Galaxy marginals poll

I heard it said recently that Galaxy had a strategy of skimping on sample sizes for early campaign polls and then going to town at the end of the campaign, in the hope that their final polls will prove more accurate than the competition – which, by and large, they have done. Sure enough, today’s News Limited tabloids have unloaded a giant survey of 4000 voters covering four marginal electorates in each mainland state. The graphic as published in Perth’s Sunday Times can be viewed here. The surprise is the mild 5 per cent swing across Bonner, Herbert, Longman and Moreton, pointing to a disappointing result for Labor in Queensland. However, it has been widely reported that Labor is on course for bigger swings in the north of the state than in Brisbane, which is over-represented in this sample of seats (though this was also true of Newspoll). Newspoll and Galaxy are also at odds over Victoria, respectively pointing to swings of 8 per cent (a likely gain of five seats) and 4.5 per cent (a likely gain of zero seats). Three of the four seats sampled were the same, the fourth being McMillan in Newspoll’s case and McEwen in Galaxy’s. Where Newspoll surveyed the four most marginal seats in South Australia, Galaxy has ditched Kingston as a lost cause to take on Sturt. One can only speculate how much that accounts for the difference between the two results: a 4.5 per cent swing in Galaxy’s case and 8 per cent in Newspoll’s. It should be noted that two Sturt polls appeared in the week suggesting Galaxy’s swing is nearer the mark than Newspoll’s. Perversely, the state where the two agencies picked the most differing sample of seats, New South Wales, is the one where they produced the most similar results – a 7 per cent swing in Newspoll’s case, slightly higher in Galaxy’s.

Then there is Western Australia, which Galaxy surveyed and Newspoll did not. Here Galaxy points to a swing of barely 1 per cent, less than would be needed for Labor to win Stirling and Hasluck, and perhaps not enough to hold Cowan given the loss of Graham Edwards’ personal vote. For more on Western Australia, see the post below on the Westpoll survey, to which I have now added a graphic from The West Australian. It gives Labor a slightly bigger swing of 3.5 per cent.

UPDATE: Chris Hammer of The Bulletin lambasts Galaxy for not waiting until the campaign launches were done before conducting its poll.

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.

111 comments on “Galaxy marginals poll”

Comments Page 1 of 3
1 2 3
  1. Assuming the poll’s accurate, all it’d take is a 1% narrowing in NSW to pull it back to a Coalition win/hung parliament situation, and we’d have 1998 all over again.

  2. Sure enough, today’s News Limited tabloids have unloaded a giant survey of 4000 voters covering four marginal electorates in each mainland state. The graphic as published in Perth’s Sunday Times can be viewed here.

    So that is 4000 people over a total of 24 electorates, less than 200 per electorate, correct me if I am wrong but I make the MOE around 6-7%.

  3. How accurate are individual seat polls? Espeically ones that have a mixture of city and towns.

    And with a national 53-55/45 polls I guess it becomes a moot point.

  4. How is it we have reached the final week of the campaign and NOW Galaxy gives us marginal seats which are at odds with months of their own National polls AND Nielsen Polls still have not done marginal seat polls. DOESTHIS MEAN ALL 4 Plling Companies National Polls INCLUDING 3 this week are irrelevant ???

  5. It’s going to be close so don’t get ahead of yourselves Labor voters!

    Biggest swings will be in NSW, Vic and SA and heres where we will pick up the most seats, Herbert is in the bag for Labor here in Queensland and so is another 2 or 3 in Brisbane.

    WA I think will have a slight swing but not enough for Labor to win more seats, but Labor will hold on to all their seats in that state.

    Tassie Labor will win back it’s 2 seats, and 1 seat in the NT.

    My prediction is between 79-85 seats to Labor. And remember we need at least 53%+ TPP Labor vote to have any chance. Every change of government SINCE WW2 has occured only if the opposition party recieved over 52% TPP

  6. Who reckons that a lot of these professional pollsters will need a good scrub to get the ‘egg off there face’s’ come Sunday the 25th!

    My call is Labor will win with anywhere from 90 min. to 105 Max. THE GURU HAS SPOKEN! 🙂

  7. William, NSW & Vic only I think; no#s…but interesting.

    ‘The latest Sunday Age/Taverner Research poll shows that 57% of the key mortgage-holder demographic will be voting for Labor, compared with only 43% for the Coalition.

