Newspoll released another of its state polls taken over the course of the past 3 months for South Australian voting intention. The poll surveyed 871 South Australian voters and returned top line primary vote figures of Liberal 43%, Labor 28%, Nationals 1%, Greens 11% and Others 17%. Newspoll extrapolates a Liberal TPP of 57% from these primary vote figures.
Just as we have done for other states, we will do a seat projection on these poll figures simply because state polls are few and far between. The seat projection is similar to previous seat projections for other states where we sum the probabilities of seat being won by Labor and the Liberals. As we are only using one poll with a relatively small sample of 871 (which equates to a MoE of 3.5%), this is taken into account in the seat projection by creating a larger standard error in the swing which ameliorates against wild swings due to polls with larger margins of error.
We have left the seats of Mount Gambier, Frome and Fisher out of the major party pool of seats as they were won by independents at the last election, We assign these seats to Independents in the projection unless specific seat polling evidence which indicates defeat for the incumbent independent comes to light and to date there has been no such poll for these three seats.
Usually we apply the uniform swing in our seat projection model, unless if there is another psephological feature which may interfere with the application of such a uniform swing. In South Australia there is such a feature at play.
2010 Election – Uneven Swing
At the last election the statewide swing against Labor was large at 8.4% but it also was greatly uneven, more so than is normally the case. This was due to Labor concentrating its resources in its marginal seats to defend its parliamentary majority. This tactic has been referred to in the media as “sandbagging”.
It was very effective in the seats were the incumbent MP was a sophomore. The deviation from the state wide swing in these seats was so large that it far exceeded the sophomore effect we usually see at state elections. In Labor’s two most marginal seats going into the last election; Light and Mawson, the party won TPP swings to it at 3.2% and 2.2% respectively.
Overall the TPP swing seemed to rack up in Labor’s safe seats. The scatterplot of the TPP swing to the Liberals and the pre-election Labor TPP is below. The seats which were held by independents prior to the election were omitted from the scatterplot.
As you can see there is a relationship between the size of the swing and the Labor TPP. However this relationship does not explain all, or even most of the variation in the TPP swing. Never the less it was a significant factor that needs to be taken into account. The three significant outliers at the bottom of the plot represent the safe Liberal rural seat of Flinders where the Nationals are a factor and the two sandbagged Labor marginal seats; Light and Mawson. The swing in these three seats cannot be explained at all by the overall trend.
However, was the swing back to the Liberals in 2010 an unwind of the pro-Labor swing of 2006? Was that swing uneven? A scatter plot of the TPP swing to Labor in 2006 and the pre election TPP is below.
Here was see a possible positive relationship between the TPP swing to Labor and the Labor pre-election TPP but it definitely is not as strong as the inverse 2010 relationship.
So as to ensure that we are not unduly reacting off a phenomenon that occurred in 2010 due to a reaction to vote movements that occurred in 2006, a 2006-2010 overall swing has been calculated for each seat where a major party probability is to be calculated. The deviation of this swing from the state-wide overall swing of 1% to the Liberals is assigned to each seat in the model.
From these deviations, 50% of the negative value of the deviation is added to the overall swing for the seat. This is done in expectation that seats which have not swung back to the Liberals will do so at the next election and vice versa.
The reason we have chosen 50% is to hedge against a total unwind of the deviation from the overall state swing. Seats may move relative to each other due to demographic reasons and we cannot say for certain that each seat will revert to its previous vote relative to the overall state.
With the addition to the overall swing we can model the swing at the next state election with reference to previous movements of the vote that were peculiar to that election.
Now on to the seat projection results.
Unsurprisingly the Liberals have a healthy majority under this projection given the have a 14% lead on a TPP basis.
Two Final Notes
- Incumbency effects have not been included in this model as we don’t know which incumbents will re-contest and which shall retire.
- The redistributed seat figures for the recently concluded redistribution have been used in the seat projection model.