What do twin adoption studies show?

by on February 20, 2011 at 7:30 am in Education, Political Science, Religion, Science | Permalink

Here is an accelerated version of a conversation I had with Bryan Caplan, noting that I do not pretend to be representing his views but rather I am writing in broad brush terms. 

"Culture" and "genes" are two major factors determining individual outcomes, toss in parenting, and if you wish call parenting and culture two parts of "environment."  It is obvious that culture matters a great deal, and this comes from knowledge which existed prior to rigorous behavioral genetic studies.

I say "soda" and people in Nebraska say "pop."  Singapore vs. southern China.  German musical tastes in 1780 vs. today.  Rural Africa vs. urban Africa.  Most concretely, if I meet someone I want to know what country he came from and grew up in; in fact that is the first thing I wish to know.  "The culture word" may be overused and abused, but still the power of culture is evident.

If twin adoption studies seem to show that parenting does not matter much, I think:

1. Matter for what and for whom?  Parenting matters a lot for language and religion and obedience and also one's sense of "how the world works," and those factors matter to parents even if they don't always matter to researchers and economists.  The word "matters" is going to carry real weight here; in my admittedly extreme pluralist view, "doesn't affect adult income" does not translate into "does not matter."  Ask yourself how easily you could marry someone from a very foreign culture -- it matters.

2. We already know that culture matters a great deal in shaping what kind of adults children become, but often individual families cannot much affect the broader culture a child is raised in.  It's sometimes the individual family which is impotent, not the surrounding culture as a whole.

3. Most parents are deep conformists.  There isn't always a lot of cross-sectional variation in adoption studies.  Even if most parenting strategies don't matter (if only because they are not varying much), if a child is raised in an Orthodox Jewish family, or in a strict American-Chinese family, or among the Amish, that probably matters, even adjusting for genes. 

The adoption studies can be showing that a) most parents don't so much shape a child's culture at the margin, or b) that environment doesn't much matter in light of the power of genes.

I see a lot of leaping at b), when to me a broader look at the social sciences -- and also some personal travel -- shows a) to be important.

While I tend to see genetic influences as very powerful, the adoption studies don't push me any extra mile in this regard.  In general, I see some proponents of the adoption studies as putting them forward as the most convincing form of evidence, and mentally working to downgrade the relevance of other forms of evidence, including the vast body of knowledge in sociology and social anthropology.  I would prefer a more pluralistic approach which draws in many bodies of knowledge at the beginning, admittedly with a messier overall understanding, but that is a good rather than a bad thing.

Now to speak on behalf of Bryan for a moment, and in a way he might object to, I wonder if he isn't leaping at b) too quickly, rather than weighing a) vs. b), using evidence which does and indeed cannot not come from twin adoption studies.  The adoption studies don't actually adjudicate a) vs. b) and they don't rule out great power for the broader environment. 

Another way to put the point is to examine Judith Harris's claim that genes and peer selection are what shape children.  Is the claim about peers -- which falls out of the statistics -- causal?  Or are peers best thought of as a sufficient statistic for the broader surrounding culture, thereby placing the causal force in that culture?

If it is the latter, the Harris evidence is simply showing, once again, that both genes and culture matter.  Which is fine, but it's hardly a revelation.  It also leaves open the possibility that parents who wish to influence their children simply need to try harder to shape their surrounding culture.  The Amish are not the only ones who succeed in that endeavor, even if most people do not succeed or wish to try very hard.

