I've devoted a lot of effort to extracting from Raj Chetty's prodigious number crunching the most interesting findings, ones that he appears to tend to obfuscate to keep from getting cancelled before getting his well-deserved Nobel. For example, here's my write-up of his 2019 paper “Race and Economic Opportunity in the United States: An Intergenerational Perspective.”
For both races [white and black], incarceration rates fall steadily with increasing affluence of upbringing. Among whites, only 0.2 percent of sons of the One Percent were in the slammer. Among blacks, the lowest percentage (1.6 percent) is found in the 98th percentile, before the incarceration rate rises in the two highest-income percentiles.
This curious anomaly could just be due to statistical noise. In the top two percentiles of upbringing reckoned across all races, there were only about 2,100 youngish black men altogether and roughly 45 of them were under lock and key.
Or it could be that this third-highest percentile is the most bourgeois among blacks, featuring, say, partners in law and CPA firms, while the top two percentiles are more loaded with black jocks and entertainers, whose sons tend to be more of a handful.
With that minor exception, why do richer kids wind up in jail less often? There are no doubt numerous reasons of nurture and nature, ranging from the wealthy being able to afford better defense attorneys, to neighborhoods without youth gangs to ensnare your son into a life of crime being more expensive, and on to genetics. In general, being wealthy is good, and you should strive for it for the sake of your kids.
The right vertical axis denotes the ratio (red line) of the black percentage incarcerated divided by the white. Unexpectedly, it rises steadily with childhood affluence.
Among men raised in the dirt-poor first percentile, blacks are 3.3 times as likely to be imprisoned.
At the 25th percentile, blacks are confined 3.9 times as often.
At the 50th percentile, the ratio is 4.5 to one and at the 75th percentile it’s 5.0.
At the 98th percentile, the ratio is 6.7, before exploding to 10.7 at the 100th.
The median black household income falls around the 27th to 28th percentile nationally, where black men are locked up 4.0 times as often as white men. So that’s probably the best summary statistic: All else being equal in terms of household income during adolescence, black men are four times as likely to find themselves behind bars as white men.
That’s a huge disparity.
For instance, black men at the 98th percentile of upbringing, the best-behaved black cohort, are jailed as often as white men at the 50th percentile. Similarly, the black rate at the national median of income is 7.2 percent, a little higher than the white rate at the single lowest percentile.
That suggests that there is approximately a two standard deviation difference in racial propensity to be prison-bound even when controlling for affluence when young.
In the social sciences, a one standard deviation difference, such as in IQ, is very large. Two is almost unheard of. Two standard deviations after adjusting for childhood income is off the charts.
Why does the black-to-white ratio get steadily worse with higher income?
I don’t know. Before seeing Chetty’s data, I might have guessed it shrank.
Is the cause racism?
Well, if it is, racism doesn’t much hinder black women. They appear to be incarcerated only about 30 percent more often than white women raised with the same family income, not 300 percent more often as with black men.
Excellent summary. I'm not an economist, but I've been writing a series of appreciations / critiques of Chetty's big studies over the last dozen years. I focus on using examples to illustrate Chetty's strengths and weaknesses. For example, here is my in-depth 2025 analysis of his study of what were the best counties for blue collar families to raise their kids in:
I would certainly agree that idiosyncratic booms and busts may very well be the most important thing in determining intergenerational mobility. But why would this random error bias the coefficient of some particular factor on outcomes? I obviously agree that ranking counties, and expecting those rankings to hold indefinitely, is silly, but you can still extrapolate from the characteristics of a region what your likely outcomes would be.
Chetty's big finding in 2015 was that the best county to grow up in the bottom half of parental income in the 1990s for having a higher income in 2010-11 around age 30 was Sioux County, Iowa, a nearly all white farm county with excellent soil and a very conservative culture (it's the home to the Dutch Reformed Church in the U.S.).
Most of the other top 25 most upwardly mobile counties were similarly white rural counties in the northern Great Plains.
The bottom 25 counties for blue collar kids were mostly heavily minority. The absolute worst was the Pine Ridge Sioux reservation in South Dakota, which has been among the most tragic places in the United States since Wounded Knee in 1890.
This is because blacks, American Indians, and Alaska Natives tend to regress toward lower mean incomes than whites.
