are three key takeaways, all of which are predicated on a key identification assumption, which I will come back to in a second, for those who may be wondering about how we're identifying causal effects here. But if you just take this at face value for a moment, I think there are three key lessons. The first is that where you grow up really matters, right? So it's not just that the kids who live in Oakland are different from the kids who live in Alameda. Apparently, if you take a given kid and move that kid, particularly at a young age from Oakland to Alameda, you see very different life outcomes for that child. Now, in order to make that strong causal claim, I have to make the key assumption that the types of families who are moving from Oakland to Alameda are more generally from a red-colored place on the map to a blue-green-colored place on the map, which is how this is identified with millions of families who are moving across such places. I have to make the assumption of constant selection. The types of families who are moving from better to worse neighborhoods, it's fine if they're selected families, but the nature of that selection cannot vary with the age of the child at the move. If the types of families who are moving to better places when their kids are young are different, more educated, wealthier, etc., from the types of families who are moving when they're older, then this is going to be confounded and we may not be able to, we may not be picking up the causal effects of place. So as you can imagine, this is really the core focus of our work in this space, trying to understand the validity of that assumption. I'll give you one simple example that makes us confident that this assumption holds. So one thing we can do is replicate this figure, looking at siblings within the same family. You have enough data here that you can look at a family that moves with, say, a two-year-old and a six-year-old from Oakland to Alameda. And what is amazing, you can look at the paper for these figures, you will, you get exactly the same figure back if you put in family fixed effects and only rely on sibling comparisons. So these effects are emerging within families, showing that it can't just be that, you know, different families are moving at early versus late ages, that kind of test, you know, rejects that possibility. And there are various other tests we do along those lines, which I'm happy to say more about. So point number one really seems like where you grow up matters. Point number two, what really seems to matter is childhood environment rather than where you live in adulthood. If you move in your early 20s or after that, we find in this study and many other studies that that has very little impact. And the third point is that there's sort of a dosage effect here. Every extra year that you spend growing up in a better environment, in Alameda in this case, or more generally in a blue-green colored place on the maps I've been showing you, the better you do in the long run. If you move to that place when you're two instead of three, five instead of 10, there's a cumulative gain from moving earlier. That result, I think, is important in light of recent policy discussions on early childhood education. A lot of focus on programs like preschool intervention, universal preschool, headstart, etc. Our sense is those kinds of programs can be quite valuable, but we don't want to take it to the extreme of saying everything is determined, you know, by the time you're five years old and there's not much value in investing beyond that. Clearly there's an enormous amount that you can do even beyond those very earliest stages. So one of the nice things in this literature as we and others have been working with these data is to see, you know, how our understanding of these questions evolves as other people do work on related topics. And so this dosage effect that I just showed you established in the U.S. data turns out to be an incredibly robust pattern established now in many different studies. I'm giving you a sampling here from different countries using different research designs. In some cases a pure experimental research design going back to a famous experiment called Moving to Opportunity that was conducted in the 1990s to other quasi-experimental designs that use public housing demolitions and so forth. There's really this quite robust, I think, consensus now in the social sciences that environment matters and it matters through this dosage effect particularly in childhood. So having established that, that this really seems to be about the causal effects of childhood environment as a key driver of economic mobility. A natural next question that you're probably wondering about is what is it that's causing these differences in outcomes across places? Why is Alameda producing higher rates of upward mobility than other parts of Oakland? And so to get at that there's now quite a large literature trying to investigate that question. Basically taking the data from the Opportunity Atlas that I just showed you and correlating it or using quasi-experimental techniques to try to understand the determinants of that variation now that we know that it's largely due to the causal effects of place. And there are many, many different factors that people have identified. I won't read all these off but ranging from things like poverty rates to the level of income inequality in an area to levels of segregation, measures of family structure and so on. And I think the picture here is complex. My reading is all of these things can be quite important and there are many different things that matter. What I'm going to do in the talk here is go into more depth on one particular factor, social capital, that many people have speculated over the years might be very important for determining many outcomes. Going back to the famous work of people like Glenn Lowry, James Coleman at the University of Chicago, my colleague Bob Putnam at Harvard, the idea that the strength of your community, who you're connected to, your networks might matter for economic outcomes. That's something a lot of people have theorized about, had small scale evidence on. But I think it's been hard to systematically measure and test partly because we've not really had data on social capital on who people are friends with on the structure of networks and so on to evaluate these sorts of hypotheses. And so I'm going to spend some time on that here both because it happens to be the most recent thing that we and others have been working on, kind of visit the frontier of the field. But also I think, as I hope to convince you, is one of the most important factors and cuts across a lot of these other explanations that are listed here. So to get into this, I'm going to show you a different map first, drawn from a completely different data source. So when we decided to study social capital, we set up a collaboration with Meta, the company that operates the Facebook platform, which of course gives you tremendous social network data on who people are interacting with. And we constructed various measures of social capital using that network data. And I'm going to start by focusing on one particular measure that I think turns out to be particularly important, which I'm going to call economic connectedness, basically a measure of the degree of cross class interaction in a society. So again, let me describe how we construct this map and then tell you what I think the lessons are. So what we're doing here is taking Facebook data on everyone between the ages of 25 and 44 in America, that's 72 million people. It's about 84% of the US population in that age range. So Facebook is not a universal data set like the tax records. It doesn't literally cover the population, but 84% it's pretty good. And our sense is there isn't a huge selection problem here. Okay, so we take that and here at the county level, for every county, we first use at the national level a machine learning algorithm to assign everyone incomes. And then in each county, we construct a simple measure of the degree of cross class interaction. We ask, if you are a below median income person, what fraction of your friends have above median income? How much are you interacting with high income people, basically on Facebook? Red colors are places where there are fewer cross class friendships, disconnection across class