Bryan Caplan  

Heritabilities Are Meaningful and Important

PRINT
The New Leisure... Progressive Nudges for Househo...
The punchline of most twin and adoption studies is an estimate of a trait's "heritability."  Heritability, usually written h2, is the fraction of variance explained by heredity.  Twin and adoption studies also estimate the fraction of a trait's variance explained by shared family environment, usually written c2.  Literally hundreds of published papers go to great lengths to estimate h2 and c2

Given the effort researchers apply to measure these variances, you'd think they were measuring something worth knowing.  But the more technically proficient critics of behavioral genetics often say otherwise.  Latest example: In the Huffington Post, Scott Barry Kaufman tells us that "The Actual Heritability Value Simply Does Not Matter."  Kaufman's main argument:
[O]ur understanding of the factors that contribute to the development of human traits in general -- and to IQ in particular -- is currently so deficient that we typically do not know if the environmental factors important in the development of a particular trait are stable across testing situations, vary somewhat across those situations, or vary wildly across those situations.
If we took Kaufman's argument seriously, it wouldn't just discredit twin and adoption research; it would discredit all of social science.  Given any social science result - or experiment! - you can always say, "Maybe the world changes so chaotically that your results are no longer valid."  The only reply to such sweeping agnosticism is to stop talking and bet.  My money says, for example, that the average adult IQ heritability estimate published in 2020 will exceed .5.  If Kaufman's isn't convinced, I'm happy to take his money.

To be fair, though, there's a narrower and more plausible complaint about heritability estimates: They're an answer to a question that no one asked.  Who cares if "heredity explains 80% of the variance of adult IQ"?  What does that mean in English?  I doubt even most practicing researchers have a compelling answer to this challenge.

Nevertheless, a compelling answer does exist.  If you delve deeply into the math of nature and nurture, you'll learn two striking results:

1. The expected correlation between identical twins raised apart equals h2.

2. The expected correlation between unrelated individuals raised together equals c2.

If you set up your study correctly, these are causal estimates that speak directly to the nature/nurture debate.  Suppose h2=.4.  Then if you were separated at birth from an identical twin, and you're 1 SD above average in a trait, you should still expect your twin to be .4 SDs above average in that trait.  Why?  Because of your shared genes. Or suppose c2=.1.  Then if you were raised with a random baby, and you're 1 SD above average in a trait, you should expect him to be only .1 SDs above average in that trait.  Why?  Because of your common upbringing.

Admittedly, these conclusions don't tell you which genes or which aspects of upbringing are responsible for these correlations.  But they do measure something both meaningful and important: the overall causal effects of genes and upbringing.  h2 and c2 aren't everything we'd like to know about nature and nurture.  Yet they're a huge advance over the preceding three thousand years of fruitless debate.


Comments and Sharing





TRACKBACKS (1 to date)
TrackBack URL: http://econlog.econlib.org/mt/mt-tb.cgi/4821
The author at Evolving Economics in a related article titled The heritability debate, again writes:
    Like the level of selection debate, the debate about what heritability means has a life of its own. The latest shot comes from Scott Barry Kaufman who argues (among other things) that: The heritability of a trait can vary from 0.00 to 1.00, depending o... [Tracked on April 1, 2011 7:33 AM]
COMMENTS (24 to date)
nazgulnarsil writes:

more specifically, you aren't supposed to talk about it in english. at least if you care about your career.

BV writes:

Is this how statistics are used in the social sciences?

Correlation refers to the amount of predictability of A given B. Zero correlation means A is not predictable given B, and a correlation of one means that A is completely predictable from B.

You make it sound as if the value of h^2 tells you what the value of the trait is going to be (such as IQ), rather than if its predictable.

Perhaps I am misunderstanding something?

Eric Falkenstein writes:

Heckman wrote a funny hit piece in the JEL after the Bell Curve came out, and argued that since IQ=f(genes,envir), not f(genes)+g(envir), it was meaningless to talk about the impact of genes. Never mind this would make almost all econometrics irrelevant.

He's now backed off on that and a bunch of other assertions in that piece.

Jason Malloy writes:

"Given any social science result - or experiment! - you can always say, "Maybe the world changes so chaotically that your results are no longer valid."

