Selfish Reasons to Have More Kids by Bryan Caplan

Bryan Caplan is the ALS Ice Bucket Challenge of authors. Readers can feel a physical shock as beliefs they hold closely are picked apart piece by piece. I half expect Bryan’s next book to show that drinking water, eating food, and breathing air are all bad for my health.Selfish Reasons to Have More Kids is no exception from this Caplanian trend. Like any Caplan book, reading it requires an open mind, and a healthy amount of skepticism.

Does having kids make you happier?

The status-quo effect of having kids on happiness is unclear. Reflectively, parents are much happier with their decision to have kids than non-parents are with their decision to abstain from kids.

When asked “If you had to do it all over again, would you or would you not have children?” 91 percent of parents say they would have children all over again. Only 7 percent said they wouldn’t. You might object that people merely rationalize whatever decision they’ve made, but the best available survey finds that nonbuyer’s remorse is common. In 2003, Gallup asked childless adults over the age of forty, “If you had to do it over again, how many children would you have, or would you not have any at all?” Over two-thirds of the people without kids confessed regret.

Pretty strong evidence in favor of being a parent! On the other hand, parents are a little less happy day-to-day after accounting for confounding factors (income, religiosity, and relationship status).

In the general Social Survey, a massive, decades-long study of Americans, every child makes you about 1 percentage point less likely to call yourself “very happy.” The difference is real, but you need a statistical microscope to detect it. Married people, in contrast, are 18 percentage points more likely to be very happy.

This roughly matches data for the 2017 Slate Star Codex survey summarized here. These studies look at the average parent, but you can be a happier parent than most, with the power of apathy. Most studies suggest that parenting doesn’t have a strong effect on children’s life outcomes. This is great for parents, because they can be lazy without throwing away their children’s future.

Parenting doesn’t matter is the ice water over the head claim. Nobody believes it the first time they hear it, and you should have a high bar for evidence to accept it.

Twin and Adoption Studies

Social scientists think about 3 factors that influence a child. Genetics, shared environment, and nonshared environment. Shared environment is any factor that is shared by 2 children growing up in the same home, and unshared environment is everything else about the child’s environment. Parental income is shared environment, but one kid joining the debate team is “unshared environment”. A mom who forces her kids to study 2 hours every night is shared environment, but one child getting pneumonia is unshared environment. Caplan thinks that shared environmental factors don’t really influence life outcomes. The evidence for this mainly comes from twin and adoption studies.

Adoption studies make two comparisons. First, they look at unrelated children related children raised under the same roof. The more similar they are, the more presumably parenting matters. They also look at related children raised in different households. Presumably, their similarities are points in favor of genetics and against shared environment. Twin studies compare identical twins to fraternal twins, typically raised in the same household. The more similar identical twins are, the stronger the case for genetics.

Health

One way that parents like to influence their children is in health. Encouraging their children to eat healthy, stay away from drugs, and get some exercise are all things parents do to make their kids live longer, healthier lives. Does it work? Caplan says no. A study looking at life expectancy for over 3,000 Danish twins found “no evidence for an impact of shared environment.” Another study of 9,000 Swedish twins found no parenting effects, though both studies found strong genetic effects. Other studies are mixed. A smaller study of Finnish twins found a moderate nurture effect on self-reported health, but no difference in doctor evaluated health. A Swedish study of 700 twin pairs did find a strong nurture effect- if you were in the 80th percentile for objective health, on average your adoptive sibling would be in the 65th percentile.

Caplan concedes there is a clear parental effect when it comes to reducing your child’s use of alcohol, tobacco, and drugs. One study of 1,000 Korean children adopted into the United States found that being raised by a family of non-drinkers lowered the adoptee’s chance of drinking alcohol by 20 percentage points.

Income and Education

Whether through reading, education, or conversation, parents try to make their kids smarter. If they’re doing a good job, it isn’t showing up on standard IQ tests. A study looking at 245 adopted children in Colorado found that ones adopted into High IQ homes did no better on IQ tests than ones sorted into average IQ homes. A similar study in Texas with 300 adopted children found the same result. Again, other evidence is more mixed. A Minnesota twin study found that if you’re in the 80th percentile for IQ, your adopted sibling is likely in the 56th percentile (a different test showed no result). This shows at least a moderate effect for upbringing.

Another metric where evidence is weak but perceptible is school. If you are in the 80th percentile for GPA, on average your identical twin will be in the 71st percentile, a strong relationship. In contrast, non-related children in the same household had GPAs no more similar than strangers. On the other hand, a Korean study found that children adopted to high education families do get statistically significantly more education, though the effect size is small.

Again looking at the Korean adoption study, the income effect is non-existent. Children adopted by the richest families went on to earn the same as children adopted by the poorest families! Another American study of 400 twins found a modest effect- if you were in the 80th percentile for income, you could expect your adopted sibling to be in the 55th percentile.

Caplan details the same pattern for lots of other things parents might care about. Religion, politics, marriage, sex, and personality all fall into the category of weak to no evidence for a nurture effect.

Bryan sums up his view from all of these studies.

Once upon a time, nature vs. nurture was a matter of opinion, but now there are hard answers: Nature wins, especially in the long run. If your child had grown up in a different family- or if you had been a very different parent- she probably would have turned out about the same.