    This is a complete reversal of the 2004 campaign, when a Taverner poll conducted then showed the Coalition enjoying an 8-point margin over Labor among mortgage holders. More than one in five mortgage holders who previously supported the Coalition appear to have shifted their allegiance to Labor.’

  8. It’s always been closer than it’s looked in my opinion. I’m still confident of a Labor win on election night, but I think 90 seats is a fantasy. I count about 12 certainties, about 10-15 tossups with a couple of them leaning Labor and then a third tier of ‘if the swing is in the right place’ seats of another 10 or so. Some seat by seat guesses would be interesting, rather than bulk numbers. I think we take down Kingston, Bonner, Macquarie, Wakefield, Makin, Bass, Solomon, Moreton, Lindsay, Eden-Monaro, Dobell, Blair. Probably Braddon as well. It’s tough to see any near certainties past there though… everything not mentioned down to the 6.3% mark of Paterson has to be rated a tossup or lean Labor, except maybe Wentworth is leaning Liberal in my opinion. Still, that’s 14 tossups, so if Labor wins 7 it’ll be a comfortable win on election night with a gain of 19. I’d say there will be 1 or 2 seats that defy the uniform swing and flip as well (i’m thinking Bowman, possibly Flynn, Hinkler or Leichhardt if the swing is unevenly spread across QLD – I used to live in Bundaberg and they are a fickle bunch in Hinkler, big swings wouldn’t surprise me there although the editor of the local paper told me Parr has said some pretty stupid things… Flynn is possible and Leichhardt too if the coalition experiences preference leakage and the retiring member’s personal vote is large enough) and possibly 1 or 2 Liberal pickups (WA), so i’ll go with Labor 79 seats with margin of error of 5 either way.

    Then again, I hope i’m wrong and we win 90 seats. It would be sweet to run the table on those tossup seats and see Howard lose Bennelong…

    I’m volunteering for Kerry Rea in Bonner, news on the ground is good from what i’ve heard. Doorknocking near the booths that went for Vasta, the response has been excellent so far. I’m going to guess Bonner will be 54-46 on the day. I’m working a booth that went 51% Liberal last time, so I think I might get a good feel for the whole election. Most of the Labor people are very confident of at least winning Bonner and Moreton in Brisbane and Bowman is very much seen as a 50/50 prospect from the people i’ve talked to.

    Oh, and hi, i’ve been reading the site for a while now, decided to start posting too.

  9. If I’ve learned anything from Possum and others since venturing into the blogosphere, it’s to observe the trend and be wary of polls with a high MoE. In any case, Galaxy has generally pointed to a lower Labor 2PP vote. This could just be sampling error, or it could be to do with their methodology. In the absence of more information, and without confirmation, Galaxy should be discounted.

  10. Actually, I am still trying to understand what they have done here.

    4000 samples should give a MOE of about 1.5%… but this is not true random sampling. They are saying that they are sampling the marginal seats, so obviously they this is not proportional to the electorate at large.

    I don’t quite know what to make of it. If the swing is smaller here, then it must be larger elsewhere.

  11. #10: If seats such as Hasluck approximate the Labor vote at 5% lower than its true value, then seats such as Moreton could just as easily be estimating the Coalition vote at 5% below its true value. In cumulative error is likely to be in the region of 2%, and some seats will be biased one way, some the other, in the end you’ve still got 4000 people from the seats that matter (this poll), versus 1000 people from all over the place (standard Nielsen/Newspoll/Galaxy national), where you could just as easily argue that the MoE in any one seat was 37%(!).

  12. If you say that about Galaxy you could say the same about Morgan always showing a higher Labor vote. I’m a natural pessimist though, having grown up in the Howard years watching Labor lose election after election, so I find it hard to believe we’re actually that far in front. I also saw John Kerry polling a solid election winning lead one day before the election… almost put money on at $2.25 then thought better of it back in 2004.

    I think 52/48 or so on election day, but enough swings in the right places to give Labor a narrow majority. I hope it’s 54-55%… the polls say it should be, but doesn’t 1-2% always fly back to the incumbent on election day?

    If it’s more than 52/48 and Labor doesn’t win… the AEC has a lot of explaining to do. Hell, if it’s more than 50.5/49.5, the AEC has a lot of explaining to do.

  13. Oliver a few Labor voters in here have had a rush of blood to their heads and aren’t giving us Mark Lathamist predictions of grand wins.

