Only accurate poll is on election
day
29 October 2004
My prediction: John Kerry will win the presidential election.
In
Elsewhere? I have no idea.
We are obsessed with guessing what will happen before it happens, with replacing surprise with certainty, mystery with knowledge. It used to be we learned a baby's gender at birth. Ultrasounds have taken away that unknown, making moot prepartum angst over pink bedrooms or blue. Gifts once were secret, revealed only with the ripping of wrapping paper. Now it's gift cards, where the recipient chooses exactly what he or she will get while it is the gift-giver, oddly enough, who has no idea of what was just presented.
And when it comes to this year's election, our oracle is not tea leaves or the karma of the Red Sox. It's polls. Clothed in science, there are hundreds of them, some weekly, others daily. Add them up and here's what they tell you.
No one really knows.
It's not that polls are worthless. It's easy for me to
predict the results in
Moreover, national polls make clear that this election is not 80/ 20 or even 60/40. It's close - and when things get close, polls can't tell you much.
There are three reasons for that. First, polling has built-in inaccuracies. There is, of course, the familiar "margin of error." Polls rely on sampling to represent the thinking of more than 130 million Americans. Mathematically speaking, there are bound to be differences.
There are other, less trumpeted inaccuracies as well. People lie or misunderstand. The way a question is worded (or even the order in which names are read) can throw off results. And sometimes the sample doesn't really represent the population as a whole. For example, as many as 6 percent of all Americans now only have cell phones. Pollsters can't reach them. A poll by Literary Digest in 1936 used the phone book to contact voters and erroneously predicted Alf Landon would beat Franklin Roosevelt. As it turned out, those who had a telephone were more likely to be Republicans. Do cell phones introduce a similar inaccuracy today?
Second, a poll is not a vote. When polled, people make guesses. When they vote, it's final. Pollsters understand this, qualifying their results by saying, "If the election were being held today." But the election isn't "today" - it's in the future, and new events can readily cause people to change their minds. That seems especially true this year. George Bush makes goofy faces in a debate and Kerry's numbers rise. Kerry calls Mary Cheney a lesbian and Bush's margin increases.
The third and perhaps most important reason polls can't predict Tuesday's results, however, is that we don't know who will turn up to vote.
Adult Americans may be entitled to the vote, but that doesn't mean they actually will cast a ballot. Blacks vote less frequently than whites. Senior citizens turn out in droves compared to everyone else. If you're a pollster, the trick in figuring out a future election lies in guessing who, really, will end up voting. This is the place where polling turns from science to art.
One way to guess is to look to the past. You might think, for example, that if someone voted in two of the last four elections, that person probably would be a "likely voter." That's not a bad measure, but it's by no means perfect. Indeed, much of the on-the- ground campaigning by both parties focuses on turning out those who normally don't vote.
Democrats, for instance, have made an enormous effort this year to persuade college students to vote. Young adults (ages 18-25) have a typical turnout of 25 percent vs. the national average of 45 percent. If those numbers rose, however, the belief is most of their votes would go to Kerry, giving him a big and - from a pollster's point-of-view - unanticipated boost.
Meanwhile, Bush strategist Karl Rove argues that 4 million Christian fundamentalists didn't vote in 2000. He thinks they're logical Bush voters and the GOP has been working strenuously to get them motivated. If they show up, their votes too will confound the pollsters.
All of which means that - unlike the birth of a baby or gifts under the Christmas tree - Election Day truly will be a day of surprise.
Unhappy with that? Can't take the gnawing uncertainty? Then perhaps you can find solace in another poll, an almost unerring one conducted for the last 64 years by Scholastic Magazine of children in grades 1 through 8. Based on its poll, Bush wins 52 percent to 47 percent.
But take heart, Kerry supporters. The kids have been wrong twice - in 1948, when they (like pollsters) picked Dewey over Truman, and in 1960, when they figured JFK to lose.
There's just a chance they could be wrong on this year's JFK as well.