Labor’s late swing curse

For all their differences, four recent Australian elections – the 2008 WA state election and this year’s SA and Victorian state elections and federal election – have followed a remarkably similar pattern. In each, Labor governments of a range of ages went into the campaign as clear betting favourites. In each, the party led in the polls for much of the leadup to the election. In each, the government was widely viewed as reasonable, although flawed, and in each the party faced some kind of public reservations about its current leader (although in the case of Julia Gillard these arose partly from the way she got the job). In each, the government was up against an opposition regarded as innocuous (and in some cases as a hopeless mess as well).

And in each case, the usual script (in which half-decent governments get swings to them on polling day if they need those swings) was ripped up savagely. Polling in the final week in all four cases showed Labor swinging down towards or below 50-50 territory. The late swing was just enough to dispose of the Carpenter and Brumby governments, while the Gillard federal government lost its majority and survives (for now) in a compromised form. The Rann government copped a massive swing and lost the two-party vote rather badly, but a brilliant marginal seats strategy (and a lot of luck) allowed it to escape with the loss of only a couple of seats.

Much has been said about the causes of the Victorian result, and I don’t propose to provide a definitive answer here. Any commentator can provide a compelling-sounding single-reason explanation post hoc, but it’s much more useful to be able to project such outcomes in advance, which very few commentators did. (Check out Peter Brent here and here and James Campbell here at least correctly picking the trend when most others thought Brumby should win pretty easily.) Some have blamed the result on an “it’s time” factor, as if the government’s defeat after eleven years was somehow inevitable, but the extreme closeness of the result alone shows this wasn’t so. The Brumby government had its share of infrastructure failings and communication problems, but it was no worse and no more tired than the Iemma (remember him?) and Bligh governments were when they were re-elected pretty comfortably. Rather, I suspect the age of the government and its many infrastructure failings, the poor campaign it ran, and the clever campaign by the opposition were all part of the puzzle. Many things had to go the Coalition’s way for them to just get over the line.

One piece of Coalition campaigning was simple but especially clever. It is common for opponents of ageing average-quality governments to say that the government has been there for too long, but not everybody will buy it. So Baillieu and co also focused the voter’s mind on what the government would have been like in another four years:

“Victorians have a choice today between 15 years of more of the same from a tired Labor Government that is not listening to us, or a vote to change.”

On surface level, the math is just plain daft. The voters would not have given Brumby fifteen more years in power had they returned him, but just four. But what the “15 years” line drew attention to was the hypothetical age that the Labor government would reach. The message was that while the voter might not really agree that the government was awful and over the hill right now, if they voted it back in it would certainly soon become so. Aware of the dreadfulness of over-the-hill Labor governments in NSW and Queensland that had been treated forgivingly by their electorates and given one term too many, Victorian voters were given the hint that they could show that they knew better and remove their government now. By a very narrow margin, that’s what they did.

Also, after an especially turbulent and quarrelsome year on the federal front, it seems that the voters had had enough of the bitchy side (ie most of it) of politics. Thus, Labor’s personal attack ads against Baillieu, most of which were poor quality and remarkably light on evidence or substance, appear to have flopped even with Labor voters. Meanwhile, Baillieu’s announcement that the Liberals would refuse to tactically preference the Greens was seen as a welcome break from the usual unprincipled what’s-in-it-for-us style of preferencing bickering. If one looks deeper, it is probably more of the same, but voters generally don’t. In short, the Liberals were sick of giving the Greens something for nothing all the time.

As I write it is likely that Labor has missed out on providing us with the fabulous fun and games of a tied parliament by four or five hundred votes in the critical seat of Bentleigh. An intriguing count is now going on in the Upper House where the Coalition is assured at least 18 seats but could win as many as 21 (which would give it a majority in both houses). One of the fascinating struggles there is Northern Metropolitan, where Crikey founder and local councillor Stephen Mayne, once among Australia’s most habitually unsuccessful election candidates, has pulled off a stunning set of group ticket preference deals. He should actually get well into the count despite polling just 1.05% of the vote, via the cascading preferences of the Carers group, the second Green, Family First, the DLP and the Sex Party. Victorian Upper House elections use a similar system to the Senate (above the line tickets with an option for below the line voting) but unlike the Senate, the voter voting below the line is not required to number (nearly) all the squares.

Currently Mayne’s fate hangs on two unclear issues. The first is whether he can get ahead of the second Green candidate after the first is elected (on present figures he is very well placed to do so, but this may change as more votes are counted). If he fails at that hurdle, it could then be a very close contest between the major parties for the crucial seat. The second test for Mayne is whether he can get ahead of one major or another at the point where only the third Labor, the second Liberal and Mayne would be left in the count. On current figures (as of Wednesday night), if the votes are treated as all being above the line (as the ABC computer does here in an oversimplified model that really shouldn’t be passed off as “Results”) then Mayne makes it with 2577 votes to spare over the Labor candidate. However, in reality, with Mayne’s feeders having 3170 votes worth of below-the-line content between them, Mayne would actually have to substantially outperform the third Labor candidate on the below-the-lines of the various minor parties to win. On past results this is looking difficult for him, and as non-booth votes are added it may well get worse, but the picture is very sensitive to tiny changes in the portion of votes for varying parties. It is similar to the Milne-Petrusma Senate battle in 2004; we might not know the answer until all the data are entered and the magic button pressed (which takes a while), or Mayne’s fate might be clearer before then as the final votes come in.

