Why McCain/Obama Polls Will Be Wrong

The way everyone follows the polls, you might think they actually mean something. Everyone wants to know who’s ahead and who’s behind. The problem is, in 2008, the public opinion polls can’t provide an accurate answer. There’s simply too many new variables this time around.

The pollsters are already trying to account for the Bradley effect. This is the tendency of some voters to claim they’ll vote for a Black candidate, but who, in the privacy of the voting booth, can’t or won’t do so. That’s why the pollsters were wrong in the 1982 California gubernatorial campaign. They showed Tom Bradley beating his Republican opponent, Attorney General George Deukmejian by a decent margin. On election day, however, Mayor Bradley’s thin margin was more than wiped out by the GOP’s groundbreaking absentee ballot campaign. (For the record, I was a Deputy Campaign Director in the Bradley campaign, so I got to experience the “effect” first hand.)

Pollsters have gained some experience in adjusting for voters saying one thing and doing another in state campaigns, but they have no experience in a national election. But the fact that Senator Obama is the first African American nominated by a major party is just one of many variables. Another is the presence of Governor Sarah Palin on the Republican ticket. Just as there are some voters who won’t vote for an African American, there are some who will hesitate to put a woman one heartbeat away from the Oval Office. And still others who will say they’ll vote for a 47 year old or a 72 year old for president, but won’t. These racists, sexists and ageists are fools, but, from a pollsters point of view, many of them are also liars. The question is, how many of these liars are there out there? And how do they balance out? That’s a lot of adjusting required.

Adjusting for fools and liars is tough enough. But the pollsters also have to account for the unprecedented influx of new voters Senator Obama’s campaign is recruiting. According to the Associated Press, in Pennsylvania the Democrats have added 375,000 voters, and Republicans lost over 115,000, since 2006. Democratis added over 165,000 voters in North Carolina, and Republicans lost over 35,000, in the past two years. In the 28 states that report party registration, Democrats have added 2 million voters while the GOP has lost over 340,000 since 2006.

And the voter registration efforts continue. With colleges and universities back in session, the Obama campaign is hard at work recruiting young voters. The pollsters will attempt to adjust their findings for this army of new voters, but the numbers are unprecedented. How many of them will actually turn out on November 4th? Will the Obama campaign pivot to an absentee ballot campaign among their core supporters sometime in the next eight weeks? If so, will the traditional assumptions pollsters use to calculate turnout in certain age groups hold up?

These factors make polling an approximate calculation at best. Throw in the fact that an unexpected incident (a candidate gaffe or health problem, a serious flare-up overseas, a terrorist attack at home) could shift voters dramatically in one direction or another, and the predictive power of polls becomes even more tenuous.

Certainly, the pundits will tout their polls. Sites that report on polls from a variety of sources will remain popular (e.g., Real Clear Politics). And candidates will tout their momentum when they have it. But in the long run, they won’t be any more accurate than they were in predicting Senator Hillary Clinton’s “surprising” win in New Hampshire back in February.

The candidates should ignore the polls and just remain paranoid. Each should assume they’re behind everywhere they need to win and act accordingly. After all, just because the polls say you’re ahead, doesn’t mean you are.


2 Responses to “Why McCain/Obama Polls Will Be Wrong”

  1. Sam Smith Says:

    Along with all the very real concerns you have expressed about the inaccuracies and falabilities of current polling data, I heard an interesting comment by an oft quoted Republican consultant last night that another problem with polls is that they are traditionally done by telephone. How many 18-25 year olds do you know who even have a land line telephone? Enough to offset the Bradley effect perhaps?

  2. Alan Says:

    Sam, you’re right. The Associated Press is reporting that the Pew Research Center for the People and the Press, “in three separate surveys this summer, contacted users not only of traditional land lines but also of cellphones. Each time, Obama came out 2 to 3 percentage points higher than when Pew called land lines alone.”

    (The AP story is on Yahoo! at: http://news.yahoo.com/s/usnews/20080929/ts_usnews/dopollsmissobamavoterswithnolandline)

Comments are closed.

%d bloggers like this: