News

What’s in a (Hurricane) Name?

Closer scrutiny reveals problems with study methodology, results


andrew

An image from the National Oceanic and Atmosopheric Administration of 1992's Hurricane Andrew at its peak. 

By Jeremy Freese

freese
Jeremy Freese

Recently, a lot of media attention was focused on a study supposedly showing that “female hurricanes are deadlier than male hurricanes.” The idea is that people do not take hurricanes named after women as seriously, and so do less to protect themselves when warned about a hurricane named (say) “Bonnie” versus one named “Andrew.”

Given that gender biases are real, the basic idea sounds at least plausible. Unfortunately, the “hurricane-names study” is a doomed one, providing no evidence about whether its clever hypothesis might be true. After analyzing the same data, I have identified four key problems:

First, it is simply not true that more people die in hurricanes named after women. There is no statistical difference in the number of deaths resulting from hurricanes with “female” versus “male” names.

Second, the study’s real claim is that the more severe a hurricane is, the more people it kills, and this rate of increase is bigger for “female” hurricanes. However, the authors’ own results do not even show this.

They measure a hurricane’s severity in two ways: How much property damage the hurricane causes in dollars, and how strong the hurricane is in meteorological terms, based on its atmospheric pressure. (Lower atmospheric pressure typically means stronger winds and more potent hurricanes.)

The paper, however, only discusses the results for monetary damages, which do fit their theory. You have to look up the fine print in the study’s notes to find the results for atmospheric pressure—and these are completely the opposite: The “male” hurricanes are more deadly.

Third, if we follow the authors and just focus on monetary damages, we also find a big problem posed by Hurricane Andrew in 1992. It caused an estimated $25 billion in damages, which at the time was unprecedented, but having taken 66 lives, it was nowhere close to being the deadliest.

Had Andrew been the second hurricane that season instead of the first, it would have been named “Bonnie.” The authors’ own statistical results predict that “Bonnie” would have killed 23,000 people instead of Andrew’s 66. Even 2005’s Katrina, the deadliest U.S. hurricane of the past eight decades, only killed approximately 1,200 people. It is hard to that believe 20,000 Floridians owe their lives to a hurricane having been called Andrew instead of Bonnie.

Now consider Hurricane Diane, which in 1955 caused three times more deaths than Andrew, but with less than a quarter of the economic damage. The study’s results predict that more than 90 percent of those who died during Diane would have survived had it been given a manly name like “Andrew.”

Fourth, perhaps Andrew and Diane are just extreme cases. What about all the other hurricanes?

That’s just it: Remove Andrew and Diane from the data set and the findings disappear. The study results depend completely on Andrew and Diane, and their wild predictions about how many deaths would have been lost or saved if only the names had been different.

The regrettable reality is that even if one does think a hurricane’s name really might affect some people’s behavior in ways that sometimes kills people, there simply have not been enough U.S. hurricanes to detect this statistically. Only if you believe hurricane names have gigantic effects—bigger than the difference between a Category 1 and Category 4 hurricane—could there be enough data to tell whether “female” hurricanes or “male” hurricanes are more deadly, or if no difference exists.

This problem is what makes the study truly a preventable disaster. The problem was compounded by an overly dramatic press release by the lead author’s university. The media did not just stumble across this study; rather, they took the press release from a well-known institution at face value and ran with it. Little surprise that a line from the press release that the research was “proof positive” and “the first to demonstrate that gender stereotypes can have deadly consequences” was often translated by reporters and editors as “sexism kills.”

We, academics, often blame the media for extreme presentations of scientists’ work. Yet we should have little patience for extreme presentations of scientists’ work that the scientists’ own employers distribute to the media. We should examine such studies with special scrutiny.

Studies about what kills real people in real natural disasters deserve far more care.

Jeremy Freese is Ethel and John Lindgren Professor of Sociology and an IPR fellow. He is also co-principal investigator of Time-Sharing Experiments for the Social Sciences and contributes to Scatterplot, a social sciences blog. “Female Hurricanes Are Deadlier than Male Hurricanes” can be found at www.pnas.org/content/early/2014/05/29/1402786111.