[Raj Chetty] Perfect. All right, well thank you so much.
Can you all hear me, is that fine? So thanks so much for having me, it’s a pleasure to be
here to speak with all of you today. I’m going to talk about how we can use modern
data sets to tackle, I think, some of the most important social and economic
challenges of our time in our own local communities building on a lot of the
data sets created here at ICPSR through the network and elsewhere. But I want
to start at a much bigger picture level by talking about the American dream
which is, of course, a complicated concept that means different things to different
people. But I want to think about it in a way that it’s been traditionally
conceptualized as a statistic that we can measure systematically in the data which
is the idea that America aspires to be a country where any child, through hard
work, should be able to go on to have a higher standard of living than their
parents did. And so in this first chart, here with my colleague David Grusky,
who’s here, and our co-authors we set about to assess the extent of which the
United States actually lives up to that aspiration by asking, What fraction of
kids go on to earn more than their parents did? Measuring both kids’ and
parents’ incomes in their mid-30s and adjusting for inflation. And we’re
looking at that data here. By the year in which children are born starting with
kids born in the 1940s, on the left, going all the way to kids born in the
mid-1980s, on the right, who are turning 30 around today when we’re measuring
their incomes. So what you can see here is that for children born in the 1940s
and 1950s it was a virtual guarantee that you were going to achieve the
American Dream of moving up, relative to your parents. 92 percent of children born
in the 1940 birth cohort went on to earn more than their parents did. If you look
at what has happened over time you see a dramatic fading of the American Dream,
such that for children born in the 80s who are turning 30 today it’s now essentially
a coin flip as to whether you’re going to do better than your parents, 50% of
children born in the mid-1980s do better than their parents did.
So that dramatic trend I think is of great interest from an economic
perspective. It’s a fundamental change in the way our economy works. I think
it’s also of great interest from a sociological perspective, from a
political perspective because I think a lot of the frustration that people around
the US are expressing, that the US is no longer a place where it’s easy to
get ahead, is basically reflected in this trend. So motivated by this broad trend
in our research group at Harvard Opportunity Insights, and what I’m going
to talk about today, we’re focused on, sort of, the very big picture question of,
How can you restore the American Dream, What is driving that trend that I just
showed you on the initial slide, and How might you go about trying to reverse
that trend through changes in policy? So I’m going to talk about a series of
studies that we’ve conducted in recent years investigating that set of issues
and give you a sense of where we are in thinking about the problem. There are going to
be three broad themes in what I discuss. First, as in much of modern applied
microeconomics we use large administrative datasets, big
data to use the popular buzzword in Silicon Valley, to study how to increase
upward mobility in this case. And I’ll talk more about the data in a second
knowing that that’s of particular interest to this audience. Second, rather
than focusing on any one particular set of intervention, so when you think about
inequality and social mobility you might naturally think about things like
education and we will think about the impacts of education but as you’ll see
our sense is that the problem actually has much more varied roots ranging from
issues related to education, to segregation, to social capital, and so
forth. And so the way we structure things is by analyzing a broad range of
interventions organized from a life course perspective from childhood to
adulthood. Finally, the starting point for a lot of
our work is that there are very sharp local differences in rates of upward
mobility and in kids’ long-term outcomes. And that is very useful both in
understanding the mechanisms for the fading of the American Dream and in
pointing towards potential policy solutions. So before I start to show you
that data, let me just sail in a little bit more
detail what types of data we use and the set of studies that I’m going to show
you. So the raw data source for a lot of the work I’ll be talking about is
anonymized Census data covering the US population, importantly linked to federal income tax returns from the IRS from 1989 to 2015.
So, of course, everyone here is familiar with Census data and various uses of it
over many decades in social science research. The key innovation that
researchers have been able to make in recent years is linking Census data,
which are of course cross-sections – snapshots at different points in time, to
tax records to turn it into a longitudinal data set. And that’s crucial
for the study of issues like the American Dream income mobility because
fundamentally there we want to study how things are changing over time as opposed
to studying in a cross-section, say what fraction of income goes to the top 1% of
the income distribution. And so the way we’re able to create that panel is
because people file federal income tax returns a year after year, and you have
their social security numbers, you’re able to create a panel where you can
essentially link the 2000 decennial census to the 2010 decennial census
using the tax data as a spine and then connect the American Community Survey to
that and do various other things. So you’re able to have this very
comprehensive longitudinal register essentially, which the Census Bureau has
now constructed and is being used for an increasingly wide set of projects. In
this case, we’re focused on issues related to economic opportunity. Now in
order to study economic opportunity and upward mobility across generations it’s
important not just to be able to follow a given person over time but also to be
able to link people across generations and these data sources are very useful
for that as well. So we’re able to link 99% of kids in the United States to
their parents through dependent claiming on tax forms. So any of you who have kids,
you’ll know that you need to write your kids social security number to claim
that child as a dependent on your tax return. And because kids are worth money,
essentially, in the US tax system virtually all kids get claimed and you
can link all kids to parents through that mechanism. And so in what I’m going
to show you here is an analysis using a data set of about 20.5
million kids, basically all kids born in the United States between 1978 and
1983, all kids either born in the US or who came to the US as authorized
immigrants while they were children. So it’s an incredibly comprehensive data set
where we’re able to follow people systematically over time on an
unprecedented scale without any issues of attrition and so forth
thanks to these administrative data. Okay so what do you get out of that data set?
