Proofiness - Charles Seife [123]
40
It is impossible if we’re talking about the mean number of sexual partners, as implied by the Associated Press’s use of the word “average.” However, the study was actually referring to the median number of sexual partners, not the mean. Technically, it’s possible to have the median number of male partners differing dramatically from the median number of female partners, but it’s highly, highly unlikely—and in fact polls that report the mean number of sexual partners also have the men-have-more-partners-than-women flaw.
41
This assumes that the number of males is roughly the same as the number of females, which of course is true.
42
The evidence of lying is clear. In 2003, a study showed that if you hook people up to a fake lie detector machine, the gap between the men’s and women’s answers narrowed dramatically.
43
When the networks discovered the error, executives said that the problem would likely make them more careful when it comes to using polls to project the winner of a race. Every few years, like clockwork, an exit poll fails in an embarrassing way. Like clockwork, the media promise to be more careful next time. And nothing ever changes.
44
For example, in January 2006 the New York Times asked people two questions about the controversial warrantless wiretapping program that was then coming to light. One question had a clause about Bush’s claim that the wiretaps were necessary to fight terrorism. The public came out in favor, 53 percent to 46 percent. An alternate wording of the question omitted Bush’s claim, and the public came out against, 50 percent to 46 percent. Few other news organizations were as careful. For example, another major polling firm, Rasmussen, asked a question that completely failed to mention anything about not needing a warrant or a court order for the wiretaps. Not surprisingly, its survey found that the public was decidedly in favor of wiretaps.
45
A fun example: a poll where you ask half your subjects, “Do you think it’s acceptable to smoke while praying?” and ask the other half, “Do you think it’s okay to pray while smoking?” Even though the questions are asking about essentially the same behavior, the polls would have very different outcomes.
46
Electronic voting is a little different, in that there need not be a physical token to count. More on this in appendix B.
47
Coleman was dogged by allegations, which turned into an FBI investigation, about improper gifts given to his wife by a major campaign contributor. However, neither Coleman nor his wife was charged with any wrongdoing.
48
There was some old-fashioned political mythmaking too. Knaak accused an election director of mishandling ballots, driving around with them in the back of her car. The accusation was repeated by the Wall Street Journal and a number of other news outlets—Sean Hannity harped on it for several nights running. It turned out to be false.
49
In fact, even 50/50 errors don’t always “cancel out” and can be expected to produce fairly large swings in favor of one party or another on occasion.
50
Whether the people whose votes are counted really represent the democratic ideal of the “will of the people” is a deeper question that will be left to the next chapter.
51
Many journalists don’t understand this. During the recount in Minnesota, Minneapolis’s ABC affiliate commissioned a poll to see who would win the election if a revote were held. I’d love to meet the genius who thought that a poll with a 4.2 percent margin of error could give any insight into an election where the candidates differed by 0.007 percent of the votes cast.
52
This particular error was (quietly) corrected roughly two weeks after I notified the nonprofit Citizens for Election Integrity Minnesota, who apparently used back channels to notify the Minnesota Secretary of State’s office about this and several other problems. Eventually, I wound up writing a four-page letter to the secretary of state and to the two campaigns outlining a large number of precincts that seemed to have made tabulation errors. Quite