Suffolk University Poll, Ohio: Tied, 47-47
Update: Reuters IPSOS Ohio: 46-46, Tie

A Suffolk University poll in Ohio just published. It has it 47-47, tied.

As far as favorability, it's also a statistical tie -- Obama 50/45, Romney 50/46.

The partisan split, D/R/I, is 39/35/27, or D+4. (Not 39-34, as my typo first had it.) Captain Ed just noted in his post on the CBS/Qunnippiac poll...

In 2008, the D/R/I was 39/31/30, while in 2010 it was 36/37/28

D+4 seems ballpark plausible, if you guess (as I would) that 2012 will be somewhere between 2008 and 2010.

20% of voters say they've already voted. CAC keeps noting that polls overstate how many people have already voted -- the people claiming to have voted early in polls has been, in past polls, something like nearly double the actual number of voters who have voted early. I don't know how 20% tracks with the actual numbers at this point.

Those who already voted say they favored Obama 54-41, but those yet to vote (80%) favor Romney 48-45.

The economy/jobs is rated as the most important issue, at 52%. But then, Obama narrowly edges Romney (!!!) on this score, 45-44.

It seems there's a bias towards people who say they've already voted. Either these people are over-eager to answer pollster's questions, and wind up being overrrepresented in samples, or Democrats are just inclined to bullshit questioners about their actual level of political involvement.

Over the weekend I spent an hour or two checking the polls for 2004 and 2008 to see how the races went then. Just to get a sense of how accurate the polls were.

In 2004, Kerry seemed to be slightly head in Ohio, or tied... until about right now. There was a late break for Bush. Then again, Bush had been ahead there before the debates, so it was a bit of a coming-home effect.

But in 2008, polls had a persistent lead for Obama through the entire season. The final RCP average was +2.5 for Obama -- but he won stronger than that, +4.6%.

If this is going to be a 2004 sort of year, we should start seeing a break towards Romney pretty soon.

Weak Voter Screen: If you check the poll, you'll see its only voter screen is a question asking people to rate their likelihood of voting. Anyone who says "Very likely" is a likely voter.

That's barely a screen at all, and that's what's causing divergences between polls with more involved screens (like Gallup) and most other polls.

Look, most people do not like admitting they are not interested in something all the Smart Set tells them they should be interested in. And they don't like admitting they're not going to do what they're told is their civic duty.

So there is a socially-favored answer here: "Yes, I'm very likely to vote."

That's why Gallup asks a bunch of questions to try to smoke out the bullshitters, asking if you've voted in the previous two elections, and if you know where your local polling place is. Likely voters tend to say yes to these questions; unlikely voters tend to say no.

It now seems that Gallup is one of the few polling companies that asks a battery of LV screen questions, rather than simply asking people to rate their likelihood of voting.

Reuters IPSOS, Tied, 46-46: Is this that late-break thing starting to happen?

The online survey of likely voters found Obama and Romney were each supported by 46 percent of the electorate as they prepare for their final televised debate on Monday night.

Note that this is an online survey. They try to be scientific about these things (unlike those useless "everyone just send your people to freep this poll" fake polls), but it's still not really the same as random digit dial polling.

I think they attempt a randomization, but it's all through internet users, so who knows whether that's really representative.

posted by Ace at 02:04 PM

| Access Comments