I’ll start with this map here which shows you the geography of upward
mobility in the United States. So the way to think about this is for the kids born
in the early 1980s, we’re essentially disaggregating that initial chart that I
showed you and asking, What the kids chances of rising up in the income
distribution look like depending upon where they grow up. Let me first describe
exactly how we construct this map and then talk about how we interpret it. So
what we’ve done here is divided the US into 741 different metro and rural areas,
what are called commuting zones, they’re aggregations of counties. And in every
one of those counties, we take the set of kids who grew up there and calculate the
following simple measure of upward mobility. What is the average income at
age 35 for kids whose parents were at the 25th percentile of the national
income distribution? So take a set of kids whose parents were earning about
$27,000 a year, that puts you at the 25th percent of the parental income
distribution, and ask where did they themselves end up in the income
distribution in adulthood, right? So it’s a simple way of thinking about upward
mobility. The map is colored so that red colors represent areas with lower levels
of upward mobility and blue-green colors represent areas with higher levels of
upward mobility. So if you start by looking at the scale in the lower right
here, you can see that there’s an incredibly broad spectrum in terms of
rates of upward mobility across the different parts of the United States. So
there’s some parts of the US, for instance if you look at the center of
the country like Dubuque, Iowa for example, kids growing up in relatively
low-income families earning $27,000 a year, on average they’re earning $45,000
a year one generation later, terms of household income when they’re in their
mid-30s. So that’s really a very large rate of upward mobility across a
single generation. If you were to compare it to data from other countries, it’s a
higher level of upward mobility than we see in any country. Typically
Scandinavian countries, like Denmark and Sweden tend to have the highest levels
of social mobility and if you were to put this on that same metric you’d see
these are much higher rates of upward mobility. On the other hand though, if you
look at a place like Charlotte, North Carolina or much of the southeast or a
city like Detroit, Chicago, Cleveland you see much much lower levels of upward
mobility there. So to pick an example of Charlotte, kids starting out in families
at the same income level, $27,000 a year, one generation later their average
household incomes are about $26,000 years. So there’s actually no progress
over a 30-year period, there’s essentially no mobility in some sense
across generations. So naturally a question of interest to academic
researchers like ourselves is to understand why upward mobility varies so
much across areas in the United States? Why is it the case that some places
truly look like lands of opportunity as we traditionally like to think of the US
but other places look more like lands of persistent poverty? And from a policy
perspective I think that question is also of great interest because if we can
understand what’s driving this variation we can potentially figure out ways to
increase levels of upward mobility in the red-colored parts of this map. So
what I’m going to do in this talk today is go through a series of explanations,
some of which might already be in your minds, about what might be driving this
variation, evaluate a series of potential possibilities, and then building off of
that research-base talk, in the second part of the talk, about potential policy
levers that we can try to pull to increase economic opportunity in the US.
Okay. So let me start with what, I think, is the first explanation that comes to
mind for many economists which is that maybe this is about differences in the
types of jobs across areas or differences in the labor market. So if
you have a booming labor market with a lot of job growth, you would think
intuitively that you might have a lot of upward mobility there where the kids in
low and middle income families might rise up. So to evaluate that, let’s start
with an example, Take the case of Charlotte, many of you might know that
Charlotte is one of the most rapidly growing cities in the US. If you take any
traditional economic indicator that we’d be used to looking at like job growth,
rates of wage growth, and so forth Charlotte would be near the top of that
list, it’s kind of like the engine of jobs in the southeast. Yet as you can see
here, Charlotte turns out to be the 50th
of the 50 largest cities in America in turn in terms of rates of upward
mobility for kids who grow up there. So how is that possible? How can you
simultaneously have the highest rates of income growth while also being the
lowest in terms of upward mobility? It’s basically, the way to think about it is
that Charlotte imports talent. Lots of people move to Charlotte to get those
high-paying jobs. But apparently, as you can see with these new longitudinal
data, kids who grow up in low and middle income families there don’t necessarily
benefit from that economic growth. So that is a pattern that holds not just in
Charlotte but more generally. If we turn to this scatter plot here. Now we’re
plotting the measures of upward mobility that I showed you on the map on the
vertical axis, against rates of job growth from 1990 to 2010 when the kids
in these data were growing up on the x-axis. And we’re showing the data here
for the 30 largest metro areas and you can see that there’s basically no
relationship between these two things. In particular, you have places like
Charlotte and Atlanta down here in the lower right which have incredibly high
rates of job growth but very low rates of upward mobility. On the other hand, you
have other places where you have much lower rates of job growth but you have
much … you still have relatively high rates of upward mobility. So the first
thing that you can see from this very simple analysis is that job growth by
itself doesn’t necessarily … is not necessarily a sufficient condition for
having high rates of upward mobility for the current residents. So why does that
matter from a practical perspective? If you think about recent discussions, for
instance about trying to get the Amazon headquarters to your city, that doesn’t
necessarily I think translate into better outcomes for people growing up or
living in that area, to begin with. It might attract a lot of people to move in