Exactly! Philosopher Neven Sesardic elaborates on this important point in Understanding Heritability.

All "behavioral genetics is worthless" arguments are unintentionally nihilistic. You can't destroy behavioral genetics but preserve anything else in the existing social sciences. Behavioral genetics is the most sophisticated set of methods in social science. So if you "prove" it is worthless or unscientific based on some idealist set of standards, you are inadvertently making a broader argument that all attempts at social science investigation are pseudoscientific.

Yet the legions of professional "genetic determinism" critics are never this consistent. Since they have a nurturist agenda, they need to reject certain kinds of evidence on "scientific" grounds, yet promote or accept other kinds of evidence supporting their preferred viewpoint. Their transparent hypocrisy is to pretend that the extreme scorched earth epistemological arguments they've advanced to undermine the legitimacy of behavioral genetics, doesn't apply to and completely undermine their own preferred evidences as well.

Jason Malloy writes:

Rather, Making Sense of Heritability:

"To attack behavior genetics with such a methodological weapon of mass destruction is a fallacy known as qui nimium probat nihil probat ("whoever proves too much, proves nothing")."

Troy Camplin writes:

Two people raised together do not have the same environment. When I was growing up, my home environment included my brother, but not me, while my brother's home environment included me, but not him. More than this, we did not live side-by-side having the same experiences. We had some common experiences, while others were different. I am older than my brother, so I had an early environment where I was a single child. My brother did not. On the other hand, when I went to college, my brother spent his high school years as a single child (Monday thru Friday, and many weekends, at least), while I did not. Also, the environment responds to the character of the child. I was more bullheaded, so was punished more than my brother. That makes for a different home environment. And this is not counting the influence of friends and classmates. And it's not counting the in utero environment, which affects brain development as well.

And then one gets to the problems of heritability, with the large number of genes involved in the brain's layout, functions, etc. This results in an extremely complex network. Left-handedness is a strongly heritable mental trait, caused by high levels of testosterone while the fetus is developing (in about half of lefties), yet only produces 10% "heritable" lefties (the other half are created through birth trauma). Heritable left-handedness runs in families, but not in any predictable pattern. Niether of my parents are left handed, but my brother and many of our cousins, and an uncle, are. My daughter is right handed (too learly to tell on the 19 mo. old son).

So on left handedness, for example, there is an inherited tendency for the fetus to produe more testosterone than usual, which can result in a switch to left handedness because of the high testosterone environment. Where is the difference between environment and inheritance?

And before you argue that uterine environment is difference, let me point out that our environments are really nothing more than a soups of chemicals, photons, and sound/pressure waves interpreted by our senses. The uterus is an environment just as much as your livingroom.

Tracy W writes:

Troy, I'm not sure what point you're trying to make with your argument.
Everyone sensible knows that both genes and environment combine to create a human being. To take the extremes:
- no food, no water, and it doens't matter how great your genes are, you're dead;
- take a rock, supply it with all the food and water and all the love and attention and anything else you think should be lavished on a human baby, and ten years later, all you'll have is a rock. Genes are needed to convert environmental inputs into something.

That said, would you disagree that, for a normal child, which language they learn to speak, given that they have the genes that allow them to learn and speak a language, is more driven by environment, which blood group they have is more driven by genes than the environment?

Or take the case of left-handedness. As you point out, a bunch of research has been done into that, and, if your summary is right (it matches with what I've read, and as a left-hander myself I have a mild interest), researchers have found it to be because of an inherited tendency for a fetus to produce more testosterone than usual, so gene-environment effects. This is a case of research disentangling how connections are formed.

That two people raised together don't share the same environment is well known, and one way researchers try to figure out the influence of parenting is by looking at differences in adult personality based on birth order, because a first-born child has a different home environment to a second-born, but we know of no reason to think that the gene mixing would differ depending on birth order, but, once you control for socio-economic effects (richer people tend to have smaller families) no consistent results show up in adult personality. This is one piece of evidence that what parents do isn't that important in forming adult personality.

Of course all this information is incomplete, and we never know what new findings are going to show up and overturn everything we think we knew, but all we can do is the best we can with the information we have to hand now.

Stefano writes:
Who cares if "heredity explains 80% of the variance of adult IQ"? What does that mean in English? I doubt even most practicing researchers have a compelling answer to this challenge.

