I attended the National Weather Association’s conference in Birmingham, Alabama, for two days. Toward the end of the second day, the main focus of the talks was the terrible death count from tornadoes for 2011, and most of the speakers were coming at the problem from the perspective of social science such as psychology. It is understandable that people would want to better understand what happened in an anomalous, outlier year such as 2011. It is understandable that people would want to find out if the catastrophe was a result of factors that can be easily changed, and that they would even be biased toward that hypothesis. (One presentation even mentioned the “optimism bias”—a concept that seems a bit strange to me as a natural pessimist, but I can readily see that it would exist in most people, and I would say that this is a perfect example of it.) My intention here is not to call anyone out. However, I think that a lot of the research is, frankly, barking up the wrong tree. There are also some very serious flaws with some of the studies themselves.
The bulk of the research involved surveys of people from the areas that were impacted by tornadoes in 2011. The surveys contained questions about NOAA watches and warnings (whether people received them, how they received them, whether they were understood) and people’s responses to these messages.
Here are some points I took away from the social science presentations:
- An overwhelming majority of people in impacted areas did receive warnings.
- A very small minority of them immediately went to shelter after receiving a warning from the first source.
- A rather large plurality sought out additional information from TV, the Internet, or personal confirmation to determine if the tornado actually existed and would potentially affect them. This was more likely in people with higher levels of education and in people who knew more about the weather. (I would like to note here that this is exactly what I did when the east-central MS EF-5 tornado of April 27 was heading my way. I did not immediately barricade myself under the stairs when I heard the warning. I looked at radar to identify a probable debris ball signature and plotted its projected path to go right over my house. I then grabbed my cat and got out of town. The tornado lifted, but if it had stayed on the ground, I could have been killed as a result of following the canned advice rather than reasoning out the best course of action for myself!)
- A minority of people chose to completely ignore a warning.
- When asked how likely they, personally, thought it was that their area (of what radius? I don’t recall if it was stated) would be significantly impacted as a result of bad weather mentioned in a warning, the most common answer was less than a 25% chance. The social scientists said that they wanted people to guess a nearly 100% chance, but in fact, the scientifically and statistically correct answer was less than 5%. Interestingly, this arguably refutes the “optimism bias” argument in that people did give a more pessimistic judgment of their risk level than was really the case, just not pessimistic enough to suit the social scientists.
The social scientists seemed to be dismayed by the fact that people were less likely to immediately dive for cover the more educated and weather-savvy that they were. Needless to say, this is an odd message to deliver to a room full of meteorologists (many of whom actively seek out bad weather in their vehicles). What is the point here? “Ignorance is strength,” to quote from Orwell? Let alone that people can’t exactly become less weather-savvy, less educated, or more paranoid about the personal impact from a storm if they already know better. This is an example of trying to close the barn door after the animals have escaped. These things are what people do, and with the proliferation of web phones with more and more features that allow people to have access to information virtually anywhere, these behaviors are only going to become more common. This means that they are the behaviors that must be worked with and planned for. Trying to force people into a state of unnecessary and statistically unwarranted fear is not going to work. Nor is it a good idea to try to bully people into not seeking out information and using cognitive reasoning. I’m no social scientist, but I can tell you that if this is attempted, the most likely reaction is a rebellious contempt for “the government” for “trying to make us not question, not think for ourselves, and do as we’re told.” I would be just about willing to guarantee it. It could backfire badly. People ultimately have to be responsible for their own decisions.
Furthermore, there was absolutely no evidence given that people who sought out more information first were more likely to be injured or die in an event, and obviously the survey methodology required interviewing people who did not die. Knowledge about what the people who died did must come from people who were with them and survived. However, I never even saw that there was a distinction made between the group of people who were in the path of the tornado and were uninjured or had only minor injuries, and those who were severely hurt or killed. It would have been useful to find out if the people who were severely harmed did anything differently from those who were more or less okay. Given that at least one of the surveys was conducted via e-mail shortly after the event in question (the Tuscaloosa tornado), I would expect that there would be very few people who were severely injured who even participated in it, because they would have been in the hospital. In effect, the social scientists gathered statistics about a control group and presented it as though it represented the experimental group. In this situation, the statistics about behavior patterns following a warning mean nothing in themselves. There is nothing (survival/non-survival, minor/major injury) to correlate them to. Implying that these behaviors caused the death toll to explode is unsupported speculation. The one survey I saw that definitely interviewed people who had lost loved ones or who were severely injured was conducted in Smithville, MS, and these authors did not make any wild inferences about how seeking out additional information had led to the deaths. There is simply no data support for it. The only situation where it might make a difference is when the lead time is basically zero and every second counts, which was not the case in the April 27 tornadoes or the Joplin tornado. (I had a lead time of about 25 minutes, which was enough for me to get my cat and laptop and go 18 miles away.)
There was one data omission that is, in my perspective, more important than any behavioral survey. One table that I did not see in any of the social science presentation was this one:
|F Scale||Killer Tor||Fatalities|
(Credit to the Storm Prediction Center: http://www.spc.noaa.gov/climo/torn/fataltorn.html)
That is, 95% of all tornadic deaths this year occurred in EF-3 or higher tornadoes, which will destroy most or all walls in a house. EF-5 tornadoes can even expose the basement and descend into it (it is a myth that the funnel would magically stop at the ground level if an open hole existed for it to twist into), sucking people to their deaths. And it gets even more significant when you dig deeper into the data. A look at the list at the top of that page shows that only 4 of the deaths from EF-2 or weaker tornadoes occurred in permanent houses. I don’t know exactly what happened there, but it could have been extremely bad luck such as a tree falling on the house, a piece of heavy furniture, or a piece of timber causing injury. It could have been a weak structure. The point is, this is very rare. The rest of the deaths in EF-2 and weaker tornadoes were in trailers, vehicles, outdoors (all highly dangerous places to be in a tornado) or were unknown.
I respect the research into this year’s terrible tornado casualty count. It is important to determine exactly why this occurred, and one question that did need to be answered was whether it happened because of bad decisions. This is the question that the social scientists have attempted to answer. I simply disagree very strongly with their apparent conclusions, as I think they are unwarranted by the questionable research methodologies, and are little more than speculation. My contention is that the catastrophic death toll is directly attributable to major, violent tornadoes, the kind that obliterate entire homes, happening to occur in a lot of populated areas this year. In short, it was a statistical outlier year. This classification does not address the underlying structural problem of the Southeast, which is that effective storm shelter is not commonly available for the most violent events, but that’s not an easy problem to resolve. Unfortunately, in my opinion, it is this hard problem, rather than comparatively easy ones regarding bad decisions, that must be answered if this type of death toll is to be prevented from ever happening again.