Criticisms

I will never hate anything as much as economist Arthur Golberger hates twin studies. Ironically, Caplan wrote a blog post in 2008 calling Goldberger the wisest econometrician in the world. Goldberger argues that identical twins are often more similar in environment than paternal twins. For instance, parents might sign them up for the same activities, and treat them similarly in ways they may not for fraternal twins.

Goldberger also goes after two twin studies cited above, the Minnesota study and the Sweden study. In the Minnesota study, the identical twins who were “raised apart” often had substantial contact with each other, casting doubt on the assertion they had no shared environment. Also, Goldberger points out, twins raised apart were often given highly similar, although separate, environments by their parents or social workers. This selection effect can’t be ignored.

In the Swedish study, the median age of twins interviewed was 59. At this point, there is good reason to doubt their self reporting. Many of the twins (between 11% and 26% depending on cohort) gave inconsistent answers about the age they were separated by more than 2 years. On top of that, 44% of twins “raised separately” in the study were raised by members of the same biological family, calling into doubt how unrelated their upbringing truly was.

Even if you accept the truth of all the studies Caplan quotes above, I think coming to a different conclusion is pretty reasonable. Bryan writes,

As shorthand, [if you are in the 80th percentile for some trait], if your counterparty in the thought experiment is in the 51st through 55th percentiles, I’ll call that a “small” effect. If he’s in the 56th through 65th percentiles, I’ll call that a “moderate” effect. Anything more I’ll call “large”.

Suppose intelligence were 100% genetic. If that were the case, having a father in the 80th percentile would mean your intelligence would be expected to be in the 65th percentile. Borderline a moderate effect, based on Caplan’s definition. In reality, intelligence is less than 100% genetic, meaning Caplan must believe that father’s IQ only has a “moderate effect” on IQ. This seems silly.

What’s really going on here? I think every variable Caplan measures, health, intelligence, education, income, is incredibly noisy. As a result, parental interventions with large effect sizes explain only a small percent of the variance in outcomes. It doesn’t change that the effect sizes are large. Combine that with a lot of studies having small sample sizes, and you make it really hard hard to rule out significant effect sizes with this methodology.

Separately, adoption studies could miss something meaningful. For example, maybe adopted kids are treated differently, or families that adopt aren’t representative in some important ways of the average family. My hypothesis here is that adopting signals a minimum enthusiasm for parenting that most families might not have. As a result, the adoption studies miss any benefit uninterested parents could achieve from becoming more involved.

Modern parenting has also changed substantially, compared to when these studies are done. Information is more accessible, so maybe it is easier to be an elite parent now, and identify which interventions can be helpful for your kids. Also, there could be prevalent harms today (maybe social media) that weren’t widespread when these studies were first conducted. Perhaps parenting today matters more than ever.

How can you be a 99th percentile parent?

My first thought when I read all of this was, doesn’t Bryan homeschool his own children? If he truly believes that parents don’t have too much of an impact on their children’s lives, why go to all the effort?

All adoption and twin studies look at parents on average. None of these studies are well powered enough to rule out the interpretation that there are some 99th percentile parents who are genuinely improving their children’s lives.

One intervention that seems to improve children’s lives substantially is private tutoring. The effect size is really large. One meta analysis showed that private tutoring has an effect size 5 times higher than any other education intervention tested when it comes to boosting test scores. Private tutoring is the perfect example of the type of intervention that wouldn’t show up in adoption studies, because so few people do it, but might have a meaningful impact on children.

Other interventions, such as taking time off around the birth of your child, have also been shown to have substantial benefits. Maybe for people who want their children to reach the absolute tops of their field, discipline and significant educational investments are worthwhile. I don’t know what the secret sauce is here, but I find Emily Oster’s substack to be a great repository for evidence-based parenting advice. For what it’s worth, Oster spends most of her time taking the Caplan view that many things are overblown.

Libertarianism

Caplan is a big libertarian. I read that subtext throughout his book. If being raised in a poor neighborhood doesn’t meaningfully change your life outcomes, maybe we don’t need to pour more money into underprivileged neighborhoods. Maybe we’re just throwing good money after bad.

The problem for Caplan, all of the best evidence for nurture effects comes from sustained programs for poor families. A mother’s pension program in the early 1900s had huge effects on recipients. The children of accepted applicants earned 14% more than the children of applicants who were rejected from the program. Another study found that each $1,000 received in EITC benefits increased the later life earnings of teens by 2.2%. Another study looked at Native American tribes before and after their family received casino income. Children whose parents received the income during their childhood were more likely to be employed at 25 than children whose parents did not.

This might all be consistent with Caplan’s arguments. Maybe the poor families just can’t afford to adopt, so they don’t show up in the data. By the time you’re stable enough to adopt, you’ve already picked all the financial low hanging fruit where you can turn money into children’s success.

I’m not accusing Caplan of making this argument, but this is important context to keep in mind. A strong belief in the power of genetics also implies an anti-libertarian argument. If you think genes are often destiny, it is more important to fund a robust social safety net, as we owe a minimal standard of living to people who are poor because of a bad luck of the genetic draw.

Conclusion

Caplan quotes astronomer Carl Sagan,

“Extraordinary claims require extraordinary evidence.” Caplan thinks “Behavioral genetics comfortably passes Sagan’s test…Parents have little long run effects on their kids.”

I think 90% of parents I know could take away an important lesson from this book. Be less neurotic about your children. There’s no need to take them to extra-curricular activities you don’t want them to go to, or fight painful battles over getting them to read their books. That being said, some parenting endeavors are useful, and parents should be cautious before casting away decades of conventional wisdom.