The Coalition’s win represents a critical blow to the idea that the betting markets close to the poll are generally a very reliable tool for forecasting election outcomes. Especially following the 2004 and 2007 federal elections, there was a lot of interest (including on my part) in these markets, which had seldom missed the mark even on a seat by seat basis, and it seemed possible to call results in well-informed betting markets more reliably than by looking at opinion polls, simply by assuming any candidate longer than about $3 had no real chance. Hiccups like WA 2008, Tasmania 2006 (where Labor retained majority government although that feat had been as long as 9-1 against) and Denison 2010 (Andrew Wilkie winning at around 10-1 after being as long as 50-1) could be explained by inadequate or misleading polling information, and as for the overall federal nailbiter, the favourite did eventually “win” under the terms required for the bet to be paid out. But now we have had an instance of a clear favourite losing a state election for which there was a large volume of good quality polling – with that taken close to the election proving particularly accurate.

Perhaps online election betting markets were all very well when they were relatively little known and most of those betting knew something, but as their early successes have become the stuff of legend, markets have become distorted by people who think you can reliably make money on elections by always betting on the favourites, to the extent that the past success of the markets has become self-undermining. But in any case, if any circumstances are going to try the ability of betting markets to forecast results, swings in the expected popular vote in the last few days of campaigns have got to be it.

Labor’s late swing curse also raises questions about the limits of polling. It’s not news that polls taken more than six months out from an election have very little predictive use in the majority of cases, but it seems that if the incumbent is a modern Labor Party the polls aren’t worth too much six weeks or even six days out; even if the party is ahead it still might find a way to blow it. It’s hard to avoid linking this to the particular style of poll-driven politics loved by Labor operatives and evident in the party’s state of political confusion in the leadup to this year’s federal poll especially. The party isn’t connecting with late-deciding swinging voters and one of the reasons for this is that it is too preoccupied with poll-driven politics and trying to woo Greens preferences, and not focussed enough on the basic concerns that will give it the best chance of being given the benefit of the doubt by the semi-disgruntled late decider whose views do not show up clearly in polls taken well out from the election. Manicuring moderate leads in the polls is no good if by doing so you win almost every poll except the one that matters. Labor needs to recover the courage that it had under Hawke/Keating and the Liberals had under Howard – a long-lived government will sometime have to take decisions that make it fall behind in the polls and place it at apparent risk of defeat. Labor under Rudd and Gillard has so far lacked that courage, and still lacks it.

Meanwhile the Baillieu victory is highly amusing given the last fifteen years of the Liberal Party’s own faction-hackery. The Liberal Party’s moderate wing was given the dalek treatment under Howard, to the point that by the end of his term it was a barely remembered political oddity accounting for a handful of backbenchers. Its most prominent figure was permitted to become leader as an experiment not long after Howard was defeated, but not to lead the party to an election. Now another Liberal “wet” has done what none of the dries or social reactionaries have been able to at state or federal level since 2004 – he has led the Coalition to an outright victory.

…

Having said the above about the limits of good-quality polls taken six days out, it’s hard to pay too much attention to another vague EMRS Tasmanian poll over three years from the next scheduled election. But, for what it’s worth, the just-released November poll ( here ) shows the Liberals with an eleven-point lead when voters “leaning” to a party are taken into account. Given that the “undecided” voters in EMRS polls are typically major party voters, Green support has not changed much since the election. An election held now would probably see a further swing to the Liberals, gaining them more seats but probably not quite winning them majority government. As an election will not be held now, this is not particularly relevant, but the up-and-down course of the last few polls suggests that voter attitudes to the new government are a bit more volatile than just random poll-to-poll variation.

The Examiner ( here ) at the time of writing misreports the preferred-premier scores as 40 for Hodgman to 28 for Bartlett and 19 for McKim; those are actually the November 2009 scores, and the November 2010 results are 39-23-21. Premier Bartlett’s preferred-premier score of 23 points represents his lowest reading ever, his biggest gap behind Will Hodgman and his smallest lead over Nick McKim. However I have pointed out before that preferred-premier is a pretty useless indicator even when taken right before an election, and that EMRS should use approval scores for individual leaders instead. Expect some journalist somewhere to incorrectly and excitedly claim that David Bartlett has a 23% approval rating.

The SMH ( here ) makes the lurid claim that the poll shows just 23% of Tasmanians will vote Labor next election (quite aside from the high undecided rating, who voters intend to vote for at the next election is not even the question!) David Bartlett ( here ) says the polling has been “two steps forward and one step back” – this one is much more like the reverse for his party, but he is right that the polls lack consistent trends. It may be a while before we know what the least disengaged three-quarters of the electorate really thinks about this coalition government.

It’s almost the end of a very long and busy year in Australian and Tasmanian electoral history! Next year will probably be quieter, though we do have the rare spectacle of four Legislative Council elections, all of them potentially interesting, on a single day coming up in May. Finally, I should mention that Tasmania’s only other regular detailed online psephology service, Peter Tucker’s Tasmanian Politics blog, is now offline as a result of Peter’s new and exciting employment. I have enjoyed working with Peter on various electoral adventures over the past few years and am sure his fine contributions to the field will be much missed by readers.