    I’m sure Rudd is going to win, in my mind he is not someone who would lose an election(if that makes senses)

    But some of the wild predictions in here of 90+ seats make me laugh… come on back to planet Earth people. Howards still got “the economy” on his side, this is purely a moral and “it’s time” vote that will get Rudd over the line.

    I also believe that the longer a member has been in power the harder it is to kick them out(in other words the swing has to be BIGGER nationally). It would be a lot easier kicking out a 1996 elected Lib member in 1998 then it would be in 2007 over 10 years later! I believe a bigger swing is needed every year to kick them out(unless they are a poor member)

    Labor needs at least 53% TPP to win government(as seen from historical elections of past a 52%+ vote is required to change government)

    Lets hope we get the 54% TPP on the day!

  14. #16: The poll isn’t made to measure the electorate at large, just the marginals, so taking a random sample from within them is perfectly acceptable, and more relevant, because a voter in Grayndler or Bradfield isn’t going to mean anything to the outcome.

  15. From the Age:
    In a statement to The Sunday Age yesterday, Vaile said he knew Ian McPhee personally, respected his professionalism and independence, and “I regret questioning the timing of the report”.

    Really Mr Vaile!

  16. Justin @ 20:

    Ok, looking at the figures, the TPP here is about 51%… we know that the TPP for the population as a whole is about 55%, so the rest of the TPP must be being felt elsewhere.


  17. I can’t see that these figures are very useful in terms of either party making a prediction and my read on it is that they would both be trusting their own internal polling at this stage (with only 6 days to go).

    The main benefit to Labor is to ensure voters don’t “mess around” with their vote and that Team Rudd maintain their discipline.
    It’s all good.

  18. By the way… i’m staying in the Brisbane electorate atm… why is Arch Bevis still paying $1.08 and in Capricornia… $1.11. Moreton is $1.12. Surely there’s a better chance of holding Capricornia than picking up Moreton? I mean, it’s very likely both will happen, but still, it’s weird. Good money on Labor incumbents? It’d be worth considering a $2000 bet or something to win $220. Infact, I should look into that on Friday if there are no changes in polling or odds before then and no time for last minute surprises. Surely that’s even better money than Labor at $1.23. Also, Franklin at $1.14 and Lyons at $1.16? Labor’s at $1.23! Surely we hold our incumbent seats in Tasmania before winning government… Eden-Monaro $1.25? We’re certainly not winning without it and it’s worth more than a Labor win. Macquarie $1.16? It’s a notional Labor electorate! Again, we’re certainly not winning overall without picking up Macquarie.

    Logically if Labor is at $1.23 (Portlandbet odds) then shouldn’t the 15th-16th most marginal seats also be close to $1.23?

    Kerry Rea’s been backed into $1.04 from $1.10 as a non-incumbent. I’ll have to tell her and see what she thinks of it next week… lol, apparently four times safer than the Tasmanian incumbents.

  19. #22: Maybe there’s been a bigger swing in safe seats.
    Maybe the true 2PP is closer to 52-52.5ish at this point in time (fits with both this and the national polls which say 54-55 with a 3% MoE)
    Maybe the methodology in this poll is completely out (if it’s fine then my MoE calculator says there’s a 99% chance that it’s within 2% of the true TPP of these seats).

    I’d favour the first or second options.

  20. What I don’t get is Labor is @ $1.19 now to win.

    You would have to be absolutely raving bonkers to put money down on those odds, when you consider you can put a bet on a single candidate in a pretty safe bet seat like Herbert @ $1.50 win for Labor.

    It’s easy for Labor to win these seats, it’s harder for them to win the election

  21. THIS IS WHAT GALAXY DO NOT SAY: the Galaxy Poll of 4,000 voters (is 200 voters in each of 4 seats in 5 states equals 4,000 voters) but Galaxy polled per Poll Bludgers attachment over 5 evenings from the 10th to the 15th November
    ie last Saturday night to last Thursday night. In between both Party’s made their Launchs. WOULD GALAXY’s POLLING PER SEAT OVER EACH OF THE 5 NIGHTS BEEN 40 VOTERS PER SEAT EVERY SINGLE NIGHT (TO MAKE UP THE 200 per seat per state ie 800 per state) ?????? if not the polling would be affected by when during the launchs the phone calls were made. does anyone know ??

  22. There is something amiss in that Galaxy table, the figures for Oct 2004 are not the final election results so what are the swings based on? For example the table shows LNP TPP of 54.5 NSW, 54.1 Qld, 55.6 Vic where the election result was 51.9, 57.1 and 51 respectively.