You will find Bryan's views on these issues in his forthcoming book.  Other readings on what twin studies mean are here (pdf), here, here (Bryan's twins spend a lot of time together!), and here (pdf, a very good piece by Kamin and Goldberger).  My favorite bit is the very last paragraph, which I will put under the fold:

"A case in point is provided by the recent study of regular tobacco use among SATSA's twins (24). Heritability was estimated as 60% for men, only 20% for women. Separate analyses were then performed for three distinct age cohorts. For men, the heritability estimates were nearly identical for each cohort. But for women, heritability increased from zero for those born between 1910 and 1924, to 21% for those in the 1925-39 birth cohort, to 64% for the 1940-58 cohort. The authors suggested that the most plausible explanation for this finding was that "a reduction in the social restrictions on smoking in women in Sweden as the 20th century progressed permitted genetic factors increasing the risk for regular tobacco use to express themselves." If purportedly genetic factors can be so readily suppressed by social restrictions, one must ask the question, "For what conceivable purpose is the phenotypic variance being allocated?" This question is not addressed seriously by MISTRA or SATSA. The numbers, and the associated modeling, appear to be ends in themselves."

George February 20, 2011 at 5:21 am

For comparing the effect of parents versus surrounding culture (e.g. peers), it might be worthwhile comparing kids who went through homeschool vs. those who went through public school. My sense is that there are systematic differences between our homeschool children and their homeschool friends relative to the surrounding culture that result from parental choice. Then again, homeschool kids are in many ways similar to each other–they may simply have a different peer group.

If that's true, then perhaps it's a mistake to refer to a monolithic culture. If we break down that assumption we can conceptualize parental influence as a matter of deciding what sub-culture (or mix of sub-cultures) predominates at home. That sub-culture (including the relevant media sources, ritual practices, other families and peers in that sub-culture, etc.) then shapes the child.

Andrew February 20, 2011 at 5:22 am

I don't believe them. For one, the pre-natal environment is a huge issue. It may be THE issue. If they ignored that, what other problems have they not foreseen?

Jim February 20, 2011 at 5:39 am

I'm with Judith: when evaluating the question of how parents vs culture affect adult personality, we should throw out everything but the twin studies. There is no way to do basic science without the control.

The corporal punishment example is illustrative. Violent parents should pass that tendency on genetically too, so you can't just say " kids learn beating from their kids."

If parerenting is so powerful it should show up clearly in the twin studies. It does not.

Shylock Holmes February 20, 2011 at 6:28 am

It seems very likely to me that the effects of parenting are highly asymmetric with parent quality. My guess is that the the difference in outcomes between an average parent and a parent that beats or sexually abuses their child is way WAY bigger than the difference between an average parent and a superstar parent. It's a lot easier to break a child than to fix one.

Related to your point about how much variation in parenting in the twin studies, I don't know how many left tail parenting events they're going to be observing – the worst possible parents are hopefully screened out by the adoption agencies.

mk February 20, 2011 at 7:00 am

I'm no expert on these adoption studies, but is it really true that they've never tried to assess correlations between "reasonable proxies for parent culture" and child outcomes?

For example, if I was designing a questionnaire one of the first things I'd put on there is "did your parents make you study >3 hours a day?" While it's only one question, isn't that a pretty reasonable signal for certain subcultures? (For example, just go through the Tiger mom book and make questions based on that.)

I think I agree with Tyler's comments, but I'm surprised about what they seem to suggest about the research (which I haven't read).

Tommie February 20, 2011 at 7:12 am

For practical purposes the answer to the nature v nurture question depends on the intended audience. Harris is read by US parents who want to know whether marginal investment in shaping their childrens' opportunities and behavior produces significant benefits on a range of indicators. That is a very different question than debating whether provision of minimal medical care in poor villages in Africa improves scores on cognitive tests.

TGGP February 20, 2011 at 7:29 am

Language is in the fact area that primarily inspired Judith Harris: kids learn the language of their peers (including its accent), while their parents can take years and never fully adopt it.

Also, twin studies are old hat. The future is genotyping siblings.

Tony February 20, 2011 at 7:54 am

Just to highlight the post above from Thomas. For some "middle minorities" like Ismailis, having a peer culture not one's own is more normal than exceptional.

As well as the national culture and the peer culture, in families with strong and longer extended family histories, there's also a tribal culture. I've occasionally heard Switzerland called "tribal" by outsiders.

ad*m February 20, 2011 at 8:43 am

I like this blog. However, professor Cowen is out of his depth with this post, and a major cause is his assumption or bias that culture has no genetic origin.

