With the debates fading in the rear view mirror and Election Day approaching quickly, the polls still show Obama ahead by a modest but meaningful and consistent margin in the Buckeye State.
Yesterday, CNN/ORG showed Obama leading by 4 points in Ohio, coming on the heels of earlier polls by similar firms conducting interviews with cell phone voters by Time and SurveyUSA showing Obama ahead by 5 and 3, respectively. Even ARG and Purple Strategies—which have tended to produce better than average results for Romney—showed Obama leading by 2 points. In ARG’s case, that was a reversal from their initial post-debate survey, which was one of the few to ever show Romney ahead.
What’s most striking is the consistency of Obama’s advantage. Even though three relatively Romney-friendly surveys showed Obama falling behind by 1 point after the first presidential debate, only one partisan poll has shown Romney leading since October 10—and two of the initial three surveys to show Romney ahead have since shown Obama retaking the lead. I suspect that level of consistency won’t last through Election Day, since most averages show Obama ahead by 2 or 2.5 points in Ohio (I’d actually peg it at just 1.9, since I include partisan surveys).
As a matter of probability, at least a couple polls should show Romney ahead in such a close race. Of course, when that poll comes, I’m sure a wave of Democratic panic and Republican euphoria will overtake Twitter, so let’s just establish in advanced that such a result should be expected. Start getting excited or concerned once the polls start showing movement that can be distinguished from static.
Romney’s chances dwindle to the risk of a systemic error in the polls if he can’t close the gap over the next ten days. Ask Michael Bennet or Harry Reid about whether that's possible, but one argument attempting to explain why one should expect systemic error in Ohio is that the polls are "oversampling" Ohio’s Democratic-leaning early voters, who have constituted as much as 40 percent of recent surveys, even though Michael McDonald’s invaluable US Elections Project’s Early Voting page shows that just 985,000 of Ohio’s voters have cast ballots—or about 18 percent of the 2008 electorate.
But even though all but one recent Ohio poll shows a 2-5 point race, the same polls show early voters ranging from perhaps as low as 20 to a high of 32 percent of the electorate. Although the Time breakdown is unweighted, they appear to show voters early voters around just 20 percent of the electorate. What poll shows 31 percent? Rasmussen--the closest poll. This suggests that random sampling, not systemic bias, is influencing the results. After all, Rasmussen shows a tied race with 31 percent of the electorate voting early and voting for Obama by 31 points—they get it back to a tie by showing Romney doing far better among Election Day voters than anyone else.
Of course, the average poll still shows early voters at 26 percent of the electorate, and that might seem too high. Part of the issue might be the 800,000 outstanding absentee ballots that have been sent to voters but haven’t yet been returned. Many of these voters may have “voted” in the presidential race by filling out their ballots, even if they haven’t returned the ballot. I don’t know how many Ohio voters would have filled out a ballot and say tell a pollster that they voted before it's submitted, it’s worth noting that the polls usually ask “have you voted” without further elaboration. What percentage of the '08 Ohio electorate has received ballots or voted in-person? 28 percent.
Speaking from experience as a Washington State voter where elections are conducted almost entirely by mail, I can say that I repeatedly told people that I “voted” in the presidential race since Tuesday, even though I completed my ballot yesterday still haven’t submitted it. After learning about this controversy, I asked a few fellow Washingtonians if they voted and they said yes. Then I asked whether they submitted their ballot, and all but one said no. So it’s possible.
It’s also possible that the polls are actually just getting too many early voters. But even if they do, it doesn’t necessarily skew the polls. For instance, early voters might be overrepresented if early voters tend to be hardcore partisans and if hardcore partisans are more likely to respond to polls. But in this scenario, potential bias should influence every state and the national polls, not just Ohio, since bias would be due to the unrepresentative character of poll respondents, not the number of early voters. For early voters to skew the polls, voters would need to be more likely to respond to a survey after voting than before. Is that possible? Perhaps, but there's not exactly much evidence and many alternative explanations.
A slightly different possibility is that the registered voter screens are too tight, not that the polls are capturing too many early voters. A certain number of "unlikely" registered voters will ultimately vote on Election Day, and just enough of them might be excluded by the polls to increase the "early voting" share of likely voters compared to their eventual share of the electorate. Of course, it would actually be big news if the Obama campaign was banking a meaningful number of "unlikely voters" in early voting.
The polls outside of Ohio didn't offer much better news for Romney. Yes, the national polls continue to show a tight race or even a Romney lead, but Obama has probably fared just as well in surveys of North Carolina over the last forty-eight hours as Romney in Ohio. Rasmussen showed Obama within two in Florida, even though they previously showed a 5 point race. And Obama led in Iowa and New Hampshire. If Obama wins Ohio, Nevada, New Hampshire, and Iowa, he can secure the presidency without Wisconsin.
Updated at 12:00PM: Time Magazine did not show early voters representing 40 percent of the electorate, but instead showed the sum of early voters and those planning to vote early at 40 percent of the electorate.