NumberMuncher: Ohio's Looking Better Than The Polls Claim

Could both the national polls (Romney ahead) and the Ohio state polls (Obama ahead) be right?

Possibly, but probably not. NumberMuncher Josh Jordan suspects the state polls of being out of whack.

It's a Read the Whole Thing deal, with numbers and such, but the math is easy, even for a Moron (TM). Here are some bullet points:

In Ohio, Republicans tend to outperform their share of the national vote: In the last six presidential elections, only in 2004 has the Republican candidate performed worse in Ohio than he did nationally, and even that was a difference of only 0.3 percentage points. In the other five elections, the GOP candidate outperformed his margin of the national vote by an average of 3.1 percentage points.

...

Democrats’ national turnout advantage is usually bigger than their Ohio turnout: Not since 1996, during Clinton’s reelection campaign, have Democrats had a larger turnout advantage in Ohio than they did nationally. In 2000 and 2008, Democrats were 2 points under their national turnout advantage and were actually 5 points under in 2004. Polls are currently showing an average Democratic turnout advantage of 6.3, which is 1.9 points higher than their current advantage nationally.

Now, about those partisan ID splits:

As mentioned above, in current Ohio polls, Democrats have a party-ID sample advantage of 6.3 points. In 2008, Democrats had a 5-point turnout advantage in Ohio. That means that while national polls have the turnout advantage down 2.6 points, in Ohio it has actually increased 1.3 points. It is almost impossible to conclude that while the nationwide party-ID advantage of Democrats has dropped since the wave election of 2008, Ohio has actually increased over the last four years.

I can't quote it all, but he suggests a culprit for these out-of-whack splits that I mentioned a week ago: Early voting's effect on the polls. Many polls (but not all) feature a Likely Voter battery of questions designed to determine who is truly a "Likely Voter" with more rigor than simply asking them "Do you consider yourself a likely voter?"

Actually -- shockingly -- some polls just use that one question as their Likely Voter "screen," making it more of a sieve than a screen. (Yeah and I know those two things are pretty similar but you get my point.) They ask just one question -- how likely are you to vote, or two questions, adding a question asking about the interviewee's level of interest in the race.

But other screens ask a few more things, again, to separate out the people who just say "I'm a Likely Voter" because it's the socially preferred answer from those who are actually, genuinely likely voters.

Now, the point is, if someone says "I already voted," you can't really ask additional questions. You can't follow up that answer up with "Are you sure you already voted?" and "No, seriously, buddy, tell me the truth, did you already vote or are you just jerking my chain?" If they say they already voted, they get put directly into the likely voter pool without those bothersome questions designed to smoke out the bullshitters.

This may also be why there are so many more "already voted" voters in polls as compared to actual state records.

NumberMuncher had similar thoughts in his last column, and also noted the Incumbent rule.

This is a rule that always receives some skepticism, but it’s very likely to benefit Romney at least some on Election Day. In 2004, late deciders broke against George W. Bush heavily, even though he was a wartime president. John Kerry beat Bush by 25 points among voters who decided in the last month, 28 points among voters that decided in the three days prior to Election Day, and 22 points among day-of deciders. Those voters were 20 percent of the Ohio electorate; while this year there are expected to be fewer late deciders, Obama cannot afford to lose among by those margins and still win.

His overall thrust is that tied or slightly behind in Ohio state polls = likely Romney lead in actual votes.

posted by Ace at 11:29 AM

| Access Comments