New York Times blogger and statistical analyst Nate Silver did his usual thorough job yesterday explaining why he’s not taking Mitt Romney’s strong performances in the Gallup tracking poll too much to heart. His piece, “Gallup v. the World” rightly pointed out that the firm’s tracking polls, which have given Romney leads of 6, 7 and 6 points in the last three days, are the most favorable yet published for the Republican. He conceded that Gallup is the most reliable of the tracking polls in that it employs the largest samples and employs a methodology for counting cell phone owners as opposed to landlines only. But he claimed that Gallup has a history of inaccuracy in recent elections that ought to cause us to take their conclusions with a grain of salt. That’s a fair point, though it should be noted that we never heard much about Gallup’s shortcomings in recent months when its results (which showed Obama with a lead) were unquestioned while the rival Rasmussen poll (which generally gave Romney better numbers) was consistently called into question.
But as long as we’re discussing methodology, it’s worth pointing out that the only surveys keeping the president’s head above water in the national average of polls are two whose credibility are very much in doubt. I wrote earlier in the week that the Washington Post/ABC News poll published on Monday that showed President Obama with a three-point lead was called into question by the sample employed by the pollsters. That poll was based on a sample that had nine percent more Democrats than Republicans; a figure that is far more than is reasonable. The same thing can be said about a new Hartford Courant/University of Connecticut poll that also shows Obama up by three but on the basis of a sample that has eight percent more Democrats than Republicans. If you adjust both of these samples to create a more representative group of Americans, even one that showed the Democrats with an edge in affiliation, it would mean they would show Romney and not the president ahead in the race.
These kind of misleading polls are not doing the Democrats any service. While these misleading numbers may help encourage the president’s supporters, they are potentially setting them up for a great fall on November 6. While, as Silver has rightly argued, party affiliation isn’t set in stone and reflects the changing fortunes of the candidates, the idea that there are more Democrats prepared to vote for Obama in 2012 than there were in 2008 when he won in a walk is ludicrous.
All polls are fallible and their results are open to interpretation. But the decision of publications to go ahead and put out polls that they already know are based on faulty samples demonstrates either bad judgment or political bias. Overcounting Democrats in polls may produce numbers that tell liberals what they want to hear. But those with more reasonable methodologies are pointing in a very different direction. With the race coming down the homestretch, it appears that some in the media are determined to portray the election in the most positive light for their side even if it means bending the truth a little.