Political Animal


October 31, 2012 8:59 AM Repeat After Me: Nobody Knows

By Ed Kilgore

The most interesting things to come across my screen early this morning were two separate pieces by two well-regarded number-crunchers from two different ideological orientations on the same subject: the continuing divergence of national and state polls of the presidential contest, with the former generally favoring Romney and the latter Obama. After conducting extrapolations of state data to a national profile of the race, both Nate Cohn of TNR and Sean Trende of RCP come to the same conclusion: the divergence really is a matter of disagreement, not a sign of an impending popular vote/electoral college split. And neither finds any ready explanation, or a clear reason to prefer one batch of polls to another. In some cycles (1996 and 2000) state polling has been more accurate, and in others (2008) the national polls have gotten it right. Nate Silver recently made the case for state polls being pretty accurate, particularly when the show a consistent lead, but he admits the evidence of state poll superiority is limited.

In any event, the Cohn-Trende consensus, if it spreads, should throw cold water on the rampant discussion, which was beginning to approach CW status, that Romney was going to win the popular vote and Obama the presidency, based on (a) blue vote falloff for Obama or (b) southern red-state hyper-fanaticism for Romney. It’s true that national polls (and the Cohn-Trende analysis) haven’t factored in the possible falloff in voting levels in the states most affected by Sandy. And for that matter, the polls aren’t yet in a position to reflect a possible bounce for Obama attributable to his administration’s emergency management performance (though Gallup’s last tracking poll, suspended on Monday due to Sandy’s impact on response levels, did show an intriguing three-point jump in Obama’s job approval rating).

But the bottom line is that the final stretch of this campaign is likely to remain something of a mystery, particularly with all the manic spin going on about “who’s winning,” especially on the Right, where confidently predicting a Romney victory or even a landslide appears to be a powerful herd obligation. There’s certainly more than a little incentive for base voters on both sides to take the trouble to vote. In particular, Obama supporters should note that most of those national polls showing Romney ahead among likely voters also show Obama leading among registered voters. GOTV is how you get the former sample to more closely resemble the latter.

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.


  • lou on October 31, 2012 9:21 AM:

    Yeah, a real Prilosec moment. No prescription for a Romney presidency.

  • c u n d gulag on October 31, 2012 9:27 AM:

    Nobody knows, and yet, that's ALL we hear about!

    How many pinhead pollsters can you fit on the head of a pin?

    I'll take Silver and Wang, since they do theirs scientifically, without over-polling Republicans, like Rasmussen and Gallup.

    And without getting naked, rubbing his body with fecal matter, and evicerating the corpse of a dead tranny hooker he killed to get to the entrails, like Dick Morris.

  • Peter C on October 31, 2012 9:37 AM:

    The popular vote is basically an irrelevant sidebar. The only thing that really matters in the short run is the electoral college.

    A split between the two (though not as good for the nation as a wave election favoring Obama) is probably beneficial; it may be the impetus to get rid of the electoral college (which is an anti-democratic complication to our elections).

  • estamm on October 31, 2012 1:28 PM:

    The only way the popular vote polling would match the state vote polling is if the popular vote polling polled each state in exactly the same proportion as the state's electoral votes. If those proportions are exactly the same, then they should match. However, whenever I see popular vote totals, the breakdown is by 'region', rather than by state, so it is really impossible to know for sure.

  • Scamp Dog on October 31, 2012 4:12 PM:

    You could do a (relatively) quick check by multiplying each state's Obama and Romney percentages by the fraction of population it has, adding them all up, and checking to see if it matches the national polling average. You'd have to talk to a real statistician to find out how closely they have to match to conclude that they track each other.

    Anybody out there have a handy spreadsheet with state-by-state polling numbers?

  • JFP on October 31, 2012 5:39 PM:

    Scamp Dog, that's basically what Cohn and Trende did. There seems to be no legitimate way to get them to agree. Either the state polls or the national polls are wrong