Improving Equality of Opportunity in America: New Lessons from Big Data

Improving Equality of Opportunity in America: New Lessons from Big Data


>>My name is Don Mar. I’m the chair of the economics department. I welcome you for braving the weather coming
on a Friday evening to hear enlightening lecture about economics. A couple of things. I just want to say something about Dr. Betty
J. Blecha. Betty Blecha was a remarkable professor here
at San Francisco State University from about 1983 to 2012, and she had this belief that
San Francisco State was a special place, but she also believed that San Francisco State,
people should be exposed to world class people, world class thoughts and world class lectures. And so she endowed this series of lectures,
which we’ve had — this is the second Betty J. Blecha memorial lecture. So we’re really pleased to offer these lectures
to the general audience. Just a couple other things about Dr. Blecha. She was actually a very multitalented individual. She was a world class photographer. She did artist and residence at Grand Canyon
National Park, and she had a variety of interests, and I will just leave you with this one thought
about Dr. Blecha, and that she would have been very happy as a Bob Dillon fan at the
current Nobel Prize for literature. Anyway, I will turn this over to my colleague,
Dr. Venoo Kakar, who will introduce our guest.>>Thank you.>>Thank you, everyone, for joining us. My name is Venoo Kakar, and I’m an assistant
professor of economics here at San Francisco State University. On behalf of the economics department, I’d
like to welcome you all to this event on the [inaudible] theme of improving equality of
opportunity in America, new lessons from big data. This is a team that has been very close to
our hearts here at SF State where we strive for upward mobility. But I’m very pleased to see the turnout on
a Friday evening, which we were very concerned when we were planning this event would it
be successful, but thank you all for making this possible. We’re delighted to have with us beyond San
Francisco State University community students and faculty from UC Berkeley, Mills College,
University of San Francisco, Diablo Valley College and California State University East
Bay. We also have two very important people in
the audience with us today, President Wong and Mrs. Wong. Thank you for coming and supporting this event
despite your very busy schedule. In just a moment, I will introduce our distinguished
speaker, Professor Raj Chetty. After his presentation of about an hour, you
will have the opportunity to ask questions. There are two mics on the floor. One here, one on my left, and Dr. Michael
Barr and myself will be helping you out with that. I also would like to mention that we are video
recording this event, and this will be made available on the SF State Youtube channel,
and finally, I’d like to turn over the slide for some fun trivia. So what do Professor Chetty and this person
have in common? And I’d like to ask you guys if you know. And you can shout. Harvard. Excellent. ^M00:03:50
[ Inaudible ] ^M00:03:52
[Inaudible]? Excellent. It is true. President Obama and Professor Chetty both
went to Harvard University. The president, of course, to study law, and
Professor Chetty to study economics. They both shrae their birthday as well, like
you said. They were born on August 4. However, Professor Chetty happens to be a
lot younger than our president. Finally, they both work extremely hard to
improve the equality of opportunity in America. Professor Chetty mostly on the research side,
and President Obama on the policy side. Professor Chetty’s work on education was cited
by President Obama in the State of the Union address in 2012. In the research that our president cited,
Professor Chetty found that high value added teachers can boost the long run earnings of
their kindergarten students, but it is to Professor Chetty’s work and to his credit
that we now know that there are huge variations in rates of upward mobility across the United
States. Professor Chetty is one of the top scholars
and prolific researchers in the world. He’s a professor of economics at Stanford
University, co-director of the Public Economics Group at the National Bureau of Economic Research. He has been named one of the top economists
in the world by the New York Times and the Economist magazine. Dr. Chetty received his Ph.D. from Harvard
University in 2003 at the remarkable young age of 23 and became an assistant professor
at UC Berkeley right after that. So I will let you guess how old or young he
is. He returned to Harvard University in 2009
as one of the youngest tenured professors in Harvard’s history and then moved to Stanford
in 2015. He has won many accolades for his accomplishments. He was awarded the MacArthur Genius Fellowship
in 2012 and is one of the youngest recipients of the John Bates Clark medal given by the
American Economic Association to the best American economist under age 40. This is extremely prestigious since 40 percent
of economists who win the John Bates Clark medal go on to win the Nobel Prize in the
economic sciences. No pressure here. He was also awarded the Padma Shri, one of
the highest civilian awards for distinguished service in any field by the government of
India in 2015. His work on tax policy, unemployment, education
and social mobility has been vitally cited in the media and congressional testimony. His articles have appeared in top rated economics
journals such as The Quarterly Journal of Economics, The American Economic Review, and
Econometric among others. In short, tonight’s going to be spectacular. So it is with great honor that I request you
all to join me in welcoming to the podium, Professor Raj Chetty. ^M00:07:01
[ Applause ] ^M00:07:12
^M00:07:34>>Dr. Raj Chetty: Thank you so much for the
very warm introduction. Venu, it’s really a pleasure to be here at
San Francisco State. Thank you all for coming out this evening. So I’m going to talk this evening about improving
equality of opportunity in America, and I’m going to focus specifically on things we can
do here in San Francisco and in California to improve opportunities for disadvantaged
youth. But I’m going to start at a much bigger picture
level by talking about the American dream, which is a complicated concept that means
different things to different people, but I want to distill it to a simple statistic
that we can measure systematically in the data. The probability that a child born to parents
in the bottom fifth of the income distribution makes the leap all the way to the top fifth
of the income distribution. So think of that as the classic Horatio Alger
version of the American dream. How common is that in the United States versus
other developed countries around the world where we have similar data? In the U.S., children who are born to parents
in the bottom fifth of the income distribution have a 7.5 percent chance of making it to
the top fifth. That compares with nine percent in the United
Kingdom, 11.7 percent in Denmark and 13.5 percent in Canada. Now, sometimes when people see these statistics
they react by saying, oh even in Canada, it looks like your odds of success aren’t that
high, right? Only a 13.5 percent chance of succeeding. But you have to remember, of course, that
no matter what you do, you can’t have more than 20 percent of people in the top 20 percent. And so the maximum value that this statistic
can reach is plausibly 20 percent. That is to say if we lived in a society where
your parents played no role at all in determining your outcomes we’d expect to see 20 percent
of kids move from the bottom fifth to the top fifth. So that’s kind of the upper bound that you
would expect this value — this statistic to take. And so relative to that benchmark, these are
actually quite large differences in rates of upward mobility across countries. One way I think about it is that your chances
of achieving the American dream are almost two times higher if you’re growing up in Canada
rather then the United States. Now, these differences across countries have
attracted a lot of public attention and a lot of concern, generate a lot of concern
that the U.S. may no longer be a land of opportunity. ^M00:10:12 Contrary to its traditional perception. But what I’m going to focus on today is that
upward mobility varies even more within the United States, and I think that’s actually
something we can learn much more from then these broad comparisons across countries. In recent work with my colleagues, we calculate
upward mobility for every metro and rural area in the United States, and we do that
by using anonymous data from tax and social security records on ten million kids born
between 1980 and 1982. So basically all children born in America
between 1980 and 1982. Now this work that I’m going to show you is
an example of a broader trend in social science towards the use of big data to answer important
economic and social questions. And so I think that’s one of the most profound
trends in our field that’s transforming not just the question I’m going to discuss today
but a variety of other problems. Much as you hear about big data being used
in the private sector, companies like Amazon and Google using huge data sets to offer better
products, likewise our vision is that you can use these data to answer social questions
and provide better evidence to policymakers to make better decisions that will improve
people’s lives. So what we do here is rank children and parents
in the national income distribution, and we look — assign children to locations based
on where they grew up, and we use that data to draw this map here, which shows you the
geography of upward mobility in the United States. So what this map is plotting is the same statistics
that I started out with, the fraction of children who start out in the bottom fifth of the national
distribution and make it to the top fifth. The map is colored so that lighter colored
areas represent areas with higher levels of upward mobility. So there are 740 areas in this map, so one
of them, for example, is the bay area represented here by San Jose, and likewise all of the
other metro and rural areas in the U.S. — and you can see if you look at the scale is that
there’s an incredibly broad spectrum in terms of rates of upward mobility within the United
States. In some places, rates of upward mobility like
in the middle of the country exceed 16 percent, higher than the numbers we saw for Denmark
and for Canada. At the other end of the spectrum, if you look
at places like Atlanta and Charlotte, the numbers are below 4.5 percent, lower than
any country for which we currently have data. So even within America, there’s this incredibly
broad spectrum in children’s chances of escaping poverty and achieving the American dream. Now in this large map, your eye gravitates
towards the broad regional variation. The much higher rates of mobility in the midwest
than in the southeast, the relatively high rates of mobility — upward mobility in the
west coast and much of the northeast. But it turns out if you zoom in further, now
