tip off
43

Election Simulation – ALP 101 Seats

Hot off the heels of today’s Newspoll quarterly aggregation in the Oz, comes the Pollytics quarterly election simulation. The way it works for those never having seen it before, is we combine the state breakdowns of both Newspoll and Nielsen over the past 3 months, weighted by sample size, to get a pooled estimate of the two party preferred for each State. That State estimate then gives us the state based swing since the last election for each state, whereby we then apply the relevant  State based swing to all 150 seats. However, we don’t just do it once (like happens with the election pendulum). We create a probability distribution for each seat where the mean swing is the state based swing, and where a standard deviation is applied that is a combination of the historical standard deviation of each state at Federal elections, as well as the standard deviation attached to the sampling error of the polling – the margin of error of the polling for each state.

We then then simulate 20,000 elections and tally up the results to get the most likely outcome were an election held in the last 3 months.

Were an election held, the ALP would have won 101 seats – a 1 seat gain since the June quarter. For the nerdy types, the distributions of the simulations – done in both “Seats in Parliament” and “Total Seats won/lost” – using the ALP come in like this:

seathist seatsim

seatsgainhist seatsgainedsim

The histograms (column charts) show the distribution of the simulations while the line charts are the results turned into a probability distribution, where you can simply choose an outcome from the bottom axis (either the ALP gaining X number os feats in Parliament, or the ALP gaining X number of seats from their current position), trace that outcome up until it intersects the red line, then trace that across to the vertical axis to show the implied probability of the ALP gaining at least that number of seats were an election held in the last three months.

We’ll be going over this quarterly Newspoll for a few days, and we’ll change the demographics in the sidebar (which is a combo of both Nielsen and Newspoll data) today or tomorrow – but in the meantime, it’s worth looking at the estimated change in voting intention since the last election, by geography, using this latest Newspoll.

alpprimsep09 lnpprimsep09

greenprimsep09 alptppsep09

The Coalition are down across the board in every state and the capital cities, only holding their own in the regions. Labor have lost some primary vote in NSW (it went to the Greens on net), and a point thereabouts in the non-capital cities (again, to the Greens on net). Qld is steady, Vic is up slightly, while South Australia and WA are up strongly.

South Australia is a major problem for the Coalition – facing a swing of 8.6% against them since the last election. Christopher Pyne must be in deep, deep shit in Sturt with the two party preferred vote in WA (oops) SA going to Labor 61/39. WA has come to the ALP party 2 and a half years later than everyone else, with Rudd pulling his highest ever primary vote in WA of 43% compared to the Libs 41%, washing out into an ALP two party preferred of 54/46 – again, the highest ever under the Rudd leadership of the ALP. In fact, the ALP has only ever achieved a primary of 43% in WA once before in the history of the Newspoll quarterly results – back in the 4th quarter of 1999.

41

Please login below to comment, OR simply register here :



  • 1
    Posted September 28, 2009 at 8:26 am | Permalink

    But The Oz tells me the quarterly breakdowns are good for the LNP….

  • 2
    Posted September 28, 2009 at 8:46 am | Permalink

    Insert some joke about Thinking.Again :-D

  • 3
    Posted September 28, 2009 at 8:51 am | Permalink

    Pyne’s seat is very marginal now isn’t it? And both Ruddock and Bronwyn Bishop have re-nominated. So much for the renewal of the LNP. They are kind of deserving of the desertion of the electorate.

  • 4
    Posted September 28, 2009 at 8:54 am | Permalink

    Mac, Pyne won Sturt last time 50.94 to 49.06.

  • 5
    Posted September 28, 2009 at 9:04 am | Permalink

    I’ll miss Chris when he goes.

  • 6
    Aristotle
    Posted September 28, 2009 at 9:04 am | Permalink

    Possum, your bar graph charts re change in votes, is that only Newspoll or Newspoll and Nielsen combined?

  • 7
    Posted September 28, 2009 at 9:11 am | Permalink

    Ari, the bar charts are just Newspoll. Sorry I didn’t make that clear – better go fix that up.

  • 8
    polyquats
    Posted September 28, 2009 at 9:22 am | Permalink

    But The Oz tells me the quarterly breakdowns are good for the LNP….

    That was the Fran Kelly take on it on RN Breakfast this morning. Funny, that.

