Hurricane Issac struck New Orleans on the 2012 anniversary of a far more catastrophic hurricane--Katrina.
A new study, published in PNAS caused some news this week.
Do people judge hurricane risks in the context of gender-based
expectations? We use more than six decades of death rates from
US hurricanes to show that feminine-named hurricanes cause
significantly more deaths than do masculine-named hurricanes.
Laboratory experiments indicate that this is because hurricane
names lead to gender-based expectations about severity and this,
in turn, guides respondents’ preparedness to take protective action.
This finding indicates an unfortunate and unintended consequence
of the gendered naming of hurricanes, with important
implications for policymakers, media practitioners, and the general
public concerning hurricane communication and preparedness.
It's a very interesting study.  Does implicit gender bias actually inform whether Americans will prepare for an oncoming hurricane? Novel idea, interesting hypothesis, original research that used Amazon's Mechanical Turk service to find volunteers.  But ultimately rather flawed.

(if the link doesn't work, it's because of PNAS's embargo policy.)

For the first part, the researchers took the 1950-2012 database. In this time period, 94 hurricanes struck the United States. Tropical storms were not included.

Two hurricanes were eliminated: Katrina, in 2005, which officially took the lives of some 1,833 people and Audrey, in 1957, which officially took the lives of 416 people. I say officially for both because it's believed both hurricanes killed more people then are officially tallied. At least 135 are still missing due to Katrina. If there were more, we may never know.

Only United States damage and United States death tolls were used, and only the western Atlantic basin's storms were used. Eastern Pacific hurricanes were not used (although many do not ever make landfall). Other basins that use gendered names for tropical cyclones such as those that fall under Australia's responsibility were not studied.

The study does NOT, repeat, does NOT state that American sexism makes residents disregard warnings. Let's get that part out of the way before we continue. We good?Let's continue.

The Archival Evidence

First, we'll examine the archival evidence. The 92 hurricanes that ultimately were analyzed cover the years 1950 through 2012 (the paper was written in 2013). Eagle-eyed folk will spot, perhaps, a huge flaw here. How can gendered names be taken into account before  1978? Between 1953 and 1978, women's names were used (the names from 1950 to 1952 were taken from the Joint Army/Naval Phonetic Alphabet system.)

In the 1953-1978 period, there certainly were some deadly storms. Agnes in 1972 killed 117 people, largely due to inland flooding in central Pennsylvania and New York. Camille in 1969 killed 256 people on the Mississippi Gulf Coast and in mudslides in Virginia. Hurricane Diane in 1955 washed away a camp in Pennsylvania and took 200 lives across the northeastern US. These storms, and their feminine names, were included in the list.

Capital Weather Gang (the Washington Post)

Researchers at the University of Illinois and Arizona State University examined six decades of hurricane death rates according to gender, spanning  1950 and 2012.  Of the 47 most damaging hurricanes, the female-named hurricanes produced an average of 45 deaths compared to 23 deaths in male-named storms, or almost double the number of fatalities.  (The study excluded Katrina and Audrey, outlier storms that would skew the model).

The difference in death rates between genders was even more pronounced when comparing strongly masculine names versus strongly feminine ones.
“[Our] model suggests that changing a severe hurricane’s name from Charley … to Eloise … could nearly triple its death toll,” the study says.

Let's use Hurricane Eloise for just a moment. Striking in September 1975 at Category 3, it was destructive along a portion of the Florida panhandle coast. Building codes changed significantly after Eloise. Eloise directly and indirectly killed 21 people in the United States, largely through inland flooding and none in Florida where it made landfall. When somewhat more powerful Hurricane Frederic approached the same region in 1979, 5 died (4 inland, only 1 at the coast). The people remembered Eloise from 1975. This is something the study does not seem to capture, even through statistical analysis. As for Charley, in 2004, it struck a less populated area of the Florida coast although a last minute jog saved Tampa-St. Petersburg from being obliterated. It was still highly damaging on its trip across the peninsula.

Another thing not captured, and something the Hurricane Center and Weather Service have been making great pains to correct, are the risks from inland flooding and storm surge. Most people in hurricanes die because of the water and not the wind and this is a global phenomenon. Many deaths in the storms mentioned were inland. Eloise's floods were in the northeast US, not in Florida and Alabama where it came ashore.

