Legend has it that Labour MP Bessie Braddock once slighted Winston Churchill by remarking on his inebriation. He replied, "And you, madam, are ugly. But I shall be sober in the morning."
Mitt Romney has lied and Republicans have cheated religiously in an attempt to defeat Barack Obama. Over the last three weeks public polls seemed to reward Romney's efforts, confirming his claim that Obama has lost his edge. Like the effects of liquor, however, the sequelae of campaigns are temporary.
America is already sobering. Come Wednesday, Obama will no longer be an embattled incumbent, just President Obama once more. Romney will still be Romney.
On Sunday evening,Nate Silver announced that averaging the 12 polls released over the previous 24 hours gives Obama a lead of 1.3 points. The Real Clear Politics average, too, now shows Obama's lead restored on average, albeit by just .4 points. These upticks are heartening, but there are even better reasons to remain optimistic.
Most public polls have underestimated Obama support. The exceptions are Democracy Corps, an independent non-profit founded by Democrats, and the Pew Research Center, an non-partisan public opinion research outfit. Democracy Corps analysis shows Obama's lead averages 7 points in the battleground states and has expanded tofour points nationally. Pew polling similarly shows Obama ahead by three, 48 to 45 percent. Pew projects that allocating the undecided voters boosts the candidates equally, making it 50 to 47 percent.
Why trust these two polls over, say, Rasmussen? Pew and Democracy Corps use people to call respondents -- not robo-calls, like Rasmussen -- and even reach voters who rely primarily or exclusively on cellular phones. Rasmussen does not call cells. The better polls incorporate interviews over multiple nights; Rasmussen makes calls on just one night, opening the poll to possible bias. Although Public Policy Polling (PPP) also uses automated calls to poll, they do so over the course of multiple nights.
Pew and Democracy Corps also interview many more people than most pollsters. Pew surveyed 2,709 voters while Democracy Corps called 1,080 as compared to the CBS-New York Times poll's 563 or CNN's 693. As a result, they are more accurate: Democracy Corps has a 2.98-point margin of error while Pew's is 2.2.
News coverage of polling often overlooks crucial details and caveats. Take for example the margin of error. The seemingly authoritative Washington Post-ABC tracking poll claiming likely voters were 48 percent for Obama and 49 for Romney. The fine print qualifies that their poll has a 3-point margin of error.
Translated, that means Obama could have majority support (51 percent). On the flip side, Romney might only be winning 46 percent of voters.
Stories offer infuriatingly scant information about the polls they cite. Gallup's tracking poll of likely voters gave Romney the lead 51 to 46 percent on October 29. The significance of the finding is substantially diminished by discontinuity in Gallup's methodology as well as the interruption of polling after Superstorm Sandy. The disparity between Gallup's numbers and those of other national polls reflects on Sandy, not Barack Obama or Mitt Romney. In other words, beware of consulting Gallup too closely between now and Tuesday.
Mainstream media polls rarely elucidate how they decide who is a "likely" voter, or the significance of the difference between polls of likely versus registered voters. Obama has led by greater margins among registered voters. Granted, the last that Rasmussen commented, in July, they had concluded that registered voter polls overstate support for Obama. The problem is, neither Rasmussen nor any other media poll uses a sophisticated method to screen for likely voters.
How a pollster decides who is "likely" to vote differs from firm to firm. For the most part, mainstream media polling does not disclose its methodology.
The best guess, however, is that these surveys just ask voters to rate their likeliness to vote. Reputable private pollsters, by contrast, often use multiple methods to determine who's most likely to vote. For example, they collect data points about individual voters, cross-correlating voters' responses and these profiles to see which variables correspond to whether they actually vote. They use that information to model turnout rates, predicting rather than guessing the composition of the electorate. In a tight race, these tiny methodological tweaks make an enormous difference.
Of course, the greatest reason to be confident is the Electoral College. Nate Silver advises, "We are approaching the point where Mr. Romney may need the state polls to be systematically biased against him in order to win the Electoral College." Obama could lose the popular vote by more than a point and still win re-election. State polls have consistently illuminated a clear, if sometimes narrow, path to re-election.
Obama's advantage in state polls is doubly significant given that state polls are more accurate predictors than national polls. Romney's challenge all along has been to find 270 electoral votes. With one day to go, it's apparent he's fallen short on that score.