but it’s not obvious that the current residents benefit. Okay. So that was
potential explanation number one. Is this about differences in jobs? That doesn’t,
you know, this is example of that analysis. We’ve done
various analyses of these types, it doesn’t really seem to be about
differences in the labor market. So a next potential explanation that you
might think about, coming back to this big map here, is that anybody familiar
with the demography of the United States would recognize that there’s a potential
connection with race here. So in particular, places that have larger
African-American populations like the southeast, like cities like Chicago, and
Detroit, and so forth those places also tend to be the ones that have lower
rates of upward mobility. Now we all know that there’s a long history of racial
disparities in the United States. And so you might wonder how much of the
difference that we’re seeing in these maps is driven by differences across
race that’s being manifested as a difference across places, and how much of
it is truly about differences across places for people of a given race? So to
get at that, we can exploit the fact that we’ve linked the Census data to the tax
data so we have information on everyone’s self-reported race and
ethnicity from the Census files. And we can construct this pair of maps here,
which now splits the data separately for black men on the left and white men on
the right. And it’s the same statistic that I was showing you before. What’s the
average income in adulthood of kids growing up in low-income families? But
now done separately for blacks and whites. And I’m going to show you the
data first for men and then women, and you’ll see why that
distinction by gender is important in a second. So if you look at these two maps.
Initially, you might react by saying, you know, it looks like in these two maps it
looks like they’ve put these two maps on two different color scales, right? It
looks like it’s on a red-yellow color scale on the left and a blue-green color
scale on the right. But if you look at the bottom, you can see that in fact the
maps are on the same color scale, it’s just that there’s such an extreme
disparity in terms of rates of upward mobility by race in the United States
that the two distributions are almost non-overlapping. That is to say, the
places that are the very best in terms of upward mobility for black men, a
place like Boston, for example, has lower rates of upward mobility than a place
like Charlotte, North Carolina, the place that ranks lowest in
terms of upward mobility for white men. So it’s almost like they’re two Americas,
one for black men and one for white men, completely disjoint from each other. So
what this shows is there’s no understating the importance of race, even
conditional on socioeconomic class, there are massive differences in outcomes by
race in the United States, you clearly can’t deny that. That being said,
there are still quite substantial differences across places even for
people of a given race. So if you look at the map on the right, for instance, you
see that there are much lower rates of upward mobility in Appalachia for white
men than there are in other parts of the country. And you’ll see that in more
detail as I zoom in to finer geographies, there’s substantial
variation across areas even conditional on race. So race matters, but so does
place. So I showed you these data for black versus white men. Let’s now look at
the same pair of maps for black versus white women. And what you see is a very
different pattern. So here the set of colors on the two sets of
maps are much more similar to each other, right? And if you look, you know, more
broadly at any statistics on mobility you don’t really find significant
differences between black and white women in terms of their chances of
moving up and down the income ladder when we’re looking at their own
individual incomes. Not when you’re looking at household incomes, because
then you bring the men’s and comes back into the picture so you basically bring
the previous set of maps back into the picture, but if you look solely at
women’s own incomes, rates of income mobility don’t vary that much between
black and white women. And so that’s interesting because it suggests that the
racial disparities that we’re seeing in the United States seem to also have an
intersectionality with gender. And so that leads you to potentially think
about a particular set of explanations that could, for instance, be related to
criminal justice and high rates of incarceration for black men growing up
in low-income families and so forth. Now in most of what I’m going to focus on in
this talk, I’m going to focus on upward mobility. How do you help kids from
low-income families rise up in the income distribution? But particularly in
the context of race, I think it’s also quite important to think about the
opposite channel of downward mobility. So take kids who start out in high
income families and ask where did they end up in the income distribution? And so
I’m going to do that here with this visual, which the New York Times constructed
using data that we had put out, and what this is showing you essentially is what
income dynamics look like for kids who start out in high-income families. So we’re
taking a set of men who start out in families in the top fifth of the income
distribution, and we’re asking where did they themselves end up when they’re in
their mid-30s ranging from the bottom fifth, to the second fifth, all the way to
the top fifth. And so the purple dots here are for black men and the green
dots are for white men. And what you see, I think, is a very disturbing and
distressing pattern about the United States, which is that if you’re a white
man and you’re born to a high income family, you tend to stay in the upper
class or the upper-middle class, those green dots kind of float towards the top.
But the purple dots, unfortunately, they cascade towards the bottom so even if
you were born to an affluent family as a black man, you have a very high chance of
ending up in the middle class or even at the bottom of the income distribution.
Your chances of ending up in the bottom fifth are almost as high as your chances
of remaining in the top fifth. And so that pattern is extremely important in
understanding the persistence of racial disparities in the US. Because one way
you might think about it visually is that for white Americans, achieving the
American Dream is like climbing an income ladder.