It seems to be standard in economics to make all sorts of broad assumptions about people so that their behavior can be captured in nice, linear equations.

However, the most important and hardest thing in the hard sciences is being aware of assumptions and bias and eliminate these as much as possible.

A few examples:

"Most concretely, if I meet someone I want to know what country he came from and grew up in; in fact that is the first thing I wish to know."

The assumption that culture is dissociated from genetic background is false for the large majority of the world's population. For someone like Cowen, from a multiethnic/'multi-cultural' society like the U.S., it may be difficult to realize that the large majority of people in the world live in highly monoethnic societies, making it almost impossible to separate cultural from genetic background.

For example, when Cowen meets someone who looks genetically 'Chinese', there is a > 1/1.04 chance that person grew up in China, and a < 0.04/1.04 chance the person grew up outside Chinese culture (there are 40 million ethnic Chinese living outside China, and more than 1 billion living in China).

So when Cowen learns someone is from China (or Indonesian or France or Poland etc), he is primarily being informed about their genetic background. Only in a small minority of people he meets would their culture be different from their genetic background. But Cowen thinks he is informed about their culture, and the probability of him meeting counterexamples where genetic background does not equate cultural background is small.

"I would prefer a more pluralistic approach which draws in many bodies of knowledge at the beginning, admittedly with a messier overall understanding, but that is a good rather than a bad thing."

Missing Occam's razor here. It mostly shows that he, a New York Times columnist after all, does not want to face the results of twin and other genetic studies.

We are discussing the association of innovativeness, culture, income, IQ, etc with genetics through twin and adoption studies.

50 or 100 years ago, discussing these same subjects, we would have brought forward their association with class and income in a Marxist or pseduo-Marxist framework.

But today, developed countries are so succesful at eliminating most environmental associations, such as malnourishment with IQ, that the genetic associations are almost all that is left to explain variance.

In Dickens' London, we might have found income explaining 80% of children's IQ variance, and genetics only 20% – while now, with the lowest incomes eliminated, income explains only 10% and genetics 80%.

"If purportedly genetic factors can be so readily suppressed by social restrictions, one must ask the question, "For what conceivable purpose is the phenotypic variance being allocated?""

Eh? Lamarckian thinking here. The genetic variance was always there, these results just shows that men and women are social animals and that laws and culture work up to a point in guiding behavior. The Scandinavian countries have very strict laws on the use of alcohol. That does not mean the propensity for heavy drinking is suppressed – as is easily shown <a href="http://www.thelocal.se/25870/20100402/">what happens when Scandinavians travel to countries with more liberal liquor laws.

If the IQ lowering effects of a genetic disease like Phenylketonuria can be eliminated through a dietary change, this does not show that PKU is not genetic in origin, it just shows how wonderful human inventiveness is.

The pernicious effect of this post and others like it is that the associations of genetics with what tends to make people uncomfortable cannot be studied.

This has stopped innovation in these fields in its tracks, innovation that might help ameliorate the effects of genetics where these are bad, as in the case of PKU.

(the problem with long comments like this is that other commenters may have addressed some of these points already, my apologies)

Bill Harshaw February 20, 2011 at 9:05 am

How much does culture vary as distance from the central city varies? Do parents control their culture by their selection of where they buy their house, or rent? Or is that just controlling education, not culture?

And on a side note, last week I saw one of those international comparisons of metrics, in which Israel ranked below the US on education test scores. What's that about? Are Israel's schools dominated by offspring of the Orthodox, or the Israeli Arabs, or recent immigrants?

Tom February 20, 2011 at 10:53 am

To some extent we can *choose*, even if we cannot *shape*, our surrounding culture.

I think the two are essentially synonymous for purposes of this discussion. The most common attempt being choosing a private school on the hope of changing the child's peers.

