I’m with Arnold here. The one thing that Chetty said on Econtalk that gives me pause, though, is that there is a “dosage” effect such that if parents move to a better neighborhood, it affects younger children more than older children. I’ve been trying to think of how selection could drive that finding.
It is hard to pick up a new sport and play competitively at an older age. You already have activities and developed interests separate from the rich kids. In other words, much less chance of the kids interacting.
Perhaps Chetty is looking only at what is being "dosed" rather than looking at what and who is being "dosed"? In other words, the moving only helps the children who can actually take advantage of the new environment, and those are the children of precisely the parents who were already capable of moving to remaining in a wealthier neighborhood in the first place. To unconfound (is that a word?) this, one would have to compare families who never would have made the move on their own with those who needed no assistance in doing so.
Yeah I just heard Chetty on EconTalk and was disappointed Russ did not bring that up.
I think the power of genetics as an explanation will mean that in a few years social scientists will be embarrassed to admit that they did not use any generic data in their studies.
Exactly what is the complaint, here? That Chetty looked at cross-class "friendships" instead of "genetics" in determining success? Or that by failing to control for "genetics" the results of Friendship are distorted?
Since genes are likely in the causal path for human behavior and performance ignoring their contribution means you are setting yourself up to be fooled by selection effects.
This is the failure to "control" for variables that work through genetics. But in principle one ought to control for these anyway. Arnold's example of "impulse control" is a good example. That could be a methodological flaw whether or not impulse control is or is not genetic.
Because implicitly people are looking for policy levers to improve people's lives. Taking genetic endowments into effects is useful in identifying the most useful policy variables, but are uninteresting in themselves and so get downplayed
If 90% of something is caused by genetic effects that you can't change, but you only study things you can change, you are essentially throwing away 90% of the truth.
That may be human nature, especially human nature of those who want to change the world for the better, but it gives you a result that is 90% false.
I don't what to exaggerate our difference in point of view so let me just try restating how I think "genetics" ought to affect social science research. I
f you are aiming a piece of artillery, you certainly do not ignore the gravitational constant, but you are interested in the angle of fire, the force of the powder in the shell, the air temperature.
Let's suppose that there is a distribution of propensity of violent behavior as there is for height. We still might need to know the number of police officers on the street, the length of sentences, the unemployment rate between neighborhoods, or maybe even who offenders and non-offenders' childhood friend are to figure out the what to do to reduce violent crime.
I agree. I am just so sick of social science policy research that says things like, "Children who grow up in homes with books do better in school than children who do not. Therefore, if we got books into bookless homes, those children would do well in school." or "Rich people's children do better in school. Therefore, rich children have an unfair advantage/poor children have an unfair disadvantage."
If the correlations are mostly caused by genetics, the "therefores" fail, as will public policy interventions based on those "therefores".
I think the way to look at most social science research (or policy, for that matter) is as a hypothesis subject to further testing. The prior distribution of books in homes is obviously not random, but getting more books into low-book homes might be cost effective nonetheless.
I'm sorry but this is going to sound mean. First you say, it's "a hypothesis subject to further testing". Then you seem to suggest that we should try it anyway without waiting for testing.
That is what I am so sick of. Because it leads to, "Well, this sounds good so let's do it. We can't wait for years of testing." And since no one wants to test while controlling for genetics, no one is going to! So we get bad policy AND a lack of good science. But people making policy feel good because they are "doing something".
It's not science if critics can't see the data to test his claims against obvious conflicts with common sense. The data is not just not public, apparently the number of people with access can be counted with two hands. At least appoint an anonymous devil's advocate defense attorney motivated to do his best to poke holes in the analysis. Absent that, it's reasonable to presume the government trusts him to publish the desired results.
Michael Porter and Peter Theil have both emphasized the importance of monopoly-like advantages and scenarios for business success, otherwise no profits over the market rate of return. Well, same goes for science, there's a lot of competition that's hard to beat. But it sure helps if you've got a permanent monopoly on the data.