  • 9
    Andos
    Posted September 28, 2009 at 9:23 am | Permalink

    So Scott, where you say “Were an election held, the ALP would have won 101 seats,” what you really mean is “Were an election held, the ALP would have a 50% probability of winning at least 101 seats,” yes?

    Also I think you meant SA, not WA here: “the two party preferred vote in WA going to Labor 61/39.”

  • 10
    Posted September 28, 2009 at 9:29 am | Permalink

    Andos,

    Sort of – it’s also the mode, the median and (nearly bang on to the decimal place) the mean simulation values. Effectively it’s the point estimate of the simulation, not only having a 50% probability of of the ALP winning at least 101 seats, but where 101 seats is the most likely result of all total results.

    Ta for the WA/SA stuff-up – I’ve fixed it.

  • 11
    evan14
    Posted September 28, 2009 at 9:37 am | Permalink

    I doubt that Turnball will lose Wentworth, and I doubt Pyne will lost Sturt(much as I’d love it to happen).
    The best scenario for the ALP in the 2010 election is something over 90 seats(which would still count as a landslide)!
    Thanks Possum, interesting analysis as always!

  • 12
    Aristotle
    Posted September 28, 2009 at 10:23 am | Permalink

    I’ve done a quick calculation with Newspoll’s and Nielsen’s quarterly results, and my results concur with yours above, Possum.

    On these figures it’s at least 19 seats to the ALP = over 100 seats.

    With the National swing at over 3%, giving up about 20 seats, you would expect the state breakdowns to deliver something similar.

  • 13
    DrMick
    Posted September 28, 2009 at 10:27 am | Permalink

    Anyone know how many of those 19 likely-to-go seats are held by the Nationals?

  • 14
    Aristotle
    Posted September 28, 2009 at 10:32 am | Permalink

    3 Nat seats on the pendulum, Dr Mick, Cowper, Calare and Hinkler.

    The rest are Lib seats.

  • 15
    Posted September 28, 2009 at 10:36 am | Permalink

    Mick,

    Cowper and Hinkler are the two Nats seats in those numbers at the moment -under the old boundaries anyway.When the redistributions are finalised by the end of the year, I’ll move the simulation across to the new boundaries where Calare will become another Nats seat thrown into the mix as well.

  • 16
    thewetmale
    Posted September 28, 2009 at 11:00 am | Permalink

    So according the The Australian a score of -0.8 is equal to a “plunge” in support.

    Poss would it be possible to have your four nifty graphs re-calibrated to show the change in numbers not from the election but from the last Newspoll quarterly breakdowns? It seems to me that the line from the Australian, reliably parroted by the lazy ABC, to be generous to the Aus, is about a change from the last poll not from the election.

    I agree with the line that the Aus is probably fishing for a good story for the coalition/something interesting in Newspoll but i’d like to know if it is based on something truthful, i.e. a significant change in Labor numbers since the last quarterly breakdown or if it is complete fluff with not even a tiny bit of logic to it.

    But yeah, as always a clear analysis and representation of the growing hole that the Libs find themselves in.

  • 17
    Aristotle
    Posted September 28, 2009 at 11:03 am | Permalink

    Possum, there must be some rounding of figures going on in the Newspoll results, esp in QLD. Based on the published pirmary votes in QLD, I get at least 52% TPP to the ALP, and with Nielsen’s results suggesting over 53% TPP, the swing to the ALP stands at over 2%.

    Hardly a plunging vote.

  • 18
    Posted September 28, 2009 at 11:09 am | Permalink

    TWM,
    In the non-capital cities, Labor dropped 4 on the primaries (from 43 to 39) and 5 on the TPP (from 54 to 49) between the second and 3rd quarters. The Coalition jumped 5 on the primaries in the non-capital cities over the same period.

    Ari,
    There’s a always rounding issues here. I too get a QLD ALP TPP of 52/47 to Labor off primaries running 43/42 that way. It doesn’t really matter though – arguing over a point here or there is pretty naff :-D

  • 19
    thewetmale
    Posted September 28, 2009 at 11:17 am | Permalink

    Cheers Poss. Although that is significant, going from a TPP of 54 to 49 but i’m sure Labor are quaking in their boots at their loss in support in more regional and rural areas. What with population growth and demographic change being what it is and all ;-)

  • 20
    JimmyD
    Posted September 28, 2009 at 12:12 pm | Permalink

    Poss – I would have thought that the bigger news than a drop in the regional vote would be the fact that Labor has opened up a 12 point lead on the primary vote in the 5 capitals, and are now leading 60-40 TPP. The OZ is clearly ignoring the fact that 60% of the country’s population resides in the five capitals.

