Political Animal


September 14, 2012 10:30 AM Blizzard of Polls

By Ed Kilgore

I thought about reacting to this week’s presidential polls in batches or according to particular themes, but gave up. As usual, Nate Silver has the best quick summary, which includes a slight dash of cold water to those Obama supporters thinking the last Marist poll from this or that battleground state means the election is over and we can all just go home:

[W]e’ve suddenly gone from having perhaps two state polls released every day to more like a dozen. (There has also been an increase in the number of national surveys.) When this happens, there can be the tendency for the news media to focus on those polls that confirm its current narrative about the race, while ignoring those that might tell a different story.
The state polls are more newsworthy, I suppose, if you had been skeptical about the notion that Mr. Obama had gotten a bounce, and needed further confirmation. Mr. Obama’s probability of winning has increased over the last few days at betting markets like Intrade and as according to bookmakers. There was not much reason to doubt that Mr. Obama had gotten a bounce, however, so this may simply mean the conventional wisdom was a little slow to catch up.

To put it another way, it’s reasonably clear now that Obama has moved into a small lead (though within the MoE of most polls) after the two conventions ended, a trend that is now being confirmed in state polling. But the size of that lead, from infinitesimal to significant, varies enough to make it important to look at all the evidence instead of a selective picture.

Certainly the behavior of the Romney campaign during the last week has given the impression of an operation that needs to take risks, as one that is nearing various signposts in public opinion where past campaigns have almost invariably failed.

And moreover, some of the new state polls—particularly the new batch out late yesterday from WSJ/NBC/Marist—show Obama hitting some major markers, as noted today by Nate Cohn:

The bigger splash came from three surveys conducted by NBC/WSJ/Marist in Ohio, Virginia, and Florida, which showed Obama leading by substantial 7, 5, and 5 point margins among likely voters, respectively. Perhaps as importantly, Obama was at 49 or 50 percent in all three states.
The underlying numbers were just as troubling for the Romney campaign. In all three states, Romney’s comeback chances start with gaining the support of undecided voters, but just 25 percent of undecided voters in Florida, Ohio, and Virginia held a favorable impression of Romney, compared to 43 percent for Obama. Now, Obama’s 43 percent tally is hardly impressive, but it would appear more than high enough to block Romney from sweeping every undecided voter in these three states, which is essentially what it would take for him to fight to a dead-heat. Overall, a prohibitively low 40 percent of Ohio voters have a favorable impression of Romney.

At some point, therefore, if current trends persist, a lot of the factors cited for a long time by Republicans as disguising an underlying Romney lead—e.g., the use of RV rather than LV samples; the antipathy of undecideds towards Obama; the incumbent’s failure to reach the magic 50% mark in either support or job approval—just don’t seem to be borne out by what we are now seeing. Add in the historical evidence (as explained by John Sides in a new article for the Monthly which will be discussed separately) that the presidential debates don’t typically have much of an impact on the results, along with the positive economic signs emerging this week, and you do get the sense Team Mitt is starting to run out of time. But the race remains close enough that it wouldn’t take a miracle, just some breaks and better candidate performance, to make the home stretch the cliff-hanger it’s long appeared to be.

Ed Kilgore is a contributing writer to the Washington Monthly. He is managing editor for The Democratic Strategist and a senior fellow at the Progressive Policy Institute. Find him on Twitter: @ed_kilgore.


  • stormskies on September 14, 2012 10:38 AM:

    That it is even 'close' at all is the real story about the nature of a large and vast amount of our fellow citizens. It takes crazy, stupid, and delusional citizens to vote for stupid, delusional, and crazy stupid candidates.

    Unfortunately that is a large part of America. And, that, indeed is the real tragedy and sad fact of our country.

  • c u n d gulag on September 14, 2012 10:45 AM:

    Polls of "Likely Voters" have at least one problem:

    They can't count the number of 'voters' who'll 'likely' be prevented from voting.

    GOTV is still critical folks!
    We can relax on 11/7, and not a second until then.

  • Celui on September 14, 2012 10:54 AM:

    Polling, by its very nature, is exclusive. It depends on the method of contact (phone, door-to-door, location and time of poll-taking, among others). So, polls by and large reflect a rather limited sample. Even with the margin of error disclaimer, the danger of modern polling methods gives the layman a false idea of who the frontrunner is. Polling, at its best, is akin to the race sheet at a horserace; who's likely to win based on mud, time of day, jockey, competition, weight of the jockey, etc. So, the latest polls are, at best, misleading due to the very nature of polling. Best polls ever seen are taken at a variety neighborhood gathering places (barber and beauty shops, 'real' delis, a shopping mall, and done face-to-face. Respondents are less likely to game the questions then, and the polling sample tends to be more representative. No poll should ever be taken at dinner time on the phone.

  • David in NY on September 14, 2012 10:59 AM:

    Interestingly, little that Celui (ci-dessus) said is true. It basically shows a profound misunderstanding of polling.

  • Th on September 14, 2012 11:27 AM:

    This all sets up some interesting debates. The Romney people for some reason think that Obama is thin-skinned and can be rattled by ridicule. And, I doubt Obama has anywhere near the respect for Romney he had for McCain.

    Who thinks Romney will be better than Obama at sticking the shiv between the ribs with a friendly smile on his face? Not me.

    Romney: Your first response to the attacks on our embassies was to apologize.
    Obama: Actually, as the President, my first response was to organize security forces in the area to protect our brave men and women serving our country. I have to leave the politicizing at a time like that to people like you.

  • Joe Friday on September 14, 2012 11:32 AM:

    Willard now utilizing the classic RightWing inoculation tactic of preemptively accusing your opponent of exactly what you are guilty of, by telling Stephanopoulos that he expects that Obama will lie during the debates.

    This from the guy that lies about his record of lying.

  • Rip on September 14, 2012 11:37 AM:

    Romney is in "need a miracle" territory, as " better candidate performance" is not going to happen at this late date. The best he can realistically hope for in the debates is that he does himself no further damage, and tops low expectations.

    It's not over, but the election is beyond Romney's ability to win on his own. He will need external events to help him pull out a victory.

  • Sisyphus on September 14, 2012 12:05 PM:

    What the Margin of Error Means:
    It is the radius of the confidence index for a poll. It doesn't mean that if Obama polls at 52% and Romney at 48, with a +/-4% MOE, that it's a statistical dead heat. The Margin of Error is applied to an individual value, not a difference. This means that Obama is the preferred candidate for 52%+/-4% (48-56%). And that Romney is preferred by 44-52%.

    Without knowing the confidence interval for the poll, the MOE is almost completely meaningless. Please, please, please read this: http://www.stat.auckland.ac.nz/~iase/publications/4/292.pdf

  • Robert Waldmann on September 14, 2012 2:14 PM:

    MOE mentum

    If you look at a bunch of polls, it makes no sense to consider the MOE of just one of them. The simple reasonable thing to do is to average the polls and look at the MOE of the average. It is not the average MOE of course. It is the (rootmeansquare) average MOE divided by the square root of the number of polls which are averaged.

    The post is very good, but there is a bit of MOE mysticism.