On the night of the 2015 Queensland election psephologist Antony Green was flabbergasted. His modelling was way off. He was sure his computer was broken, it was giving him results he couldn’t understand. As we now know, unprecedented swings tipped out a first term LNP government. The big mistake Antony and everyone else had made was looking at two party preferred polling, and using the previous State election preferences as a guide. This was ill-advised in the extreme. Antony should have realised preferences in the 2012 election were badly skewed by the record result for Newman, reducing Labor to an amazing seven remaining MPs. This doesn’t happen without an enormous swing across the board. An adjustment for preferences was needed.
Now we are entering the final phase of the federal election campaign, and pollsters who aggregate many polls are more cautious. Most are introducing “adjusted results” where more accurate polls are weighted positively whereas erratic polls are given less weight in the final aggregated prediction.
In the campaign’s fifth week, after Essential Research’s latest result recorded the gentlest of nudges in favour of Labor, William Bowe’s BludgerTrack poll aggregate has gone the other way on 50.5-49.5 to the Coalition. This was mostly due to a ReachTEL result that would have come out at 52-48 in favour of the Coalition if previous election preference flows had been used, as opposed to the headline result of 50-50.
The Guardian is also predicting a narrow Coalition victory due to the LNP needing only 49.1 per cent of the 2PP vote to win. In campaign week 5 their aggregated result is sitting, in adjusted terms, at 50-50.
With Antony Green’s Election Calculator, you can get any result you want, just by adjusting the swings to suit. One factor which becomes increasingly obvious when you play with the sliders, is that New South Wales is where the election will be won or lost, simply because it has the most seats. If you set the calculator to adjust regional swings, you will probably come to the conclusion that Labor needs around a 4 per cent swing in NSW to win, a tall order indeed. According to polling aggregated by Wikipedia, the Coalition vote in New South Wales is down only 2.2 per cent, Victoria down 2.9, Queensland down 5.1, Western Australia down 8.7, South Australia up 1.8, Tasmania up 1.7, and ACT and NT down 3.1, for an overall decline from the 2013 result of 3 per cent.
Wikipedia graphs an Australia-wide aggregated poll without adjustment, just charting the raw figures. Their 2PP figures are 50.5 for Labor to 49.5 for the Coalition.
Sportsbet is said to be a reliable guide, but as we saw in the Queensland State election it can be wildly wrong. Presently it has the Coalition as the odds-on favourite at 1.17 while Labor has drifted out to fives (5.0). ‘Hung parliament’ has also gone off the boil at Sportsbet. It’s interesting that the gamblers have gone in the direction opposite to that of the raw polls.
With recent history showing upsets and great volatility, there is no reason to suspect that this election will be different, so expect a few upsets.
- Poll Bludger »
- The Guardian’s Poll of Polls »
- Antony Green’s election calculator »
- Wikipedia’s aggregrated poll »
- Sportsbet »
- Polls and the pendulum »
- How can Australia reduce the risk of another ‘systemic polling failure’? »
- Over the last 30 years, a fifth of polls have called the wrong winner. Here are 3 things poll-watchers need to understand »
Record field for Queensland Senate
Meanwhile Cameron Atfield reports in the Sydney Morning Herald (11.6.16) that Queenslanders will have a record number of Senate candidates to choose from at next month’s federal election, following the ballot draw in Brisbane on Friday afternoon.
The white ballot paper for the full Senate election will feature 122 names, 103 of whom would appear in grouped columns for above-the-line voting, which was a 49 per cent increase from the previous poll.
See the the link below for the full list.