Conversely, the Coalition shares were 53 per cent in 1995, 49 per cent in 1998 and 40 per cent in 2001.

So there was a two-party preferred vote swing to Labor of four per cent in 1998, delivering minority government to Peter Beattie. Then there was a further swing to Labor of nine per cent to produce the landslide of 2001, bringing the combined swing from 1995 to 2001 to a whopping 13 per cent.

The same electoral boundaries applied for three general elections, in 1992, 1995 and 1998. A major redistribution of seats followed in 1999, though the number of seats was left at 89. The February 2001 and February 2004 elections use these same boundaries.

If checking my pendulum against official statistical returns, you would notice 71 of my 89 seats are clearly based on the two-party preferred vote, Labor versus National or Labor versus Liberal.

There are some cases where I have changed the actual scale of the seat into the conventional (Labor versus Coalition) scale. While Inala is not Labor's safest seat against independent candidate George Pugh, it is Labor's safest seat against the Liberal Party.

Likewise, while Warrego was the weakest National Party seat against the independent candidate Wally Gleeson, it was the strongest National Party seat against Labor.

Of the six independent seats (all shown on the non-Labor side of my pendulum) two were won at by-elections, Surfers Paradise in May 2001 and Maryborough in April 2003. But in each case the swing figure shown is that based on the general election vote.

Surfers Paradise was won at the general election by then National Party leader Rob Borbidge, who polled 55.3 per cent of the two-party preferred vote to Labor's 44.7 per cent.

When Borbidge resigned from parliament in March 2001 and the May by-election gave the seat to independent Lex Bell, I merely struck out ``National'' on my then pendulum and inserted ``independent'' in its place. So the Nationals then had 11 members.

At the February 2001 general election, Darling Downs went to independent Ray Hopper, who switched to the Nationals in December that year, restoring them to 12 members.

At the general election Gympie was won by Pauline Hanson's One Nation candidate Elisa Roberts, but she became an independent in April 2002.

So, in the last parliamentary term Labor numbers remained at 66 throughout and the Liberals at three. The National Party began at 12, dropped to 11 when Borbidge resigned and went back to 12 when Hopper joined up.

Pauline Hanson's One Nation began at three but finished at two. The number of independents began at five, rose to six, dropped to five and finished at six.

A Maryborough by-election in April 2003 was caused by the resignation of one independent, John Kingston, but was won by another, Chris Foley, the present member.

However, my swing figure is based on the general election vote when Kingston secured 50.5 per cent of the two-candidate preferred vote to Labor's 49.5 per cent.

Now for some predictions. I believe Labor's share of the aggregate two-party preferred vote this time will be 53 per cent to 47 per cent for the Coalition. In other words, I am predicting a swing against Labor of seven per cent.

Pendulum theory tells me such a swing would reduce the Labor majority of 43 seats in 2001 to 15 seats in 2004. I do predict that.

My belief is that the effect of such a swing will be the same as if the swing were uniform.

Let me begin with the non-Labor side. I predict all six independent members will retain their seats of Nanango, Nicklin, Surfers Paradise, Gladstone, Gympie and Maryborough. I predict that both the remaining members of Pauline Hanson's One Nation will retain their seats of Tablelands and Lockyer.

Subject to one reservation, I predict the Nationals will retain their 12 existing seats and the Liberals their three.

The one reservation is Keppel, where the popular sitting member, Vince Lester, is retiring. If Labor were to gain one seat, it would be Keppel.

Now to Labor losses. Because I think the swing is likely to be relatively uniform I predict the Nationals will gain the following eight seats from Labor: Burnett, Burleigh, Toowoomba North, Charters Towers, Broadwater, Thuringowa, Burdekin and Redlands.

On the same basis I predict the Liberals will gain the following six seats: Noosa, Clayfield, Kawana, Indooroopilly, Aspley and Mudgeeraba.

I am aware that the cancellation of deviations effect may overcome any uniform swing effect, so it would not surprise me if the Nationals failed to win Redlands but won Glass House instead. Likewise, it would not surprise me if the Liberals failed to win Mudgeeraba but won, instead, either Gaven or Currumbin.

In any event, my prediction for overall numbers is 52 for Labor and 37 for non-Labor. That would leave Labor with a majority of 15 seats.

On my predictions the 37 non-Labor members would be 20 Nationals, nine Liberals, six independents and two for Pauline Hanson's One Nation.

Predictions are inherently dangerous but there is one I can make which is totally safe. No seat will change hands between the National and Liberal parties, given the Coalition agreement to field only one candidate in each seat.

Associate Professor Malcolm Mackerras teaches in the school of politics at the Australian Defence Force Academy in Canberra