How much does social mobility ever change?

Here is Dylan Matthews interviewing Gregory Clark about his new book The Son also Rises:

Another remarkable feature of the surname data is how seemingly impervious social mobility rates are to government interventions. In all societies, what seems to matter is just who your parents are. At the extreme, we see in modern Sweden an extensive system of public education and social support. Yet underlying mobility rates are no higher in modern Sweden than in pre-industrial Sweden or medieval England.

There was one case where government interventions did seem to promote mobility, which was in Bengal, in India. There the strict quota system in educational institutions had benefited significantly people with surnames associated with the Scheduled Castes.

But the bizarre element here is that these quotas did not help those truly at the bottom of the social ladder. Instead, the benefits went to families of average social status whom the British had mistakenly classified as Scheduled Caste. These families have now become a new elite. The truly disadvantaged, such as the large Muslim community, have been correspondingly further burdened by being excluded from these quotas.

Interestingly, in China, the extreme social intervention represented by the Communist Revolution of 1949, which included executing large numbers of members of the old upper class, has not resulted in much of an increase in social mobility. Surnames of high status in the Imperial and Republican era continue to be overrepresented among modern elites, including Communist Party officials.

The families that have high social competence, whatever the social system is, typically find their way to the top of the social ladder.

The interview is interesting throughout. And you will of course note the new Chetty results — created with entirely different methods and data — showing economic mobility has not much changed in the United States for decades.

For the initial pointer I thank Samir Varma.


The problem is that intervention itself is endogenous - one, that middle classes only agitate for redistribution when their own ambitions are threatened; and two, that particularly severe stratification leads to revolts where the new elite simply do not identify themselves with the old ethnic-national identity, so checking the movement of surnames is nonsensical. Elite Muslims in India have a relatively minimal national presence because all the descendants of elite Muslims are now in Pakistan or Bangladesh.

*nonsensical in those instances where national borders have changed relatively recently. That does relatively minimal damage to some methods of analyzing intergenerational mobility (it shouldn't affect Chetty's... not obviously, at least), but Clark has a tendency toward unwarranted optimism in analysis.

Not only national borders, but also immigration. What % of the Swedish population emigrated to the US, for example? Probably close to 20%.

"Instead, the benefits went to families of average social status whom the British had mistakenly classified as Scheduled Caste."

Sounds like the U.S.

As Henry Louis Gates and Lani Guinier complained in 2004 about racial quotas at Harvard, they were set up to benefit the descendants of the victims of American slavery, but quotas wind up disproportionately benefiting immigrants or their children, including Africans descended from slavesellers, and/or individuals who identify as black even though they have a white parent or grandparent and may have been raised in a wholly non-black privileged environment. (Does this sound like anybody you've ever heard of?)

Ironically, Gates and Guinier pointed this out a couple of months before Harvard LS grad Barack Obama became Presidential Timber at the 2004 Democratic Convention.

It's not only about quotas. The descendents of the victims of American slavery recently voted overwhelmingly for a candidate having zero share in their heritage on the basis of moderate cutaneous melanin.

They've been doing that for quite some time now, some say even before 1964, on the bases of even less cutaneous melanin.

Elizabeth Warren is another example. No group in America is more disadvantaged than the aboriginals so it's cool to rise up from those humble beginnings even if the story is fictitious. The Indians should be more careful about who they let into their teepees.

Warren's alleged Indian heritage was likely only the cherry on top of the sundae. She was hired at Harvard, despite her unremarkable career, because (in this order): 1) she married a Harvard professor, 2) she had a vagina, and 3) she was a liberal who fought against GOP bankruptcy reform - helping deadbeats avoid paying their debts.

She is not the first underqualified +1 appointment to a top school. Another notable +1 became Fed chair, although at least she had the right pedigree.

Im still waiting for someone to explain what credentials make Yellin qualified for Fed chair. And her prior post is hardly qualifying since she wasnt qualified for that either. I see nothing remarkable in her CV except for papers co-authored with her brilliant husband. I will happily accept persuasive arguments that I am wrong.

Liberal women I know were swooning over her before they knew she was speaking with forked tongue. Going by the small sample of moonbat women I know, the dishonest injun stuff only helped her cause. They were convinced Scott brown was going to sew their wombs shut before Warren came out of her teepee to run for the seat. But, having a senator that is a pathological liar with the IQ of a goldfish is standard fare for Massachusetts. At least she has not murdered or raped anyone that we know of.

