RealClearPolitics: Romney on top for the first time this year

RealClearPolitics shows Romney ahead in the popular vote for the first time this year

This is their "poll average." It attempts to measure national popular vote.

The electoral vote estimates still show Obama with a solid (but shrinking) lead:

• Real Clear Politics calls it 294-244.
• 538 calls it 297-241 based on the average result of multiple sims. Obama's high water mark was 321-217. The site has established that Romney has a 28.8% chance of winning, but that has risen steadily from 12.9% on October 4th.

Reader comment:

I'd note that I'm profoundly suspicious of the average numbers in this case. Check out the historical poll data here:

You'll have to click "All General Election: Romney vs. Obama Polling Data" in the top table or scroll down like a common animal.

You've already discussed Romney's best score in the Pew poll. It traditionally skews Democratic but something is spectacularly wrong with the shift in partisan identity.

Take a look at the other polls that comprise the average as I'm writing this, and the previous polls by that company.

Rasmussen actually shows a 2 point increase for Obama since the debate.

Gallup has dropped 2 points for Obama since the debate and 1 additional point since the poll before that. However, Gallup has applied its LV filter between the most recent poll and the present one, and LV filters generally favor GOP candidates. Additionally, Gallup was far below the median in terms of Obama support prior to the debates. Its results have generally synced with Rasmussen IIRC.

Investor's Business Daily/TIPP has also had a swing of 4 but it to has just applied a LV model. They're previous to last poll had Obama up by 7 but this was well before the debates. ("CSM", which I'm guessing is the Christian Science Monitor, stopped polling with them. It'd be interesting to know why.)

Politico/GWU/Battleground shows a net decrease of 1.

CNN/Opinion Research is down 3 but it was conducted prior to the debate.

Now look at all those companies that previously had very good Obama numbers that haven't repolled yet.

I'm an emotionally invested Obama supporter, but it looks to me like the following has happened:

1) Certain polling outfits owing to survey methods, electorate modelling or whatever, have been showing a pronounced Pro-Romney bias or house effect, coming in below the average Obama support numbers. Some of these, like Rasmussen and Investors Business Daily have a management and clientele that skews Republican. Gallup probably does not.

2) For whatever reason, these companies were most keen to measure the immediate effect of the debate fallout. Gallup because of its bellweather reputation, the rest likely because they wanted to be the first to show a Romney upsurge as a selling point for other GOP folks who might commission polls from them.

3) Coincidentally or by design, two of them switched from RV to LV while this was happening.

4) Thus firms with a Pro-GOP bias or house effect constitute a disproportionate number of the pollsters who contribute to the latest RCP Averages.

5) By some freakish coincidence, the only poller with Dem bias or house effect gave the phone number randomizer a hit on the crackpipe this week, skewing the results even further.

6) The result is that if you

a) Throw out the obvious garbage from Pew; and b) Account for the changes brought on by the application of LV to Gallup and IBD; then you're looking at a situation where the net drift towards Romney is basically zero. When the polling from the more neutral or Dem leaning pollsters comes in the next two weeks, it'll snap back up to the pre-debate levels.

This makes sense to me because the narrative that the media is playing "Romney's debate performance has won 4-5 million voters over" is absolute fucking batshit insanity. Would these be the same voters who bailed on him for calling them parasites (basically)? What points did he make that were so persuasive 10% of the viewing audience switched their support from Obama to him? The narrative that he blew up his support base among seniors and poor whites with the 47% comment made perfect sense. What kind of voter was going to have been persuaded by his debate performance? And are there actually enough of these voters to have had the stated effect?

Scoop's comment:

We're probably getting into too much depth for my jokey little nudie-and-sports blog, not to mention my jokey little brain, but I would add a comment on the party self-identification shift in the Pew poll. That baffled me. Their September poll showed that 717 of 2424 respondents called themselves Republicans. In October it was 403 of 1201. If we establish the confidence level at 95%, that variation is too great to be caused by chance. Assuming chance to be eliminated as a cause, I studied the two polls bit-by-bit to see if Pew made a change in methodology or something, but they did not, and all their numbers were kosher. Nothing else changed significantly in the composition of the samples except the % of people who identified themselves as Democratic or Republican. I was baffled by this result until I read Nate Silver's comment on this on the 538 blog. He pointed out that party self-identification is itself fluid. If Nate is right, then it is possible that Pew could have run into that exact same shift in party ID even if they had called the exact same people as in September.

In other words, the percentage of people who identify themselves as Republican and the percentage who prefer Romney are correlated, but neither is fixed, and no causality can be presumed. We have been assuming that more people preferred Romney because more Republicans were sampled, but it may be that more people now call themselves Republicans because they now prefer Romney.

I dunno.

538 still lists Romney's chance of winning as only 28.8%, reflecting statistically what you summarized in prose, i.e. let's not go too far off the deep end until we see more evidence over a greater period of time, because (1) these polls may be misleading; (2) even if they are not misleading and the shift is real, it may be ephemeral; (3) even if the effect is neither misleading nor ephemeral, the election is decided in the Electoral College, not by the popular vote, and we need to know more about Florida, Virginia and Ohio.