This is Chetty's cooked, aggregated data. The most specific level of aggregation is high school which is ~10^3 individuals, and the data offered is Chetty's constructs (connectedness, cohesiveness, civic engagement, exposure to high SES individuals and so on; see data description in the README file). It's no doubt interesting and useful for reference and possibly for practical purposes such as deciding which school district to move to, but for testing Chetty's claims it is worthless. For that, one needs access to the raw individual data - if not the data dump itself, then at least a complete description of the raw data's structure and a way to execute aggregated queries against it remotely. The latter approach is similar to the way ordinary public gets access to DALL-E, GPT-3 and other big machine learning models. Given sensible limitations on the form (basically number of outputs) and number of queries users are permitted to run, this should be sufficient to completely reproduce Chetty's results and test alternate hypotheses, but not to deanonymize the individuals whose data Chetty cadged out of IRS and Meta.
Impulse control is part of what psychologists call "executive function", which is known to be almost entirely genetic. Based on that, we can discard most hypotheses that involve other factors*, such as friendships or socioeconomic status, having significant effects on impulse control
It can, but in order for it to be able to provide this function, it has to be robust, and that unfortunately means exactly the kind of thick, hierarchical interpersonal relationships which very modern, liberally minded people find especially icky, even when they tolerate and practice them to an extent within their nuclear families and communities in the manner Charles Murray berated them for in *Coming Apart* ("practice but don't preach").
Compare two recents studies by John A. List and colleagues. (Note: What List et al. call non-cognitive traits includes, among other things, impulse control.):
1) "The Social Side of Early Human Capital Formation: Using Field Experiment to Estimate
the Causal Impact of Neighborhoods" (December 2020):
"This study leverages insights from sociology to explore the role of neighborhoods on human capital formation at an early age. We do so by estimating the spillover effects from a large-scale early childhood intervention on the educational attainment of over 2,000 disadvantaged children in the US. We document large spillover effects on both treatment and control children who live near treated children. [...] spillover effects are localized, decreasing with the spatial distance to treated neighbors. Perhaps our most novel insight is the underlying mechanisms at work: the spillover effect on non-cognitive scores operate through the child's social network while parental investment is an important channel through which cognitive spillover effects operate." (Abstract)
"We find that non-cognitive spillover effects are significantly larger for Blacks than Hispanics. [... .] African American adolescents are significantly more likely than Hispanics to (i) know most people in their neighborhoods, (ii) stop on the street and talk to someone from the neighborhood, and (iii) use recreation facilities in the neighborhood." (pp. 4 & 5)
" [...] non-cognitive spillover effects are more likely to operate through children’s rather than parents’ social networks." (p. 4)
2) "How Experiments with Children Inform Economics" (May 2021):
"[...] skills and economic preferences result from a nature-nurture (i.e., gene-environment) interaction. In our framework, children are born with a 'band' of potential skills or preferences, and the level expressed within that band depends on the environment, including parenting or interventions. [... .] genes explain some - but not all - of the variation in economic preferences and outcomes " (pp. 10-11)
"[...] the effects of a parenting program were most concentrated among students with high executive function skills at baseline" (p. 14)
"There is some evidence for 'reversion to the mean' in the sense that many early childhood interventions see fade-out of skills learned in the years following a program" (p. 14)
"[...] there is some evidence that [cognitive and non-cognitive] skills move in tandem" (p. 15)
"[...] a review of data from interventions targeted at various age groups have shown benefit-cost ratios that decline with the age of the targeted group" (p. 15)
"[...] low-SES parents are less likely to believe that their investments in their baby can affect brain development as compared to higher-income and higher-educational attainment parents." (p. 16)
"Children who believe they can change their intelligence (growth mindset) achieve more than children who believe that they cannot do so (fixed mindset)" (p. 16)
"[...] time preferences measured at age 3-5 are associated with disciplinary referrals years later in elementary school. Importantly, these preferences are distinct from cognitive skills (e.g., reading, writing, math) and executive functions (e.g., inhibitory control and working memory) in that they have an independent role in predicting disciplinary referrals." (p. 26)
"[...] time and risk preferences measured in adolescence are correlated with educational and labor market outcomes later in life." (p. 29)
One context where I think there is likely to be a causal effect to different SES people mixing together also happens to connect to your impulse control objection.