Warren is a first class idiot, but she was ten times the lawyer that Obama and Hillary were, and that really isn't saying much at all. She's far worse a politician, but in her state that isn't a high hurdle. Not being a criminal is high praise for a democrat.

You are wrong, Willits. Yellen could not be more qualified for the Fed Chair ... but hey, you are less wrong than Z, so there's that.

Wow, a several thousand word essay that is 100% fact free. Free of relevant facts, that is.

I asked for an intelligent articulation of her qualifications, and you provide a gushing puff piece filled with the opinions of her admirers.

The article talks at length about research she co-authored with her husband. NB that he won a Nobel Prize and she did not. NB that she FAILED to get tenure at Harvard. Why? Presumably because her research wasnt up to snuff.

Before I posted, I took a look at a selected list of her publications. The only ones that were ranked in top econ journals were co-authored with her husband. The rest were all puff pieces, low ranking journals, and invited pieces.

Z is absolutely correct. Yellen was being hoisted on the shoulders of liberals before Summers had been rejected. My belief is that he was NEVER the intended candidate. He was cannonfodder so that Yellen could pass through relatively unscathed.

So, LMGTFY, you have failed. You havent even attempted to point out original work of hers or a consistent pattern of independent research that makes her qualified to vote on monetary pokicy, much less lead it.

FWIW, IMO, Christina Romer would have been much better qualified. My only criticism of her is that she was goaded into unrealistic forecasts of economic recovery by the huckster from my home town.

My only fail was trying to respond to you. It's sad but not surprising in this echo chamber. I won't make the mistake again. Enjoy your ignorance, thankfully it doesn't matter what you think.

It should be too obvious to state that in socialist, communist, fascist--any top-down statist--systems, social connections become all-important. In such environments good ideas and hard work get you nowhere if you don't find favor with someone in power.

social connections become all-important

As opposed to what?

In such environments good ideas and hard work get you nowhere if you don’t find favor with someone in power.

Proven results.

As opposed to bottom-up capitalist systems, in which one can advance using either social connections or bootstrapping or myriad other methods.

Like China, and everywhere else, it seems this was even in the case in Soviet Russia:

Using surnames has a tremendous survivorship bias. Until recent decades it has been difficult for a family to keep it surname. Especially in public records such as censuses. Ask any genealogist. So what's measured here is the social mobility with an upper-class.

It used to be the practice of the rich to incorporate their maternal lineage as a middle name.

Also, a noble name in and of itself can attract new talent: e.g., the Churchill (Duke of Marlborough) name was in the doldrums a half dozen generations after Jack Churchill, until Winston's strong-willed paternal grandmother married his boring grandfather, and then his American heiress mother was a tigress, too.

Not just the rich; common throughout Scottish society.

but the big question is *why* does intergenerational mobility appear to be so stable ... and mind you in the US that is even in periods of rising inequality, those who move up now have to make even more money than their counterpart decades ago. it's like there is a natural amount of churn in a society (though it does differ across countries). one interesting result in the Chetty et al study is that probability of upward mobility varies across the US (grow up poor in the Southeast, much less likely to become rich) BUT downward mobility does not vary across the US (grow up rich anywhere, same chance of ending up poor). It was on an unrelated topic, but I credit Dick Thaler with teaching me if you see what looks like a 'natural law' in human behavior (in this case stable mobility), there is likely an institution driving the result. (I think stable mobility in the face of rising inequality undercuts attempts to tie this to biological differences.)

"probability of upward mobility varies across the US (grow up poor in the Southeast, much less likely to become rich) BUT downward mobility does not vary across the US"

At first glance, this doesn't seem mathematically possible - the quintiles gotta stay the same size (10 folks move down => 10 had to move up). But the quintiles aren't based on the local population's income, they are national. So what you are really saying is (a) there is now more income equality in the Southeast as everyone gets equally poor, or (b) the upwardly mobile move away.

And the corollary is that somewhere in the US there are more people than before in the top quintiles.

It is entirely possible that the factors which help a low income person "climb the ladder" are different (and distributed differently across the country) than those that allow a high income person to "fall down the ladder." Also they look at measures of absolute and relative mobility, though you are right the outcome metrics are nationally defined.

Here's the executive summary of the geographic mobility paper:
And the research website which also has the paper referred to in the post:

Actually it's impossible for a low income person to "climb the ladder" without one or more high income people "falling down the ladder", so no the factors would be the same. I suppose it could be possible for a few low income people to shoot way up and most high income people to just go down a little, rather than a few high income people plummeting.