lines, blue-green colors here are places with more cross class interaction. So you probably immediately recognize that this map looks incredibly similar to the map of upward mobility from the tax records that I started out with. And so indeed, you can do a scatter plot of the data on upward mobility from the tax records against the data on cross class interaction from the Facebook data. And you can see those two things are incredibly highly correlated with the correlation of 0.65, very different from that job growth scatter plot that I started out with. So there is indeed a tight relationship between these two variables. Now, you'll note, I've picked one particular measure of social capital here. Those of you familiar with this literature might recognize there are many other ways you can think of measuring social capital, how cohesive networks are, how many friends you have, how much people are volunteering, levels of trust. There are many different concepts that people have talked about over the years. In this paper, we construct many other measures that have been suggested in the literature and we correlate all of those measures with the data on upward mobility. I showed you the correlation of 0.65 in the previous slide with this measure of economic connectedness. It turns out there's a very stark result here. All of these other measures of social capital that sociologists, network theorists and others have proposed over the years are basically uncorrelated with levels of mobility. So this is not to say they are not important for other outcomes. I want to be clear on that, but at least for the topic we're focused on here today, it seems like this cross-class interaction phenomenon in a predictive sense is particularly important. So why is that? Coming back to this plot here. So we have seen and we've shown in this work that if you do that same movers analysis I was showing you earlier, if you move to a place where there's more cross-class interaction at a young age, you have better outcomes in adulthood. And you can tell various causal stories for why that might be true. If you're connected to people who have higher incomes, we know many jobs in America are obtained through referrals. So maybe that kind of network helps you directly get a referral or an internship that makes a difference. But I think even more importantly, a very plausible mechanism here is that your aspirations or what you aim to do in life, what you think is possible is greatly influenced by who you're around. If you've never met anybody who's gone to college, maybe you never even think about the possibility of going to college. If you live in an environment where some of your friends' parents went to college, became scientists, became lawyers, you know, whatever it might be, those possibilities start to become things you aspire to do, start to become things you work toward. So that I think is a plausible explanation for at least part of this correlation. But there are other explanations for this correlation that have nothing to do with the causal effects of networks or social capital. So to give you an example, take San Francisco on the far right there, which looks like a place with a lot of cross-class interaction overall and very high rates of upward mobility. One potential story is that there's a causal effect of those cross-class interactions on upward mobility in San Francisco. Another possibility is that relative to other places in America, like Indianapolis, the Bay Area is a relatively rich place. It's going to have more funding for schools, other types of public goods, other types of resources. Maybe it's just the average levels of income, which are what are leading to these greater interactions with high-income people to begin with, but maybe it's just the average levels of income and the resources associated with that that are leading to higher levels of upward mobility. So to show you how we can disentangle between those explanations, I want to turn to this chart here, which I really see as the key thing from this analysis of social capital. So let me walk through it in a couple of steps. So what we're doing here is each dot represents a different zip code in America. And we're plotting first just the share of high-income friends that low-income people have in that zip code. So our measure of economic connectedness from the Facebook data against just how rich people in that zip code are. Median incomes in that zip code. And you can see there's a very clear upward-sloping relationship here, as you might expect intuitively, because you tend to be friends with the people around you, if you live in a place where people are richer, you tend to have more high-income friends. So that kind of makes sense and is a bit mechanical. Now notice, despite that relationship, there's still a lot of dispersion vertically in this chart. So at a given, with a given income distribution, there's still big differences in the share of high-income friends that people have in some zip codes versus others. So now I want to get to the key point. Let's now color these dots by the data from the Opportunity Atlas, tax data measures of upward mobility, where let me remind you red colors are places with lower levels of upward mobility and blue colors are places with higher levels of upward mobility. So what you see in this chart I think is a very clear and striking pattern, which is that if you take any vertical slice of this chart, you see that the colors change systematically from red to blue as you move up. So that is to say if I take a set of zip codes, all of which have median household incomes of about $50,000 a year, they all have similar resources in some sense, but I go from the zip codes where low-income people are not connected to, not interacting with high-income folks to the places where there is a lot of cross-class interaction, I see rates of mobility changing systematically. In contrast, if I do the converse and take any horizontal slice of this graph, I go from a place that has much fewer resources to a much richer place holding fixed the level of cross-class interaction, there's no change in levels of economic mobility. So that suggests quite strongly that what matters is not just average incomes in an area, but actually the degree of cross-class interaction conditional on the level of income, which points more in the direction of the view that this cross-class interaction really matters. Now this, of course, is just a non-parametric depiction of a two-variable regression. That's what we're doing here. And so to show you that this phenomenon actually turns out to be pretty robust across different explanatory variables, I'm now going to, in a more compact way, just show you a couple of regression results that I think further illustrate the power of this cross-class interaction variable. So I'll give you another example of a similar exercise. There's a long literature in economics that shows that there's a cross-sectional link between the level of income inequality in a given generation and rates of mobility across generations. So this was first established by an economist named Miles Korak and popularized by the late Alan Kruger in what he coined the great Gatsby curve, the link between inequality and mobility. And that's what's replicated in this first regression here of upward mobility on the Gini coefficient within a county. In the second column, we repeat that regression, but now control for the level of economic connectedness, this new Facebook-based variable. And you can see that two things happen. First, economic connectedness is a very strong predictor of mobility. And second, the relationship between inequality and mobility basically disappears, exactly like I was showing you in the previous chart with poverty rates. So in that sense, statistically, connectedness explains the link between inequality and mobility that's been widely discussed in the public discourse and the prior academic literature. Another example. So my colleagues, David Cutler and Ed Glazer have a famous paper in the quarterly Journal of Economics in 1997 where they noticed using more limited survey data that correctly, segregation is extremely harmful for blacks. If you look at black kids growing up in racially segregated neighborhoods, large black share, they tend to have lower rates of upward mobility. But they know it in their paper in the