Consider the LDS communities in Utah, which score quite well for upward mobility in Chetty’s data. In the church community, poor people receive a lot more direct attention and assistance from rich people than in other settings, in terms of money, time, and free counseling. Mitt Romney essentially had a part time volunteer job helping struggling Mormons on the side of his career in the private sector. Now what happens when Mitt Romney meets a young member of his ward who is poor and aimless, but also very bright and capable? I suspect he makes a big effort to connect him to a job or schooling opportunity that will point him in the right direction. I’m sure this has a meaningful effect, and have indeed met several people like this who were helped by less famous versions of Mitt.
Won’t LDS communities select on impulse control? Surely to some extent, since standards of abstinence from alcohol and pre-marital sex are harder to fully keep without it. But many people who struggle with these things also select into the church and I think they benefit from it. Even if we suppose that only the higher impulse control individuals see the persistent, meaningful gains I think that there is a lot more value than most people realize to having high status people around to encourage and recognize talent. Many people would not accomplish nearly as much without that experience.
I wish Chetty would emphasize this connection to religion more explicitly when he talks to the New York Times (he did in his CWT), since I don’t think just living in the vicinity of rich people does very much. But despite partly agreeing with objections made by yourself and Jim Heckman, I think there’s a causal effect of some sort here, not just selection.
Compare John A. List et al., "How Experiments with Children Inform Economics" (May 2021, forthcoming JEL):
" There is some evidence for 'reversion to the mean' in the sense that many early childhood interventions see fade-out of skills learned in the years following a program. This suggests that children who increase their skills through early childhood interventions may need to be scaffolded with additional interventions in the middle-childhood years to attain
persistent gains" (pp. 14-15)
In the spirit of Infovores' comment, I would say that religion can be what List et al. call "scaffolding," for persistent, continuous "interventions," if religion involves an institution, a place, shared norms/expectations, and positive peer effects.
A quick note from the abstract of the second paper: "We show that about half of the social disconnection across socioeconomic lines—measured as the difference in the share of high-socioeconomic status (SES) friends between people with low and high SES—is explained by differences in exposure to people with high SES in groups such as schools **and religious organizations.**"
Seems like you're thinking in the same frame as what the study finds. Though, you're right, he doesn't emphasize this component in his public-facing appearances, at least not that I've seen. Probably something to do with who he thinks the audience is and what they want (more school, less church?)
Here's a relevant section for others who are interested (I have slightly adjusted the formatting, removing reference numbers and explaining EC):
"Friending bias is lowest on average in religious groups, in which friending bias is −0.03, implying that low-SES people tend to form friendships with high-SES members of their religious groups at a rate that is slightly higher than the share of high-SES people in their religious groups. Friending bias is negative in religious groups because religious-group friendships do not exhibit substantial homophily by SES—a finding that is consistent with previous research using survey data—and because high-SES people make more friends than low-SES people. Holding fixed exposure, people with low SES are about 20% more likely to befriend a given high-SES person in their religious groups than in their neighbourhoods—a large difference, comparable in magnitude to the 22.4% under-representation of high-SES friends on average among low-SES individuals. Put differently, if friending bias in all settings was reduced by an amount equal to the difference in friending bias between neighbourhoods and religious groups, most of the disconnection between low-SES and high-SES individuals in the US would be eliminated.
Since religious groups are highly segregated by income, as shown in Fig. 2b, their low friending bias does not currently translate to a high level of EC [economic connectedness] (Fig. 2a). Efforts to integrate religious groups by SES may be particularly effective at increasing EC if friending bias remains low as they become more integrated. This assumption is not innocuous—as illustrated by the challenges faced in efforts to integrate college classrooms—but it is bolstered by the fact that religious groups exhibit low levels of friending bias at all levels of exposure (Supplementary Fig. 1b)."