The one mostly white county in the bottom 25 was Horry County, SC, home to the immense Myrtle Beach golf resort. Myrtle Beach was booming in the 1990s, but then a huge recession hit the golf industry from 2001 until covid brought golf back into fashion. Not surprisingly, Myrtle Beach adolescents whose parents were prospering constructing golf courses and condos in the 1990s were struggling in 2010.
Similarly, Chetty got a lot of publicity for reporting in 2013 that Charlotte, North Carolina was the worst big metro area in the country for upward mobility. But a recent study by him finds it got much better lately.
The obvious explanation is that Charlotte prospered in 1994-2000 due to home building. Charlotte was one of the hubs for mortage banking (e.g., Wachovia Bank), for home construction, for golf course second homes, for lumber, and for furniture making. The collapse of the housing bubble in 2008 left young people in Charlotte really badly off. But, over time, Charlotte's economy has improved and young people in Charlotte now enjoy more upward mobility than they did at the depths of the Great Recession.
Bizarrely, Chetty is not averse to taking some credit for Charlotte recovering from the bursting of the Housing Bubble, as I pointed out last year:
By the way, another reason the northern Great Plains did so well for blue collar upward mobility from the 1990s to 2010-2011 in Chetty's data was the unexpected development of fracking technology that set off a huge energy boom in North Dakota, which brought in blue collar workers from surrounding states who were tough enough to work out doors in the oilfields in winter.
Fantastic overview which happens to gloss over the most important question, one of political economy. How did he manage to get access to all of this data?
Is he missing a first born child? Is his middle name Faust?
"Well, as you might imagine, Michael, it was one step at a time, and there were many steps to goin the journey – a long and winding path. It took us, I would say, the better part of five or six years. I mean, I've lost track at this point from beginning to end. Part of it was just an enormous effort of collaboration with partners and data merging and analysis. We spent a year or more crisscrossing the country, trying to convince college presidents and admissions officers and other people in universities to share data with us and to partner with us more broadly. We shared some data back with them and so on. And that required a lot of work, but I hope it was worth the effort. I mean, we lost some time to the pandemic as many people did, and at times I thought, maybe with this research, we'd never see the light of day, but it finally did. And for that, I'm very grateful both to all the people on the research team, including my co-authors, and our research assistants, but also, importantly, our partner institutions, without whom this research would not be possible."
David Deming: Sure. Yeah, Ralph. Well, it was quite a process as you said. We started this project in 2017, so six years ago, and we started sort of one school or one university system at a time asking people if they were interested in participating. Of course, as you might imagine, a lot of people said no, some people said yes. And then having worked with some early partners like the University of Texas, Austin, and the University of California system and other schools like that, once we got a few people saying yes and showed that we were serious, we had a conference called the Climb Conference also in 2017, which led to a surge of interest and some partners. And then it was kind of like a snowball rolling downhill. I mean, again, we didn't get universal participation, but we spent a lot of time shaking hands, knocking on doors, whatever metaphors you want to use. And then eventually we ended up recruiting, I think, more than 400 colleges and universities encompassing about three and a half million students per year. It's about 15% of the undergraduate population in the US. So we ended up with quite a few partners and we're really grateful to them. Without them, this research wouldn't be possible.
Raj Chetty: And just to add a bit more color on how we got into this, a lot of this data collection started after an earlier paper we had released together with our collaborator, John Friedman at Brown University, on just measuring rates of economic mobility across colleges. And that was not using internal college admissions data, it was using data from federal tax records in the Department of Education to measure the parental income distributions at every college and how kids did after college. And so we were able to put those statistics out publicly and lots of people were able to see how their college stacked up relative to other colleges. And that led to a lot of interest in understanding how we can create more mobility at our colleges. And so that naturally led to springboard to have the types of conversations that Dave described where we were able to say, "If you really want to answer that question, we need more data," which is what researchers usually say in order to go deeper.
I've devoted a lot of effort to extracting from Raj Chetty's prodigious number crunching the most interesting findings, ones that he appears to tend to obfuscate to keep from getting cancelled before getting his well-deserved Nobel. For example, here's my write-up of his 2019 paper “Race and Economic Opportunity in the United States: An Intergenerational Perspective.”
For both races [white and black], incarceration rates fall steadily with increasing affluence of upbringing. Among whites, only 0.2 percent of sons of the One Percent were in the slammer. Among blacks, the lowest percentage (1.6 percent) is found in the 98th percentile, before the incarceration rate rises in the two highest-income percentiles.