Well, some people in the past had a very clear idea of what that would mean, namely that low-IQ people shouldn't be allowed to reproduce. For the good of the Nation, of course.


What is more interesting, in my view, are the value of c2 of specific nurture practices. For example, how much gain in IQ can be obtained using a kind or another of day care, or breast feeding, or small school classes? What are the costs/benefits? Is it worthwhile? Should the society as a whole subsidize it?

Petrik runst writes:

In the evolution if childhood by konner, which mostly supports the claim that 50% of the variation is explained by heredity it says that there is one important exception: attschment and anxiety in young children. The one twin study done so far found ZERO effect of genetics. So that serves as some mild evidence against the hypothesis that parents do too much worry.

Troy Camplin writes:

Tracy,

My point is that environmental and genetic interactions are far more complex than we imagine. There are things, like blood type or eye color, that are purely genetic. But the immune system and the structure of the brain is very much determined by both genes and environment simultaneously. The example you give of language is actually a good example. We have a genetic propensity to language, but not to learn a particular language. However, there is some evidence that particular languages map on to different brain regions, and that different languages affect to some degree how we think. That would occur only by affecting the brain's structure. And that is in part affected by genes. The genes allow for a range of ways of thinking, but language affects what happens within that range.

The issue of personality as being more determined by genes than parents is often brought up as proof that parenting doesn't matter that much. But personality has never seemed to me to be that important, or something many parents have been that interested in affecting (perhaps because our children act like us, and we think we're just fine). My daughter is into books because she is raised in a book-rich environment. There can be no genetic propensity for loving books per se.

Tracy W writes:

My point is that environmental and genetic interactions are far more complex than we imagine.

A rather sweeping statement. How do you know how complex my imagination is? Environmental and genetic interactions might be more complex than you imagine, while still being less complex than I imagine. (To be honest, until you said this, I would have guessed that you were the more imaginative of the two of us, but even if I am less imaginative than you, that still leaves a lot of people on the planet who could be more imaginative than either of us).


However, there is some evidence that particular languages map on to different brain regions, and that different languages affect to some degree how we think. That would occur only by affecting the brain's structure.

As everything we learn has to affect our brain structure, this is not a very impressive statement. My memory of what I had for breakfast affects my brain's structure. (Dualists may disagree, but that's another argument).

The genes allow for a range of ways of thinking, but language affects what happens within that range.

Correction: The genes allow for a range of ways of thinking, and language affects what happens within that range.
It's not that there's a conflict between genes and environment, it's that the two combine together to result in us. Even in the case of blood types - no food, no blood of any type.

The issue of personality as being more determined by genes than parents is often brought up as proof that parenting doesn't matter that much. But personality has never seemed to me to be that important, or something many parents have been that interested in affecting (perhaps because our children act like us, and we think we're just fine).

As Bryan Caplan points out, the lack of effects of specific parenting behaviours, once you control for genetic relationships, is far broader than just personality. It seems implausible that parents don't care how much their children earn as adults, or whether they acquire a cirminal record, or how religious they are. And it's telling that where parenting behaviours can have major effects is by their ability to affect their childrens' peer groups, eg moving to France and sending your kids to a French-speaking school.

My daughter is into books because she is raised in a book-rich environment. There can be no genetic propensity for loving books per se.

This example I think indicates that my imagined environmental and genetic interactions are more complex than what you imagine. I agree with you that there can be no genetic prospensity for loving books per se. My ability however to imagine complex environmental and genetic interactions however suggests to me that there might be a genetic prospensity to be able to tolerate staying in one place for long periods of time, a genetic prospensity to enjoy the structure of stories, and to be attracted by new ideas.
Likewise, I don't imagine a genetic prospensity to engage in criminal behaviour, more the combination of a genetic prospensity for risk-taking, a genetic prospensity for excitement and a disprospensity for taking into account long-term effects of behaviour, a genetic prospensity for aggression. (And of course, some environments can be so rule-filled as to turn virtually everyone into criminals, eg as I understand it, in Cambodia under the Khmer Rogue everyone had to break the rationing rules to survive, genes and environment are both vital.)
Of course, just because I can imagine more complex gene-environment interactions than you doesn't mean that my imaginations are complex enough. Possibly no one's imaginations are complex enough. But we're not exactly at quantum mechanics level of complexity here yet, I think to say that we can't imagine this is prematurely defeatist.