Well obviously. I very much doubt that it was Chetty's hypothesis that it was the wealth per se that caused the increase in positive outcomes of the children. But rather some basket of other attributes that the wealth either caused, or which are correlated with wealth. I mean obviously these young children are not running around spending mommy and daddy's money. There are other factors at work here
Depending on the cost and difficulty, trying to promote less class segregation in education is probably still a good idea. My Bayesian prior of effectiveness is not zero.
An obvious factor for friendship among children is sports and activities. How many poor kids play lacrosse? Hockey? The equipment cost is prohibitive. How many rich kids play football now? Soccer and basketball probably have a mix of players, but the leagues are different. Travel teams and clubs are costly in time and money. Activities like scouting are now mostly rich kids. My son has enrolled in the Russian School of Math since 3 (he likes it), not many poor kids there. How many poor kids ski? Summer camps are definitely divided by class. Not many poor kids are in my son’s sailing camp. I just went to New England county fair with my son, one I used to attend as a child. As Boston resident immersed in the biotech/academic world, it was like visiting a foreign country - there were Trump flags for sale!
The obvious reason for the divide - the rich are building their children’s resume for college admission and do everything they can to differentiate them.
"The obvious reason for the divide - the rich are building their children’s resume for college admission and do everything they can to differentiate them."
I never thought of that. Another reason to hate the college admissions process.
I’m with Arnold here. The one thing that Chetty said on Econtalk that gives me pause, though, is that there is a “dosage” effect such that if parents move to a better neighborhood, it affects younger children more than older children. I’ve been trying to think of how selection could drive that finding.
It is hard to pick up a new sport and play competitively at an older age. You already have activities and developed interests separate from the rich kids. In other words, much less chance of the kids interacting.
Perhaps Chetty is looking only at what is being "dosed" rather than looking at what and who is being "dosed"? In other words, the moving only helps the children who can actually take advantage of the new environment, and those are the children of precisely the parents who were already capable of moving to remaining in a wealthier neighborhood in the first place. To unconfound (is that a word?) this, one would have to compare families who never would have made the move on their own with those who needed no assistance in doing so.
Why is it where academic and economic success is studied, that above all genetics must at all costs be ignored as a major determining factor?
Because that's what the people with the carrots and sticks want.
Yeah I just heard Chetty on EconTalk and was disappointed Russ did not bring that up.
I think the power of genetics as an explanation will mean that in a few years social scientists will be embarrassed to admit that they did not use any generic data in their studies.
Exactly what is the complaint, here? That Chetty looked at cross-class "friendships" instead of "genetics" in determining success? Or that by failing to control for "genetics" the results of Friendship are distorted?
Since genes are likely in the causal path for human behavior and performance ignoring their contribution means you are setting yourself up to be fooled by selection effects.
This is the failure to "control" for variables that work through genetics. But in principle one ought to control for these anyway. Arnold's example of "impulse control" is a good example. That could be a methodological flaw whether or not impulse control is or is not genetic.
Because implicitly people are looking for policy levers to improve people's lives. Taking genetic endowments into effects is useful in identifying the most useful policy variables, but are uninteresting in themselves and so get downplayed
If 90% of something is caused by genetic effects that you can't change, but you only study things you can change, you are essentially throwing away 90% of the truth.
That may be human nature, especially human nature of those who want to change the world for the better, but it gives you a result that is 90% false.
I don't what to exaggerate our difference in point of view so let me just try restating how I think "genetics" ought to affect social science research. I
f you are aiming a piece of artillery, you certainly do not ignore the gravitational constant, but you are interested in the angle of fire, the force of the powder in the shell, the air temperature.
Let's suppose that there is a distribution of propensity of violent behavior as there is for height. We still might need to know the number of police officers on the street, the length of sentences, the unemployment rate between neighborhoods, or maybe even who offenders and non-offenders' childhood friend are to figure out the what to do to reduce violent crime.
