Probability theory and the 2019 election

What do the latest Federal opinion polls mean? The superficial answer derived from political commentary is: almost anything – with the list of possibilities at least as long as the list of journalists reporting on politics.

A better approach is to use probability theory, which the blog has written about before, particularly in the light of Nate Silver’s work on probability – work derived from Bayesian theories and his long experience in sports and poker.

Another approach – illustrated by Tim Colebatch in an enlightening piece of analysis of the recent Victorian State election in Inside Story – demonstrates that while State and Federal elections may be different they can still be significant indicators of possibilities – especially given that the ALP could win government Federally just with outcomes in Victoria. Of which more later.

Silver says in his book The Signal and the Noise “that you’ll see pretty much everything at least once.” The blog hopes that means it will one day see Collingwood win a Grand Final by a point with a kick after the siren. Although, perhaps in a parallel with the Collingwood situation, Silver says he wouldn’t question anyone “who says the normal laws of probability don’t apply when it comes to the Red Sox or the Chicago Cubs.”

Silver sums up the experience of seeing everything at least once as: “Play enough poker hands, and you’ll make your share of royal flushes. Play a few more, and you’ll find your opponent made a royal flush when you have a full house.”

Silver copped some flak in the last US election when he said the probability was that Hilary Clinton would win the last Presidential election. He was on the money in some respects given the overall popular vote and the freakishly narrow results which delivered Trump the Electoral College. Silver did defend himself, however, by asking if any of his critics would have played Russian roulette at the odds he was offering on a Trump victory. It was, as the US and the world now knows, an apt analogy.

But Silver has produced an excellent chart (to which the blog has referred in the past) which seems very useful given the forthcoming NSW and Federal election. It describes the probability of a US Senate candidate winning, based on the size of the lead in polling averages at periods from one day to one year. Naturally the longer the period before the possibility of the poll being an accurate predictor is about the same as the odds of a coin toss.

Now readers will say US elections are different – which is an understatement – but the Silver analysis could be robust when applied here because of compulsory voting and some other demographic factors – such as the fact that a poll for a Senate election in California, Texas, Florida or New York is based on a population size greater than, or close to, Australia’s population.

If you take the Silver chart, and extrapolate the recent opinion polls in NSW and nationally, it’s probably a toss up in NSW although the ALP’s chances could be as high as 58%. However, volatility in National Party seats and recent by-election results complicate (significantly) the outcome.

Using an extrapolation from the Silver chart for the Federal election – now presumably a bit over two months away – the probability of a Labor victory is around 80%. It is a probability and not a certainty but one would rather be in Shorten’s place than Morrison’s. And just in case you are seduced by the argument about Bill Shorten’s unpopularity – it didn’t save Jim Callaghan against Maggie Thatcher and on some measures Tony Abbott was one of the most unpopular Opposition leaders in history when he was elected Prime Minister.

When it comes to State and Federal elections there is no strong historical correlation between the outcomes. However, in Victoria Labor has been in government for most of the time since Jeffrey Kennett lost and Victoria seems to be becoming more progressive. John Howard dismisses the State as Australia’s Massachusetts conveniently overlooking the fact that the US State has a Republican governor and introduced a precursor to the Obama health care plan under another – Mitt Romney. The Howard comment is as silly as comparing Queensland to Mississippi even if that did not once seem  an outrageous comparison.

There has been some individual seat polling in Victoria including some which showed Higgins and Kooyong in doubt and, while this seems possible at present, Tim Colebatch points out that “individual seat polling in Australia has proved erratic.”

In contrast to this approach he has produced “the first published estimates of what the 3.7 million actual votes cast in Victoria’s state election would mean if replicated on the new federal boundaries.”

He concludes: “The Liberal Party could be reduced to holding just seven seats in Melbourne if the federal election sees Victorians vote as they did at last November’s state election. Transposing the state vote to federal boundaries would see the Liberals lose five seats in Melbourne and a sixth, Corangamite, in the hinterland of Geelong.

“Of their other seven seats, three would be left on a knife edge, with another three potentially within Labor’s reach. On state voting figures, Labor would have won the federal Liberal seats of Corangamite, Dunkley, Chisholm, Casey, Higgins and La Trobe. It would also have come within 1 per cent of winning Deakin, Flinders and Goldstein, and within 2.5 per cent of winning Kooyong, Menzies and Aston.

“On the state voting, the Liberals’ only safe seats in Victoria are now the Western District seat of Wannon and the West Gippsland seat of Monash (the renamed McMillan). All of the party’s other twelve seats in Victoria are at risk of being lost,” he said.

So, take your pick of the extrapolations and keep in mind that the bookies (heavily influenced by insider betting) are showing the same likely outcome.