Blue Mass Group

Reality-based commentary on politics.

Keller’s Poll is Skewed

By

Jon Keller just released a new poll showing some surprising results, but how accurate can it be? 

Nearly 40% of the poll sample has graduate degrees.  Thats not even close to being representative.  I haven’t fully examined the results yet, but at first glance it seems like there are a LOT of problesm with it.

Please share widely!
fb-share-icon
Tweet
0

Comments

  1. that might be a problem if the Grad School voters (25-53-18) had significantly altered the overall results, but the College and Some College results (30-47-20) almost exaclty line up with the Overall poll (29-45-21).

    <

    p>
    Good news for Patrick’s campaign, and not at all surprising seeing as its the first poll since before Deval Patrick started advertising.

  2. I never completely discount any poll. I think you need to try to decipher what it is telling you.  This one is a bit of outlier relative to all others that I have seen. Depending on the poll, everything I have seen recently has one of two orders depending upon the methodologies, weightings, and assumptions.

    <

    p>
    Either
    1.) Gabrieli
    2.) Patrick
    3.) Reilly

    <

    p>
    or

    <

    p>
    1.) Patrick
    2.) Gabrieli
    3.) Reilly

    <

    p>
    Since Thursday’s debate, the general trend has been Gabrieli up, Patrick up, Reilly down.  In the end, the composition of the electorate (% of independents vs. % of democrats) will be key.  Also, it’s not just turnout, but where is the turnout coming from as there are real pockets of strength for each candidate more pronounced than “normal” (e.g. Deval in Cambridge vs. Gabrieli in Quincy).

    <

    p>
    All of that said, I see a number of significant issues with this poll’s makeup.

    <

    p>
    60% of the poll is from Boston

    <

    p>
    40% of the people earn 80k or more

    <

    p>
    40% have PhD’s/graduate degrees – as noted earlier in above post

    <

    p>
    All of these factors would tend to favor the return for Deval based upon what I have seen and know of the broad demographic trends in this race. 

    <

    p>
    Deval is also up with elderly and Independents – not consistent with most other polls.  For example, Zogby’s horserace results with Healey were just released this week and showed the following.

    <

    p>
    Gabrieli 57.8%
    Healey 28.7%

    <

    p>
    Patrick 57.5%
    Healey 33.0%

    <

    p>
    Reilly 46.6%
    Healey 35.7%

    <

    p>
    So in the end, the WBZ poll is good news for Deval and his supporters just like Zogby and others that I am not yet at liberty to publicize paint a better picture for the Gabrieli Team.  I look forward to the only “poll” that matters on Tuesday night!

    • my point about the tilted distributions is that if you look at other education brackets or income brackets, Patrick is strong or in the lead: there’s a similar trend across the demographics, which I think tends to minimize the problems your talking about.

      <

      p>
      Also, the Zogby poll just indicates that Gabrieli and Patrick poll very well in 1-on-1 against Healey, which I think is the biggest problem of those Zogby polls: not including Mihos, only having 2-way races. But regardless, taken on the results, how could Zogby be significantly better for Gabrieli than Patrick?

      • Good Point Alex. 

        <

        p>
        My focus is also on the geography.  The geographic distribution of the poll “feels” way off and that is more critical than on most other occasions (i.e. it’s not enough to just be right on the demographics; the distribution down to the exact city/town distribution is also key and that is very hard to poll and sample with a significant sample size).  There are areas where Patrick’s support is off the chart.  There are others where his support is virtually non-existent.  The same goes for the other candidates.  The margins for one candidate or the other at the extreme local levels that I have observed are bigger than have typically been observed in 3 or more person races – perhaps indicative of the strength of each team identifying and “marketing” to their areas of strength.  This accounts for the fact I am staring at three poll – this one, one with Patrick up 3, and another with Gabrieli up 2.  My guess is that State House News will have Patrick up 5-12 based upon the pollster and their methods later today. I may be wrong, but that’s my guess.

        <

        p>
        With this being the case, correctly projecting the weighted turnout based upon region/city/town will result in the best poll.  The question is, who has the best projection?  I can tell you that Boston – while significant – will not account for 60% of the electorate next week.  More importantly, which areas of Boston are represented in polls are key.  If the weighting is off at the “town” level (e.g. Southie, West Roxbury, Charlestown, Roxbury, etc.) that can have a significant impact on the Boston results.  You need to get that right to get a good read on Boston. 

        <

        p>
        I am also intrigued by the results observed with independents and elderly.  They run counter to the others polls I have observed in terms of the magnitude (e.g. as we move “right” on the spectrum better for Gabs/Reilly, as we move left, better for Patrick, as we move up in age, better for Reilly, but not bad either for Gabrieli and Patrick). 

        • is it the City of, or “Greater” Boston?

          <

          p>
          given that the African-American sample is 6% of the entire poll, which is meager given the dynamics, looks like there are a few quirks as you get to the crosstabs.

          <

          p>
          But remember, most people this week saw:

          <

          p>
          The Reilly-Gabs joust at the debate, rather than the whole debate;

          <

          p>
          The Globe endorsement for Deval (Herald is more recent), and;

          <

          p>
          A pretty good, positive Deval ad down the stretch. (much better for swing voters than the previous one)

          <

          p>
          So how negative and for how long will the other guys go? (conventional wisdom says you end on a positive note)

  3. I think 45% is unrealistic. If it had him in the high 30s I think it would make sense, but remember SUSA has been polling a bit higher for Patrick than the other polls so far so I would expect to see numbers in the high 30’s for him in the other polls.

    <

    p>
    However note, not only has Patrick started his TV blitz since the last poll, but also the campaign has been in high gear. At least in our area we have been canvassing every weekend and phonebanking every night for the past several weeks. This may be paying off. Who knows even the Globe endorsement might have helped.

    <

    p>
    Either way it is good news for Patrick because it show a significant increase in his polling numbers since the last SUSA poll (which if I remember correctly had him at 34%).

  4. The  undecided % in the fast track poll is too low.  There will be more than 4% undecided when we wake up on the 19th (i.e. more than 4% will make up their minds that day, on the way to the booth, and/or at the booth on election day).

    <

    p>
    Most numbers that I have seen place the undecided percentage anywhere from 15% to 25% depending upon the source and methodology used.

  5. It’s a stratified random sample, designed to get at likely voters (not just registered voters, but voters likely to show up at the polls).  The sampling is designed to mirror the demographics of turnout in a Democratic primary.

    <

    p>
    If you recall past polls, Deval voters are more committed voters.  (I’m for Patrick and I am absolutely going to the polls to vote for him.) Gabrieli voters are less committed to their candidate, and less likely to show up at the polls.  (Yeah, I like Gabrieli, if I get to the polls on Tuesday I’ll probably vote for him.)

    <

    p>
    Gabrieli’s vote is soft, and it’s their campaign’s task to harden them – to keep them from switching in the next week, and to get them out to the polls.

    <

    p>
    Interesting.

  6. Question: Does this poll result mean that Gabrieli is going to go negative, especially with only a week left to make an impact?

  7. Reward passionate support, which I’ll concede Deval has more of. The main reason to wait on the line and answer those recorded questions is because you want your answers recorded in the results.

Archives