conclusion, but we do not have an exact understanding of why this is true. So again, we replicate their analysis, but now control for the level of economic connectedness cross-class interaction. And once again, you see that exact same pattern, right? The lack of connectedness in those predominantly black communities can, in a statistical sense, explain why you see much lower levels of mobility there. And so based on that set of results, our sense is that this measure of cross-class interaction is extremely important in understanding economic mobility in combination with other factors like the quality of schools and other things for sure. But this is certainly something that warrants attention. And so I want to make one final set of points on this before turning to a couple of other things and concluding today's lecture. So I've shown you that there are these big differences. There's this link between economic connectedness and mobility. And so you might wonder, okay, if we think this cross-class interaction is important, how can we get more of it, right? So I don't have a definitive answer for that, but I think it's useful to start at a conceptual level by thinking about two different factors that matter for the level of cross-class interaction or connectedness in a society. The first is just simply exposure. So to give you a simple visual example, imagine you've got two schools and imagine that all the high-income kids shown with the green circles go to the first school and all the low-income kids go to the second school. So since you can't be friends with people you never meet, obviously this is going to generate a very disconnected society across class lines. So that's one potential explanation for why we have stratification by class in terms of social networks in America, just segregation. But that's not the only possibility. You could also have this situation, a situation with what we're calling friending bias where you have perfectly integrated schools, yet if you look at the friendships that are formed shown here by these lines, you still don't have any cross-class friendships because all the high-income kids in a school hang out with each other and all the low-income kids in a school maybe hang out with each other. Which of these two things is driving the lack of connectedness is extremely important to understand from a policy perspective because if it's the former then you can think about policies like changing zoning laws, redistricting, busing, etc. that create more exposure. If it's the latter that's not going to you know that's not where the problem is. You've got to figure out what's happening that's creating a lack of interaction within a given room. So the power of the social network data, the Facebook data, is that we can actually start to disentangle these phenomena. And the way we do that is by taking the friendships that we see, the billions of friendships, and turns out that you can you actually have enough information to make a pretty good guess about where those friendships were formed. So we can make a good guess about whether you made your Facebook friends in a high school, in a church, in a specific college, etc. by looking at you know where you both were at certain times and so on. So I'll again spare you the details. So we do that for all of the friendships and using that we're able to construct these very precise measures. Here I'm showing you the data by high school of these two key axes that matter for connectedness. The frending bias measure, so conditional on a level of exposure. How many high income friends do you make? One way to think about it statistically is how much are you sampling non-randomly from the set of friends, set of peers you have in your school. That's what's on the y-axis. And the x-axis is just how many high income students you have in your school, just a measure of exposure. So again, take a local example, Berkeley High School. Berkeley High School on the surface looks like a very diverse high school. It's got a pretty mixed income distribution. If you look in the Facebook data, Berkeley Public High School is one of the schools in America with the greatest level of frending bias. It's the most separated by class among schools in the United States. And so you see that pattern more generally among big public diverse high schools where at some level you have integration, but in practice you actually don't have integration when you look at what the networks look like. And as we've seen in the slides I was showing you earlier, it's really the actual cross-class interaction, not just being in the same building that's predicting good outcomes for kids from low income families. So using this kind of data, and again we've made this publicly available, so for the students here, if you're interested understanding what's driving this variation and so on, I think there are lots of interesting questions here, can be analyzed with these publicly available data. We can take that and go back to the question I raised initially. How much of the social disconnection in America by class is due to exposure versus frending bias? And it turns out if you do various reweighting exercises, basically the answer is 50-50. Half of the disconnection is due to a lack of exposure, the fact that poor and rich kids live in different neighborhoods, go to different schools, go to different colleges, and half of it is due to frending bias. So even conditional on exposure, there being disconnection in terms of who people interact with. Now at some level this all seems maybe a little bit abstract and we're figuring out seems like we're figuring this out with these new data, but at another level I think these are totally familiar ideas to anyone just in their own life from introspection, just to give you a sense of that. I want to give you a quote here from Carmelo Anthony, the famous basketball player from his recent memoir, which I think captures this idea in very simple terms. He talks about his experience growing up in Baltimore where he notes that millionaires could live on one side of a street and the projects could be on the other side. In that sense it was a very integrated place. If you looked at like that census tract it would be very integrated by income, but he notes those two worlds would never cross, never make friends, never acknowledge each other. In our jargon he's basically saying there's a lot of frending bias. Everybody was okay with it especially the rich and I think it's that kind of situation that explains why Baltimore in particular was actually a place with really poor outcomes for kids from low-income families. So I think this frending bias phenomenon how you actually get people from different backgrounds to interact with each other is very important to try to think about how we address going forward, much as how I think we devote a lot of attention and policy circles to thinking about issues of exposure and segregation. There are numerous policies focused on zoning laws and other kinds of things that are about getting people into the same sorts of neighborhoods, but I think the latter component is also important. I want to show you one final piece of data here which is I don't want to leave the impression that this frending bias is somehow some intrinsic thing that cannot be changed through policy and only exposure is manipulable. And to give you a flavor for that and I'll come back to this more in the second lecture as well, just want to show you a couple of pieces of evidence that I think support that view. So one point to note is that the level of frending bias varies substantially across settings. If you take a given set of people and look at the friends that same set of people make in different settings, they exhibit much more frending bias in certain settings like the friendships they make in their neighborhood or their college relative to say their religious institution or recreational group where people are much more likely to make friendships that cut across class lines. Another example relevant to the school's context and consistent with the example of Berkeley High School that I gave you, if you look at the size of a school or the size of a group more generally and the level of frending bias, people tend to come apart in big groups. So in very big schools you find the other kids who sort of look like you and make your own subgroup in smaller