An interesting piece from the Boston Globe on “genoeconomics”:

Though the name wasn’t coined until 2007, genoeconomics flickered briefly into existence once before. In 1976, the late University of Pennsylvania economist Paul Taubman published the results of a study in which he followed the financial lives of identical twins, and found there were curious similarities in how much money they made as adults. Taubman concluded that between 18 percent and 41 percent of variation in income across individuals was heritable.

It was a startling conclusion, and one that Taubman’s fellow economists didn’t quite know what to do with. One joked that Taubman’s findings meant the government might as well shut down welfare, since clearly some people would remain poor no matter what.

….After Taubman, the idea that genes had an important role to play in decision-making was largely abandoned in the world of economics. But with the completion of the Human Genome Project in 2000, the first full sequence of a human being’s genetic code, people started wondering if perhaps it would be possible to push past broad heritability estimates, of the sort that Taubman generated, and figure out what part of a person’s genome influenced what aspect of his behavior.

…Over time, social scientists started coming to terms with the fact that even the most heritable of traits, such as height, were influenced not by one or two powerful genes, but by a combination of hundreds or even thousands—and that environmental factors, like a person’s upbringing, play a complex role in determining how those genes are expressed. “Every single direction has proved to be less promising than people originally expected,” said Laibson.

… hope lies in a new approach to data-gathering that is only just getting underway, wherein researchers look for patterns among thousands, and even millions of people—numbers that are only just becoming possible thanks to massive collaborations linking gene studies being conducted all over the world.

The researchers in question, Daniel Benjamin, David Laibson, David Cesarini and others, seem worried about the possibility of tracing attributes and behavior to genetics. Most of the big news is out already, however, and more easily observed in phenotype than genotype.

For more on the new approach see The genetic architecture of economic and political preferences.


"One joked that Taubman’s findings meant the government might as well shut down welfare, since clearly some people would remain poor no matter what.".

Sounds like a reason to expand welfare to me.

This is the difference between (my labels) the conservative versus liberal views of it. Had a discussion once with a lefty who said we should give money to the homeless to buy alcohol. We could never understand eachother.

Was that wood grain alcohol?

Depends what you think the purpose of welfare is. Is it to provide a guaranteed minimum standard of living, even if you don't/can't work at all? Or is it to try to get people who are down on their luck or otherwise having some problems, pick themselves back up again and re-enter productive society?

I tend to think that it's both. If there were people who are genetically doomed to forever be poor (and note that I'm not agreeing that there are), they should still be prevented from starving. Where you should set this minimum standard of living, though, is definitely up for debate. Perhaps reflecting my home (Canada), I tend to believe that the ideal for this is left of the right-wingers in the U.S. who write about this stuff (think: Heather Mac Donald), and to the right of northern Europe.

File under: Sailer bait

The paper is excessively turgid. No simple, clear summary up front. Too much jargon.

Not really clear what it means, either. So if I make $100,000, then my child is likely to make at least $40,000. Is that what they're saying? Or is it something else? And if they are, is that an insight?

To some people, the notion that anything more than appearance is heritable is unthinkable and possibly evil.

Confirming things we already know is sometimes just as important as making breakthrough insights.

The evil is in making prejudgments when RCTs can't even show significance. And I mean "evil" in the objectivist sense.

Queue "Thus Spake Zarathustra". Economists have discovered intelligence.

Baaaa baaaa baaaaaa..... pim pum! (bim bum bim bum bim bum)...

An unobserved attribute constant to an individual, like "innate ability" or intelligence, can be accounted for by the inclusion of a fixed effect term. People have been doing that for a long time. Attributing that unobserved ability to a gene is something new that can only happen once data on genes becomes available, but the idea that "intelligence" (or an innate ability or characteristic) plays a role in various outcomes is hardly new and hardly just becoming discovered.

But mainly (see my other comment), the economic studies I've seen that look into genes are, simply put, really terrible studies. They've basically amounted to "Let's throw in a whole bunch of data into a regression, see what we get, and entirely ignore the dozens of problems entirely invalidate our results." They often strike me as examples of people who know little about genes and little about econometrics combining the two in utterly meaningless ways. They're often great examples of "garbage in, garbage out." It's never even come close to a moment worthy of being soundtracked by Richard Strauss.

Didn't take the time to read it, is there an orphan control group? And 18 to 41%, pretty hard to deny that kind of mathematical proof.

I once saw a "genoeconomics" paper presented at a conference that was probably the worst paper I'd ever seen. It may have even been this one (I didn't bother to read the linked paper just in case it was that same one. I don't want to waste more time than I already have.)