I agree. I am just so sick of social science policy research that says things like, "Children who grow up in homes with books do better in school than children who do not. Therefore, if we got books into bookless homes, those children would do well in school." or "Rich people's children do better in school. Therefore, rich children have an unfair advantage/poor children have an unfair disadvantage."
If the correlations are mostly caused by genetics, the "therefores" fail, as will public policy interventions based on those "therefores".
I think the way to look at most social science research (or policy, for that matter) is as a hypothesis subject to further testing. The prior distribution of books in homes is obviously not random, but getting more books into low-book homes might be cost effective nonetheless.
I'm sorry but this is going to sound mean. First you say, it's "a hypothesis subject to further testing". Then you seem to suggest that we should try it anyway without waiting for testing.
That is what I am so sick of. Because it leads to, "Well, this sounds good so let's do it. We can't wait for years of testing." And since no one wants to test while controlling for genetics, no one is going to! So we get bad policy AND a lack of good science. But people making policy feel good because they are "doing something".
It's not science if critics can't see the data to test his claims against obvious conflicts with common sense. The data is not just not public, apparently the number of people with access can be counted with two hands. At least appoint an anonymous devil's advocate defense attorney motivated to do his best to poke holes in the analysis. Absent that, it's reasonable to presume the government trusts him to publish the desired results.
Michael Porter and Peter Theil have both emphasized the importance of monopoly-like advantages and scenarios for business success, otherwise no profits over the market rate of return. Well, same goes for science, there's a lot of competition that's hard to beat. But it sure helps if you've got a permanent monopoly on the data.
https://data.humdata.org/dataset/social-capital-atlas
This is Chetty's cooked, aggregated data. The most specific level of aggregation is high school which is ~10^3 individuals, and the data offered is Chetty's constructs (connectedness, cohesiveness, civic engagement, exposure to high SES individuals and so on; see data description in the README file). It's no doubt interesting and useful for reference and possibly for practical purposes such as deciding which school district to move to, but for testing Chetty's claims it is worthless. For that, one needs access to the raw individual data - if not the data dump itself, then at least a complete description of the raw data's structure and a way to execute aggregated queries against it remotely. The latter approach is similar to the way ordinary public gets access to DALL-E, GPT-3 and other big machine learning models. Given sensible limitations on the form (basically number of outputs) and number of queries users are permitted to run, this should be sufficient to completely reproduce Chetty's results and test alternate hypotheses, but not to deanonymize the individuals whose data Chetty cadged out of IRS and Meta.
Impulse control is part of what psychologists call "executive function", which is known to be almost entirely genetic. Based on that, we can discard most hypotheses that involve other factors*, such as friendships or socioeconomic status, having significant effects on impulse control
See https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2762790/.
*With some exceptions for rare and extreme factors (e.g. severe malnutrition or brain trauma).
I was going to make this point and link to the same study. Fairly clearly, religion can provide an external substitute for internal executive control.
It can, but in order for it to be able to provide this function, it has to be robust, and that unfortunately means exactly the kind of thick, hierarchical interpersonal relationships which very modern, liberally minded people find especially icky, even when they tolerate and practice them to an extent within their nuclear families and communities in the manner Charles Murray berated them for in *Coming Apart* ("practice but don't preach").
