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

Patrick by the numbers

Jon Keller weighs in with a neat bit of media criticism, observing that today’s Boston Globe story on Gov. Deval Patrick’s relative popularity in Western Massachusetts is based on a poll of just 70 people — “a sample so small, it’s meaningless,” as Keller notes.

Previous

Meeting my match

Next

From the Eliot to J.J. Foley’s

6 Comments

  1. Aaron Read

    IS it actually meaningless? I've seen Arbitron surveys based on samples not much bigger…and a heckuva lot of money rides on those.I've heard it said that "Statistics is the fine art of taking the unquantifiable and trying to quantify it." But the math behind statistic sampling is both complex and usually quite justifiable. If you have valid assumptions about your demographic you're sampling, you can get a representative sample from a very surprisingly small number of people.That said, on the face of it…70 people seems rather small. Even for relatively-sparsely-populated western Mass.Sooooooo…it begs the question: why is the Globe running a puff piece so desperate to make a point that it's using fishy numbers? If Morrissey Blvd is hoping for a bailout, it ain't coming from Patrick.

  2. bostonmediawatch

    From Keller, quoting North Adams State Rep. Dan Bosley:"there’s not as much vitriol in the press. And there’s no talk shows." Translation: ignorance is bliss."Translation: Bosley and/or Keller thinks people need scandal sheets and talk shows to tell them how their life is going.I guess the poll was in yesterday's paper; there's nothing in the online version that says how the poll was conducted or what the margin of error is.To simply dismiss a sample size of 70 as meaningless without mentioning some fault with the methodology is ridiculous.It's unfortunate that WBZ markets Keller as a snark master – they could use a straight up political reporter with his chops ala John Henning.

  3. mike_b1

    bostonmediawatch is dead-on: To simply dismiss a sample size of 70 as meaningless without mentioning some fault with the methodology is ridiculous.A poll conducted with half that number of subjects could return a very high degree of confidence if properly designed (and executed).Keller may be correct in his analysis of Patrick's popularity, but if he is, it's not because he understands the poll.

  4. Lance

    Keller's bigger problem is that he argues that high unemployment in Leominster/Fitchburg/Gardner and Athol render the result of the poll meaningless. Of course, Leominster, Fitchburg, Garnder and Athol are all in Worcester County and are therefore in Central Mass. In the poll CMass gives Patrick his highest unfavorables (60%). Keller's point that an accurate poll would give the guv bad numbers was correct. It did. It is Keller's geography that is wrong, not the results of the poll.

  5. Steve

    Aaron – is it actually meaningless? Assuming it's a random sample, I calculate the traditionally cited "margin of error" in a poll with 70 respondents to be plus or minus 12%.So it's not exactly meaningless. More like meaning-not-so-much.

  6. Dan Kennedy

    Lance: I subscribe to the old Cleveland Amory theory — that anything west of Dedham is The West.

Powered by WordPress & Theme by Anders Norén