  • 21
    Posted September 28, 2009 at 12:15 pm | Permalink

    I’m getting to that Jimmy D! This place has more gazumpers than Dodgy Brothers Real Estate! :-D

  • 22
    JimmyD
    Posted September 28, 2009 at 12:19 pm | Permalink

    lol – I was not aiming my fire at you Poss, but at the OZ

  • 23
    bob1234
    Posted September 28, 2009 at 12:44 pm | Permalink

    and I doubt Pyne will lost Sturt(much as I’d love it to happen).

    Having lived in Sturt for ages, i’d have to agree with you on both points.

  • 24
    Nipper Quigley
    Posted September 28, 2009 at 12:57 pm | Permalink

    Any idea of the moe for the non-caps? Given the wide variation between the states and the apparently huge drop in support for Labor (Coalition primary up 5 on last time), this could throw some light.
    Or is it real and our country cousins backing Barnaby to the hilt?

  • 25
    JimmyD
    Posted September 28, 2009 at 12:59 pm | Permalink

    evan14 and bob1234 – Pyne holds Sturt with a 0.9% margin and SA is swinging by 9% to Labor according to Newspoll – you do the maths.

  • 26
    Posted September 28, 2009 at 1:00 pm | Permalink

    Nipper,

    The MoE for the non-caps should be just under 2%.

  • 27
    Posted September 28, 2009 at 1:02 pm | Permalink

    I’m with Jimmy – if SA swings by even 3 or 4%, the tide of that swing will wash Pyne out of Sturt.

  • 28
    Andos
    Posted September 28, 2009 at 1:13 pm | Permalink

    Thanks for your explanation at 10, Scott. That makes it clearer.

  • 29
    fredex
    Posted September 28, 2009 at 2:35 pm | Permalink

    Hmm.
    Since the election:
    ALP has dropped .8% in non-caps.
    Greens have risen .9% in non-caps.
    Should just about cancel.
    Apparently not tho’, last graph has ALP 2pp of -.4% since the election.
    Is that due to “Others”?

  • 30
    Posted September 28, 2009 at 2:47 pm | Permalink

    Fredex,

    Yep, Others and the fact that Greens prefs only flow 70-80% to Labor (so there’s a bit of leakage)

  • 31
    Tri$tan
    Posted September 28, 2009 at 5:38 pm | Permalink

    Thanks Possum. Always enjoy the simulation days.

    Does Australia have enough data to take this fowards into not “if the election were held in the last 3 months” but a prediction what the actual outcome will be at the next election?

    I understand this is more difficult, but Nate Silver was doing this over at Five Thirty Eight for the US election (with good accuracy it turned out). I think/assume by looking giving some parametres in possible changes over time by increasing the error bars. http://www.fivethirtyeight.com/

    Maybe I was fooled by his fancy numbers, and it’s obviuosly more difficult to predict accurately than reflect on what would have happened. My guess is that you probably are not close enough to the election just yet – but this might be more possible the closer you get.

  • 32
    Posted September 28, 2009 at 5:47 pm | Permalink

    Poss: Is there a pattern to the way the independents tend to fall? Or do they fall to a donkey vote? (so to speak)

    How much influence, if any, do State politics have on general elections? If the answer is heaps then Victoria would bring home the bacon for the ALP. No bad thing. I’m absorbed, totally, like watching, through a window, the sight of one’s neighbour about to be done in, the astonishing performance of the Coalition. It’s as if they are all standing in a deep viscous glue and immobilized. Of course, they had it coming but it’s so self-inflicted.

    The thought of Christoper Pyne-Hissy Pyne-being flung out of Sturt is one of the best concepts I’ve heard this year.

    More, more!

  • 33
    Posted September 28, 2009 at 10:10 pm | Permalink

    I’m with Jimmy – if SA swings by even 3 or 4%, the tide of that swing will wash Pyne out of Sturt.

    But…But he has “a hide like a rhinocerous”!!

    http://www.abc.net.au/news/stories/2007/11/27/2102684.htm

  • 34
    Posted September 29, 2009 at 8:03 am | Permalink

    Venise.

    [I need to add here that once you remove the Greens…]

    If you add up the preference flows of all minor parties, they generally flow between 55% and 60% in the Coalition’s favour. Sometimes it might be slightly less, sometimes slightly more – but on average they generally favor the conservatives slightly.