Whereas for black Americans, it’s more like being on a treadmill. Even after
you’ve climbed up in one generation, there are tremendous structural forces
that tend to push you back down in the next generation, as you can see here. And
then that fact leads to a persistence of disparities across generations. So why is
that potentially important from a policy perspective? I think there is rightly a
lot of focus on how we can improve opportunities for disadvantaged black
youth growing up in lower-income neighborhoods and so forth but what we
find in these data is that disparities by race persist even among families at
the very top of the income distribution, even families in the top 1%, even among
kids who go to the best schools in the United States you still see huge
differences in the outcomes of black men versus white men. And so what that suggests is that we should also be thinking from a
policy perspective about how you reduce those disparities even in affluent areas.
How do you help black families stay in the middle class or the upper-middle
class once they’ve reached there? That’s equally important in figuring out how
you narrow the race gap in the long run. So in what I’ve shown you so far, I’ve
shown you mostly national and regional statistics, at a fairly broad level. What
I want to do next to get a better sense of what’s going on is show you how these
differences across areas emerge at a much more granular level. And so in order
to do that, I’m going to toggle over to this tool that you can freely access on
the web called the Opportunity Atlas, which we constructed using the same data
that I’ve been describing here. And so way the way this works is it’s showing
you the same statistics in the national view that I started out with, the rates
of upward mobility by area. But what we can do with this tool is now very much
like a Google map, enter in any address you’d like. So here I’m going to go to
530 Sutter Avenue in Brooklyn which is going to illustrate some results that
I’d like to talk about. And so basically you can zoom in to any city you’d like,
any neighborhood you’d like. And so now what we’re doing here is looking at
the data census tract by census tract within New York City. So as you all know,
there are about 4,000 people on average per census tract in the United States. So
this is a much much finer level of geography than what we were looking at
before. We’re able to construct these statistics because you have data on the
entire US population, you have big enough sample sizes to get reliable estimates,
even at this level. And so the first thing I’d like you to see here is that
the spectrum of colors that you’re seeing on this map is the same as what
you saw on the National map, you go from the deepest reds to the darkest blues
within a couple of miles in New York City. So one way to think about that is
you can drive two miles down the road in New York and it’s like you’re going from
Alabama to Iowa in terms of rates of upward mobility. Right? So it shows you
this is an incredibly local phenomenon. It’s not about differences across
regions, or states, or cities even it’s about differences across neighborhoods
that are very near each other. Now I entered this particular address
530 Sutter in Brooklyn because this turns out to be a very large public housing
project in New York called the Van Dyke Public Housing Projects. And I’m going to
show you what is an interesting pattern if you look at black kids in this
area, which is called Brownsville in Brooklyn, so if you look right here
around this area, you see very poor outcomes for black kids growing up in
the Van Dyke Public Housing Projects, average incomes in adulthood
of just $18,000 a year. But you see kind of an interesting contrast, which is if
you look at this set of tracks right up here which is on the north side of a
road called Dumont Avenue, you see relatively poor outcomes. But then if you
cross right over that street on the other side of the street, you see
distinctly better outcomes with average incomes of something like 26 or 30,000
dollars in these tracks just south of that area that’s
highlighted. And so the reason I pick out that particular example is we see lots
of cases like that across the United States. You can look at your own city
where you’re going to see patterns like this very sharp variations across nearby
areas. This particular case is interesting because when we put out the
study a few months ago, NPR did some investigative reporting to try to
understand exactly what was driving these differences around Dumont Avenue.
And it kind of illustrates how you can learn from these data about what is
going on. So I’m going to go back here to the slides and let’s see what they
learned. When people find out where Audre Palacio is from, they often react in
disbelief. [Audra] “How could you come from there and you live there?” It’s, like, almost
as if it’s, like, I can’t believe you made it out. [narrator] Nearly 40% of Brownsville lives
in poverty. If you look at the Opportunity Atlas and zoom into
Brownsville, a lot of it is exactly what you’d expect: black kids raised in the
area 30-some years ago now make about $17,000 a year, same as their parents. But
once you head across Dumont Avenue everything changes.
Black kids from the same exact background are doing better than their
parents making around $26,000 a year. [Raj] So why is that the case? Why do we see
these sharp differences on two sides of the street? [narrator] In the 80s New York City had
been hard hit by a recession, then the crack and HIV epidemics. There was a part
of Brownsville that was totally abandoned, the other
side of Dumont. The New York City government sold over 16 square blocks of
Brownsville to the East Brooklyn congregations for $1. Those blocks were
dilapidated, rundown. The City agreed to build infrastructure and provide cash
subsidies for over a thousand affordable homes. They would start selling at
$30,000 each. They were called Nehemiah Houses after the man in the
Bible who rebuilt parts of Jerusalem. [new voice] The family was growing and we needed
something that was much better for the children. I didn’t like elevators, up and
down the elevators, for my children because it was a lot of people living in
the housing projects. [narrator] Audra Palacio was six when they bought
the house. [Audra] I remember when we moved. It’s in the Nehemiahs. We were so excited, we had rooms, we had space, we had a backyard. [narrator] Here’s Reverend Brawley.
