As you might have seen elsewhere, the two new California polls out today show wildly diverging results: Zogby has Obama up 49-36, while SurveyUSA has Clinton up 52-42. As Josh Marshall says, somebody's gonna end up looking pretty stupid. Most likely they both will--I don't expect the final tally to be more than five or six points in either direction. But SurveyUSA will probably end up looking worse.
I took a look at the crosstabs of the SurveyUSA poll (I can't seem to find them for Zogby--if any commenters can, please link!). Unfortunately, one of the most relevant things to look at is the relative proportion of African-American and Latino votes in the electorate (since, by all accounts, Obama will win the black vote overwhelmingly and Clinton will win the Latino vote easily, though by a smaller margin). In the 2004 Democratic primary, 16 percent of voters were Latino and 8 percent were black. SurveyUSA estimates the 2008 Democratic electorate as being 26 percent Latino and 9 percent black. To be frank, I would be shocked if those numbers end up being correct--particularly given the disproportionate surge in black turnout we saw in South Carolina. If I had to guess, I'd say the proportions will be about 19 and 11--which is more in line with what other pollsters predict. This alone would swing the result several points toward Obama from what SurveyUSA has.
In addition, SurveyUSA has Hillary beating Obama by one point in the Bay Area, which defies all conventional wisdom. The Field Poll, which has more expertise polling California than anyone, had Obama winning in the Bay Area 41-31--and this was just at the beginning of what looks like a late Obama surge in the state. Hillary will win the state if turnout in Los Angeles and the Inland Empire are large enough, but, again, I'd be very surprised if she won the Bay Area (full of latté liberals and African-Americans) outright.
One final point: SurveyUSA doesn't say what percentage of Democratic primary voters are unaffiliated with either party (or "decline to state" in California parlance). But most models, I think, are underestimating these voters (who break heavily for Obama) as a share of the electorate. Field had them at 13 percent of Democratic primary voters, and far be it from me to question the oracle. But consider: decline-to-state voters can only vote in the Democratic primary, and it's easy for them to do so--they just show up at the polling place and ask for a Democratic ballot. They constitute 31 percent of eligible Democratic primary voters--that is, if you add up the number of Democrats and decline-to-state voters in the state, the latter make up 31 percent of that total. Now, of course, some of those folks lean Republican, and decline-to-state voters are less likely to vote in a primary generally--so the total decline-to-state share of the electorate won't approach 31 percent. But given the salience of the race and the enthusiasm it seems to be generating, I'd bet decline-to-state voters will make up closer to 20 percent of the electorate. This would add another couple points to Obama's total. Without having seen the crosstabs, I'd bet that Zogby has Obama up by so much because (1) he has a more favorable black/Latino composition for Obama, and (2) he has a more generous estimate of the decline-to-state share of the vote.
Bottom line, the race remains very, very close--Clinton's advantage among voters who have already mailed in their ballots should not be underestimated. But I will make two predictions. First, Obama will win among voters who go to the polls today. And second, Clinton will not win by double digits. (Obama probably won't either, but there's an outside shot.)
--Josh Patashnik