National Geographic

Why Have Female Hurricanes Killed More People Than Male Ones?

Here’s a simple fact with an uncertain explanation: historically, hurricanes with female names have, on average, killed more people than those with male ones.

Kiju Jung from the University of Illinois at Urbana–Champaign made this discovery after analysing archival data about the 94 hurricanes that hit the US between 1950 and 2012. As they write, “changing a severe hurricane’s name from Charley to Eloise could nearly triple its death toll”.

Why?

The names certainly don’t reflect a storm’s severity, and they alternate genders from one to the next.

Jung team thinks that the effect he found is due to unfortunate stereotypes that link men with strength and aggression, and women with warmth and passivity. Thanks to these biases, people might take greater precautions to protect themselves from Hurricane Victor, while reacting more apathetically to Hurricane Victoria. “These kinds of implicit biases routinely affect the way actual men and women are judged in society,” says Sharon Shavitt, who helped to design the study. “It appears that these gender biases can have deadly consequences.”

But Jeff Lazo from the National Centre for Atmospheric Research disagrees. He’s a social scientist and economist who has looked into the public communication of hurricane risk, and he thinks the pattern is most likely a statistical fluke, which arose because of the ways in which the team analysed their data.

Let’s look at each of these arguments in turn.

First, Jung’s team asked nine people to rate the name of US hurricane on a scale of 1 (very masculine) to 11 (very feminine). They found that the more feminine names were linked to more damage (normalised to today’s value) and deaths. (They excluded Katrina because that was such a huge outlier.)

To test their hypothesis about gender biases, the team ran six experiments. (For stats junkies, here’s the table showing all the numbers behind the experiments; note that each one involves a fresh group of volunteers.)

When the volunteers saw a list of hurricane names, and nothing more, they guessed that male storms would be more intense than female ones. After reading a more detailed scenario about an incoming hurricane, they predicted that the storm would be riskier and more intense if its name was Alexander rather than Alexandra.

After reading another similar scenario, they were more likely to say that they would evacuate their homes if Hurricane Christopher was hypothetically bearing down upon them, than if Hurricane Christina was doing so. Likewise, if they read a voluntary evacuation order, they were more likely to comply in the face of Hurricanes Danny, Victor or Alexander than Hurricanes Kate, Victoria, or Alexandra respectively

These differences aren’t due to explicit sexism. When the team asked people directly if male or female hurricanes would be more dangerous, the responses were evenly split. “This suggests that the effects in the main experiments are implicit in nature,” says Shavitt. In other words, gender stereotypes influence our thoughts and behaviour, whether or not we buy into them outright.

But 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. [Update: To clarify, I mean that there's already a huge amount of evidence that individuals respond to gender, not that the biases themselves are no big deal. Of course, they are.]

All of this matters because Jung’s conclusions, if they’re right, could have implications for how hurricanes are described. “It may make sense to move away from human names, but other labels could also create problems if they are associated with perceptions of mildness or gentleness,” says Shavitt. “The key is to provide information and labels that are relevant to the storm’s severity.”

Lazo says, “If there’s reasonable validity to their findings, it is worth further exploration with real hurricane-vulnerable subjects. That would be the proper conclusion to their study, and absolutely not any specific policy recommendations about changing naming conventions!”

Update: The authors have responded to Lazo’s criticisms in the comments below (see also this PDF). You can also find other critical viewpoints at Mashable, Slate, and indeed, in some of the comments in this piece.

Update 2: Bob O’Hara and GrrlScientist have written another rebuttal at the Guardian, pointing out flaws in the paper’s model. Check it out. I’d also pay attention to comments from Harold Brooks below.

Reference: Jung, Shavitt, Viswanathana & Hilbed. 2014. Female hurricanes are deadlier than male hurricanes. PNAS http://dx.doi.org/10.1073/pnas.1402786111. If the link isn’t working, this is why.

More on stereotypes:

There are 77 Comments. Add Yours.

  1. Peter Apps
    June 2, 2014

    This is an explanation looking for something to explain. Divide any set of data similar to human deaths in hurricanes into two classes, and one class will have more instances in it than the other. What would really need an explanation would be male- and female-named hurricanes having killed the same number of people.

    Very dodgy to have deliberately excluded Katrina, which single-handedly reversed the difference.

    [No, Katrina killed the most people, so if you include it, it just strengthens the higher fatality trend for female names - Ed]

  2. Huh
    June 2, 2014

    > When the team asked people directly if male or female hurricanes would be more dangerous, the responses were evenly split.

    Wait, what? Is this written confusingly, or did this question did not allow for a “it would make no difference” option? I’m going to be extra suspicious of any conclusions from this survey if they have badly formed questions like that which exclude reasonable answers.

    As if the exclusion of Katrina wasn’t already a bit iffy. “This data point ruins our stereotypes and headline potential… maybe we should just mark it as an outlier.” I know that real scientists don’t think that way, but I have not yet been convinced that this studies authors are, in fact, real scientists and their methodology is suspect enough to leave that in doubt until I have more evidence.

  3. Huh
    June 2, 2014

    I was wrong about Katarina. I retract that point. I’m still suspicious here, but I did misunderstand that.

  4. Nancy Curtis
    June 2, 2014

    The article was ok, BUT THE TITLE WAS AWESOME!!!! ;)

  5. kylefree
    June 2, 2014

    I would love to see a follow-up that looks at media reports in the run-up to hurricane landfalls since 1979. It’s a sufficiently large data set, and a media bias could explain a difference in people’s preparedness.

  6. Shecky R
    June 2, 2014

    So many poorly-controlled variables here…and certainly the 1950-78 data, when only female names were used (and involving some devastating storms), shouldn’t be included. I agree with Lazo, “a very big stretch.”

  7. Kiju Jung, Sharon Shavitt, Madhu Viswanathan, and Joe Hilbe
    June 2, 2014

    We appreciate your careful attention to our study but suggest that looking closer at our reported results would have answered some of the questions raised in your article. Specifically:

    1. We are of course aware that all hurricanes had female names from 1953 through 1978. In 1979, they began alternating the gender of the names. However, our analysis primarily focused on the femininity-masculinity of names, not only on male/female as a binary category. Even during the female-only years, the names differed in degree of femininity (compare two female names: Fern, which is less feminine to Camille, a rather feminine name). Although it is true that if we model the data using only hurricanes since 1979 (n=54) this is too small a sample to obtain a significant interaction, when we model the fatalities of all hurricanes since 1950 using their degree of femininity, the interaction between name-femininity and damage is statistically significant. That is a key result. Specifically, for storms that did a lot of damage, the femininity of their names significantly predicted their death toll.

    Is this a statistical fluke? Lazo says, “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.” But no, that is not the case according to our data and as reported in the paper. We included elapsed years (years since the hurricane) in our modeling and this did not have any significant effect in predicting fatalities. In other words, how long ago the storm occurred did not predict its death toll.

    What’s more, looking only at severe hurricanes that hit in 1979 and afterwards (those above $1.65B median damage), 16 male-named hurricane each caused 23 deaths on average whereas 14 female-named hurricanes each caused 29 deaths on average. This is looking at male/female as a simple binary category in the years since the names started alternating. So even in that shorter time window since 1979, severe female-named storms killed more people than did severe male-named storms.