Cliff if the rich vs poor bins are defined nationally then it is entirely possible that the chance of moving up vs the chance of moving down varies sub-nationally. And the factors may well vary by the direction. For example, health shocks might play a role in downward mobility but access to quality public schools might be the key for upward mobility. For the US as a whole these flows have to add up (for relative measures) but that does not mean same factors at play in both directions. Complicated.

The Chetty study of social mobility by region, which finds dusty dying Great Plains towns to be beacons of hope, and Charlotte and Atlanta to be pits of despair, has massive methodological and interpretive flaws:

Steve, you first point is wrong. They find that own race does not matter ... it is the racial composition of the community. In fact, the associations they find with mobility are community, not individual specific attributes. The family structure on is interesting ... growing up in a community with lots of single adults lowers your mobility even if your own parents are married. So I do think you first point holds up. Your second point is a matter of inference. The locations are set in childhood and not necessarily where they live as adults. Migration would be something to think more about but it does not undermine their chart. I really encourage you to read both the papers. The interpretation is not altogether clear, but their methods are sound.

oops ... I meant: "So I do NOT think you first point holds up" (at the very least it is not going to explain their geographic variation)

"They find that own race does not matter … it is the racial composition of the community."

No, Chetty is being disingenuous here to avoid being Watsoned. He isn't using individual data by race so he couldn't find what you said he doesn't find. If you look at Chetty's map of low social mobility it's very much like a map of where NFL cornerbacks played high school ball.

The main thing he's discovered is regression toward the mean. The children of blacks in the bottom quintile regress toward the black mean while the children of whites in the bottom quintile regress toward the white mean. Thus, heavily white places like the Great Plains and West Virginia come out in his map as oases of social mobility, while fast-growing but heavily black areas like North Carolina are seen as having terrible social mobility.

Chetty should withdraw his paper from last summer because it has muddied the waters of media understanding so badly.

from the paper p 31

"Figure IXa confirms that areas with larger African-American populations do in fact have sub- stantially lower rates of upward mobility. ...

The correlation in Figure IXa could be driven by two very different channels. One channel is an individual-level race effect: black children may have lower incomes than white children conditional on parent income, and hence areas with a larger black population may have lower upward mobility. An alternative possibility is a place-level race effect: areas with large black populations might have lower rates of upward mobility for children of all races. To distinguish between these two channels, one would ideally control for race at the individual level, essentially asking whether whites have lower rates of upward mobility in areas with a larger black population.

Unfortunately, we do not observe race in our data. As an alternative, we predict race based on the parent’s 5-digit ZIP code (in the year they first claim their child as a dependent). We use data from the 2000 Census to measure racial shares by ZIP code. ... Most notably, even in this predominantly white sample, rates of upward mobility remain low in the Southeast and are much higher in the West."

Yes, it would be better to have individual race measures; however how do you explain the white zip codes? If whites really have a different mean then we should see it there too. Also the family structure result uses actual reports of own parents (married or single) not just estimation like the race analysis.

One should always be skeptical of empirical work; however, asking for uncomfortable or puzzling results to be withdraw is poor form.

"but the big question is *why* does intergenerational mobility appear to be so stable"

Imagine a society that was obsessed with achieving perfect intergeneration mobility. The most extreme solution would be to abduct every child at birth and put them in a common orphanage, where all ties to parents are severed. This would completely eliminate any advantage a parent's position in society, economic class or connections would confer on their children. Yet even in this situation we would not achieve perfect intergenerational mobility. Because the children of the wealthy and intelligence would still hold onto those genes. Naturally over time they would outperform their peers, even without knowing their own background.

Decades of twin research has shown that of the heritability of intelligence and economic success 80-100% is due to genetics, with only a very small to nil portion due to family environment. The vast majority of generational class immobility is due to genetics. It doesn't matter what kind of society we live in: feudalist, capitalist, socialist or communist. The logic of genetics remains the same. Society and the state can expend tremendous effort trying to increase generational mobility, but at the extreme it can only shave down that <20% that's due to family environment. Efforts to this end are largely pushing up against a biological invariant.

yeah, yeah ... super genes FTW. but my puzzle is why does mobility stay constant even as inequality increases? I find it much harder to say that our genes hard wire us into our relative positions (regardless whether those positions drift apart in absolute terms), maybe but I do not see that in your explanation. And here's another kicker from the paper ... own race does not matter but the racial composition of the community you grew up in does. Square that with genes. Something more complicated is at play that your biology-centric story.

