 Welcome everybody. I'm Susan Denarski. I'm a professor here at the Ford School and also at the School of Education and the Department of Economics. And I'm co-founder with Brian Jacob of the Education Policy Initiative. Thanks so much for joining us today. This is our first EPI talk of the season, and it's great to see this wonderful turnout. EPI is a program of coordinated activities to apply rigorous research methodology to the evaluation of education policy. We sponsor speakers like this. We work with some of the best practices to practitioners. We train graduate students. We've got postdocs. We have all sorts of fun. So thanks for being here for the first event of our season. Today we're pleased to present Sean Bearden, a professor of education and sociology at Stanford University, and director of the Stanford Interdisciplinary Training Program in Quantitative Education Policy Analysis. Sean's done important work in a variety of areas of education research, including the effects of education policy, uneducational and social inequality, causes, patterns, trends and consequences of social and educational inequality, and applied statistical methods for education research. Most recently he's written about the widening achievement gap between children from high and low income families and the growth in residential segregation between high and low income families over the past 40 years. So a really interesting set of research topics. Today Sean's going to discuss his research examining the relationship between income and income inequality with educational outcomes. He's going to talk for about an hour, and then we're going to have Q&A. So if you can hold questions until the end, that'd be terrific. So let me thank the Gessner Fund, the Charles H. & Susan Gessner Fund, the Institute for Education Sciences, and RACM for their support of this event. And anyone here who's in my stats class will recognize what I'm going to say next, which is if at all possible, if you can break yourself free, close your electronics so we can focus on the speaker. I know whenever I'm giving a talk and I see people on a computer, she's transcribing, so she gets to. But I assume people are checking Facebook or surfing porn or something. You don't want people to worry about that. Now I have that image in my head. They always tell you when you speak to people, imagine the audience naked and makes you less nervous, but now I'm not imagining that. Anyway, thank you, Sue, for the invitation and the introduction. I'm really happy to be here. I'm going to talk to you today about a body of work, a couple of studies that I've been doing with some colleagues over the last few years that are all around the issues of understanding the relationship between family income, income inequality at a kind of a national or a contextual sort of level, and how well children do in school, how they do an achievement test, how far they go in school, and things like that. All sort of to help think about a sort of big set of questions that are of interest, both to sort of people who study education, sociologists of education, economists of education, but also sort of just from a sort of broad social perspective because I think they illuminate or help us at least think about some of the issues of the current era. So one of the best well-known sort of stylized facts in all of the sociology of education is that children from higher income families do better in school. They perform better on academic tests. They are more likely to go to college. They're more likely to go to selective college. They're more likely to get an advanced degree. In any sort of educational measure, children from higher income families tend to do more. Now that's been true since the Coleman report. It was true long before the Coleman report, but certainly it was true when Coleman studied the issue. But the strength of that relationship, how well family income predicts educational success is not a fixed thing. So it might vary across time. It might vary across societies. It might vary sort of in relation to features of education policy or social policy or economic conditions. And so I want to sort of think a bit about that today. So the big questions that you might sort of keep in the back of your mind is what role does schooling play in socioeconomic inequalities and social reproduction, social mobility? Is schooling help us understand the extent to which people from low income backgrounds have high social mobility, can move upward in the sort of economic ladder? Under what conditions does schooling do that? And to what extent does broad social inequality play a role in this? And there's sort of, in some ways there's sort of question that people like to ask all the time about education. That is, is education sort of equality producing or inequality producing? Does it make things more or less unequal? Or does it not play a role at all? Is it just sort of economic conditions that matter? I'm not going to be able to answer all of those questions. Really, any of those questions say, but they're good questions still. I'm going to sort of give you some answers to some smaller questions that are going to help think about that, I think. So one is, how big are socioeconomic inequalities in educational outcome? How big are they now? How have they changed over the last several decades? And that's going to sort of help present some sort of stylized descriptive facts to sort of think about the relationship that might exist between income inequality and education. And then I'm going to look at some international evidence and some preliminary work that looks at across countries what features of countries or economic conditions in countries are related with more or less educational inequality. So how much does economic income inequality matter in a country in producing educational inequality, and why? So I'm going to talk about three different things. So we already did the introduction, so we're moving on here. I'm going to first tell you about income and educational outcomes in the U.S. based on a couple of different studies. And then I'm going to talk about this cross-national work that looks at income inequality and education outcomes in OECD countries, developed countries. And then I'm going to sort of end by sort of discussing some reasons why we might see the patterns we see. So first, let's look at what income inequality looks like in the U.S. So this is a now famous graph by Thomas Piketty and Emmanuel Saez. And this sort of charts the development of income inequality in the U.S. over the last 90-something years. So the red line tells you the percentage of all income earned in the U.S. that went to the top 10% of earners. Back in 1929, before the stock market crashed, and up through really 1940 through the start of World War I, economic inequality was really high in the U.S. The top 10% of earners earned more than 45% of all the income. That plummeted after World War II, a result of dramatic changes in the economy, the rapid economic growth of the 50s and 60s. And so through the 50s, 60s, and 70s economic inequality was low. It wasn't zero. The top 10% of earners earned about a third of all income in the U.S. And you can see that it took off dramatically in the last 30 years or so. And now the top earners earn nearly 50% of all income in the U.S. So economic inequality now is, always as high or higher than it's ever been in the last century. So we might wonder what are the consequences for that big upward trend there for educational outcomes. Here's another way to look at income inequality that's going to be useful later on. So this, instead of saying what share of income do the top earners have, this looks at the ratio of the income of a family at the 90th percentile of the income distribution to the income of a family at the 10th percentile of the income distribution. So back in 1970, that high earning 90th percentile family earned about five times more than the 10th percentile family. And now that's about 11 times difference. So the sort of gap in income between the bottom and top of the income distribution is more than twice as high as it used to be. In sort of today's dollars, the 90th percentile family now, this is the 90th percentile family with children earned somewhere around $170,000, $165,000, and the 10th percentile family earned something like $16,000 roughly. So that's that 11-to-1 gap. Now if we break this down a little bit, instead of looking at the ratio of the 90th to the 10th percentile family, this is the ratio in the blue line here of the 50th to the 10th percentile family and this is the ratio of the 90th to the 50th. So now we're sort of the blue lines comparing the family at the median to the family at the bottom and the red lines comparing the family at the top to the family at the middle. And you can see that all the growth, most of the growth in the 70s and 80s in this measure of income inequality