We have very foggy election forecasts heading into next Tuesday’s voting, and unless something changes in the next several days, one or more of the nation’s top polling organizations is going to be flat wrong. According to Rasmussen and Gallup, Mitt Romney is winning. According to Quinnipiac/CBS/NY Times, President Obama is winning. They can’t both be right.
Those findings are based on national tracking polls. But even at the state level the polls differ sharply. Rasmussen finds that Romney has a two-point lead in the critical state of Ohio, while Quinnipiac finds that Obama has a five-point lead there. And when it comes to measures such as early voting the polls are again divided. Gallup shows an edge for Romney in swing states with early voting, while CBS/NY Times surveys show Obama up in the states that matter.
What’s happening here is that they are using different models for predicting who is actually going to show up and vote next Tuesday, and in what numbers. Quinnipiac, CBS, and other such as Time magazine and Survey USA, are using turnout models that mirror closely to 2008 and predict that young voters, minorities, and union voters will combine with enough elderly voters and urban/suburban voters to carry Obama over the top in key states such as Ohio, Virginia, Nevada, and Colorado.
Ramussen and Gallup see it differently. Their predictions forecast a turnout more like 2004 when Bush defeated Kerry, and their preliminary survey findings lead them to weight Republican turnout higher than Quinnipiac does. For example, Rasmussen and Gallup find that Romney leads among Independents, and leads significantly on economic, finance, and tax-related survey questions. Their model also disagrees with Quinnipiac on the numbers of young voters, low-income voters, and minority voters that will actually turn out to vote on Tuesday.
By contrast, the other major respected polling group, Pew Research has it tied, at 47 – 47. In that model, independent and late-deciding voters will determine everything.
It’s unusual for the major polling firms to be this divergent so close to an actual presidential election. You have to go back to at least 2000 and probably 1980 or 1976 to find the polls this far apart. But it provides someone with a real opportunity to be right; and of course it also means someone is going to look really bad next Tuesday night.