let’s take a look at the data here within the bay area, there’s a lot of variation even
across counties within a community or even across relatively nearby neighborhoods. So if we look at the data for the bay area
here, again looking at the same statistic, what fractions of kids make it to the top
starting at the bottom, we see a remarkable amount of variation across counties in the
bay area where for kids growing up in the 1980’s, a remarkable 18 percent of kids growing
up in San Francisco or San Mateo made it to the top 20 percent. Remember, relative to that benchmark of 20
percent perfect mobility, this is really an incredibly high rate of success. In contrast, if you go across the bay bridge
to Oakland, that number falls nearly by a factor of two to 11.4 percent. So even within relatively narrow geographies
we see really sharp differences in the rate at which kids achieve the American dream. So naturally, the question of interest to
us as researchers and to policymakers is to ask why upward mobility varies so much across
areas and what ultimately we might be able to do from a policy perspective to improve
rates of upward mobility more broadly. There are two very different explanations
for the variation in children’s outcomes across areas that we’re seeing in these maps. The first is that there’s heterogeneity. That is, different people live in different
places. Maybe the types of people who live in Atlanta
are just different from the types of people who live in Iowa. We know that that’s true at some level, and
maybe this is not really so much about different places but rather about different people and
different places. The second possibility is what sociologists
and economists call neighborhood effects, the idea that places have a causal effect
on upward mobility for a given person. That is, if I take a given child and put that
child in San Francisco instead of Oakland or in the bay area instead of Atlanta, I’m
going to see really different outcomes for that given child. There’s been a lot of work over the past 30
years or so in sociology and economics to try to distinguish between these two very
different explanations. But that work has been quite controversial
and somewhat inconclusive. Recently, thanks to big data, modern data
sets were able to discriminate between these explanations much more precisely. So let me explain how we do that by stepping
back and first asking what is the ideal experiment that you would run to test whether where you
grow up really matters, really effects how you do in the long term. So the ideal experiment you would run would
be to randomly assign children to neighborhoods and compare their outcomes in adulthood. So if you were thinking about this from a
scientific perspective, you’d like to randomly assign kids to each of those 740 areas I showed
you on the map and compare how they do in adulthood depending upon where they grew up. So naturally, it’s very difficult to run an
experiment like that in practice. And that’s, I think, the key challenge in
social science that we’re not able to run experiments as scientists are typically able
to do. So instead what we do is approximate that
experiment using what I’ll call a quasi-experimental design. We study five million families who move across
areas in observational data that we have from tax and social security records, and the key
idea of approach is to explore variation in the age of the child when the family moves
in order to identify the causal effect of environment. So let me show you how this works with a simple
example rather then going into the statistical details. So let’s take a set of families that start
out in Oakland. And to make things concrete, suppose if you
grow up from birth in Oakland you earn $30,000 on average when we measure your income at
say age 30. And now consider a set of families that move
from Oakland to San Francisco which as we saw in the map seems to produce better outcomes
for kids in low income families. So let’s say pick a number again. If you grow up in a low income family in San
Francisco, when you’re 30 years old, you earn $40,000 on average. So now what I want to do is take a set of
families that move from Oakland to San Francisco with kids of different ages, and lets start
by looking at children who move when they’re exactly 9 years old from Oakland to San Francisco. What we do is we track these kids who moved
when they were nine forward 21 years and measure their own earnings when they’re 30 years old. And we ask how are they doing relative to
the kids who grew up in Oakland from birth and the kids who grew up in San Francisco
from birth. And what we find is you see in this [inaudible]
here is that they end up about halfway between the kids who grew up in Oakland from birth
and the kids who grew up in San Francisco from birth. That is, they’re earning roughly $35,000 when
we measure their incomes as adults. So that’s for the kids who moved when they’re
exactly nine. So now let’s look at the data for kids who
moved when they were ten, 11, 12, 13 and so on. What you see is a very clear declining pattern. The later you make that move from Oakland
to San Francisco, the less of the gain you get. If you move after you’re 21 or so, you get
essentially no gain at all. If you move after that point as a young adult,
there’s absolutely no effect. So what does this chart show you? I think it has three very important lessons. First, it shows you that place matters. It’s not just that the kids who live in Oakland
are different from the kids who grow up in San Francisco. Apparently, if you take a given child and
move that child from Oakland to San Francisco, that child’s outcomes change very significantly. That, I think, is a very encouraging message
because it suggests that these differences in upward mobility are malleable. It’s not just that some people have worst
prospects of moving up the income ladder than others. There’s actually something about the environment
that can really effect children’s outcomes, and this shows that we can do something about
the problem potentially to improve outcomes in places like Oakland and other areas in
the U.S. Second, you see that childhood environment really seems to be critical. If you move as an adult, it doesn’t do much
for you. If you move as a child, there are really significant
gains from moving to a better environment. Third, you see that every extra year of childhood
exposure to a better environment matters roughly equally. If you move when you’re nine instead of ten,
or 15 instead of 16, each incremental year contributes to greater long term success. That’s very important especially in light
of current policy discussions about early childhood education where there’s a lot of
focus on the idea that we need to invest at the very earliest years. ^M00:20:21 I think those kinds of investments
can be very valuable, but these data show you that there’s no reason to give up on kids
after they’re 5 or 6 years old. If you’re in a better environment even at
age 10 or even at age 15, that continues to have really substantial long term returns. Okay. So we think that place seems to matter. That is, where you grow up really has profound
effects on long term success. So now I want to turn to think about what
that means for policy. So what did we learn from that in terms of
how we can go about improving upward mobility in practice? So I think there are really two different
strategies that one can think about. The first is what I’ll call the choice based
approach, which is to help people move to better areas. You can step back and say, look, we know on
this map that some places like San Francisco apparently, at least in the 1980’s, produces
better outcomes than other nearby places like Oakland. So maybe we can just help low income families
move to these brighter colored areas on the map in order to generate better outcomes. And so that’s one approach that I’ll talk
about. Now second, recognizing that moving people
around has limits, you can’t possibly move everyone out of Oakland or a city like Atlanta
to a different place. You really need to think about a second set
of approaches, which I’ll call place based approaches, which focus on investing in low
opportunity areas in order to replicate the successes we see in areas that currently have
higher levels of upward mobility. So let me talk about each of these in some
more detail. So let’s start with the first appoach of moving
to opportunity or the choice based approach. So one way we might go about doing this is
to give low income families subsidized housing vouchers to move to better areas. As some of you might know, the U.S. already
spends about $45 billion per year on affordable housing. But most of that affordable housing importantly
is in low opportunity areas. So we have about two million families that
get section 8 housing vouchers in the U.S., but 80 percent of those vouchers are used
in relatively high poverty, low opportunity areas. And so there’s a real potential there in our
view to change that program so that more families are using the money they’re already getting
from the government to live in some of these areas that we identify as being higher opportunity
places. The second thing to note when thinking about
these policies is that we’re not just talking about uprooting a set of families who are
happy in the area where they live and trying to force them to move somewhere else. Instead, the way I think about it is that
lots of low income families, something like 20 percent, already move every year. And so the way to think about these policies
is when they’re making that decision to move, maybe we can encourage them to move to a area
that will be better for their kids in the long run. Redirect where people are moving rather then
getting a whole new set of people to move each year. Now this idea of moving to opportunity is
something people have thought about for some time, and in fact, there was an experiment
conducted in the 1990’s under the Clinton administration, which gave vouchers to families
to move to better — that is lower poverty areas using a randomized lottery. So this experiment involved about 5000 families
in five large cities in the United States in the mid 1990’s. And the way this experiment worked can be
illustrated in this map here, which focuses on the New York site of the experiment. They took a set of families that lived in
the Martin Luther King Towers, a high poverty, public housing project in Harlem, and some
of those families were given purely through a lottery a voucher that allowed them to move
to a lower poverty area. And many of those families who got that opportunity
ended up moving to a place called Wakefield in the North Bronx, which is about ten miles
away from the Martin Luther King Towers in Harlem. So basically, you can think of this experiment
as asking how do the kids who just out of pure luck got the chance to grow up in the
Bronx do relative to the kids who grew up in Harlem? And in recent work with my colleagues where
we track the long term impacts of this moving to opportunity experiment, we find that children
who move to lower poverty areas like the North Bronx when they were young below the age of