was because the middle pulled away from the bottom fell out but the difference between being at the sort of middle class median income and being at the 10th percentile got a lot bigger over that time period and then has been pretty flat since then and the income ratio of the 90th to the 50th percentile has sort of slowly and steadily increased over this whole time period. Keep this in mind because there's going to be a quiz later because I'm going to come back and refer to this a little bit later but the important thing here is that although income inequality by lots of measures grew at the top, when we look at families with children, a lot of the growth in income inequality over the last 40 years happened between the middle and the bottom and that's important to think about later on when we get there. So hold that in your mind. Now let's look at the gap in achievement in standardized test scores between children at the high and the bottom end of the income distribution. So what I'm going to do to look at this is I took all data I could find for the last 50 years that included nationally representative samples of students, standardized test scores and information on family income. So there's 13 big nationally representative studies I found. They go back. The first one is project talent. It was 1960. They tested 400,000 high school students and the most recent one is ECLSB, this early childhood cohort, these children born in 2001 who were tested when they were in kindergarten in 2006, 2007 and so about 50 years of data, cohorts of children born as early as in the early 1940s and as recently as 2001. We're going to look at each of those cohorts of kids and look at the relationship between their family income and their test scores. So here's sort of how we do it. This is one sample. So we line up everyone by their income, their rank in the income distribution. So this is data from 1988. These were eighth graders in 1988. So in 1988 if your family made between $15,000 and $20,000, if you were between the 20th and 30th percentile or so of the income distribution and you had average test scores that were about here, these are sort of standardized test scores and so we plot all these scores and we look at this nice red line I drew through here and I'm going to take the average test score at the 10th percentile, so 10th percentile kid right here and I'm going to take the average test score at the 90th percentile and I'm going to take the difference in those average scores and that's going to be the 9010 income achievement gap. It's the average difference in test scores between a child whose family makes at the 90th percentile of income and the child whose family makes at the 10th percentile of income. So it's sort of analogous to those 9010 income gaps but now it's a test score and so for every one of these studies we're going to sort of do the same exercise and compute this gap. So here we would say the gap is three-quarters of a standard deviation, 0.37 minus negative 0.38 so about three-quarters of a standard deviation test score gap. Here's another example. This is from 2006 eighth graders. Same story. We draw the picture and you can see the gap is a lot bigger. Now it's almost 1.2 standard deviations. So I'm going to do that for every one of these studies. I won't show you the picture for every one of the studies but I'm going to give you a quick idea of how we compute this. And when I put all the studies together I get this lovely picture. So on the bottom here we have the birth year of the cohort. So the earliest study has children born in the 1940s, most recent in 2001 and then for each of those cohorts I get an estimate of the gap. That's what the dot is and then these standard error bars are confidence intervals around it. And then we fit this curve through it and the story here is the achievement gap is relatively flat for cohorts born in the 50s, 60s and into the 70s and then starts to widen in the mid-70s through 2000. These data are somewhat unreliable so I don't put a lot of stock in this early rise because it's a very old test and we don't have too much information about it but the rest of it is a pretty credible test. So this is the gap in reading. We do it for math and the picture is the same but I didn't want to show you the same picture twice basically. But here's that same picture but I took away all that extraneous dots. So this is the sort of story about the income achievement gap. It's gone from about 0.8 or 0.9 standard deviations for cohorts born in the 1970s to about one and a quarter standard deviations for cohorts born more recently. So that's an increase of roughly 40%. So that's quite sizable. How big is that in terms and comparison to other things? Here's the black-white achievement gap over the same time period. So for kids born in the 50s and 60s the black-white achievement gap was much bigger than the achievement gap between high and low income children but the black-white gap went down a lot. We made enormous progress closing the black-white gap over the several decades of kids born in the 50s, 60s and into the 70s. That progress has largely stalled but at the same time that that progress stalled the income gap has gotten much bigger. So now this income gap is more than 50% larger than the black-white gap which is sort of the opposite of the story here. In some ways you could look at that picture and read the history of late 20th century America as we went from a society in the 1950s and 60s characterized by historically low levels of income inequality and very, very high levels of racial inequality in every sort of domain of life whether it was education or health or mortality or what income. So we went from that era where sort of racial inequality dominates economic inequality in some ways to an era now where racial inequality is certainly not zero, these gaps are far from zero and racial gaps are not zero in many other domains of life but they're much smaller than they have been and they're at historic lows in many domains. But we're also in an era where economic inequality is at historic highs. So we've gone from this sort of era when racial inequality was in some ways is more extreme than economic inequality to an era where economic inequality is more extreme in some domains than racial inequality. Don't get me wrong, racial inequality hasn't gone away. Some people think I'm saying that, I'm certainly not saying that and when this gets to zero then I'll talk to you about racial inequality going away but you can see we're still a long way from that. But economic inequality has become increasingly important and if you look at this picture and you think about those earlier pictures you might think well, economic inequality that is income inequality went up at the same time this sort of achievement inequality went up. Are they related? Well, they have the same shape so that's sort of a start but part of what I want to talk about for much of the rest of this is how much rising income inequality explains this rising achievement gap. So let's go back to this. Now we're going to divide this 90-10 achievement gap into two halves. One which is the 90-50 achievement gap and one which is the 50-10. That is the red part of this is the difference in achievement between the 90th percentile income child and the median income child and the blue is the difference between the median income child and the 10th percentile income child. And the story here is that the gap between the child from the median income family really changed over this time if anything it got a little bit smaller. But the gap between the 90th percentile family and the 50th percentile family child is where all the increase grew. This gap is about almost twice what it was back here for kids born in the 70s or 60s. So most of the growth in fact all of the growth is in this income achievement gap is really driven by children of high income families pulling away from middle-class children, not from middle-class children pulling away from low-income children or low-income children falling further and further behind. If you look at test scores over the last 40 years on average test scores in math for example have gone up very dramatically in the U.S. over the last 40 years. You might not know it if you listen to all the rhetoric about how terrible our schools are but the average nine year old now has test scores equal to the average 11 year old in 1978 that is sort of two years of progress over the last 30, 35 years. And if you look at the test scores of low-income children they've also gone up but not as much and if you look at test scores of high-income children they've gone up dramatically. So this widening gap isn't driven by the sort of bottom falling out of the test score distribution or low-income students sort of falling off the map. It's really because everyone's been going up but the scores of high-income children have been going up much, much