With fewer than three weeks left, election-obsessed Americans in both parties have turned into 24/7 poll-watchers, reading every number for some sign that the election has turned decidedly for Mitt Romney or President Obama.
I urge you to stop reading every poll. Here’s the reality: barring some major incident (a terrorist attack, news one of the candidates has engaged in an extramarital affair), the race will remain close and within the margin of error in most polls. Obama’s big post-DNC convention lead was surprising and probably always going to disappear at some point, as Republican-leaning independents came back home to Romney. Two of the last three American elections (2000 and 2004) were effectively ties, and this election appears headed to a similarly-close result. In 2008, President Obama won in states like Indiana and North Carolina when the Republican Party was at perhaps its lowest moment in the two decades, and he is very unlikely to repeat that kind of electoral wipe-out.
It’s not that the polls are wrong or biased. Gallup, whose surveys have been criticized as too pro-Romney this year, predicted an 11-point point win for Obama in 2008 in its final poll. He actually won by seven. The Pew Research Center, which has had wild swings in its surveys this year, had a six-point margin in its final poll in 2008, as did Rasmussen, almost exactly dead-on.
And pollsters are right to separate “likely voters” from “registered voters,” as the latter includes people who are very unlikely to show up on Election Day. (In Pew’s 2008 survey, Obama led by 11 among registered voters, much bigger than his overall advantage. Younger voters, who lean Democratic, are in particular low-turnout voters.)
Instead, there are two real challenges to polling the race. The first is that knowing who exactly will vote is hard to predict. Most pollsters estimate the electorate will be about evenly divided between men and women and minorities will compromise between 25 and 30 percent of voters.
But a one- percentage point drop or increase in the black vote in Ohio would be a significant shift, and it will be very hard for most national or even state polls to catch that. In an election like 2008, when it was clear Obama would win, those kinds of details don’t matter as much. But they will this time if the race is as close as expected.
Secondly, it’s a bit too early. Polls in the days after big events, like conventions and debates, are catching swings in the electorate because undecided voters are still sorting out how they feel. (Yes, you might not be undecided, but between 5 and 7 percent of likely voters are, a number that means up to 9 million Americans still aren’t sure.) By the Sunday before the election, barring big news, those feelings should be largely settled. Look at the polls again then.