Three weeks after I hit reset on BludgerTrack (a fact now represented on the sidebar charts, in which the Gillard and Rudd epochs as separate series), the results remain sensitive to weekly variation as the overall pool of data is still very shallow (eleven polls in all). This week we have had Nielsen’s monthly result, the poll which appeared last week from newcomers AMR Research, and the usual weekly Essential and Morgan. The state relativities have been updated with last week’s result of federal voting intention in Queensland from ReachTEL, along with breakdowns from Nielsen and Morgan (the latter of which pleasingly looks to have become a regular feature).

What this all adds up to is a move this week from minor to major parties, one consequence of which is that the Greens have recorded what by some distance is their worst result since BludgerTrack opened for business in November. This may well portend a further decline born of the leadership change and the tightening focus on the major party contest, but I would want more evidence before I signed on to that with confidence. It’s certainly clear that the return of Rudd has been bad news for the combined non-major party vote, but the scale of it is a bit up in the air at the moment. So far as this week’s result is concerned, the shift has enabled Labor to both handily break through the 40% primary vote barrier while going backwards slightly on two-party preferred, on which the Coalition recovers the narrowest of leads.

Tellingly, despite two-party preferred being a mirror image of the 2010 election result, the seat projection still points to a continuation of Labor in office, albeit that it would rely on Andrew Wilkie (whom ReachTEL suggested to be on track for victory in its Denison poll last month) and Adam Bandt (who will continue to be designated as the member for Melbourne until polling evidence emerges to suggest he will lose, which will by no means surprise me if happens) to shore it up in parliament. This points to the crucial importance of Queensland, where there are no fewer than nine LNP seats on margins of less than 5%. So long as the swing in that state remains where BludgerTrack has it at present, Labor could well be in business.

However, as Kevin Bonham notes, there is an obstacle facing Labor on any pathway to victory that runs through Queensland: eight of the nine marginals will be subject to the effects of “sophomore surge”, in which members facing re-election for the first time enjoy a small fillip by virtue of acquiring the personal vote which is usually due to an incumbent. In seven of the nine cases this comes down to the LNP members having won their seats from Labor last time, although Leichhardt and Bonner are a little more complicated in that the members had held them at earlier times. The other two LNP marginals are the Townsville seat of Herbert, which stayed in the LNP fold in 2010 upon the retirement of the sitting member, and Fisher, which as Kevin Bonham notes is a “fake marginal” and an unlikely Labor gain.

The BludgerTrack model has sophomore surge effects covered, with adjustments of between 0.4% and 1.9% applied according to whether the seat is metropolitan or regional (the latter being more susceptible to candidate effects generally) or has what Bonham calls the “double sophomore” effect, in which the challenging party also loses the personal vote of its defeated member from the previous election. Other factors used in the model to project a seat’s result are the existing margin, the statewide swing as determined by the poll trend, and a weighting to account for an electorate’s tendency to swing historically. These results are then used to calculate a probability of the seat being won by Labor, and the sum of the various seats’ probability scores determines the statewide seat total shown on the sidebar. Sophomore surge effects are currently reducing Labor’s Queensland total by about 1.3 seats, which means they will be down one seat for about two-thirds of the time, and down two seats for the remainder.

Finally, sharp-eyed observers may note that the projection has Labor down a seat in New South Wales, by the narrowest of margins, despite a small swing in their favour on the two-party preferred. The loss of sitting members in three loseable seats (Dobell, Kingsford Smith and Barton) is playing a part here, but it also represents the fact that the model rates Labor as having been slightly lucky to have won a twenty-sixth seat there at the last election.