Talkback: "Do Baby Girls Cause Divorce?"

I listen to a lot of podcasts.  And there are so many times when I wish I could step into the podcast and start asking the speaker all kinds of questions... or presenting my arguments...!  So I decided that this blog will sometimes feature my talkback to my podcasts!

Today's post is a talkback to Freakonomics, one of my absolute favorite podcasts, and a recent episode called "Do Baby Girls Cause Divorce?"
In the episode, the ineffable Stephen Dubner interviewed Dr. Enrico Moretti, an economics professor at Berkely, who did a study recently on how the gender of a baby may affect the marriage of its parents.  According to his findings

• Couples who conceive a child out of wedlock and find out that it will be a boy are more likely to marry before the birth of their baby.
• Parents who have first-born girls are significantly more likely to be divorced.
• Fathers are significantly less likely to be living with their children if they have daughters versus sons.
• In any given year, roughly 52,000 first-born daughters younger than 12 years (and all their siblings) would have had a resident father if they had been boys.
• Divorced fathers are much more likely to obtain custody of sons compared to daughters.

This raises so many questions!

After looking into this issue, I do have some points to make. The first one is the use of the word "significantly."  Actually, the term is referring to "statistical significance," not necessarily the way we usually think of significance ("a huge, general problem!")  The effect is ~2.5% of the population.  Most people wouldn't consider something that is 2.5% to be generally significant, but it is enough for statistical significance.

Overall, I was disturbed by some of the assumptions raised in this podcast:

• Unaddressed assumption #1: The divorce/refusal to marry in these cases was the dad's idea.  This is not always the case.  About 66% of divorces in this country are initiated by the woman.  That doesn't necessarily mean the divorce is the woman's "fault" (she might be trying to get away from a real jerkface, for example...), but the chances that the gender of the baby affects the woman's choice is unlikely, since she is usually going to get custody of the children anyway.  The causes for divorce are legion; it is impossible to pin any divorce to just one cause, like the sex of the firstborn baby.
• Unaddressed assumption #2:  Boys are easier to raise than girls. This is a weird idea.  I've usually heard the opposite, actually, but I don't know which opinion is more prevalent in society.  And how would one go about defining "easy to raise"?  From an objective standpoint, males, statistically, have more behavior problems, are more aggressive, take longer to potty train, take longer to learn to talk, have a harder time sharing, and are more likely to develop autism than females.  This isn't to demean males in any way (I am madly in love with my two boys!!), but my point is that if there is an overarching societal attitude that boys are easier to raise than girls, it is not an attitude based on fact.  (Realize, I'm talking about the middle of the bell curve, here.  Of course, many girls are very difficult, and many boys are easy.  In fact, I know one parent who told me that after having such a hard time with her extremely strong-willed daughter, that she vowed never to have another child.  But statistically, she's probably an outlier.)  Whether these gender differences are caused by biology or cultural expectations, or some of both, doesn't affect my point.
• Unaddressed assumption #2: These differences explain a gender preference among American parents.  Well, actually, the differences found explain an effect that seems to be linked to gender, but this effect is not necessarily related to preference.  It COULD be linked to gender attitudes and stereotypes, such as "boys need dads more than girls do" (and therefore, this divorce is OK, because I have a daughter not a son.) This possibility wasn't raised in the paper.  The authors of the paper came to the conclusion that these data reflect gender preference, when there could be other gender-related explanations as well.  Actually, their third set of data, about how fathers are more likely to get custody of sons than of daughters, supports my hypothesis, that these effects are linked to stereotypes about what boys vs. girls need, more than preferences.

One question I had was "Why is this study focused on marriage?  Many people have children and choose not to get married.  I understand that this would make the data much more difficult to obtain, because there is no legal paper trail in such arrangements. But I wondered.  So I opened the actual paper.  One of their footnotes said that, according to the Census data they were studying, only 4% of children are living in homes with two unmarried heterosexual adults.  I find that interesting; I'd expect the number to be higher.  I wonder what the data is for how many children are living in homes where the mother never married at all, and how do those statistics relate?

But then I saw something that made me really surprised. The data they used for this study comes from U.S. Censuses from 1960-2000! 

1960!! 

What!?  Anyone with any knowledge of U.S. history whatsoever knows that a LOT has changed in American culture since 1960!  Why didn't the authors of the study focus on more recent Censuses to get an idea of where we are NOW?  They did say:

For completeness, in Table A1, we report the effects of a first-born daughter broken down by Census year, decade of birth, race, and education. These estimates by subgroup are necessarily less precise because they are based on smaller samples and therefore are not the main focus of our analysis. 

Well, I would think that if you are wanting to try to measure whether America HAS a problem with parental gender preference that is big enough to affect something as important as whether parents stay together based on it, you'd want to look at current numbers.  I did go to Table A1, and there are indeed different numbers from decade to decade, but I am not trained as a statistician, so I don't really know how to interpret the numbers very well. So I won't comment further on that point.

The bottom line:  I'm not trying to disparage Freakonomics, nor the authors of this study, but I don't see anything that is solid enough in this study to really base any kind of alarm on.