I am always stunned when people – especially media pundits – fail ti understand the basic statistical concept of variability. I was hit by this awareness again as I hear the Iowa Caucus results interpreted.
One poll reported that 41% of Iowans were still undecided when they walked into the caucus. So 4 in 10 votes were cast by people who planned to be swayed by the emotional speeches of candidates’ supporters: an emotional state that would cease to exist within minutes of the vote. The dominant pattern in the results – more than any of the leading candidates – was the pattern “swayed in last minute choices.
A situation of high variability indicates that most people do not have firm support or commitment for anyone. If “bring us more candidates” had been an option in Iowa, I suspect it would have won.
Nonetheless, donors are crossing some candidates off their list, and the candidates are dropping out. If the caucus had been a week earlier, Newt would be explaining what he did to take the lead and Santorum would be reconsidering his options. If it were a week later, others would have ganged up on Rick and dropped his numbers.
Students in statistics class place a great deal of emphasis on the mean, and have a very hard time grasping the fact that the confidence interval really conveys much more important information. The media pundits and, yes, the American voter seem to have the same problem.
God help us. This is going to be a very long year.