I had been meaning to pen a longer response to Bryan Caplan (he is the one with "a theory of everything" in this area, not I, his theory just happens to have few variables), but I'll focus on two of his claims, as they are indicative of the larger disagreement:
Parents strongly affect what you say your religion is, but have little long-run effect on your intrinsic religiosity or observance. I don't discuss language, but it's pretty clear how a twin or adoption study would play out: You can make your kid semi-fluent in another language with a lot of effort.
Both claims are false, at least at many commonly available margins.
Take Jews. If a group of children are born to Orthodox, Conservative, Reform, or liberal parents, their later religious observance will be predicted by both peers and parental upbringing. (Perhaps genes are a factor too.) An Orthodox Jewish boy, with Orthodox parents, growing up in an otherwise non-Orthodox Jewish community of peers, is more likely to stay Orthodox than a Reform Jew from the same community is likely to become Orthodox. And that of course correlates with levels of observance. I have no formal study to cite, but can I just stamp my feet and scream this is true? Because it is. You can imagine numerous variants on this tale, even if it isn't true for all religious denominations.
Bryan has a tendency to concede environmental factors by noting something like "Of course parents can lock a kid in the closet and affect him that way." He is less likely to admit that a lot of less extreme influences can matter too and that those influences are missed by twin adoption studies, for whatever reason.
Or take language. Yana speaks Russian. She learned Russian from Natasha (her mother, and it wasn't hard for her to speak Russian at home), and note that Yana left Moscow before she was two years old. This is again a common pattern. The parents matter, even though in most American families you won't see enough cross-sectional variation (most people speak English at home) to always pick this up. Travel around India for more examples of this phenomenon.
Presumably the twin studies have in their data sets Jews and possibly some Russian immigrants as well. And yet the twin studies, with their ultimately macro orientation, miss micro mechanisms such as these. Parents can matter more than the studies suggest.
By treating those studies as an epistemic trump card, Bryan is led to make claims which are indefensible on the face of it. I stick by my earlier points. The evidence Bryan is citing for twin adoption studies is simply…the studies themselves. Where is consilience when you need it?