He says the Nehemiah houses in Brooklyn gave children a space to do homework, a
good night’s sleep. [Brawley] When people have ownership of their properties, ownership
of their community, you have a better chance of addressing all core issues,
such as education and quality of life. [narrator] After I leave the family, I walk just a
few blocks to Dumont Avenue. According to the Atlas it’s the dividing line. On the
map it looks jarring but in person it’s completely unspectacular.
People bustle on their way to work, cars zoom by. Just another New York City
Street, it means nothing. But what side you’re on means everything. Jasmine Garsd NPR News New York. [Raj] And so what you can see there
is that we’ve gone from starting out with the American Dream at a national
level and seeing it fading over time, to drilling down to the state level, city
level, and now the neighborhood level. And you can see that the roots of the
American Dream really play out, I think, on different sides of a given street,
across particular blocks. As you saw from this particular example, in that case, it
seems like it has something to do with the development of mixed income housing
on one side of the street relative to having much higher rate of concentrated
poverty on the other side of the street. But the broader theme, I think, is that
through modern data, we can really zoom in on understanding what the roots of
complex problems like income ability at the national level are. So what I want to
do next is show you a few more results from our research in terms of
understanding the mechanisms for what’s driving these differences and outcomes
across nearby neighborhoods and then I’ll conclude by talking about policy
implications so a natural starting point in thinking about these differences
across areas is to wonder how much of the difference that we’re seeing in this
case now I’m showing you a snapshot of the opportunity outlets from Seattle
where you see a similar pattern very sharp differences and outcomes across
nearby areas how much of that is driven by different types of people living in
different neighborhoods versus the causal effect of growing up in one
neighborhood versus another so as you all know there’s been a long literature
in social science trying to understand that question it’s really at the heart
of sociology how important our neighborhoods in driving economic
outcomes and it’s central in terms of trying to figure out what one should do
from a policy perspective so if neighborhoods really have very large
causal effects if I take a given child and put a child in a different
neighborhood if I see very different outcomes for that kid then that means I
might want to think about interventions at a neighborhood level whereas if it’s
just about different types of people living in different places you might
want to have more of a people based solution so in order to get at that
issue what are the causal effects of neighborhoods and why two neighborhoods
if they matter why do they matter for kids outcomes we’ve done a series of
studies in which we look at kids who move across different neighborhoods and
rather than getting into the statistical details of these studies I’m gonna
describe what we find in the context of a simple example so here and this is why
I’m showing you the map here in Seattle we’re doing some work there which I’ll
come back to if you look here at Seattle I’m gonna use an example where we think
about kids moving from the Central District in the middle of the city where
kids who grew up in low-income families from birth or in about $26,000 a year to
a place called Normandy Park south of the city where you can see kids who grow
up in low-income families they have much higher rates of upward mobility they’re
earning in the mid 40 thousands when they grow up so what I want you to do is
think about a set of families that move from the Central District to Normandy
Park with kids of different ages starting with families who move when
their child is exactly two years old so what we do with the tax data and I’m
simplifying this in the context of this example but we’re gonna take kids who
make this move exactly at age two track that child forward 30 years in the tax
data measure that child’s income at age 30 and what you see in this first dot
here is that on average kids who move when they’re two years old from the
Central District to Normandy Park they have average earnings in adulthood of
about 39 thousand dollars a year okay so that’s for the kids who move when
they’re exactly two now we’re gonna repeat that analysis looking at kids who
move when they’re three four five and so on and what you see is a very clear
declining pattern the later you make that move from the Central District to
Normandy Park or more generally the later you moved from a red colored part
of the map to a Bluegreen colored part of the map the less of a game you get
and if you move after you’re in your early 20s the relationship is completely
flat and you get no gain at all so what do you see from these results I think
there are three key takeaways first apparently where you grow up
really matters it’s not just that there are different kids living in different
areas if you take a given child and that child moves to a different area
you see very significant changes in their life outcomes now a particularly
persuasive piece of evidence to me that seemed to suggest that this really was
about the causal effects of neighborhoods is that you can repeat the
chart that I’m showing you here comparing siblings within the same
family so imagine a family that moves with a 2 year old and a 7 year old
remarkably if you identify solely off of that variation you find that the
two-year-old is doing better than the 7 year old exactly in proportion to that
five year age gap like this curve so that suggests it’s not about differences
in the types of families moving to different places at different ages it
really seems like it’s about the causal effects of the neighborhood environment
second what you see here is that what really seems to matter is childhood
environment not where you’re living as an adult
so it’s about human capital formation while you’re growing up that’s the
mechanism through which it looks like neighborhoods matter and third every
extra year of exposure to a better neighborhood improves kids long-term
outcomes it’s not just the earliest years that matter most so many of you
will have heard of policy discussions focused on preschool interventions for
example like Head Start programs we think that could be very valuable but
what these data show you is that moving to a better environment when you’re 10
instead of 15 or 15 instead of 20 still helps quite a bit and so environment
seems to matter throughout childhood not just at the very earliest years so what
I’ve shown you so far is it really seems like where you’re growing up matters for
human capital development so forth now what you’re probably wondering is what
