Who to believe?

Written By: - Date published: 11:41 am, August 31st, 2011 - 38 comments
Categories: polls - Tags: ,

38 comments on “Who to believe? ”

  • Comments are now closed

  1. ianmac 1
    31 August 2011 at 11:54 am

    Wow. Perhaps someone like Hooten would argue that Fairfax was much more reliable and that the 18.6% was just a margin of error or a misprint or an indication that Phil Goff is being challenged by ummm….

  2. Josip Blow 2
    31 August 2011 at 11:55 am

    Herald poll surveyed 750 , Fairfax around 1000, both with margin of error of 3%.
    Now come to the the rescue if your statistically minded, but looking at Auckland only i guess no more than a third of those surveyed would be from Auckland would that make the margin of error 9% ?, which sort of bridges half the gap. Be a good question for the respective pollign co’s to attempt to explain.

    • wtl 2.1
      31 August 2011 at 12:28 pm

      I did a two proportion z-test between two results, assuming that the samples sizes for Auckland were 1/3 of the total (i.e. 250 for Herald, 333 for Fairfax), and the resulting Z score was about 4.9. The p value for this difference is ~4.79e-7. In other words, it is very very unlikely that the difference can be accounted for by the sampling error of the polls.

  3. Craig Glen Eden 3
    31 August 2011 at 11:57 am

    Yeah one has to start calling in to question these polls I mean seriously is polling a tool for the future?
    And if it is what shape will it take to give a more accurate story.

  4. Ari 4
    31 August 2011 at 12:00 pm

    Polling has some accuracy when you’re taking a nationwide picture and focusing mostly on the two main parties.

    When you start getting into regional concerns or trying to estimate parties under 10%, the only useful thing you can measure is the trends- the actual numbers have such a high ratio to inaccuracy that it’s ridiculous to speculate on them.

    Most likely one or both of the numbers is/are rogue.

  5. mickysavage 5
    31 August 2011 at 12:07 pm

    Labour polled from memory 36% in Auckland last election.
     
    My very strong impression from spending hours and hours doorknocking and talking to people is that there is a return of support to Labour from those persuaded last time to vote for Smile and Wave.
     
    The three Auckland by election results are also evidence of this.  On the polling Labour should have lost Mt Albert and have been trounced in Botany and Te Tai Tokerau.  There was a swing to Labour in each.
     
    These bloody polls just provide diversions and weapons.  Note every right winger, Hoots, Farrar, the slimy one et al are talking about this latest poll as if the results are carved in stone.
     
    For them these polls are weapons to bash the left with no matter how dodgy their figures are.

  6. kriswgtn 6
    31 August 2011 at 12:16 pm

    cos of continually delivering thjis crap ,read it enough times and hearing it enough times will hopefully make some left voters not bother going out to vote which boils down to
    1= The fascist right are worried…..

    I say keep up the good work-not a supporter of Goff- im a supporter of policies that benefit all not a select few

    So keep the lies/bullshit the right are promoting going out on Twitter facebook-I index alot of these stories on my facebook page and joined a few groups etc etc

    The word is out there regarding these tories dont you worry and its working

  7. Danyl Mclauchlan 7
    31 August 2011 at 12:21 pm

    Maybe the polls are completely wrong. But we went through this all last election, with Labour supporters insisting that the polls were lying, that they didn’t sample cell-phones and that their sense from door-knocking was that things were going really well. And they turned out to be completely wrong and the aggregates of the main polls were a really accurate predictor of the final outcome.

    http://en.wikipedia.org/wiki/Opinion_polling_for_the_New_Zealand_general_election,_2008

    • McFlock 7.1
      31 August 2011 at 12:37 pm

      Not 3 months out, though.
       
      And my position is still that National will probably take a bump below 50% in the actual campaign, and if they do then they are pretty much fresh out of friends – a few 1-3 MP coalition allies, creating not so much a “five headed monster” as a “four-pimpled teenager”. If National drop to the mid-40%, they might well not be able to form a government. Then the pressure will be on the Greens as to whether they want to support Labour or National.

      • Lanthanide 7.1.1
        31 August 2011 at 1:01 pm

        If they have a choice of realistically creating a Labour government (eg one that isn’t Labour + Greens + NZ First + Mana + Maori) then they’ll go that way.

    • wtl 7.2
      31 August 2011 at 12:48 pm

      See my comment above (2.1). Perhaps the aggregate of the polls will be correct this time as well. Averaging polls should certainly give a better picture than one poll. But the inconsistency between polls is telling us that there is something very wrong with their methodology (in at least some polls) – we are in NOT looking at random error introduced by sampling error. Therefore, we have no idea of the true ‘error’ associated with the polls and cannot be sure that there is no systematic bias.