faster than the scores of middle-class children. All right so the next thing I want to talk about is well what happens during school? Does the gap get bigger as children progress through school? So here's data from two studies one is the early childhood longitudinal study of kindergarten and cohort so these are kindergartners they were in kindergarten in 1998 and they were tested seven times over the next nine years Fall, spring of kindergarten Fall and spring of first grade third grade, fifth grade and eighth grade and this is the gap between the 90th percentile income child in kindergarten and the 10th percentile child and this is the gap between those children in eighth grade. You can see that the gap got a little bit bigger but not appreciably. It's huge when kids enter kindergarten over here come from another study of a birth cohort who were tested in their kind of school readiness at age four and then tested when they were in kindergarten and you can see even at age four the gap is quite big on these school readiness measures that measure kind of pre-literacy skills. So it doesn't look like the gap is sort of growing during the K-12 school in years the gap is really big at the time kids enter kindergarten or two before. So this suggests that maybe schools aren't the reason why this gap is growing because if schools were the reason if kids come to school and then high-income kids get access to really good schools and good teachers and learn a lot and lower-income kids get access to less good schools and don't learn as much we would expect to see this gap widening during the K-12 years but it doesn't. If anything, there's a little bit of evidence that it narrows in schooling because from fall to spring the gap gets smaller while kids are in kindergarten and then it grows a bit over the summer and then from fall to spring first grade it narrows again. Now we don't have any other fall to spring measures here that we can do this with to see if it continues. There's some evidence from some other more local studies, a Baltimore study that suggests that pattern sort of continues but they're sort of small samples I don't think you should read too much into this you shouldn't walk away saying schools narrow the achievement gap and then summer widens it and that's the story but there's nothing here to make you say well schools are really the culprit schools are widening it now how much of the gap here is between the high and the middle and how much is between the middle and the low here's the same data but now I broke it down again into this 90-50 and 50-10 gap and you can see it doesn't change a whole lot about a little more than half of the gap is between the bottom and the middle and the rest is between the middle and the top of the income distribution and that's pretty steady over time at least after kids enter kindergarten so there's no there's no evidence that the sort of gap is being disproportionately widened or narrowed at one part or the other during the schooling years here's data from a bunch of other longitudinal studies just to show you that it's not unique to that study so the blue solid line is the one I showed you before the reading from ECLSK but there's a bunch of these longitudinal studies with national samples that are followed over time and generally all of them show the gaps kind of relatively flat over the period of time you know the age period that students are observed this is the only one that's different and we don't really believe this is a kind of a really lousy test this prospects test so this is probably not to be trusted but even so there's just this sort of one data point that suggests gaps narrowed during school for the most part all these longitudinal studies suggest no substantial widening during the K-12 years alright so what you should be thinking about now is okay so over the last 30 years or so the gap in achievement between high and low income kids is widened now let's look to see what's happened in terms of access to college enrollment in college particularly we did a small little study recently where we looked at enrollment in the most selective colleges and that was related to family income so oh first I have this slide from Martha Bailey and Sue Dynarski right here so Martha and Sue looked at college enrollment and this is the fraction of students who complete a four year degree I believe right and so there's two cohorts they looked at one cohort born in the mid sixties and one cohort born in the early sixties for this cohort only about 5% of low and it's quartile students completed college that went up modestly to 9% by the later cohort but for high income kids the fraction completing college went up much more in sort of percentage terms and so this appears from this evidence to be a sort of widening disparity between high and low income kids in who's completing a four year college then we looked at data from a couple studies this is from the high school class of 2004 the study known as L's and we looked at where students enrolled in what type of college students enrolled in after after high school based on these barons ranking so we assigned people to one of nine categories they either weren't in high school I mean they either hadn't graduated from high schools this is two years after they should have graduated they have a high school degree but they're not in college and haven't enrolled yet in college they enrolled in something less than a four year college some kind of community college typically or they were enrolled in a four year college of one of six selectivity rankings so the ones are the most selective institutions the Michigan's for example and then down here the sixes are the least selective four year institutions typically broad access four year colleges that don't have any admissions requirements so the blue is the upper income quintile students they're not strictly quintiles but families with incomes more than $75,000 in 2001 you can see there's this very steep gradient where in the most selective colleges you've got lots and lots of high income kids and very few low income kids you do see this interesting pattern and you see this in other data too that the most selective colleges actually have more low income students in them than the next tier colleges and that's most likely partly because they compete for those students colleges want diversity and the most selective colleges can attract the high achieving low income students there just aren't that many of them to go around so we looked at this a different way depending on your family income percentile what's your probability of attending a highly selective college and by highly selective college we mean one of those top two categories so about about four percent or six percent of students are going to these most selective schools so if you graduated in the class of 1982 your probability of going to one of those most selective schools obviously increases as your income goes up but increases very sharply if you're at the end so the 90th percentile family is a much higher chance of having a child go to one of those colleges than even the 70th percentile family so most of the action is up there we look at this in 1992 and the probabilities increased and increased substantially particularly at the high end now I think you should be a little bit worried about over interpreting this increase because one of the things about the 1992 cohort is they were born in the around 1974 which was kind of the low point of the baby bust that is there's the baby boom and you've got big cohorts born in the 50s and through the mid 60s then the cohorts get smaller and smaller and smaller so by the mid 70s you've got historically small cohorts and so your odds of finding a seat in one of these colleges are greater just because there's fewer of you and the same number of seats or roughly the same number of seats so part of the reason why these probabilities go up and go up everywhere is because the cohort is smaller so you want to keep that in mind and then this is it in 2004 where the cohort is more comparable to 82 and so I think the 2004 to 82 comparison is probably more valid than you might want to ignore the 92 but what it suggests is that the relationship between income and your probability of going to one of these selective colleges appears to have risen particularly at the high end and so access to those selective colleges seems to depend much more on family income than it used to. Now one of the obvious mechanisms which it depends on that is your achievement since achievement is a big part of what gets you into these schools and is becoming a more important part according to your colleague Michael Busteto there as the income achievement gap goes up so we would expect to see a greater increase in the role of the relationship between income and elite college access as we do right so I think that's part of the same story. I don't think it's