What's Up With the Polls?

Isn’t it fun reading all the polls? Except when your candidate’s numbers are tanking! I’d like to take one recent development in the nationwide tracking polls and use it to illustrate a few things about the business of evaluating the public mood.

Surely you noticed that after the third debate, over the course of several days, Barack Obama lost a couple of polling points to John McCain. The race tightened across a number of surveys. What does that mean?

First, real changes in polls only happen when something substantive changes in the campaign atmosphere. In this case, it is the increasingly angry and strident tone that John McCain took in the third debate and that his campaign has continued since. Among other things, McCain is now saying that Obama is a “socialist” who will “spread the wealth.”

Who do you think is persuaded by that kind of talk? The conservative base! Lo and behold, there were still a few right wingers out there who hated McCain so much that as recently as October 16 (the evening of the third debate) they were telling pollsters that they were undecided or even for Obama. Those right wingers finally shrugged off their hatred—which was going to happen sooner or later. (See for yourself. Compare DailyKos poll internals for October 16 [1] with internals for October 20 [2]. Republicans moved from 84 percent for McCain compared to 8 percent for Obama and 4 percent undecided to 89 percent for McCain compared to 6 percent for Obama and 2 percent undecided.)

But substantive changes in the campaign atmosphere can boomerang. The selection of Sarah Palin, for example, drove McCain’s numbers up after the GOP convention, but her unfavorable ratings now seem to be dragging the ticket down. Similarly, the new, more strident tone of McCain’s campaign will turn off independents (which seems to be happening already according to the latest Zogby poll).

What else do current polls show? A lot of voters who say they’re undecided are liars! Right now, about 7 or 8 percent of voters say they are undecided about the presidential race. Just two weeks before the most important election in memory, don’t you find that shocking? How could anyone who cares enough to vote be so clueless at this point in the race? Well, like those right wingers who finally came out for McCain, some aren’t really undecided. But others aren’t really voters. Someone who is truly undecided is not very motivated to vote—and a lot of them won’t.

They get counted because at the beginning of any random-digit-dial poll of “likely voters” each respondent has to say that he or she is going to vote in this election. Some people exaggerate their likelihood of voting because they’re embarrassed to admit they won’t vote. There is a solution to this sampling problem. Instead of deciding who is likely to vote by asking respondents, pollsters can use a list of registered voters that has been coded to show who has a history of voting. As a political consultant, that’s what I always did whenever possible. The problem is, that kind of voter list doesn’t exist nationwide and it rarely exists statewide. So random-digit-dialing is what we’ve got to work with.

Finally, surveys from different pollsters don’t compare well to each other. Theoretically, polls are highly scientific, accurate measurements. If 20 polls were each based on truly random samples of Americans who will actually vote in the upcoming election, the results of 19 out of 20 would be grouped closely together.

But it’s impossible to get a truly random sample of voters. Some voters are more likely to answer the phone than others, and some are more likely to cooperate than others. So pollsters have to make assumptions about the makeup of the electorate—how many of each party, gender, race, and age will actually vote. And, quite frankly, nobody knows. Different guesses by different experts account for most of the differences among the polls.

When you’re looking at a lot of polls, the numbers aren’t as important as the trends. If most of the polls are trending up or down, or if they are mostly stable, that’s real. With the exception of the recent strengthening of Republican support for John McCain, the presidential race has been exceptionally stable throughout October. (Here’s the DailyKos trendline [3], for example.) Except for Obama’s huge fundraising advantage—which allows that campaign to deliver more messages and mobilize more field staff—there’s nothing going on that’s likely to change the stability of the race.

The writer is a Senior Fellow at Campaign for America’s Future and author of the recent book, Framing the Future: How Progressive Values Can Win Elections and Influence People [4].