Reader Challenge
Here's a challenge to readers of this blog: can you identify a state in which Obama isn't doing better among early voters than he is among the electorate as a whole?
It's not an idle question. For the past few days, I've been pushing a consistent line of attack against several of the tracking polls - outfits like Gallup, Rasmussen, Franklin & Marshall, and Hotline/Diageo. These polls have shown Obama running about as well among early voters as he is among the electorate as a whole. In some cases, the claim is restricted to those who have already voted; in others, it includes all voters who tell pollsters they intend to vote early; and with Gallup, apparently both.
My problem with this claim is the mounting body of evidence that it's flat-out wrong. There's another set of polls - Newsweek, ABC/WashPost, Zogby and Pew - which show Obama leading among early voters by massive margins, far outperforming his showing with voters at large. Since, on average, this second set of polls shows a substantially larger Obama lead than the first set of polls, there's good reason to think that much of the underlying disagreement relates to likely-voter screens and demographic modeling. And if that's so, the first objective data on turnout that we've seen this year seems to give the edge to the second set of models.
Just today, for example, the partisan edge among the third of North Carolinians who've already voted increased to 55-28%, and a PPP poll found Obama leading among early voters, 63-36%. That's an extreme example of a general trend - of the nine states that have released partial or complete returns by partisan registration, eight have shown a greater proportion of Democrats among early voters than among the electorate at large. The exception is Florida, where an initial GOP lead among absentee voters is quickly being overwhelmed by a surge of Democratic early voters. By the time the numbers are updated again this morning, I expect it will fall into line. (Not to mention that Suffolk has him up 60-40% among early voters; in general, the partisan turnout edge is just a fraction of the total lead, a margin padded by support among Independents and defecting Republicans.)
The polling tells a similar story. Even in McCain's home state of Arizona, a recent poll put Obama up 47-46% among the third of early voters, but down 49-42% among the two-thirds who have yet to vote. The closest I've found is Indiana, where SurveyUSA puts Obama up 50-46% among ballots cast, but just 48-45% among those yet to vote.
I'm sorry if this is getting boring for those who have read through my stream of posts on this subject. But I'm still hoping for an explanation of how four prominent tracking polls can continue to claim that the early electorate resembles, in its preferences, election day voters. We've spent months arguing whether or not the polling this cycle is accurately reflecting what's happening at the ground level. Now, we have a hint of an answer.
So my challenge to readers is to document counter-examples, places where Obama's performance among early voters seems to lag his overall showing. And my challenge to pollsters is to explain themselves and their models, or to adjust them to conform to emergent reality.
Afternoon Update: Pew and ABC/WashPost double down; the latter finds the largest leads (74-25% already voted, 69-29% plan to vote) in the battleground states. Gallup and Rasmussen, on the other hand, maintain their silence (even as Gallup's traditional voter screen now shows an implausible 49-47% race). Can there be any doubt that there's something wrong with their modeling? Why aren't poll-watchers (Blumenthal, Silver) demanding an explanation of what went wrong with their projections of the early vote, and how they plan to adjust for it?
Advertisement
Personally I feel their main discrepancey has to deal with likely voters. Those just cannot be accounted for. Obama has so many people voting for the first time it's just known, but not calculated. That's why there will be at least one suprise state come November. Maybe a GA, that these pundits will have no idea we would have won.
October 28, 2008 9:18 AM | Reply | Permalink
Polls are weighting their results to account for expected bias, such as self-selection (i.e. who will respond to questions at all). But a real number is the difference between actual registered party members' requests for early or absentee ballots. That is hugely one-sided, in Democrats' favor. Another is new registrations, again hugely Democratic.
October 28, 2008 10:15 AM | Reply | Permalink
Also people have to tell them what they did and I suspect many will lie...
October 28, 2008 1:32 PM | Reply | Permalink
No doubt. But that's a general hazard of polling. To use that explanation to account for the discrepancies I've outlined above, you'd have to explain why Obama supporters are twice as likely to lie to one set of pollsters about early voting as to another. Also, why only Obama supporters seem to be lying. And finally, why the actual numbers now being reported by states would seem to confirm those lies.
In short, though I don't doubt that there's the usual degree of prevarication, I don't see any reason to think it's producing these discrepancies.
October 28, 2008 1:40 PM | Reply | Permalink
"The exception is Florida, where an initial GOP lead among absentee voters is quickly being overwhelmed by a surge of Democratic early voters."
You probably know this, but to clarify, the absentee voting in FL began much earlier than the "early" (polling place) voting. So the short-lived Rep lead was baked in the cake.
October 28, 2008 6:08 PM | Reply | Permalink
I really don't get the concept of "lying" about who you voted for. If someone asked me who I just voted for I would have to vomit if I said it was John McCain. I think the "Bradley Effect" isn't a factor for two main reasons:
1. Decades have passed, and times have changed. We have seen Colin Powell, Condoleeza Rice, and many others, confront and conquer the question of confidence in black leaders among the most reisistent of our populace. Most of us have had black supervisors and co-workers, which was NOT the case during the time of the Bradley election.
2. Those who truly are prejudiced against a black President can give other reasons for their choice: experience, religion, choice, war, etc. They can stand up and say they are voting for the other and never admit that they have a racial motivation for it, so they don't have any reason to pretend that they are voting for Obama while doing otherwise.
OK, I am watching him now. WE are so going to win this! I just wonder how republicans will manage to stay out in the sun after this -- they need to slither under rocks
October 28, 2008 9:05 PM | Reply | Permalink