NEWS AND VIEWS
Monday, 13 August 2007
Pay Your Money and Take Your Choice
Subscribe to Crikey and you read on Friday that the latest Morgan Poll has the ALP two party preferred vote at 58.5 to 41.5% for the Coalition; a drop of 0.5% for the ALP since the last Morgan telephone poll. Fancy the Sunday Telegraph as your journal of record then how about a move of seven percentage points to Labor's Maxine McKew shown by Galaxy in the seat of Bennelong giving her a lead over John Howard of 53% to 47%. Put your trust in the Fairfax papers and you find this morning that the Coalition, while still well behind, has improved its position by three points to only be behind 45% to 55% with Michelle Grattan telling us that in Western Australia the positions are reversed with the Coalition on 56% to Labor's 44%. That must be confusing to believers in the Westpoll published in the West Australian on Saturday which had the numbers on Saturday as Labor 54% to the Coalition's 46%. In despair perhaps you would prefer to revert back to last Tuesday's Newspoll in The Australian and settle for the Labor national figure of 44% for the Coalition and 56% to Labor.
Or perhaps you would prefer to block out all stories about opinion polls or at least all those that try and pretend there is some significance in the myriad movements up and down. For the truth is that all samples of opinion by all the pollsters fluctuate. A margin of error is the statistical nature of the beasts and journalists earnestly searching for a trend invariably find something that is not really there at all.
Here is a graph of movements this year in the AC Nielsen measurement of the two party preferred vote of the Coalition.
It is easy enough to plot a trend line or two and we show a linear (straight line) version as well as a lovely curved one known as a polynomial but the statisticians have a little test to see if they actually are significant called r squared. The closer that figure is to 1 the greater the significance of the trend line but when it is only around 0.5, as in the AC Nielsen example, you would not want to be relying on the trend line.
Plot the same kind of graph for Newspoll and it is hard to see a trend at all in this year's figures.
If you are a straight line believer then you would have the Coalition flat lining around a 43% share of the two party preferred vote. Be an adventurous polynomial type and you might convince yourself that John Howard and his team are on a step rise. Check the r squared number and you should realise that there is no significant trend line at all.
Perhaps the best we can say about the polls between now and election day is that all the major ones are pointing towards a Labor victory although the recent results give us no clue as to whether that will continue to be the case between now and polling day.
For my part I will be sticking to the collective wisdom of the market which provides the basis for the Owl's Election Indicator. This morning it shows the Coalition with a 38.8% chance of being returned to office compared to a 60.2% chance of a Labor victory.
Well, Perhaps She Could
The election indicator was in advance of the opinion poll in signaling that Maxine McKew was looking better in the battle for Bennelong. The day after we posed the question "she couldn't win, could she?" out came the Sydney Sunday Telegraph with its Galaxy Poll showing the Prime Minister, John Howard, running second in the two horse race. Confirmation of the PM's concern about his future came with the backdrop he used at the weekend for his little grabs for the television news. Forget entreaties for people to vote Liberal, the message was vote Howard for Bennelong. The new poll did move the Indicator even further in Labor's favour.
The latest AC Nielsen national voting intention figures out this morning did little to move the Owl Indicator with Labor now at a record high point.
We will now start updating the Owl Indicator daily on our website which will also contain election betting information.
The Market Still Lagging Behind the Opinion Polls
The movement might be in the direction of Labor but the market is still much less optimistic than a believer in opinion polls would be.
The average split of the two party preferred vote according to the four national polls is Coalition 44.3% and Labor 55.8%. The break even result in terms of seats won comes if Labor gets 51.3%. Past history suggests the standard deviation of the polls this far out from an election is around 4 points. That puts the probability of a Labor win at around 87% rather than 61% as the market currently has it.