Yesterday, the Australian Electoral Commission performed an act which in a rational world would have excited no interest. Since last weekend the commission has featured a “national two party preferred result” on the front page of its Virtual Tally Room, which has assumed tremendous psychological interest as Labor’s margin has steadily eroded from 0.6 per cent to 0.4 per cent. However, the tally had a flaw which biased it in Labor’s favour: there were no Labor-versus-Coalition figures available from strongly conservative Kennedy, Lyne, New England or O’Connor, where the notional two-candidate preferred counts conducted on election night involved independents. This was only balanced out by left-wing Melbourne, where Labor and the Greens were correctly identified as the front-running candidates for the notional count. For whatever reason, the AEC decided yesterday to level the playing field by excluding seats where the notional preference count candidates had been changed since election night, which in each case meant left-wing seats where the Liberals had finished third to the Greens (Batman and Grayndler) or Andrew Wilkie (Denison). The result was an instant 0.4 per cent drop in Labor’s score, reducing them to a minuscule lead that was soon rubbed out by further late counting.

In fact, very little actually changed in yesterday’s counting, which saw a continuation of the slow decline in the Labor total that is the usual pattern of late counting. The media, regrettably, has almost entirely dropped the ball on this point. Mark Simkin of the ABC last night reported that Labor’s lead had been eradicated by the “latest counting”, as opposed to an essentially meaningless administrative decision. Lateline too informed us that Labor’s two-party vote had “collapsed”, and Leigh Sales’ opening question to Julie Bishop on Lateline was essentially an invitation to gloat about the fact. Most newspaper accounts eventually get around to acknowledging the entirely artificial nature of the 50,000-vote reversal in Labor’s fortunes, but only after reporting in breathless tones on the removal of votes that will eventually be put back in.

The reality is that nobody knew who had the lead on the two-party vote yesterday morning, and nothing happened in the day to make anybody any the wiser. The Prime Minister equally had no idea on election night when she made her ill-advised claim to the two-party majority mantle. Only when all seats have reported Labor-versus-Coalition counts, which is probably still a few weeks away, will we be able to say for sure. The best we can do at present is to construct a projection based on the votes counted and our best assumptions as to how the gaps in the vote count data will be filled when all the figures are in.

At present we have completed “ordinary” polling day totals for all electorates and advanced counts of postal votes in most cases, but there has been no progress yet on absent or pre-poll votes in roughly half. Where counting of any of these three categories has been conducted, I have projected the party results on to the expected total of such votes (derived from the “declaration vote scrutiny progress” for absent and pre-poll votes, and from the number of applications for postal votes discounted by 16 per cent as per experience from 2007). Where no counting of a particular category has been conducted, I have compared the parties’ 2007 vote share in that category with their ordinary vote share, and applied that difference to the ordinary vote from this election. For example, the 2007 Liberal two-party vote in Canberra was 7.19 per cent higher than their ordinary vote share, so their 40.54 per cent ordinary vote at the current election has been used to project an absent vote share of 47.73 per cent.

For Batman, Grayndler and Denison, I have used the figures from the two-party Labor-versus-Liberal counts that were conducted in these seats from ordinary votes on election night, calculated the swing against the ordinary vote in 2007 and projected it over the expected absent, pre-poll and postal totals. For Melbourne, New England and Kennedy, where no Labor-versus-Coalition figures are available, I have used preference shares derived from the Labor-versus-Coalition counts from the 2007 election to determine the swing on ordinary votes, and projected that swing through the other categories. It’s with Lyne and O’Connor that things get crude, as we have no case study of how Rob Oakeshott’s or Wilson Tuckey’s preferences split between Labor and Nationals candidates. For O’Connor, which has at least been a Labor-Liberal-Nationals contest at successive elections, I have crudely arrived at a 7.9 per cent swing against Labor derived from the primary vote swing plus moderated by a 70 per cent share of the swing in favour of the Greens. The best I could think to do for Lyne was average the two-party swings from the neighbouring electorates, producing a 5.14 per cent swing against Labor.

Plug all that in and here’s what you get:

Labor 6,313,736 (50.02 per cent)
Coalition 6,307,924 (49.98 per cent)

In other news, Andrew Wilkie says the two-party vote total is “not relevant” in determining which party he will back. Good for him.

UPDATE: An Essential Research poll has it at 50-50, which is “unchanged” – I’m not sure if this is in comparison with the election result or a previously unpublished Essential result from a week ago. Basically no change on preferred prime minister. UPDATE 2: The 50-50 from last week was indeed an unpublished Essential result from their rolling two-week average, which they understandably felt was not worth publishing under the circumstances.