It is with growing concern that I have been following the Rasmussen daily tracking poll over the past two weeks. Starting on March 15th, a survey that included interviews from March 11-14, John McCain broke out of a tight contest and took a 47%-42% lead over Barack Obama. From that day forward, McCain's lead has been a minimum of 6% over Obama, a lead that he replicates over Hillary Clinton and which has also been reflected in Rasmussen statewide polls.
Now, while Rasmussen was terribly wrong in 2000, and while it is headed by a Republican, its track record in both 2004 and 2006 was pretty solid. Also, while the other seven Obama vs. McCain non-tracking polls conducted entirely since March 11th show Obama ahead by an average of 1.3%, there is an important difference between Rasmussen and those seven polls. Specifically, Rasmussen is an automated, IVR style poll that does not use live interviewers, while the other seven non-tracking polls all utilize the more traditional, live-interview survey methodology. As such, I am worried that we are seeing a "Bradley effect" or "Wilder effect" between the IVR polls, and the live-interview polls. Perhaps people are telling an automated survey something that they would not tell a real life person.
Is there a hidden McCain advantage, based on a Bradley effect, where he is actually ahead by 5-8%? It is one possibility, and I find it quite worrying. It is also entirely possible that Rasmussen simply is wrong, given that their national polls have almost always been out of step with other national polls this year. For example, on the even of Super Tuesday, Rasmussen showed Mitt Romney tied with John McCain nationally, although the actual voting bore out very difficult results. It might also be possible that Rasmussen is using far too tight of a "likely voter" screen, especially given that the general election is so far away and that the Democratic campaign in 2008 has consistently demonstrated an ability to bring out some less than likely voters.
It is very, very difficult for me to believe that John McCain can win this election, given the enormous structural advantages Democrats have this year (fundraising, partisan self-identification, the general mood for change). My gut tells me that once Obama emerges as the presumptive nominee in the minds of the electorate, that he will gradually start to pull away from McCain and win comfortably. However, another part of me also worries that what Matthew Yglesias wrote today is true, and that in terms of electability, we were always deluding ourselves that either a woman or an African-American were ever really all that "electable" nationwide:
I heard a liberal Obama skeptic remark a couple of months ago that it would be a strange day in America when the correct answer to the question "who's the most electable" was "the black guy." I think that's right, and it's a reminder that though the cliché is to say that Democrats are torn between two very strong candidates, in some ways we're torn between two very weak ones.(...)
Given the extreme strong underlying pro-Democrat fundamentals, it's very hard for me to imagine how a "generic Democratic white dude" like Chris Dodd or John Edwards or, indeed, John Kerry would lose in this environment.
I don't know how likely it is, but the Rasmussen poll suggest that it is indeed possible that Obama does not hold a small lead over McCain, and instead actually faces a decently sized, 5-8% deficit at this point in the campaign. Might a live-interview "Bradley effect" be inflating Obama's (and Clinton's) numbers outside of IVR polls? Might Obama and Clinton always have been further behind McCain than live-interview polls would suggest? It is certainly a possibility, and a worrying one to consider.