    2. Another question raised was whether it’s appropriate to look at both direct and indirect deaths. Please note that many of NOAA’s monthly weather reports that we used to obtain fatality data do not distinguish between direct and indirect categories. Direct and indirect deaths are often grouped together. The issue of indirect deaths has been addressed here: http://www.slate.com/articles/news_and_politics/explainer/2012/10/hurricane_sandy_how_to_count_the_fatalities.html That article reads in part: “Fatal car accidents caused by torrential rains or flooding are indirect deaths, but storms can also be blamed for so-called ‘natural’ deaths.” Deaths due to car accidents caused by washed out roads, or fires started by downed power lines, or heart attacks or other adverse health events that result from the storm may reflect preparedness. We believe these deaths should count and are appropriately included in the dataset.

    3. Hurricane names versus other factors that affect preparedness: We cannot claim (nor did we claim) that gendered naming is more important than the other factors that Lazo mentions. Those other factors certainly matter, as well. But that doesn’t mean we should ignore the apparent impact of the femininity of the names. Meterologists and hazard communication specialists have called for more attention to social science factors that predict how people respond to hazard warnings. Implicit biases represent an understudied factor that makes a difference.

    4. Policy Implications: We are not suggesting that policy be changed based on one study. As we wrote to Ed when he emailed us last week, we will leave such decisions to policy experts. What we are suggesting is that this finding merits further investigation. Our goal is to add to the knowledge in this area and to the ongoing policy conversation.

    Thank you,
    Kiju, Sharon, Madhu, and Joe

  8. Cleveland K. Evans
    June 2, 2014

    The variable these researchers were using in correlating the number of deaths with the names was not the “male” or “female” designation of the names by the Weather Bureau, but data from ratings of the “masculinity” or “femininity” of all the names used since 1951. Although male names are obviously going to average out much more “masculine” than female ones, there will also be differences within the categories. Among pre-1979 names which were retired, for example, most modern Americans would probably rate Camille, Celia, and Audrey as being more “feminine” than Agnes, Ione, and Carmen, even though they are all “female” names.

    The “asking people directly” also does requires more explanation. What they did there was take all the names on the 2016 and 2017 lists and create contrasts between all the pairs starting with the same initial (Alex vs. Arlene, Bonnie vs. Bret, down to Walter vs. Whitney) and have people rate in each of those pairs which hurricane they thought would be the stronger one. They didn’t get a significant difference between ratings of male and female names that way. I myself am not sure that was the best way to judge “explicit” beliefs about gender name differences, though.

  9. Kristine
    June 2, 2014

    This was yet another study where the “researchers” already had a conclusion they wanted to make and then set out to make the data fit their desired conclusion.

    Personally, I have never met a single person who cared more about a storm name than the storm’s location and severity.

  10. Hanna
    June 2, 2014

    “This was yet another study where the “researchers” already had a conclusion they wanted to make” It’s called a hypothesis and most scientific research has one.

    Having said that, I find the explanation for these findings difficult to believe but serendipity does exist.

  11. Mike
    June 2, 2014

    “What’s more, looking only at severe hurricanes that hit in 1979 and afterwards (those above $1.65B median damage), 16 male-named hurricane each caused 23 deaths on average whereas 14 female-named hurricanes each caused 29 deaths on average. This is looking at male/female as a simple binary category in the years since the names started alternating. So even in that shorter time window since 1979, severe female-named storms killed more people than did severe male-named storms.”

    Talk about small-number statistics. That’s really not a big difference when we’re talking about 30 storms totaling about 800 deaths and not nearly as big as the one you’re claiming in your banner headline. If you look at all your post-1978 data, based on the spreadsheet available on the abstract, I find that male and female storms each averaged 16 deaths.

  12. Harold Brooks
    June 2, 2014

    From above comments:

    “What’s more, looking only at severe hurricanes that hit in 1979 and afterwards (those above $1.65B median damage), 16 male-named hurricane each caused 23 deaths on average whereas 14 female-named hurricanes each caused 29 deaths on average. This is looking at male/female as a simple binary category in the years since the names started alternating. So even in that shorter time window since 1979, severe female-named storms killed more people than did severe male-named storms.”

    This result depends entirely on Sandy. Without that event, even the data the authors give in the comment, the mean death rate in male-named storms is higher than in female-named storms. If they had done the analysis 20 months ago, the conclusion would have been that more people die in the male-named storms, if mean is the statistic of choice.

    The mean is, in my opinion, a poor choice to use as an estimate of central tendency given its sensitivity to outliers. That’s made clear in the very first sentence: “”Estimates suggest that hurricanes kill more than 200 people in
    the United States annually….” The second highest death toll for a year in the NWS data since 1972 has been 51 and it’s only gotten to 200 4 times in the last 74 years. (http://www.nws.noaa.gov/om/hazstats/resources/weather_fatalities.pdf)
    You have to think that the indirect death toll rate is many times that of direct to say that there are 200 deaths in the US in hurricanes in all but a small fraction of the years. If you think that, there’s a big problem, in my mind, to attribute it to gender of names because most of the indirect deaths occur after the hurricane and I have a real hard time imagining people think less of the damage caused by female-named storms than male-named storms when they are actually looking at the damage.

  13. Will Holz
    June 2, 2014

    I want to test this!

    Let’s give a bunch of hurricanes really harmless names and then a bunch of others really scary ones.

    If hurricanes Fluffy and Cuddlebutt end up killing far more people than Hades and Murder-Death-Kill then the data will be even stronger.

    Also funnier. Except for the dead people part.

    Bad fluffy!

  14. Michael Cook
    June 3, 2014

    Is there any actual link found between preparedness and deaths? I know it sounds silly to suggest otherwise but has it actually been shown?

    Is the distribution of hurricanes for name and location been shown to be random? Is there any link between hurricanes with male names occurring in certain areas more than others and vice-versa? Its not enough to say this would happen due to the names being alternated but does need to be shown.

    An odd study thats already been spoken about on my local radio. The media LOVES stories like this and thus itas even more important to get it right.

  15. Laura Hall-Schordje
    June 3, 2014

    I am curious how you could control for differences in meteorological advances since the early 1950s. When all hurricanes had female names, we also did not have Doppler radar, nor even satellite photo capabilities. It would seem that the lack of more effective warnings might also affect the numbers of those killed.

  16. Greg Ledger
    June 3, 2014

    Just wondering why gender-neutral names are not included in naming hurricanes: Chris, Pat, Sydney, etc. Or pet names, or robot names. That would be cool. And provide more interesting data for this story (NG why so boring?)

  17. LK
    June 3, 2014

    There’s precedent for gendered language affecting cognition: http://www.stanford.edu/class/linguist156/Boroditsky_ea_2003.pdf

    Just a quick example, the word ‘key’ is masculine in German but feminine in Spanish. Germans tend to describe keys as, “…heavy, jagged, metal serrated…”. Spanish speakers instead described a key as, “…golden, intricate, little, lovely…”. That’s just one example, lest you think all keys in Spain are tiny and made of precious metals.

    Granted, making a potentially life-altering decision is a lot different than the tasks given in those linguistic experiments. But a lot of people made a bad decision to stay home for Katrina despite warnings. The point of this paper isn’t that someone said, “Oh this storm is a lady, so I’ll be great right here even though its a category 3 coming in with a giant storm surge”, it’s that some small unconscious bias based on the gender of the language may well have combined with a bunch of other factors.

  18. Peter Apps
    June 3, 2014

    My bad on Katrina – late night after a long day !

    “What’s more, looking only at severe hurricanes that hit in 1979 and afterwards (those above $1.65B median damage), 16 male-named hurricane each caused 23 deaths on average whereas 14 female-named hurricanes each caused 29 deaths on average.”

    This hints at a good control – compare deaths per million dollars of property damage to correct for any tendency of “masculine” hurricanes to have hit unpopulated areas, while “feminine” ones went for cities and shopping malls. Also compare whether “masculine” and “feminine” hurricanes do tend to hit cheaper or more expensive areas.

