Jun 22, 2010

Marginal Seat Newspolling

Newspoll released today via The Oz a set of marginal seat polling that they undertook along with their usual nationwide poll over the period from last Friday to Sunday. It was based on

Possum Comitatus — Editor of Pollytics

Possum Comitatus

Editor of Pollytics

Newspoll released today via The Oz a set of marginal seat polling that they undertook along with their usual nationwide poll over the period from last Friday to Sunday. It was based on three separate samples where the Qld seats of Longman, Dawson and Flynn were one, the NSW north coast seat of Page was another and the western Sydney seat of Lindsay was the third sample. Each of these samples had 600 respondents, giving us a margin of error that maxes out around the 4% mark.

If we look at the results and compare it to the national poll that Newspoll undertook at the same time, this is what we get:


The first thing we need to do here is pay absolutely no attention to the Lindsay result, as it was polling most of the same people that were voting on the weekend in the NSW state byelection in the seat of Penrith! Whoever thought that was a smart idea needs a cup of tea, a Bex and a good lie down. This polluted the sample to the point where it isn’t even useful to treat the results with a little additional uncertainty – it goes to the heart of why federal governments go out of their way to avoid holding elections around the same time that a state election is being held. It would probably be more efficient to do so in terms of deploying political party resources and the opportunity cost of your average voter, but the political meme streams get crossed in the public brainspace, to the point where grievances and perceptions of State and Federal politics become too intertwined with too many people, making the fortunes of one level of politics interdependent on the fortunes of the other.

So, ignoring Lindsay, what we find is that the results are entirely expected and simply not unusual.

If we order all seats in terms of their margin from highest to lowest for a given party – giving us an electoral pendulum – when a national swing of, say, 3% occurs, seats sitting on a margin of less than 3% fall and seats sitting above the 3% margin don’t fall – with one important caveat.

For every seat that swings by more than 3%, there will be another seat that swings by less than 3%  and vice versa simply to make up the maths of a national swing – this is the variance of our swing and it usually comes in as a standard deviation of between 2 and a bit and 3 and a bit in federal elections. Not always, but usually.

So what we end up with is an approximately bell curve shaped distribution with a mean of the national swing, and where individual seats fall down approximately equal on either side of that national swing.

According to the last Newspoll, the national swing to the Coalition is about 1%, so with the marginal seat polling, what we would expect to see is some seats swing more and some seats swing less. That – as it happens – is exactly what the marginal seat polling produced, with the Qld marginals swinging more to the Coalition and Page swinging less (in fact, swinging to the ALP).

All this really tells us is that these Qld seats at the moment are likely to be swinging slightly stronger to the Coalition (although we can’t tell the strength of the swing in any single one of those three seats), meaning that other seats somewhere are swinging less to the Coalition. Where would those seats be? Well Page looks like one, but there would be others as well – we’ll have a better idea when we see the Newspoll quarterly breakdowns giving us State based federal voting intentions over the next few days, where we can aggregate them with Nielsen and some other polling to take a closer look.

(Visited 1 times, 1 visits today)


Leave a comment

22 thoughts on “Marginal Seat Newspolling

  1. TheTruthHurts

    [What are Truthy’s latte belt seats?]

    The latte belt seats include: Fowler, Prospect, Weriwa, Chifley, Reid, Blaxland, Lowe, Banks, Grayndler, Kingsfor Smith and Sydney.

    You can check out the map here:

    In these seats Labor is very very very popular, but these seats are irrelevent because they hardly ever change hands and are Labor strongholds. It’s the marginals like Townsville that Rudd has to try and win, and other Queensland marginals which he must try and hold.

  2. Possum Comitatus

    It certainly does Ken – especially in Qld where they do actually have a history of voting for 3rd parties when they say they’re going it. The preference allocation issue can get problematic when you get a big surge in the minor party vote and they only thing we can really do is keep an eye on Nielsen poll data where they let respondents allocate their own preferences.

    As for what Truthy’s “latte belt” seats are – one presumes anywhere with a Banjo ownership level of less than about 40% 😛

  3. Ken Macpherson

    Poss, surely combined greens/third party votes of 21 and 24% in the Qld marginals and Page make these polls even more difficult to interpret? Would Newspoll’s practice of allocating preferences as per the last election be an appropriate way of calculating the 2PP for this big chunk of votes? What are Truthy’s latte belt seats?

