How close can a referendum get?

There was an interesting story on Friday in the channel nine newspapers, where Australian electoral commissioner Tom Rogers told us that the result of the Albanese government’s “Voice” referendum – supposed to be held later this year, although no date has yet been set – might not be clear on the night if there are a lot of postal votes to count.

It’s not made clear what the occasion for this opinion was, whether he held a press conference or he and reporter Lisa Visentin just happened to be chatting (there’s no sign of a press release on the AEC site). But while it’s not impossible that Rogers was being verballed, the quotation from him is pretty explicit: “It could well be that there is not a result on the night depending on how close that result is and how many people voted by postal [sic].”

You might think, as I did, that this is an odd thing to say. Of course it’s conceivable that the result, like any result, could just happen to be extremely close (see, for example, Florida 2000). But is it really likely that a referendum in Australia would turn on such a small number of votes that we would have to wait on the postals before being confident of the result?

Federal elections are sometimes close. There are always a few seats, sometimes as many as a dozen, where the winner is not clear on the night, and three times in the last ten elections – in 1998, 2010 and 2016 – the overall result was so close that it was impossible to be sure who had won without resolving some of those doubtfuls. That said, only 2010 was really knife-edge: in the other two cases the element of doubt was pretty small.

As increasing numbers of people use postal and especially pre-poll votes, this is becoming more of an issue, as seen in this year’s New South Wales election. While it was clear from early on that Labor had won, there was some doubt on the night as to whether it would have a majority in its own right. In fact the postal and pre-poll votes changed the complexion of the count quite significantly, and Labor ended up with only 45 of the 93 seats.

But a referendum, you might think, is quite different. There only two things matter: the majority of votes overall, and whether the proposal can win approval in a majority of states (that is, four out of six). Whereas any number of individual seats in a close election might turn out to be critical, in a referendum you would expect the median state to be obvious from the start of counting, if not before.

And with a question like the “Voice”, which looks like resolving itself into pretty much a straight left/right or progressive/conservative contest, it’s really only the majority-of-states issue that we’ll need to look at. If it’s anywhere near losing on the overall total, then it has surely lost: because the smaller states tend more conservative, there’s no way it’s going to win a majority of states without a majority (probably quite a large one) nationwide.

You can see what I mean by looking at the last referendum question whose outcome was in any doubt on the night, the so-called “simultaneous elections” question of 1977 (actually an attempt to abolish fixed terms for the Senate). There was no doubt at all about the overall majority: the “yes” vote had 62.2%. It also had clear majorities in three states (New South Wales, Victoria and South Australia). The question was whether it could snare a fourth state; in the end it failed narrowly, with 48.5% in Western Australia and 47.5% in Queensland.

That’s the only close result in the last seventy years, spanning twenty referendum questions. With a fair amount of agreement this time that the tipping-point state is going to be either Tasmania or Western Australia, it might seem that the chance of one of those turning out to be so close as to not be clear on the night is pretty remote.

There is, however, some support for Rogers’s view in the earlier history of Australian referenda. In the first fifty-odd years of federation, close results did occur with some frequency. In 1913, for example, a question on trusts lost with 49.8% overall and 49.7% in its fourth state. In 1946 a question on marketing had 50.6% overall and 48.7% in its fourth state. And in 1951 the referendum to ban the Communist Party had 49.4% overall and 48.7% in its fourth state.

By my count, 11 of the 24 questions* asked in that time were decided by a margin of less than two percentage points, either overall or in the tipping-point state. And depending on how many people vote by post, and how much the pattern of the vote departs from an ordinary election – that is, to what extent taking two-party-preferred vote as a proxy for the referendum vote will enable us to match swings to polling places – it’s quite possible that a two-point margin would leave us still in doubt by midnight on referendum day.

So the commissioner’s suggestion is not as far-fetched as it might seem: the fate of the “Voice” might indeed depend on waiting for the postals. But I remain sceptical.

.

* Or 26, counting the two conscription referenda of 1916-17. Both were close, but not quite that close.