The question is: if peer effects are important, why don't they show up in adoption studies?

Generally speaking, when I search for literature on peer effects, the information is sparse and confusing. I'm not too surprised, since such effects are much more difficult to disentangle than heritability and shared environment.

My working hypothesis is that:

  1. Peer effects matter a lot, but only up to a certain threshold of peer quality, and this threshold is basically what people intuitively perceive as sufficiently respectable company for their kids. So, basically, underclass peers will ruin your kids, but upper-class or genius peers won't improve things relative to the company of ordinary middle-class kids. (Just like downright abuse will ruin them, but helicopter parenting won't improve them.)

  2. In order to quality for adoption, people must pass through sufficiently strict checks that they are highly unlikely to provide an environment below this threshold. So there aren't any good natural adoption experiments that expose kids to underclass peer groups.

I'd be curious to hear about any contrary evidence, though.

I sometimes get a very "politics" vibe from your comments [...] For example, [the middle paragraph in the above comment] reads a lot like standard political rants.

Maybe it does, but this really is my honest impression of what the situation looks like to a typical person aspiring to a middle-class lifestyle these days. I'm curious if you would disagree with any of the following statements, which seem to be roughly equivalent to what I wrote above (all given in the context of contemporary North America):

  1. A house in a place where your kids will grow up with -- and, in particular, go to school with -- kids from respectable middle-class families is very expensive. In many places, and especially prosperous centers of economic activity that offer good career opportunities, it is somewhere around an order of magnitude above the median yearly household income.

  2. Unless one is extraordinarily wealthy, to obtain such a house, one has to get into debt that is, just like the house price, enormous relative to one's income.

  3. Such debt, due to its sheer size, can't be repaid in any time shorter than several decades. Just to pay the interest, let alone to make any dent in the principal, one must part with a significant part of one's income during this period. In this situation, a plausible bad luck scenario like job loss, health problems, etc. can easily push one into insolvency.

  4. Worse yet, this situation implies that the bulk of one's net worth is completely non-diversified and invested in a single asset -- of a sort that is notoriously prone to bubbles and price crashes. Even worse, the occurrence of such crashes is positively correlated with bad economic conditions that make job loss and decreased earning power especially likely.

  5. Even with a minimalist approach to parenting, raising kids is expensive. Each additional kid makes it less likely that one will manage to remain solvent under the above described conditions.

  6. Sufficiently bad financial ruin can plausibly put one into a situation where one is no longer able to afford to ensure a peer group for one's kids that will be above the threshold where bad peers exercise significant bad influence. Also, generally speaking, below a certain class threshold, all sorts of social pathologies are rampant to a degree that seems frightful to a typical middle-class person -- and, again, bad financial ruin can make one unable to afford to insulate oneself from people that fall below this threshold.

  7. Taken together, (1)-(6) makes for a rather stressful existence, in which having more kids will seem to a lot of people like an additional burden in an already difficult situation, and an additional risk in an already uncomfortable gamble.

I'd be really curious to see where exactly our opinions diverge here.

Showing 3 of 7 replies (Click to show all)

I agree with you about peer effects but I think you assume without cause that they are lasting. Twin studies, despite their flaws, would seem to be the best way to establish that, whatever their influence now, differing peer groups and adolescent environments will not lead to different adult characteristics. Any efforts that parents take NOW to improve children's peer groups, in order to improve current well-being, are valuable. But if the effect fades over time, then the harm is less than most parents think it is, and therefore on the margin children are less costly and Caplan's conclusion is warranted.

1Ghatanathoah8yPerhaps another reason peer effects don't show up is that situations consisting of one kid of upper class background completely surrounded by lower class kids and having no other options but them as a peer group are relatively rare. In most cases there are a number of other middle classish kids in the same boat to form a peer group with. I base this conclusion on two pieces of evidence, the first is anecdotal, my own school background. My school had a variety of kids that included a large amount of lower class kids from a nearby trailer park and a large amount of respectable kids. For the most part nothing the trailer park kids did rubbed off on me or any of the other respectable kids because we rarely socialized with them, we naturally tended to interact with the kids we had something in common with (although most of the trailer park kids were friendly enough in class, I can't really say that most of them were unpleasant to be around). The only long-term impact they had on me was to help me realize that the underclass are usually trying to be nice people, even if they fail at it a lot. The second piece consist of articles (mostly by Thomas Sowell) I've read about various immigrant communities in poor neighborhoods and how kids from groups with middle-class values (i.e. Chinese, Jewish immigrants) tended to cluster together and interact with each other and not the poorer kids around them. So it seems plausible to me that the ability to form small clusters of like-minded peers might mitigate peer effects. Also, I second jsalvatier's points about some of your comments having a "political" feel. In particular it seems like you have a tendency to work in angry-seeming statements about how awful and unpleasant poor people are that can be rather off-putting, to say the least.
2juliawise8yThis may be true in suburbs, but not everywhere. My mother grew up in a small college town in Kentucky. Her parents, and the parents of her friends, were mostly college educated but below national median income. They lived in small, inexpensive houses. Local kids had access neither to ballet and karate lessons nor drugs and gangs. Her social life focused around church and folk dancing. From what I understand, it was a high-quality, low-cost childhood.

Review: Selfish Reasons to Have More Kids

by jsalvatier 3 min read29th May 2012260 comments


This is a review of Bryan Caplan’s book Selfish Reasons to Have More Kids. Co-written with Walid.


Adoption studies indicate that differences in parenting styles have mostly small impacts on long term life outcomes of children, such as happiness, income, intelligence, health, etc.. This means that parents can put less effort into parenting without hurting their children’s futures. If you think kids are neat, then you should consider having more. 


Note: We think this is a pretty useful book, and it has changed our minds on how many children we want to have, though neither one us has any children yet. Also, neither of us are experts on twin or adoption studies.

Caplan argues that parents drastically overestimate their ability to improve the adult lives of their children. His argument is driven by adoption studies, which suggest that there is very little that parents can do beyond techniques employed by the average parent that would get them better results with their children. Specifically, the following areas are identified as areas where differences in parenting don’t seem to matter:

  • No effect on life expectancy, overall health (as measured by the presence/absence of particular health problems and self reported health), height, weight or dental health.
  • No effect on intelligence.
  • No effect on various measures of personality: conscientiousness, agreeableness or openness (not certain about extroversion or neuroticism).
  • Little or no effect on marriage, marriage satisfaction, divorce, or child bearing.

But that is not to say that styles outside of the average do not matter at all -- there are a few areas where parenting differences do seem to have an effect:

  • A small effect on adult drinking, smoking and drug problems.
  • A small effect on educational attainment, but no effect on grades in school or on income.
  • A large effect on political and religious labels, such as whether you call yourself democrat or republican or Christian or Muslim but small effects on actual political and religious attitudes or behavior.
  • A moderate effect on when girls start having sex (but not boys), but no effect on teen pregnancy or adult sexual behaviors. 
  • Possibly a small effect on sexual orientation.
  • A moderate effect on how children remember and perceive their parents.

So how do adoption studies lead to these conclusions?

Adoption studies (If you have a link to a better overview or discussion of adoption studies, we'd appreciate it) help find out the influence of parenting differences on adult outcomes by comparing adoptees to their adopting family. If adoptees systematically tend to be more like their adopting family than like other adoptees along some measure (say religiosity or income), that implies that parenting differences affect that measure.

When an adoption study finds that parenting does not affect outcome X, it does not mean that parenting cannot affect it, just that the parenting styles in the data set did not affect it.

The evidence Caplan talks about is primarily long run life outcomes. Shorter run life outcomes often do show larger effects from parenting, but these effects diminish as the time horizon increases.

If parenting doesn’t matter, what does?

Caplan references twin studies in showing that genetics have relatively big effects on all the measures previously mentioned. This explains why we see strong correlations between parents’ traits and children's’ traits. He specifically uses it to call out attributes that we would commonly ascribe to parenting, but may actually have a much larger genetic component.


Once Caplan has argued for the stylized fact that parenting has only small effects on major life outcomes, he explores some of its implications. 

Don’t be a tiger parent

One big implication is that you should put less effort into trying to make your kids into great adults and more effort into making your and your kids’ lives more fun right now.

For example, parents probably spend too much energy convincing their children to eat their vegetables and learn the piano, given that it won’t affect whether they will eat healthy as adults or be more intelligent. No one likes fighting. If you want your kid to learn the violin so they’ll have fun right now, it may very well be worth it, but don’t do it because you think it will increase their future income or intelligence. If neither you nor your child likes doing an activity, consider whether you can stop doing it.

Adoption studies provide good evidence that most activities don’t have a much of a long term effect on your children, so you need good evidence to start thinking that an activity will be good for your kids future. The odds are against it.

Have more kids

Focusing more on making your and your children’s lives more fun means that overall, having kids should be more attractive. If having another kid no longer means fighting about finishing their broccoli every night, maybe it’s not such a bad idea. On the margin, you should consider having more kids. If you were planning to have zero kids, consider having one. If you were planning to have 3 kids, consider 4, etc.

Other Topics

In much of the rest of the book Caplan gives common sense advice for making parenting easier for the parents. A couple of these, such as the Ferber method for dealing with infant sleep problems, are empirically based.

Here are some other topics Caplan discusses in his book:

  • Happiness research on parenting. Caplan argues that although being a parent seems to make people less happy, the effect is small (Ch 1).
  • Child safety statistics. Children are many times safer than in decades past (Ch 4).
  • Many of the benefits of having children come later in life (e.g. having people who will come and visit you, etc.), which makes it psychologically easy to ignore these benefits (Ch 5).
  • The externalities of children. He argues that on net, extra people have large positive externalities (Ch 6), so you shouldn’t feel guilty for having more children.

What parts should I read?

We wholeheartedly recommend reading the first 5 chapters (121 pages) of Selfish Reasons To Have More Kids as these have the most useful parts of the book; the rest of the book is less valuable. 

Criticisms of Selfish Reason To Have More Kids

There are a number of criticisms relevant to Caplan’s arugments. For example:

  • Nisbett claims that heredity is much less important for IQ than thought (see also counterclaims posted below).
  • Will Wilkinson claims (one, two) that the cost of parenting plays a small role in people's family size decisions, thus it's not a very strong reason to have more kids.
  • Jason Collins likes the book but would like it to discuss the research on non-shared environment (i.e. that not explained by genetic or parenting differences, such as peer effects) (link).