Jan 4, 2010

New Year Election Simulation

Over the weekend, Newspoll via The Oz brought us their quarterly demographic breakdowns which you can see in their entirety

Possum Comitatus — Editor of Pollytics

Possum Comitatus

Editor of Pollytics

Over the weekend, Newspoll via The Oz brought us their quarterly demographic breakdowns which you can see in their entirety over here – although beware, that’s an 11MB pdf file. Those Newspoll results cover the period from October through to December, giving us a mega sample on the vote estimates of 8059. We can combine those results with the Nielsen polls over the same period – 3 polls with a total sample of 4000 – to give us a rather large pooled sample for the October-December period of 12059 respondents. On the sidebar to the right, the new voting intention estimates for various demographics have been calculated using this new pooled Newspoll/Nielsen data – which probably makes it the most accurate voting intention data by demographic cross-tabs in the country. It's certainly worth having a bit of a squiz at. The other thing we do with the quarterly data is run a monte carlo simulation to give us a reading on what a new Parliament would most likely have looked like if an election were held over the Oct-Dec period and where the result was consistent with the polling. The simulation works by using the state level breakdowns of the two party preferred voting estimates. We first calculate what the swing currently is in each state – so using Victoria as an example, the combined Newspoll/Nielsen data currently has Victoria sitting on an ALP two party preferred vote estimate of 59.4%, a 5.1% swing towards them since the 2007 election. For every seat, we create a swing for that seat that is a probability distribution - where the mean is the state level swing (for example, 5.1 in Victoria) and give it a standard deviation that is a combination of both the margin of error of the pooled state sub samples and the historical standard deviation of federal election results in that state. We then draw a random number out of that probability distribution and add it to the two party preferred result in that seat at the last election. When we do that once for all 150 seats, it gives us 1 simulated election where we add up the number of seats that are now sitting in the government’s column. We then do that another 20,000 times, tally up the numbers, turn them into a probability distribution and we have our simulated election result. If an election were held over the October-December period and the results were consistent with the polling, the most likely outcome would have been a new Parliament where the ALP had 107 seats, the Coalition had 40 seats and the Independents held 3 seats. In those 20,000 simulated elections, the worst result for the government was 92 seats while the best was 121 seats. The median and mode results were 107 seats while the mean result was 107.2 seats. The distributions of the result come in like this.

simseatfreq simcumulative

The first chart is just a histogram of the results, while the second chart gives you the probability (left hand axis) of the ALP getting at least a given number of seats (bottom axis) in the new Parliament were an election held last quarter.

Free Trial

Proudly annoying those in power since 2000.

Sign up for a FREE 21-day trial to keep reading and get the best of Crikey straight to your inbox

By starting a free trial, you agree to accept Crikey’s terms and conditions


Leave a comment

11 thoughts on “New Year Election Simulation

  1. Martin C. Jones

    Nice methodology, Poss! How are you combining the margin of error of the pooled state sub samples and the historical standard deviation of federal election results in that state to get the standard deviation? And what sort of probability distribution are you assuming for the swings?

  2. praha

    This is bad. This is very bad.

    I’m not a fan of the Liberal Party, but it is really not good to have any party win over two-thirds (and nearly three-quarters) of the seats in parliament. I don’t have any solution, but this really is a problem for Australia. We need some diversity in parliament, and an effective opposition which is a credible “alternative government.”

    Forty to forty-five seats out of 150 doesn’t cut it. You can’t be an effective opposition with that many seats, let alone be a credible alternative government. If we get the result that Possum is projecting (and I can’t honestly seeing it being significantly different), we’ll effectively be choosing to be a one-party state. And that’s not good.

    I wish there were some realistic alternatives.

  3. tomasso

    I can’t make out a LP (or coalition) strategy since MT was ousted. Is there any? Is it something for later in the year? Is it a rebuild on JWH foundations?

    I see plenty of tactical jousting, but without careful observation and some sense of direction I expect there will be more tears…


  4. Michael McG

    The Coalition Militia, aka Abbott’s Shadowy Ministry, are all falling over themselves so quickly to get on the Abbott Train that they, in the style of Scott Morrison, have lost the little ‘l’ that used to characterise some of them as liberal Liberals. The stupidity is, that in the absence of contemporaneous polling over the Christmas/New Year break, they have no idea whether the “train” has even made it out of the station. I expect, when voters begin to register their responses to all that Abbott has had to say so far, and the locomotive is revealed to be more loco than motive, then we shall see the coat-tail riders like Morrison looking for another platform to stand on, eyes pleadingly cast down the line for the next leadership express to roll in.

  5. shepherdmarilyn

    With Abbott spraying all over the place about anything and everything and bringing back the three dead people walking it is no wonder they are getting nowhere.

    I just got an email from Scott Morrison who seems to think “everyone has the right to seek and enjoy asylum from persecution in other countries” is MY VIEW and he begs to differ.

    Shows how out of touch with reality these people really are.

  6. Seats at Risk – Quarter 4 2009 – Pollytics

    […] Politics, elections and piffle plinking Skip to content « New Year Election Simulation […]

  7. Possum Comitatus

    Alex – there was a 100% certainty that Labor *would have won* if the election result was within the combined sampling error and historical variation of the poll results.

  8. Peter J. Nicol

    So, that looks like about another 5 seats from the last one, iirc?

    Pollytrack will indeed be interesting to see, but this looks like a complete blowout.

    Who is giving what odds that people skills doesn’t make it to the election?

  9. alexander white

    So, does this mean there is a 100% certainty that Labor will win?

  10. Possum Comitatus

    Sure is Cud – it’ll pop up sometime over the next few days.

  11. cud chewer

    Any chance of an updated Pollytrack?

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