      On that basis, we seriously have to be questioning their results and stop thinking they can give us any indication of the level of support for various parties beyond a rough guide. That is, I have no doubt that the support for Labour is ‘low’ at this time and support for National is ‘high’. But exactly how ‘low’ and how ‘high’ is anyone’s guess. Perhaps we should just forget the polls, go through the election campaign, and see what happens on Nov 26?

    • Blighty 7.3
      31 August 2011 at 12:57 pm

      I don’t think you have to be a head in the sand ‘the polls are all wrong’ type to acknowledge that something is seriously wrong with one of those polls.

    • Sanctuary 7.4
      31 August 2011 at 2:08 pm

      It is not that the polls are wrong – it is now a question of which one is right.

      As far as I know, the last time round the polling companies held up how they were all in agreement with each other as evidence of their accuracy. Now we are seeing wild variations that are not just at the edges of the margin of error but well outside them. This Fairfax poll in particular is so at variance that I can only think that it has been designed to serve an agenda – to panic Labour voters and deliver another “2002” style result, only in reverse.

      three years is a long time in technology, especially when incomes at the bottom have been falling and squeezing things like the fixed line phone bill off the agenda. Personally, I regard these huge variations in results as indicating the polling methodology of Fairfax polls at least is now very suspect.

      • Vicky32 7.4.1
        31 August 2011 at 5:25 pm

        especially when incomes at the bottom have been falling and squeezing things like the fixed line phone bill off the agenda.

        What on earth just happened? I wrote a long answer to this about why it was reasonable to assume that low-income people would actually prefer a fixed-line phone – and it vanished! WTFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF?

        [There’s nothing pending or in the trash – it never made it to the server. See your comment on Open mike re connection “burp” – that will be it. r0b]

    • Tiger Mountain 7.5
      31 August 2011 at 3:52 pm

      @ Danyl: Both are possible. The “FearFacts” polls are bent (and wrong) and the ultimate election results somewhat mirror the poll predictions. You do not need to be a genius to work out how that could happen.

    • GINA 7.6
      4 September 2011 at 9:24 pm

      Apparantly in the labour strongholds in the last election turnout was very poor and you have wonder why.Was it dissatisfaction with Labour or was it that a lot of poor kiwis work saturdays to make ends meet and thought theres no point voting because the polls give it to National. Apparantly 200000 labour voters didn’t even turn up. So a large part of the result was not swinging labour voters but absent labour voters.

      In the 2008 election National only got 43% of the vote but becuase Winstone missed the 5% by a whisker everyone got more seats proportionally. Then act managed to get into parliament well below the 5% level and that threw the numbers. The last election was not a massive vitory for the National Party but a series of tacticle manouvers with the aid of media manipulated scandal by the NZ Herald and the police etc.

      Had those 200000 ( 7% of voters ) Labour voters had made it to the polls and Winstone not had bullshit police enquiries hanging over him and got his 5% the election results would have been a very differrent story, possibly with a hung parliament.

      Remember the 2004 was being called for National and they lost didn’t they. Cellphone only households amoungst the poor have grown quite a bit since then.

      These polls that are swinging wildly between polling companies and obviously must have a far greater margin of error than the 3% claimed.

  8. monty 8
    31 August 2011 at 12:44 pm

    Amongst my friends the Fairfax poll would have to be optimistic for Labour and the Herald poll fantasy

    • Monty 8.1
      31 August 2011 at 10:23 pm

      piss off fake Monty, find your own name to post under.

      And for what it is worth, I would guess that support for Labour in Auckland is somewhere between the two polls , and no matter what Labour have no hope this 2011 election.

  9. Puddleglum 9
    31 August 2011 at 12:50 pm

    Same conclusion as wtl but I went about it slightly differently:

    If the Auckland sample in the Fairfax poll is approximately one third of the sample of people who expressed a party vote preference (i.e., 1000 – 16.2% of 1000 (i.e., the ‘undecideds’) = 838) then that gives a sample size for Auckland of about 279 (i.e., 838/3).

    On an expected binomial percentage of, 20% support Labour and 80% don’t, then the Estimated Sampling Error table I’m looking at suggests that, for that sample size of between 200 and 300, the sampling error should be somewhere between 5.7 and 4.6% %. (If our estimate of Labour support is 50%, the error would be between 7.1 and 5.8%).

    For the Herald digipoll, given that there would be approximately a 250 sample from Auckland (I forget whether or not the ‘undecideds’ were included in the 750 sample reported) then the sampling error, based on a 40%/60% estimate (i.e., the 38% found in the poll), would be within the following sampling error bounds – between 6.9 and 5.7%. (Ditto for the other estimates, as above).

    So, if we assume that we can ignore the finite population correction factor (which we can – Auckland’s a big city), and also assume that the Herald poll was – at the 95% confidence limit – on the upside (i.e., 5.7% more than it ‘should’ be) and that the Fairfax poll was an underestimate at the lower edge of the 95% confidence level (i.e., 6.9% less than it ‘should’ be) we could explain 12.6% of the gap between the polls.

