What Is A Three Party Preferred?
With the historically close result of the seats of Brisbane and Macnamara in 2022 many may be quite a bit more familiar with this subject than I thought.
They are no longer publishing it but for a time the AEC was conducting a “three candidate preferred” (3CP) indicative count of ballots in the Division of Brisbane in the 2022 election. Conducting this is not unlike the “two candidate preferred” (2CP) indicative count carried out in every seat but rather than assuming the exclusion of all but two candidates the AEC assume the exclusion of all but three candidates. Here is a point in this count compared to the 3CP count in the final distribution of preferences in the same seat in 2019.
| 3CP Brisbane | LNP | ALP | GRN |
|---|---|---|---|
| 2019 | 50.89% | 25.37% | 23.74% |
| 2022 (83% counted) | 42.09% | 28.62% | 29.29% |
| 3CP swing | -8.80% | +3.25% | +5.55% |
In order to win a seat in Australia’s Instant Runoff Voting system a candidate must never be in last place. In 2019 the Greens’ candidate Andrew Bartlett was in last place at this 3CP stage and was thus excluded. The LNP 3CP vote share was too high for anyone else to win anyway.
In 2022 it appears certain that the Greens’ Stephen Bates will instead exclude the Labor candidate and because the LNP 3CP vote share has dropped precipitously he will win on the preferences of ballots in this Labor 3CP vote share (which are almost entirely Labor primary votes, but also a few from 4th party candidates).
This is one particular pathway to victory for a Greens candidate, but imagine an alternative seat of Brisbane where 10% of the 3CP vote share was shifted to the Labor from the LNP or vice versa with the Greens staying the same. In either case the Greens would be nowhere near winning the seat. It’s a bizarre quirk of the voting system that the vote share required to win changes depending on the relative popularity of other parties. 29% may be good enough in one race and nowhere near in others.
One more point I’ll make before jumping into the deep end. Notice in the above chart that the three parties’ 3CP vote shares add to 100%. This will always be the case which means that if you know two of the parties’ vote shares you can easily calculate the 3rd.
Three Party Preferred Landscape
I came up for the idea of this graph:
On the x-axis is the Coalition 3CP vote share. On the y-axis is the Greens 3CP vote share. Every point on the graph represents a potential 3CP vote share in a three-corner-contest of the three parties with the coordinates (x,y) corresponding to an election with Coalition on x% 3CP, Greens on y% 3CP, and Labor on (100-x-y)% 3CP. The shaded areas are which party wins on those 3CP numbers (for the purposes of the demonstration I have assumed 80% of Labor and Greens 3CP ballots preference the other party and a 70% of Coalition 3CP ballots preference Labor).
On the left hand side of the graph the Coalition vote is low and they lose to either Labor or the Greens. The boundary between Labor and the Greens is quite complicated. On the far left (ha!) the Coalition is being excluded in 3rd place and their preferences are improving Labor’s chances of defeating the Greens. In the middle whoever of Labor or the Greens come 3rd is out of the race and allow the other party to win. On the right (ha again!) the Coalition vote is high and they win.
The distance of an electorate’s result from changing from one shaded region to another is in some sense its “margin” but unlike the traditional “two party preferred” pendulum model this is a two-dimensional not one-dimensional margin.
Here is a zoom in with with the Greens three QLD gains mapped upon it (if these graphs look ugly try opening on a different browser or a different device):
The three seats all saw a swing away from the Coalition which corresponds to a right to left movement on the graph. Each also saw a swing toward the Greens which corresponds to an upwards movement on the graph. Brisbane and Ryan moved from the blue shaded region to the green one while Griffith moved from the red shaded region to the green one, ie- Brisbane and Ryan were lost by the Coalition to the Greens and Griffith was lost by Labor to the Greens.
I must give credit to my friend Alex Jago who after I shared my idea whipped it up into existence with his programming skill. Have a play around with the graph at his site here. You can edit preference assumptions which is quite cool and makes the tool relevant to any 3rd party or independent (if you have some idea how ballots for other parties in the three-corner-contest may flow).
What Is To Come
In future elections I will be going through the Greens seats, graphing them, and determining what exactly that position on the graph means for what vote changes are required for the Greens to win. Coming soon blog posts will be:
A breakdown of the VIC state election Greens potential target seats
A look at the final distribution of the 2022 federal election to see the true 3CP results
A breakdown of the NSW state election Greens potential target seats
