Will Undecided Voters Break Toward Romney?

Like Gen. Francisco Franco in 1970s episodes of Saturday Night Live, Mitt Romney's momentum is still dead. The eight tracking polls I've been reporting on since they started reflecting post-debate polling show a net drift, since Monday's final pre-debate polling, of 0.9 points toward Obama (compared with 1.0 yesterday, 0.6 the day before, and 0.9 the day before that).

But this stasis could get disrupted any day now. About 5 percent of voters are still telling pollsters they're undecided. And traditionally,* late-deciding voters -- which this 5 percent certainly qualifies as -- wind up mainly voting against the incumbent.

The most interesting analysis of this phenomenon I've seen lately -- and the most alarming for Obama supporters -- comes from Bob Krumm (no, not Bob Shrum, and for that matter not Robert Crumb). Take a look at this collation he's done of this year's polling results and the polling results from 2004, the last time we had an election featuring an incumbent. The dashed purple line and the solid blue line represent the incumbents in, respectively, 2004 and 2012, and the other two lines represent the challengers.


Note the relative fortunes of the challenger and the incumbent in polling done during the final week of 2004. That looks like more than 1 point of net gain for the challenger. That's the kind of thing that could erode the peace of mind of Obama supporters. Here's another such thing: New York Times polling guru Nate Silver -- whose forecast currently gives Obama a 70-something percent chance of winning and is a daily source of reassurance for many Democrats -- has a model that fails to take adequate account of the dynamic on display in 2004. At least, that's what Krumm says. He says Silver's model is too heavily informed by the 2008 election, which was unusual in featuring neither an incumbent nor a presidential candidate who was the sitting vice president. [But see update below for a link to Silver's questioning the conventional wisdom that undecideds regularly break toward the challenger.]

Right now pretty much everybody has Obama as the favorite--not just Silver, but all the betting markets and the statistical aggregations done at the Huffington Post and at RealClearPolitics. This consensus, as I explained here, probably makes sense even in light of the fact that Romney has a close-to-one-point lead in national polls. However, if Romney's lead in the national polls gets beyond two points, and polls in battleground states change by a comparable amount, the consensus may start to fray.

[Note: Here are the latest numbers from the tracking polls that underlie the 0.9 figure cited in the first paragraph. The numbers for each poll represent the change relative to polling done on Monday, the day of the last pre-debate polling. Gallup, no change; IBD/TIPP, no change; IPSOS/Reuters, Obama +1; PPP, Obama +2; RAND, Obama +4.4; Rasmussen, no change; UPI/C-Voter, no change; Washington Post/ABC, no change. For reasons I explained here, it's tempting to exclude the RAND poll, which would reduce Obama's net gain since the debate from 0.9 to 0.4 (and would reduce all the other numbers cited in the first paragraph as well). But either way you get a net drift toward Obama since the debate, and either way the drift is so small that it doesn't make sense to say either candidate has momentum. All these polls can be found (in regularly updated form) here, except for RAND, which is here.]

*Update, 10/27, 11:55 p.m.: Commenters zic and mikewaz point to two relevant posts -- respectively, (a) a recent post by John Sides at The Monkey Cage doubting that undecideds will break significantly toward Romney; and (b) a post from July by Nate Silver questioning the conventional wisdom that undecideds in general break significantly toward the challenger.