13, for example, do much, much better as adults. They’re earning 30 percent more, which translates
to $100,000 gain in lifetime earnings. They’re 27 percent more likely to attend college. They’re 30 percent less likely to become single
parents. So on a broad set of outcomes, you see really
substantial improvements in children’s long term success. Interestingly, moving had little effect on
the outcomes of children who were already teenagers. So if you move at older ages, there was hardly
any improvement. And moving had no effect on the parent’s earnings. So if you think about this pattern that I’m
describing on the slide, it lines up exactly with that declining pattern that I showed
for people moving from Oakland to San Francisco. If you move at young ages to a better environment,
you get significant gains. If you move at older ages, you don’t gain
all that much. And so what that data shows you is once again
what really seems to be critical is exposure to a better neighborhood during childhood
as opposed to moving there as an adult. And that’s really important when thinking
about policy because it suggests that the kinds of factors that matter are not things
like the availability of jobs in an area or the conditions in the local labor market,
but rather things that effect you while you’re growing up. So these data illustrate that moving to opportunity
can really be an effective way to improve children’s outcomes. Now, there are a couple of concerns you might
have about this sort of approach that policymakers often have. One is the cost of such an approach. So you might say, “Isn’t it going to be really
expensive to move all these families to better areas? Where is that money going to come from?” Well, it turns out that the increase in the
tax revenue from the children’s higher earnings, because remember, these kids are earning about
$100,000 more over their lives when they move to these better areas, that increased tax
revenue actually in and of itself more than offsets the incremental costs of the program
relative to keeping the families in the Martin Luther King Towers and the public housing
projects. So the government actually saves money on
that in the long term by having the sort of voucher program relative to traditional public
housing. A second very prominent concern that comes
up in the political debate I think is an important issue here in San Francisco, for example,
is that integration via decentralized housing vouchers might end up hurting the rich. So many people are worried about these kinds
of policies because they say, “Look, I’m a person in the middle class or from a higher
income family. I don’t want to have people from lower income
backgrounds living next to me because I’m concerned that that might increase rates of
crime or effect my own kids’ outcomes adversely.” Now, it turns out when we look at the data
that in fact if you look at mixed income neighborhoods where low income families are living near
middle income and higher income families, if anything, the outcomes of children from
rich families in those areas look better than they do in segregated neighborhoods where
the rich are living separately from the poor. So there doesn’t seem to be any clear evidence
in the data that having more integration, having more low income families live in proximity
to high income families actually has any negative effects on the others around them. And I think that’s very encouraging in terms
of generating political support for these types of policies. Okay. So that’s one approach that we think concretely
can work, improving our housing voucher program to provide more low income families affordable
housing in high opportunity areas. But there are, of course, limits to the scalability
of such policies, right? Policies that move people can only go so far
because you can’t possibly, as I was saying earlier, move everyone. And so you also need to think about what types
of policies can improve existing neighborhoods. The first step in identifying such policies
is to understand the characteristics of areas with high levels of upward mobility. So ultimately what you’d like to be able to
do is identify the exact recipe that cities that have high levels of upward mobility are
using. What exactly are places like San Francisco
or Salt Lake City or much of the midwest using to generate high levels of upward mobility
and how could you translate that to an environment like Atlanta? So we don’t know exactly that that recipe
is. We and many other researchers are studying
that question at the moment. But what I can tell you in the meantime is
a set of correlations. What types of factors really seem to be strongly
associated with these differences in mobility across areas? So we looked at a variety of factors that
might correlate with differences in mobility, and I’m going to show you here in the interest
of time the five strongest correlations that we’ve identified. The first is segregation. We find that places that are more segregated
by race or by income are associated with significantly lower levels of upward mobility. Now, there are many different ways in which
you can measure segregation statistically. But it turns out that the patterns are so
stark here that you don’t actually even need to worry about exactly how you go about measuring
segregation. ^M00:30:12 You can actually just see this
result visually. So let me give you some examples. This map here depicts the degree of racial
segregation in Atlanta. The way it’s constructed is that each person
in Atlanta is represented by a dot using data from the census, and the dots are colored
so that whites are blue, blacks are green, Asians are red, and Hispanics are orange. And you can see immediately regardless of
how you measure it, it’s clear that Atlanta is an incredibly segregated city. The green dots are completely separated from
the blue dots. The blacks and whites are living in completely
different parts of the city. And what we see is that cities like Atlanta
are — have some of the lowest levels of upward mobility, cities that are as residentially
segregated as Atlanta tend to have the 4.5 percent or four percent rates of moving from
the bottom quentile to the top quentile. Now compare Atlanta with Sacramento, which
has the same minority share as Atlanta, the same fraction of blacks and Hispanics as Atlanta. And you can see immediately that Sacramento
is a much more integrated city than Atlanta. The dots are much more interspersed. The colors are much more interspersed. Sacramento’s not perfectly integrated. There’s still significant segregation even
in Sacramento, but it’s considerably more integrated than Atlanta and corresponding
to that, we find that cities that look like Sacramento tend to have much higher rates
of upward mobility. So that’s the first strong correlation. We find higher levels of segregation are associated
with lower levels of upward mobility. Now why might that be? I think one plausible hypothesis is that this
is about role model effects or differences in aspirations or exposure. So think about if you live in a city that
looks like this, you’re much less likely to come into contact with people from different
socioeconomic backgrounds who might show you other pathways to success that you’re not
aware of in your own family or among your own friends relative to a city like this where
you’re much more likely to have that kind of exposure. Another potential explanation is that in a
city that’s more integrated, you benefit from a tax base to fund local public schools, for
example, that’s better than if you live in a much more segregated city. So there are many potential mechanisms for
why segregation might be associated with lower levels of upward mobility. We don’t know exactly what the mechanism is
that’s driving this relationship. But what we can say is it’s a really strong
robust pattern in the data that higher levels of residential segregation are associated
with lower upward mobility, and I think that’s very concerning in a city like San Francisco
as we have rising housing prices here and increasing levels of segregation, we have
to wonder about whether San Francisco will continue to be an area of opportunity as it
was traditionally in the 1980’s. Second, the second strong correlation we find
is with income inequality. We find that places with a smaller middle
class tend to have much lower levels of mobility so the degree of inequality within a generation
is strongly related to the extent to which people are able to move across income levels,
across generations. The extent of social mobility across generations. Again, this is potentially concerning as we
live in a society with growing income inequality, this relationship suggests that we have to
worry that children’s chances of achieving the American dream of climbing up the income
ladder also might deteriorate over time. The third and fourth factors come more from
sociology than economics. We find very strong associations with measures
of family structure. In fact, the single strongest correlation
we find in the data is that areas with more single parents have significantly lower levels
of upward mobility. Now, in understanding this relationship, it’s
very important to note that this correlation is not directly driven by the fact that children
who grow up in one parent households fare worse than children who grow up in two parent
households. And the way you can see that is that if we
look at the subset of kids whose own parents are married, that is the subset of children
growing up in two parent households, if such a child lives in an area with a larger fraction
of single parents, they are less likely to climb the income ladder. So even if your own parents are married, if
you grow up in an area with more single parents, you have a lower probability of rising up. So that shows you again that it’s not an individual
level difference. It’s again, some community level factor that
this variable is picking up. Related to that is the concept of social capital,
which we also find to be very strongly associated with differences in mobility. So the way I think about social capital is
the old addage that it takes a village to raise a child. The idea of social capital is how cohesive
as a community with someone else help you out in your area even if you’re not doing
well. Salt Lake City with the Mormon church, is
thought to be a classic example of a city with a lot of social capital, and correspondingly,
it’s a city with very high rates of upward mobility in our data. Now this concept of social capital was popularized
by Bob Putnam, a sociologist at Harvard, in a well known book called Bowling Alone. And the reason for the title of Bob’s book
is that he used the number of bowling alleys in an area or whether people were bowling
alone in particular as a proxy for social capital, which is a very hard thing to measure. Now, I was amazed to find in our own data
that the number of bowling alleys is actually very highly correlated with these differences
in upward mobility that I’ve been showing you. That actually turns out to be true. But the reason I mention it here is that it
