Distinguishing rogues

The line only one poll counts is right but it doesn’t mean that those interested in politics don’t take notice of opinion polls between elections.

However, there are reliable polls and there are rogue ones and Kiwiblog has a very goodguide on how to distinguish between them.

It’s something the Sunday Star Times ought to have applied when reporting on the latest Horizon poll and to be kept in mind by anyone reading:

Two-thirds of voters polled last month by Horizon Research believe National will be able to govern after the election, expected to be held in November.

But when asked to reveal their personal voting intentions, the result makes the election too close to call. National is by far the single most popular party, with 40.4% of the 1718 people polled saying they will give it their party vote, compared to just 28.3% for Labour. But the poll also predicts 8.9% for both the Greens and New Zealand First, 2% for Act, 1.7% for the Maori Party, and 1.2% for United Future.

When those parties are taken into account, a Labour-Greens-New Zealand First alliance (46.1%) would be a whisker ahead of a National-Act-Maori Party-United Future coalition (45.3%).

In every other poll in the past year National is around, and often above 50%, Labour rarely gets above 33% and New Zealand First, thankfully is safely under 5%.

It isn’t impossible, but highly unlikely, that National will continue to keep enough support to govern alone as the election approaches. It’s not impossible that Labour could lead the next government and that NZ First would be part of it.  

But as Whowouddathort shows this poll is so inconsistent with others it can’t be taken seriously.

26 Responses to Distinguishing rogues

  1. Cheers for the link to my post.

    Enjoy reading your posts.

    The lefties are going on about this poll like it is a serious one. Shows how desperate they are for any hope to hold on to.

    Like

  2. robertguyton says:

    Latest poll sends National packing!

    Rightwingers rally to ‘3-monkey’ the results.

    The Greens have certainly captured the public mood and they’re on the rise!
    Watch out Key’s mob!
    The tide! The tide!

    🙂

    Like

  3. http://whowuddathort.blogspot.com/2011/01/sst-goes-rogue.html

    RG, I don’t think that National are unbeatable.

    But I certainly don’t trust polls so far from other respected polls.

    If you look at the breakdown of their numbers they make for interesting reading. Supposedly over 11% who voted progressive last time are now voting ACT.

    Also, if you work out what 40.4% of the total people who answered National in this rogue poll, then take the “32.7%” of that number who said that last time they voted ACT and you find that they make up 13% of the sample size.

    But did ACT get 13% of the party vote at the 2008 election? I think not.

    Rogue poll.

    Like

  4. david winter says:

    This poll might well be biased, but it’s no good comparing it to other polls because, as the SST article says, it explicitly uses a different method.

    Most NZ pollsters just throw out undecideds and presume that will break in the same way as the rest of the populaton. There is no particular reason to belive this is true. On the other hand, Horizon asks them to pick one side and, if they still can’t decide, reports the undecideds (which discounts all the other totals for obvious reasons). It would be helpful is Horizon provided numbers that are comparable to the other pollsters (ie, preferences from before they tried to force undecideds to pick) but even without that we know that the Nats have 43% of decidededs and those with a preference, which isn’t so far from other pollsters…

    Oh, and whowuddathort, you are misreading those tables. They actually say 11% of ppl who voted for the progs last time (about 2 ppl) said they’d vote act this time . Similarly, that 32.7% is the number of Nat voters who voted for ACT last time. The total number of respondents who say they voted ACT last time is 32, which is about 1.8%. ACT got about 1.5% in the lat election.

    Like

  5. Actually David I think you may have misread the data.

    it asks: Who would you vote for THIS time. Then it says who did you vote for LAST time.

    ACT got 3.65% in the last election.

    Like

  6. david winter says:

    I know it’s counter-intuitive. But look at the table and ask yourself, do the rows of the columns add up to 100?

    Like

  7. david, if i take your logic from the numbers then the following results come out:

    National: 41.3%

    Labour: 32.9%

    Greens: 6.4%

    NZF: 3.78%

    ACT: 1.8%

    Maori: 2.03

    Like

  8. david winter says:

    Right, but you have to divide each by 0.915 since about 8.5% of respondents didn’t vote in the last election. I suspect this panel has actually been chosen to be representative in this regard…

    Like

  9. divide what by 0.915 David?

    Like

  10. david winter says:

    All the numbers you listed, since only 91.5% of the sample voted in the last election.

    Like

  11. so if i have understood you correctly i need to divide for instance Nationals 41.3% by 0.915 etc etc?

    If that is the case then the numbers go as follows:

    National: 45.13%

    Labour: 35.95%

    Greens: 6.99%

    NZF: 4.13%

    ACT: 1.96%

    Maori: 2.21%

    Which still doesnt work out with your theory?

    Like

  12. david winter says:

    I’m not sure I understand what you are talking about.

    Aren’t those numbers very very close to the 2008 election result? Doesn’t that suggest this panel is representative with regards that variable?

    Like

  13. Nope. Here are the 2008 election results:

    National 44.93%

    Labour 33.99%

    Greens 6.72%

    New Zealand First 4.07%

    ACT 3.65%

    Maori 2.39%

    Sorry David. I may have lost the point that you were making orignally and the reason why you mentioned the 0.915 in the first place.