    Is it possible that the gender assignments of names are influenced by the severity of hurricanes – has Katrina been judged as more “masculine” since hurricane Katrina ?

  19. Joe
    June 3, 2014

    Maybe I’m missing something, but if the genders alternative, maybe this just means that even (or odd) numbered storms are particularly bad.

  20. Cerina
    June 3, 2014

    Kristine: So, do you think the secret evolution and global warming evidence-planting cabal masterminded this conspiracy too?

    Love that juicy strawman in your second comment, by the way.

  21. Character
    June 3, 2014

    Id like to see how they determined what old names were feminine and masculine despite being all female names. It sounds littered with subjectivity. I highly doubt they took into consideration the masculinity or femininity of the times which would be a factor in how people responded to them according to their hypothesis. Don’t forget, Leslie and Pearl used to be men’s names. Typical political science nobs anachronistically making judgments. Take a damned history class!

  22. Allan Bird
    June 3, 2014

    It’s painful to see junk science — and make no mistake — this is junk with a dose of scientism added. The lack of rigor is appalling. Even more appalling is to think that this got past a set of peer reviewers. The sampling procedure for the six studies is pure junk. The appropriate sampling pool would be persons living in hurricane corridors. But getting that sample would require hard work. The authors instead used convenience samples that are worthless. Taking the easy out on the sampling sends a clear signal that the authors are interested in the appearance of rigor, but not the application of rigor.

    As for the subjective “feminine-sounding” names classification scheme. Simply not credible.

  23. Harold Brooks
    June 3, 2014

    “Just wondering why gender-neutral names are not included in naming hurricanes: Chris, Pat, Sydney, etc.”

    You mean like Sandy?

    If you surveyed people who lived a significant chunk of their lives prior to the musical Grease and Olivia Newton-John on screen, I’d wager that Sandy is much more likely to considered a male name than it is by undergraduate students, e.g., Sandy Koufax. Just look at the wikipedia entry on the name.

  24. Heather Dewar
    June 3, 2014

    This is completely ridiculous research. Before 1979 all hurricane names were female, and hurricane forecast accuracy was poor, so people were frequently unable to take adequate precautions. The change to alternating gendered names happens to coincide with advances in modeling that greatly improved forecast accuracy and saved lives. Anyone who’s followed hurricanes for a few decades could tell you this.

  25. ferdberple
    June 3, 2014

    the more feminine the storm’s name, the more people it killed.
    ============
    this makes perfect sense. hurricane fatalities are decreasing due to better forecasting and notification. male names have only been used recently. what you have is the effects of a decreasing trend on two samples of different lengths.

    the shorter sample will appear on average to be lower (less fatalities), simply because its average is more recent (thus contains mostly lower data points)

  26. William Osler
    June 3, 2014

    Great article. It is saddening how often statistics are misused. To try and support a conclusion in a discussion with anyone but statisticians, using a P=0.07 is almost fraud. It suggests that they went into the study with desired results, not searching for truth but support.
    “They use statistics like a drunk uses a light pole, for support not illumination.”
    The corruption of science for policy or arguement is dangerous and as such is worthy of distain.

  27. ferdberple
    June 3, 2014

    Also, over time hurricanes will become less dangerous due to adaptation. Hurricane prone structures get destroyed. Hurricane resistant structures survive. Over time the percentage of resistant structures will increase, further reducing the trend in fatalities.

    since male names coincide with more recent trends, this will skew the average of male names lower than female names.

  28. Harlan
    June 3, 2014

    It’s also worth noting regarding their experiments that even if you assume that they were all properly done, that no other experiments were done but not reported, and that no other “experimenter degrees of freedom” were involved, the effect sizes that they report (eta-squared of 0.03 to 0.06) are on the small side. Best case, there’s a small tendency, in the lab, for at least some subjects to take gender into account when they are asked to guess about something they know little about. Even in Ed’s writeup, where he says “they guessed that male storms would be more intense than female ones”, that’s too strong. It should be that “they tended to guess that male storms would be slightly more intense than female ones.”

    I agree with most of the other commenters — it seems unlikely that the effect in real hurricanes will hold up with another 20 years of data, and it seems like the lab effect is likely a very subtle effect with minimal real-world implications.

    [Good point, Harlan, well made. - Ed]

  29. Mike
    June 3, 2014

    Astrophysicist friend pointed out that if you do a Kolmogorov-Smirnov test of the data, you’ll find that the male and female named hurricanes are basically drawn from the same distribution. I really don’t think they tested the null hypothesis at all here.

  30. chris moffatt
    June 3, 2014

    I would propose a hypothesis – namely that the degree of preparedness is directly dependant upon one’s experience of living through a hurricane. Survivors will rarely if ever fail to evacuate, if they can, after surviving their first hurricane experience. If another hurricane comes my way I’m getting out of town no matter what its name is.

  31. Ivar Wind
    June 3, 2014

    And what does the data say about the actual strengths of the hurricanes or (unavoidable on a 24-hour timescale) material damage as a function of the hurricane names? With a sample of this size, I’d be greatly surprised if there was no noticeable difference.

    (Of course it goes without saying, that any part of the data and analysis that includes hurricanes before 1979 when names began alternating, must be excluded. No scientist would have included such data to begin with.)

  32. Shannon
    June 3, 2014

    Why does the study only use US deaths? Seems silly to exclude the Caribbean and Latin America, where there are significantly more deaths from hurricanes.

  33. Eric Panzer
    June 3, 2014

    It’s good that the study’s authors limited the analysis to storms that did greater than $1.65 in damage, but this doesn’t tell the full story. If through random chance, the male-named storms in their data set happened to affect a larger area and/or affect areas with greater population density, then one would naturally expect to see higher death tolls from the male-named storms. Perhaps the authors accounted for this, but I have so far not seen any evidence that they did so.

    Ideally, one would want to “normalize” the number of deaths associated with each storm by absolute number of people affected (i.e., the number of people living in the area that experienced tropical-storm forced winds or greater). This would be a challenging data set to create, but I have a sneaking feeling that if one did so, and changed the metric to “deaths per number of people affected by each storm,” you might actually see the association get even weaker.

    I’d read the methods myself, but don’t particularly feel like paying $10 for the privilege.

  34. Eric
    June 3, 2014

    I’m a little disappointed by the tone of the article and the comments. Practically all the points that have been raised are addressed in the article (and clarified in their later response). Lazo himself seems happy to “star” conclusions at the .1 significance level in his own papers. There seems to be a fair amount of resistance to the idea that gender might matter in ways that are worth studying. My sense is that it has more to do with the conclusions than the evidence.

    • Peter Apps
      June 3, 2014

      Eric wrote; “My sense is that it has more to do with the conclusions than the evidence.”

      Given the way the study was carried out there are only three possible conclusions; 1) “feminine” hurricanes are more deadly, or 2) “masculine” hurricanes are more deadly, or 3) what a hurricane is called makes no difference to how many people it kills. Given that hurricanes (almost certainly) do not know the names that we give them, conclusions 1 and 2 make no sense at all unless people’s response to hurricanes depends on their names, and the death tolls depend on people’s responses to the threat. Both these links, especially the first, are speculative at best, and until they are established with proper evidence the most parsimonious explanation is the pre-1979 bias in names.

  35. Eric
    June 3, 2014

    “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.”

    Just to clarify, this is the whole point of statistical methods. A whole variety of factors may affect hurricane preparedness. You “control” for them to estimate the effect of gender specifically. And you test for statistical significance in order to deal with having a small sample size.