    If I’m not mistaken the Galaxy Qld prediction of 51/49 ALP against election 2004 results of 42.9/57.1 is actually an 8.1% swing and not 5% as you stated William.

  23. Justin @ 25:

    Hrrmmm… bigger swings in “safe” seats…

    This will be fun to watch.

    You don’t believe for a second that the TPP is about 52% do you?

    I don’t think there has been a single poll in 6 months or more that has reached that low…

  24. If Labor wins by 2 seats then that is fine – the job is done. We will all be happy. AND there is nothing we can do anyway really, except help out the booths and sit and watch.

    In reality all the polls all year have pretty much agreed with each other and even now point to a 53 – 55 result. Possum’s model points to 55.1

    The most disturbing thing for LNP must be the AC Nielsen since in every election it has been pretty good on the ALP primary and, over estimated the LNP vote each time by a few percent coming into the last week. If it runs true to form again then Possum’s wont be far wrong.

    And this has been a terrible last week for Howard – it is all falling apart. Today’s headlines are not going to help and, as has been noted by Anthony Green statistically undecided voters tend to break 2/1 with the trend. The 2001 electioni had a similar break.

    With LNP and Howard starting to look like lame ducks in this last week the undecideds will follow their usual pattern which, could actually increase Labor’s position.

    You can sit and worry or sit and be confident – it wont change the result. If Labor and Rudd can’t win under these circumstances it will be hard to see when they could ever win.

    My original prediction was for 84 seats – I now would not be surprised if it goes 90+ but would be happy with any win.

  25. I don’t believe the nationwide TPP is 52%, no. It’s at the very least 53, but it could well be 52 in these seats given there’s no past polls to say this is an outlier (and the sample size indicates it’s unlikely to be one, anyway) and a lot of the swing could be in safe seats where just about nobody is out campaigning (from the Right side of the fence, anyway).

  26. Let it end,
    You may be confusing the tpp in the 4 seats in each seat, with the state tpp. The figures of tpp do add up to the collective 4 seat tpp.

  27. This poll gives 18 seats to Labor, but I don’t think they’ll win Bennelong, Wentworth, Dobell or Paterson, so that’s 14 seats. If Howard takes Cowan, that’s 13 nett.

    On this poll I am giving it to Howard by the barest of margins, and real problems from Georgiou, et al, next term threatening to cross the floor.

  28. You can have an error in one or two polls but when you have a group and series of polls from different pollsters pointing to the same thing you would have to accept the reality.

    The current TPP is certainly between 53-55 and with upward pressure – not downward.

  29. Justin @ 31:

    Glad to understand you properly!

    It is so hard to understand these marginal seat polls. I know Possum on the other thread poured scorn on these types of things, and from a layman’s perspective, I would have to agree.

    What is to be made of them? In and of themselves, they are not so terribly useful, methinks.

    There are larger forces at work, and these particular marginal seats polls, are just a subset (if accurate) of the larger whole.

    They are such a strange beast. 800 samples from one seat would be much more useful, to my way of thinking. What actually is this poll representative of?

  30. The Green vote would also suggest that the sleeper of the campaign is Global Warming, whilst most recognise that Howard has totally stuff his IR campaign he has looked shocking on GW. GW is an interesting one as it could increasingly be a ‘vote’ changer.

  31. The very best result for Labor is just enough votes to win the most marginal seat by one vote, the next most marginal by one vote.. and so on.

    With this sort of voting pattern Labor could win with *less* than 50% of the overall 2PP but *given* that they are likely to get about 54-55 percent overall *and* that the swing goes up the less marginal the seat is…

    Well.. you do the maths 🙂

  32. Galaxy’s Poll produces a 2 PP of 50.8% Labor & 49.2% for Liberals. Each state had 800 polled & the results FOR LABOR 2 PP were :
    WA 50%
    NSW 53%
    Vic 49%
    Q’ld 51%
    SA 51%

    WHERE is Labor’s 2PP of 54% per Nielsen and 54% per Newspoll
    In safe Labor seats ?????

  33. You have a Newspoll of marginals pointing to a mini-disaster for the LNP and Galaxy of marginals point to an 18 seat loss as well. AND individual seats can be difficult to sample properly. You have national polls pointing to a disaster for the LNP and some want to give it to Howard?

    Howard will most certainly lose Bennelong. Wentworth will probably be saved. Howard will need a miracle to save this election.

