During my interview blitz last week, I was asked by multiple reporters to tie the recent Moore tornado to a presumed overall upward trend in the number of U.S. tornadoes. This was a poorly posed query, for two reasons:
Firstly, it is illogical to conflate a single event (e.g., a car accident) with an trend in those events (e.g., an increase with time in the number of car accidents).
Person A: “I got in a car accident last week, therefore the number of car accidents each year is increasing.”
Person B: “I didn’t get in a car accident last week, therefore the number of car accidents each year is not increasing.”
Neither of these statements make any sense! You need at least two data points to begin to discern a trend, and many, many more to discern a statistically significant trend.
Secondly, although media coverage of tornadoes has undoubtedly increased in the last few decades, to the best of our (researchers’) knowledge, the actual number of tornadoes has not. When adjusted for population growth (read: more eyeballs looking for tornadoes), the trend since ~1950 is essentially flat. Here we see an example of what is known as the availability heuristic: People can more easily recall recent tornadoes than those farther back in the past, especially now thanks to saturation coverage of tornadoes. Viewers may reach an erroneous conclusion that the total number of tornadoes each year is increasing. I’d imagine the same holds true for shark attacks, child abductions, and deaths from falling coconuts.