hacs writes:

"If you set up your study correctly, these are causal estimates that speak directly to the nature/nurture debate. Suppose h2=.4. Then if you were separated at birth from an identical twin, and you're 1 SD above average in a trait, you should still expect your twin to be .4 SDs above average in that trait. Why? Because of your shared genes. Or suppose c2=.1. Then if you were raised with a random baby, and you're 1 SD above average in a trait, you should expect him to be only .1 SDs above average in that trait. Why? Because of your common upbringing."

This is the point Bryan. The implicit hypothesis is that there is a T (trait intensity; whatever), a fixed number (or some moment of some stationary stochastic process), which is estimable from data. But, there are some experiments saying that T is a function (or some moment of a non stationary stochastic process) changing during the lifetime according to the environmental history. Of course, parents "determine" the function (non stationary stochastic process), but it is weaker than usually thought.

hacs writes:

Moreover, it seems to me that another implicit hypothesis is assumed in those experiments. The parents know who are their biological offspring, and who are not. Are you completely sure about the irrelevance of this fact?

Tracy W writes:

Hacs - if the implicit hypothesis you refer to was false then estimates of inheritance would be biased downwards, by the number of children who were thought to be biological, but aren't (I presume the number of people who unknowingly adopt their own biological children is low). So therefore Caplan's figures would underestimate the importance of inheritance.

Troy Camplin writes:

Tracy,

That was a pretty ungenerous reading of what I said about gene-environment interactions being more complex than we can imagine. And yes, they are more complex than even you can imagine. How do I know? Because we haven't teased out all of the gene to cell to environment to cell to gene to cell to environment to cell, etc. bipolar (positive and negative) feedback and feedforward interactions. For example, did you know that your gut bacteria affect behavior? Or that there are environments I experience whose effects can be passed on to my offspring through epigenetics? And what about the role of mirror neurons, which cause us to mimic those around us, and make us prefer to mimic those around us almost perfectly? There are 30,000 genes, a third of which are dedicated to the brain, another third of which are also used in the brain. How many permutations is that? How many environmental elements interact? There are 10^10,000 possible brain states. Can you imagine how each of those would interact with each possible environment? THe good news is that since our brains are self-organizing networks, they tend to reduce to a set of human universals, yet even so, variations in the social environment affect their expression, with the result of an incredibly wide variety of expressions of those universals. How many languages, how many religions, how many ways of cooking are there? How many will there be?

I'm very impressed if you can in fact imagine all of that. I know all of this and find it hard to imagine.

Tracy W writes:

Troy, I am sorry, I misread you. I didn't realise that you were merely claiming that we couldn't imagine 10^10,000 gene-environment interactions at once, I thought that you were claiming that we couldn't imagine many of the possible ways in which genes and the environment could interact to affect each other. If you had mentioned the 10^10,000 ways directly, I would of course agreed with you. It was without that, and your apparent ruling out of a genetic propensity for book reading, that resulted in me wrongly thinking that you were claiming that we couldn't imagine some ways of more complex gene-environment interactions. My apologies.

That said, I still don't understand what point you were trying to make in your initial response to Caplan's post. Yes, we cannot possibly imagine *all* of the possible interactions between genes and the environment, but that doesn't mean that we can't test for statistical relationships, and look at what we know of the real world and draw tentative conclusions from that. Indeed, the cases you cite, such as that of relationships between our gut bacteria and our behaviour, the epigenetic relationships, are surely dependent on this sort of research being capable of being carried out.

hacs writes:

Tracy - those estimates of inheritance could be biased upward also, if parents adapt their nurturing to maximize the performance of the biological offspring.

Troy Camplin writes:

As I said, the result of all of these interactions is, fortunately, a set of strange attractors exhibited as human universals. To a certain extent, these are your "statistical relationships." These are still incredibly complex, to such a degree that we actually had people proclaiming there was no universal human nature, and that it was literally impossible for the various cultures to understand each other. That is a step too far, of course -- but it's not an entirely unreasonable response to human levels of complexity.

