5 ms·

So, they do this by comparing correlations in MZ (genetically identical) and DZ (typical sibling genetics) twin outcomes. Doesn't that assume that the fact of

So, they do this by comparing correlations in MZ (genetically identical) and DZ (typical sibling genetics) twin outcomes. Doesn't that assume that the fact of being DZ vs MZ does not itself create environmental (non-genetic) effects? The identical twins I know face particular social pressures to respond to their twin identity, which could drive towards outcomes that are more correlated -- or not! But it seems weird to base a branch of literature on the assumption that it's not. Is this addressed?

tomp 7y ago
You can also compare identical twins that grow up in the same family, vs. identical twins that were separated at birth... although I imagine you wouldn't have a lot of samples.
Symmetry 7y ago
That's a reasonable sort of question but would you say that MZ twins are treated differently enough to explain 40% of variation in earnings? If they were all locked in cellars out of shame I could see that being the case but that doesn't square with the identical twins I've known. But it actually is addressed by other studies such as adoption studies that show generally the same results. And, of course, there are potential problems with adoption studies too. But wouldn't it be suspicious if every[1] method of studying the problem we tried gives fairly close numbers but only because they all happen to be biased by the same amount in the same direction? [1]GWASes give lower numbers but we know that there are many sorts of genetic interaction they don't cover so this is expected.
barry-cotter 7y ago
This is an empirical question and as far as we can tell if there’s pressure to be more similar or more different it washes out. Identical twins reared together are no more similar than ones raised apart on a huge host of traits.
oli5679 7y ago
Yes that is valid point about the limitation of this approach. There are 3 methods to address it, none of them are perfect: (1) Statistically compare the population of twins vs the population of non-twins. The paper I linked to has more specifics on this, but twins earn similar amounts of money, live similar lengths of time, do similarly well in exams vs. non-twins, so it any twin-specific genetic effect has to simultaneously be strong enough to cause this discrepancy, but not strong enough to impact population means. (2) Compare adoptees to genetic offspring of parents. This has the same issue that you pointed out (maybe there is something specific about adoptees) but the two together find roughly similar effect magnitudes, which reenforce each-other. I'm actually a bit more concerned about adoption being less 'clean' an identification strategy than twin studies, because it's a complicated social process. (3) Look at genetic variation at population level and find genetic markers that explain variation, using Mendelian randomization as identification strategy - this is technically very challenging, and has only recently become feasible. So far research have only been able to explain around 7% of IQ for example with specific genetic markers, but it will be interesting to see if this creeps up as the field matures. The big puzzle here is the large variation that we between cultures - moving time-periods, or countries can significantly change your life-expectancy, earnings, test-scores or even IQ. I don't know how to reconcile this with the strange finding of this research strand is that family level environmental effects (shared environment) seem to be much less important than the conventional wisdom would suggest. Links https://www.sciencedirect.com/science/article/pii/B0080427073001759 https://www.sciencedirect.com/science/article/pii/B008042707... (gated) https://en.wikipedia.org/wiki/Mendelian_randomization https://en.wikipedia.org/wiki/Mendelian_randomization https://www.newscientist.com/article/2163484-found-more-than-500-genes-that-are-linked-to-intelligence/ https://www.newscientist.com/article/2163484-found-more-than... https://en.wikipedia.org/wiki/Flynn_effect https://en.wikipedia.org/wiki/Flynn_effect

tomp 7y ago

You can also compare identical twins that grow up in the same family, vs. identical twins that were separated at birth... although I imagine you wouldn't have a lot of samples.

Symmetry 7y ago

That's a reasonable sort of question but would you say that MZ twins are treated differently enough to explain 40% of variation in earnings? If they were all locked in cellars out of shame I could see that being the case but that doesn't square with the identical twins I've known. But it actually is addressed by other studies such as adoption studies that show generally the same results. And, of course, there are potential problems with adoption studies too. But wouldn't it be suspicious if every[1] method of studying the problem we tried gives fairly close numbers but only because they all happen to be biased by the same amount in the same direction? [1]GWASes give lower numbers but we know that there are many sorts of genetic interaction they don't cover so this is expected.

barry-cotter 7y ago

This is an empirical question and as far as we can tell if there’s pressure to be more similar or more different it washes out. Identical twins reared together are no more similar than ones raised apart on a huge host of traits.

oli5679 7y ago

Yes that is valid point about the limitation of this approach. There are 3 methods to address it, none of them are perfect: (1) Statistically compare the population of twins vs the population of non-twins. The paper I linked to has more specifics on this, but twins earn similar amounts of money, live similar lengths of time, do similarly well in exams vs. non-twins, so it any twin-specific genetic effect has to simultaneously be strong enough to cause this discrepancy, but not strong enough to impact population means. (2) Compare adoptees to genetic offspring of parents. This has the same issue that you pointed out (maybe there is something specific about adoptees) but the two together find roughly similar effect magnitudes, which reenforce each-other. I'm actually a bit more concerned about adoption being less 'clean' an identification strategy than twin studies, because it's a complicated social process. (3) Look at genetic variation at population level and find genetic markers that explain variation, using Mendelian randomization as identification strategy - this is technically very challenging, and has only recently become feasible. So far research have only been able to explain around 7% of IQ for example with specific genetic markers, but it will be interesting to see if this creeps up as the field matures. The big puzzle here is the large variation that we between cultures - moving time-periods, or countries can significantly change your life-expectancy, earnings, test-scores or even IQ. I don't know how to reconcile this with the strange finding of this research strand is that family level environmental effects (shared environment) seem to be much less important than the conventional wisdom would suggest. Links https://www.sciencedirect.com/science/article/pii/B0080427073001759 https://www.sciencedirect.com/science/article/pii/B008042707... (gated) https://en.wikipedia.org/wiki/Mendelian_randomization https://en.wikipedia.org/wiki/Mendelian_randomization https://www.newscientist.com/article/2163484-found-more-than-500-genes-that-are-linked-to-intelligence/ https://www.newscientist.com/article/2163484-found-more-than... https://en.wikipedia.org/wiki/Flynn_effect https://en.wikipedia.org/wiki/Flynn_effect