By Dan Kennedy • The press, politics, technology, culture and other passions

Dial “Z” for poll numbers

I’m allergic to math, but even I can understand why Barack Obama’s eight-point lead in the Zogby tracking poll is bigger than it looks.

Zogby has consistently had John McCain doing better than in many other polls. As Nate Silver explains, it’s because Zogby weights party affiliation based on exit polls from the previous election.

So if the electorate is more Democratic today than it was in 2004 — and it is, by quite a bit — then Obama’s lead is actually understated by Zogby. That is, if the tracking poll is accurate in the first place, which is always a question worth asking.

Discover more from Media Nation

Subscribe to get the latest posts sent to your email.


Outraged Liberal reveals identity


Double digits


  1. Doug Shugarts

    The latest national Pew poll is wildly optimistic, but a bit shocking, too.52-38, Obama.

  2. Suldog

    What amazes me concerning almost any of these polls is that, at this late date and in a race so clearly delineated, the pollsters are still able to find 5% or 10% “undecided” voters.Of course, they could be voting for “others”, but I optimistically fell for that line of thinking too many times, when I was a Libertarian campaigner, to be lured by that siren again :-)They could be “refused to answer” folk, but that still seems a bit high.

  3. Michael Pahre

    (Sorry for the long comment…)While Zogby’s methodology might, at first glance, appear wrong, there is a statistical reason that you might use a methodology such as this in analyzing poll data.While analysis of data from a “closed box model” of voter registrations might be wrongly biased by the Zogby methodology, voter registrations involve new people brought into the process — and others leaving the sample (e.g., dying) — such that corrections like Zogby’s can, at some level, be justified.I use the term “closed box model” to mean that both the old (2004) and new (2008) polls sample the exact same 100 people — just that some of those people now identify themselves as Democratic rather than Republican. If the same 100 people are in both samples, then you don’t need to re-weight the data in any way — and Zogby’s is a flawed methodology.But voter registrations are not a “closed box model”: voters age (conventional wisdom indicates they become slightly more conservative with time); older voters die and don’t vote (unless buried in Chicago or New Jersey); and younger people reach 18 years-of-age and register for the first time.If the new registrations of younger voters are disproportionate with the other aging trends of the registered voter population, then you’ve got a problem because younger voters are known to turn out at rates lower than older voters. A phone survey of registered voters could become biased by a large number of younger (= Democratic) registered voters, most of whom won’t turn out on election day. (Then again, maybe the younger voters will turn out at record rates this year…)Zogby’s methodology is one way by which you can correct the poll data to account for such trends, although it is not necessarily the best method nor the most accurate — particularly since changing party affiliation from Republican to Democratic could be occurring simultaneously with the disproportionate new registrations.So there is reason to continue to watch the Zogby poll data without dismissing it altogether.

  4. acf

    i just made a similar comment about pollsters’ party weighting over at TPM. It makes it impossible to compare polls from different pollsters, and without knowing the weighting, it is difficult to draw any rational conclusion about the published results. One thing that annoys me is the fact that I heard pollsters change their weighting for polls at different times during the campaign. It makes believing them a crapshoot, at best.

  5. Dan Kennedy

    Michael: I was perhaps too glib. I should have pointed out that Zogby obviously believes his is the best, most accurate method to use. But the possible flaws are clear.

  6. O'Reilly

    Dan.Do you have an interest in commenting or posting a thread on this topic?Link to U-Hub comment.

  7. O-FISH-L

    I like the Nickeloden kids poll that finished with Obama 51% and McCain 49% with something like 2.2 million presumably child voters. What are the kids hearing at home that their parents are afraid to tell the pollsters for fear of being labeled racist?Not sure if the kids on-line ballot had pics beside the candidates names, but in my community the Kids Vote ballot does have pics and the kids (K-5) overwhelmingly choose the younger candidate in every race, no matter how much of a longshot and even if the older candidate is a can’t lose incumbent. By huge margins, kids select the youngest person in the pictures. Now if this ballot had pics of McCain at age 72 and Obama at 47, and McCain still came within a whisker, that’s a hoot!Incidentally, Kerry won the same poll, going away, in 2004.

  8. Brad

    What amazes me concerning almost any of these polls is that, at this late date and in a race so clearly delineated, the pollsters are still able to find 5% or 10% “undecided” voters.“Facts are stubborn things, but statistics are more pliable.”Still, think of how many people still think Saddam Hussein had WMD, or was directly involved with 9/11…yeah, quite a few reeeeeeally dumb people.I used to say “ignorant”, meaning perhaps they were just misinformed. As I’ve gotten older, I’ve realized it’s really WILLFUL ignorance; the stubborn clinging to misinformation despite substantial evidence to the contrary. That speaks to a closed and narrow mind, which is the same as being a goddamn idiot.With that in mind, only 5-10% of the US population being so freakin’ DUMB that they can’t tell the difference between Obama and McBush? I start to wonder if that’s too LOW.

  9. mike_b1

    Not to be rude, o-fish, but you’re off the deep end. Did you even visit Nick to see how the poll was done? The methodology couldn’t have been worse had they conducted it in McCain’s War Room.

  10. Peter Porcupine

    “With that in mind, only 5-10% of the US population being so freakin’ DUMB that they can’t tell the difference between Obama and McBush? I start to wonder if that’s too LOW.”DK – and you wonder why people say they are undecided?The hysterical venom in this race, beginning with the buyers remorse that set in when Obama failed to have a 20-point lead from the outset, will skew polls hugely. NOBODY is answering accurately, afraid of being ‘push-then-bludgeoned’ polled.Like you, I think it will be close. I also think that polling may have reached the end of its usefulnesss in predicting the outcome of races.

Powered by WordPress & Theme by Anders Norén