There is a commonly held perception that for the ALP to win the 2007 Federal Election, they must gain somewhere in the vicinity of 51% of the Two-Party Preferred (2PP) vote.  This perception is based on the 1998 result, where Kym Beazley exceeded 50% 2PP, but did not win.  Recent election results have indicated that the electorate has changed, and the result will not be this way in 2007. 

By analysing past election results, from 1996-2004, we can use probability formulae to predict the potential result for any given seat, and count up the predicted seat wins for each party, based on these past results.

New South Wales

 The graph for New South Wales shows that it is a fairly even state for both major parties.  Whichever way NSW goes, so goes the country.  This can be seen by comparing this graph with the national graph further down on this page.  The two independent seats of Calare and New England have been placed in the "in doubt/other" category.

Victoria 

 Victoria is no doubt a happy hunting ground for the ALP.  For the Coalition to win Victoria, this analysis shows they must get around 54% of the 2PP vote nationally. Note also the relatively small number of "in doubt" seats in comparison to New South Wales and Queensland (below).  This indicates a "divide" of sorts in Victoria, where seats are fairly predictable for one side or the other.

Queensland

 As much as Victoria is a Labour heartland, Queensland is even more so for the Coalition.  This graph is a veritable sea of red, and shows that the ALP must get around 55% of the 2PP nationally just to break even in the Sunshine State.  Even if the ALP has a landslide 56-44 result, the Coalition is still predicted to win 11 of the 29 seats.  If the reverse occurs, say a 56-44 Coalition landslide, the ALP would only hold one seat in Queensland (Rankin).

Western Australia

In WA, we find a more even battleground, although still slanted slightly towards the coalition, particularly if the election goes strongly their way.

South Australia

South Australia is a slightly more volatile state, showing more significant changes as the 2PP slides from one party to the other.  Still, it can be argued that this state also leans towards the Coalition.

Tasmania, the ACT and the NT

 Grouping Tasmania, the ACT and the NT together, we can see that these states, with a combined seat count of 9, are definitely Labour territory, no matter what the election result is. 

The Nation

 Nationally is where the real trend becomes apparent.  A 50% result for the Coalition shows 63 seats predicted for the coalition, 75 for Labor, with 13 in doubt/other.  Only at 51.5% of the 2PP does the Coalition actually gain a majority of seats.  This is totally the reverse of 1998 results. 

Here is the table of the actual seat count predictions: 

 On these numbers, the Coalition will not gain enough seats if they only claw back the 3-5% that they have been trailing by for most of the year.  Even from the recent Galaxy poll of 53-47 (October 19), they would need to gain a further 4.5% to win the election.  If the polls creep further out from here, say back to 55-45 where they have been for most of the year, the task gets even harder.

The Nation - with "in doubt" seats distributed 

 The difficult task that the Coalition faces can be seen further by distributing the "in doubt" seats evenly according to the proportions of definite gains.  This shows that a 51.5-48.5 result would only give them 76 seats, with the ALP having 75, and 2 independants.  Thus, a win by the barest of margins, even at 51.5% of the 2PP vote.

The final seat count table for each 2PP result is:

It will definitely be interesting to see if the past four federal elections are indeed an accurate guide to the 2007 result. 

- Scott McCamish

Comments (51)

2. Explanation Written by This e-mail address is being protected from spam bots, you need JavaScript enabled to view it , on 21-10-2007 23:24 The difference between the two models are:    1. Antony green's calculator applies a swing evenly across all seats, regardless of the current margin in that seat.  2. My model uses past 2PP results to establish a longer term trend on a seat-by-seat basis.    Because swings are never uniform from one seat to the next, I aimed to produce a more granular model, focussed not on applying uniform swings, but looking at how the 2PP vote for a given seat compares to the state 2PP result. I found, in most cases, remarkably similar results from one election to the next.     These trends are then plugged into a probability formula, and that probability result is used to say that a certain seat is "likely" to fall one way or the other. Whether it does or doesn't is immaterial. My model is designed to predict the overall seat count, and not an individual seat result.    I intend to publish a table of seat 2PP predictions in the last week of the election campaign, base on an average of the major polls of that week. We'll then have a checklist, and see how this model goes in terms of predicting actual seat results.    Thanks for the question. Report