also illustrates very well that everything that I’m showing you on the slide are correlations
rather then causal effects. So I think it’d be pretty surprising if the
policy lesson we should draw from this analysis is that we should build more bowling alleys
to try to increase upward mobility in America, right? So that shows you that it’s very important
to distinguish correlation from causation, and so the way you should think about these
factors that I’m showing you here is that they give us hints about where to look, but
they don’t necessarily tell you the exact recipe for how to create success in areas
that currently have low levels of mobility. The fifth and final factor that we identify
as you might expect intuitively is the quality of public schools in an area, places that
spend more on local public schools, places that have smaller classes, places that have
higher test scores tend to have higher levels of upward mobility. Now here, unlike with the other factors, unlike
with social capital, we think there’s very clear evidence that there are causal effects
of improving the quality of schools rather then just correlational evidence. And so I’m going to next show you what some
of that evidence looks like by talking about one specific way in which we might try to
improve the quality of public schools. I’m going to talk about the impacts of teachers
drawing upon some other research that we’ve done in our research group. So in this case, we’re using big data again
to study teachers long term impacts, and the way we do that is by linking data for all
kids who went to New York City Public Schools between 1989 and 2009 — so that’s about 2.5
million children who’ve written 18 million tests over that period. We take that data and link it to the federal
income tax returns that I was using for the previous analysis that I was showing you so
that we can look at things like how much kids are earning when they’re 30 years old, which
college they went to, whether they had a teenage birth and so forth. So by linking these two data sets, you can
ask a question like how does the quality of your third grade teacher effect how well — how
much you’re earning when you’re 30 years old. Now, in order to answer that question, the
first thing you need to do is define a measure of the quality of teachers that you’re going
to analyze, right? So if I want to show you that higher quality
teachers have an impact on kids long term success, I have to have some way to measure
the quality of teachers. Now there are many different ways in which
you might think about measuring teacher quality. I’m going to talk about one prominent measure
and the current policy debate what are called teacher value added measures or test score-based
measures of teacher effectiveness. The concept of teacher value added while there’s
some complicated statistical issues in the background really boils down to a very simple
question. We ask how much does a teacher raise his or
her students test scores on average? So for example, if I’m a fourth grade teacher,
we’d calculate my value added by taking my students test scores on average at the end
of fourth grade, minus their test scores at the beginning of fourth grade on average,
and if that number was very positive we’d say I’m a high value added teacher, and if
that number is very negative, we’d say I’m a low value added teacher. Now, the question we’re interested in answering
is how being assigned to a high value added teacher effects children’s long term outcomes. And so once again, in studying questions like
this, I find it useful to step back and ask what’s the experiment you would ideally conduct
as a scientist to answer that question? So what you’d like to do here is take a set
of teachers and use historical data on their students to estimate each of their value added
based on the method I just described. ^M00:40:02 And then you take a new batch of
kids and tell their parents — look, we want to answer this really important question of
whether teacher quality effects children’s long term outcomes. So we’re going to run an experiment with your
kids where we’re going to assign some of them to teachers we’ve identified as being low
value added and some as being low value added and some as being high value added, and let’s
see how your children do in 20 years. So you can imagine running that experiment
is extremely difficult in practice, and so now what I’m going to illustrate is how even
though we can’t really run that experiment in practice, we can essentially approximate
it pretty well using big data. So to describe what we do here, again I’m
going to use an example. We’re going to exploit the fact that there’s
a lot of turnover within schools. There are lots of teachers who enter or exit
a school and so let’s think about a specific school in New York City where we’re going
to track kids who get to fourth grade in different years. For example, children who arrive in fourth
grade in 1993, 1994, 1995 and so on. So these are different cohorts of kids. So what I’m going to do here is plot the average
test score of each of these different sets of kids, and you can see that they’re kind
of bouncing around the median around the 50th percentile. And then what happens at the end of the 1995
school year in this example is that a new teacher comes in who’s rated in the top five
percent of the value added distribution. So we’ve identified this teacher based on
historical data as being really excellent in terms of their value added. So lets see what happens to test scores right
when that teacher comes in. You can see that immediately test scores jump
up, and they continue to stay high as that teacher teaches subsequent cohorts of children. Okay? So this data suggests that the entry of that
high value added teacher really significant effects on how much students learn as measured
by their performance on standardized tests. Now, we all know that when we run an experiment
like from a middle school science class, you want to have a treatment group or something
changes, but you also want to have a control group so that you can make sure that nothing
else is contaminating your analysis, that nothing else is changing at the same time
that’s leading to this big jump in test scores in fourth grade. So very natural control group in this context
is to think about test scores in third grade because if the better teacher came in in fourth
grade, she shouldn’t have any impact on how well students are doing in third grade, right? And so if we plot test scores in third grade
over time, we see that in fact they’re totally flat around the entry of that teacher exactly
as you’d expect if the teacher’s having a causal effect on children’s learning. Now in this example, I’m focusing on an excellent
teacher entering the school, but this turns out to work totally symmetrically in the opposite
direction. So if you have this teacher who’s in the bottom
five percent of the value added distribution and doesn’t want to be there, he immediately
pulls down fourth grade test scores when he starts teaching relative to the prior grade. So this works completely symmetrically across
the distribution. So what these results show you is that teachers
have really significant impacts on how much their students learn as measured by standardized
tests, but now to tie this back to what we were talking about earlier, upward mobility,
what we’re ultimately interested in is not how students do on standardized tests. We’re interested in their long term outcomes. How much are they earning? Do they have a good job? Do they have stable family situation and so
on. And so we repeat exactly the type of analysis
that I’ve been showing you here, looking at long term outcomes, looking at earnings, looking
at college attendance, to see how based on teachers entering and exiting, what are their
effects on children’s long term performance. Let me summarize the impacts of that analysis
with this hypothetical policy exercise here. So let’s take the bell curve of teacher quality,
and suppose we identify the teachers who are in the bottom five percent of the distribution
of value added highlighted in yellow here. Suppose we either train those teachers or
hire a new set of teachers, letting these teachers go such that we bring the quality
of these teachers up to the average in the school district. So for instance, we let those teachers go
and hire a new set of teachers who are of average quality. What would the impacts of such a policy be? Our estimates based on the analysis that I
just showed you imply that this simple policy change of having a teacher of average quality
instead of quality in the bottom five percent for a single year would increase the lifetime
earnings of a child — the average child by $50,000, which translates to an earnings gain
of $1.4 million per classroom for a student of average size in the U.S., about 28 students. If you view that as an investment and discount
that back to present value with a five percent interest rate, bring it back to money at age
ten or 12 for instance, that’s worth about $250,000 of cash on hand. So you can see the teachers have incredibly
large impacts on students’ long term outcomes and can really have a profound effect on upward
mobility. More generally, there’s other research by
a number of other scholars that’s shown that other aspects of schools, for instance, certain
types of charter schools or the Harlem Children’s Zone program can have really significant effect
on children’s long term outcomes. And so I think education out of the five factors
that I listed is one area where we can really see from a concrete policy perspective that
we can have important effects on upward mobility in the long run. I want to show one final set of data that
approaches these issues from a different perspective, which thinks about equality from the — equality
of opportunity from the perspective of economic growth rather then principles of justice. So the traditional argument for greater social
mobility or greater equality of opportunity that many of us come to these issues with
is concern about justice, the idea that everyone should have a shot at the American dream no
matter what family they’re born into. But what I want to show you here is even if
you set aside those concerns about justice, and you’re just interested in maximizing GDP,
maximizing the size of the economic pie, you should still potentially be interested in
improving opportunities for upward mobility. To illustrate that idea, I’m going to focus
on one specific pathway to upward mobility, which is particularly relevant here in the
bay area, which is innovation. So in order to show you this analysis, what
we’re going to do here is study the lives of 750,000 patent holders in the U.S., and
the way we do that is by linking the universe of patent records to tax data so that we can
track the lives of — and inventors and see where inventors come from and how we might
be able to get more inventors in America. So let me start with this chart, which shows
you patent rates versus parent income, the probability that a child goes on to become
