There's been several diaries today on the latest SurveyUSA poll of Ohio showing Obama beating McCain by 9 points, 48%-39%. I caution you guys from getting too excited, to first take a look at the cross-tabs in the poll.
It assumes 52% of the voters will be Democrats, only 28% will be Republicans, and only 18% will be Independents.
Is this realistic?
The 2004 Ohio exit poll shows that the electorate was 40% Republican, only 35% Democratic, and 25% Independent. This is a far cry from the 52% Democrats this month's poll shows.
Update: Some people asked about the 2006 numbers, given how much Ohio changed that year. Well, from the 2006 exit poll, the electorate that wiped out Blackwell and DeWine was only 40% Democrat, 37% Republican, and 23% Independent. That's an 8-point swing in our favor from 2004, but still nowhere near 52% Democratic turnout.
Furthermore, in a troubling number, among Ohio Independents, SurveyUSA shows Obama losing to McCain 46%-36%. Obama only wins the state based on the ridiculously high percent of Democrats in their poll. Skew the sample enough one way or another, and you can have any result you want. If I were to do a poll of 1,000 people in California, and have 800 Republicans, only 100 Democrats, and 100 Independents, guess what? It'll show McCain winning handily, and it'll be worth shit too.
Now, SurveyUSA's poll from April had it at 49% Democrats, 31% Republicans, and 17% Independents. And that one had Obama losing to McCain 47%-45%. And the one from March had 44% Democrats, 34% Republicans, and 18% Independents. There, McCain was beating Obama 50%-43%.
So SurveyUSA is including more and more Democrats in their sample. Now, there may be a reason to do so, if that's the kind of response they're getting. The Des Moines Register's pollster Anne Selzer did this with the youth vote in Iowa, because that's what their polling was showing, and she was really the only one who nailed the caucus results, thanks to showing the massive increase in the youth vote, when other pollsters were going by the outdated models of years past.
Obviously, a lot has changed since 2004. I would certainly expect more than 35% of the electorate in Ohio to be Democrats this time around. (Back in March, 68% of voters voted in the Democratic primary. But then, the Republican primary was essentially over at that point, so that's hardly an accurate snapshot of how and where Ohioans will break in November.) But 52%??? And where did all the Independents go? Are they really dropping below 20% of the total electorate in November? The March SurveyUSA poll seems more accurate, with 10 percentage points more Democrats voting than Republicans. I could buy that. Let's say those are the correct ones, then re-adjusting this month's SurveyUSA data would give us:
Obama: .12*(34) + .73*(44) + .36*(18) = 42.68
McCain: .81*(34) + .14*(44) + .46*(18) = 41.98
If you assume 44% Democratic turnout and 34% Republican turnout, as the March poll had, then you'd have Obama leading McCain in Ohio by less than 1 point, 42.68%-41.98%.
It'd be nice to hear something from SurveyUSA on this sample, and how they got 52% Democrats.