Labor's new federal president, Carmen Lawrence, has been in politics a long time and would have seen a pea and thimble trick or two. She can expect more to cross her desk over the next year.

Those tricks are the party polls that point to electoral oblivion.

Make no mistake, Simon Crean is a lacklustre leader and the published polls show it. But the party's internal polls, at least the versions launched into the public domain via friendly journalists, are too shonky to be believed.

Take the most recent, in The Australian in September under the front-page headline "Crean's New Poll Shocker".

Said the report: "Internal polling by the Queensland [Labor] party shows all seven of its federal Labor MPs in the lower house would be wiped out, with the ALP's primary vote slumping to just 32 per cent. Support for the coalition had soared to a 52 per cent primary vote."

Read those sentences carefully. Taken together they are bunkum.

At the last election, Labor did very poorly in Queensland. Its primary vote was 35 and the coalition's 46. After distribution of preferences, that went to 45 versus 55 (two-party preferred) - a 10-point lead for the government, giving it 19 seats to Labor's seven and one independent.

Now, according to September's internal poll, the government's position has improved. But enough to take every Labor seat?

We have to tread carefully with aggregate survey data. Parties can go backwards in votes but pick up seats. There's no such thing as a uniform swing.

But to make any use of such polls, uniform swings must be assumed. It's not perfect but it's the best methodology available. Now, Oxley, with a margin of 8 per cent, is Queensland's safest ALP seat, so if every seat is to fall the swing must be at least that amount. And here the poll gets wobbly.

An 8 per cent swing on the 2001 result would give the coalition 63 to Labor's 37 per cent (two-party preferred - that's what decides seats, remember). But can we get 63 to 37 from primary votes of 52 to 32? Not in this universe. It would mean that, of the 16 per cent voting for neither major party, the coalition eventually receives preferences of 11 and Labor gets five.

That won't happen. No one disputes preferences will strongly favour Labor at the next election.

In fact, a reasonable two-party preferred estimate from the leaked numbers might be 58 to 42 in the government's favour.

That's a swing of 3 per cent and, assuming it's uniform, an extra three seats to the government.

A horrible scenario for the ALP, but not total annihilation.

This was just the latest in a line of dodgy leaks through various channels. You can't blame the leakers for trying it on, but journalists who are willing conduits might apply some rigour to the process.

Peter Brent is editor of mumble.com.au, a website that looks at electoral behaviour.