groups, you kind of end up interacting with everyone by the end and that leads to more of these connections. This is not to say, you know, this is in and of itself the solution, you know, maybe creating smaller cohorts which actually turns out Berkeley High School is trying to do at some level. You know, maybe that can make a difference but more generally my point here is this seems malleable to some extent and we should be thinking more about how we can make a difference on that dimension. Okay, so I want to open it up to questions but before doing that I want to show you one last set of data which I think is particularly relevant in this context which is about the role of higher education in economic mobility. So to this point I've been focused largely on disaggregation across neighborhoods and schools, K through 12 schools and so on. Of course a natural next important junction in the pipeline to opportunity is where we're all sitting right now, right, institutions of higher education. Many people view institutions of higher education as the realm where the playing field gets leveled where you can really create equality of opportunity and give people pathways to upward mobility. So the last set of data I want to show today is to explore whether that's actually the case in practice. Try to understand how colleges affect economic mobility and here we're going to use data again from the anonymized tax records that we've used in the other studies linked in this case to Department of Education records and data on everyone's SAT and ACT scores for everyone who took the SAT and ACT in the US. So let me start with this chart here which comes from a paper with Emmanuel and Danny Yegan and others where first I'm just going to show you some data on what mobility looks like across colleges in America. So when we're thinking about colleges there are two dimensions that matter for a college's contribution to economic mobility. The first is what I'm going to call the upward mobility rate. If you take a set of kids from low-income families in the bottom 20% of the income distribution and ask what fraction reached the top 20% of the income distribution that's a measure of how upwardly mobile the student body is. Like how much are you helping the low-income kids on your campus rise up when you measure their incomes 10 years after college using tax data and on that measure you can see that a lot of the highly selective colleges in America like my own institution currently Harvard, Stanford, Princeton etc. look terrific as does UC Berkeley but that's not the only measure that matters for a college's contribution to economic mobility. What also matters of course is how many low-income kids you have on campus to begin with and on that measure if you look at places like Princeton and Harvard only two or three percent of kids at those colleges are coming from the bottom 20% of the income distribution you're about 80 times more likely to be at Harvard if you're from the top 1% than if you're from the bottom 20% so these colleges are enormously skewed towards kids from high-income families. Berkeley has more low-income kids than Harvard it's not enormously large but it's still significantly more more like eight or nine percent as opposed to three percent but then as you go over to the right side of this distribution you can see that there are many colleges in America each one represented by a different dot that do serve many many low-income kids right but there you have a different challenge which is if you look at a lot of those institutions typically two-year institutions or community colleges you don't see very good outcomes at those places and so at some level one way to think about what the problem is in the higher education system in the United States in terms of contributing to economic mobility is that we basically don't have many dots in the upper right here we don't have a lot of places that serve a lot of low-income kids and have excellent outcomes now one reason that might be the case is that you know there's a limited capacity for colleges to do anything about these issues because of all the other disparities that we've talked about starting at birth from the neighborhoods you live in to who you interact with the schools you attend so forth and so on you know one view you might have is maybe there's just very little capacity for these colleges to admit more qualified low-income kids while maintaining their selectivity standards and there's just limited capacity for these colleges to do better given all the challenges kids have faced to that point so to interrogate that further one way we can do that is by using data on SAT scores to get a sense of kids qualifications kind of where they are at the point that they are applying to an entering college and so at some level there is quite a bit of truth in that view there are a lot of disparities that emerge by the time you're 18 if I plot for instance the fraction of kids who are scoring above 1500 out of 1600 on the SAT that puts you in the top 1% of test scores on the SAT by parental income it is indeed the case that there's a very steep gradient here you're about 30 times more likely to have an SAT score above 1500 if you're from the top 1% relative to being from the bottom 5th and so at some level you know you take a place like Berkeley trying to admit very highly qualified kids because of this fact whatever you think about whether the SAT is biased or not and whether we should go test optional I'll touch on some of those issues in the next lecture but there's some content in this as I'll show you direct data on in the subsequent lecture but you know given that there's going to be a limit to how much you can do at colleges but that being said I don't think we should let colleges completely off the hook because there are still significant differences in attendance rates even conditional on SAT scores and so just to give you a quick flavor of that let me toggle over to one last thing here so I'm going to go over to this website that the New York Times has constructed to look at these college level data and here what you can do is type in the name of any college so let's type in Berkeley so first what we're going to do is just look at attendance rates for kids who attend Berkeley for kids at Berkeley in general by parental income based on information from parents incomes from tax records right and so you can see at Berkeley there's kind of a flat distribution and then it does take up pretty sharply at the top where Berkeley has many more kids from high income families than from middle class and lower income families but what you can now do is ask using the SAT data suppose we take a set of kids who have the same qualifications when they're applying as measured by the SAT we can click on this button here and see what happens to the distribution at Berkeley turns out it totally flattens so at Berkeley there doesn't seem to be much of a difference in attendance rates conditional on SAT scores in some sense there's a limit to what you see Berkeley can do further in terms of creating equity unless of course one wants to have sort of class-based affirmative action and have more kids from lower income families admitted but Berkeley and a number of other public institutions are somewhat distinct in that way so now if I type Stanford okay so Stanford looks completely different where you can see Stanford even conditional on SAT scores has this kind of u-shaped distribution with much fewer kids from the middle class and then a real uptake at the very top of the distribution where even if you take two kids with the exact same SAT score you are far more likely to be attending Stanford if you come from families making more than $600,000 a year than if you're from the middle class or a lower income family and so just to come back to that here and wrap up for today you know if you look at that more systematically and look at Ivy League Ivy Plus colleges the Ivy League in schools like Stanford elite private institutions look at your probability of attending these colleges four kids all of whom have say exactly the same SAT score an SAT score of 1510 which is exactly the 99th percentile threshold you see this very clear upward sloping pattern where you're like two or three times more likely to attend these colleges even holding fixed SAT if you're from a high income family