"I find it much harder to say that our genes hard wire us into our relative positions"

Economic success is basically a function of IQ (okay, maybe some other traits, but we can add them altogether and call it "IQ" all the same), plus some noise. The distribution of IQ remains invariant across generations as the gene pool stays the same. However the economic environment changes. In some periods the returns to IQ are much higher than other periods. For example we know in 1960 the demand gap between skilled and unskilled workers was quite small. By 2010 the demand gap between had substantially increased. Hence high IQ workers started earning a lot more relative to low IQ workers. Yet the high IQ of 2010 are pretty much just the descendants of the high IQ of 1960, and same for the low IQ. Inequality increases but intergenerational mobility remains fixed.

In other words, inequality is pretty much just a function of the skilled to unskilled wage premium. This is largely derived from economic conditions and changes over time. But mobility is a function of the IQ distribution and its intergenerational correlation. This is pretty much genetically and determined, and remains invariant despite changing economic conditions. Hence why you see changing inequality, but static mobility.

while I do not agree with several statements you make I do appreciate that you sketched out a theory that addresses the data puzzles I mentioned. thanks.

"And here’s another kicker from the paper … own race does not matter but the racial composition of the community you grew up in does."

The best predictor of economic success is IQ, which research has shown is basically due to genetic components. Hypothetically the ultimate way to predict future mobility would be to fully sequence everyone's genome and identify every single gene that affects IQ. Of course that's just science fiction, so now we're limited to ways to proxy for genetic IQ. One very good way is to look at a person's parents, as the correlation will be very high.

Absent that another, someone decent way is to look at a person's race. Any reasonable person pretty much knows that different races have different average genetic IQ set points. Yet individual variance within races is much higher than between races. So simply knowing a person's race is by no means a perfect proxy. Another, perhaps better proxy, is to look at the racial composition of an area. For example there are many white families who are dumb, these families are probably more likely to have low success and live in poor areas. These poor areas are more likely to contain high numbers of (in aggregate) under-achieving minorities. And vice versa, there are many smart and successful black families, and on average these families are likely to live in areas with fewer of the (in aggregate) under-achieving minority groups. (Think of the Jeffersons: "Moving on up, to the East Side!").

Quite simply an individual's race tells you very little about that individual's IQ. But due to the law of large numbers a community's race tells you a lot about the community's aggregate IQ. Paradoxically the racial composition of a person's community can therefore be a better predictor of that person's IQ than his or her individual race.

Mobility, as it's commonly defined, is about moving between quintiles. Inequality, as it's commonly defined, is about the shape of the distribution. They're just not as connected as you think they are.

The discussion is always on upward mobility, but it seems to me it's quite possible to focus on downward mobility. Upward mobility seems to mean children switching career paths from their parents: the barkeeps son becomes a politician and eventually a million a year lobbyist, or the occupational structure changes (professional athletes start getting 1% salaries, IT IPOs). Downward mobility seems to me to be more individual, a matter of dissipation, poor judgment, or choice to abstain from the competition for money. But I suppose it could also reflect changes in occupational structure: CEO's of shoe and textile manufacturers no longer earn what they did in the 50's.

Clark has always argued that downward social mobility was what made the industrial revolution happen in the UK, because downward mobility enriches the lower ranks of society.

I recall reading someone who made the claim that the Black Plague laid the groundwork for the enlightenment and Industrial Revolution. There were simply no enough people left to sustain feudalism. The ruling elite had to pluck from the lower classes to fill their ranks, thus refreshing the smart fraction. I'm not sure I buy it, but it is a nice story.

The black plague damaged feudelism in Europe beyond repair in the 12-1300s. In combination with the development of the horse collar that made people able to farm 50% more land the limited labor of the farms allowed for a vast change in the power dynamics of the old system.

You may be running your events together too much for any causal relevence - the Black Death is halfway through the 14th Century (1348) and the horse collar was already in general use in the 11th C (1000s).

"Downward mobility seems to me to be more individual" - that's my observation, in a family I know well.

"... a matter of dissipation, poor judgment, or choice to abstain from the competition for money." No: in that same family it sticks out as being a matter of IQ and of health.

Clark is understandably shy of making bold pronouncements about causal factors in this area, but it seems pretty clear to me that he's a direct intellectual descendent of Francis Galton.

What have the direct intellectual descendants of Francis Galton, such as Pearson, Fischer, Keynes, Louis Terman, Fred Terman, Wright, Haldane, Shockley, Hamilton, Wilson, Trivers, and Dawkins, ever done?

Communist regimes discriminated very strongly against "black" family background people, while energetically seeking children with potential from the lumpen and the underclass. I cant imagine that these massive efforts had no effect and that descendants of the Russian aristocracy were allowed to study and are back in leading positions in society.

