Be careful with those numbers

Australia’s election result is now clear, at least in the House of Representatives. With Labor holding Gilmore and Macnamara, the Liberals holding Deakin and the Greens taking Brisbane, the numbers have come out at 77-58-4 with 12 others, just as I foreshadowed a week and a half ago. The Senate is less clear, although the combination of Labor, Greens and independent is still set for a majority.

Even in the House, however, counting is not quite finished. Although the deadline for receipt of postal votes has now passed, some of the late-arriving ones are yet to be processed, as are varying quantities of absentee and provisional votes. That should all be tidied up this week.

The fact that counting extends over such a long period means that figures need to be interpreted with some care. While the electoral commission provides a wealth of detail for those interested (see official site here), some aspects of it can easily trap the unwary.

An obvious case is turnout. Both the aggregate tables and those for individual divisions show a figure labelled “turnout” – as of this morning, the national one is 88.6%. That looks like a noticeable drop from the 2019 figure, which was 91.9%. Last week, when most seats were showing numbers back in the 70s, there was a wave of internet comment from readers who noticed that turnout seemed to have fallen off a cliff, and wondered why the media had not picked this up.

The answer is that in the context of incomplete figures, “turnout” doesn’t really mean turnout: it’s just the number of votes counted so far, expressed as a percentage of the total on the roll. So those numbers have been steadily climbing and will continue to do so; they may (or may not) finish a bit below those from three years ago, but there’s clearly no radical difference.

Incidentally, even the final turnout figure will be not quite what it appears, since, while enrolment is theoretically compulsory, we know that it is far from universal. The turnout figure gives the proportion of those who enrol, not those who are eligible to enrol, and the relationship between the two is not constant. So if the AEC conducts a successful drive to enrol more people, and (as seems likely) those new enrolees are less enthusiastic about voting than the average of the population, then measured “turnout” will go down, even though the real underlying turnout has gone up.

A more subtle problem of interpretation comes with the two-party-preferred vote, always a subject of some interest. The national summary table gives you the primary vote for each party, and then at the bottom the “National two-party preferred” figure (currently showing 51.8% for Labor). Only the very alert will notice that this comes with a different and much lower “turnout” figure: currently 74.1%.

The problem here is not just that there are lots of votes that have been counted for primaries but not for two-party-preferred. It’s that they come from very specific places, and therefore may not be at all representative of the total – making the progress figure, which many pundits have already seized on, potentially very misleading.

The places that haven’t had a two-party-preferred count done yet are those divisions where the main contest (that is, who are expected to be the final two candidates in the count once everyone else is eliminated) is something other than between Labor and the Coalition. That means that the indicative count there, or “throw”, which is the only sort of preference distribution that has happened so far, won’t tell you anything about the national two-party-preferred vote, so nothing from those seats has yet made it into the table.

The commission calls these “non-classic divisions”: there’s a table of them here. There are 26 of them, or one seat in six. Six of them are ALP vs Greens, three Coalition vs Greens, two ALP vs independents and 15 Coalition vs independents or minor parties.

That’s a big number.* As support for the major parties has declined, non-classic divisions have been on the rise. In 2007 there were only three of them; that rose to eight in 2010, 11 in 2013 (when I wrote a post about them), 17 in 2016 and 15 in 2019.

And of course the more of them there are, the greater their possible effect on the overall two-party-preferred figure. If, for example, the non-classic divisions were on average six points more favorable to the Coalition in two-party-preferred terms than the rest, that would bring the Coalition’s two-party-preferred total up by a full percentage point, making the election outcome look quite different. (And conversely if it was more favorable to Labor.)

Eventually the commission will get around to doing a two-party-preferred count in all these seats for information purposes, enabling us to get an accurate total. It may take a while; in 2013 (when there were a lot fewer of them) it took two and a half months, although as William Bowe pointed out to me, in 2010, when the election was extremely close, they were able to do it in a couple of weeks.

For what it’s worth, it looks to me as if adding the non-classic divisions will boost Labor’s two-party-preferred vote a little, but probably only by a couple of tenths of a point. Shortly after election day, Bowe and Antony Green estimated Labor’s final two-party-preferred share at 51.8%; I think it’ll be higher than that, but certainly in the same ballpark. Until we see the preference flow from the large number of teal independents, there’s not much more we can say.

The very high minor party and independent vote also raises the question of just how meaningful that overall two-party-preferred vote is. That’s something we can have a look at after the counting is complete.

.

* And it may grow; it’s possible that in the course of distributing preferences, the commission will find it was wrong somewhere about who the final two will be. Hume looks to me like such a case, and there could be others. But for now those seats have two-party-preferred counts, so for present purposes there is no problem.

.

ADDENDUM Wednesday 15th: As I flagged in the note above, another seat has been added to the “non-classic” list – Groom, in Queensland, where independent Suzie Holt has leapfrogged the ALP to finish as runner-up to the sitting Liberal National member, despite starting out in fourth place with just 8.3% of the vote. That brings the total to 27; if Hume joins them, it will make 28.