Saturday, August 27, 2011

More Freakonomics causality

I tuned in to APM's Marketplace recently and heard the following from Stephen Dubner:
Since the introduction of the ultrasound in Asia, in the early 1980s, it's often been used to determine the gender of a fetus -- and, if it's female -- have an abortion. In a part of the world with big populations, these sex selection abortions have had a big, unintended consequence.

Hvistendahl: I mean there are over 160 million females missing from the population in Asia, and to put that in perspective, it's more than the entire female population of the United States.

So, what happens in a world with too many men? For starters, there's more sex-trafficking, more AIDS, and a higher crime rate. In fact, if you want to know the crime rate in a given part of India, one surefire indicator is the gender ratio. The more men, the more crime. Now, the ultrasound machine didn't create these problems, but it did enable them. So, you have to wonder. What's next?

The hypothesis that increasing the ratio of men to women would produce "more sex-trafficking, more AIDS, and a higher crime rate" is entirely reasonable, but like so much observational data there's a big self-selection factor here. Families and women not involved in the sex trade tend to avoid rough neighborhoods and red light districts. There's also a question about outliers -- a few very bad areas with very high male to female ratios.

Once again, the suggestion that changing gender ratios would have significant social consequences makes perfect sense, but if you want to go from sensible suggestion to well-supported hypothesis, it's not enough to mention a fact that points in the right direction; you also have to show how other explanations (self-selection, sampling bias, etc.) can't explain away your fact. That, unfortunately, is where Freakonomics and many other economists-explain-the-world books and articles fail to make the grade.


  1. This one they might have better evidence for. There is a not insignificant branch of anthropology that goes back the the 1960's that has come to similar conclusions re: Sex Ratios. It is true that you can't do a controlled experiment (due to timeframes and ethics), but even American and European culture have seen these shifts (and the direction seems right unless there is a powerful third variable).

  2. Just to be completely clear, I have no problem whatsoever with the conclusion. That said, this example seems to have an obvious potential for self-selection. If Dubner were to mention this, even dismissively, I'd be satisfied. Lots of popular econ writers have a tendency to treat a good example as a rigorous proof.

    On a different tangent, I'll bet you could find some pretty good natural experiments that demonstrate this effect.