  36. BEdge
    June 3, 2014

    As several people have said, it is doubtful that the relative femininity of the name would have as much effect on residents of hurricane prone areas as a random sample of people who have never experienced a hurricane.

    Also, the after effects of storms are somewhat dependent on the length of time since the previous major storm. Most of the examples the authors mentioned could be considered part of the storm’s effect. But when it has been a long time since a very severe storm, there will be much more tree damage (due to the trees not having experienced a storm while growing). This results in many more people who think they know how to operate a chain saw causing serious injuries to themselves or others. It also results in more cleanup injuries in general. This can hardly be considered a direct indicator of the strength of the storm.

    The time of expected landfall (in relation to tides and time of day) probably has more of an impact on deaths than level of preparedness due to name femininity.

  37. Drew
    June 3, 2014

    He’ll hath no fury like a woman storm

  38. Eric
    June 3, 2014

    Peter Apps wrote: “Both these links, especially the first, are speculative at best”

    I don’t think it’s “speculative at best” to claim that people’s responses to hurricanes affect the death tolls! Now, the deeper question is whether people’s responses depend on the name of the hurricane. This is precisely what the study is trying to determine.

    This argument doesn’t seem that far-fetched to me. Psychologists and marketers study how subtle differences in framing can affect which products we buy. As one commenter amusingly pointed out, people might take “Hurricane Fluffy” a little less seriously, media commentators might downplay it’s ferocity, and on the margins (1 in a thousand perhaps), some extra people might stay at home. Now, I am not saying this study is the final word on the subject, but it’s definitely strong evidence that there might be something to their argument.

    • Peter Apps
      June 3, 2014

      Hi Eric

      Admittedly, it seems very likely that people’s responses affect death tolls, but where are the hard numbers ? At the individual level, someone who takes a threat seriously and evacuates out of the hurricane’s path is less likely to be killed by the hurricane (and more likely to be killed in an RTA) but is there a measurable effect at population level ? (which is what the study was about). Considering the tens or hundreds of thousands of people in a hurricane’s path the death tolls are mercifully small and very variable, and so it would take a huge change in the number of people fleeing for high ground to have an effect on mortality that would be detectable. Your example of 1 in 1000 extra people staying at home when a girl hurricane is approaching, would increase the death toll by 0.1% which is far too small an effect to be detected by this study (indeed it does not even represent one extra death for the means of 23 and 29 deaths cited previously).

      The problem is that for “gender perceptions” to have the impact on death tolls that is claimed, they would have to have an implausibly large impact on people’s behaviour.

      • Harold Brooks
        June 3, 2014

        ” (indeed it does not even represent one extra death for the means of 23 and 29 deaths cited previously).”

        More critically, if the study would have been done 1 Sept 2012, before Sandy, the male storms would have had more deaths than the female storms. Adding the 1 event and including the indirect deaths that in Sandy’s case were very odd (most were hypothermia and CO poisoning a week after the hurricane when power was still out and a winter storm came through-I have a hard time thinking people behaved differently about the power outage that happened a week before because it was associated with a female name rather than a male name) changes the sign of the effect of the gender of the name. Part of that is that they used the mean as their measure, which is not resistant to outliers, but it should be a red flag when adding a case changes the sign of your result.

        I admit I also have a hard time with coders identifying “Sandy” as a more feminine name than “Opal” or “Betsy” or “Beulah.”

  39. Lugo
    June 3, 2014

    “Lies, damned lies, and statistics” in that order.

    Having dated a Fern, I can attest that it is a very feminine name… Saying it is less so than Camile is subjective to the point of absurdity.

    Give me $$$ and a staff and I’ll show statistics that indicate storms with long names cause tinted windows and babies named Bob.

  40. C
    June 3, 2014

    The authors comment here “so even in that shorter time window since 1979, severe female-named storms killed more people than did severe male-named storms”, because one has a mean of 23 and one a mean of 29.

    Why are you using the mean? Also, do you understand confidence intervals? (Post 1979, I get a mean of 15.3 +/- 4.3 for male names, and 17.0 +/- 6.4 for female names… Equivalent! Although no on in their right mind would use Gaussian statistics for this distribution.) Also, as someone commented above, why does a KS test indicate there is no effect here any way you cut it: considering all data, just post-1979 data, only N_deaths >= 10 storms, only N_deaths >= 25 storms…? I literally can’t even do a post-hoc subgroup analysis that finds a significant effect.

    This entire study relies crucially on a tiny number of extreme outliers and bizarrely choosing statistical techniques that are extremely sensitive to outliers.

  41. Juan Manuel Contreras
    June 3, 2014

    There’s also the fact that their main significance tests do not correct for multiple comparisons (i.e., on average, 1 out of 20 tests on junk data will show an effect). With a Bonferroni correction, only 2 of their experiments show statistically-meaningful results.

  42. Michael Kraus
    June 3, 2014

    Many of the complaints here about the first study in the paper (on the actual hurricane data) quibble with the inferential statistics used (e.g., the interaction p-value and the outlier hurricanes). The data were in fact population data (all Hurricanes in the US or all Hurricanes in the US since 1979). Thus, if Female named hurricanes killed more people than male ones in an absolute sense, then we don’t need statistics to conclude this–all we need is our eyes to tell us that female hurricanes killed more people. That’s it. Outliers and p-values and bonferroni correactions deal with inferences made about a population based on a sample. Since we have population data already, we need not worry about these statistical anomalies in the data.

  43. Rudge
    June 4, 2014

    Is very interesting and important the aspects of perception for communication of hurricane risk.

    It will be interesting to know how the validated the online data coletect?

    And also why they used this web plataform for the survey?

    http://blogs.ua.es/percepcionderiesgo/autor/

    Thank you

  44. Namrata Panigrahi
    June 4, 2014

    In our part of the world, cyclones and hurricanes are not given any names, so we do not have any such bias. If you give both masculine and feminine names to hurricanes and add up the death tolls from both, one is bound to surpass the other, and it is absurd to seek an explanation for the same.

  45. Amy
    June 4, 2014

    Katrina didn’t kill the most people. The unnamed hurricane of 1900 killed an estimate 12,000 people. Various Pacific typhoons have killed hundreds of thousands. The very suppositions of this “study” are incorrect. First off, no hurricanes had a male name before 1979 and then only every other year, which , as Ed points out, makes the dataset very small. But from 1950-1979 ALL hurricanes had female names. And before that, none of them had names at all. And some of the worst storms occurred in the era before names (Galveston 1900). Most of the worst storms ever occurred in other parts of the world, like Bangladesh in 1970, in which an estimated 400,000 dies. So this study is bullshit from the get-go. It is about cherry-picking the data to show exactly what you want it to show, and then following up with a bogus study of how the names made people “feel.” Utter BS.

    You cannot rule out the possibility that more deaths occurred in an era before good forecasting methods, when meteorologists relied on the weather reports of ships at sea that happened to be close to a storm. Predictions about landfall and strength have gotten so much better, as have building codes in storm areas, which means, that in most of the modern era (Katrin excepted) hurricanes have resulted in fewer deaths, relatively speaking. So male-named storms, occurring only in the modern era, would result in fewer deaths. So this study is bogus all around. And the idea that people take female-named storms less seriously is a just-so story and they haven’t proven that to be the case at all.