because income itself sort of independently buys you more access it probably operates largely through the stronger relationship between income and achievement so that's my kind of quick tour through some descriptive stuff about the relationship between family income and educational success in the U.S. over the last 30 or 40 years. Now I want to tell you a little bit about this international comparison. So in the U.S. we just have this one time series and in it you see income inequality going up and you see this income achievement gradient going up and you think maybe they're related to each other but maybe they're not so let's look at other countries and see if the income inequality in a country is related to the size of this income achievement gap. So what I did, so this is work that I've been doing with Katen Chemaluski who's over here, who's a postdoc at Michigan State right now and was a student at Stanford until this past year and so Katen and I have been working on this study and so what we did was we looked at wealthy OECD countries so there's the U.S. and 18 other OECD countries that we could have we got data from the PISA or Pearl study that had family income, not all the countries in PISA gave the family survey that included family income measures so we don't have the full sample of all the PISA countries, we just have the ones that gave this survey and then the U.S. didn't give the survey but we have data from the same cohort of students who we have information from the ECLSK so it's a comparable cohort, comparable age and so in each country we're going to compute these 90-10 income achievement gaps and we're going to see how they relate to some basic features of the countries so for each country we construct four different indices to kind of summarize what we think are some features of the countries that might be related to the size of the income achievement gap so first is this poverty inequality index so we look at the level of income inequality in the country, the child poverty rate how much income segregation there is between schools in the country, low birth rate rate and teen childbearing rate, all of these factors are highly correlated and so we load them all together, the reason we have to make these indices is because we only have 19 countries and we don't have a lot of degrees of freedom to work with so we're going to kind of lump stuff together in potentially displeasing ways but it's sort of the best we can do with the cross-national stuff then we have this social welfare index which says how much does the country spend on things that might particularly help low income families like public health spending, public spending on family benefits pre-primary school spending, things like that so that's a kind of what we call the social welfare index then we have this parental support index which is a measure of the kind of social policies in the country that might support families, particularly families with young children so maternal and paternal leave policies and then we have this educational differentiation index the argument being in countries where you have lots of tracking and lots of sorting of students among schools or classes by ability you might expect there to be wider income achievement gaps if income is correlated with it and so high income children are more likely tracked into higher tracks or in a case like Germany into sort of separate schools where there are more academic tracks versus vocational tracks and so we might see that sort of differentiation of instruction might lead to wider income achievement gaps in contrast if you have a country where sort of everyone gets the same curriculum the same instruction in the same sort of context you might think there would be less of an income achievement gap because there's sort of less possibility for sort of differential school environments to create disparities so we're going to look at these four indices and how they relate to the size of the income achievement disparity in each country first here's the size of the income achievement disparities across the country so there's three different data sources the Perl's study which is fourth graders in each of these countries tested in 2001 and the two PISA 06 and 09 studies which have 15 year olds tested these are all reading tests so the U.S. here in red you can see it tends to rank on the high end of the size of the income achievement gap and there's a lot of variation Iceland Norway, Sweden, Netherlands some of these Scandinavian countries have relatively small income achievement gaps that is there's not as much difference in the achievement of high and low income kids in those countries U.S., Greece, Germany, Portugal have other ones so another case of U.S. exceptionalism up there so here's the relationship between the size of the income achievement gap in each country and this poverty and inequality index which is a sort of combination of income and equality poverty rates things like that and we drew two lines the red one which uses all the data and the blue one which makes the U.S. out U.S. is a big outlier here so we sort of didn't want U.S. to be driving it but you can see we get basically the same relationship between income and equality and the achievement gap whether U.S. is in the data or not lots of the Scandinavian northern European countries are down here with both low achievement gaps and low income inequality we also took all of them out thinking maybe Scandinavia is just different and you still get the same sort of a relationship there so it doesn't seem to be driven by any particular country here's the relationship between our educational differentiation index and the achievement gap so remember the differentiation sort of says does kind of everyone get the same kind of schooling or are there tracks and separate kinds of schools or separate tracks within schools and here we get an upward slope so it's somewhat different depending whether we keep the U.S. and the U.S. is sort of unusual in that it's not very differentiated by our measure which mostly looks at how early tracking happens and how much between school tracking and things there are but it's still upward and then we did some regression models so you should think of these as very descriptive models we're not trying to make big causal claims here about whether if you changed inequality or changed educational differentiation you'd narrow these gaps or something but from a sort of descriptive point of view when we put in all four of the indices together the educational differentiation and the poverty inequality index are significant so we get rid of these two and then we break the poverty inequality one into sort of its component parts to see maybe if is it driven more by income inequality or child poverty rates or school income segregation and any one of those separately has the same size of effect together they're also collinear that they kind of just fight with each other and none of them are significant so we sort of think this is the best model because it's not really any different of this but basically the story is both income inequality and educational differentiation seem to be related to how big the income inequality is now the income inequality one you might not think is surprising but I think the fact that this big features of the educational system are related to the size of the gap might suggest that the educational system does play some role in the size of the income achievement gap that is countries that have more differentiated systems so the extreme here Netherlands is kind of an outlier but so think of Germany in Germany students are tracked into different kinds of schools at a young age at age what 14 or younger right 10 so and they might go to the the gymnasium there's a more academic track or the help shula you know the more vocational track and so they that's very early tracking and very you know you know major tracking it's not sort of you're in the same school but you take a math class it's more advanced and a you know French class it's less advanced right it's not it's a very serious sort of structural tracking so places like Germany that have that kind of big differentiation tend to have bigger gaps even that of income inequality so again I don't think you should think of this as causal but it's suggestive I think that maybe the educational system will play some role in the relationship between income and achievement there's a lot of caveats to that you know it's only in 19 countries and we only have four variables and you know we didn't measure everything terribly well and so on but you know it is what it is it's you know we're trying to see if we can improve a few things but you know it's hard to do in this cross-national stuff alright the last thing I want to sort of talk about is why is this happening why is income increasingly related to children's educational success and so you know we want to think