Two party preferred margins

If we've learnt anything it's that winning the two party preferred vote doesn't guarantee victory. And that the pendulum is a useful tool for going into an election, but it's not perfect. For example, the pendulum going into the last federal election would have "predicted" that a two percent swing to the government would boost its majority by 14, but when Howard did get that swing his seat majority only increased by two.

But in a way the pendulum is more useful in hindsight, in describing how close the last election was. Hawke Labor's safest election result was actually with its third highest two party preferred vote - 50.8% in 1987, with a 24 seat majority. In uniform terms a meagre 47.4 percent two party preferred would have been enough (delivering a one seat majority).

The then government managed this by annoying the heartland no end but impressing marginal seats - specifically, those outside the city. The western suburbs of Melbourne and Sydney swung most savagely to the Coalition, with the detested Treasurer Paul Keating's electorate of Blaxland going by 7.4 percent, but most regionals went the other way. Those classics of the genre Eden-Monaro (NSW), Herbert (Qld) and Bendigo (Vic) shifted to the government by 2.5, 3.2 and 1.4 percent respectively.

Look at this this way: after any election you could take those results and say that in three years time such and such a uniform swing will result in a change of government. (It's what, after allowing for redistributions, pundits do now. They use the results from the last election.) But of course in that three years the government may have alienated certain sections of the community, impressed others, and be on the nose in particular states and so on, so limiting the pendulum's usefulness. However, if you use the pendulum to analyse the previous result, it has more applicability, because the uniform swing assumption is more realistic.

So here is a table with federal results from 1983 to 2001. It shows the actual two party preferred result and the result that would have been needed for a different election outcome (assuming a uniform swing).

* Assuming uniform swing. So for example in 1983, had the Fraser government received 49.0% (2.2% more than it did) or more it would have won; in 2001 Kim Beazley would have needed 50.7% (which was 1.7% more than he got).

The closest election over the last two decades was, from the table, in 1993, and the safest win was 1996, followed by 1987. (Both the latter involved John Howard). 

Surprisingly, 1987 was safer then either 1983 or 1984. Labor started the '80s needing 51 percent of the vote to win, but by the election in '84, 49.6 percent would have done it. The 1987 figure of 47.4 was its lowest, probably in history, while 1998's 51.9 was possibly its highest ever.

Note that at the last election in 2001, 50.7 percent would have given Labor victory. That's the figure pundits now use for the upcoming poll. This also means that had the ALP held off the pro-government swing to 0.4 percent (Labor's vote going from 51.1 to 50.7) or less, it would probably have won. That's not as strange as it sounds: Labor was coming from the 1998 position record vote majorities in western Sydney seats, and those seats looked to swing strongly to Howard in 2001, but not enough to actually deliver seats to the Coalition. Meanwhile, other demographics would (and did) swing to the Opposition.

Keating's 1993 swing was also "battler" heavy, which was why it also delivered few extra seats (and took Labor back to its 51 percent requirement for victory).

Expanded table here