an inventor by his mid 30’s or so versus his or her parents income. The way this chart is constructed is that
on the x axis, on the horizontal axis is the parent’s household income percentile so there
are 100 dots here, one corresponding to each percentile of the distribution and what we’re
plotting is the number of kids who go on to have a patent by their mid 30’s in each of
these 100 groups. You can see that there’s an incredibly strong
relationship between your parent’s income and your probability of becoming an inventor. If you happen to be born to parents in the
top one percent of income distribution, you’re ten times as likely to become an inventor
as if you happened to be born to parents at the median of the income distribution. Now, why is there such a strong relationship
between your parent’s income and your probability of having a patent? One possibility is that it’s the type of issues
I’ve been talking about in this lecture. That it’s about differences in childhood environment
or schools or types of resources that you have access to while you’re growing up. A completely different possibility is that
this is about differences in ability. So maybe the children who are born to parents
in the top one percent, presumably their parents had to be talented to get to the top one percent
of the income distribution and ability is transmitted genetically across generations
so maybe that’s why they’re more likely to become inventors, have patents themselves. So to discriminate between those two very
different explanations, I’m going to turn back to that New York City test score data
that I described earlier as a measure ability early on in childhood. So this next chart here plots patent rates
versus third grade math test scores. So each dot here represents five percent of
the test score distribution. And you can see an interesting pattern, which
is if you’re below the 90th percentile of your third grade math class, you’re not very
likely to have a patent. And so that again shows you the strong predictive
power of these early childhood test scores, but then the probability of becoming an inventor
really shoots up at — if you’re at the top of your third grade math class, which I think
is intuitive in terms of how you think this might work. Now, more interestingly for the topic today,
suppose we split this chart up, looking at kids at low versus high income families separately. So this chart replicates the one that I just
showed you, but we’re breaking kids into two groups based on their parent’s income. The blue series is for kids from low and middle
income families, kids who have parents below the 80th percentile of the income distribution. The red series is for kids from high income
families, families that are above the 80th percentile in the top fifth of the income
distribution, and what you see is a really striking pattern, which is that high ability
kids, kids who are from the top five percent of their third grade math class are much more
likely to become inventors if they’re from high income families. If you’re from a low income family that is
in the blue series here, your odds of becoming an inventor aren’t all that much higher if
you’re from — if you’re in the top five percent of your third grade math class. ^M00:50:37 So to put it differently, what
these data show you is that in America, you need two things in order to become an inventor. You need to be smart as measured by your test
scores, for instance, early in childhood. And you need to be from a rich family. And that, I think, gives you a very different
perspective on equality of opportunity. It suggests that if we can bring more of these
kids who are in the blue part of this chart, these smart kids from lower income families,
if we can bring them through the innovation pathway and help them become inventors, that
will not only enhance their own incomes and increase their prospects of upper mobility,
it will also help the rest of us because they might end up discovering the next blockbluster
drug or the next iphone that’s going to change everyone’s life and increase GDP overall. So I think even setting aside concerns about
justice, we should all be interested potentially in improving equality of opportunity. Now, how can we actually go about trying to
get more kids from low income families to become inventors? I’m going to show one final set of data on
this shown — that ties back to the geographic variation that I started out with. So this map here shows you the origins of
inventors in America. We take each of those 740 areas that I showed
you before and calculate the fraction of kids who grow up in each of those areas who go
on to become inventors, who go on to have a patent. And the map is colored so that the red areas
are the hot spots of where inventors come from. The red areas are the places that generate
more kids who go on to have patents. So you can see here a very clear pattern which
is if you grow up in an area that is a hub of innovation — for example, look at the
bay area, that’s where you see the reddest colors, the greatest number of kids who grow
up to become inventors. If you grow up in the northeast, for instance,
you’re also much more likely to become inventors. Another striking example, look at Texas, which
in general has pretty low rates of kids going on to become inventors with one example, with
one exception essentially right in the middle that’s Austin, Texas, growing up around again
the university innovation hub. So that data suggests that again this is something
about exposure. If you grow up in an innovative area, you’re
more likely to become an inventor yourself. What’s striking is that that pattern holds
even more sharply across the type of invention that you have, so inventions are classified
into different technology classes, computers versus biological patents for instance or
the specific type of computing equipment that you invent, what type of semiconductor did
you develop and so forth? And what we find is the exact area in which
you become an inventor is highly influenced by the environment which you grow up. So let me give you an example of this. Let’s say you take two kids who currently
live in Boston. Say one of them grew up in the bay area, which
has a lot of computer innovation, and let’s say another grew up in Minneapolis, which
happens to have a lot of medical device manufacturers. It turns out that if you take these two kids
in Boston, the kid who grew up in Minneapolis is much more likely to have a medical device
patent, and the kid who grew up in the bay area is much more likely to have a computer
patent. That type of result holds really systematically
in the data, and it gives you a little bit of insight into what might be going on in
driving that big gap in innovation between low income and high income kids. Kids from high income families are exposed
to lots of different opportunities through their parents or through their networks, through
their neighborhoods, internships, science programs and so forth. Kids from lower income backgrounds typically
don’t have any of those exposures, and the importance of exposure as seen in this map
suggests that that can be a critical pathway in generating this big gap in innovation by
socioeconomic status. Also suggests that if we can increase opportunities
for exposure for kids from lower income backgrounds through targeted internship programs or outreach
by Silicon Valley companies, for example, that can really have a profound effect on
those kids’ lives and on aggregate levels of innovation more generally. So let me conclude by talking about some policy
lessons that emerge from this body of work. The first lesson, I think, while we often
talk about issues of inequality and the American dream at a national level, we should really
be thinking about tackling upward mobility at a local level, both by investing and improving
opportunity in areas that currently offer less upper mobility, cities like Atlanta or
Baltimore places like Oakland, for instance. But also by helping low income families with
young children move to higher opportunity areas. And in particular, I think we should be focusing
on encouraging moves to areas that one might call opportunity bargains, which is something
we’re focusing on in our current research. So to illustrate how this works, let me show
you this chart here, which shows some preliminary data for New York where we’re constructing
these measures of opportunity for every zip code in New York. So on the y axis is an opportunity index,
the effect on a child’s earnings of growing up in a given zip code. So at the top, if you grow up in that zip
code, you earn 20 percent more than the average place. And if you — at the place at the very bottom,
like the Martin Luther King Towers that we talked about earlier, you lose 20 percent
relative to growing up in the average place where plotting that data on opportunity versus
the average monthly rent in the zip code. So as you might expect intuitively, the places
that offer better opportunities for kids tend to be more expensive. So to take one example, Park Slope, which
is where Mayor De Blasio used to live is a very expensive part of New York, and kids
do quite well there. But interestingly, it turns out there are
a lot of opportunity bargains to be found. So take a look, for instance, at Rock Away
Park or East Midwood, other neighborhoods in New York, they offer almost the same level
of opportunity by — as judged by our data as Park Slope. But they cost about half as much as living
in Park Slope. And so by focusing on these opportunity bargains
and helping families with housing vouchers move to those areas, which is a project we’re
currently working on on a large scale with the Housing and Urban Development agency and
various public housing authorities, we think we can really improve the prospects of kids
from low income families. So that’s the first broad lesson I think. Think local. We can actually all make a difference in our
own local communities on these important problems. The second lesson is that we should focus,
I think, on improving childhood environments because that really seems to be the critical
driver of these differences in opportunity. This is not just about spending more money,
so often I think economists think if we have a problem like schools that are not of good
quality, we should increase funding for schools. So obviously money can be helpful. But it’s important to note that the U.S. already
spends more on schools than many other countries, for example, Finland or other Scandinavian
countries that have much better outcomes. So instead, my view is that we should focus
on key inputs like improving the quality of teachers or expanding certain high performance
charter schools. Think about how we can spend the money. We’re already spending more smartly rather
then just increasing the total budget, which is difficult in a time of tight government
budget constraints. And finally, I think at the broadest level,
as I hope I’ve illustrated here, harnessing big data to develop a scientific evidence
