False dawn

The first heartening opinion poll to emerge for the Coalition during the Queensland election campaign turns out to be not worth the paper it’s printed on. The Herald Sun reports the Roy Morgan findings under the heading "Beattie approval rating plummets", and asserts that the figure in question has "slumped by 13 percentage points". It then transpires that this is based on a feeble sample of 268. No detailed breakdown of voting intention figures is provided, but the Roy Morgan site gives a two-party preferred result from the sample of 52.5-47.5 in Labor’s favour.

Author: William Bowe

William Bowe is a Perth-based election analyst and occasional teacher of political science. His blog, The Poll Bludger, has existed in one form or another since 2004, and is one of the most heavily trafficked websites on Australian politics.

17 comments on “False dawn”

  1. Given that Morgan has always scorned telephone polls(sometimes refering to the competition as”just telephone polls) why did he make some extravagant claims on just such a telephone. He has said that Beattie will do much less well than expected..we’ll we’ll soon know!
    He also makes similar claims a about Victoria ,on a small phone poll sample ?

  2. Morgan has a habit of generalising to a population from statistically meaningless samples. It really is wrong for him to do it. Here this item http://www.roymorgan.com/news/polls/2006/4070/ from a few weeks ago making claims about how Tasmanians feel about state governments fighting corruption. This is from a sample of 50, which Morgan does not with an asterisk, but he shouldn’t be drawing conclusions in the commentary. The Tasmanian media, of course, ran headlines on the story, and opposition parties made a meal of it. No mention of the sample size, of course. Then I heard Gary Morgan giving a radio interview where linked the Tasmania result of his poll with a recent ministerial scandal.

    All this based on a meaningless survey. Journalists don’t know better, but Gary Morgan does.

  3. The Morgan Polls haven’t been credible since the infamous survey just before the 2001 election that predicted a Beazley landslide. I’d take this latest effort with a grain of salt(and why is it being reported in a Melbourne newspaper, of all places?).

  4. Health is always a big issue, because people can see themselves being sick. Voters think the Queensland opposition can do a better job on health. Regardless of how Morgan reached this figure I think it will be closer to the final outcome than some of the other polls. If it is I suppose it will be blamed on a ‘protest vote’ but this will be meaningless.

  5. I very much doubt voters think the Qld opposition can do a better job on health. Based on the polling Graham Young and I have been doing, which is discussed on Currumbin2Cook, most voters accept that health is a complex area not amenable to quick fixes, and are sceptical of the quick fixes on offer from both sides. But the competence issue works very much against the Libs – based largely on Flegg’s disastrous campaign.


  6. Morgan is playing the publicity game. What better way to get attention for your polling company, particularly from essentially a conservative press, than by predicting against the accepted opinion, that being that Labor will be returned comfortably. To do this of course you need to poll fewer people and provide as little information about the poll as possible. As Piers Ackerman said, pollsters believe that around 1000 people or more is needed to get a fair sample of the opinions of the population. If Morgan was being more responsible and doing his job correctly and getting the same results as other polling organisations, he would not be rating a mention in the media. Most papers and radio ignore him for most part.

  7. If Morgan polled 268 – and the article admits 10% were undecided so he only has 244 responses – the MoE is +/- 6.5%. That suggests Beattie is only half pulling our legs when he says there’s a discernible swing against Labor. Does the chance of a ‘rogue’ poll increase with small sample size?

  8. I agree with Gary Bruce that it seems to be a good way for Morgan to get free advertising…

    When I follow the link to Roy Morgan, I find at the bottom of the introduction:

    “Voting Intention
    The Morgan Poll conducted on August 29/30/31 with a sample of 165 Queensland electors found the ALP with a 5 percentage point lead on a two-party preferred basis — ALP 52.5%, L-NP 47.5% (published in Saturday’s Melbourne Age [September 2]). This is despite 73% of those surveyed who think Labor will win next Saturday compared with 13% the Coalition and 14% undecided.

    More details are available on a subscription basis.”

    Morgan seems to be saying further down the report that Beattie’s approval rating drop is in some way linked to Wayne Goss’ rating drop in 1995-6. I think he is trying to imply that there is indeed a ‘sleeper’ anti-Beattie vote out there.

    However in 1995, Queenslanders had a competent Opposition to vote for.

  9. I think the unknown 10% is still a worry for the ALP. I do agree however that the sample size is ridiculous, it is however no worse that the one the Courier Mail presented on Saturday. After polling 1000 voters across five seats it qoutes the National’s primary at 9% but fails to mention they are only contesting in one of the 5 seats polled. (ie 45% of the primary for that seat)

  10. “Does the chance of a ‘rogue’ poll increase with small sample size?”

    yes it does purely because the more people you sample the less likely you are to get an uneven sample.

    ie if you flip a coin ten times its much more likely to get 6 heads than if you flick a coin 100 times and get 60 heads.

  11. If I have correctly totalled the Poll Bludger’s predictions he suggests that the Nationals will gain 2 (Bundaberg & Gympie), the Liberals gain 1 (Kawana, Mudgeeraba & Noosa, but forgoing Chatsworth & Redcliffe) … leaving a final tally of ALP 58, Ind-O.N. 5, Lib 8, Nat 18.

    I will with amateurish enthusiasm boldly predict that the Nationals will also pick up Hervey Bay, Toowoomba North, & Keppel. With more trepidation also that the Liberals will still manage to win Clayfield, Indooroopilly & Barron River, (despite themselves). Final tally… ALP 52, Ind-O.N. 5, Lib 11, Nat 21.

  12. Graeme,

    Here’s a half-informed layperson’s view. There’s a lot of subtlties in the correct use and description of probability, so forgive me if I put my foot in it.

    if you have a sample of 1000, and if that sample is representative of the population (that is, it’s a truly randomly selected sample), you can say that your results have a margin of error of +/- 3%, 19 times out of twenty. In other words, if you take twenty different representative samples and compare them to the result for the entire population, 19 of the 20 will be within 3% of the population’s result.

    The Confidence Interval and MOE are linked, so the 3% MOE with 95% Confidence is equivalent to about 4% at 99% or 2.6% at 90%.

    So, assuming two different polls are reported with the same confidence interval, and that both have taken a representative sample, then the chance of a ‘rogue poll’ is the same.


  13. looking at these sums it is possible that the alp wins
    Currumbin,gavan and retains Bundaberg
    this would be 61 alp which is where matters stand at the moment approx

Comments are closed.