Tyler Cowen at Marginal Revolution linked to an interesting article that claimed "Of those in the contiguous U.S. killed by tornadoes from 1985 to 2007, almost half [43.2%] were in mobile homes." (In support, Phil Miller at Market Power cited one of his 2005 posts which reported that in 2000, "68.3% of all deaths attributed to tornadoes occurred in mobile homes".)
This claim prompted some interesting discussion. One of Tyler's commenters wrote, "There has always been a running joke that tornadoes tend to touch down mostly in trailer parks." Another one of Tyler's commenters wrote, "It’s almost as if tornadoes deliberately target trailer camps."
But nobody seems to have mentioned--yet--the elegant theory proposed by noted economist R. Preston McAfee. Ladies and gentlemen, The Predator Theory of Tornadoes (Competitive Solutions: The Strategist's Toolkit, pp. 244-45):
It is often observed that tornadoes usually damage mobile homes. This observation has given rise to the predator theory of tornadoes--mobile homes increase the food supply of tornadoes, which brings tornadoes.
. . . every standard-deviation increase in mobile-home sales--about 3,317 mobile homes--tends to increase tornadoes by 0.37 standard deviations, which turns out to be 94 tornadoes. This works out to be one tornado for every 35 extra mobile-home sales.
For the humor-impaired, he's kidding. But McAfee's discussion provides a memorable example of the statistical problem of an omitted variable.