  46. Doug Smith
    June 4, 2014

    I haven’t read the details of the study, but from what I’m reading here there are assumptions that aren’t stated that I’m very uncomfortable with. The two assumptions that bother me most about the determination of the ‘femininity” and “masculinity.” are that the characteristic is static with time and place. I’ll offer Miley as an example. I suspect it’s gone from somewhat masculine as a diminutive of Miles to a cutesy girls name when Hannah Montana equated with Miley Cyprus to what after the release of Wrecking Ball. Another example would be Sandy. How I ranked the question would range from somewhat masculine if I had just been talking to an older man about baseball pitchers to somewhat feminine if I’d just been talking to a friends brother-in-law about his ex wife. That means you’d get a different answer on different days but even on different hours on a particular day.
    I’m not even sure how many people would of thought about whether a name was more or less “feminine” in 1950. I also suspect that how a name is viewed also depends on where you live and your social-economic class. I’d at least want to confirm that a given name had a similar rank in the parishes around New Orleans as it did in Boston around Cambridge. I see designing a study to try and get around these two assumptions as being resource intensive. It would have to involving a large number of people answering a significant number of indirect questions about their views on a name. I’d work along the lines of What would have you thought if your son had married a Cindy opposed to Name of Person he did marry.
    It also appears to me that the diminutive forms of female names are more common in the list than those for male names. This year we have Dolly, Sally, Vicky as diminutive female names and just Teddy as a diminutive male names. Take Candy for incidence. I’m not sure that Candace would be reliably ranked less feminine than Candy, but a Candace would be taken more serious than Candy.
    Finally I wonder how consistent indirect deaths are reported from one storm to the next. I mean if a person gets electrocuted by a down power line it makes sense to include them in the indirect death total. On the other hand If a person is killed in an auto accident where there is a light not working is that really an indirect death? There are fatal accidents all the time from drivers doing stupid stuff all the time. The challenge is to say which ones to include in the storm count and to be consistent about it. While looking at variance over time will eliminate some of this, it won’t eliminate it because of location.

  47. Kyle Altis
    June 4, 2014

    To repeat what has been said repeatedly before: this is not about the actual severity of the hurricane. This is about whether the gender value of a hurricane’s name subconsciously “nudges” people into perceiving a hurricane as a greater or lesser threat, which in turn would affect their personal level of hurricane preparedness and possibly their willingness to evacuate.

    (Many forget that evacuations themselves often kill people, and that hurricanes don’t always strike where they are forecast. As always, it is a comparison of relative risk.)

    To those who claim that there would be different reactions to hurricanes in those with prior hurricane experience, probably — but bear in mind that experience can cut both ways (eg. “only” a category 1 *shrug* no big deal).

    Bear in mind also that much of the U.S. Southeast has not been struck by a major hurricane in decades. Most of the population of Florida has never experienced a major hurricane at all.

    The level of significance found here is not high. However, the vast majority of studies have levels of significance which are no higher — including, as noted earlier, several studies by the rebutter in this article.

    I strongly suspect the reason so many people are up in arms about this particular finding is because we really don’t want to conceive of ourselves reacting in subconscious ways at all, let alone to gender. I have seen similar reactions any time any research hit the mainstream media which involved subconscious “nudges”. This is a marketing term — and indeed marketers make very heavy use of this, which tend to be even more effective because so many people refuse to believe they can be manipulated this way and are convinced all their choices are fully conscious and rational. (So which brand beer do you drink? Do you think you could identify it in an anonymous lineup? Most diehard loyal beer drinkers can’t — and it has nothing to do with their level of intelligence.)

    I notice the authors are behavioural psychologists. Psychology abounds in studies measuring word non-tangibles (such as emotive value) in order to establish reliable scales with repeatable results. Gender is a common non-tangible in these scales.

    The authors themselves point out that this finding only opens a door to further research. Now that initial research has demonstrated that there *may* be a slight nudge based on gender names, subsequent research can select precisely based on hurricane category (wind speed), or storm surge, or rainfall, or any other desired criteria. The samples should be U.S.-specific if the findings are to have U.S.-specific applications.

    Incidentally, Shannon, if you exclude Haiti, the Caribbean does not have significantly more deaths from hurricanes than the U.S. — quite the opposite, in fact. Hurricanes which hit mainland Latin America have the complicating factor of heavy rainfall + landslides, which requires the kinds of mountainous slopes the Atlantic hurricane-prone coast of the U.S. does not have at all.

    Finally, if further research does demonstrate a link, however slight, that link is measurable in lives. Does it really matter whether the extra lives saved could be counted in single digits, or double digits, or hundreds of digits? This would be a case where a completely no-cost-to-the-taxpayers change could save those lives.

    There are multiple options for no-pay list name changes. However, the easiest is probably the same system already used by the western Pacific, which has a non-personal name system for naming its hurricanes.

    Seriously, peoples — how often do we discover a case where there is actual evidence that a small change — at no taxpayer cost — won’t harm lives and may well save lives? Why are we arguing so hard against anything of the sort?

    • Harold Brooks
      June 4, 2014

      The reason I’m up in arms is that they don’t show any real evidence that there’s a difference in the deaths by gender of the name. The very weak evidence they give depends on 1 storm and, in particular, the indirect deaths associated with a winter storm that happened a week later. If they had done the work before the last event in the database, they’d be explaining why male storms kill more people.
      The strongest thing they should have said was that their participants perceived the names differently, but that doesn’t appear to have any impact on death tolls, perhaps because real decision makers have a lot of forecast information in addition to the name.

    • Harold Brooks
      June 4, 2014

      The reason I’m up in arms is that they don’t show any real evidence that there’s a difference in the deaths by gender of the name. The very weak evidence they give depends on 1 storm and, in particular, the indirect deaths associated with a winter storm that happened a week later. If they had done the work before the last event in the database, they’d be explaining why male storms kill more people.
      The strongest thing they should have said was that their participants perceived the names differently, but that doesn’t appear to have any impact on death tolls, perhaps because real decision makers have a lot of forecast information in addition to the name.

  48. david
    June 4, 2014

    As a scientist, I’m troubled by the tone of this discussion. Let’s take a step back and try to understand what exactly we’re discussing. First, a group of scientists conducted a study, went through the peer review process, and published a paper. Second, a journalist (presumably) read their paper and published a critical review of it. Third, another group of people read that summary and are very critical of their findings in an internet comment section.

    The troubling part is that the journalist either a) Didn’t carefully read the paper, or b) carefully read it but didn’t understand it.

    So, unless you read the paper yourself, you’re reading a summary of an article that the authors of the article indicate is not a good summary of what they actually did and what their results suggest- including policy implications, which the authors do not even speculate about.

    Read the paper, folks! 99% of these criticisms are either a) addressed with data or b) listed as a limitation of the study design. The authors qualify their results based on these limitations. That how science works. Like it or not, people who want to understand this phenomenon are much more likely to read the comments here than the paper itself.

    Your criticisms might be legitimate, but you’re being unfair to the authors if you do not read what they wrote first.

    [To address the bit of this that's relevant to me: yes, of course, I read the paper and, yes, I understood it. I stand by my piece. - Ed]

    • Eric Panzer
      June 4, 2014

      OK, I’ll bite. After reading this comment I went ahead and obtained the full text of the paper. In your comment, David, you complained that people had not read the paper and you wrote that “the authors do not even speculate about [policy implications].”

      The discussion section of the paper goes on at length (two paragraphs) about the policy implications of the paper’s findings. Here is the key paragraph from the original paper:

      “For policymakers, these findings suggest the value of considering a new system for hurricane naming to reduce the influence of biases on hurricane risk assessments and to motivate optimal preparedness. For media practitioners, the pervasive media practice of giving gendered descriptions of hurricanes should prompt a reconsideration of the use of ‘he’ or ‘she’ when communicating about hurricanes. Finally, making members of the general public aware of the impact of gender biases on subjective risk perceptions may improve preparedness in the face of the next Hurricane Fay or Laura.”

