Don't Count On Polls
by David Keene

Is President Bush ahead by 13 points, as one poll suggested last week, or is the race a dead heat? The experts don't seem to know, and one can sympathize, given the chaotic nature of polling these days, but casual observers and those following the race for the presidency have to be totally confused by this point.

It does seem clear that while Sen. John Kerry failed to get the "bounce" he needed from his convention, Bush did get one from the show his people put on in New York.

The question is whether his "bounce" explains the lead some say he now enjoys or whether something else is happening. The answer is that he did accomplish what he needed to accomplish in New York and is running against an opponent who just doesn't seem to be able to get his act together.

Having said this, it would be foolish and inaccurate to assume Bush enjoys anything approaching a double-digit lead at this point. The nation remains fairly evenly divided, and while it is possible that the thus-far ineptly run Kerry campaign will lose by a bigger margin than some expect, no one should bet on that happening.

The polls are all over the place for a variety of reasons. Polling is getting more and more difficult because it is harder in today's world for pollsters to draw a reliably random sample with literally millions of Americans using cellular rather than regular phones and more than ever refusing to cooperate with pollsters who do manage to reach them.

These problems have, of course, been widely reported, as pollsters and those who rely on their work whine about the difficulty of getting really accurate results or try to explain why one poll differs so much from another taken during roughly the same time period and seeking answers to virtually identical questions.

What most of us who pore over the results of every poll that appears often fail to realize is that political and other surveys are only scientific to a point. Any student of public-opinion research can theoretically draw a random sample, but such a sample isn't particularly useful in gauging how people who will actually vote on Election Day feel or will act.

To get this information, the pollster has to look at the demographic and partisan makeup of the sample drawn to see how it stacks up against the actual electorate.

Thus, if the sample contains too many more Republicans, women or minorities than the electorate, the pollster usually adjusts by weighting the opinions of some respondents differently than others or goes back and tries to adjust for the differences in other ways.

Sometimes this works, and sometimes it doesn't.

But that's only part of the problem. The real problems come when pollsters try to "screen" for "likely" voters, those who we can be pretty sure will vote Nov. 2. There is no scientific way to do this, so pollsters look to history by seeking respondents who have voted in past elections, rely on the respondents' expressed intention, and use guesswork.

It is the guesswork that is a problem. The pollster has to develop a screen that takes into account the nature of the campaign, the intensity on each side and the relative motivation of different demographic and other groups in the electorate. It's easy to guess wrong at this point, and being wrong can make a huge difference.

This year, both parties are mounting get-out-the-vote efforts unlike any we have seen. If one of the parties does better than the other in dragging its voters out on Election Day, the outcome could differ substantially from predictions and the candidate behind on the eve of the election could win. Other external events not associated with the campaigns themselves could also affect this.

Everyone involved on both sides assumes Florida, for example, will be almost as close as it was four years ago, and each side has worked to put together a ground army in that crucial state to deliver on Election Day. Their troops, however, have just managed to survive a series of horrendous hurricanes that have disrupted everything in crucial parts of the state. Both campaigns must now wonder whether people they have been counting on who have been traumatized by what's gone on down there will perform as they've hoped.

The answers to these questions cannot be answered by Gallup or any other pollster.

What a combined look at the polls and the other evidence we have indicates is that this is a close race and will be decided by the people who actually turn out Nov. 2.

It's trite to observe, as politicians who find themselves behind are wont to do, that the only real poll that matters is the one in which people vote on Election Day, but this year that couldn't be truer.

David Keene is chairman of the American Conservative Union and a Washington-based government affairs consultant.

Email the Editor