and what that shows you then is that actually there is probably scope going back to this plot to do something at the college level to move these dots to the right and then in a different vein to think about how you move those dots upward and potentially amplify the impacts of colleges on upward mobility as well so you know more broadly you know the point here is I think piece by piece by looking at this pipeline with these granular data we can figure out where the shortfalls sort of are and make progress in tackling these issues going forward and so I'm going to skip a couple of things here and conclude for today by just highlighting three takeaways I've shown you lots of different data but you know I hope you'll take three main messages away first you know I think childhood environment plays a central role in shaping prospects for upward mobility through a dosage or exposure effect second I think a lot of the reason environment matters is because of social capital often in economics and in policy debates we focus solely on financial resources but I think there's an important complementarity between financial resources and social capital that can be very influential and you'll see how that guides a lot of the policy interventions we're thinking about going forward in tomorrow's lecture and then finally in some separate work that I didn't talk about here our sense is focusing on creating equality of opportunity can be very useful from the perspective of reducing inequality which is what I think motivates a lot of people to be interested in these issues but can also have payoffs in terms of increasing economic growth this is not a zero sum game you know you bring more people through the pipeline they start new businesses they invent new things that benefit everyone and I think there's clear data supporting that as well so that's what I have to say for today tomorrow I'm going to take the set of findings shared today and try to think about what we can do to increase upward mobility on the ground talk about some policy pilots and interventions that we and others are implementing in this space and I'll focus on three areas reducing segregation making strategic place-based investments to kind of turn the red colored places into blue colors on the maps that I've been showing you and what we can do to improve higher education the last set of topics that I focused on but let me stop there for today and thanks so much so if anybody has any questions we have a microphone up in the front if you could please stand in line up there and we'll take them thank you and please try to make them brief as a question hi um so if I look at what we at an institution of higher education can do um sack bleemer has recently been coming out with data which shows that uh limiting the enrollment in lucrative majors like economics like computer science actually causes seems to be causally related to racial stratification and limitation of upward mobility um you know in a way that is correlated with that racial stratification so what's your thought about that and about how that might give us a way to move this in a different direction yeah great thanks so much so you know we're zack as a terrific graduate of the berkeley phd program we're actually working with zack on these issues now uh with these data and you're absolutely right so you know as you start to interrogate what is going on at some of these colleges where you're seeing lower levels of mobility part of it is about the types of things that people are majoring in which may be partly about major restrictions like zack documents in one of his papers it may also be about other factors that lead people for example to be discouraged from pursuing a class you often see gender disparities for instance emerge where women if they get it there's some evidence which is just that if women get a bee in an introductory class they may choose something else men are happy to plow ahead you know with the bee or whatever the great is and so i think there are lots of things underneath the hood like that to try to unpack i'll show some evidence on this tomorrow take city university of new york which is a little bit of an outlier in that graph that i'm showing you what are they doing they have a number of interesting programs that try to really support kids in the pipeline help them graduate help them choose majors that can work well for them uh etc so i think all of these things are the types of things to be thinking about hi i'm carissa i'm a student in the econ department here i think your research is so interesting and so necessary and what i see in it what i identify with is an enormous love for this country anyone who hated america wouldn't spend nearly this long thinking about it so i wonder after all of this at a very high level if you think there's any vision that you have of a new american dream or if you think um we should move away from it altogether yeah thanks so yeah i mean on my view is the us has traditionally been very focused on being a capitalistic engine kind of maximizing growth and that can have certain benefits of course in terms of creating more innovation more discovery entrepreneurship and so on as i think we're all well aware the fruits of those returns are very unequally distributed as shown clearly in work by immanuel sayaz here and gibral zikman and and others and part of what i think we're trying to show here is that thinking about the dynamic process the equality of opportunity in particular can be a useful lens to try to restore the american dream in some sense and as i'll emphasize in tomorrow's lecture i think that way of thinking about it as opposed to focusing on inequality of outcomes the distribution of income which may also be important obviously in its own right when you focus on equality of opportunity it can be a broad tent that brings people together so in my experience you find people on the right people on the left all of whom are very interested in trying to figure out how to tackle this problem because i think a lot of people believe it's just a fundamental ideal in america and beyond that we want to be a place where people can rise up whether you know one can get back to where we were in the 1950s i don't know but my view is step by step with the kind of evidence we have here making scientific progress i think one can make progress in the right direction my question has to do with that first graph that you showed in terms of the decreasing american dream and i'm curious over that time from the 1940s until i guess it was the 1980s a larger percentage of generations were going to college so since college connects with income level is there something where you're kind of tapping up against how educated the children can be right yeah great question so part of what i think is going on here at a macroeconomic level there's a nice book by my colleagues claudia golden who just won the novel prize and larry cats called the race between education and technology that kind of captures what you were getting at in your question where they point out that if you look at trends in levels of education just measured say by the number of college graduates or total years of education up until 1980 it was going up steadily and then it completely plateaus and falls off the trend line and so the way they think about it is there's constantly technological progress globalization there are these forces that sort of human beings have to compete with in terms of getting wages up until 1980 we were keeping steady in that race after 1980 we fell behind and you know one way you can look at it is that is one of the core reasons the american dream has really faded here and my view is it's not just about years of education but the quality of education the nature of the childhood environment america has become more segregated there are many elements related to what i've been showing you in the more modern data that i think have changed in the time series as well now you raised the question of whether there's a ceiling i don't think there's a ceiling here that kind of says you know we've hit the limit now only 50 percent of people are going to do better than their parents and the way you can see that and we actually show this in this paper is the overall size of the pie in america is still growing very rapidly growth rates are very large average incomes are going up