Compare two recents studies by John A. List and colleagues. (Note: What List et al. call non-cognitive traits includes, among other things, impulse control.):
1) "The Social Side of Early Human Capital Formation: Using Field Experiment to Estimate
the Causal Impact of Neighborhoods" (December 2020):
https://bfi.uchicago.edu/wp-content/uploads/2021/01/BFI_WP_2020187.pdf
Excerpts:
"This study leverages insights from sociology to explore the role of neighborhoods on human capital formation at an early age. We do so by estimating the spillover effects from a large-scale early childhood intervention on the educational attainment of over 2,000 disadvantaged children in the US. We document large spillover effects on both treatment and control children who live near treated children. [...] spillover effects are localized, decreasing with the spatial distance to treated neighbors. Perhaps our most novel insight is the underlying mechanisms at work: the spillover effect on non-cognitive scores operate through the child's social network while parental investment is an important channel through which cognitive spillover effects operate." (Abstract)
"We find that non-cognitive spillover effects are significantly larger for Blacks than Hispanics. [... .] African American adolescents are significantly more likely than Hispanics to (i) know most people in their neighborhoods, (ii) stop on the street and talk to someone from the neighborhood, and (iii) use recreation facilities in the neighborhood." (pp. 4 & 5)
" [...] non-cognitive spillover effects are more likely to operate through children’s rather than parents’ social networks." (p. 4)
2) "How Experiments with Children Inform Economics" (May 2021):
https://bfi.uchicago.edu/wp-content/uploads/2021/05/BFI_WP_2021-62.pdf
Excerpts:
"[...] skills and economic preferences result from a nature-nurture (i.e., gene-environment) interaction. In our framework, children are born with a 'band' of potential skills or preferences, and the level expressed within that band depends on the environment, including parenting or interventions. [... .] genes explain some - but not all - of the variation in economic preferences and outcomes " (pp. 10-11)
"[...] the effects of a parenting program were most concentrated among students with high executive function skills at baseline" (p. 14)
"There is some evidence for 'reversion to the mean' in the sense that many early childhood interventions see fade-out of skills learned in the years following a program" (p. 14)
"[...] there is some evidence that [cognitive and non-cognitive] skills move in tandem" (p. 15)
"[...] a review of data from interventions targeted at various age groups have shown benefit-cost ratios that decline with the age of the targeted group" (p. 15)
"[...] low-SES parents are less likely to believe that their investments in their baby can affect brain development as compared to higher-income and higher-educational attainment parents." (p. 16)
"Children who believe they can change their intelligence (growth mindset) achieve more than children who believe that they cannot do so (fixed mindset)" (p. 16)
"[...] time preferences measured at age 3-5 are associated with disciplinary referrals years later in elementary school. Importantly, these preferences are distinct from cognitive skills (e.g., reading, writing, math) and executive functions (e.g., inhibitory control and working memory) in that they have an independent role in predicting disciplinary referrals." (p. 26)
"[...] time and risk preferences measured in adolescence are correlated with educational and labor market outcomes later in life." (p. 29)
One context where I think there is likely to be a causal effect to different SES people mixing together also happens to connect to your impulse control objection.
Consider the LDS communities in Utah, which score quite well for upward mobility in Chetty’s data. In the church community, poor people receive a lot more direct attention and assistance from rich people than in other settings, in terms of money, time, and free counseling. Mitt Romney essentially had a part time volunteer job helping struggling Mormons on the side of his career in the private sector. Now what happens when Mitt Romney meets a young member of his ward who is poor and aimless, but also very bright and capable? I suspect he makes a big effort to connect him to a job or schooling opportunity that will point him in the right direction. I’m sure this has a meaningful effect, and have indeed met several people like this who were helped by less famous versions of Mitt.
Won’t LDS communities select on impulse control? Surely to some extent, since standards of abstinence from alcohol and pre-marital sex are harder to fully keep without it. But many people who struggle with these things also select into the church and I think they benefit from it. Even if we suppose that only the higher impulse control individuals see the persistent, meaningful gains I think that there is a lot more value than most people realize to having high status people around to encourage and recognize talent. Many people would not accomplish nearly as much without that experience.
I wish Chetty would emphasize this connection to religion more explicitly when he talks to the New York Times (he did in his CWT), since I don’t think just living in the vicinity of rich people does very much. But despite partly agreeing with objections made by yourself and Jim Heckman, I think there’s a causal effect of some sort here, not just selection.
Compare John A. List et al., "How Experiments with Children Inform Economics" (May 2021, forthcoming JEL):
" There is some evidence for 'reversion to the mean' in the sense that many early childhood interventions see fade-out of skills learned in the years following a program. This suggests that children who increase their skills through early childhood interventions may need to be scaffolded with additional interventions in the middle-childhood years to attain
persistent gains" (pp. 14-15)
In the spirit of Infovores' comment, I would say that religion can be what List et al. call "scaffolding," for persistent, continuous "interventions," if religion involves an institution, a place, shared norms/expectations, and positive peer effects.
