Skip to main content

View Diary: Polls That Exclude Cell-Phone-Only Homes Have Growing GOP Bias (48 comments)

Comment Preferences

  •  People have been going on and on about this for a (1+ / 0-)
    Recommended by:

    at least three years now. It hasn't been a problem for the polls in 2008, 2009 or 2010 so far. No reason to assume it would suddenly pop up in November.

    Support Dennis McDonald and Montana Democrats in the 2010 election!

    by twohundertseventy on Thu Oct 14, 2010 at 05:48:17 AM PDT

    •  au contraire. . . (4+ / 0-)

      Actually, the polls are getting consistantly more wrong. Look at just this election cycle. From Brown to O'Donnell, polling consistently missed the trend that resulted in the end result, until nearly the last minute. You can bet that Castle and That Woman in Mass would have loved to have accurate polling right along.

      Even above that, the polling is even more skewed when one considers that it has become an 'almost only the old who vote' sample. Pollsters compensate by formulas based on reasonable models of how the results would change if they could actually contact the cohorts that cannot, but that is all guess work, and clients do not like it. They want real numbers from real people.

      Polling has become a substitute for journalism in elections, and this fact is going to bit a lot of people in the ass, someday. That day may be the first Tuesday or this coming November

      •  No, but polls can't PREDICT things, they're just (1+ / 0-)
        Recommended by:

        a snapshot in time. Brown and O'Donnell surged late, and the polling showed exactly that.
        The polls in neither race were off much, and if you extrapolate the surge they showed from the field dates of the last polls until election day they were exactly spot on.

        I'd bet that if the DE-SEN elected had been suddenly held a month earlier with just a day prior notice, Castle would have won 70-30. But O'Donnell had late momentum, and the polls reflected that, showing her gaining rapidly.

        Support Dennis McDonald and Montana Democrats in the 2010 election!

        by twohundertseventy on Thu Oct 14, 2010 at 06:12:02 AM PDT

        [ Parent ]

    •  Growth in cell-only over that time frame (1+ / 0-)
      Recommended by:

      From the first half of 2006 to the last half of 2009, cell-only households went from 10.5% to 24.5%.  Extrapolating on that probably means about 28% now. In addition, the cell-only population is aging a bit. Turnout percentage increases with age group, so the missed segment is becoming more significant.

      There is no meaningful difference politically within a demographic cohort - gender, ethnicity, age group, region - when looking at phone status. But the cohorts where the Dems are strong are more likely to be cohorts with high cell-only percentages. I don't think properly weighted polling is off by that much, but the issue is starting to undermine established methodology.

      •  Well, it SHOULD undermine established (0+ / 0-)

        methodology. Except for it doesn't.
        I agree with you theoretically. Cell-phone demographics skew Democratic, there was an extensive Pew study on that. Excluding them SHOULD result in a GOP bias. Yet, polls at large, averaged together, don't exhibit much of a bias either way, and in 2008 most cell-phone pollsters were way too Dem-friendly and showed mostly a D+9-10 race, if I recall correctly, while landline-only pollsters had the race right at D+6-7 where it ended up being.

        Support Dennis McDonald and Montana Democrats in the 2010 election!

        by twohundertseventy on Thu Oct 14, 2010 at 06:14:37 AM PDT

        [ Parent ]

        •  Weighting (0+ / 0-)

          Pollsters are having to rely more on weighting to correct for this issue, so their established methodology is being tweaked. The more variables that must be massaged, the less likely the results are correct. Throwing in the cell-only variable into an election with atypical turnout by age made 2008 a tough one.

        •  It won't until it does (0+ / 0-)

          I mean, it's one of those things that isn't true until, all of a sudden, it is true.  Eventually, the discrepancy will be large enough that it's difficult to adjust for and impossible to write off.  

          So I don't know whether that would be this election cycle, the next, or whatever.  

          I do know that it wasn't true in 2004!

          •  Eh. No, there's no reason to believe that the (0+ / 0-)

            discrepancy would show up all of a sudden. Why would it? The increase of the cell-phone-only population is a gradual process, so you'd expect the cell-phone-effect in polling to increase gradually as well. Except for-- there is none so far at all. Despite >25% of the population being cell-phone only.

            IMO low response rates among landline users are a greater problem than cell-phone users, but pollsters have been able to deal with that well.

            Consider this: I'm pretty sure that the proportion of cell-phone-only users has been greater in the Democratic Primary electorates of 2010 than it will be in the General Election in November, because the Democratic primary electorate skews toward cell-phone-only demographics.

            Yet, pollsters didn't do worse in these primaries than they usually do.

            Support Dennis McDonald and Montana Democrats in the 2010 election!

            by twohundertseventy on Thu Oct 14, 2010 at 07:06:25 AM PDT

            [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site