base for economic and social policy can be quite valuable. So all of the statistics that I’ve been talking
about today are publicly available on a website called equalityofopportunity.org, which any
of you can go look at and use for followup research projects or to study how your own
community is doing. And our hope is by constructing data like
this, local mayors and cities and other interested advocates will be able to monitor trends in
mobility and study which policies work. Not just in improving opportunity in cities
that currently don’t look so good, but also I would stress in maintaining opportunity
in places like the bay area, which in the 1980’s looked really good in terms of rates
of upward mobility. That’s what we saw in the maps that I was
showing you, but it’s not clear that that will be the case for the current generation. So I want to end on that point. So many people ask, “How does the bay area
look for the current generation of kids in terms of upward mobility?” It’s difficult to answer tht question directly,
right? Because we’re not going to see the earnings
of kids who are currently growing up in the bay area for another 20 years or so or until
they start working. However, one thing that you can do, which
is an exercise that was done by journalists at the Atlantic magazine, an interesting article
recently using our data is to ask, “What would you predict upward mobility will look like
in the bay area based on how those five key factors that I described are trending in this
community? So things like segregation, income inequality,
levels of social capital, quality of schools. We can measure all of those things using current
data, not data from the 1980’s but data from the past few years. ^M01:00:06 And you can ask given how things
are changing in the bay area, how would you expect upward mobility to look for the current
generation of kids? And the answer that they reached is summarized
by the title of this article, which is the place where the poor ones thrived, which is
what they called the bay area. They point out that because the bay area has
increasing segregation, higher levels of income inequality, potentially declining quality
of public schools. All of these factors suggest that for the
current generation of kids, the bay area is no longer going to be the land of opportunity,
and I think that’s very worrisome and shows that all of us should be thinking seriously
about these problems even in an area that traditionally was really a land of opportunity. And so I think the tools that we’ve identified
here and hope to identify in ongoing research will hopefully point to additional directions
through which we can revive the American dream. Thanks very much. ^M01:01:00
[ Applause ] ^M01:01:17
>>Thank you, Professor Chetty. You gave us many things to think about here. At this point, we’ll take some questions and
answer them. I think Dr. Kakar set it up so that there’s
two microphones. So if you have questions, if you could give
them — they’re on the mic, and then we’ll be able to do this. Thank you.>>I was wondering on the map of Sacramento
if you were to do an overlay of income instead of race, what that map of Sacramento would
look like?>>Dr. Raj Chetty: Yeah. Very good question. So if you were to look at segregation by income
instead of segregation by race, what are the patterns? The reason I showed you segregation by race
there is that segregation by income actually looks quite similar. So cities that are very segregated by race
tend to be very segregated by income as well. That’s not one for one true in every single
city, but there’s a very strong association between those two things. So Atlanta is much more segregated both by
income — low income people in Atlanta live in very different parts of the city then high
income people then Sacramento where there are more mixed income neighborhoods. It’s a little bit harder to measure income
segregation then racial segregation, which is why people focus on racial segregation
measures, but you see very similar patterns when you have the data.>>So my question is you focused a lot on
the availability of opportunity as the measure for mobility, and there seems to be a lot
in the news lately about the availability of opportunity for millennials and their ability
to achieve that opportunity particularly in housing markets. So I’m curious if there’s any insights in
the data about how the availability of opportunity has actually trended across different generations?>>Dr. Raj Chetty: So as best we can tell,
and the data are somewhat limited here historically because we don’t have the type of huge data
sets that we’re using here with millions of kids going far back in time, but as best we
can tell in many places the availability of opportunity is falling over time as this article
kind of suggests, right? That you have growing segregation. You have growing inequality. The various factors that I was talking about
suggest that we’re going to have declining levels of opportunity for subsequent — for
the current generation and subsequent generations. And as a result, you’re going to have more
chronic persistence of poverty across generations. And so that’s why I think it’s important to
think hard about what types of levers can be used to push in the opposite direction
as we have growing segregation? How can we create more affordable housing
so we have mixed income neighborhoods? If we think that colleges, community colleges
and state colleges can play a big role in upward mobility, how can we preserve those
institutions so that kids of the next generation continue to have the same opportunities? So I think the trends are not necessarily
in a good direction. But one shouldn’t lose hope because there
are things one can do to [inaudible] that trend.>>Thank you.>>This is more of a follow up to the speaker
before me on this side. Since you mentioned before that segregation
and social status has a large factor in terms of how far one can move socially, and then
you mentioned how teachers like you recommend putting teachers into neighborhoods, but — and
I’ve been told in multiple sociology classes that you are most likely going to be in the
same social class as your parents, which is consistent with what you’ve been saying. And you’ve also said that you would recommend
spreading teachers around and to follow up with that, how would you — since you would
say to improve teacher equality all around, how exactly would you do that? Would you prefer putting teachers in say a
neighborhood that’s like less segregated, which you say has more chance of a social
mobility or would you prefer just putting in a place with minimal high segregation but
minimal social mobility? What would you think is best?>>Dr. Raj Chetty: Good question. I mean, I think — well there are — I think
there are a couple of elements of your question. So first, how can you improve the quality
of the teaching workforce in the U.S. both in segregated and less segregated areas? How do you go about doing that? So I think one of the ways to think about
the problem is that teaching is not a profession in America that many people aspire to go into. I think the way it’s structured despite its
importance is not ideally suited towards drawing and retaining the most talented people who
I think really should be going into teaching given the profound effects that teaching has
on so many people. And so I think part of that has to do with
salaries. Part of that has to do with prestige. So if you look at countries like Finland that
have the best education systems, what you hear is a child who’s doing really well in
school in Finland, their aspiration is to become a teacher because that’s really viewed
as the best job you can get. And that’s not the case in most American cities,
and I think that’s something we really need to fix. Now, if you have a good set of teachers, do
you want to put them in the most segregated areas or the less segregated areas? That’s a difficult question to answer. My instinct would be you want to tackle — you
want to try to improve conditions into places where we have the biggest challenges. So in these cities with very high levels of
concentrated poverty, which often have the schools that are struggling the most, I think
having effective teachers, which is what, for instance, the Teach for America program
tries to do, can be really effective. I think the challenge with Teach for America
is we have excellent teachers in the classroom for a couple of years, but they don’t end
up staying for many years. And that is what we need. We need teachers who continue to want to stay
in these high poverty urban schools so that they can help kids do well in the long run. So that’s where I would try to emphasize improvements
in teacher equality.>>So you mentioned [inaudible] his question,
but you mentioned [inaudible] essential for [inaudible] success and for mobility around
the housing area. So I think [inaudible] was proposed [inaudible],
which was failed to put in practice in the states. What would you think would be a better to
— policy that will be potential to propose in the future that can ensure equality of
the education?>>Dr. Raj Chetty: As I was saying, I think
early childhood education can be very valuable, but these data show that improving children’s
environment not just in the earliest years in childhood but throughout childhood and
even in the college years can be very valuable. And so I think providing greater access to
high quality education and high quality childhood environments, it’s not I think purely about
education. It’s also about where you live. If you live in an area where the people around
you are involved in crime or you don’t see a lot of opportunity, I think that can really
influence a child’s perceptions and influence their long term outcomes. I don’t think it’s just about tax credits
or specific investments purely at the early childhood level. I think it’s a broader shift that’s required
in environment and education throughout childhood. How you’d go about doing that, I think, comes
back to questions like how do you improve the quality of teachers? How do you design more integrated neighborhoods? How can you create more affordable housing
so that you don’t have some people living in one part of the city and other people living
in different parts. I think I would take a more — solution that
has many different elements rather then one single tax credit or try to tackle the problem
in one specific way.>>Hi. Thanks for the wonderful lecture. It’s really fascinating. I really gained a lot. My name is Chen. I’m an econ grad student here. I’m also international student. So I came to San Francisco three years ago,
and because of the more opportunities here and [inaudible] education level, so I want
— and you’re talking about upper mobility at local level and national level. So I want to ask you what do you think about
upper mobility at an international level?>>Dr. Raj Chetty: And so do you mean at an
international level in terms of immigrants or do you mean how does upward mobility compare
in the U.S. to other countries?>>Probably more like immigrants or international
students — how they perform better before — ?