One problem is that "genes" aren't the constant and steady thing people make them out to be. Your genes will change over time. They've studied twins over long periods of time and their genetic differences do increase as the years go on (via mutations). Another problem is that genes can work in combinations. You'd have to add in interaction terms. Looking at the presence of single ones doesn't get you much. More problematic is the idea "gene expression." Certain gene clusters can become activated, or do very little. This can change over time and genes can be triggered into action by environmental factors. Just seeing whether or not someone has a gene doesn't matter nearly as much as whether or not that gene has been activated.

Also, while everyone knows that correlation does not imply causation, lesser known is the equally true statement that causation does not imply correlation. There may indeed be a true causal relationship between two variables and your regression won't even pick up a simple correlation between the two. I think genes may be a good example of how that could happen:

Say Y is a variable for entrepreneurship or whatever. Say X is the gene that causes this. Now say there's a factor A that represents whether this gene has been activated. If it has been, A =1, if it hasn't been, then A = -1. Say that whether or not it's been activated is entirely random, a 50/50 change of either. Then you have y = Ax. Even if we already know that X causes Y, a regression would find zero correlation between the two. And, of course, if we didn't know x caused y, we'd likely never suspect it does since they're not even correlated.

Looking at single genes at a single point in time thus ignores three huge issues...1. that they can change over time. 2 that combinations of genes matter, and 3. whether or not that gene is even active.

Throw in a badly defined dependent variable (entrepreneurship), measurement error, selection issues, etc. and things start getting even worse. Plus, remember that for every 20 completely meaningless covariates you throw in, 1 should be statistically significant purely by chance if you're using a 95% C.I. If I throw in thousands of genes, dozens will come out as significant purely due to randomness. If you include interaction terms (like one should given how genes interact), the number of covariates increases exponentially, leading to an even higher number of purely random things coming up as significant.

In short, every paper I've seen in this realm has always struck me as utter bupkis.

Great summary! May I add:

"Under the key assumption of no environmental confounding" (p.2)

Sometimes I think researchers get their priors about identical twins' life experiences from TV shows.

So, because the data is hard to analyze, genes don't affect behavior?

I believe for height and IQ they have actually attempted to test whether there are significant "interaction" effects, or whether you can just treat the huge number of genes as having small additive linear effects, and the latter approach works quite well. Of course, Robyn Dawes pointed out in "Rational Choice in an Uncertain World" that simplistic functions of variables with arbitrary (maybe uniform) weights are quite good at prediction as long as the sign for each is correct.

These results help others to understand the genetic flexibility of human foetuses and and their ability to withstand an adverse prenatal environment.

As an identical twin boy among four boys, we had a number of disadvantages.

We received significantly less nutrients and oxygen and were severely cramped in our small mother's womb and as infants, we only received half our mother's attention our brothers received in our early years.

Our standardised aptitude test results in sixth grade for my twin and I were nearly the same.

Yet, in the genetic lottery, each of us outperformed our brothers in education accomplishment, income and career goals.

Twin-to-twin transfusion syndrome is a major and common problem in identical twin pregnancies. Wikipedia says 5.5–17.5% of all monochorionic pregnancies, which themselves are about 75% of identical twin pregnancies. It leads to asymmetric outcomes up to and including the death of one twin, which often kills the other. Until I start seeing twin studies that account for it, I'm going to believe that genetic heritability is underestimated by twin studies, because a large fraction of identical twins go through an environmental divergence in the womb.

This is probably out of date given new research in fields like epigenetics.

As Tyler implies at the end, for a few generations, we've had twin, adoption, and even separated twin studies of real world outcomes. They've tended to suggest, stylizing hugely, that nature and nurture are about fifty-fifty in importance.

Along come various concepts like epigenetics, which announce that "genes change in relation to the environment," but what's seldom discussed is that the reasoning goes like this: "Genes are actually more important that twin studies suggested, maybe 80-20, except that the environment causes genes to vary, so we end up with 50-50, just like the twin studies said all along."

To think that genes play a role in how much money one makes is, to me, a bit ridiculous. With this logic, we should see Bill Gate's or Warren Buffet's kids to follow their parents footsteps and be just like them, earning ridiculous amounts of money or inventing life changing technology. It is possible, although, that these kids will inherit certain traits which were a factor for their parents success. They will see how hard they worked or pick up on a few characteristics, but to say that they are born with those genes that allows them to be successful seems a bit far fetched.

What would genoeconomics really do for us anyways? We cant change or alter our genes. Its something that we are born with. Knowing that I do not possess a gene to be successful because my parents or grandparents lacked it may encourage me to drop out of school and not even try; whats the point in trying? This would be very bad for the economy and for every aspect of our society. Finding out this information is not only useless but is probably taking up a lot of resources, resources that should be going to better studies that will help improve our lives.

I'm going to take a wild guess and say that you've never heard of "regression to the mean". Or you're being sarcastic.

güzel metin teşekkürler... thank you so much...

Comments for this post are closed