is it about the blue green colored places on the map that’s actually
leading them to produce better outcomes than the red colored parts of the map
and how might we replicate the successes in those high opportunity areas in other
places so that is a very difficult question to answer that’s one we’re
continuing to study in our research group and a number of others are
studying as well what I’m gonna do here is instead summarize a set of
correlations that we’ve looked at where we’ve looked at a variety of theories
that sociologists and economists have proposed over the years and asked what
are the factors that are most predictive of these differences in upward mobility
that we’re seeing a neighborhoods and in the interest of
time I’m just going to summarize the four strongest patterns we find which
are that neighborhoods with high rates of upward mobility tend to have less
concentrated poverty so like that NPR example places that have more mixed
income communities tend to have higher rates of upward mobility second places
that have more stable family structures so more two-parent families in
particular those types of places tend to have much higher rates of upward
mobility third places with greater social capital so what is social capital
I think of the old adage that it takes a village to raise a child is capturing
what social capital is about how connected as a community Salt Lake City
but the Mormon Church is thought to be a place with a lot of social capital
consistent with that Salt Lake City has very high rates of upward mobility in
our data and other places that look like that tend to have high levels of upward
mobility and then finally as you might expect intuitively places with better
schools K through 12 school systems that measured in various rough ways tend to
have higher levels of upward mobility so what you can see here is that it doesn’t
really seem to be about any one single factor we’ve never found that it’s just
about schools or just about segregation really seems to be about a combination
of things at least from a predictive point of view and so the way in which we
should think about improving upward mobility going forward has to be
somewhat holistic I don’t think it can really focus on one exclusive set of
interventions and so motivated by that I now in the remaining time want to turn
to word talking about how we can use these types of data and this type of
research to actually inform what we should do from a policy perspective to
increase upward mobility so in our group we’ve organized our efforts in that
direction around three pillars so if you think about what I’ve shown you so far
from a research perspective I if you think about the American Dream in the
United States one way to visualize it is that there are many distinct pipelines
of opportunity the neighborhood in which you grow up from birth to something like
age 22 or 23 and those pipelines they vary in quality dramatically across
different parts of a city some of these pipelines are very good we see kids from
low and families doing very well there are other
places though where they’re not functioning so well so for whatever
reason we’re not seeing kids growing up in low-income families rising up in
those places so if you take that view of the world that there are these lots of
these separate small pipelines to opportunity we estimate that it’s
something like a half mile radius around your house that’s the the size at which
neighborhood seems to matter for kids outcome so it’s incredibly granular so
if you have that view of the world I think there are three intuitive ways in
which you in which you might think about intervening to increase upward mobility
the first is to reduce segregation so you know one simple way to look at it is
if there’s a better pipeline to opportunity two miles down the road than
where I’m living now maybe one thing we can do to increased rates of upward
mobility is to help kids growing up in low-income families move to those better
pipelines to those neighborhoods that are gonna give them better opportunities
to rise up it’s the way I think about that is that’s basically trying to
reduce the amount of segregation that we have in American cities
lots of low-income families are segregated into high poverty low
opportunity areas can we tackle that problem and create more integrated
environments second recognizing that you can’t possibly help people move to
different areas and achieve impact on scales solely through that approach you
also have to figure out how you increase opportunity in what are currently the
low opportunity areas and so there I think of a place-based investment
approach where we try to figure out what is not working in some of these red
colored parts of the map and increase upward mobility there and then finally
recognizing that after age 18 the key touch point for children is not their
childhood home but where they’re going to college we it’s quite useful to think
about how you can amplify the impacts of higher education institutions like the
University of Michigan for instance on increasing upward mobility so let me
talk briefly about each of these three things in turn and then I’ll conclude so
in the context of reducing segregation let’s start with that so I want to go
back to Seattle where we’ve done some work on this issue and you could see how
the data that we’re able to construct using modern tools can really be used in
a practical way here so when we these opportunity outlets Maps we got a
lot of inquiries from people working in local housing authorities asking whether
they might be able to use these data to improve the services that they offer low
income families so in particular some of you might know that in the u.s. we spend
about forty five billion dollars a year on various affordable housing programs
which have the intent of giving families access to better opportunities to higher
opportunity areas now we started out by looking at whether families are actually
using that assistance from the federal government to move to find housing in
high opportunity places so what we’ve done here is overlaid on the opportunity
atlas map for Seattle the most common locations where families receiving
housing vouchers from the federal government what used to be called
section 8 vouchers but are now called Housing Choice vouchers where do these
families typically live in so the 25 most common locations in Seattle and
King County are shown by the green dots here and what you can see is that those
dots tend to be clustered in the red and yellow parts of the map not the blue
green parts of the map right so even though families are getting this rental
assistance which in principle is supposed to give them access to higher
opportunity neighborhoods in practice 80% of families are using these vouchers
to live in neighborhoods where we think their kids are likely not to have great
outcomes where poverty is likely to persist across generations so motivated