    A more likely explanation is that one or both of the polls are systematically flawed in ways that compromise their results. 

    Then again, my stats skills aren’t what they used to be. Happy to be corrected.

  10. Tom Gould 10
    31 August 2011 at 12:53 pm

    A local National Party apparatchik has explained to me the Fairfax poll giving labour 20 percent support in Auckland is the accurate poll because the Herald is getting more and more left wing since Len Brown won the mayorality and so are their polls. Complete nonsense, but as good an explanation as any I guess.

  11. side show bob 11
    31 August 2011 at 1:03 pm

    Don’t believe either poll, Labour’s support is way lower then that.

  12. the pink postman 12
    31 August 2011 at 1:24 pm

    Well if gallop-polls are so accurate why have a general,election. Have a gallop-poll and save thousands of dollars., Of course it would mean the Nats would win every time ,But that could cause a revolution which might be the answer.

  13. tsmithfield 13
    31 August 2011 at 1:47 pm

    So, this would be the Herald-Digipoll that the Herald reported as:

    Three months before the election, the Labour Party’s support has dropped again in the Herald-DigiPoll survey to its lowest this term.

    Inspiring stuff.

  14. Carol 14
    31 August 2011 at 2:35 pm

    The big issue for me is that this is another example of the MSM focusing more on the game,, polls, strategies and presidential style leadership comparisons, than on discussing polices & party teams.

    • Carol 14.1
      31 August 2011 at 6:10 pm

      Oh, thanks, Ben. I’ve had an accident & will be typing with my left-hand only for a while – and I’m right handed, so it’s a slow business.

  15. felix 15
    31 August 2011 at 4:42 pm

    If Labour get their shit together and start working as a team right now we’ll get rid of these tory pricks in November.

    If they don’t, we wont.

    And that’s my 2c on what the polls are worth.

    edit: Carol, how many times have I told you to put your tags away when you’ve finished playing with them? 😉

    • Carol 15.1
      31 August 2011 at 5:37 pm

      Yes, I know, Felix. I didn’t check immediately after I posted, then it was too late.

      • wtl 15.1.1
        31 August 2011 at 5:38 pm

        Someone really needs to put an end to all these italics. Did that help? Edit: Nope. We need a moderator.
        [Ben: fixed]

  16. Georgy 16
    31 August 2011 at 8:11 pm

    The polls simply cannot be trusted. While they may have the total number required to reflect a snapshot of the nation, what we don’t know is who that number actually is. Is it random? If a third of the population do not have landlines [as many schools will indicate] then can the total in the poll be truly reflective of the wider community?

  17. Brett Dale 17
    31 August 2011 at 9:05 pm

    Big difference, these polls are about as accurate as the ones that says the greens are at 11%

  18. Puddleglum 18
    31 August 2011 at 9:37 pm

    Here’s Keith Rankin on the importance of the ‘undecideds’ in polling and how polls are often misleadingly represented in the media. (Press ‘2’ on the video to go directly to Rankin’s interview).

  19. HC 19
    1 September 2011 at 12:58 am

    Before the polls are conducted, the water supplying companies all over NZ get “secret instructions” to immediately put an undisclosed substance into the water supplies. It comes right from the top. The substance is alleged to have drug like effects, causing people to fall into a relaxed, light-hearted, trusting and fulfilled mood, leading them to feel unconcerned about everything and forgetting most of their worries and problems. The substance is supposed to contain a mixture of plasma ocytocin and prolactin.

    Hence the polled group of people give the typicla kind of answers that are representative of the poll results we get presented by the media.

    There is a corresponding effort made in most TV, radio, print and other media to emphasize the desired positive and negative aspects in political debate, planning and decisionmaking.

    Some suspect that there may be involvement by an overseas based advisory agency, which has successfully led other politicians and parties to decisive victories.

    Don’t worry, be happy, it is all in good hands!!!???

    • Jim Nald 19.1
      1 September 2011 at 7:23 am

      The invisible hand that is putting us at the front of the queue for some good screwing. Yippee!

    • Jum 19.2
      1 September 2011 at 8:14 pm

      HC

      As usual, when you try to tell people the truth nobody listens…

      • HC 19.2.1
        2 September 2011 at 2:54 am

        Jum: Truth is by mistaken persons or wrongdoers associated with IMMENSE DISCOMFORT, requiring rethinking, changes of mind, realisations, wake up scenarios, facing reality and making amends for previous wrong decisions and actions. So it is natural that few will endeavour to do this. It is also the dictator’s problem, because facing reality is associated with becoming honest an facing justice, which may in return pay a very harsh and serious price for the wrongdoings committed before. Who volunteers for this?

  20. Bored 20
    1 September 2011 at 8:03 am

    If South Auckland, West Auckland, PTown, Aranui and the rest of the Poly / TW / Westie suburbs are feeling the pinch of hard times on election day no amount of “polling propaganda” is going to help National.