  34. follow the preferences, you know something, despite the press being determined to label Rudd as me-too the reality is that most sensible voters (forget the ‘swingers’ from four corners for a moment) see Rudd as the clear winner on climate change.

    can’t wait for some of the latest seriously scary data about climate change to filter through in the media this week…

  35. I wouldn’t get carried away with this Galaxy in opposition to the Newspoll marginals which actually is in synch with the national polling by all the polsters.

    The analysis has already been done – the smaller swings are in Labor seats, then LNP marginals and then LNP safe seats.

    Mad cow is right – this is why it could actually turn out to be a masacre if the planets line up right.

    Next is the Morgan marginals.

    Galaxy’s Poll mathematically equals a 2 PP of 50.8% Labor & 49.2% for Liberals. Each state had 800 polled & the results FOR LABOR 2 PP PER STATE were :
    WA 50%
    NSW 53%
    Vic 49%
    Q’ld 51%
    SA 51%
    This works out at 2032 actual 2 PP votes to labor out of the 4,000
    which equals 50.8% 2 PP

    1/ WHERE is Labor’s 2PP of 54% per Nielsen and 54% per Newspoll
    In safe Labor seats ?????
    2/ Galaxy’s poll is over 5 nights in 20 seats equals 40 polled per seat per night
    but the 5 nights were last sat night to last thurs night…would not the poll be affected by WHEN in the week voters per polled ???

  37. Go to Possum’s site – he did a thourough mathematical analysis of where the swings were taking place some time ago.

    Look for the series of articles called ‘Pollycide’

    THE bad news for the LNP was that the swings were not happening so much in Labor seats but in marginals and especially Liberal safe seats. This is no doubt why you see Howard and Rudd campaigning in LNP safe seats – they are also up for grabs.

    If these Galaxy figures are near correct then Howard has some severe problems because not only is he going to lose a bunch of marginals but a very big bunch of safe seats.

    It is intuitive in anycase that Labor safe seats will not swing much since they already have a saturation of labor voters. The bigger swings will more likely take place where there are the least Labor vote no doubt.

  38. Laborvoter #19

    Labor needs at least 53% TPP to win government(as seen from historical elections of past a 52%+ vote is required to change government)

    When, in history, has a government kept power on less than 49% of the vote after an election? I can’t find any such event in postwar elections.

  39. More I think about it the more I just don’t get this poll. As mentioned in Williams Westpoll comment Lateline reported last Friday night this Galaxy would show Labor TPP of 53/47 and yet it has 50.8/49.2. Very strange.

    Hard to believe Vic has the lowest ALP primary vote of 37 for any state, that’s 7 below Newspoll in 3 out of 4 of the same seats.

  40. To me there is something horribly wrong with doing a “marginal” seat poll and then not polling all the most marginal seats. Even worse than that, they have thrown a number of non-marginal seats into the mix – Robertson at 6.9%, Longman at 6.7% and Sturt at 6.8%.

    I would suggest that if the result was 53/47 even though it contains these less marginal seats then it points to a big Labor win – it shows that Labor will pick up all the most marginal seats and that there will be a few ‘second tier’ seats in play as well.

    Also, not all states were polled either, so if you add Soloman, Bass and Braddon to the mix then it makes the position for Labor look even better.

  41. Milne’s assessment of this poll states – “While Galaxy did not poll in Tasmania or the Northern Territory last week, the survey results would also give Labor Tasmania’s Bass and Braddon, and Solomon in the NT.”
    Can anyone explain how this could possibly be?
    Also, Galaxy covers their butt by saying that just a very small further swing in Queensland would see an extra 8 seats fall. Other seats would fall in Victoria with a similar further small swing.

  42. There are some big questions here:

    1) Newspoll shows that NSW has moved for the ALP by + 2% in just two weeks and their last figure now cross-checks welll with Galaxy. Why the shift? was is the interest rate hike?

    2) Newspoll and Galaxy differ bigtime in Victoria. The idea that LNP are still ahead 51/49 in Victortian marginal seats (Galaxy), when Newspoll has the whole state for the ALP at 59/41 = WTF?

    3) Both Newspoll and Galaxy show QLD and SA to have moderated thier swing to the ALP dramatically and agree the net gain for ALP will be only 5 or 6.

    It wasn’t that long ago the polls were showing QLD would deliver up to 10 seats. Does this mean the LNP campaigning has done some great things?

Comments are closed.

Comments Page 1 of 3
1 2 3