In light of prime minister Julia Gillard’s announcement today of an election date (14 September, in case you’ve been hiding under a rock), Antony Green has a post up about how the notice given for elections has varied over the years. In it he mentions, quite rightly, that there’s been a trend recently to fixed terms in Westminster systems:

In Australia, four states and the two territories already have fixed election dates. Only Queensland, Tasmania and the Federal Parliament still have variable terms where the head of government determines the date.

Many Canadian provinces have adopted fixed terms, as has the current UK government, Prime Minister David Cameron having already announced the date of the 2015 election.

In 2011 New Zealand Prime Minister John Key gave nine months notice of his country’s election to avoid speculation on the election clashing with the Rugby World Cup.

But it’s worth noting that even before today’s announcement, Australia already had something a lot like fixed terms, due to the rather interesting way that our two houses of parliament interact. So here’s a quick guide.

The House of Representatives, in theory, has fully flexible three year terms. But the Senate runs to its own fixed timetable: new senators take office on 1 July every third year (half of them at a time; I’m ignoring those from the territories), and the election for their places has to be held within the preceding twelve months – except in the case of a double dissolution.

No government wants to hold a half-Senate election separate from the House of Representatives, because voters would treat it like a gigantic by-election. (The last one was in 1970, and it helped to bring down John Gorton.) So that means elections for the lower house have to stick to the three-year timetable as well, unless the option of a double dissolution is available (as it hasn’t been in the present parliament).

Even if a double dissolution is possible, it’s a risky move, for three reasons:

  1. It can only be held more than six months before the regular election is due, so the government has to cut short its term and upset voters with an early election;
  2. After a double dissolution the new Senate terms are backdated, so the following election has to be held early as well; and
  3. Because the senators are elected all at once, quota for election in a double dissolution is almost halved, so it’s much easier for minor parties and independents to get up – which can make life difficult for the government in the future.

So governments are mostly stuck with, at most, a 12-month window for holding an election. In practice it’s much less than that, because summer and winter are both unattractive times for sending people to the polls; elections are usually spring or autumn. And once you’ve held a spring election, you have to keep doing it – it’s hard to switch to autumn without the House of Reps going over its three-year limit (it’s only been done once, in 1909-10).

From 1955 to 1972, seven successive elections were held within a narrow range, from late October to early December. (The dates are all here.) But one of them, in 1963, was held a year early – without the Senate, thus putting the two houses out of alignment. After double dissolutions in 1974 and 1975, Malcolm Fraser got things back in order with elections late in the year in 1977 (December) and 1980 (October).

Another double dissolution in 1983 upset the calendar again, forcing Bob Hawke to hold the next election early (December 1984), before then holding his own double dissolution (the last, to date) in August 1987.

That allowed the next election, in 1990, to be held in March, as were the following two. From the point of view of a government’s flexibility, autumn elections were ideal: there was no long period to wait before the new senators took up their positions, no football finals to complicate things, and the option of going a few months earlier was always there just in case the political situation looked especially favorable.

But when John Howard exercised that early option by holding an election in October 1998 he severely constrained his options for the future (as I explained in 2007); it was then impossible to go back to autumn. And Gillard made things worse by going even earlier in 2010, effectively limiting her options for this year to August, September and October.

She has now made a virtue of necessity by renouncing early in the piece a flexibility that was mostly an illusion. (Although an illusion that the media, for their own purposes, were happy to feed.)

This may well become the fashion. Unless a government is willing to risk a double dissolution or a separate half-Senate election, or else shift the dates gradually later over a few terms until it can make the jump from December to March, we’re going to keep having spring elections. Since there’s not much to be gained by fiddling with the date by a few weeks, and since Australians are mostly used to fixed term elections (having them in most states), why not commit yourself early?

And although Australia’s way of reaching it is unusual, the result is relatively common around the democratic world: an election date that is decided by the government of the day, but which, either by law or by very firm convention, must occur within a quite narrow range – unless there are exceptional circumstances.

That’s the case, for example, with the most notable election expected this year, in Germany. The election is constitutionally required to be held between 46 and 48 months after the parliament first met, although in special circumstances it can be dissolved early. So we know that Germany will go to the polls sometime in September or October; the government can choose exactly when, but that doesn’t give it much of an advantage.

The big thing about fixed terms is getting rid of the prime minister’s ability to manipulate election dates for political advantage: to hold snap elections months or even years early in the hope of catching the opposition napping. By luck as much as design, Australia has pretty much done that anyway.