quite a bit it's just that the distribution of that income growth is far more skewed than it used to be in the past and so you know in the limit if it's just one person getting much richer then of course you're going to end up in a situation where 50 percent of people do better than their parents but it's not like we've hit the ceiling in terms of economic growth it's just that we've gotten closer to that situation of one person capturing all the growth and we show in the paper that if you were to distribute the modern growth rates more equally like we did in the past maybe through some of these types of solutions that i've been talking about here creating more equal opportunities you would reverse two-thirds of the decline in mobility showing you that most of this is not about hitting a ceiling in terms of education or potential progress hello my main question was on your propositions when it comes to in specific the importance of integrating people based on CSU economic levels not just on race and one of my questions because i think one of the main pushbacks especially behind closed doors when we're trying to push this forward from a policy perspective is is there a negative effect when you bring in people from lower income demographics into the communities that are from higher income because i think the higher income families i think in the back people's heads what we see on camera is very different than what we vote behind closed doors and so i was wondering in terms of bringing this to light and sharing this with people is there a negative correlation if you introduce people from lower socioeconomic groups into higher and then does that affect those children's abilities to still be high performers and further generation excellent question and of course a critical one in understanding whether any of this is going to be actionable going forward and so what can we say in the data setting aside what people's perceptions might be so a couple things come out and you can look at this for yourself and the opportunity out there so if you look at those maps that i've been showing you you know supporters i was to show you the same map not looking at kids who started out in low income families but kids who start out in high income families what would it look like spatially fact one is that there's much less dispersion across areas in outcomes for kids who grow up in high income families than low income families if you look at kids who grew up in high income families in Atlanta for example they do just as well as kids who grow up in salt lake city or the bay area and i think that's intuitive basically what we find is where you grow up matters much less if you're rich than if you're poor and it's kind of because you can insulate yourself from the local conditions you can send your kids to the best schools etc you have networks that cross many boundaries and so forth so that's one way of looking at it that the outcomes of the rich sort of are always good regardless of of the environment to some extent but then more directly you can ask if we look at these more integrated places we see low income kids are more likely to rise up do we see worse outcomes for the high income kids there and the answer there is a little bit more nuanced so if you just look at that in the raw data you see slightly worse outcomes for kids from high income families that turns out to be driven largely by the fact that the average incomes are different in that place why might that matter in the us with property tax local financing of schools if you live in a more integrated place obviously your school is going to have less funding and that might have a negative effect larger classes etc for kids from high income families once you control for that in the way that I showed you with that color dot plot hold fix the level of resources and look at changes in composition there's absolutely no effect of more cross-class interaction on the outcomes of kids from high income families while kids from low income families do better so it is not a zero sum game by any means and I think trying to convey that publicly now coming to the core of your question to actually change people's perceptions I think that would be a very valuable thing to figure out how to do thanks for a fascinating talk Raj I have two quick questions the first relates to what this young gentleman just talked about when you had that 50 50 split between how much of it is segregation and how much of it is friend bias yeah then at least it suggests that creating less segregation more cross area busing or policies like that will build off of that 50% and do something yeah it's possible however that the areas where there's less of that are areas with you know intrinsic attitudes so that if you would mix them the friend bias would become much more severe right so it's not clear that we would get 50% if we did that right and in fact you see that in the data so in the more diverse communities that are are more mixed income you see more friending bias and so there's kind of a catch 22 and that's what shows you you need to think of it's not just about integrating people you need to think about how you actually foster that interaction this is speculation moving beyond data in a way that usually I don't feel comfortable doing but one hypothesis is that people look for something in common like if you look at this data on religious groups or recreational groups if you find kind of the common denominator the shared faith or the shared sports team it overcomes other forms of stratification and so maybe it makes sense to think intentionally about that when one thinks about integrating the second question when you look at that you shape with the you know so you compare Berkeley to Harvard I'm wondering if a big part of that isn't part of the say private elite university's business model of legacy admits yeah because that is a tremendous engine for donations yeah and if I were the president of Harvard I would look and then say I'm probably not changing anything if I want to keep on getting the money I'm getting so Steve I'll get into this in more detail actually in tomorrow's lecture to decompose exactly what's driving that uptick because I think that then points to potential policies one might undertake to create more equity there you're absolutely right that legacies are part of it turns out it's not all of it it's about 40% of it but let me emphasize one key thing I think about legacies anticipating what I'll say tomorrow which is at the scale of these private institutions at this point the vast majority of legacies were admitted quite honestly do not have enough money to make a genuine difference for an institution that has a 55 billion dollar endowment and so yes there are some people who give 350 million dollars and that you know does obviously matter to an institution but that's a tiny sliver of the population and it relates fundamentally to the work in manual and tomopakety and others have done that the distribution is of wealth is so skewed you really need to admit 10 kids to get the vast majority of your donations and so my view is it's not obvious even from that very economic perspective that this makes sense we can talk more about that thanks again Raj thank you so much so glad to see you here at Berkeley again so my question has to do with the value added effect of a university or college if my memory is correct maybe it's not I think there's a paper by Alan Kruger that said that once you control for the individual and where the individual you know the course courses taken the majors of the individual that the college effect almost disappears and how do I square that with your value added yeah yeah well thanks rich so I feel like the questions you all are asking perfectly set up for the next lecture where I'm going to tackle that as well so our most recent paper actually is about the causal effects of colleges and why kids from high-income families are more likely to attend some of these colleges and just in a nutshell our view is that we've updated our views since Alan Kruger's paper for two reasons it turns out to what Alan Kruger was doing in this very famous paper is taking kids who got in to say their local state flagship school in a highly selective private school comparing the outcomes of a child who chose to go to the private school versus a child to go to the chose to go