This curious anomaly could just be due to statistical noise. In the top two percentiles of upbringing reckoned across all races, there were only about 2,100 youngish black men altogether and roughly 45 of them were under lock and key.
Or it could be that this third-highest percentile is the most bourgeois among blacks, featuring, say, partners in law and CPA firms, while the top two percentiles are more loaded with black jocks and entertainers, whose sons tend to be more of a handful.
With that minor exception, why do richer kids wind up in jail less often? There are no doubt numerous reasons of nurture and nature, ranging from the wealthy being able to afford better defense attorneys, to neighborhoods without youth gangs to ensnare your son into a life of crime being more expensive, and on to genetics. In general, being wealthy is good, and you should strive for it for the sake of your kids.
The right vertical axis denotes the ratio (red line) of the black percentage incarcerated divided by the white. Unexpectedly, it rises steadily with childhood affluence.
Among men raised in the dirt-poor first percentile, blacks are 3.3 times as likely to be imprisoned.
At the 25th percentile, blacks are confined 3.9 times as often.
At the 50th percentile, the ratio is 4.5 to one and at the 75th percentile it’s 5.0.
At the 98th percentile, the ratio is 6.7, before exploding to 10.7 at the 100th.
The median black household income falls around the 27th to 28th percentile nationally, where black men are locked up 4.0 times as often as white men. So that’s probably the best summary statistic: All else being equal in terms of household income during adolescence, black men are four times as likely to find themselves behind bars as white men.
That’s a huge disparity.
For instance, black men at the 98th percentile of upbringing, the best-behaved black cohort, are jailed as often as white men at the 50th percentile. Similarly, the black rate at the national median of income is 7.2 percent, a little higher than the white rate at the single lowest percentile.
That suggests that there is approximately a two standard deviation difference in racial propensity to be prison-bound even when controlling for affluence when young.
In the social sciences, a one standard deviation difference, such as in IQ, is very large. Two is almost unheard of. Two standard deviations after adjusting for childhood income is off the charts.
Why does the black-to-white ratio get steadily worse with higher income?
I don’t know. Before seeing Chetty’s data, I might have guessed it shrank.
Is the cause racism?
Well, if it is, racism doesn’t much hinder black women. They appear to be incarcerated only about 30 percent more often than white women raised with the same family income, not 300 percent more often as with black men.
You can read the whole thing at:
https://www.takimag.com/article/americas-black-male-problem/
Thanks.
Excellent summary. I'm not an economist, but I've been writing a series of appreciations / critiques of Chetty's big studies over the last dozen years. I focus on using examples to illustrate Chetty's strengths and weaknesses. For example, here is my in-depth 2025 analysis of his study of what were the best counties for blue collar families to raise their kids in:
https://www.takimag.com/article/moneyball_for_real_estate_steve_sailer/
I identified a number of weaknesses:
'In summary, Chetty’s data still suffers from crippling problems with:
"– Regression toward the mean (especially among races)
"– Temporary booms and busts
"– Cost of living differences.
"Yet, these should not be impossible challenges for him to overcome in future iterations."
I would certainly agree that idiosyncratic booms and busts may very well be the most important thing in determining intergenerational mobility. But why would this random error bias the coefficient of some particular factor on outcomes? I obviously agree that ranking counties, and expecting those rankings to hold indefinitely, is silly, but you can still extrapolate from the characteristics of a region what your likely outcomes would be.
Chetty's big finding in 2015 was that the best county to grow up in the bottom half of parental income in the 1990s for having a higher income in 2010-11 around age 30 was Sioux County, Iowa, a nearly all white farm county with excellent soil and a very conservative culture (it's the home to the Dutch Reformed Church in the U.S.).
Most of the other top 25 most upwardly mobile counties were similarly white rural counties in the northern Great Plains.
The bottom 25 counties for blue collar kids were mostly heavily minority. The absolute worst was the Pine Ridge Sioux reservation in South Dakota, which has been among the most tragic places in the United States since Wounded Knee in 1890.
This is because blacks, American Indians, and Alaska Natives tend to regress toward lower mean incomes than whites.