      I’m sorry to be blunt, but clearly you are among those who have not actually read the full paper. After looking at the Materials and Methods section, I feel that my criticisms of the study still apply, though I give credit to the authors for disclosing what data were not available. The original paper states:

      “However, maximum wind speed data were not available until 1979; therefore, this variable was excluded from the data analyses…Data on many other factors potentially responsible for hurricane fatalities (e.g., width of hurricane, route of hurricane) were unavailable…Total deaths had the strongest association with normalized damage (r = 0.555, P < 0.001), among other variables such as minimum pressure (r = -0.394, P < 0.001) and hurricane category. Perhaps this is because it reflects other unobserved factors potentially responsible for hurricane fatalities, such as population density, route, and duration of hurricane, indicating that costlier hurricanes are much deadlier."

      Without considering the route and size of a hurricane (and therefore the absolutely number of people potentially affected) the value of the paper's analysis is drastically diminished. Damage values, especially ones normalized in the manner described by the paper's authors, likely correlate well to the number of people affected by a particular storm, but this is still a suboptimal metric. Add to this the fact that the analysis (understandably) didn't consider warning times or accuracy of forecasts ahead of landfall, and you're left with the conclusion that the results remain very much open to dispute.

      • david
        June 4, 2014

        Eric- you’re right. I have not read this paper. However, I am not denouncing it as “junk science” before reading it.

        My problem is with people saying things like “obviously this study is nonsense because the authors didn’t consider ____” without reading the paper.

        My point is not about this paper in particular, but about how science is reported and consumed by non-scientists. Science journalism seems to want to drum up controversy- and often that’s not necessary. Sure, this paper has limitations, and sure not all scientists agree with the results, but the implicit assumption in most of these comments is that this paper is rubbish because the person the journalist interviewed found fault with the paper.

        And no, the authors did not speculate on policy changes based on their data. Suggesting that it is worth considering policy changes is not the same as speculating on policy changes- and certainly not the way Lazo and the author of this article implied.

        • Eric Panzer
          June 4, 2014

          With regard to policy speculation, I’m going to once again quote a key phrase from the original paper itself:

          “For policymakers, these findings suggest the value of considering a new system for hurricane naming to reduce the influence of biases on hurricane risk assessments and to motivate optimal preparedness.”

          In the case of the above statement, implying a substantive difference between “suggesting that it is worth considering policy changes” and “speculating on policy changes” seems to me to be a disingenuous semantic argument. The authors are both putting forward a specific potential policy change (a new naming system for hurricanes) AND explicitly stating that the policy change could serve ” to reduce the influence of biases on hurricane risk assessments and to motivate optimal preparedness.” How is this in any way not “speculating on policy changes?”

          • David
            June 4, 2014

            Ok. you all convinced me that if I’m going to berate folks for not reading a paper, I should read it myself.

            And BTW, Eric & Harold- you weren’t the target of my original comment.

            I do think there’s a difference between saying “… suggest value of considering a new system for hurricane naming” and suggesting specific policy changes. Their statement reads as an open-ended question. My guess is that this was carefully worded to avoid suggesting specific changes. This article, and many comments, imply that the authors are saying we should change the way we name hurricanes based on the results of their study- that’s not the way I read it.

            Harold- I think your criticism of hurricane Sandy being a female storm is misplaced with regard to its impact on the analysis. In my reading of the paper, they treated masculine vs. feminine as a continuous variable, and Sandy would likely be somewhere in the middle. If anything, giving a “female” storm a relatively masculine name “Sandy” would have decreased their statistical power.

            [David, 2.58pm, 4 June: "So, unless you read the paper yourself, you’re reading a summary of an article... Read the paper, folks! 99% of these criticisms are either a) addressed with data or b) listed as a limitation of the study design... Your criticisms might be legitimate, but you’re being unfair to the authors if you do not read what they wrote first.

            David, 6.21pm, 4 June: "Eric- you’re right. I have not read this paper."

            This is the best thing I've read today. - Ed]

          • Harold Brooks
            June 4, 2014

            The inclusion of Sandy as a female name is not really my point. It was rated a 9 on a 1-11 scale with 11 as purely feminine, so Sandy is a lot more feminine than Beulah or Opal or Wilma or Agnes. Of the big damage female storms, it’s right in the middle of the female.

            There are two issues with Sandy. There are >50 indirect deaths from Sandy that are due to power outages that led to hypothermia and CO poisoning in a winter storm more than a week after Sandy. If those deaths aren’t included (and I have a really hard time thinking people treated a power outage for a week differently because of a feminine name), the mean deaths from male and female storms are the same, so their result depends on deaths that happened more than a week after landfall.

            More importantly, if the analysis was done the week before Sandy, the mean deaths from male storms would have been greater than from female storms. In fact, the difference would have been larger than the female-male difference with Sandy and its indirect deaths included. So, if they had started this work 2 years ago, they would have been trying to explain why male storms kill more than female storms. That’s just way too much sensitivity to a single event.

    • Harold Brooks
      June 4, 2014

      I certainly read the paper, downloaded their dataset and analyzed it, as well as looking at the official NWS database, and I’m convinced that no one who knows the hurricane death database was a reviewer on the paper. The first sentence (estimates of more than 200 deaths, referenced to a popular, not a scientific book) wouldn’t have made it through review. Only one year since 1972 has had as many as 200 deaths. The 30-year mean is <50. I don't know, nor do I care if the social science experiments are done well. The death record does not support the idea that more people die in female-named storms. Sandy's the only thing in their dataset that marginally makes it not more people dying in male-named storms. The second sentence of the abstract is not supported by the data. Without that support, there's no basis to do the experiments.

  49. LizR
    June 4, 2014

    Because…

    WHEN the Himalayan peasant meets the he-bear in his pride,
    He shouts to scare the monster, who will often turn aside.
    But the she-bear thus accosted rends the peasant tooth and nail.
    For the female of the species is more deadly than the male.

    When Nag the basking cobra hears the careless foot of man,
    He will sometimes wriggle sideways and avoid it if he can.
    But his mate makes no such motion where she camps beside the trail.
    For the female of the species is more deadly than the male.

    When the early Jesuit fathers preached to Hurons and Choctaws,
    They prayed to be delivered from the vengeance of the squaws.
    ‘Twas the women, not the warriors, turned those stark enthusiasts pale.
    For the female of the species is more deadly than the male.

    Man’s timid heart is bursting with the things he must not say,
    For the Woman that God gave him isn’t his to give away;
    But when hunter meets with husbands, each confirms the other’s tale—
    The female of the species is more deadly than the male.

    Man, a bear in most relations—worm and savage otherwise,—
    Man propounds negotiations, Man accepts the compromise.
    Very rarely will he squarely push the logic of a fact
    To its ultimate conclusion in unmitigated act.

    Fear, or foolishness, impels him, ere he lay the wicked low,
    To concede some form of trial even to his fiercest foe.
    Mirth obscene diverts his anger—Doubt and Pity oft perplex
    Him in dealing with an issue—to the scandal of The Sex!

    But the Woman that God gave him, every fibre of her frame
    Proves her launched for one sole issue, armed and engined for the same;
    And to serve that single issue, lest the generations fail,
    The female of the species must be deadlier than the male.

    She who faces Death by torture for each life beneath her breast
    May not deal in doubt or pity—must not swerve for fact or jest.
    These be purely male diversions—not in these her honour dwells—
    She the Other Law we live by, is that Law and nothing else.

