The idea that there are “federal implications” of state elections doesn’t hold up under examination. But that doesn’t mean the Gillard government will hold on in this year’s federal election.

Time to kill another myth: “federal implications”.

I won’t reel off every journalist and politician who has insisted there are “federal implications” to the Western Australian state election, either directly using that phrase or otherwise opining on its meaning. I don’t need to. We now live in a political culture where professional journalists seriously discussed the “federal implications” of a state by-election in New South Wales in 2010.

What do “federal implications” actually mean? This is where it gets tricky. Does it mean voters are likely to vote the same way federally as they have previously at state level, so a state result is an accurate gauge of a federal election occurring shortly afterwards? What about “state implications”? Do voters vote the same way at state level as they have federally? National political journalists generally don’t bother with “state implications”, but there’s no reason why, once you assume voters are unable or unwilling to distinguish between state and federal issues, it shouldn’t run both ways.

And is it the be-all and end-all of voter choice? If the Labor brand is “toxic”, how important is that as a factor determining votes? Is it a dominant factor in determining vote, or just a nebulous addition to other factors like economic management? And what about candidates who defy voting trends through strong local work?

So let’s put it to the test. What evidence is there that voters regularly vote the same way federally as they vote at state level — or vice versa?

There were two state elections before the federal election in 2010, held in South Australia and Tasmania, five months before the August election, i.e. closer than the WA election to the federal election this year. In neither case did the performance of the major parties match their performance in those states at the subsequent federal election: in South Australia in the state election, Labor suffered a huge 7%+ swing in the lower house, which went directly to the Liberals. At the federal election five months later, both Labor and the Liberals suffered swings against them in the House of Reps, with the Greens, who’d only picked up 1.6% at the state election, benefitting from it. And in the Tasmanian election, Labor suffered a big swing, most of which went to the Liberals, but at the federal election the Liberals suffered a big swing against them, with Labor picking up votes; the Greens also picked up a swing, but somewhat less than at the state election.

If you look at the Victorian election held three months after the federal election, state Labor suffered a big swing in the lower house and lost government; the Liberals, the Nationals and the Greens all picked up swings. But three months earlier, both major parties suffered swings against them and, as in South Australia, it went to the Greens.

But the “state implications” idea works better for the NSW election in March 2011, admittedly many months after the federal election, but the swing against Labor and to the Liberals in NSW at the federal level was magnified at the state election.

“Out of 11 state or territory election held adjacent to federal polls since 1998, only three had similar results in terms of swings.”

What about in 2007? There was a NSW election in March that year, a long time before the November federal poll. The Liberals picked up a small swing from Labor at that poll, not enough to defeat Morris Iemma. Of course, in November there was a big swing to Labor in NSW under Kevin Rudd.

The 2004 election? We can test the “state implications” thesis in WA, where an election was held in February, not long after John Howard won his final term. At the 2004 federal poll, WA voters had swung hard to the Liberals, with a swing of over 6% in the House of Reps, partly from Labor but also because One Nation had imploded. But in February, West Australians returned Geoff Gallop for a second term with a swing of over 5%; the Liberals picked up a swing from the One Nation vote, too, of 4%. But the Greens, oddly, went slightly backward at state level after picking up votes in the House of Reps federally.

The ACT had a poll just a week after the 2004 federal election. And that appears to reinforce the “state implications” thesis: Labor and the Liberals both picked up a swing in the ACT just as both parties had a week earlier, mainly because the Democrats almost entirely vanished, leaving 5-6% of House of Reps votes up for grabs. In fact, the Democrats suffered almost exactly the same swing in both elections.

Let’s go back further. South Australia went to the polls early in 2002, not long after the 2001 election. That election reversed what had happened only a few months before: Labor had gone slightly backwards federally in SA but picked up a swing to secure government for Mike Rann. The sizable swing to the Liberals federally entirely vanished. The ACT had a poll three weeks before the federal election. Labor picked up a big swing to put Jon Stanhope into office, but just three weeks later Labor went backwards in the territory by nearly 4%.

Back further still, in 1998, Tasmania went to the polls a few weeks before John Howard sought re-election. It proved a rare predictor: Tasmanian voters performed the same at both elections, with Labor picking up a large swing at the expense of the Liberals. But the other, more famous election that year, in Queensland, was a poor predictor: One Nation smashed the conservative vote and handed Peter Beattie a win, but the Liberal and National vote was far stronger federally, with One Nation’s +20% swing reduced to 14.4%.

So out of 11 state or territory election held adjacent to federal polls since 1998, only three had similar results in terms of swings, and two of those state elections mirrored federal results, not the other way round. In another three state polls at least one of the major parties secured a swing in the same direction as federally — but again two of those were state elections after the federal poll.

It also seems that the closer state and federal polls are, the more likely voters are to vote the same way.

From all that, you might conclude that “state implications” is a thing — there is some evidence voters vote the same way at the state level as they do federally when they vote at a state poll not long after a federal election. But “federal implications” — voters signalling how they’ll vote federally by how they vote at a state level — doesn’t appear to hold up. That’s unless you assume that in all but one case, federal politicians were skilled enough to recognise the “federal implications” of a state result and work successfully to reverse them in time for their election. Say, by dumping a current prime minister for a former prime minister.

That’s not to say that a state-federal division will always exist. The group most assiduous in seeking to blur the boundaries of state and federal issues in recent years has been federal politicians: centrists like former PM John Howard and federal Opposition Leader Tony Abbott tried to win votes by taking over state hospitals; Prime Minister Julia Gillard spent last week selling law-and-order policies on the streets of western Sydney like she was campaigning in a state byelection. Perhaps after a time voters will decide that there’s no distinction between federal and state issues, and vote accordingly.

It’s also not to say that the Gillard government isn’t in deep, deep trouble and headed for a massive defeat. On current polling, the WA election looks likely to be an accurate predictor of a federal poll later in the year, except in understating the size of Labor’s defeat. But it’s a rare exception.