The Australian carries a consolidated Newspoll today. Usually these aggregate five or six surveys over three months, each with a sample of about a thousand. But this one contains two fat 1700 strong polls.

This sentence stands out: “Individual state sample bases range from 655 to 679 voters”. That’s odd. If Newspoll had randomly sampled (or sampled proportionately for each state) you would expect numbers to range from, say about 300 (for South Australia) to 1200 (NSW).

But instead we now know that Newspoll’s two recent surveys, one showing 58 to 42 to Labor and the second 54 to 46, deliberately over-sampled the smaller states and under-sampled the larger ones.

This means that the state data published today have meaningful sample sizes. This is good. But it also might help explain the fluctuation between the two weekly polls.

Obviously Newspoll would have weighted their data to take account of state sizes – that is, the NSW sample would count for four times the South Australia one. But that means that small fluctuations from one week to the next in the 300-350 strong NSW (or Victorian) component would have been magnified by the weighting.

Or maybe I’m wrong. Something for the statisticians, perhaps.

Also, Dennis Shanahan rightly makes a big deal of the age breakdown (although only half the sample was polled after John Howard had announced his $4 billion package).

There is an interesting story there. In broad terms (according to Newspoll), Labor’s primary support with the under 50 year olds has jumped about ten points from the Beazley leadership to the Rudd one, while those aged 50 and over have moved by about half the amount. The shift is slightly more pronounced in the 18-34 year olds than the 35 to 49ers.

But Labor’s improvement hasn’t all come from the Coalition. Indeed, among over 50s the Coalition’s vote hasn’t much changed from this time last year, and among the under 50s, the ALP’s jump is only about half accounted for by a drop in the government’s.

Most of the rest has come from the Greens. Their support has almost halved since the days of Beazley.

Now, Newspoll is the only outfit which doesn’t include Greens in its “which of these parties will you vote for?” question; respondents must come up with the party themselves. This probably understates Green support; certainly all the other pollsters are registering it as higher.

But this aside, two repercussions flow. First, a low Green vote lessens both their chances in the Senate, and their preference directing power in the House of Representative. It means they have less to offer Labor in a preference deal, but they need one more.

Second, most of the Green votes under the Beazley leadership were always going to flow to Labor in preferences – something not all commentators understood. Labor always needed to take votes off the government, not the Greens, and Rudd, so far, seems to be doing both.

But it also explains why Rudd’s primary vote has leapt, while his two party preferred support is currently only a few points above where Beazley’s was in the second half of 2006.

If the Greens do make up ground between now and the election, it will probably be at Labor’s expense.

But losing support to the Greens won’t hurt Rudd’s election chances much.