My point was that it is at best extremely difficult to disentangle the influences of genes and environment, since each influences the other. And they do so through bipolar feedback.

hacs is also correct. For example, my daughter is taking piano lessons. She's 4. The teacher has noted that she is very good at following rhythm -- much more advanced than she should be (to such an extent that the teacher said she was going to have to find a rhythm book to help her). Why might this be? My wife's father is a professional musician, so it might be genetic. But my wife and I have had her in Music Together classes since shortly after she was born. Further, I'm a poet, and I make up poems and songs all the time as I play with my children. So perhaps it is environment. Or perhaps the environment is enhancing the gene-based tendencies she was born with. Without the stimulating environment, such tendencies wouldn't be(come) quite as developed. Thus, the genetic component is enhanced by the environment -- but one could easily point to the fact that her father is a poet and one of her grandfathers is a musician, and conclude it must all be in the genes. I am doubtful of that.

hacs writes:

Tracy (cont.) - smarter parents could do a better job (maximizing the performance of their biological offspring). If optimal nurturing is a child-specific technology, then smarter parents could get closer to the optimum nurturing technology, separating points (i.e. this technology is not so good for other children). It is just a hypothesis.

Tracy W writes:

Troy: I agree with you that these influences are hard to disentangle, but as I said in my first response to you, do you doubt that for a normal child, which language they learn to speak, given that they have the genes that allow them to learn and speak a language, is more driven by environment, which blood group they have is more driven by genes than the environment? You yourself confidently said earlier that there are things like blood type and eye colour, that are purely genetic.

As for the example of your daughter: You don't mention the possibility that the genes she has are influencing how she shapes her environment, eg that she seeks out musical stimulation. She could either get it from her parents, or from other sources. If she was born in a culture without music based around rhythm, her genes couldn't do this (genes and environment.)

To tell a different story, one of my mother's cousins was adopted out at birth, unbeknownst to anyone in our family (the blood relationship to us was through her father, who was in jail at the time she was born. She was adopted by a small-town NZ Pakeha family, where the best that a girl could hope for was to leave school at age 15 and get a job in a bank. Despite this, she had this massive lust for more education, and managed to persuade her adoptive parents to let her finish high school and then go to teachers' college (teachers' college as there was a job at the end of it). Then she turned 21 and tracked down her birth family, and discovered the whole bunch of us with degrees and heaps of teachers in the family. Would she have become a teacher 200 years ago under similar circumstances? Almost certainly not, but the environment of a small NZ country town in the 1970s, including government support for education, allowed her genes to express in this way - genes and environment, but apparently not adoptive-parental environment.

I note that the studies of parental influence do typically find that parental environment does affect kids' personalities and other characteristics, it's as the kids get older and turn into adults that the measures of parental influence on behaviour fades. (see https://notes.utk.edu/bio/greenberg.nsf/0/6cfebd6304576bf085256c7900642f84?OpenDocument)

Hacs - presumably if your hypothesis was right, we should see identical twins brought up together by their biological parents more similar in personality than identical twins brought up separately. But any such effects appear to be very minor, identical twins bought up together are very much as similar as identical twins brought up separately. See http://rc.usf.edu/~jdorio/Personality/Personality%20Similarity%20in%20Twins%20Reared%20Apart%20and%20Together.pdf

Troy Camplin writes:

Blood group is not plastic, like the brain. When it comes to behavior, only the brain is of any real concern. The neurons in our brains can and do adjust to their environment. Further, our brains grow new neurons all the time. If we use them through continued learning, they become integrated into the brain (otherwise, not). Surely more neural connections and more brain cells, resulting in a denser neural net, affects our mental abilities. More than that, there is evidence, from the work of Clare Graves and others that, depending on our environments, we develop through ever-greater complexities of thinking, with changing values as a result.

One can give just as many counterexamples. Like my wife wanting to go to college even though there is little evidence of anyone in her family wanting to go, and her grandparents actively discouraging her. Her intelligence got her into a magnet school, and everyone there was going to college. That's environment.