    I’m not a big believer in State government performance impacting on Federal government ratings (though plenty of others believe it’s the case)- I think people are generally smart enough to tell the difference, and for those folks that aren’t, their voting intention is so erratic anyway if they arent “rusted on”, that it all comes out in the wash, depending on what their favourite gripe may be this week.

    If you add the Greens in, the minor party pref flows favour the ALP around 65/35

  • 35
    Posted September 29, 2009 at 8:10 am | Permalink

    Tri$tan,

    Unfortunately we don’t have enough data of the type the US has to do that type of 538 regression program with demographic variables. The US has state based demographic data and really regular polling data going back 50 years that can allow for some good, dense demographic modelling – especially since each state is effectively a single electorate.

    In that respect, Australia is really the data equivalent of 1 US state in terms of the total data we have available. Even if we break Australia down into States where the polling is less regular, each state is made up of large numbers of electorates which tend to wash out any given demographic issue anyway.

    For us to do the same type of modelling that Nate does, we’d really need either regular electorate level polling, or small groups of aggregated electorate polling (say, 5 electorates combined).

    But it’s food for thought, and when we get a little closer to the election it might be worth throwing open the gates and seeing what we can all come up with.

  • 36
    Posted September 29, 2009 at 8:27 am | Permalink

    Venise – I should add that *once you remove the Greens* the minor party preferences flow that way.

  • 37
    Aristotle
    Posted September 29, 2009 at 9:32 am | Permalink

    Tim Gartrell on the Oz’s poor analysis.

    http://www.thepunch.com.au/articles/the-one-conclusion-from-newspoll-is-turnbull-is-cactus/?from=scroller&pos=4&referrer=article&link=text

  • 38
    Tri$tan
    Posted September 29, 2009 at 11:35 am | Permalink

    Thanks for the response Possum. I suspected that there wouldn’t be sufficent data – but not undertanding them as intimately as you it’s good to get confirmation.Your data analysis is second to none for Australian politics…

    …which leaves a general question for me: it seems you tend to shy away a little from ‘what does this mean’ questions. Is this intentional? Have I interprete this correctly?

    Especially, ‘what does this mean for the future’ and ‘what does this mean for political tactics’. I suspect this is the schtick of your blog (maximum analysis, minimal opinion, comment mainly on the state of play).

    However, maybe, you could pose some questions and leave it open to the comments section to attempt to answer them? For example, my question on the above is armed with this infomation, and the infomation on your previous post on sentate vote correlation with primary vote, is ‘how should this infomation influence the possibility of a DD election?’.

    Just a thought.

  • 39
    Posted September 29, 2009 at 1:16 pm | Permalink

    That’s intentional Tri$tan – there’s plenty of that sort of navel gazing stuff going around without me having to add to it, especially since the folks reading this are more than capable of coming to their own conclusions that are usually far better than mine anyway!

    On posing a few questions though, that’s not a bad idea at all.

  • 40
    Aristotle
    Posted September 29, 2009 at 1:59 pm | Permalink

    Tri$tan @ 39, I think you’ve highlighted the difference between fact and opinion.

    There’s plenty of sites with plenty of opinions, but precious few with plenty of facts.

    I agree, following a presentation of facts, informed analysis and discussion is worthwhile, but unfortunately it often ends up as uninformed and far from analysis. It’s just a collection of unsubstantiated opinion. Have a look at the responses to Tim Gartrell’s piece in today’s Punch (linked above@38) and you’ll see what I mean. Gartrell, too, asks for informed discussion, but he just gets (mostly) a bunch of diatribe.

    Opinions are like bums, everyone’s got one, but only a few are really worth looking at!

  • 41
    Tri$tan
    Posted September 29, 2009 at 3:42 pm | Permalink

    Point taken Aristotle. Those comments are garbage.

    There might be a better chance here as it at least starts with some numbers and the readership appears different.

    But still, it could be a can of worms.

2 Trackbacks

  1. …] original here:  Election hypothetical: who would win? Related Posts:Question Time has always been a farceWomen at war, the mother of political […

  2. By Election Simulation Under New Boundaries – Pollytics on September 30, 2009 at 2:58 pm

    …] The results – the most likely outcome is the ALP winning 102 seats in Parliament, a 19 seat gain on their current position. The charts come in like this and can be read the same as last time. […

Please login below to comment, OR simply register here :