With a year (and bit) to go what do the polls actually mean?

What do the current Federal election poll numbers actually mean for the result of the next election – particularly now we are a bit more than 12 months out?

Nate Silver’s use of Bayesian theory to predict election results has been referred to by the blog from time to time. Last time it was to use Australian data and Silver methodology to give a rough prediction of the probability of the Victorian ALP winning last year’s State election. The blog, based on Silver, put it at about 65%. Bookies were even more confident and started to pay out on the ALP before the election.

In his book The Signal and the Noise he says: “the FiveThirtyEight models (reference to the US Electoral College membership) provide much of their value (in assigning probabilities to opinion polling). It’s very easy to look at an election, see that one candidate is ahead in all or most of the polls, and determine that he’s (sic) the favorite to win. (With some exceptions this assumption will be correct.) What becomes much trickier is determining how much of a favorite he is. Our brains, wired to detect patterns, are always looking for a signal, when instead we should appreciate how noisy the data is.”

In his book Silver produces a table on the probability of a US Senate candidate winning based on the size of lead in polling average according to the time before an election. For instance a 5% lead one day out gives a 95% probability of winning and a one point lead one day translates into a 64% probability of winning. Now the normal provisos apply: the US poll data is much richer and the data base larger; and, voting is not compulsory.

However, it is a useful rough and ready guide to Australian elections even though there is an important proviso – polling results, despite the comments of Australian political pundits, are largely irrelevant predictors more than 12 months out. Nevertheless that magic 12 month figure is rapidly approaching in Australia and in just three weeks we will be two years from the 2013 federal election.

So taking an average of the latest Morgan, Essential, Fairfax-Ipsos, ReachTEL and Newspoll results (with adjustments for differences between actual preference flows in the last election and polling on how people are now saying they will allocate their preferences) the current Australian situation is 55% ALP and 45% LNP.

Now let’s go back to Silver’s table. A lead of between one and 5% one year out translates into a probability of between 52% and 59% of winning although, as Silver says, that’s “barely better than a coin flip”. But when the difference is 10% the probability is 67% – which is much, much better than a coin flip. Of course there are variables – the Libs might change leaders; events dear boy, as Harold MacMillan would say, might intervene; and, the polls might be wrong. This last explanation is the one favoured by Crosby-Textor after the recent Conservative win, which some seem to think they “masterminded” although George Osborne probably thinks he made a somewhat larger contribution. And most voters would say Ed Miliband and the SNP might have had something to do with it as well.

Anyway, with the general robustness of aggregated polls in Australia, plus compulsory voting, Tony Abbott would be wise not to place too much credence in the Crosby-Textor hypothesis even though he does tend to be more a man of faith than facts.

So – 12 months out is just three weeks away and the probability, according to the Silver data, of Labor winning is 67%. Labor might be like the Boston Red Sox or Chicago Cubs, the prime Silver examples of how probability theory can come unstuck, and as Labor has had a touch of the Sox and Cubs about them from time to time the outcome might be the same. But the probabilities suggest the ALP would be a better bet than Abbott – although the best betting odds currently being offered don’t reflect that and may provide a good opportunity. The odds are: bet365 at $2.37 for the ALP and $1.53 on the Libs. Tom Waterhouse is offering $2.05 ALP and $1.77 on the LNP and Sportsbet is offering $1.65 on the LNP and $2.25 on the ALP. Sportsbet is also offering odds on who will be the Coalition Leader at the next election and the odds there are: Abbott $1.40, Turnbull $4.50, Morrison $5.00 and Julie Bishop $7.00. By any measure the Morrison and Bishop odds are extremely attractive although the $251 being offered on Sophie Mirabella is mean beyond belief.

Meanwhile the Canadian election has just got interesting with Stephen Harper starting the October 19 election campaign on August 2. The rationale is that it worked against Michael Ignatieff in the last election and, more importantly, it stops the recent tide of anti-Harper advertising being run by unions and others. Harper’s Conservatives are expected to spend more than the NDP or Liberals so that is an extra advantage. However, long election campaigns don’t always favour the incumbent. In the 1984 Hawke-Peacock campaign Hawke opted for a very long campaign imagining, apparently, it would be like a coronation tour rather than a campaign. As sure as night follows day, as Peacock said so often during that campaign, it wasn’t.

In Harper’s case the most recent poll suggests his Progressive Conservatives have fallen into third place behind the NDP (33%), Liberals (28%) and Harper’s lot (27%). NDP Leader, Tom Mulcair’s preferred Prime Minister numbers are ahead of both Trudeau and Harper although Trudeau’s have improved significantly since the election was announced. In contrast the Progressive Conservatives poll numbers have dropped since the election announcement and Harper’s approval rating is now 32% and his disapproval rate is 62% – up there with his soul mate Tony Abbott.

Why hark on Canada? After all, Harper has been in power for a while and longevity might be playing against him. But it’s mainly because Harper and Abbott are so alike in terms of personality, approach and policies – just as Canadians and Australians are alike in many other  respects. And a loss for Harper would be interpreted by some as a significant omen for Abbott.

Incidentally, if you are thinking there might be an early Australian election the odds of the LNP winning an election in the next three months, based on the polls is about 13% according to the Silver table. And the Sportsbet odds on the Canning by-election are ALP $4.00 and Liberals $1.35 which, despite the 2013 Canning result, are reasonably attractive for someone wanting a wager on Labor winning.