As for my daughter, we cannot know because we cannot replicate the experiment, but there's no way she could have sought out music unless there's music around. She couldn't get to Music Together classes as an infant, and there is little doubt she wanted piano lessons because of Music Together. The genetic component certainly may allow her to become a better musician than if she did not have it. Humans do seem programmed to enjoy music. But there are some who are better at learning and making it than others. But without the proper environment, those genetic tendencies will often go undeveloped. I love music, for example, but I was discouraged from learning music as a child, so I never learned to play a musical instrument. That musicality only found expression when I started writing formalist poetry. I read novels growing up, as did my mother. But not poetry. That was a later development, and there were no relatives on either side who wrote poetry or did music. The only artistic element on either side was an ability to draw. My brother is a visual artist, and I have the ability to draw anything you could want. But I don't draw much anymore, as I did when I was a child (though my daughter does draw quite a bit, and quite well for a 4 yr old). Is there a general tendency for art? Why would mine get directed into language? There is a strong shamanistic element to literary production, but there is but one example, a great uncle, of religious leadership in my family. Considering the religiosity of my family, why wouldn't I have become a pastor (which would have thrilled my family) instead of first going into biology and then becoming a poet and playwright?

I think that a lot of these studies involve people who have not wandered too far away from their families' behaviors in the first place -- for whatever reason. Outliers cause problems with the data, so get excluded, either consciously or unconsciously. My wife and I are, in many ways, outliers to our families. I suspect, though, that our children will be more like us, because they are likely to go to college.

On another personal note, I wrote a weekly article for a newspaper in Kentucky for over a year. It was read by friends and relatives of my parents. Someone noted that in my sense of humor, they saw elements of my dad, elements of my mom, and something that was distinctly me. Where did "me" come from?

Tracy W writes:

Troy, I'm rather puzzled. Nearly everything you say here I agree with, yet I get the sense that you think I differ from you.

I agree with you utterly that the neurons in our brain adjust to our environment. That's why I said, earlier: "My memory of what I had for breakfast affects my brain's structure."

And as I've been telling you, it's genes and environment. I 'corrected' you when you implied "genes but environment". I've even been bolding the and. Is there anything I else could possibly say that could convince you that I already agree with you that it's genes and environment? (It's not just environment, those neurons come from somewhere in the first place, and how they change in response to environmental inputs also has to come from somewhere, to see wh,y buy a computer, hook up a webcam and a microphone to the CPU and the hard drive, and you can blast it with environmental input for decades, but unless there's something in there that can learn, e.g. a neural net, it's not going to learn anything).

On the issue of your daughter, in the last comment I made, I pointed out that parental influence on children does show up, it's just that it fades as the kids grow up and become more able to control their own environments.

I agree with you that a problem with a lot of these studies is that they involve people who have not wandered too far away from their families' behaviours in the first place. Judith Harris points that out as a problem, that to really understand what is going on, you have to look at

For the further questions you raise, about why the individual things happen, eg why you didn't become a pastor, may I suggest reading Judith Harris's "No Two Alike"? She suggests a set of hypotheses there for why humans develop individual, though she proposes it as something to be tested, by someone other than the proposer. I think you'd still find the book very interesting. And she, and I, do agree with you that a problem with a lot of studies is that they don't look at outliers, but it's the outliers that can disentangle what is really going on.

hacs writes:

Tracy - "presumably if your hypothesis was right, we should see identical twins brought up together by their biological parents more similar in personality than identical twins brought up separately. But any such effects appear to be very minor, identical twins bought up together are very much as similar as identical twins brought up separately."

My hypothesis does not imply homogeneity. In fact, my hypothesis just implies that less smart parents use more inefficient technologies (perhaps, more robust/unspecific ones) and smarter parents use more efficient technologies (more child-specific ones). In both cases, there is no personality homogenization.

Smarter parents and less smart parents optimize their technologies wrt their biological offspring, assuming that what is good for them is good for others. The inefficiency of the less smart parents is less harmful for adopted children.

Tracy W writes:

Hacs: I didn't say that your hypothesis implied homogeneity, I said it implied that identical twins brought up by their biological parents would be more similar in personality. I am not certain if when you say "my hypothesis does not imply homogeneity" you're disagreeing with my expectation of what your hypothesis implies.

And, I don't understand how your hypothesis can be right and yet, simultaneously, we not see any increase in the similarity between identical twins brought up by their biological parents compared with identical twins brought up separately. Perhaps it would help if you explain how your hypothesis is testable.

Comments for this entry have been closed
Return to top