to the local state school and found that on average their incomes weren't that different turns out that if you use the tax records with now a tremendous amount of precision and you look at the fraction of children who reach the upper tail of the distribution there's an enormous difference in your odds of reaching the top of the distribution if you go to one of these highly selective private colleges versus less selective public schools so in that range it matters tremendously and then there's other work including by Zach Blamer and other research designs we're using where if you look at the state flagships versus say like the Cal States or the CUNYs the next year of schools even on average incomes you start to find really significant differences and so our sense is that you know what has actually become quite a popular takeaway a lot of people have read that paper it's been incorporated in the general view like maybe college doesn't matter so much I actually don't think that that's right we're able to replicate those findings reconcile them with more modern data and my sense is there is actually a real value out of here and I'll show that data more in tomorrow's lecture thanks hello my name is Leah I'm an undergrad I actually was assigned one of your lectures to watch on YouTube in like freshman year and econ one and I've been a fan ever since so I'm excited but so as many people will know and recognize when we use meta platforms I have a question regarding your economic connectedness variable we use meta platforms a lot of the time we get a little you know some little box that says people you may know and I think this contributes to what people discuss in literature and just in pop culture is the echo chamber effect and so I'm wondering given this very significant coefficient you got and your research overall like how does the echo chamber effect impact the significance and the meaning of your economic connectedness variable and that data yeah great question so you basically wonder you know how much of the friendships we're seeing on the Facebook data are sort of generated algorithmically because of the algorithm used to recommend friends or because of the nature of online interactions to generalize a bit and so on and so we're concerned about that as well and we try to tackle that in various ways so one example is we try to look at your closest friends so you can use various proxies like the messages people are exchanging and where they're located various groups they participate in things like that to get a very clear sense of who people's closest in person friends are and who are kind of broader contacts who may have been algorithmically recommended and it turns out if you subset to those five or ten closest friends you get results very similar to what I've been showing you so why is that that might not seem totally intuitive you know average person has like 350 friends on Facebook why does it turn out if I look at your 10 closest friends I get a similar picture if I look at the 350 basically it's an empirical result in the data that if I tell you the socioeconomic status of your first 10 friends it is very highly predictive of the 350 friends you have and so you end up getting a pretty sharp signal of probably the contacts who genuinely matter not the very broad network and so that's why our sense is despite those kinds of issues this is actually giving you a pretty reliable segment okay and a follow-up on the the economic connectedness we talked about that you talked about the channels through which this variable is impacted we have like the friending bias and exposure but I'm wondering about potentially other channels like time to use social media that people who work full-time jobs might not have and how that's going to also impact your data if there are other channels that you're currently discussing yeah totally you know and I think a very interesting question for students and others here interested in these issues is what is the effect of online networks themselves on these kinds of connections and how is that going to impact outcomes down the road we are basically using the Facebook data as a proxy to get an offline interaction more or less as I was just saying but you know one can think going forward in some sense there's a lot of potential to overcome exposure constraints you're going back to Steve's question if you're not bound by physical geography and you can connect with people anywhere in the world in principle you can have much more diverse connections in practice as you just said you end up getting a lot of echo chambers and so are there ways that we can connect people meaningfully overcoming these geographic boundaries I think that's a fascinating question with social media thank you very much hi I'm Sean I'm a physics undergrad here incidentally physics is the major here with the most imbalanced gender ratio in favor of men uh can you speak a little more about the gender disparities we see in the maps yeah absolutely you know maybe the way I will do that is just briefly show one slide here that I skipped which is looking at where kids become inventors in America is very relevant to fields like physics so uh what you see is on many dimensions by gender by race by income there are many more kids who become inventors from certain backgrounds relative to other backgrounds more men than women more kids from high income families and low income families white folks than underrepresented minorities etc and to answer your question in an empirical way we can try to understand you know why that's happening and what's driving gender disparities in particular and I'll show you again a geographic disagregation that turns out to be useful in figuring that out so here what we've done is linked patent data on the universe of patent holders in America to the tax records so we can look at the lives of who goes on to become an inventor in America and here we're looking at where inventors tend to grow up and you'll notice you know a lot of inventors grew up in the bay area a lot of inventors grew up in some other cities in the midwest what is this blip here in Texas that's Austin Texas right around UT Austin so you probably see a pattern it's around places of kind of where there is a lot of innovation and knowledge production among the adults living in that place so that's consistent with other ideas I talked about in this lecture that exposure might matter if you're growing up in a place where people are doing a certain thing like science and innovation you might yourself have that on your radar screen but tying this now to your question about gender you see some very sharp patterns with this that speak for example to the origin of gender disparities so it turns out that as a girl if you grow up in an area with a lot of men who are innovating who are patent holders in a particular field it has very little impact on your probability of becoming an inventor if you're a boy growing up in a place with a lot of male inventors you're much more likely to become an inventor but as I was just saying if you're a girl growing up in a place with a lot of male inventors it's actually totally irrelevant if you're growing up in a place with a lot of female inventors it has a great influence on your probability of going into science and becoming an inventor and it turns out that also happens in a field specific manner so you can classify patents into the technology class like what type of patent you have and this happens in a very technology class specific way so if you grew up in the Bay Area you might be much more likely to have a patent in computers if you grew up in Minneapolis which has a lot of medical device manufacturers you tend to have a lot of patents in medical devices even if you yourself are living in some other place in adulthood very consistent with this exposure idea and it's totally gender specific and so coming back to your question you know why might we have so few women in physics I think part of it might be this intergenerational transmission where we've had so few women in physics and that itself leads to so few women in physics going forward and this is the type of problem that I think we really need to try to address going forward so if you look at the fraction of female inventors over time what fraction of people who are getting patents are women by birth cohort you can see that