about whether or not schooling produces this and I think there's two contradictory kinds of evidence I gave you here they're just both sort of suggestive one is in the US the gap doesn't widen as kids progress through school and so that suggests schooling is not a big part of it but in the cross-national stuff the size of the gap is related to these big features of the educational system that suggests maybe schooling is part of it so I don't think there's sort of a clear obvious answer here in as to whether schooling plays a big role or a minor role but so why is it so one answer is it's just income inequality in the US income inequality has risen so if money matters for how will your child does in school and if the difference in the amount of money between the high and low income families got bigger then high income families have a lot more money money matters therefore the difference in achievement is going to be better or it could be because income has gotten more strongly associated with achievement than it used to be so that a sort of a dollar income matters more or is more strongly associated with achievement than it used to be if that's the case why so let me do a little bit of math for you so if we assume that your achievement is related to your income and the coefficient beta tells us how much the sort of doubling of your income relates to an increase in your achievement in a descriptive sense then we can sort of take the average achievement of the 90th percentile minus the average achievement at the 10th percentile and we can write it out like this and we know the rules of logarithms mean we can divide this and this is the 90-10 income ratio this is the thing I showed you before that went up in the US and so the gap between the 90th and 10th percentile family depends both on what beta is and it depends on the income ratio and so the gap could get bigger because the income ratio got bigger and we know that happened but it could also get bigger because beta changed and so what we want to think about is it all because the income ratio got bigger or did beta change does income sort of matter more or in some ways than it used to and so think of the first thing as a mechanical thing income matters a certain amount rich families have more of it than they used to relative to poor families so their kids do that much better that's sort of a mechanical relationship where beta doesn't change and the other is this sort of contextual story where beta change that is as income and equality grows maybe for some reason the association between income and achievement gets bigger so we want to think about why beta might change so here's the picture I showed you before remember this is the 50-10 ratio and it's going up a lot in the early period and then flat and this is the 90-50 ratio if beta stays the same and if we look at our 90-50 income gap it ought to kind of go slowly up and our 50-10 income achievement gap ought to kind of go up a lot and then flatten out right if it's all driven by the changes in the income ratio so what I did was I estimated beta from each of those studies and then we're going to just look and see if beta stayed the same but I'm going to show you two betas one is the beta for the 90-50 part of the data so I just take the top half of the income data and I estimate beta so how much does an extra amount of income matter among high-income families and then I take the bottom half of the data and estimate to the bottom how much does extra income matter among high-income families so here's the this is the coefficient beta it's how much your difference in achievement with a doubling of income over time and this is for the above median income families so this is the high income beta so it's gotten a lot bigger over time even if we ignore this it's gotten bigger over time when we fit this curve and when we look at the low income beta it actually looks like it's gone down over time so what this says is that the reason why the income the income achievement gap has gotten bigger at the high end isn't just because the income ratio has gotten bigger at the high end but because income sort of matters more at the high end that is families with twice as much income as other families in the top half of the distribution are seeing a lot bigger difference in their children's achievement than they used to so why would that beta be bigger there and the sort of opposite story is at the bottom that income matters if anything sort of less than it did at the bottom I'm going to skip that so what else has changed that might help explain this that why beta would change well this is family expenditures on children over the last 40 years spending on children has increased substantially over this period low income families spending hasn't increased much so high income families appear to be not only do they have more money but they appear to be spending more of it on their children than they used to so that might be part of it might be different patterns of investment I'm going to skip that for a second another possibility is that parents think about what they're supposed to do as parents differently than they used to and it differs with social class so there's this argument that Annette LaRoe makes that middle class families think of parenting as a process of what she calls concerted cultivation so your children are orchids you cultivate them they're very delicate but they'll bloom beautifully but it takes a lot of work and attention you hover over them and water them from a teardrop maybe a teardrop anyway she says that's sort of the modern middle class version of parenting and the working class version of parenting is what she calls the accomplishment of natural growth you think of your parents you think of your children like you think of trees they grow but you don't spend a lot of time watering them and nurturing them it rains the sun comes out so that's what children do they grow they develop you don't sort of have to make it happen through your hovering kind of attention now that may be a kind of relatively recent pattern and there's some argument Julia Ridley and Mimi Shab that the parents have gotten sort of increasingly focused on cognitive development over the last half of the century partly because schools have told them that's what they should focus on schools have said test scores are important and so parents have sort of focused on that schools have said going to college is important and so parents have started to worry about it more and more but that's differentially affected the sort of middle class so as sort of evidence of this there's this nice study by Julia Ridley it's old so it would be nice to sort of see what the new data would look like but she looked at magazine articles aimed at mothers over the course of the 20th century her graduate students did this let's be honest thousands of articles over the 20th century and they looked at articles like these are magazines like Good Housekeeping Ladies Home Journal whatever the equivalents of those would be at sort of different historical eras they looked at articles aimed at giving advice to mothers about what they should do with their young children so a lot of these would be written by doctors or experts and what would they say and so she categorizes the advice into five categories so some of them they need nutrition medical attention, fresh air apparently there's no more fresh air you can see it's big in the 20s intellectual stimulation and social emotional development but you can see the early part of the 20th century the advice is all about nutrition and medical care make sure they're healthy and they stay alive infant mortality was not trivial in the beginning of the 20th century so that's not unreasonable advice to give to new parents but by the 50s and 60s cognitive and social emotional development start to dominate if this sort of is a leading indicator of what parents do and particularly middle class parents who might be reading these kinds of magazines then it's suggested there's been this sort of shift over time towards an increasing view of what you're supposed to do as a parent to be about nurturing your child's cognitive and emotional development and if middle class families are doing more of that over time than working class families to sort of change in the upper income or middle class families then this could lead to more sort of investment in early childhood more attention to how well kids do in school and all the sort of things that go into it so it might not be so much that money is changing but that parenting behaviors are changing in ways that are sort of correlated with income that might be driving this there's another argument that says money it's things correlated with money but this is more what I think of as the sort of Sarah McClanahan argument that is that family income family structure has gotten really polarized over the last 40 years and so family