>>Dr. Raj Chetty: That’s a great question. I think another important aspect of the American
dream that attracts a lot of people to this country including my own parents is the idea
that you can have much greater opportunities an immigrant here then elsewhere. It’s difficult for us to study that directly
in the data because what we’re doing is comparing kids incomes to their parents incomes, and
we can’t see parents incomes if the parents are not in the United States quite naturally,
and the tax data that we’re using we only see parents incomes if the parents are also
in the U.S. So actually in the analysis that I was showing
you for that reason, we restrict the sample to U.S. citizens. So everything I was showing you was for citizens,
but my intuition — this is not based on data, but just my instinct — is that you’d see
very similar patterns for immigrants. So I think immigrants who come to the United
States do so rightly at least traditionally because the U.S. really does offer great opportunities
relative to other countries. ^M01:11:11 There are other studies, for instance,
that can track immigrants better using Swedish data, for example. Sweden also has a lot of immigrants showing
that immigrants who come to Sweden do dramatically better particularly if they get there at younger
ages. And so I think all of these issues likely
apply to immigrants in the same way that we were talking about here. It’s just that we can’t directly analyze immigrants
in the data we have so far.>>So in your earlier slides, you were showing
that — you were showing that numbers — I was assuming the numbers were based off of
pure income.>>Dr. Raj Chetty: Yes.>>So you were showing that if you say grew
up in Oakland and then at some point moved to San Francisco, and would you then be staying
in San Francisco for the lifetime of the data?>>Dr. Raj Chetty: Not necessarily.>>Okay. Because the thought I had was incomes are
different in different areas but also cost of living is different.>>Dr. Raj Chetty: Good question. Absolutely. Cost of living couldn’t matter. It’s important to note that when we talk about
children’s locations, we’re talking about their location where they grow up, which is
not necessarily the location where they’re living in adulthood. So if you grew up in Oakland or San Francisco,
certainly within the bay area people move around a fair bit. But more generally, take the center of the
country where we saw really high rates of upward mobility for kids who grew up in Iowa,
for instance, the kids who are really succeeding in Iowa, lots of those kids are living in
Chicago or New York when we’re measuring their incomes in adulthood because lots of the most
successful kids end up moving elsewhere to get that higher paying job. And so I think that actually illustrates a
really interesting pattern in places like Iowa, which is that they are extremely good
in terms of rates of upward mobility despite having kind of a brain drain problem where
the most talented successful people are constantly moving out, new people are moving in yet they
continue to generate these fantastic outcomes. So figuring out what’s going on in those places,
I think, is really important for exactly that reason. But yeah, I think it’s not just driven by
local differences in cost of living because this is about where you grow up, not where
you’re working in adulthood.>>So you mentioned a couple ways to improve
people’s social mobility aspects by either moving them or investing in the neighborhood. How would moving them to a place like San
Francisco even work because in your example we talked about Harlem to Bronx and real estate,
gas prices, property, grocery prices are not even the same when we look at San Francisco. So how would the voucher system work here?>>Dr. Raj Chetty: Right. So affordability is clearly an issue. I think that’s why it comes back to this slide
that I was showing here where you’ve got to think about not moving people to the most
expensive parts of San Francisco or New York, moving people to Park Slope, which is not
practical. But rather identifying what we’re calling
these opportunity bargains, places that would be affordable to families with housing vouchers
even though vouchers are never going to cover the full cost of an apartment in a prime part
of San Francisco. So I think San Francisco’s a particularly
challenging example because for most of the city, it’s not really plausible that a family
with a voucher would be able to rent here. There are outlying areas maybe where you’d
see bargains like this where you have pretty good outcomes for kids that are more affordable. One of the things we’re hoping to do with
this very large data is identify precisely what are those places where families with
vouchers can move that are kind of these hidden gems, right? So everyone knows that if you move to the
place with the best school district. In New York, San Francisco, it’s great for
kids and it’s a great place to live, but it’s also totally unaffordable. So that’s not all that useful exactly as you
say. But identifying these bargains, I think, is
really where we can make a difference . And my view is given that we already spend about
$40 billion on programs like this, we might as well use that money more effectively to
try to help families get to those areas. So I think moving is by no means the only
solution, but given our current policy framework, it’s something we should be thinking about
doing more intelligently.>>Hi. You alluded to social capital as a really
important factor and also to the difficulty of measuring it. I was wondering if among your colleagues or
in the literature if you’ve kind of seen any promising ways that you might actually kind
of solve that problem and measure it. And particularly, I was kind of thinking about
in this era we live lin, public social media data, all these things about someone has a
network. These people are their friends, etc.>>Dr. Raj Chetty: Excellent question, and
you’ve anticipated exactly the path we’re pursuing to try to study that question more
carefully, which is to use data on networks. So we’re starting a collaboration with Facebook
to understand the role of networks in social mobility more directly, and our sense is networks
might give us greater insight in a more grandular level into exactly how the social capital
or role model pure effects are working in practice. So the logic of the project we’re starting
is to stay take a city like Atlanta versus San Francisco. Does the structure of the friendship network
in Atlanta look very different then the structure of the friendship network in the bay area
or in Salt Lake City, places that have higher levels of mobility. In particular, do you see more low income
kids having friends from higher income backgrounds? And is that integration in terms of across
social groups, does that seem to be a key mediator of children’s later success? We’re also interested in understanding the
determinants of networks and social capital using these data. So one of the hypotheses, if you think back
to the moving to opportunity experiment that I described where the kids who moved at young
ages from Harlem to the Bronx did much better than the kids that moved at older ages. One prominent hypothesis for that is that
the kids who moved at older ages didn’t really change their friendship or their social group
at all. They basically took the subway back to Harlem
and hung out with their old friends and nothing really changed. So we’ll be able to test that directly by
using Facebook data today to see whether the friendship networks or the kids who move at
young ages versus older ages look very different. So that gives you a sense of how we’re trying
to dig into these issues with modern data more precisely.>>Hi there. Good evening. Thanks a lot for giving your talk tonight. I wast wondering — unfortunately, the wealthy
do have a lot more decision-making power than the poor, and as you mentioned, they do have
a concern with mixed neighborhood housing vouchers that they might lose their slice
of the pie. And you mentioned that it may in fact be more
beneficial for the wealthy as well. I can totally understand how that may be the
case intuitively. I’m wondering how you came to that conclusion. Was it intuitive or did that bear itself out
in the data as well? Thank you.>>Dr. Raj Chetty: Yes. So we reached that conclusion empirically
in the data in the following way. So the maps that I was showing you were focused
on kids from low income families. What did their outcomes look like based on
where they grew up? So we could analogously draw the same type
of map for kids growing up in high income famillies. Can say how well the high income kids in Atlanta
do and how well the high income kids in San Francisco do, etc. And so as I pointed out, if you’re in a more
mixed income area, we see the kids from the low income families doing better. Analogously now if we look at the data for