Picking through the list of named storms I can find if the researchers had done further digging, they'd have found interesting stories. One can pick Hurricanes Alicia and Allen. Both storms struck Texas in the early 1980s but Alicia was deadlier. However Allen struck a sparsely populated region north of Brownsville, and Alicia rolled right over the densely populated Houston metro. Ike struck the Houston metro in 2008 and was far deadlier than both Alicia and Allen (in the US) combined, despite being a weaker hurricane. Both Ike and Alicia highlighted a serious problem with evacuating the city of Galveston--for a variety of reasons, people refuse to leave (complacency) or actually cannot leave (poverty) that city when under threat. Neither complacency or poverty were controlled for in the study.

The two hurricanes that struck the United States in 2003 were Claudette and Isabel. Claudette struck a sparsely populated part of the Texas coast. Isabel roared right up the Chesapeake, and its media blitz made sure everyone knew it was coming well in advance. The study does not seem to normalize for population distribution or "If there are more people in the way of a hurricane, more people will die."

The study also doesn't seem to normalize for the fact that forecasts have become much better since 1950 and even more so since 1978. Forecasts for storms like Sandy, which for a full week was forecast to do what it ultimately did, are still outliers, but they will become more common as the skill at forecasting gets even better and better. Looking at the 1950 through 2012 data you can clearly see, regardless of gendered names, that hurricanes kill far fewer people after the 1970s than before (and yes, Katrina is an extreme outlier). Hurricanes killed far more Americans before 1950 as well.

Forecast fatigue is also perhaps not accounted for in the study. When Allen missed Houston and went far to its south in 1980, residents of Galveston Island were less likely to take action when Alicia threatened in 1983. Take Irene and Sandy. They both hit New Jersey and New York, roughly a year apart. Irene was heavily hyped in media for the New Jersey coast and for New York City. It even got a How I Met Your Mother episode dedicated to it.  Irene was not as damaging along the coast as it was inland and this affected how coastal residents dealt with Sandy, whose media hype began while the storm was just a collection of thunderstorms south of Jamaica. Because Sandy was forecast to be and ultimately was an incredibly complex storm the Weather Service and National Hurricane Center took steps that may have hampered warnings to the public for that storm. Suffice to say New Yorkers will remember Sandy now. However a postscript suggests that people only remember the worst effects of a hurricane at their location for about seven years. This isn't a gendered names problem. It's a sociological problem due to how people perceive risk and in the US, it's compounded by the fact that a Category 3 or higher storm hasn't hit since 2005.

Sociologists estimate, however, that people only remember the worst effects of a hurricane for about seven  years (B. Morrow, personal communication). One of the greatest concerns of the National Weather Service's (NWS) hurricane preparedness officials is that people will think that no more large loss of life will occur in a hurricane because of our advanced technology and improved hurricane forecasts. Bill Read, current Director of NHC, as well as former NHC Directors, have repeatedly emphasized the great danger of a catastrophic loss of life in a future hurricane if proper preparedness plans for vulnerable areas are not formulated, maintained and executed
I can't entirely fault the researchers for knowing the above stuff. None were meteorologists. That's okay, but I think it would have enhanced the study and perhaps they'd have received a different result.

The Other Experiments

Now the experiments are a bit more interesting.

The researchers used Amazon's Mechanical Turk as well as University of Illinois students to evaluate responses to masculine or feminine names. Each experiment has a new group of people evaluating the experiment. The table above summarizes the findings.

The effects with modeled hurricanes appear slight but they are pretty clear. Male-named storms get a slightly stronger response than female-named ones---meaning male named storms were perceived as more dangerous than female named ones. This is where the study is novel, and neat, and original; no one has considered this.

The results from the model storms suggest the naming system may be working against storm preparation by tapping into the implicit biases that people carry. We see these in other arenas: in the US names that appear black will get a resume bounced; overseas, in places like Sweden foreign names will get a resume bounced. We've discussed at length #YesAllWomen, and noted how the media covers missing young women. Implicit and explicit biases are everywhere.

I'm not sure what the researchers have found here, but they seem to have found something in modeled imaginary storms. While I'm highly skeptical with storms that have happened, I think there might be something in the modeled storms that is worth looking further into to see if it can be replicated globally.

The Very Valid Criticisms

There are quite a few, documented in Mashable and by Ed Yong.

First, Mashable interviewed a number of people.

But outside experts who were not involved with the new study told Mashable that, though the study presents interesting results, the findings are questionable and do not justify making any changes to the naming system.

 If you had not told me that it was a real paper I would have assumed it was satire," said Robert J. Meyer, co-director of the Wharton Risk Management and Decision Processes Center at the University of Pennsylvania.

Meyer said the study ignores the improvements in the quality of storm warnings over time. The accuracy of storm warnings, he said, is one of the strongest predictors of storm-related fatalities.