income is much more correlated with other stuff that might matter for kids development than it used to be so increasingly we live in a society where there's really just two kinds of families there's the family with two earners both well educated either one's employed or both are employed if high income the mother had her first child when she was relatively old meaning mid 20's or later so that family has a lot of income but it also has a lot of human capital it has two educated parents and what not or you have families of single moms who are less educated unemployed or underemployed and had their first child when they were quite young so that's a low income family but it's also a family with relatively low levels of education and human capital and so it might be that this polarization of family structure that Sarah McClanahan writes about means that income gets more correlated with other features of the family that matter so there's also this increase in what Schwartz and Merrick call assorted of mating you might know it as marital homophily whatever you call it most of you are probably doing it it means you hang out with people who are like you that is you marry people and you have children with people who have similar levels of education as you do and so as that's increased it means that a high income dad increasingly has a well educated high income mom in the household too and so income gets increasingly correlated with these other sort of human capital things in the family so again it's a story that says it might not be income but this increasing correlation of stuff with income that matters for young children we did these models where this is using all of our data again and we looked at the we did regression models where we just predict your test scores on the basis of family income and parents education really simple descriptive models don't read too much into these things but it's the income that gets much bigger over time not the education coefficient and so this suggests to us that it's not that sort of education is starting to matter more but it's that income is mattering more even net of sort of its increase in correlation with parents income so that suggests maybe the McClanahan story isn't the dominant story so another possible story is residential segregation so not only the rich families have more money than they used to but they live near richer people than they used to so Kendra Bischoff and I did these studies recently and there's others by Tara Watson and Paul Drogowski that look at similar things showing that income segregation rose a lot over the last 40 years so this is stuff Kendra and I did so we classify neighborhoods into sort of six types affluent near affluent high middle low middle low income and poor and we look at how many families live in these different neighborhoods over time and you can see there's an increasingly move of families either living in very poor neighborhoods or very affluent neighborhoods and fewer or fewer families live in these kind of middle income mixed in kinds of neighborhoods there's lots of other ways to look at this but all the data tells the same story income segregation has gone up so if you think neighborhoods matter or schools matter for kids well high income families increasingly live in places where they have access to high income neighbors and collective investment in sorts of things like high quality preschools and parks and schools and all the sort of things that go with an affluent community that might play a role now the neighborhood effects research has not been overly convincing on this point in terms of sort of saying that neighborhoods really matter so I'm not saying this is the story but it's certainly income has not just gotten more correlated with other stuff in the family but it's gotten more correlated with other stuff in the environment so high income kids live in much more affluent environments than they used to and low income kids live in more low income environments than they used to and they also therefore sort of go to schools that are more segregated which then could add into this so my job here is really to tell you a bunch of possible explanations for this but never to settle on any one of them you can all go do some research and tell us what the answer is it seems like inequality alone doesn't explain the widening achievement gaps because the coefficients have gotten bigger and so why is it so here's my sort of provisional story I'm not saying this is right this is how I'm thinking about it sort of today and for the last few weeks so it's not it's not totally off the cuff so a couple things have been happening so for young workers the returns to a college degree have doubled or gone up even more depending who you believe on this at least the observed returns there's a contentious literature about this but people have the story in their head that college has gotten more and more important in terms of success in the labor market and upward social mobility and so as people start to sort of think about education as being increasingly important parental behavior and parental investment in their children changes it changes how parents think about their children it changes how what parents do with their children how they spend their money and it also changes how they think about the role of schools I think if you went back 40 years you wouldn't find nearly as many people talking about whether schools are good at producing test scores which is largely the conversation we have today the in fact it's kind of striking how much nothing else about sort of what schools are supposed to do as part of the national kind of policy conversation about schooling we really talk about schools now as places that produce test scores in children and if that's the message you're getting as a parent that's kind of convenient because test scores are at least a thing you could observe and see if your kids are doing better on if they're good at producing democratic citizens it's not very easy to tell whether or not your school is doing that or whether your children is doing that or something so as we sort of move towards this idea that schooling is more important now we move to this idea that schools are really about test scores then the natural and reasonable parental response is to try to do what you can to make sure your kid does well in school which means getting good test scores and so you invest your time and your energy and the money that you have in trying to make sure your kid does well on test scores well as income and equality grows there's increased I mean as this grows there's sort of increased competition for sort of educational success if you think of educational as a positional good that is it doesn't really matter how much you know it matters where you rank you kind of want to be up at the top of that distribution wherever the distribution is then you sort of compete for position in that and then as income and equality grows some families have a lot more money essentially to buy the things that help their children compete well in that and so it's not that income and equality on its own drives this it's that in an era of growing competition for educational success widening economic inequality facilitates the success of the people at the very top of that and makes the sort of the competition sort of that much more brutal and harder to succeed at if you're sort of not at the high end of the income distribution so I think the big worry here is that as the link between income the link between your family's income and your child's achievement grows and the link between the increasing returns to cognitive skills and earnings you get this feedback cycle where the children of the rich do well in school and those who do well in school get rich and so the sort of idea of the American dream of sort of social mobility that if you just work hard and do well you sort of rise up gets harder and harder to achieve and studies of sort of social and economic mobility in the US suggest that that's true that we're getting we're sort of seeing less and less upward social and economic mobility and so I think this is all consistent with that story of this kind of increasingly sort of rigid stratification in the relationship between income and achievement and I don't think it operates primarily through schools but schools it could be a mechanism for trying to sort of buffer against that if designed well, if the system were designed well but but probably hoping that schools alone can sort of solve this problem is hoping for too much I mean when income inequality has gone up as fast as it has it's unlikely to think that a single social institution like schools can undo that and we probably need more redistributive if I'm allowed to say that word economic policies and what not so I think I'm going to I think I'll stop there so we have time for questions I think right are we good on time either