It's fairly clear that President Obama holds a healthy-sized lead in the race for president. A look at the RealClearPolitics average has Obama by about 4 percentage points, with some polls higher and some lower. That's what we'd expect given random sampling. Averages work because the center of a bell-curve is usually right.
Yet, it is the pollsters who are showing the closest races – that is, in this contest, those the friendliest to Mitt Romney – who are, not surprisingly, receiving the most attention from conservatives. One of the two pollsters with consistently good Mitt Romney numbers compared to the average has been Rasmussen Reports.
Rasmussen polls are mostly ignored by the mainstream media. Whether it is because of Scott Rasmussen's Republican connections, a Republican bias in 2010, or the pollster's usage of interactive voice response instead of live interviews (which I personally think is a bogus reason for lack of attention), Rasmussen remains polling's version of the black sheep of the family.
The other poll consistently favorable to Romney has an illustrious history. The Gallup poll has been around longer than most of us have been alive. Starting in 1935, Gallup has surveyed in over 140 countries on many different issues. You could argue that the word Gallup is to polling as Coke is to soda.
Gallup passes the press's tests for a "legitimate" poll, which Rasmussen fails. You don't see Gallup editor-in-chief Frank Newport going on conservative cruises as Scott Rasmussen does. Gallup uses live interviewers. Gallup calls cellphones. One prominent Washington political analyst I've spoken to has called the Gallup tracking poll the political horse-race equivalent of "crack".
The question is whether Gallup deserves the outsized attention it is receiving this year? The answer in my opinion is an unequivocal no. Here are three reasons why:
1). Gallup has if anything been less accurate in the past than other pollsters in presidential elections
Sometimes, certain pollsters know something others don't (for example, Ann Selzer and the Iowa Caucus). I went back into the record books, thanks to the National Council on Public Polls and Real Clear Politics, to see if Gallup works the national equivalent of Selzer's Iowa magic. The answer is no, and it's not even close. Here's Gallup's record over the past five presidential elections:
Gallup has been worse than than the average pollster in three out of the last five elections. On average, Gallup has been worse by 1 percentage point.
The average of all the polls (as opposed to comparing Gallup to each pollster, we pool all the pollsters, including Gallup's data, together to form a RealClearPolitics-like average) has had an even better record. The aggregate of all polls beats in Gallup four out of the last five elections. Its average error has only been 2 percentage points – nearly a point and a half better than Gallup.
Gallup has, in other words, been below average compared to other pollsters and is almost always beaten by aggregate of all the polls.
2). Gallup in 2010 also had a Republican tilt and it was wildly off-the-mark
We've all heard the saying "Fool me once, shame on you. Fool me twice, shame on me." In the 2010 midterm elections, Gallup projected the Republicans would run up a 15-point margin in the national House of Representatives vote. Only two other pollsters were even close to this prediction of a mammoth lead: Fox News at 13 points, and Rasmussen at 12. Though the majority of evidence pointed away from Gallup, Gallup's previously solid record in House elections meant it couldn't be ignored.
The election result was a Republican win by 6.8 percentage points, which means Gallup was off by a little over 8 points.
Gallup has, apparently, done little to change its polling methods since. They included cellphone surveying in 2010 and still do. Its 2010 demographic samples were, according to Gallup, typical of a midterm election. They have fewer minorities than other samples in 2012, but that hadn't previously impacted the accuracy of their polls. Overall, Gallup seems to be doing most everything right.
Yet, why should we trust Gallup and its Republican House effect, when they showed the same effect in 2010 and it was so inaccurate?
3). Gallup leans more Republican than you think
Gallup is currently one of the most pro-Romney polls in the RealClearPolitics average, but it's the only poll whose population is registered voters. The registered voter population tends to lean more Democratic than the actual election day electorate that pollsters try to measure through "likely voters". Even with sky-high Democratic enthusiasm in 2008, Gallup's likely voter model was still 2 points more Republican than its registered voter results. Gallup's likely voter model was 3 and 4 more points Republican in 2000 and 2004, respectively.
When Gallup turns on its likely voter screen, past experience dictates that, instead of the tied race it currently sees, Gallup will find Romney ahead by a few points. That would make it the most pro-Romney polling data by a few points. That would also make it more friendly to Romney than Rasmussen.
Conclusion
Both Gallup and Rasmussen continue to show a close race that most other pollsters don't see. That doesn't mean they should be ignored; they should be examined as part of a larger pool of polls that indicate a clear Obama lead.
Gallup was a pioneer in the polling industry, but first doesn't equal best. Gallup may have the right ingredients, yet the sauce is just a little off. There's no reason to think they are right this time around while everyone else is wrong.