Political Animal


December 28, 2012 10:36 AM Polling Takeaways

By Ed Kilgore

Nate Cohn’s only real competition for numbers-cruncher wunderkind (since Nate Silver’s now a grizzled veteran!) is The Guardian’s Harry Enten. So I’m all ears when Harry offers five lessons on polling from the 2012 campaign, each with a big caveat. The most useful is his first: given the obvious difficulty pollsters are having in constructing accurate Likely Voter screens (Gallup’s enduring problem), perhaps in future we should just pay attention to polls of registered voters, so that poll consumers are in a position to make apples-to-apples comparisons instead of wondering if the LV screens are too tight or reflect outright bias.

Enten goes on to suggest that ignoring cellphone voters is no longer acceptable (though PPP has effectively compensated for the problem by adjusting for the expected electorate); Internet polling is beginning to become credible (unless you’re Zogby!); internal polls should be ignored (unless the pollster is Mark Mellman!); and state polls are usually more accurate than national polls (though they weren’t particularly good this year in Colorado, Michigan, Nevada or Virginia).

So there’s enough ambiguity in his findings that Enten and others won’t have to worry about putting themselves out of work before the next cycle.

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.


  • c u n d gulag on December 28, 2012 11:54 AM:

    OT - no offense, Ed, and WaMo, but I'm still having problems with this site.

    Who do you have to handle your computer problems?
    "Secular Humanist I. T." - aka: S.H.I.T.?

    Maybe it's time for some PC religion, and call J.C.I.T. - Jesus Christ I. T.?

    Oh, and weren't they supposed to have a kinder, gentler, CRAPTCHA ready for us, like any day now - back in the SUMMER?!?!?!?!

    Just kiddin,' IT boiz & gurlz!

  • mb on December 28, 2012 1:11 PM:

    I find it remarkable that these two venerable polling outfits, Zogby and Gallup, have become so hackish. It'd be interesting to know the inside scoop on their decline. At some point it seems they went from trying to describe reality to trying to shape it.

  • Doug on December 28, 2012 1:37 PM:

    How hard is it to take a poll of registered voters, screen THAT using the voter break-downs from most recent Presidential and general elections? Then issue the poll with the appropriate caveats. Ta-daa!
    Or does that do too much to de-mystify our new seers?

  • kahner on December 28, 2012 2:04 PM:

    There's no reason to get rid of LV screens, people just need to be aware of how they work and why they might be misleading. The reason Gallup's LV results were so bad is that they LV screen was terrible. It asked 7 questions and the repondent had to answer 6 correctly to be counted as a LV. The questions were:
    Thought given to election (quite a lot, some)
    Know where people in neighborhood go to vote (yes)
    Voted in election precinct before (yes)
    How often vote (always, nearly always)
    Plan to vote in 2012 election (yes)
    Likelihood of voting on a 10-point scale (7-10)
    Voted in last presidential election (yes)

    Another pollster, I think PPP, just asked "Are you planning to vote" and their LV screen was far more accurate. So no need to throw the baby out with the bathwater here. LV screens, if well designed, can add valuable info.