the kids growing up in high income families, we see that they don’t do any worse in those
areas. If anything, they do slightly better. So in practice when you look at the data,
it is not true empirically that in more mixed income areas you see children from higher
income families doing worse. Now, that doesn’t necessarily rule out the
possibility that if we have some radical policy change that completely shifts the structure
of cities we could get very different outcomes, but the data we have don’t point in the direction
that higher income families should be incredibly concerned that their kids are going to do
much worse if we have more integrated neighborhoods.>>Thank you.>>Hi. Thank you for your time tonight. And thank you for coming here. I learned a lot tonight. My question is regarding your mentioning that
there are two main ways to change people’s mobility. ^M01:20:04 One is to move them and one is
to improve neighborhoods. Regarding the initiative of improving neighborhoods,
my question is what’s the best way to improve a neighborhood but keep it affordable for
the people living there? And my reasoning is that if the neighborhood
is improved, then the demand for that neighborhood goes up, and more people want to move there. And so how do we address that problem?>>Dr. Raj Chetty: Excellent question. I mean, your question gets to key issues of
gentrification and how when you improve a given area how do you maintain opportunity? Maintain affordability for the families that
are already there? I don’t have a concrete answer to that beyond
saying that again if we have more integrated neighborhoods where you — you have zoning
laws. Take the example of zoning, which I think
illustrates this concretely. Some communities that have excellent schools,
for example, impose things like minimum lot size requirements that basically serve to
keep out lower income families that don’t want to have the one acre of land and can’t
afford that. And so those types of laws serve to increase
segregation and create — and make it hard for lower income families to benefit from
things like better public schools. So as you improve the quality of public schools,
if you’ve retained the option to have lower income, high density housing in a given area,
for instance, by not allowing zoning laws that prevent the building of such property,
then I think you can potentially continue to maintain affordability and opportunity. So I think one has to again think about a
combination of different policy changes while you’re changing the school system thinking
about zoning laws at the same time to tackle problems like the one you described.>>I just have a follow up. So I believe that there’s a policy in San
Francisco where forever square footage of luxury housing built, I believe that the developer’s
responsible for at least partially subsidizing some kind of lower income housing project. So is that something that is a good idea?>>Dr. Raj Chetty: Yeah. I mean, I think one of the key challenges
San Francisco faces in particular — we just had a forum this morning at Stanford on housing
policy in the bay area, and the big challenge in the bay area is just purely on the supply
side. There are insufficient housing units in the
bay area given the tremendous job growth here, and that’s what’s driving up house prices
so much. And so I think things like credits for developers
to build affordable housing can be helpful, but it’s also very important that the city
actually permit more building. So at the end of the day, there are regulations
in San Francisco that basically prevent builders from building any kind of housing in great
quantities, restrictions on how tall buildigs can be and how many new permits can be issued
and things like that. And so I think it’s very important to think
not just about subsidies but just allowing more building in a city where there’s so much
job growth is critical, I think.>>Thank you very much.>>One last question.>>Thank you again. The lecture was amazing. Hopefully, I can close this out with a twofer
question. One is a methodology question. In the party [inaudible] the data where you
looked at the movement from city to city from below probabiltiy to a high probability of
mobility city, were you able to control for some of the factors that you identified as
being correlated with high mobility cities between those cities? So when you look — you looked at one city,
and they may have had the same integration factor or — that’s the first question. The other is to get your thoughts on the use
of big data. Obviously, this whole area and your whole
study is based on the use of big data and the analysis. There is kind of the dark side of the force
possibility of using big data in many areas where the algorithms that are used develop
profiles on people that are used for credit scores, are used for profiling for criminal
activity, etc., and I’m wondering what your thoughts are on the use of data and how it
should be used and how particular algorithms that impact people’s lives and their potential
upward mobility should be used in society?>>Dr. Raj Chetty: Yes. Thanks. Those are great questions. So on the first question, we didn’t think
about approaching the mover’s analysis exactly in that way because what we were kind of interested
in understanding is if you move let’s say from a more segregated area to a more integrated
area, what are the effects on your own children’s outcomes, right? So your question is if we take two places
with the same level of segregation, same level of all thees other factors, what ends up happening
if a child moves from what still appears to be a better area or worse area? An interesting idea. We haven’t looked at that directly. One thing that I can tell you is the amount
of variation across these areas once you control for those factors actually becomes quite small. That is, those five factors explain more than
85 percent of the variance in outcomes across areas so the vast majority of the story here
is really in those five factors, not in the part that’s left. And so that’s the reason we didn’t think about
controlling for those factors so much as understanding what happens when you move to a place that
looks different on those five dimensions. On your second question on big data, I mean,
big data is — all the talk as you all know, both in academia and in the private factor
at Stanford, there are all kinds of courses on big data, new methods being developed and
so forth. I think you highlight an important trade off
in the use of big data. I think in everything that I’ve been showing
you this is not so much about an individual person. I was showing you data aggregating over thousands,
if not millions of people, to understand aggregate patterns, stories that we could draw out that
matter for all of us. Not so much singling out and profiling any
one person. And so I think the use of big data in the
private sector sometimes goes in a very different direction of really trying to figure out how
we can market things in a better way or screen people for jobs and things like that. And I agree with you. Some of those approaches can actually have
a pernicious negative effect precisely for the reasons that I’ve been talking about here. They could potentially lead to further discrimination
or basically further separation across groups that exacerbate some of these problems of
inequality and a lack of opportunity for certain groups. So I think like any other tool, you can’t
view big data as just a [inaudible] that’s going to improve things universally. I think it’s about how do you use that tool? Do you use it to try to do better things? I think one of the very important uses of
these data can be to implement better social and economic policy. My view is that we’re spending billions if
not trillions of dollars on government programs to try to combat problems like poverty, increase
opportunity and so forth. But we’re doing that with incredibly limited
evidence. And when you step back and think about it
from like a scientific perspective, it’s incredible how little evidence we have on what actually
works relative to the amount of money we spend on these problems. And to me, that’s really the potential of
big data to shed light on what actually works and what doesn’t, and I think if we can use
big data in that way, we can all benefit greatly from it despite the concerns one might have
about confidentiality and profiling and so forth, which I think if handled in the right
way can be addressed while still reaping the benefits of this data for — in terms of knowledge.>>I want to thank Professor Chetty again. One of our economics association students,
Esteban, is going to give us a token thing in Asian culture. You always have to have a gift. So we’re going to leave you with this one
gift. To the audience, thank you very much. Have a good evening and a safe trip home.

Leave a Reply

Your email address will not be published. Required fields are marked *