  4. David Richards

    hear hear Poss

    methinks thou doth protest too much, Truthy

    Could it be the Liberals who are in trouble in their marginals?

    That is one point that seems to have escaped the MSM shark’s feeding frenzy – while there may be some ALP losses, there are also bound to be some Lib marginal losses (Mincing Poodle being put down?)

    If there is any net gain, it is more likely to be to the ALP than the Libs.

    The Libs have not only to gain seats, but not lose any.

  5. Possum Comitatus

    [I’ve covered this in the other blog… Labor is losing marginals all over the place, but is polling strongly in the latte belt(seats Labor has held for years and don’t decide elections).]

    That you did Truthy, and it was as silly over at Pollbludger as it is here.

  6. TheTruthHurts

    I’ve covered this in the other blog… Labor is losing marginals all over the place, but is polling strongly in the latte belt(seats Labor has held for years and don’t decide elections).

    Rudd’s in big trouble. He needs to win the marginals, not decades safe Labor latte seats.

  7. David Richards

    Polling Polling Polling
    Keep Them Pollsters Polling
    Though you’re disaproving
    Keep Those preferences moving

    Vote ’em in Vote ’em out Vote ’em in

  8. David Richards

    If the ALP don’t use clips of one of the mining execs signing a new deal with China in their ads, they are missing a big opportunity.

    They should also throw the mining bosses’ own words back at them, where they contradict their current stand.

    The MCA campaign is so full of lies, the Trade Practices Act should be invoked.

  9. DaveM

    MoE and statistical caveats aside, wouldn’t seats like Dawson and Flynn particularly be right in the middle of the mining tax shitestorm?
    And Longman is where the LNP picked a 19 year old candidate, right?

  10. Possum Comitatus

    Jimmy – that is just business as usual at The Oz with poll reporting.

  11. JimmyDoyle

    Poss – then shouldn’t you argue that the trumpets at the OO aren’t taking this poll for what is?

  12. Possum Comitatus

    There’s nothing wrong with it John – it will give you a polling estimate for the combined population of those three seats.

    Whether that has any actual political value is arguable, since they are three seats separated by some serious geography – but statistically its fine as long we treat it for what it is.

  13. imacca

    But Poss, according to the great and all seeing Dennis this polling means that “ALP is living on false hopes”, and:

    “The Newspoll survey of the marginals will confirm those darkest fears”

    All very spooky isn’t it?

    Hmm, for poll analysis who to believe, Dennis or the Marsupial??

    No brainer there.

  14. john

    Is that a legit thing to do? I mean, statistically?

  15. Lastest Politics News | Political News Channel.com

    […] Marginal Seat Newspolling Newspoll released today via The Oz a set of marginal seat polling that they undertook along with their usual nationwide poll over the period from last Friday to Sunday. Read more on Crikey […]

  16. Possum Comitatus

    I don’t think it would have been exactly 200 per seat- what they seemed to have done was pooled all the phone numbers in those three seats, and chosen 600 people randomly from that pool.

  17. john

    So 200 per seat?

    That moe would be huge, surely?

  18. john

    Was the sample of 600 from each Qld marginal, or 600 all up?

  19. Tweets that mention Marginal Seat Newspolling – Pollytics -- Topsy.com

    […] This post was mentioned on Twitter by Possum Comitatus and Trevor Cook, Brett Peterson. Brett Peterson said: @AustraliaVotes 600 respondents per electorate, per @Pollytics (http://bit.ly/cCoML7) […]

  20. David Richards

    Also, in the case of Lindsay, you would have poll responders who could differentiate between State and Federal politics, but mistakenly thought they were being polled with regard to the by-election, and responded accordingly. A natural assumption to make. Sometimes people don’t necessarily listen too well, and just assume things. Phone rings. Voter picks up phone. Pollster asks if they would be willing to participate in polling. As there is a by-election for the state seat imminent, voter assumes it is for the state government and answers accordingly.

Share this article with a friend

Just fill out the fields below and we'll send your friend a link to this article along with a message from you.

Your details

Your friend's details