one I'll call this guy over here in green since we're taking we're going to have you asking the mic hello thanks for coming out today my question is regards to self learning and kind of like self schooling curious as to your opinion of what the role of that of inducing a passion to learn within students and empowering students to learn in their own will kind of be able to reverse the feedback cycle that you alluded to and how that plays into the education reform you think it's effective so I think all children probably have a natural desire to learn but I suspect most children don't have a natural desire to learn in school or or to do homework or you know that schooling is not the institutional arrangement that most children would organize for themselves if left to their own or most adults for that matter so I don't know why we do it but so the trick I think is in I mean all schooling no matter who it's for you want children to sort of not lose that desire to keep learning and give them the skills that they can kind of keep learning on their own the question is whether or not the schools that serve lower income children are as good at doing that as schools that serve higher income children or whether or not the economic conditions create the incentives that might make people want to do that if it looks to you like no matter how hard you try you're never going to get ahead because nobody around you gets ahead and schooling doesn't seem to have helped anyone around you then it gets sort of hard to feel like it's really worthwhile to invest a great amount of energy in the work of schooling even if you like to learn so so I think in some ways we'd like all schools probably to do more of that I'm not sure it alone is the solution to the problem I think you need the mic this may touch on the previous question it's not clear to me where you put the role of the school at the end of the neighborhood and let me do it in a very specific example here in Ann Arbor we have paired a couple of schools to increase the diversity within each we could save money on transportation by doing away with that pairing thus have more money for instructional services and less diversity in each of those schools what should we be aware of or watching for as we make that choice and then most strongly of all do you just know what the right choice is based on your work no I mean I think in a society where there's lots of residential segregation which is the society we live in you face this if we had a society that wasn't so residentially segregated everyone could go to school in their neighborhood with a diverse mix of students and they could walk down a safe street to get there and come home and play in the sand lot as they used to call it across the way I mean but given the amounts of residential segregation you end up having to choose as a policy sort of perspective kind of aiming for diversity but spending money on transportation or something or you know aiming for segregation not aiming for it but you know accepting segregation and spending money on other things so there's a choice in the long run we ought to think much harder about housing policy but in the short run that's not going to solve the problem so I think it also depends on the extent to which you think the goal of schooling is to sort of as a it depends whether you think if you think of sort of schooling as a consumer good that parents are buying then parents sort of want they want their kid to go to school in their neighborhood with the peers that they paid to live next to and things like that if you think that the role of schooling is partly to create sort of citizens who are able to participate in a diverse democracy then you sort of think about there's a role to sort of make sure that children experience diverse environments and learn to function well in them even if that's not the first choice of every parent but that's a harder sell in today's environment where we think of schooling as a kind of consumer good that parents sort of are buying for their kid in a sort of individualistic way there is I think some interesting research on the benefits of diversity some of it done by Scott Page here at University of Michigan there's an interesting study I was talking to someone about today by Heather Schwartz that children in low income families in Maryland were randomly assigned to live in different low income or mixed income housing projects depending on sort of availability and whatnot and and the families that were assigned to housing units where the child would go to a more diverse or middle class school the children in those schools so much better long term outcomes than children who were assigned to go to schools that were more homogeneously low income and so it's a nice study because it has this sort of random assignment experiment nature and suggests that going to a more socioeconomically diverse school was was beneficial even in terms of test scores for the low income children now there aren't enough of those studies that we can say convincingly that that's a system thing and there are other studies would suggest that peer context and stuff doesn't matter so I don't think it's a sort of settled question from a kind of empirical scientific point of view whether or not diversity has benefits in terms of academic achievement but I think we also ought to bring into the conversation the issue of whether we think diversity has benefits for other things we value in a democracy other than academic achievement and in some ways I think framing the conversation that way makes it easier to have to think sort of thoughtfully about the trade offs you raise rather than to sort of end up sort of with everyone fighting for you know the thing that's best for their kid without thinking about the sort of collective good who's got microphone you and then some of your people in the back next yeah Finland like in the past 20 years or so really overhauled their schools and essentially focused on equality in public schools and along with that they made private schools pretty much non-existent do you think that something like that could work in the United States so there's almost as many explanations for why Finland scores as it does as there are Finns so I don't I mean there's a lot of ways Finland's different in the US and so I don't know what it is that makes it different there's certainly things like you point out also things about how they train the teachers how much they pay teachers there's a bunch of big structural features about Finland is educational system not to mention Finland as a society so I don't know why but would getting rid of private schools work in the US I don't know I mean I don't know I don't think it's doable in the first place I mean we have this sort of long tradition of private schooling and sort of two traditions of it one a kind of religious reason that you have all sorts of fights about if you try to not allow there to be private religious schools and then a kind of elite opt out private system that is you know I'm willing to pay $40,000 a year for my child to go to what I think is a better school than the local public school and and I think you have a hard time disallowing that somehow so I don't I don't think it's feasible to get rid of private schooling in the US and I'm it's not clear to me it would solve the problem because I think you just find ways to re-stratify within the public sector like one thing in higher education as big public institutions got more and more got sort of bigger and bigger and in order to sort of compete they just end up stratifying inside themselves so almost every big state institution has an honors college inside of it so that they can essentially have a kind of small liberal arts private school experience for a select group of students inside the big public institution American institutions are enormously nimble in finding ways to sort of stratify themselves under any form you give them so I don't it's not clear to me that getting rid of private schools even if one could do it within the long run change things there were some back here I just had a question on the institution of the school itself so I've heard for example 100 years ago if someone would be transported through time to now the only thing they'd really recognize in society is the school building we still teach the same way we have one person who teaches to a classroom and so my question is looking at how the school could play a role and I guess not as much how much of a role do they play but instead of looking at the school itself and the way it's the way we are taught now do you have any specific suggestions or alternatives that you yourself think could help specifically lower income students for example with the role of new technology identifying that usually there's been evidence that shows that perhaps boys learn differently than girls things that we could do inside the school that