Writing in Commentary, Andrew Ferguson on the press, gullibility, and surveys:
As the election approached, Miller-McCune bristled with research about politics. Below the glistening surface, though, the data are almost always murky. Here’s an example from early October, chosen at random. “Experimental research shows,” one story announced, “when [voters] say they intend to do something [like vote], they are more likely to do that.” Thus if a campaign calls voters before an election and doesn’t just implore them to vote, but actively asks whether they plan to do so, voter turnout will increase by as much as 23 percent. A nifty little datum, perfect for a walk-up article before Election Day, and totally certified by science!
Except it isn’t certified by much of anything. The “experimental research” consisted of an exercise conducted by 13 Ohio State undergraduates who polled 60 of their classmates by phone shortly before the 1984 election. The results (and that 23 percent) have been cited ever since, even though later experiments with larger samples have failed to produce the same effect.