is indeed going up but it's going up at a rate of a quarter of a percentage point per year which means if you extrapolate it's going to take another 118 years to reach gender parity in terms of innovation and I think it's this kind of process that somehow one needs to figure out how to break out of to have an impact as an epidemiologist and a data nerd your talk was a real treat thank you my question actually relates a little bit to looking at this so virtually your entire talk had us the primary outcome the longitudinal income objective but if you want to talk about the american dream or a humanistic dream income may not be the only relevant variable it's very convenient but I I'm interested what other variables outcome variables you'd looked at yeah like invention or or the bouton happiness index or yes great question you know despite being an economist happy to conceive that many things matter beyond income and you know we've we've tried to make some progress on that you saw that with the innovation work something like happiness and well-being is of course more difficult to measure on scale in the way that we're trying to do but here's another example of the type of outcome we're trying to study more systematically which is health life expectancy for example and here you know from some other work we've done here we're just plotting life expectancy in america at age 40 versus income and the point I want to make here is there's an enormously strong link between income and life expectancy right so the poorest men in america we estimate using population social security death records live about 15 years shorter lives than the richest men in america which is a shocking disparity if you put it in context so the cdc estimates that if we were to eliminate cancer as a cause of death life expectancy would go up by 3.2 years so think about 15 years relative to that as you as you know well and so you know my takeaway from this is while income is not in and of itself surely the measure we care about there are lots of other things like health and other measures of well-being that are very strongly correlated with income and so it's a useful place to start and we find with when we start to look more directly at these other outcomes there's some nice work done by Stephanie DeLuca Johns Hopkins for example showing that that same kind of dosage effect I was showing you for income when people move to better neighborhoods also emerges in the context of health and adulthood so we're finding similar patterns for this constellation of variables but I think more work can be done on those other dimensions as well yeah I think this is a last one. Hi my name is Mike um I'm trying to reconcile a couple of big picture things with the kind of the social engineering kind of focus at your yeah because it seems pretty obvious we move a kid into a a situation where there are a greater resources available for their self-improvement that they will improve that seems fairly obvious um I mean why I went to college we didn't have tuition I went to UC there was no tuition yeah now they're paying what 15,000 20,000 dollars a year so that seems to be a mechanism that would weed out some of the middle class people like me that had the advantage but the question I have is in the last 40 years there's been a huge shift in income and two years ago I think I read is the first time in U.S. history that the majority of income national income came from return on capital rather than return on labor in other words investments dividends capital gains interests which obviously in that same 40 year period there's been a significant decline in the taxation of those forms of tech those forms of income so so and at the same time there's been a disinvestment in the public space public education public health all those kind of public investments have declined dramatically so that people have access to fewer resources outside of their income so how do you how does this uh how do you connect those the broad trend yeah which is away from investment in in in kind of mechanisms that might even up the which is driven by macroeconomic economic policies taxation and accumulation of capital through investment how do you reconcile those with what your goals are is to even the playing field yep yep yeah great great question so I think some of those macroeconomic factors contribute to the trends that led to the very first draft that I showed right of the fading of the American dream in my view partly through the mechanisms that I'm digging into in more detail here so for instance take the example you gave the colleges that people attend we think have very significant causal effects on their future outcomes in the past it was maybe easier if you wanted to to attend an institution like UC Berkeley certainly a private college now that's become much more difficult to do partly for financial reasons but partly for other reasons as well so if you take like the u-shaped graph I was showing you know with the ivy league colleges here this is actually not largely driven by an issue of income and affordability because at this point many of these colleges have basically made it free to attend these institutions if you have an income below something like 80 thousand dollars or even a hundred thousand dollars because of financial aid yet those kids are not attending those colleges at the same rates and in this case it's because of who's getting in as opposed to just financial aid so my view is those macroeconomic factors that you're describing no doubt like the reduction in funding for public institutions have created some of the greater challenges that we're seeing they've led to more segregation they've led to a more unequal distribution of resources that limit opportunity but at the same time I think there are also other margins that are relevant in terms of creating more opportunity like who's admitted to these colleges to take one example I'll give other examples and tomorrow's lecture of specific policies where we're spending billions of dollars to say try to reduce segregation in America but I think much less effectively than we could if we understood better you know what mechanisms are driving why people locate in different places so I don't view it as one hypothesis versus the other I think these things are connected and there's a value in addressing it at both levels hi I really enjoyed your talk one thing I was really moved by was the graphic showing outcomes of black and white men from rich families and where they ended up in the in the percentile of income and it made me think of the three generation curse which is like the I don't know if you're familiar the phenomenon where nine and ten families in third generation squander the wealth in that family and I was wondering what role of any do you anticipate your work in economic mobility in ending the three generation curse especially in families of color yeah great yeah great question so yeah you're exactly right that you're mapping out the dynamics not to two generations but three generations if you kind of carry that process forward if you started out rich or even less likely to stay at the top when you look forward three generations now you used as I think as commonly used the term squander which kind of seems to implicitly I think people assign the responsibility to the families in the sense of you know you could have kept this wealth but somehow you ended up not keeping that wealth and my view is when you see those systematic patterns like I was showing for black Americans that it's not so much a choice or some sort of irresponsible spending or behavior they're more they're deeper structural factors that are leading to these vast differences and outcomes in a single generation further and subsequent generations and I think digging into what is causing that will help improve outcomes in a single generation in multiple generations and importantly in the long run in the steady state to use the term economists would use would lead to a convergence of incomes so I think that's absolutely right and that's by thinking about these various solutions and getting into some of the policy solutions as we'll do tomorrow is valuable thank you so please join us tomorrow for the second of the lectures which will deal more with policy and thank you all very much for coming and thank you professor Chetty for a great lecture