I am a very heretical rationalist, because I believe in an effect of parenting on children.

Among relatively homogeneous samples, about half of variation is genetic, while about half of variation is the product of nonshared environment. However, much of nonshared environment may be the product of measurement error: more careful studies suggest, for example, that 85% of variation in personality is explained by genetics and only 15% by non-shared environment, suggesting that a full 70% of non-shared environment is actually measurement error.

Some people conclude from that study that, in fact, nothing has an effect on anything and your entire personality is based on genetics. I conclude from that study that this is not a very reliable test.

It is very very difficult to pick up small effects with unreliable tests. You need a huge sample size to do so. Twin studies don’t generally have huge sample sizes, because there just aren’t that many twins in the world. And even social science studies that don’t have any twins are regularly underpowered, because journals don’t care about statistical power in the same way that they care about p-values.

It definitely looks like we’re in the sort of world where parenting could have effects. Studies of children who grew up in Romanian orphanages have shown that spending more than six months in the orphanages– in which they had inadequate food, poor hygiene, little social or intellectual stimulation, and no love or affection– results in cognitive impairment, autismlike symptoms, and emotional problems. There is a shared environment effect in heterogeneous samples– such as those which include both rich and poor parents– which is presumably not solely because of the salutary effects inherited wealth can have on the personality. If we know that extreme differences in parenting have effects we can detect, and we know that there are lots of small effects we wouldn’t be able to detect even in principle, then it seems to me that we should assume those small effects exist.

Believing in zero role for shared environment also fails to pass the sniff test. Lots of people, I think, will agree that whether or not you talk to your baby might not have any long-term consequences. But the “zero role for shared environment” position requires yourself to commit to the position that all of the following have either zero correlation between siblings or no effect on children’s psychology whatsoever:

  • Lead poisoning.
  • All other forms of air and water pollution.
  • Prenatal nutrition.
  • Drinking during pregnancy. (Not just light drinking, I mean slamming down ten shots a night every night for your entire pregnancy.)
  • Smoking during pregnancy.
  • Illicit drug use during pregnancy.
  • Secondhand smoke exposure during childhood.
  • Childhood nutrition.
  • Childhood sleep deprivation.
  • School quality.
  • Whether there are drugs being sold in front of your house.
  • Childhood mental health treatment.
  • Childhood physical health treatment.

Sure, maybe you can quibble about one or another of these. Maybe you think that sending your kid to Andover will result in exactly the same consequences as sending your kid to a school where they learn that evolution is a lie because the Loch Ness Monster is real. Maybe you think there’s no real harm in feeding your child lead paint for dinner and that fetal alcohol syndrome is a lie made up by people who just want you to stop having fun, man. Maybe you think no one ever smokes through more than one pregnancy. But it seems really implausible to me that every single one of those either doesn’t have an effect or isn’t correlated between siblings.

People who don’t believe in shared environment sometimes propose mechanisms for what non-shared environment is. “Maybe siblings have different friends!” they say. Okay, and your claim is that there is literally zero correlation between who siblings are friends with? No parents ever bring all their children to another child’s house for a shared playdate? Poorly funded inner-city schools contain exactly the same set of people to be friends with as wealthy schools in the suburbs? “Parents treat different children differently!” they say. Okay, and your claim is that there is literally zero correlation between how parents treat one kid and another kid? No similarities based on a parent being a really angry person, or a really demanding person, or a single mom working three jobs who barely has time to take care of herself let alone a child?

The only reasonable course here is to grant “okay, maybe lead poisoning might increase criminality, and children who grow up in a house with lead paint are more likely to have lead poisoning than other children.” That opens the door for other small effects of parenting as well.

The interesting consequence of this argument, however, is that we don’t have much scientific evidence what the good forms of parenting are. There are few randomized controlled trials of parenting strategies. You can do an encouragement design, in which some parents are given free Baby Einstein videos or sent to educational classes about how you shouldn’t spank your children. Unfortunately, since many parents won’t change their behavior even with a class or a free video, you have to have a huge sample size, which is very expensive. So most people just do observational studies, which are cheaper.

Essentially all observational studies of parenting strategies suffer from healthy user bias. The sort of people who follow parenting advice are different from the sort of people who don’t follow parenting advice. They’re more conscientious. They put more effort into parenting. They care about doing right by their children. If nothing else, they’re more likely to follow all the other parenting advice, and that includes obviously correct things like “don’t let your child eat lead paint.” It’s very difficult to control for all of these factors. So if a study finds that following conventional parenting advice makes your child healthier, more talented, and kinder, it doesn’t actually tell you a heck of a lot. (On the other hand, if a study tells you that not following conventional parenting advice makes your child healthier, more talented, and kinder, you should probably listen to that study.)

One heuristic I’ve been using is thinking about the effects I can expect to have on my spouse. There are a lot of similarities: we live together, we spend a lot of time together, and we’ll do so for years. I think our intuitions about our effects on our spouses are less biased than our intuitions about our effects on our children, because we don’t have a ridiculous guilt-inducing culture that believes that if we just do everything right any spouse can make a six-figure income, live to age 100, and be ecstatically happy at all times. These are my intuitions about my spouse:

  • My husband’s basic temperament is set. He is always going to be basically himself, no matter what I do.
  • I have a huge effect on how much my husband likes me and how good our relationship is.
  • I could traumatize him if I chose to behave in an unconscionable way.
  • I definitely have it in my power to make my husband miserable, but it’s not in my power to make my husband happy– brain chemistry and other life influences have their role.
  • I am only one influence on my husband’s behavior. The effect of my actions is real, but I cannot automatically cause anything to happen.
  • I can share information with my husband that changes his behavior.
  • I can model behaviors I’d like my husband to pick up. If my husband sees me exercising, he’s more likely to exercise himself.
  • I can reward my husband for doing things I want him to do, but I shouldn’t expect that he’d keep doing the thing if I stopped rewarding him.
  • I can introduce my husband to experiences he otherwise wouldn’t have but that he enjoys.
  • I can support my husband and allow him to achieve things he wouldn’t be able to otherwise.
  • None of those five things work 100% of the time. How well they work on any given subject depends on my husband’s preferences, temperament, and choices.
  • Perhaps most of the effect I have on my husband’s personality does not come from my deliberate efforts. It comes from how he responds to my everyday behaviors.
  • If we stayed married for twenty years and then he divorced me, I’d expect my effects on his personality to decrease over time, but there’d always be an effect that came from me.

Of course, children and spouses are different in many ways. For one thing, not loving an infant causes far worse long-term consequences than not loving a spouse. But I think as a first-pass heuristic for what effects you can and can’t have, and how you can have them, this is pretty reasonable.