In the lead up to the next Australian election on May 18, I’ll be releasing a series of articles explaining the Buckley’s & None approach to forecasting elections.
To start, I thought it might be helpful to do a quick introduction to the quirky voting system in Australia. We only forecast the House of Representatives so we’re going to ignore the Senate for the time being. To quote wikipedia:
The [Australian electoral] system presently has a number of distinctive features including compulsory enrolment, compulsory voting, majority-preferential instant-runoff voting in single-member seats to elect the lower house, the House of Representatives.
Simple, right? Not unless you’re Antony Green.
Let’s start with the easy ones – compulsory enrolment and compulsory voting. In Australia, if you are eligible to vote you have to enrol with the Australian Electoral Commission. Compulsory voting means you also have to rock up on election day and get your name marked off the list, otherwise you could get fined. Australia ❤️ voting.
During the 1911 debate on compulsory enrolment Senator George Pearce said:
Too often [voting] is looked upon merely as a privilege, because people throughout the world have had to fight for it – in some instances under distressing conditions … but I venture to say that in a country like Australia, where we recognise that every man and woman should have the right to vote, that right becomes more than a privilege – it becomes a duty.
Who would have thought such a beautiful idea would have lead to so many 🍆 being drawn on election papers. It’s also worth noting that indigenous Australians were excluded from voting for many years after compulsory voting was established in 1912.
Compulsory voting in Australia is a big point of difference with US presidential elections where voting is voluntary. This adds a layer of complexity for forecasters in the US who need to estimate who is going to vote (“turnout”) and how those individuals are going to vote. In my view, this makes forecasting Australian elections a little bit easier because turnout is mandated by law.
Alright, onto some of the trickier ideas – “preferential instant-runoff voting”.
Preferential instant-runoff voting refers to the fact that when Australians vote in the House of Representatives they number candidates from 1 (the most preferred candidate) to the end of the list (the least preferred candidate). These preferences are then distributed so that if a voter’s first preference candidate is unlikely to win, their ballot can be re-examined to determine their next preference. This process is repeated until a candidate has more than 50% of the total vote.
Preferential voting is definitely not the most common voting system. In most countries voters get one vote, for their most preferred candidate, and that is it. The point of preferential voting is to elect the candidate that can build an absolute majority of support in the electorate. This contrasts with “first past the post” voting systems where the candidate with the most initial votes wins and no preferences are recorded or examined.
Australia’s system of preferential voting means that the “two-party preferred” vote is often more informative than the primary or “first preference” vote when estimating how many seats a party will win. The two-party preferred vote refers to the percentage of vote each of the final two candidates have after preferences have been distributed and the other candidates have been excluded from the count. An individual can have the most first-preference votes but lose the election, and the two-party preferred vote, when another candidate receives more non-primary (preferential) votes.
We can illustrate the importance of two-party preferred vote by examining the relationship between a party’s primary vote and the number of seats they win at each election. We will use the percent of total seats won, rather than number of seats won, to account for the change in the total number of seats from 75 in 1901 to 151 in 2017.
Primary Vote vs. Seats
There does not appear to be a strong relationship between the primary vote a party wins and the proportion of seats they win at an election.
As you can see the primary vote of a party does not give us a clear idea of the number of seats they will win at an election. For example, in 1977 the ALP won 39.7% of the primary votes and lost power with 30.6% of the total seats available. However, in 1990 the ALP won slightly less of the primary vote (39.4%) but won government with 52.7% of the total number of seats. This would make it difficult to predict the number of seats a party will win based on their primary vote.
In contrast, when we look at the relationship between two-party preferred vote and proportion of seats won there is a clear relationship. The more two-party preferred vote a party wins, the more seats they win. Let’s look at the same two elections. In 1977, the ALP two-party preferred vote was 45.4% so it’s no surprise they lost the election. And in 1990 the ALP two-party preferred vote was 49.9% and they scraped enough seats to win a majority.
Two Party-Preferred Vote vs. Seats
There is a strong relationship between the national two-party preferred vote and the number of seats a party wins at an election
I explore the relationship between a party’s two-party preferred vote and the seats they win more in part two. To summarise, in Australia you have to vote and you do this by numbering candidates from most to least preferred. In other countries, voting is voluntary and voters usually just pick their favourite candidate. This makes a party’s two-party preferred vote more relevant than their primary vote when examining elections in Australia.