Steve Sailer February 20, 2011 at 2:14 pm

I mostly agree with Tyler. I made similar criticisms of Judith Rich Harris's overreaching in my 1998 review in National Review of her book The Nurture Assumption:

"In contrast, her third assertion — parents don't matter — is plausible only within her narrow, arbitrary boundaries. To fully explain human behavior, everything matters. Anything conceivable (whether genes, peers, parents, cousins, teachers, TV, incest abuse, martial arts, breastfeeding, prenatal environment, etc.) influences something (whether personality, IQ, sexual orientation, culture, morals, job skills, etc.) in somebody.

"To show that peers outweigh parents, she repeatedly cites Darwinian linguist Pinker's work on how young immigrant kids automatically develop the accents of their playmates, not their parents. True, but there's more to life than language. Not until p. 191 does she admit — in a footnote — that immigrant parents do pass down home-based aspects of their culture like cuisine, since kids don't learn to cook from their friends. (How about attitudes toward housekeeping, charity, courtesy, wife-beating, and child-rearing itself?) Not until p. 330 does she recall something else where peers don't much matter: religion! Worse, she never notices what Thomas Sowell has voluminously documented in his accounts of ethnic economic specialization. It's parents and relatives who pass on both specific occupations (e.g., Italians and marble-cutting or Cambodians and donut-making) and general attitudes toward hard work, thrift, and entrepreneurship.

"Nor can peers account for social change among young children, such as the current switch from football to soccer, since preteen peer groups are intensely conservative. (Some playground games have been passed down since Roman times). Even more so, the trend toward having little girls play soccer and other cootie-infested boys sports did not, rest assured, originate among peer groups of little girls. That was primarily their dads' idea, especially sports-crazed dads without sons.

"While millions of parents sweat and save to get their kids into neighborhoods and schools offering better peer groups, Mrs. Harris redefines this merely as an "indirect" parental influence. She claims modern studies can't find predictable relationships between "direct" influences (i.e., different child-rearing styles) and how children turn out. But that may be merely an inherent shortcoming of these non-experimental analyses. For example, she asserts (not necessarily reliably) that studies prove it doesn't matter whether mothers work or not. But the same methodology would report that it doesn't matter whether you buy a minivan or a Miata, since purchasers of different classes of vehicles report roughly similar satisfaction. In reality, women don't randomly choose home or work; they agonize over balancing career and family. They tailor their family size to fit their career ambitions and vice-versa. Mothers then readjust as necessary to best meet their particular families' conflicting needs for money and mothering. For instance, a working mother might quit when her second baby proves unexpectedly colicky, then return when the children enter school, then shift to part time after her husband gets a big raise. That's bad for these studies, but good for their kids."

Mercy February 20, 2011 at 3:54 pm

"Culture" and "genes" are two major factors determining individual outcomes, toss in parenting, and if you wish call parenting and culture two parts of "environment."

There's an important qualifiers missing here: in the modern day first world. There is a third factor of physical environment which is important in any environment where malnutrition, heavy metal poisoning or aflatoxins are prevalent, possibly much more important than the two factors you cite if the Flynn effect is any judge. Twin studies necessarily control for these effects: you don't adopt if you can't afford to feed your kids.

Tracy W February 21, 2011 at 12:33 am

Is the claim about peers — which falls out of the statistics — causal?

Yes, it's a causal claim. And there's some studies she has found supporting it. They are, from memory:
– A study of elementary-age children who shifted their group of friends during a school year, finding that their attitudes towards things like homework changed towards that of the group they moved into. It seems unlikely that parenting or teacher attitudes changed during a school year.
– A study of Japanese executives who were posted to the US for a couple of years, with their families. The children of these families went to American schools, the parents knew they were going back to Japan so had no incentive for their children to acclimatise to American culture, but the children's behaviour did shift towards American norms, then back towards Japanese norms when they moved back to Japan (but the shift was easier based on age).
– That Hebrew took off as a revived, non-academic, language once the motivated parents sent their kids to schools where Hebrew was the only language they had in common, while Esperanto has never taken off.
– That tall men earn more is widely known, but a study by a couple of economists found that earning more was even more highly correlated with height at age 16 than with height as an adult, implying that it was the behaviours men formed with being unusually tall as a kid that employers were paying for.