    Anyway, Look forward to further banter with you over the coming year. Check out my blog if you want:

    http://whowuddathort.blogspot.com/

    Like

  14. gravedodger says:

    And we assume that the respondents are honestly answering the questions, have a landline connection , are home at the chosen moment, will actually get off their collective gluteus maximus and are not obtuse bastards like me and deliberately answer opposite to their beliefs.
    Farrar’s analysis on polls was enlightening for Moi for sure.

    Like

  15. robertguyton says:

    GD – “And we assume that the respondents are honestly answering the questions”.

    So all political polls are suspect?
    It’s as I suspected.

    Like

  16. gravedodger says:

    Meanwhile the strains of land of hope and glory accompany the glorious announcement of the Horizon poll result as the saving of the nation from those dastardly Nats.

    Like

  17. homepaddock says:

    Whowuddathort – if you’re wondering why your 5.16 comment has only just appeared it’s because it got directed to spam and I only just found it.

    It would have gone there because of the link.

    Robert@ 8.16 – when pre-election polls are compared with election results some are usually reasonably accurate, but no-one knows ahead of time which ones will be and it would take someone with a better grasp of polling and stats than I have to tell you if that’s just a matter of luck.

    GD @ 8:04- I wouldn’t have picked you for a contraian 🙂

    Like

  18. Home, No worries. I assumed it would make it eventually. 🙂

    Like

  19. adam2314 says:

    ” no-one knows ahead of time which ones will be and it would take someone with a better grasp of polling and stats than I have to tell you if that’s just a matter of luck. “..

    HP. With the rising tide of Sub-Continent voters, how can you be so sure that ” no one knows ” ??..

    Sounds like a Pakistan Cricket Score to me ..

    What ever happened to those local body election double, double, double, double, dippers ???

    Like

  20. david winter says:

    I know this is a whole day later, and in internet terms now ancient history, but…

    whowuddathort,

    The point was one of the reason you cited for dismissing this poll was that it massively over-represented people who voted ACT in 2008:

    Also, if you work out what 40.4% of the total people who answered National in this rogue poll, then take the “32.7%” of that number who said that last time they voted ACT and you find that they make up 13% of the sample size.

    In fact, the poll slightly under-represents 2008 ACT voters but over all it’s closer to being representative of that election than I imagine you’d get by chance. So that particular criticism doesn’t hold much weight.

    The deal about dividing by 0.915 was because the numbers you calculated your percentages from were with respect to the total sample not just from those who actually voted last time (only 91.5% of the sample).

    There are plenty of reasons to be cautious about the results of this poll, but, as with most things, both sides of the debate seem to react more strongly to the result than the underlying method.

    Like

  21. cheers david,

    get what ya mean now. Some good points there.

    Like

  22. No clothes on says:

    robertguyton says:
    January 9, 2011 at 1:34 pm

    Latest poll sends National packing!

    Rightwingers rally to ’3-monkey’ the results.

    The Greens have certainly captured the public mood and they’re on the rise!
    Watch out Key’s mob!
    The tide! The tide!

    Looking at the political compass the Greens position on civil liberties is an outlier. If it were true the the Greens have captured the public mood, people in some neighborhoods wouldn’t be talking of “cowering in their homes at night”*.

    *where working class and middle class meets riff-raff.

    Like

  23. robertguyton says:

    People in some neighbourhoods are cowering in their homes at night?
    Has Law and Order not improved?

    Like

  24. AB says:

    Good lord, when are people going to read the Horizon polling correctly?

    It’s results cannot be directly compared with other recent polls.

    Because it says it is taking not only the decided vote, but the preferences of the 14% or so who are undecided, then filtering that again by intention to actually vote.

    What value is there is all these other polls which give you results for only the 69% or so who have made up their minds, when the outcome rests on thos who have not – and whether or not people turn out?

    Horizon is frighteningly accurate, when comparing the main parties’ votes at 2008 with today.

    National got 41.3% in 2008, Horizon says it has 40.4% now (with a 2.2% margin of error). They should be very pleased with that, given the amount of water running under the biridge, and can probably thank John Key for that. Labour got 32.9% of votes cast, Horizon has them constantly falling and at 28.3%.

    No-one’s disputing its results on how Kiwis feel about political leaders. I see Key is well ahead in imparting hope http://www.horizonpoll.co.nz/News

    It’s time the commentators comapred apples with apples in these poll results, and well past time the media (and others) started published results for ALL respondents to their surveys. Otherwise people are debating the margin of error when considering the all-important minor parties’ vote shares.

    As for some pollsters suggesting new entrants to the business shouldn’t be allowed, where would that have left YouGov in the UK, now one of the biggest – and most accurate – online pollsters?

    And why don’t they follow Horizon’s methodology in determinig who’s most likely to vote and for whom, as best-practice pollsters do in the same way in the USA?

    And when are the other pollsters here going to be as open as Horizon about their polling methodology – including whether or not they weight on 2008 party vote which one would have thought quite essential to getting a represetnative sample in a political poll?

    The detailed HorizonPoll results on the “potential vote” I found here:
    http://www.horizonpoll.co.nz/page/89/potential-vote-detailed-results

    Like

  25. AB

    to corret your stats:

    National got 44.93% at the last election, NOT 41.3% as you say and Labour got 33.99 NOT 32.9 as you say.

    Get your point about comparing polls. But it clearly doesn’t relect anything.

    If you look at other polls, they were more accurate in predicting the election result.

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: