She’s a Hurricane: Evidence That Gender Bias Is Not All Fully Rational Statistical Discrimination

I have gotten much more interested in the issue of gender-bias since working with my (still anonymous) coauthor on our column “How Big is the Sexism Problem in Economics?”  It is often devilishly hard to tell whether discrimination is statistical discrimination that would be rational for an unbiased individual based on genuine differences between groups or evidence of actually being willing to pay a price to discriminate. Sometimes it is possible to identify directly the price someone is paying to discriminate by the higher profits or otherwise better deal to be had by dealing with disfavored groups. But there is another type of evidence for gender bias: situations in which everyone ought to know there is no genuine difference by gender in which people treat males differently than females. Such is the case with hurricanes. I was pointed to a report by Nicholas Kristof in the New York Times on the following research:

Researchers find that female-named hurricanes kill about twice as many people as similar male-named hurricanes because some people underestimate them. Americans expect male hurricanes to be violent and deadly, but they mistake female hurricanes as dainty or wimpish and don’t take adequate precautions.
The study, published in the Proceedings of the National Academy of Sciences, underscored how unconscious biases shape our behavior — even when we’re unaware of them.

The article should have completed the logic by stating that male and female names are in fact assigned to hurricanes in a way unrelated to severity, but I assume that is the case, since meteorologists on their own would not want to deal with the criticism from discriminatory naming of hurricanes. 

Update: Of course, everything I say above depends on the names of hurricanes being randomly assigned. One reader pointed out that all hurricanes used to have female names, so if they didn’t correct for that, the conclusion I make above is not warranted. But in that case, the data should be reanalyzed for the later time period when hurricane names were randomly assigned. 

Make sure to read the comments below, which cast doubt on the care of the analysis in the paper behind Nicholas Kristof’s article. In the light of the criticisms, I have to apologize for relying too much on Nicholas Kristof’s vetting of the empirical paper, just as I had to apologize for relying too much on Carmen Reinhart and Ken Rogoff in “An Economist’s Mea Culpa: I Relied on Reinhart and Rogoff.”