CAN Labor win majority? Will we have a hung parliament? Just what will the election result be? Well, we could all just chill out and wait three weeks … but that’s no fun.
The most recent TasPoll published in The Mercury has added nothing to our crystal ball gazing. Kevin Bonham has saved me a good few paragraphs by pointing out what the journalists fail (or aren’t willing) to grasp: the sample size for individual electorates is too small for any meaningful interpretation. Below is a table comparing the electorate divisions and the state totals for both EMRS and TasPoll. At electorate level the results are all over the place yet the whole-of-state results are (almost) identical.
I think this is what we know so far:
• Support has dropped off Labor and gone principally to the Greens. For the 18 months or so following Jim Bacon’s departure in mid-2004, Morgan Polls consistently showed Labor in the mid 40s, the Liberals in the low 30s and the Greens in the mid to high teens. The above table demonstrates that, in the shadows of the election, the Liberals have not budged while four percentage points or so has shifted from Labor to Green.
• The undecided and swing vote is so high that the election is still to be won or lost by Labor. Although the 25 per cent undecided records by EMRS is way above what most polling outfits normally return, TasPoll’s eight per cent does confirm that there are many voters still to make up their minds. And for those undecideds …
• … the “defining moment” is still to come. Sometime during most election campaigns there is an event or moment upon which the outcome is seen to turn. (Granted, such moments are often only apparent in hindsight.) For the last federal election it was the PM’s forestry trump card, and for Western Australia it was Colin Barnett’s grand canal. The Mercury’s Sue Neales may have been less than responsible in her poll coverage item, but is more perceptive here by saying the campaign so far has been predictable and hum-drum; the fire is still to be lit.
Several months ago I made an attempt to convert election predictions into probabilities. Since then I’ve messed about with the model a bit and sought some advice from two experts (http://econrsss.anu.edu.au/~leigh/pdf/ElectionForecasting2005.pdf) in modelling election outcomes, Andrew Leigh from ANU and Justine Wolfers from the University of Pennsylvania . The main limitation with applying probability analysis to elections is that the outcomes are highly correlated across electoral divisions (whereas probability theory requires each event to be independent). This means the model cannot be used in single seat elections; well, not without some serious statistical allowances being factored in for both correlated and independent forecast errors. (Maybe I will get around to that one day.)
But I still think the model has some validity for Hare Clark because the fifth seat outcomes can display a relatively high degree of independence. These contests are often dog fights where personalities, local campaigning and intra-party rivalries dominate.
Anyway, it is worth giving it a go to see what happens. In the table below I have modelled the probabilities of Labor winning the fifth seats. The assumption I have made, to simplify the model, is that Labor will already have won two seats in each electorate. (OK, some would argue with that; but most commentators are willing to concede Labor will win at least two seats in each division.) The probabilities I have assigned, of course, are entirely mine. I have had the benefit of two excellent “form guides” here
and here although that is not to say the authors of those would come up with the same quotients as I have settled on here.
What I have done is largely ignore the electorate based polling (for reasons mentioned above) and just used some judgment. In Braddon, Labor should win the fifth seat, providing the Greens miss, and I give the Greens less than even money to make it. Bass does seem a foregone 2/2/1, but some slight allowance must be given to Labor winning a third seat if the voters swing strongly their way in the final weeks of the campaign. The other three electorates, to varying degrees, give Labor a chance at a third seat. I think Labor deserve to be favourites in Lyons where they traditionally do well. Denison looks very hard because the two issues where Labor is really struggling, health (particularly the RHH) and Paul Lennon’s personal style, will bite the hardest in the capital city. It is not that much better in Franklin, although Labor will have the advantage of three ministers on the ballot paper.
Labor must win a third seat in at least three out of the five electorates to gain their majority. The table shows there are 32 possible win/loss combinations across the five electorates. Sixteen of those provides Labor with the result they need (that is, three seats or more). The table shows that the sum of probabilities gives Labor a 32.67 per cent chance of a majority. Expressed as odds that is about $3.20 (or 9/4 for those who still bet with the satchel swingers). Centrebet
is accepting bets on the Tasmanian election, which would have to be a first. A Labor majority outcome was $4 last time I looked. Good odds? Maybe.
The technique has its limitations, as noted above. Also, there are sensitivity considerations: “What happens to my results if I make reasonable adjustments to my assumptions?” Relatively small changes to the probabilities assigned can make a large difference to the outcome.
Nevertheless, for better or worse, this is my best guess. Do you agree with my probability assumptions? If not, grab a calculator and a glass of whatever (you may need the whole bottle) and have a go yourself.
Peter Tucker is a PhD candidate at the University of Tasmania’s School of Government and a past advisor to the Tasmanian Liberals.
7 Bridge Street
Taunton
Somerset UK
[email protected]
First published Feb 27, republished Feb 28 with active links.