In China, studies based on family names are useless and misleading.

On the other hand German Civil Service is full of "von" and other "aristocrat families". Like current minister of defense. But I think most are fake.

You're correct that the Soviet Union tried to keep the old aristocracy down. What is surprising was how hard a time they had of it.

In Poland as well.

It turns out that man is not a formless hunk of clay that can be molded by the right laws and right rulers. Maybe those population geneticists are onto something after all.

People (generic "people") seem to be confusing mobility and inequality. If the distribution of income is more uniform, mobiliity is less important (i.e., has a smaller effect on actual outcomes). In many of these papers, how close the income quintiles are to each other is not part of the analysis.

Of course parents matter. Just ask the kids who where unfortunate enough to be born in Bangladesh. Who is responsible for bringing their economic status up to even the lowest rungs of US society. Where exactly does ones responsibilities to income equality begin and end. Obtaining income equality across the word would require US citizens to transfer 98% or more of their wealth to others in the world. Just asking how many people are willing to do that in the US.

You poor naive soul. It's not about ending inequality, it's about using it as a bludgeon to implement policies progressives have been pushing since long before inequality was the cause of the moment.

It may be about building a case to grab some of the stuff from the rich.

State lottery ticket buyer seem not to be so concerned about inequality seeing that when the pot goes up they buy more tickets.

Obtaining income equality across the word would require US citizens to transfer 98% or more of their wealth to others in the world. Just asking how many people are willing to do that in the US.

How much of that 98% is held by 1% of American society? If you're equalizing, some are going to drop a larger percentage than others...

It seems that no discussion of socioeconomic mobility addresses what ideal mobility would look like. The studies all say that x percentage of people move up but they don't compare it to anything except past numbers. Additionally, for every person who moves up, someone has to move down (in the ranking at least, not so much in absolute terms), so even "perfect" mobility would leave us with the same number of people in each quintile, which would certainly be construed as a bad outcome by people who don't understand math... The reason mobility matters is that we want people who deserve to move up to move up, but that requires deciding who "deserves" to move and how to measure that (and then comparing actual mobility to this new ideal mobility number). I think it matters tremendously how many people actually take concrete steps in their lives to move up the ladder. After all, everyone "wants" to move up the ladder but if someone is perfectly happy earning $10/hour (perhaps just my hypothetical person, but still) and never even attempts to earn more, it doesn't strike me as unjust that he/she doesn't move up the ladder.

This is true in the US as well.

Our most successful students, our engineers, and doctors today have the surnames Chen, Mohammed, and Patel.

Anyone who points out that the most successful of one hundred years ago in this country had different surnames is obviously a racist, misogynistic homophobe.

To what extent is social mobility tied to geographic mobility? (Which doesn't imply causation- those who move to improve their fortunes may be more motivated- and perhaps more talented- than those who don't.).

But just a different environment (and out of one's comfort zone) might encourage people to consider choices they would never have thought about in the old neighborhood.

People of lesser means should go (Gestation) Galt and not have any more kids.

It's already happening for people in the "over" middle in very many countries, so why not accelerate the process with some sort of propaganda campaign?

Firstly, when people are interested in social mobility then yes, of course they are interested in whether individuals can climb, not surname groups. So it is odd to me for Clark to say that essentially "We know social mobility is lower and less variable across societies because surnames do not rise and fall that much across societies".

Secondly, seems like the reification of whatever allows a family to climb in social status as high social competence, as if it were some f*cking innate quality in all societies, for sure, rather than, ooh, let's think, a chance advantage due to circumstance picking up some cultural speciality or resource (which gradually gets eroded over time) is rather loathsome.

This is something certain classes of reviewers seem to be keen on, while Clark seems sensibly agnostic, and to address this pretty fairly in his interview -

"Why is social status so sticky? This is intellectually the most intriguing part of the story, and it's one that's hard to make a lot of progress on. It is clear that families are very powerful determinants of children’s outcomes. But what do parents transmit to their children? Is it mainly some type of culture? Or is it mainly genetics? The data does not exist to provide any conclusive answer to this question."

I expect there the answer is different at any time and place - social mobility (the max and min) is probably limited by common structural factors of sophisticated societies beyond which they could not operate well, but how people actual divvy up the spoils of what social mobility there is, that probably has no fixed rule (i.e. it is not always "meritocratic" in the sense of favoring people with the best "natural" physical and mental health and ability).

I also like that the interview makes clear the difference between the distribution, relative position and real inequality, absolute position.

what about selection bias?

Comments for this post are closed