    She can bring no more to living than the powers that make her great
    As the Mother of the Infant and the Mistress of the Mate.
    And when Babe and Man are lacking and she strides unclaimed to claim
    Her right as femme (and baron), her equipment is the same.

    She is wedded to convictions—in default of grosser ties;
    Her contentions are her children, Heaven help him who denies!—
    He will meet no suave discussion, but the instant, white-hot, wild,
    Wakened female of the species warring as for spouse and child.

    Unprovoked and awful charges—even so the she-bear fights,
    Speech that drips, corrodes, and poisons—even so the cobra bites,
    Scientific vivisection of one nerve till it is raw
    And the victim writhes in anguish—like the Jesuit with the squaw!

    So it comes that Man, the coward, when he gathers to confer
    With his fellow-braves in council, dare not leave a place for her
    Where, at war with Life and Conscience, he uplifts his erring hands
    To some God of Abstract Justice—which no woman understands.

    And Man knows it! Knows, moreover, that the Woman that God gave him
    Must command but may not govern—shall enthral but not enslave him.
    And She knows, because She warns him, and Her instincts never fail,
    That the Female of Her Species is more deadly than the Male.

    (…or maybe because no one worked out that Sandy is a gender neutral name.)

  50. Dewi Morgan
    June 4, 2014

    Where I’d be interested to see this taken in the next study is to scrap the whole masculine/feminine thing for a start. It confuses the issue too much.

    Instead, have people rate how threatening the names sound, as a storm’s name (and possibly, as a person’s name: this may be significantly different for some of the populations listed below).

    Separate those people into:
    – Americans in the hurricane areas who were alive when the named hurricane hit.
    – Americans outside the hurricane areas who were alive when they hit.
    – Americans who were not alive when the hurricanes hit.
    – Same for non-Americans.

    If there are no differences in the above populations, then this can be ignored in future studies, otherwise, the scale of the effect can at least be compensated for somewhat.

    Using these, you can then find whether threatening storm names cause more deaths-per-dollar-damage or deaths-per-landfall or deaths-per-storm-size than non-threatening ones, which will say whether people take it into account when preparing for storms, and also tell whether there’s a tendency to allocate more threatening names to more potentially dangerous storms or vice versa.

    Once you’ve established whether threatening storm names are a factor, *then* you can look into whether the genderedness of those names plays a role. But it feels like, first you have to establish that there’s a relationship between percieved risk, and real risks taken, before you claim the less-direct linkage of gender -> percieved risk -> risks taken.

  51. Mike
    June 4, 2014

    “In other words, how long ago the storm occurred did not predict its death toll.”

    Pre-1979 hurricanes killed 27/event, post-1979 16/event. If your model is giving you an answer that contradicts everything we know about hurricane death tolls and simple math, I would suggest your model doesn’t work.

  52. Harold Brooks
    June 4, 2014

    One final observation. If you take the storms with at least $1.65B in damage since 1979 (Sandy included) and randomly pick one from the male group and one from the female group, there’s a 52% chance the one from the male group will have more fatalities, a 45% chance the one from the female group will have more, and a 3% the death tolls will be equal.

  53. J
    June 5, 2014

    “Pre-1979 hurricanes killed 27/event, post-1979 16/event. If your model is giving you an answer that contradicts everything we know about hurricane death tolls and simple math, I would suggest your model doesn’t work.”

    It’s not quite as simple as you claim. If you correlate years elapsed versus deaths, you get r=0.032, which is statistically insignificant (p=0.762). The simple comparison of 27 vs 16 ignores the large spread of each of these distributions.

    Of course, there is more a priori plausibility for hurricanes getting less dangerous over time (as compared to, say, a far-fetched theory about the genders of hurricane names), so I wouldn’t rule it out.

  54. Sharky
    June 5, 2014

    re: “It may make sense to move away from human names, but other labels could also create problems if they are associated with perceptions of mildness or gentleness.”

    Simple remedy–avoid human names, and instead adopt a suitably terrifying nomenclature such as: Hurricane Gonna Kill You, Hurricane Puppy-Mangler, and Hurricane Oh, Sweet Lord, NO!

  55. Kyle Altis
    June 6, 2014

    Ah, the gender scale actually in the article has finally been discovered. One way to improve future research in this area is to more rigorously test that scale — or possibly develop a better one — by using responses specifically from population samples in hurricane-prone zones, preferably by analysing responses by region. (Appropriately detailed analysis based entirely on region-internal data is still heavily limited by sample size.) Limiting the sample to post-satellite era hurricanes should limit the history effect, as well as the possible effects of forecasting within their current methodology (although much of that last was covered by their internal analysis of the earlier subset of hurricanes).

    “The very weak evidence they give depends on 1 storm and, in particular, the indirect deaths associated with a winter storm that happened a week later. If they had done the work before the last event in the database, they’d be explaining why male storms kill more people.”

    Or, alternately, they could have chosen to include Katrina and Audrey in their data set … which would have given them a much stronger correlation, but then people would be complaining that they chose *those* storm extremes; and people less familiar with statistical methods would suspect that the analysis results then depended entirely on Katrina.

    Several of the complaints are that Sandy’s hypothermia deaths are included. However, unlike the traditional U.S. hurricane-prone regions, which extend from S. Carolina to Texas, more northerly tropical storms and hurricanes commonly include hypothermia-inducing weather on their backside, up to and including snow. Sandy was an extreme example, but other hurricanes have also produced extremely cold rain and (less commonly) snow. As a result, hypothermia deaths are probable consequences of mid-latitude tropical storms and hurricanes, as they are not for hurricanes in more traditional areas.

    This brings up an internal irony, in that there is quite an extensive body of modelling research which suggests a probable increase in storm severity specifically in mid-latitude storms which corresponds with global average temperature. Any determined omission of hypothermia linked with tropical storms will thus also tend to artificially reduce the mortality of mid-latitude storms specifically.

    Incidentally, I am not saying that the researchers found a particularly strong result. If this were a stand-alone study, the results could well be abandoned.

    However, the point is that this should not be a stand-alone study. The results are strong enough to merit further research into the phenomenon using other hurricane criteria. The sample size is constantly growing, such that in subsequent research, it should be possible to use only hurricanes after the satellite era, which should ease the effects of that particular confounding variable.

  56. david
    June 6, 2014

    Ed- as one your merely makes a living by criticizing other people’s science, rather than producing anything himself- I do not expect you to understand my point about reading the article. Your post hoc insertion of yourself into the peer review process is my problem.

    I’m glad that you found the time to cut and paste my argument in such a way as to amuse yourself. It seems that this is what you do for a profession.

  57. david
    June 6, 2014

    To clarify my comments above- This story is one of the worst covered science stories in recent memory. To Ed’s credit- this particular story is one of the least bad of them all. The tone of 90% of these stories is negative- if not condescending to the authors. Everyone has a theory about why the authors are wrong. My only major problems with this version of the story is that it is presented as CONTROVERSY: SCIENTISTS DISAGREE ABOUT IMPLICATIONS OF FINDINGS.

    The authors of the original article politely told Ed that if he had more carefully read their paper, he could have included their counterarguments in his story.

    My problem with this type of article is that this is the watered-down version of science that the public consumes. When policy decisions are made, it is these types of discussions that are influencing ultimate decisions.

    Harold Brooks- you’re a scientist- you know that PNAS regularly prints criticisms of from other scientists. Write an article and submit it- give the original authors a chance to explain why the included Sandy and indirect deaths. That’s how science efficiently progresses- not through internet discussions.

