Skip to main content

Presidential Polling an Inexact Science

In America, we don't just want things, we want them now. Drive-thru McDonalds, Google on iPhones, overnight shipping more expensive than the pair of shoes. ... Naturally, it was only a matter of time before the national quest for instant gratification leaked into politics, and so began our obsession with the presidential preference poll.

Indeed, Gallup's daily election polling results pop up everywhere online and in print, and its Web site now hosts an "Election 2008" page that has support for the two candidates broken down into everything from the ubiquitous race and gender to education, marital status, and even "purple" states. It may be tempting to grasp at the elusive "know now!" promise, but as Don Rose pointed out in his article for several weeks ago, polling results are, more often than not, misleading.

Taking a look at a Gallup Poll Accuracy Record for presidential elections of the last 72 years, that conclusion would appear to be true, most notably in 1948, when Thomas E. Dewey was predicted to win by five points and instead lost by roughly that amount. As recently as 1992, predictions were off by as much as six points. And though for the past few elections the deviation has stuck to one or two, with elections so close, even two percentage points can spell loss instead of victory.

Which means that sometimes, all there is to do is just wait and see.