Why Can't The Presidential Polls Just Get Along?

Are you one of the people checking the Web several times a day to see how the presidential race is going? It's a lot like people watching the market fluctuations on the business channels or baseball fans sneaking peeks at the playoffs on their BlackBerrys.

If you're one of us obsessives, you probably already know the presidential race index known as the "RCP average." It refers to the average gap between Barack Obama and John McCain in the polls listed on the Web site Real Clear Politics.

Recently, that RCP average raised a lot of eyebrows by indicating that the gap in percentage points declined from 8.2 on Oct. 14 to 5.0 on Oct. 19. It even dropped below 5.0 in what you might call "intraday trading," suggesting Obama had lost nearly half his lead. Looking at the individual polls in the average, several had reduced Obama's lead from high single digits to low single digits.

Needless to say, this was a heartening trend for McCain supporters, who have not had much good news from the polls (or anywhere else) since mid-September. McCain supporters and surrogates were quick to suggest the voters were cooling on the young Democrat while the Republican's new message on taxes was resonating.

They had a feeling that Obama's "spread the wealth" comment (made impromptu to the now-famous Joe Wurzelbacher) was triggering fears of tax-and-spend Democrats swarming Washington. On Fox News, a pollster for Strategic Vision opined that "McCain was finally sounding like a Ronald Reagan conservative" in the last debate. Michael Barone wrote a column for National Review Online asking whether "Joe the Plumber" could "turn it around" for McCain.

But no sooner did this story line take hold than a new wave of polls came out this week showing Obama's lead essentially unchanged. In fresh polls taken of registered votes since the last debate, Obama was holding his lead by 11 points (Gallup Daily Tracking Poll) and 14 points (Pew Research Center).

So who's right? Could it be none of the above? All of the above?

Truth is, all these polls are legitimate and arguably correct, by their own lights. But they are describing slightly different parts of the same landscape. So you need to know a little about polling to put together the full picture.

The RCP average dropped suddenly last week for two reasons. First, the index dropped several polls that had begun their field interviews more than a week earlier. Second, the seven polls still in the index were all polls of likely voters (LVs) only. Likely voters are a smaller core within the larger world of registered voters (RVs), which is itself a subset of the total voting-age population.

The RCP ticked upward this week in part because it included a new poll of likely voters, done by ABC and The Washington Post, that pegged the Obama lead at 9 points.

Obviously, the likely versus registered argument is important.

For many years, pollsters have tried to improve their accuracy by picking the likelier voters to interview. In 2004, for example, John Kerry ran ahead of George W. Bush among registered voters in polls, but behind among likely voters. On Election Day, the likely voter number proved more predictive.

How do pollsters decide who is likely to vote? By far the best predictor of voting likelihood is voting history. If you haven't voted before, the research says you are not likely to vote this time either. Newly eligible voters and people who have never developed the voting habit are historically the least likely to vote.

Those most likely to have voted (and so most likely to vote again) are also more likely to be older than 30, better educated, better paid, married and white. So, the preferences of older, more affluent voters will be better represented among the LVs than the RVs. And on Election Day, the actual results have been closer to polls among LVs than RVs.

Will this dynamic hold in 2008?

That is the essential question underlying this election. If younger voters and people of color turn out in historic numbers, as the Obama campaign insists they will, they will not only defy history; they will change it.

And they may change the presumptions of polling as well.

In recognition of this possibility, Gallup now divides its results on likely voters into two reports. One uses the conventional definitions of voting likelihood and is called "Gallup Traditional" (rather like Coke Classic). The other allows the possibility of expanded turnout among nontraditional voting groups and is called "Gallup Expanded."

So on Oct. 18, for example, the tracking poll of RVs for the previous three days gave Obama an 8-point lead, the "Gallup Expanded" measure of LVs gave him a 4-point lead and the "Gallup Traditional" a 2-point lead.

That 2-point lead in the Traditional metric was often cited as evidence McCain was closing the gap. But three days later, all three of the Gallup measures had moved 3 or 4 points in Obama's favor.

You can check out the changing numbers in all three of these polls at the Gallup Web site.

And you can do it as many times a day as you like. See you there.

Comments

 

Please keep your community civil. All comments must follow the NPR.org Community rules and terms of use, and will be moderated prior to posting. NPR reserves the right to use the comments we receive, in whole or in part, and to use the commenter's name and location, in any medium. See also the Terms of Use, Privacy Policy and Community FAQ.