  58. Martin Robbins
    June 6, 2014

    “Ed- as one your merely makes a living by criticizing other people’s science, rather than producing anything himself- I do not expect you to understand my point about reading the article. Your post hoc insertion of yourself into the peer review process is my problem.”

    …but why is he wrong? Can you answer using sensible phrases that aren’t insulting or rooted in your intellectual insecurity? And why should criticism stop at peer review, exactly? I bet you bloody love Ivan Oransky.

    Your post hoc insertion of yourself into Ed’s comments is undoing all of the good work accomplished by the caffeine in my bloodstream this morning.

  59. Doctorknow 007
    June 6, 2014

    Interesting discussion; Thanks Ed!!

  60. JohnR
    June 6, 2014

    Once again, Shakespeare said it best: Much Ado about Nothing. Also, David, the old adage comes to mind here – when you find yourself in a hole, stop digging.
    Simply on the basis of the response and the discussion (and not even considering the amusing original paper), this was a good post, Ed!

  61. JohnR
    June 6, 2014

    Also, too, @LizR: deadlier, yes (probably), but also a whole lot cuter and cuddlier (on average, plus or minus a standard deviation, and taking into account non-parametric statistical methods).

  62. JamesM
    June 7, 2014

    Actually, Martin, if you take a look at the researchers’ response to this piece you can see a point by point discussion on why Ed is wrong. I’m not positing that the original article is a solid piece of scholarship nor am I suggesting that the mere fact that it went through the peer review process absolves it of scrutiny. However, as David already pointed out, the issues that Ed raises here were already addressed in the original paper–particularly the silly point about the shift in naming practices, which is the criticism I keep seeing reposted and it’s driving me up the wall!

    I’m joining the chorus of people here troubled by the tone of this conversation, and I say that even after being convinced that the original article is a bad case of junk science. Harold Brooks here seems to know the field and has identified a litany of issues that undermine the credibility of the researchers’ conclusions. But here’s the rub: I’m choosing to take Harold’s word for it when he says that he downloaded the original data set and conducted his own analysis, and I’m also trusting in the fact that he did so in a competent manner. His methodology has not been published for peer and public review–his reasoning is sound and he makes a convincing argument (he convinced me!) but the fact remains that there’s something of a leap of faith taking place here. The fact that Ed would, in an update to his original piece, refer to a faceless commenter to take down a peer reviewed journal article is a sad indictment of the state of science reporting today.

  63. david
    June 7, 2014

    What did Ed do wrong? He inserted himself into this story in the role of editor, when he- having as far as I can tell, has never participated in the peer review process. He is woefully unqualified for this role. He did not consider the authors arguments- likely because he did not understand them. The authors of the article are being polite when they suggest that he could have more carefully considered his work.

    Then, there is the issue of Harold. I believe him when he claims to be an expert in hurricanes- but he certainly is not in statistics. He suggests that the authors should have used s median instead of a mean. This is s nonsensical suggestion because the authors’ primary analyses where in a regression framework- i.e., not mean comparisons. His analyses, which were much less sophisticated, were mean comparisons. A real editor would have set him straight on this.

    Why should Ed endorse this comment?

    I’m not alone in this-my colleagues are similarly troubled by the way this story was reported.

    Look at the response from folks who jumped to Ed’s defense. He certainly is an accomplished journalist- but that says more about his entertaining style of writing (e.g., turning a story like this into a controversy, as if, the same exact story couldn’t be written about any paper. Scientists always disagree- that’s a fundamental part of science), than his critical thinking skills.

    The fact that so many people seem to think this was a useful discussion illustrates my point. I know I’m being a jerk here, but this was not high level discourse.

  64. cleverity
    June 10, 2014

    There may be solace in the fact that they sourced non-coastal persons to produce a gender continuum for hurricane names: That those likely less affected previously by those names as hurricanes [and hurricanes as names] also being those more likely to make mistake [naively] were they ever to be so affected.

  65. david
    June 11, 2014

    It seems I’ve unwittingly stumbled on the classic ivory tower vs. science as infotainment debate. To be perfectly honest (and feel free to mock me again for something I admit), I hadn’t given much thought to the way science is covered in the media before this story.

    Is it common practice for science infotainers to mock people who comment on their articles on twitter? If so, mark that down as another reason your model is flawed. 140 characters, surrounded by sycophantic “followers” is the appropriate response? Makes sense.

    After reading your “rebuttal,” I’m stuck in the same situation as the authors of this article- and probably most of the scientists you cover as well. Did he read my comments? Did he read and not understand them? Or did he deliberately misrepresent my point for entertainment value?

    My point is that- the way this article is written- you are attempting to mediate a dispute between two scientists. And you are functioning in the role of the action editor (I’m assuming you know how that works in science. If I’m wrong, I’d be happy to explain how the peer review process works).

    Based on your mocking comments on twitter- this is supposedly the role of a science journalist. I’m troubled by that. As an example- your point about a p-value of .07 “not really counting” isn’t entirely accurate. p-values are dependent on sample size, whereas effect sizes are not. Thus, the exact same effect could be less than .05 with a large sample and greater than .05 with a larger sample. That is a mathematical fact. There is debate about whether .07 should count. Problems with null hypothesis significance testing aside- statisticians disagree. Personally, I think the effect size is more important. One of the comments you endorsed suggested the effect sizes were too small to be considered important. However, this effect size in this study is a measure of loss of human life. A small loss of human life is more important than a large correlation.

    My point is not that science writers shouldn’t criticize science. Or that they shouldn’t interview other scientists who are critical of the work. (however, I’d strongly argue that the scientist you interviewed and the one who commented on this blog, should write a reply and subject the reply to peer review as well).

    My point is that you shouldn’t make the decision of what is “true” and what is not true when two scientist disagree. And, the tone of your article is to do just that.

    You can take this criticism one of two ways:

    1) Some of the scientists you cover are elitist jerks who think only other scientists are smart enough to criticize their work (i.e., the ivory tower idea)

    or

    2) Some scientists think the way this story was covered- with science writers inserting themselves into the story to mediate disagreements among scientists. And, that this model, while entertaining, is not the best way to educate the people interested in science.

    My comments were never about this paper. This paper is a cute study, but it is unlikely to have a long term impact on any field it touches. My comments were about the way this story was covered in the media.

  66. Mav Rick
    June 19, 2014

    I wonder if the results of this study are true or lean a bit more on the side of “correlation”. My psych professor gave an example of this with a study done that discovered that as popsicle sales went up more drownings occurred. At first this seems unbelievable, but it must be true since it was done by an academic study. Come to find out, both popsicle sales and drownings increased because of the time of year (summer) and were not directly related as a cause and effect. In the same way, are hurricane names directly related to the number of deaths? Are people biased to gender in names?
    Jeff Lazo brings up a good point: it could be that more people died in hurricanes because more people died on average before they started using male names in 1979. Ed Yong’s analysis points out other flaws in the study, such as deaths that occurred during debris clean up that were also counted.
    While people do have biases regarding the attributes associated with a certain gender-related name (Susie vs. Bill), I think it is debatable on whether or not people respond differently to hurricanes based on the gender (or degree of that gender). In our changing society, many of the lines are blurred when it comes to differences in gender. Others are advocates of the empowerment of women (which, remember, in the past was seen as the “weaker” gender). This leaves me with a question: should hurricane names be changed to non-human names to avoid any kind of bias? But that leaves us with trying to decide which non-human names will avoid biases that will cause people to underrate the nature of the storm (Sparrow vs. Echidna anyone?). The issue here is not whether people have a bias towards the characteristics of names, but that they do and how to give hurricanes names that won’t lull people into complacency, nor freak them out.

Add Your Comments

All fields required.

Related Posts