Can’t see full online results anywhere at this stage, but GhostWhoVotes and The Australian relate the publication of the latest quarterly Newspoll figures providing breakdowns by state, gender and metro/regional from the past three months’ polling. The state figures are always the most interesting from my perspective, and it’s only Newspoll and Nielsen which offer this level of detail. Nielsen publishes full breakdowns for its monthly polls, but these are from much lower samples than the quarterly Newspoll. To even the playing field, the following discussion uses quarterly averages of Nielsen’s results.

The resulting samples are substantial for the biggest states – over 2000, in the case of Newspoll for New South Wales – but correspondingly smaller for Western Australia and especially South Australia. It is presumably no coincidence that for the two biggest states the two pollsters are currently in agreement, with swings of 8 per cent in New South Wales and 6 to 7 per cent in Victoria (although as the charts below show, it was a different story in the previous quarter). It is also agreed the swing in Western Australia is around 4 per cent. With Queensland however, a gap emerges: Newspoll says 6 per cent, Nielsen says 10 per cent. There is a still bigger gap in the case of South Australia, but this can be put down to small samples and the latest obviously anomalous result from Nielsen.

To establish whether there has been any consistency in these distinctions over time, the charts below show the Labor swings recorded in each quarter since the election. Despite poll-level peculiarities, both broadly suggest that Labor enjoyed a post-election dead cat bounce in the resource states. In the case of Western Australia, this gave Labor a buffer which is still evident in the relatively slight current swing. On the Nielsen chart however, the most recent result sees the lines for New South Wales and Victoria cutting across Queensland’s – remembering that the prevous quarter’s results for these states were very different from Newspoll’s, the only serious interruption to a broadly similar picture for these two states since the election.

Conveniently, Galaxy has also conducted one poll of 800 respondents in each quarter in Queensland, and these accord perfectly with the Newspoll and Nielsen results from this state. In each period, Labor is slightly higher in Newspoll and slightly lower in Nielsen with Galaxy in between, and there’s not much in it in any case. In the current quarter, Galaxy’s 63-37 two-party preferred splits the middle of the previously noted four-point gap between Newspoll and Nielsen. The only other state-level results I’m aware of are two Western Australian polls of 400 respondents conducted by Patterson Market Research. One of these was as long ago as October last year, which accorded with Newspoll and Nielsen of that time in showing a Labor recovery. However, an unpublished poll from two months ago was solidly worse for Labor than either, pointing to a swing of about 7 per cent.

What the polls would appear to indicate then is a big enough swing in New South Wales to account for Greenway, Robertson, Lindsay, Banks, Reid, Page, Eden-Monaro, Parramatta, Dobell, Kingsford Smith, Werriwa, Barton, Richmond and McMahon, and a slightly smaller swing in Victoria that would take out Corangamite, La Trobe, Deakin and possibly Chisholm (UPDATE: I originally included McEwen, but as noted in comments, the redistribution has made this safer for Labor). Since Galaxy splits the middle in Queensland, it seems best to apply its 8 per cent swing there – which, as was noted at the time the poll was published, would leave only Kevin Rudd standing in Griffith. Gone would be Moreton, Petrie, Lilley, Capricornia, Blair, Rankin and Oxley. In Western Australia, Labor currently holds Brand on 3.3 per cent, Fremantle on 5.7 per cent and Perth on 5.9 per cent: the Newspoll and Nielsen poll swings would put the first in danger while sparing the second and third.

Results from South Australia are small-sample and inconsistent, except that they have broadly been at the higher end of the national spectrum – perhaps around 8 per cent. However, this is coming off the high base of last year’s election, which gave Labor very handy buffers in a swathe of traditionally marginal seats. The lowest Labor margins are 5.7 per cent in Hindmarsh (where Labor has weakened relatively over the last two elections), 7.7 per cent in Adelaide, 12.0 per cent in Wakefield, 12.2 per cent in Makin and 13.9 per cent in Kingston. The last three seats, remarkably, were all in Liberal hands as recently as 2007.

Owing to insufficient sample size, neither Newspoll nor Nielsen provides state-level breakdowns for Tasmania. We did however have an EMRS poll from Bass a month ago which pointed to a 9 per cent Liberal swing, but this was from a small sample of 300 and there were questions raised about its methodology. A swing of that size would nonetheless be enough to take out Bass (6.7 per cent) and its neighbour Braddon (7.5 per cent). The territories of course are pretty much excluded from the polling picture altogether, although Warren Snowdon’s hold on Lingiari in the Northern Territory would have to be open to question given its margin of 3.7 per cent.

None of this should be read as a prediction: first term governments notwithstanding, its a rare government that doesn’t plumb mid-term polling depths far removed from the result eventually produced by the election. This is especially so in the modern environment, when weakening party loyalties have produced an ever-swelling contingent of swinging voters. Even so, the drumbeat consistency of dire results for Labor since April is hard to ignore, and it has no precedent for any government which lived to tell the tale. Labor’s leads during the early part of Mark Latham’s shooting star trajectory were never higher than 55-45; only once in early 2001 did Kim Beazley get as high as 57-43, and was usually solidly lower; and the relevant Newspolls for the great Houdini act of modern federal politics, Paul Keating’s win in 1993, look fairly benign compared with Gillard’s recent numbers. The Fightback! polls which toppled Hawke at the end of 1991 were in the order of 56-44 and 57-43, and Keating wrestled them back to the low fifties by March. Only from November 1991 to February 1992, after John Hewson remodelled his GST to exclude food and clothing, did the Coalition reach such peaks again.

Another lesson from history is that when the electorate ejects Labor from office, it tends to do with a force which the conservative parties are spared. With few exceptions (a handful of those in New South Wales plus Brand, Lingiari and arguably Oxley, which Pauline Hanson won in 1996 as a disendorsed Liberal), the seats listed as Labor losses on the current results have all been lost to them before.