Electoral Polls: Conflicting and Confounding

Sen. Barack Obama has a huge lead. No, Sen. John McCain, III is closing the gap. The next poll shows Obama running away with the election. With McCain, it is a 2-point race. No, Obama is 7 points ahead.

This election season features conflicting polls galore — four times the number in 2000, close to three times the number in 2004. We can’t figure it out when we see one poll saying Obama is 16 points ahead of McCain, while another cites a statistical tie — on the same day.

To obtain a reasonably accurate sampling of American voter sentiment, we must know the precise composition of the electorate. One poll or survey, whichever term is used, may have included all respondents, regardless of whether they were registered to vote. Another may have included only registered voters. Yet another may be based upon a group of respondents evaluated most likely to vote. Hence, we could have three very differing results.

When I was a campaign consultant, we usually took a survey at the beginning of the race to determine where the candidate stood in terms of name recognition, positives, negatives, match-up with the likely opposition, etc. Around the middle of the campaign, we would survey the same folks to determine whether we had made progress in swinging voters in our direction. We also surveyed a week out so we would know what we needed to do to close the gap and win.

This year is hardly like that. Every local and national media outlet wants to poll. Inasmuch as survey research is only a snapshot in time, trends are important. If our candidate is ahead by 5 points but the significant trends are contrary our candidate easily could lose if corrective steps are not taken.

The polls this year appear to be something else. They appear to be driving results. The media favor one candidate for president very much. By overly sampling or choosing a polling organization, perhaps a university, that has had little polling experience, those commissioning the poll may secure the result they wish. A series of quickie polls all showing one candidate way ahead can demoralize the opposition.

In addition, but very important, the geography of a presidential poll is very significant. Everybody knows the District of Columbia and Massachusetts will vote overwhelmingly Democratic, Alaska and several small Mountain and North Central States will vote Republican (if not necessarily as overwhelmingly).

Thus, polling of swing states is much more meaningful.

Cliché that it is, it’s worth repeating: There is only one poll that counts: Election Day. Democrats and Republicans appear to be neck and neck in states with early voting. We well may be in for a long night Tuesday.

Paul M. Weyrich is chairman and CEO of the Free Congress Foundation.