A quick note from the abstract of the second paper: "We show that about half of the social disconnection across socioeconomic lines—measured as the difference in the share of high-socioeconomic status (SES) friends between people with low and high SES—is explained by differences in exposure to people with high SES in groups such as schools **and religious organizations.**"
Seems like you're thinking in the same frame as what the study finds. Though, you're right, he doesn't emphasize this component in his public-facing appearances, at least not that I've seen. Probably something to do with who he thinks the audience is and what they want (more school, less church?)
Absolutely, it’s completely consistent with their findings, and they do mention it. But when it reaches the press it gets kind of buried IMO.
Here's a relevant section for others who are interested (I have slightly adjusted the formatting, removing reference numbers and explaining EC):
"Friending bias is lowest on average in religious groups, in which friending bias is −0.03, implying that low-SES people tend to form friendships with high-SES members of their religious groups at a rate that is slightly higher than the share of high-SES people in their religious groups. Friending bias is negative in religious groups because religious-group friendships do not exhibit substantial homophily by SES—a finding that is consistent with previous research using survey data—and because high-SES people make more friends than low-SES people. Holding fixed exposure, people with low SES are about 20% more likely to befriend a given high-SES person in their religious groups than in their neighbourhoods—a large difference, comparable in magnitude to the 22.4% under-representation of high-SES friends on average among low-SES individuals. Put differently, if friending bias in all settings was reduced by an amount equal to the difference in friending bias between neighbourhoods and religious groups, most of the disconnection between low-SES and high-SES individuals in the US would be eliminated.
Since religious groups are highly segregated by income, as shown in Fig. 2b, their low friending bias does not currently translate to a high level of EC [economic connectedness] (Fig. 2a). Efforts to integrate religious groups by SES may be particularly effective at increasing EC if friending bias remains low as they become more integrated. This assumption is not innocuous—as illustrated by the challenges faced in efforts to integrate college classrooms—but it is bolstered by the fact that religious groups exhibit low levels of friending bias at all levels of exposure (Supplementary Fig. 1b)."
Seems like LDS is just such a religious group. Here's a link to their interactive map sorted by zip code and looking at salt lake city: https://www.socialcapital.org/?dimension=EconomicConnectednessIndividual&dim1=EconomicConnectednessIndividual&dim2=CohesivenessClustering&dim3=CivicEngagementVolunteeringRates&geoLevel=zcta&selectedId=#map-selection-view-link
Fun to explore how the data "looks"
Someone should ask Emil Kirkegaard for possible factors like IQ and verbal tilt distorting this result of parental correlation vs environmental.
Well obviously. I very much doubt that it was Chetty's hypothesis that it was the wealth per se that caused the increase in positive outcomes of the children. But rather some basket of other attributes that the wealth either caused, or which are correlated with wealth. I mean obviously these young children are not running around spending mommy and daddy's money. There are other factors at work here
Depending on the cost and difficulty, trying to promote less class segregation in education is probably still a good idea. My Bayesian prior of effectiveness is not zero.
An obvious factor for friendship among children is sports and activities. How many poor kids play lacrosse? Hockey? The equipment cost is prohibitive. How many rich kids play football now? Soccer and basketball probably have a mix of players, but the leagues are different. Travel teams and clubs are costly in time and money. Activities like scouting are now mostly rich kids. My son has enrolled in the Russian School of Math since 3 (he likes it), not many poor kids there. How many poor kids ski? Summer camps are definitely divided by class. Not many poor kids are in my son’s sailing camp. I just went to New England county fair with my son, one I used to attend as a child. As Boston resident immersed in the biotech/academic world, it was like visiting a foreign country - there were Trump flags for sale!
The obvious reason for the divide - the rich are building their children’s resume for college admission and do everything they can to differentiate them.
"The obvious reason for the divide - the rich are building their children’s resume for college admission and do everything they can to differentiate them."
I never thought of that. Another reason to hate the college admissions process.