by this pattern we in collaboration with the King County and Seattle housing
authorities asked you know why why do we see this see that families aren’t
choosing to move to these high opportunity areas is it on the one hand
that it’s about preferences maybe they don’t want to live in Bellevue or on the
other side of the lake even if they could afford to do so because it’s
farther away from their family or their jobs there could mean many good reasons
you don’t want to move to those places or you feel you don’t fit in etc or is
it about barriers in the search process where maybe you have a hard time renting
a unit in a higher opportunity place even if you want to do so because you
lack information or assistance in the search process you kind of don’t have a
broker to help you maybe landlords don’t want to rent
to you because dealing with the housing voucher system can be complicated maybe
you don’t have the liquidity you need a little bit of money to pay the security
deposit any fees and so forth and so we conducted a randomized trial called the
creating moves to opportunity program in Seattle which was designed in
collaboration with these local housing authorities where we tried to remove
some of those barriers that families face in moving to high opportunity areas
in the form of providing housing search assistance recruiting landlords to
participate in the program simplifying the inspection process giving them an
insurance fund saying that if anything goes wrong we will cover that providing
a little bit of short-term financial assistance and this whole program just
to give you a sense of magnitude it costs about $2,500 per family to provide
this kind of customized support so that is a non-trivial amount of money upfront
per family but it’s a small amount in comparison to the amount we spend on
vouchers already per family so the average voucher payments they’re about
$1,500 a month for many years so we estimate that this program is only a 2%
incremental cost relative to what HUD is already spending on the voucher program
okay so this was done through a randomized trial where families that
come in to apply for a voucher and Seattle they either get these additional
services we tell them what the high opportunity areas in Seattle are and so
forth and help them find housing in those areas if they want to move there
or a control group that just received standard services and so here’s what we
found so this is the fraction of families that move to high opportunity
areas in the control group you can see that’s 14% the vast majority of families
are living in lower opportunity neighborhoods as I showed you in the
initial map in the treatment group that number jumps up to 54% the majority of
families are moving now to these neighborhoods for we estimate that on
average kids who grow up there from birth will earn an additional
$200,000 over their lifetimes based on the analysis that I was showing you
before and so you can see how this plays out visually if you look at map go back
to the map of Seattle the blue shaded area here
which is a little bit hard to see on the screen it’s the higher opportunity parts
of the map that I was showing you before you can see that the families in the
control group which are shown by the red pins they are concentrated outside that
area they tend to live in the lower opportunity neighborhoods in Seattle
in contrast the families in the treatment group shown in the green pens
they are concentrated predominantly in the higher opportunity areas and
importantly notice that those green pens are scattered all over Seattle so it’s
not like all of these families are moving to the same neighborhood if that
was the case you would worry that you then change the character of that
neighborhood that could lead to people potentially leaving that neighborhood
and so forth what’s very encouraging about this is that where people choose
to move is incredibly dispersed so it really seems like through this
relatively simple intervention you’re able to create substantial integration
in the city so that’s a very encouraging result in my view because it shows that
the segregation that we see in American cities is not due to deep-rooted
preferences on among tenants or among landlords actually through a relatively
simple set of interventions that HUD and other housing authorities are now trying
to scale throughout the US with our team support you can potentially have quite a
substantial impact on segregation and economic mobility in the US okay so let
me now briefly talk about the the other two pillars so as I said you know some
families or they choose to move to different neighborhoods we think that
can provide their kids good opportunities to succeed but we
recognize that many families choose to stay where they are for good reasons and
so we need to figure out how we can bring opportunity to those lower
opportunity areas so we’re also doing quite a bit of work in that dimension so
in particular we’re focusing on the City of Charlotte where when we put out the
initial statistics that I was talking to you about Charlotte ranks 50th among the
50 largest cities in America in terms of upward mobility that simple result
itself motivated a lot of change in Charlotte there was a lot of concern
among the CEO of Bank of America lots of local leaders how could we both be one
of the most rapidly growing cities in the u.s. that we’re very proud of yet be
failing our children essentially and so they set up a task
horse and a commission to try to look into this figure out how they can invest
in the local neighborhood and so forth now the big challenge as our team and
others try to support these efforts is that we don’t yet know exactly what
works in terms of increasing upward mobility in areas that currently have
lower levels of Economic Opportunity so we have a sense that it’s primarily
about programs that try to increase human capital improve the childhood
development process it’s not fundamentally about bringing more jobs
to an area things like that but within that space of trying to improve kids
outcomes there are many many different programs at different ages that people
have tried so their early childhood interventions their mentoring programs
their ways to try to improve schools their programs that try to connect kids
to colleges and so forth we think all of these have promised we’re trying to
evaluate a number of them the fundamental challenge is that we haven’t
had data historically in a longitudinal form that allows us to really understand
which of these programs are most effective and so in order to address
that problem in collaboration with Trent Alexander here at Michigan KD genetic
and David drusky we are doing some work to to remedy that problem and create a
new data infrastructure that we think can be really useful in studying these
and many other issues we’re calling that the American Opportunity Study and what
that is is construction of an anonymized longitudinal data set by linking census
