A by-election circus is upon us in Aston, Victoria, where former minister Alan Tudge is retiring. The carnival metaphor derives from the fact that commentators and party insiders set up these events as “tests” of leaders and parties. They are even seen as predictors of general election contests. This one played second fiddle to last weekend’s NSW state election, but has since come into its own.
Some opinionistas believe by-elections don’t just measure the electoral temperature but also influence the future. So, for example, the Financial Review’s Philip Coorey writes that the July 2001 by-election in the same electorate — which the Liberals retained after a modest swing to the opposition — didn’t simply show that the Howard government was still competitive but also “created a momentum swing” all the way to the November general election.
It’s because the political class obsesses so much over these electoral events that they can actually influence the future — at the political level. Just ask Malcolm Turnbull, whose prime ministership was fatally damaged after a — wait for it — 3.8 per cent two-party-preferred swing to the opposition in Longman in July 2018. And it was the current Liberal leader who led the charge against him.
But even here the political bubble can’t get its story straight. It made a massive fuss about Longman but found little to see in the larger, 4.8 per cent swing in Bennelong seven months earlier. (Is there an irony in the fact that Labor won Bennelong but not Longman when it took office last year?)
By-elections are like massive surveys with non-random samples, except we don’t really know the question being asked. It’s certainly not “which party would you vote for at a general election?” And because nothing as important as who will govern is at stake, electors are liberated to cast their votes based on other matters. Candidates can make more of a difference in by-elections, as can mini-scandals. Voters can send a message to one side or the other. And a lot of other things we can’t identify also play their parts. For elaboration of the meaninglessness of by-election results, see my piece before Eden-Monaro in 2020.
Now, it’s true that very unpopular governments tend to suffer big by-election swings. But the intuitive extension — that well-regarded ones, as the Albanese government can reasonably be characterised, can expect to do well — is not borne out by the statistics. (Blogger Kevin Bonham has found “a reasonable correlation between federal government polling and by-election results” — see Twitter thread starting here — but his line derives all its power from the unpopular government sub-sample. The left side of his graph, containing data points with governments performing well in national opinion polls, has rather random by-election swing outcomes.)
One hundred and fifty-eight federal by-elections have been held since Federation, with another two vacancies uncontested. (These include supplementary elections and re-elections — in fact, any House of Representatives vote held apart from a general election.) It makes sense to identify a two-party-preferred swing in eighty-nine of those held since the introduction of preferential voting in 1918. (Only in those eighty-nine were Labor and the Coalition or its predecessors left in the final count for both the by-election and the previous general election.)
The swings in these seats range from 13.4 percentage points towards the government in the Australian Capital Territory in 1970 to 20.1 points in the other direction in Wakefield, South Australia, in 1938. Both were triggered by deaths, both were under Coalition governments and neither portended much about the next general election. (“Coalition” here means the various versions of the anti-Labor major parties. Pre-1983 two-party-preferred figures are often estimates.)
The average swing across all eighty-nine is 4.6 per cent to the federal opposition. But if we disaggregate between government-held and opposition-held, and between resignations and deaths, the subset in which Aston sits — resignation in an opposition-held seat — has an average swing of 2.3 percentage points to the opposition.
(My table is here. I categorise resignations due to section 44 of the Constitution separately: they shouldn’t produce the level of resentment a normal resignation does, nor the loss of the sitting MP’s personal vote, and including them only changes it to 2.2 points overall.)
In recent decades, governments have waxed and waned in their tendency to contest opposition-held electorates at by-elections. The Hawke–Keating governments (1983–96) ran in its first six, then from 1989 sat out the remaining six. The Howard government (1996–2007) ran in none out of six. Rudd Labor threw its hat into the Gippsland ring in 2008, got a hefty 6.1 point movement against it for its trouble, and stayed out of the next four. The Abbott government ran in Griffith after Rudd’s retirement and managed a decent 1.2 point swing, but not enough to take it. It declined to contest Batman, Perth and Fremantle, but weighed into Braddon and the aforementioned Longman.
Of course, some by-elections are just more winnable than others. It made sense for the Morrison government to contest Eden-Monaro near the height of its Covid-fuelled popularity in 2020. Once again it managed a small swing, though not quite enough. (Eden-Monaro was declared a “test” for opposition leader Anthony Albanese. Apparently he “passed” because the swing was less than the 0.86 points required to change hands.)
Labor’s decision to run this weekend in Aston is fair enough: it’s a possible, if unlikely, win. The seat has been redistributed by a notional 2.3 points towards Labor since the 2010, when Tudge replaced Chris Pearce, scraping home with just 0.5 points. But 2010 was a very strong year for Labor in that state, indeed its second-best Victorian two-party-preferred vote in the history of preferential voting. If Tudge had retired at the 2022 election it would’ve been up for grabs.
The evidence suggests that Tudge built up a very big personal vote, peaking in 2019, but deflating somewhat in 2022 for reasons anyone not living under a rock would be aware of. Had he not attracted that odium, the on-paper margin today would be bigger than 2.8 per cent, and perhaps Labor wouldn’t be tempted. (And, of course, Tudge might not be pulling up stumps.)
But it must also be said that neighbouring seats also swung quite solidly to Labor last time. Most have substantial Chinese-Australian populations, and in Sydney and Melbourne this cohort seems to have particularly swung to the Liberals in 2019 (turned off by Labor’s economic policies) and back to Labor in 2022 (absence of said policies, plus the Morrison government’s thunderous drum-beating on China.)
And/or: perhaps Labor ran a brilliant campaign in the seat in 2022. In 2023 it is running the same candidate, Mary Doyle.
This time, sightings of “Test for Peter Dutton!” claims have been much more prevalent than for the prime minister, but a big swing to the Liberals would transform the chatter. The Albanese government’s mooted superannuation changes aren’t particularly tailored to inflict pain on Aston (the wealthier Victorian seats these days are held by teals and, in the case of Higgins, Labor), but details aren’t always important. The “one in ten” (in thirty years’ time) statistic that the Coalition–News Corp tag team came up with a few weeks ago is potentially effective.
As well as dominating the coverage, New South Wales also seems to have soaked up media organisations’ polling budgets, and so far they have published no surveyed voting-intention figures whatsoever for Aston. In terms of the next general election, tomorrow’s result will only matter insofar as whoever is the sitting member in 2025 will enjoy an advantage. But interpretations of the swing can mean a lot for leaders’ fortunes, of course. And because, unlike in Longman in 2018, the opposition is behind in the national polls, it’s Dutton who will most be sweating on the result. •