"I have no problem believing the core result that when you put a female name on objects — a gun, a car, an earthquake — there is an implicit tendency among naïve subjects to ascribe less aggressive traits to it," Meyer said.

"But to take this, very small, effect observed in a web study and use it to conclude that there would have been fewer deaths in Katrina had it been named 'Ken' is simply ridiculous."

And the last part, I happen to agree with. Katrina was especially deadly because so much failure happened at the local government level in Louisiana. Katrina was also deadly because of choices made in the construction of canals decades prior, the poor construction of floodwalls, and the loss of wetlands all of which funneled the surge in to New Orleans, where the majority of deaths occurred. Poverty kept people from leaving the stricken city. I don't see how these failures would have been any different had Katrina been named Kenneth.

 I recommend reading the entire thing at Mashable, obviously, I can't blockquote the entire thing.

Ed Yong writes:

But [Jeff] Lazo thinks that neither the archival analysis nor the psychological experiments support the team’s conclusions. For a start, they analysed hurricane data from 1950, but hurricanes all had female names at first. They only started getting male names on alternate years in 1979. This matters because hurricanes have also, on average, been getting less deadly over time. “It could be that more people die in female-named hurricanes, simply because more people died in hurricanes on average before they started getting male names,” says Lazo.

Jung’s team tried to address this problem by separately analysing the data for hurricanes before and after 1979. They claim that the findings “directionally replicated those in the full dataset” but that’s a bit of a fudge. The fact is they couldn’t find a significant link between the femininity of a hurricane’s name and the damage it caused for either the pre-1979 set or the post-1979 one (and a “marginally significant interaction” of p=0.073 doesn’t really count). The team argues that splitting the data meant there weren’t enough hurricanes in each subset to provide enough statistical power. But that only means we can’t rule out a connection between gender and damage; we can’t soundly confirm one either.

Other aspects of the team’s analysis didn’t make sense to Lazo. For example, they included indirect deaths in their fatality counts, which includes people who, say, are killed by fallen electrical lines in the clean-up after a storm. “How would gender name influence that sort of fatality?” he asks. He also notes that the damage a hurricane inflicts depends on things like how buildings are constructed, and other actions that we take long before a hurricane is named, or even before it forms.

Then, there are the six experiments. As is common in psychology, the volunteers in the first three were all college students. “There is no reason to think that University of Illinois undergraduate students in hypothetical scenarios would have any relation to real-world decision making to populations in hurricane vulnerable areas,” says Lazo. The participants in the last three were recruited via Amazon Mechanical Turk—an online platform for finding volunteers.  Again, it’s unclear how representative they were of people who live in coastal, hurricane-prone towns.

Finally, Lazo says that there’s a lot of evidence on how people respond to hurricane threats, and how their decisions are influenced by their social situation, vulnerability, culture, prior experience, sources of information, when the hurricane makes land, and so on. “Trying to suggest that a major factor in this is the gender name of the event, with a very small sample of real events, is a very big stretch,” says Lazo. And if the archival analysis isn’t as strong as it originally seemed, then what the team has basically done is to show “that individuals respond to gender”—hardly a big deal.

I think the same way. People do seem to react, but not very strongly in either direction whether it's a male named storm or female named storm---look at the chart above. In the archive data the effect is far weaker if you just use post 1979 storms (you can play with the data here). Again, the US is better at storm preparation now than it was in 1950, despite a far larger population along the coasts that are hurricane-prone. Building codes were strengthened after Hugo in 1989 and Andrew in 1992, and again after the hurricanes of 2004 in Florida.

The fact that the study is US only is disappointing. I'm curious as to why this was limited to just the US. Caribbean culture reacts differently to hurricanes, as they are in the path of them almost every year. The Atlantic Ocean is not the only basin that uses gendered names. The South Indian Ocean does between Mauritius and Madagascar, as well as the east and west coasts of Australia (including the South Pacific.) The US is responsible for forecasting in the Eastern Pacific and on rare occasions a tropical storm force storm makes into inland Arizona. I'll be more willing to be less skeptical if the results of this study were replicated there and perhaps they were---Mechanical Turk workers can be anywhere. But we don't know.

Not bad. Neat study. It's something no one's considered. But I think ultimately flawed. Still, it's worth adding to the toobox, and the Weather Service is actively looking into the sociological reasons people react to weather. This flawed study may have found something that's worth a closer look, or not.

Originally posted to SciTech on Tue Jun 03, 2014 at 04:59 PM PDT.

Also republished by Community Spotlight.

Your Email has been sent.