But as Judith Harris notes, it's not a hypothesis that's been thoroughly tested, it needs more research. She's well aware of the dangers of over-confidence in science.

Tracy W February 21, 2011 at 12:55 am

Steve Sailor: "In contrast, her third assertion — parents don't matter "

It's wise to read a book before criticising it. Judith Harris specifically did not claim that parents don't matter, and is annoyed by newspaper articles claiming she did.

She also doesn't assert that "studies prove it doesn't matter whether mothers work or not", she asserts that studies haven't shown that it matters whether mothers work or not. Judith Harris is defending the null hypothesis, that there's no link, as the one that should be believed until data is produced showing that there is a link.

You say: "Worse, she never notices what Thomas Sowell has voluminously documented in his accounts of ethnic economic specialization. It's parents and relatives who pass on both specific occupations … and general attitudes"
I think that Judith Harris's response would be that how do you know that the parents were causally important? How did Sowell control for the possibility that it was the children's peer group (which in many cultures includes a lot of cousins and younger aunts and uncles) rather than their parents?

albatross February 21, 2011 at 6:29 am


Yeah, that's an important point.

Imagine we have a three-dimensional graph–the z axis is child outcome measured somehow (IQ, income, crime rate). The x axis is parental input, the y axis is environmental input.

Those twin studies tell us something about the partial derivative dz/dx in a particular small range of the space, where both x and y are fairly high thanks to being first-world parents who were also pre-screened for income and quality. It's really hard to use that to say much about dz/dx at some other part of the space, where parental input and environmental input are both much lower.

In statistical terms, this kind of study tells you about what happens in the middle of the distribution, but probably not in the tails. How big an impact does it have on measurable outcomes of kids to go from Pap Finn to Amy Chua? As far as I can tell as a non-expert, we don't really have data on that.

Geoffrey February 21, 2011 at 6:59 am

Holy Moley – cat, meet pigeons!

I think I was a lot more impressed by Judith Harris' work before the really interesting studies of the long term effects of the Tennessee STAR project came out – essentially, we really had no evidence that teachers or class sizes made any differences until we started fully randomly mixing people and studying the results over many decades. The twins studies purport to do this, but the mixing is far from random (Cosma Shalizi has some fascinating points on this) and more importantly, we don't have the advantage of tracking 50-100 children per parent, the way we do with the teachers in the STAR data, thus we have no way of really getting our arms around the impact of individual parents.

Tracy W February 21, 2011 at 8:32 am

Geoffrey – why would the STAR study make you discount Judith Harris's theorising about parenting effects or the lack thereof? If it takes studies of 50-100 kids per parent to show the impact of individual parents, then that impact can't be very large or very important.

Finch – perhaps the twin and adoption studies mostly haven't been identifying the children's peer groups, so this influence gets lost? Judith Harris does discuss a Dutch study, I think, looking at adopted children that found that, amongst adopted children whose biological parents had criminal records, the area in which the kids were bought up was more strongly correlated with the children later having a criminal record than the adoptive parents' criminal record.

ohwilleke February 21, 2011 at 11:07 am

Twin studies and related heritability research tries to fit variation into "hereditary," "shared environment" and "non-shared environment." The broad general finding is that the non-hereditary component of variation is overwhelmingly due to "non-shared environment." But, assuming that low levels of "shared environment" translates into low parental impact, may be inaccurate. It could be that parental impact is potentially great but not universally utilized.

Where does that come from? Mostly it comes from non-twin siblings not having any more similarity than the hereditary component of twin relationships would suggest. Put another way, if non-twin siblings adopted into different families we wouldn't expect them to be more different than non-twin siblings raised in the same family.