instead of just change about the income but the actual way we are taught it's a good question other than kindergarten schools really don't look any different probably than they did 100 years ago so I think all the new kind of excitement around technology and the possibilities of sort of online learning and flipped classrooms and whatever the sort of thing of the week is I mean are great right there's lots of potential there both in higher education and in K-12 but I think if what we need to do is sort of think carefully about how to organize those so that they're sort of equity producing I think because to the extent that sort of some of the actors and players in the market are interested in profit they're going to have a very different agenda than a kind of social equity agenda potentially and to the extent that there's just kind of a unregulated let's just sort of see what happens the people with the most say in society and the people with the most sort of voice and resources are often going to be the most and so a lot of things end up tilting in ways that end up favoring the more advantage unless there's a kind of conscious discussion about how to make sure that things are sort of equity producing so so all I would say is I think we ought to really we ought to be excited about all these kind of new possibilities that the kind of new technological environment offers but we ought to think about trying to make sure that we do them in ways that favor the more disadvantaged population rather than sort of tilt towards the more advantage population you do you guys up there next? How much time do we have? I don't want to take about seven and then maybe over here next if someone can take the microphone over there hi one of the things that I thought was the most striking or one of the most striking things about your presentation was that it seemed like a lot of the gap in achievement by income existed prior to students even getting to kindergarten and I was just wondering if you had any thoughts about why that may exist particularly as it relates to maybe things that are happening in the home and how we could possibly address those things using public policy given that we don't get those students until they're actually in the education system yeah I think you know in terms of where we ought to invest limited money and limited energy early childhood investments have got to be sort of number one from an educational perspective like getting kids to school ready to learn and equally ready to learn regardless of background would do more to change all forms of inequality in society than anything else you can probably think of I mean I think we ought to think also about housing policy and segregation issues but you know sort of one thing I was allowed to do I was elected president I had four years and I can only focus on one policy and I'd already done healthcare you know you know I would I would invest big time in early childhood and lots of different kinds of early childhood things things like the nurse family partnership that helps first time low income moms learn how to be a parent and deal with all the stress and anxiety unfamiliar what am I supposed to do with this young person kind of things like that have shown to have long term impacts for kids things like high quality early childhood programs, childcare preschool all that stuff pays off you know the investment you got to think of it as investment not a cost right it's because I think the payoffs would be enormous if we sort of invested substantially early childhood stuff we don't we have a pretty loose public involvement in early childhood role and I think that's why we see the gaps so as big as they are I wonder about the difference in outcomes between the white non-Hispanic white traditional sort of population and the black population and the immigrant Hispanic Asian and so on population because it seems to me that you know one of your conclusions is about the negative feedback loop where the rich the children of the rich get you know do well in schooling and the children of the poor do poorly in schooling and therefore you know you're not seeing the traditional social mobility the American dream but I think that that's still there for the children of immigrants you know who do a lot better by and large than their parents unless their parents were professional immigrants of which and also the children now increasingly there is a new and growing black middle class so I wonder if the people who are actually stuck with that disappearing middle if it's not more the traditional non-Hispanic white population that is experiencing that yeah so a couple of things are let me just a few data facts so if we look at this story just for whites if I showed you these pictures just for whites it would look the same that is the gap between the 90th percentile white and the 10th percentile white has gotten bigger by the same amount over this time period if we look at it just for blacks it also is the same though it's harder to estimate because the samples are smaller and it's hard to look at for Hispanics over this long time period so the kind of widening gap seems to be there the black and the white population it's unclear about the Hispanic population when you look at the black white achievement gap and the Hispanic white achievement gap when kids enter kindergarten they're both quite big in fact they're equally big and economic disparities between Latinos and whites are about the same size as economic disparities between blacks and whites but something interesting happens the Hispanic white gap gets smaller in the first years of school and the black white gap gets bigger in the first few years of school so even though they sort of they're these sort of similar economic conditions in a very sort of crude sense and the gaps are sort of the same when kids get to school the trajectories once they're in school are quite different and one obvious reason is that Latino students are acquiring English the part of the reason their gaps are big is their lower levels of English skills are better in school and do better on tests and also that potentially black students are actually in schools that are often pretty lousy and that's why the gap between blacks and whites is getting bigger most of the research suggests that after you control for family income the black white gap is not zero but is a lot smaller than it is now there's a new paper by Jesse Rothstein where he uses sort of better measures of long term family income and he says that the black white gap goes away but that's the only paper to date that has really convincingly argued that income alone is enough to explain the black white gap most other papers suggest income alone isn't enough there's not sort of a comparable study of the relationship between income and the Latino white gap that I know of so that would be an interesting area to look more at I think do we have time for more? one short one I think there was somebody over here waiting thanks my question to Paul that particular graph we had the black white achievement gap reducing over the last half of the 20th century the income achievement gap increasing I thought that was really interesting I wondered if there was any conversation about looking in the public policies and that time frame that helped to reduce sort of the racial gap and whether or not those same types of policies but retargeted for income might have similar effects or no effects yeah good question so what happened in so these are kids born between sort of 1955 and 75 for which the black white gap narrowed dramatically what happened then there was the war on poverty there was the civil rights act there was desegregation of schools in the south there was desegregation of hospitals in the south there was a whole bunch of social policy that was aimed at increasing at reducing discrimination and disparities between blacks and whites particularly in the south but also most of that gap is from the south over that time period so it wasn't just that there was some education policy that tried to make schools serving black students better you know desegregation of hospitals in the south had a huge impact on black infant mortality and black infant morbidity disease and sickness which has implications for later on achievement things like head start the war on poverty all those things sort of helped the black population in lots of ways other than sort of directly educationally and so we would you know I guess the implication that would be we need some sort of you know wrap around targeted kind of set of social policies if we wanted to see the sort of same size of effect on the income achievement gap I guess and we're not really that's not really part of the national conversation we're kind of trying to take money away from them and services away from them rather than you know invest more in them on that lighthearted note help me thank me thank