But I’m particularly interested in what we’ll learn tomorrow about the purported Bradley, cell phone, and ground game effects.
With all the aggregation of polls this year, we’ve got a pretty good sense of where polls have the race. So the actual results may give a reasonably good read on several questions that have been raised this race.
For example, one of the only ways McCain is going to win even a few of the states he needs is if Scottish Haggis is correct that some people have lied to pollsters about how they will vote–he’s simply too far behind on all the polling. And in Pennsylvania, which has become a make or break state for McCain, Obama is above 50% in all but two out of the 13 polls conducted in the last week, with few undecideds remaining. So if McCain is going to win, he going to win by getting support from people who are currently telling pollsters they are going to vote Obama.
Nate has pretty much debunked the Bradley effect here and here, though the only place he found a hint of Obama underperforming was in the Northeast, so it might be a concern for Pennsylvania. And my gut feel–from seeing white working class people who once supported Obama blare their support for McCain–is that if people were going to flip because of McCain’s fear-mongering, they already would have.
Still, I think the pundits are still factoring in a Brady effect in their fairly conservative calls on EV predictions. Without a Brady effect–assuming polling averages are at all indicative of the true state of the race–then McCain’s going to be blown out.
Cell Phone Effect
This morning, Nate showed that national polls that include cell phones in their sample show a a 4.4 higher lead for Obama, on average (9.4 versus 5), than those polls that don’t include cell phones (this is an even greater margin than the 2.8 point margin Nate has calculated before, so it probably also reflects a likely voter model that incorporates early voting). If Nate is right, then Obama is going to have one hell of a blow-out tomorrow. Even the lower 2.8 point difference would put Georgia and Indiana over the top (Obama is currently behind 2.2 in Georgia and 0.9 in Indiana in the Pollster averages), and the higher numbers might put Arizona (Obama’s behind by 5.2) over the top–though of course the numbers may be lower if many of these polls are also including cell phones in their samples.
Most interesting, though, is the possibility that the polls are predicting Obama’s results too conservatively because they’re not taking into account ground game. Not all pollsters are even adjusting their likely voter models to account for the huge number of people–significantly weighted to Democratic turnout in every swing state but Colorado–who have already voted. One that has, though, is Gallup; it’s two likely voter models have converged, partly because of the large number of African-American voters who have already voted. It’s worth noting, then, that Gallup has the most optimistic numbers for Obama of all of Pollster’s recent polls: 53% to 42%.
But taking the Democratic advantage in early voting accounts for just one part of the equation. For example, Gallup shows the race among registered voters–rather than likely voters–to be 53% to 40%.
The gap in voter support for Obama versus McCain is slightly wider (53% to 40%) when the vote preferences of all registered voters are taken into account. The likely voter model typically shows a reduction in the Democratic candidate’s advantage, as has been the case with Obama this year. Nevertheless, Obama has been able to maintain a significant lead over McCain in recent days, ending in the 11-point lead in the final poll. It would take an improbable last minute shift in voter preferences or a huge Republican advantage in Election Day turnout for McCain to improve enough upon his predicted share of the vote in Gallup’s traditional likely voter model to overcome his deficit to Obama.
In other words, even with its huge margin for Obama, Gallup is calculating that Obama will lose a significant number of voters to either not voting or not having their vote count, and assuming that McCain will be able to improve on his relative weakness through the greater reliability of Republican voters.
Now, I’m not suggesting that Obama’s going to improve his turnout tomorrow over what they’ve already done in early voting, except perhaps among youth voters. But I think likely voter models that presume Republicans will reliably turn out may turn out to be wrong, particularly since McCain’s rallies today are attracting one tenth of the crowd they expected, since Republicans are underperforming Dems in early voting (though still voting early at higher rates than in 2004), and since McCain has cannibalized his GOTV funds to dump into advertising.
In other words, though Gallup’s likely voter models converged, its model(s) still assume healthy GOP turnout. But there are lots of reasons to think fewer people who say they’re support McCain will show up than Gallup and other pollsters think.
Don’t get me wrong–I think even the 11% Gallup is predicting is larger than Obama’s lead will be. I certainly don’t think Obama’s going to beat McCain by 13% or more.
But I do think pollsters may be using overoptimistic numbers for GOP turnout, particularly at the state level. Which, again, might make the difference in states like Missouri, Florida, North Carolina, and North Dakota (where Obama slightly leads in the polls), and Georgia,Indiana, and Montana, where McCain has slight leads in the polls. (To say nothing of Arizona… Update: Jeebus, even AK is closing, though ground game won’t affect it.) On ground game, I’ll be watching Indiana particularly closely. Ann Selzer, the pollster that accurately predicted the Iowa primaries this year, polled the state at 45.9% Obama and 45.3% McCain last week–basically a tie (and her polls do include cell phones). While Zogby and Big Ten have both had much bigger margins recently, it otherwise appears to be virtually tied in IN. So I’ll be very curious to see whether McCain can get his voters out to win the close one there.