A possibily measurement based source of this is that adoptive families and first world, non-immigrant, middle class families that end up in these studies are "abundant" and "permissive". They are "abundant" in the sense that they are not subject to the environmental deprivations that make IQ more environmental than hereditary in the poor than in the rich. They are "permissive" in that middle class adoptive parents (and American middle class parents generally) may feel less of a perogative and right to try to force a child away from a "natural" tendency to develop a certain phenotype.

Also, adopted familes, in general, may have a great deal in common with each other in terms of values and parenting styles. After all, they are expressly screened for eligibility on those measures.

In contrast, one might expect the poor generally, to suffer high levels of environmental deprivation that prevents people from developing their natural tendencies (e.g. someone with strong musical or sports talent might never know that until it is too late), and one might expect particularly non-permissive parenting for children who are being raised to fill some particularly role – like a first born son of a Japanese business owner being raised to carry on the family business, or someome born into a rigid caste society.

One possible way to measure the amount of potential parental influence is to look a situations like the differential intermarriage rates of second generation immigrant men v. second generation immigrant women. In the U.S. the boys are often under strong pressure to retain culture, while the girls are not (if you doubt this look, for example, at gender and ethnicity specific name lists where the top names of Hispanic boys in the U.S. are typically very Latin, while names like Jane and Sarah top the list for girls). Similarly, boys intermarry in the first generation much less often than girls, perhaps because parents see traditional cultures as offering benefits for boys but American culture as superior for the prospects of girls (in other words, immigrant parents may believe as their daughters do that traditional homeland husbands are jerks towards their wives). The models used for heredity read this as "non-shared environment" (girl v. boy), but another way to look at it is "potential parental impact" which is exercised for some children, but not for others, even in the same family. Likewise, I wouldn't be surprised to see higher parental impact for earlier birth order children than for later birth order children since may key succession goals have a primogeniture aspect — only one child will carry on the family business.

Tracy W February 21, 2011 at 12:39 pm

one might expect particularly non-permissive parenting for children who are being raised to fill some particularly role

But why would you expect the children in question to fall in with that role? For a start, children are not clones of their parents, the combination of genes we each might inherit, including recessive ones that never expressed themselves in either Mum or Dad, affect what we can do with our lives. If Dad was big and intimidating and you're short and skinny, it might be a good idea to learn how to talk your way out of trouble, no matter what Dad has to say about it. Furthermore, when genes are inherited, the sorts of parents who have the strong-willed genes to force things on their kids can be the sort of parents who have strong-willed children who refuse to let themselves be forced. And outright defiance is not the only response a child can make to their parents, another option is to follow their rules in their house and do whatever you darn well please outside it.

Secondly, the world might well have changed from when your parents were growing up. A friend might be naturally better in the group at what your Dad did than you can be, and thus you need to develop a niche somehow else. Most traditional cultures practice out-marriage, so one of your parents is quite possibly not from your home-town and thus perhaps not the best person to copy from. Perhaps the chief of the village in your Dad's youth liked unhesistating obedience while the current chief likes someone who stands up to him. Even in societies where arranged marriages are the norm, your parents are unlikely to find an exact clone of the opposite-sex one of them, so you'll need to figure out how to get along with someone different to mum/dad.

Similarly, boys intermarry in the first generation much less often than girls, perhaps because parents see traditional cultures as offering benefits for boys but American culture as superior for the prospects of girls (in other words, immigrant parents may believe as their daughters do that traditional homeland husbands are jerks towards their wives).

Or perhaps boys and girls both rationally look at the options and make up their minds which is better for them, and would do so the same way even if their parents didn't give a damn one way or another. I remember when I went to university and got my first email account, and was convinced that this was world-changing and that all my friends and relatives should get one right away, and browbeat them into it. I certainly didn't get that from my parents, they didn't know about email until I started telling them. Why wouldn't a immigrant girl see men from her culture acting like jerks, American men opening doors for her, do a brief cost-benefit analysis and decide "hey, I want me one of those"?

Comments on this entry are closed.

Previous post:

Next post: