In a recent column, I argued that the most plausible reason that couples with daughters divorce more often than couples with sons is that parents, on average, prefer boys. Since then my e-mail box has overflowed with objections, alternative theories, and requests for clarification. In the meantime, I've changed my own mind. I think I have a much better way to explain the facts. So, I want to revisit these issues.
First, to the e-mail. The easiest ones to dispose of are those that question the facts, particularly the claim that a one-child American family with a daughter is 5 percent more likely to get divorced than a one-child family with a son. The facts are accurate. The divorce numbers come from a study by Gordon Dahl and Enrico Moretti based on more than 3 million observations from the U.S. Census. With that kind of sample size, there's no way the 5 percent difference is "just a coincidence," as some readers suggested.
Other readers accepted the reality of the 5 percent difference but questioned the conclusion that daughters cause divorce. After all, marriages differ in all sorts of ways that might be relevant—financial stresses, infidelity, emotional distance. The phrase "correlation does not imply causation" popped up a lot.
But in this case, correlation does imply causation, and here's why: If you take 3 million people, have them all flip coins, and divide them into two groups according to whether their coins came up heads or tails, then the two groups are going to look statistically identical in every way—same average income, same average intelligence, same average height. That's called the law of large numbers, and it works for two reasons—first, the sample size is huge, and second, coin flips are random. Now do the same thing, dividing your 3 million people according to the gender of their last-born child. The same thing happens—parents of boys are going to be statistically identical in every way to parents of girls, because you've still got a huge sample size and because the sex of a child is as random as a coin flip. Since everything else is equal, the only thing that can be causing the difference in divorce rates is the gender of the children.
"Why not just ask people why they got divorced?" inquire several e-mails. Answer: You can't feasibly ask 3 million people why they got divorced. You could feasibly ask 3,000, but then your results would be statistically suspect because of a small sample size.
In fact, there are published surveys indicating that mothers of boys are happier on average than mothers of girls. These surveys are directly relevant to the point at issue, but I chose not even to mention them in the original column because Gordon Dahl convinced me that their statistical significance was suspect.
A number of readers offered the comment that, evidence be damned, they would simply never believe that the children's gender could be relevant to a divorce decision. My favorite of these came from a therapist in Iowa—it would probably be inappropriate to mention her name, so let's just call her "Bozo the Therapist"—who took me to task for the "archaic notion" that children ever have anything to do with divorce. Unless she's been practicing not in the state of Iowa but on the planet Iowa in some distant solar system, Bozo must win the prize for "least observant therapist in human history." The fact is that children do affect divorce decisions; if we didn't know this from statistical evidence, we'd still know it from common sense. And to a smallish but non-negligible extent, girls cause more divorces than boys do. Ignoring those facts won't make them go away.
The facts are clear and worth reporting, but there's legitimate controversy about what they mean. There are three key facts: 1) Parents of daughters are more likely to divorce than parents of sons; 2) in multichild families, parents of daughters are more likely to try for another child than parents of sons; 3) divorced mothers of daughters are less likely to remarry than divorced mothers of sons.
I originally said that all three facts point to a parental preference for boys. Several readers pointed out that the third fact—that divorced mothers of daughters are less likely to remarry—admits a better explanation: Mothers don't want to expose their daughters to a potentially predatory stepfather. Excellent point. So, I now think the evidence on remarriages is ambiguous regarding whether second husbands prefer boy stepchildren.
Other readers offered what they thought was counterevidence: Adoption agencies report a higher demand for girls. But this is exactly what you'd expect in a world where parents prefer boys. In such a world, boys will tend to be put up for adoption when there's something seriously wrong with them, but many girls will be put up for adoption simply for being girls. So, if I'm looking to adopt a bright healthy child, of course I'll choose a girl: I expect that among children put up for adoption, girls are on average brighter and healthier than boys. I could well make this choice even if I prefer boys to girls, as long as my preference for bright and healthy is stronger.