The one mostly white county in the bottom 25 was Horry County, SC, home to the immense Myrtle Beach golf resort. Myrtle Beach was booming in the 1990s, but then a huge recession hit the golf industry from 2001 until covid brought golf back into fashion. Not surprisingly, Myrtle Beach adolescents whose parents were prospering constructing golf courses and condos in the 1990s were struggling in 2010.
Similarly, Chetty got a lot of publicity for reporting in 2013 that Charlotte, North Carolina was the worst big metro area in the country for upward mobility. But a recent study by him finds it got much better lately.
The obvious explanation is that Charlotte prospered in 1994-2000 due to home building. Charlotte was one of the hubs for mortage banking (e.g., Wachovia Bank), for home construction, for golf course second homes, for lumber, and for furniture making. The collapse of the housing bubble in 2008 left young people in Charlotte really badly off. But, over time, Charlotte's economy has improved and young people in Charlotte now enjoy more upward mobility than they did at the depths of the Great Recession.
Bizarrely, Chetty is not averse to taking some credit for Charlotte recovering from the bursting of the Housing Bubble, as I pointed out last year:
https://www.stevesailer.net/p/raj-chetty-advises-kamala-on-how
By the way, another reason the northern Great Plains did so well for blue collar upward mobility from the 1990s to 2010-2011 in Chetty's data was the unexpected development of fracking technology that set off a huge energy boom in North Dakota, which brought in blue collar workers from surrounding states who were tough enough to work out doors in the oilfields in winter.
Fantastic overview which happens to gloss over the most important question, one of political economy. How did he manage to get access to all of this data?
Is he missing a first born child? Is his middle name Faust?
I think it’s pretty clear on how. He had a track record of truly exceptional work for years.
If I may interrupt, professor Deming detailed the process of their paper on legacy admissions here https://econofact.org/wp-content/uploads/2023/07/EFChats-Transcript-Deming-The-Top-1-Percents-Admissions-to-Highly-Selective-Colleges.pdf
"Well, as you might imagine, Michael, it was one step at a time, and there were many steps to goin the journey – a long and winding path. It took us, I would say, the better part of five or six years. I mean, I've lost track at this point from beginning to end. Part of it was just an enormous effort of collaboration with partners and data merging and analysis. We spent a year or more crisscrossing the country, trying to convince college presidents and admissions officers and other people in universities to share data with us and to partner with us more broadly. We shared some data back with them and so on. And that required a lot of work, but I hope it was worth the effort. I mean, we lost some time to the pandemic as many people did, and at times I thought, maybe with this research, we'd never see the light of day, but it finally did. And for that, I'm very grateful both to all the people on the research team, including my co-authors, and our research assistants, but also, importantly, our partner institutions, without whom this research would not be possible."
And also him and prof Chetty here https://www.hks.harvard.edu/faculty-research/policycast/legacy-privilege-david-deming-and-raj-chetty-how-elite-college-admissions-policies#transcript-legacy
David Deming: Sure. Yeah, Ralph. Well, it was quite a process as you said. We started this project in 2017, so six years ago, and we started sort of one school or one university system at a time asking people if they were interested in participating. Of course, as you might imagine, a lot of people said no, some people said yes. And then having worked with some early partners like the University of Texas, Austin, and the University of California system and other schools like that, once we got a few people saying yes and showed that we were serious, we had a conference called the Climb Conference also in 2017, which led to a surge of interest and some partners. And then it was kind of like a snowball rolling downhill. I mean, again, we didn't get universal participation, but we spent a lot of time shaking hands, knocking on doors, whatever metaphors you want to use. And then eventually we ended up recruiting, I think, more than 400 colleges and universities encompassing about three and a half million students per year. It's about 15% of the undergraduate population in the US. So we ended up with quite a few partners and we're really grateful to them. Without them, this research wouldn't be possible.
Raj Chetty: And just to add a bit more color on how we got into this, a lot of this data collection started after an earlier paper we had released together with our collaborator, John Friedman at Brown University, on just measuring rates of economic mobility across colleges. And that was not using internal college admissions data, it was using data from federal tax records in the Department of Education to measure the parental income distributions at every college and how kids did after college. And so we were able to put those statistics out publicly and lots of people were able to see how their college stacked up relative to other colleges. And that led to a lot of interest in understanding how we can create more mobility at our colleges. And so that naturally led to springboard to have the types of conversations that Dave described where we were able to say, "If you really want to answer that question, we need more data," which is what researchers usually say in order to go deeper.