and tax data covering all Americans from 1950 to 2020 so a 70 year longitudinal
panel that’s essentially extending the type of data that I was describing here
in recent years much further back historically that’s as you can imagine
incredibly complicated project that involves digitization of records held on
microfilm optical character recognition record linkage a bunch of different
things that will take time and will be conducted at the Census Bureau and the
plan then coming back to the application here is to use these data to study the
impacts of place-based interventions on the prior residence
in a given neighborhood so the fundamental problem in previous studies
of place-based policy is you can do an intervention in a particular place and
it looks like that place has gotten a lot better you might see lower poverty
rates lower rates of crime and so forth the problem though is you don’t know if
you actually help the people who are living there to begin with or if you
just displace them and gentrified the neighborhood which is obviously not you
know typically the intent of these sorts of policies we haven’t been able to
figure that out in the past because we couldn’t track the people who were
living in these areas to begin with over time and this type of data
infrastructure effort will solve that problem and we think allow a systematic
evaluation of basically which of the many many policies implemented in the
war on poverty are most effective in increasing economic mobility so last
piece improving higher education so I’ve been mainly talking about neighborhoods
as a potential unit of change but as I was saying I think there’s also a lot
that can be done in our higher education institutions and so I wanted to show you
a quick slide that illustrates that which shows you similar types of data
that we’ve constructed and are now publicly available by college in the
United States so when you think about social mobility at colleges and what
colleges are contributing to social mobility there are two dimensions that
matter the first is kids outcomes so we’re
showing that here on the y-axis plotting a simple measure of upward mobility for
students at each College so what we’re doing here is a statistic which is just
what fraction of kids who start out in low-income families in the bottom fifth
of the income distribution what fraction of them reach the top fifth of the
income distribution so a simple notion of upward mobility and so you can see
for instance that places like Harvard and the University of Michigan look
terrific on that on that dimension kids from low-income families who attended
schools they have great outcomes as you might expect intuitively however what
matters for upward mobility is not just the outcomes of the low-income kids you
have but how many low-income children you actually have at your college right
if you don’t have any low-income kids you can’t be contributing a whole lot to
upward mobility no matter how good your outcomes are and so unfortunately
if you look at Harvard and maybe somewhat more surprisingly the
University of Michigan as well they essentially serve no low-income kids so
only 3% of kids at University of Michigan come from low-income families
it’s the same thing at Harvard and now people get all of these other dots
there’s one dot here might be a little bit hard to see on the screen for each
of the colleges in the United States and you kind of see a downward sloping
pattern where the there are some colleges that serve lots of low-income
students like community colleges for example but the outcomes don’t look
great and then there are other colleges like Harvard in the University of
Michigan that I have excellent outcomes but don’t serve many of our income kids
so in order to amplify the impacts of colleges and upward mobility you really
need to think about how you push colleges to the upper right of this
chart right how do you increase access at places like Harvard in the University
of Michigan there’s some great work being done here at Michigan itself by
Soudan darsky and her collaborators and what’s called a hale program that is
actually led to a significant increase by recruitment of applications from kids
from lower-income families that’s an example of the type of intervention that
can make a difference and analogously how can you improve outcomes at some of
these colleges that are serving lots of low-income kids how do you help them
achieve better outcomes these types of data I think can allow you to take a
more granular look at that and figure out what you need to do in different
places and that’s something we’re working on so let me conclude by talking
about hopefully I’ve Illustrated how these modern data I think can be really
useful in understanding major social problems in a more granular and precise
way I want to talk briefly given this audience about how I think we can build
on this to use data for social science in the era of big data and the modern
information age and so I think there’s three key things to think about so the
first is linking administrative and survey data sets at secure repositories
I think has incredible value so data of librarians could potentially help
researchers tremendously by identifying new data sources that could be linked so
there are secure protocols that we and others have developed where you can
enjoy best data that might be held by a
company or local governments or many many other sources school districts and
so on and I think there’s great value as I hope I’ve Illustrated here in linking
different types of data so I think that takes a lot of work but it would be
incredibly valuable public good second I think constructing and disseminating
publicly available aggregate statistics via icpsr or other repositories would be
very useful so the opportunity Atlas data for example that I was showing you
or the statistics by college those things are all publicly available on
this opportunity insights website if you just go to that URL or google that you
can download all these different datasets and I think there are a number
of research studies that can be done just using those publicly available
aggregated data sets that would have had a lot of value and then finally I think
the dissemination of aggregated data usually we think about the intended
audience as academics and researchers but I think there’s great interest in
the media the general public the policymakers in using these data to
understand exactly what’s going on in their communities in ways that can be
very impactful so let me end by coming back to all of you and just thanking you
for the great work you’ve all done I think to make these types of data
available to researchers like myself and many many others I think that’s really
foundational for the work a number of people are doing from basic science to
apply it impact to actually helping kids grow up in different neighborhoods and
I’m very grateful for that effort thank you