I have a live results feature standing ready for service this evening, which will presume to offer projections according to a somewhat experimental method that will be explained further below, together with individual results pages by electoral division featuring booth results displayed in table and map form (click on “activate” at the bottom of the individual seat pages for the latter). It may be the only place where the latter data will be available online, not counting the Australian Electoral Commission media feed from which it will be extracted.
Some final polls are in, with pollsters collectively offering a fairly wide spread that means the contest for post-result bragging rights is wide open:
• The Australian has a Newspoll result that concurs with others in finding movement back to yes – though by an entirely insufficient three points to 37%, with no down one to 57%. The 1413 sample from this poll has been combined with 1225 sample poll from last week to produce state breakdowns with no leading 54-41 in New South Wales, 51-43 in Victoria, 65-30 in Queensland, 65-28 in Western Australia, 60-33 in South Australia and 55-38 in Tasmania. The new batch of polling means we are also treated to a second set of Newspoll federal voting intention numbers in a week, in this case putting Labor ahead 54-46 (53-47 last week) from primary votes of Labor 36% (up two), Coalition 35% (down one), Greens 12% (steady) and One Nation 6% (up one).
• A JWS Research poll in the Financial Review has no at 52% and yes at 39%, converting to 57-43 after exclusion of the undecided. The poll was conducted Friday to Monday from a sample of 922.
• The Australian related yesterday that a poll conducted from October 1 to 9 by DemosAU, whose director George Hasanakos had a polling analysis website back in the day called Poliquant, had no leading 57% to 30% in both Queensland and Western Australia.
The projection model in my results will use the seat-level estimates from Focaldata’s multi-level regression with post-stratification exercise as a baseline for measuring the results as they are reported. However much the results that are in differ from what Focaldata predicted will then be projected on to Focaldata’s overall results at state level. Doubtless this will be noiser than the booth-matched swings methods that can be applied at elections, but it should at least go some way towards correcting for the peculiarities of the early numbers.
My attention this evening will be focused on the referendum, but I will have some sort of a post up following the progress of counting in New Zealand, a dedicated thread for which is here.