As the minor party and independent vote has increased, we have seen the voting method used for the House of Representatives take on more complex and difficult to predict features. Nearly every MP has been elected without a majority of the vote, and an independent or minor party made the final count in one sixth of all seats. More broadly, the AEC was not able to call as many races based on the primary vote and two-candidate-preferred vote, with a mathematical chance of the distribution of preferences changing the outcome in about half of all seats.
Most close races typically focus on the two-candidate-preferred (2CP) count, but there have been an increasing number of races where the order of elimination comes into play, with the three-candidate-preferred (3CP) count deciding who makes the final two (and in some cases the 2CP is a foregone conclusion).
Greens candidate Stephen Bates won the seat of Brisbane despite narrowly coming third on primary votes, overtaking Labor and then winning the race with relative ease over the Liberal National Party with the benefit of Labor preferences. There were also questions about the order of elimination in Macnamara and Richmond.
Traditionally the AEC has conducted two-candidate-preferred counts as a matter of course in every seat, starting as soon as possible on election night, but in 2022 they also conducted ad hoc three-candidate-preferred counts in Brisbane and Macnamara which clarified the winner well in advance of a full distribution of preferences.
The AEC has now published the full distributions of preferences for all 151 seats, although unfortunately they are in a user-unfriendly PDF format, not yet in the tidy CSV format available for the 2004-2019 elections. I have produced a spreadsheet with just the percentages for the last three candidates at the second-last round of the count (the 3CP, in other words) which you can download here.
The three-candidate-preferred count will become a more important part of our electoral system, so this blog post is about analysing who makes that count, and how it has changed over time.
First up, let’s look at how the gap between second place and third place has been changing over the last two decades. I originally analysed this gap on primary votes, but actually the 3CP is the crucial round. It accounts for any preference flows from minor candidates, and it’s the actual round which decides who makes it into the top 2. If the 3CP gap between second and third is close, that means a seat can be more marginal and the result less clear.
This first chart shows how many seats have a particular gap at each of the last seven elections.
The mode (the most likely outcome) was in the high 20s back in 2004, and had slowly shifted to the left. You can see an increasing number of seats with a 3CP gap of less than 10%.
This next chart shows the same data but boiled down to the average and the median, and both have been shrinking steadily, from 25-26% in 2004 to around 15% in 2022.
There are ten seats where the gap between second and third on the 3CP was less than 4%:
- Maranoa - 0.20%
- Groom - 0.33%
- Macnamara - 0.64%
- Hume - 0.83%
- Brisbane - 1.66%
- Mallee - 2.07%
- Canberra - 2.20%
- Page - 2.39%
- Richmond - 2.53%
- Sydney - 3.06%
Of the 151 seats contested this year, 126 saw no change in the rankings between the primary vote and the 3CP, and four others saw the first- and second-placed candidates swap places, but they still both made the 2CP. In 18 other seats, the fourth-placed candidate on primary votes made it into the 3CP, and in one of those 18 cases they actually made it into the 2CP. There were a further three seats where the second-placed candidate dropped to third place on the 3CP.
Thus there were four seats where the candidates in the 2CP were not the same top two from the primary vote: Brisbane (Greens overtook Labor), Macnamara, Richmond (Coalition overtook the Greens) and Groom (independent Suzie Holt came from fourth place to overtake Labor and One Nation).
Eleven parties made the three-candidate-preferred in 2022, as well as independents (eight if you count the Coalition as a single unit).
- Coalition - 153
- Labor - 150
- Greens - 89
- Independent - 26
- One Nation - 20
- United Australia - 10
- Lambie Network - 2
- Katter's Australian - 2
- Centre Alliance - 1
There were two seats where both Liberal and Nationals made the top three. In Nicholls, they made the top three along with an independent. Nicholls was also the only seat where Labor didn't make the top three. In Durack, the top three was between Labor, Liberal and Nationals.
Exactly one minor party reached the 3CP in 150 out of 151 seats. The only exception was Durack, but if you treat the WA Nationals as a minor party, you have a complete set of 151.
This next chart shows how this has changed over the last two decades.
It's interesting to see how the number of Greens 3CP entries dropped from 2010 to 2013 and has never really recovered.
The Palmer United Party made the 3CP in almost a third of all seats in 2013, but they went away in 2016.
The Nick Xenophon Team made the top three in all 11 South Australian seats in 2016, and then at the last two elections we've seen an increasing number of independents making that count.
This next chart shows the median percentage at the 3CP point for each of these parties over the last seven elections.
While the Greens has made the 3CP in fewer seats, their 3CP vote in the remaining seats is going up, which is consistent with the Greens making up a smaller share of a larger non-major party vote. The median percentage for independents in the 3CP has also shot up markedly in 2022, which is consistent with their results.
Finally, I've made a map which I think gives an interesting perspective on the relative strength of the minor parties across the country. The map shows which minor party made the 3CP in each seat. The Greens, One Nation, United Australia and independents have their own colours, while the other parties, all of which only made the 3CP in 1-2 seats in a specific geographic area, are all coloured purple.
If you scroll around, you see plenty of green, but the patches of One Nation orange and UAP yellow in outer suburbia is quite clear.
This is my last blog post for now on the federal election. Stay tuned for some new content on other elections, coming soon.
What was the process that would’ve happened in Groom to establish that Holt made it into the 2CP?
I suggested to the AEC IT people that they ought to tweak their systems (and reporting of it via Media Feed) to make 3CP an option for Election Night operations at a National, Divisional and PP level. Given that they had to rejig the 2CP 13 times during Election Night, they might be amenable to processes that streamline things for the DROs.
They are thinking about it.
Of course, there’s no particular logical reason why we need stop at 3CP … maybe they should go for an nCP option 🙂
In theory that’s how preferential voting is meant to work – jumping to 2CPs or 3CPs is just a shortcut!
I also think the “election pendulums” need to be tweaked now as well, to show what their margin was at their closest point of elimination, rather than only the 2CP margin.
Otherwise you have seats like Macnamara listed as “Safe” with a 12.5% margin when their real margin at their closest point of elimination was actually only 0.64%.
I agree with Trent. Labelling seats like Macnamara as “safe” is misleading. If the Greens had gotten 0.64% more primary votes or votes from preferences, Labor would be eliminated and ironically, it would be a “safe” Greens seat. I assume most Labor votes preference the Greens.
Fun fact. Suzie Holt (IND) in Groom scored 8.26% and came 4th but ended up in the 2CP. This must be a record. This also should be a wake up call to the major parties that with tactical voting, preferences and a high minor party vote, a major party could lose to a minor or independent with a low primary vote. I remember Andrew Wilkie won in 2010 with 21% of the primary vote.
Votante and Trent, I would argue that both 2CP and 3CP numbers show the dynamic of the contest (for Macnamara at least). It is a 3 way tie, competitive between the Greens, Labor and Liberal.
However, the seat is considered safe on a left/right axis as only 1/3 of vote is sitting with right wing parties (most of it tied up in the Liberal vote) and the rest captured by the left leaning parties (mostly Labor and Greens).
Therefore the 3CP margin (0.64%) shows that the contest between candidates is indeed close, but the 2CP margin (>10%) is more indicative of the general partisan lean, and that Macnamara is always a progressive seat.
@Geoff I would even suggest that it’s time for the AEC IT people to finally rebuild the Media Feed system and turn it into a REST API with a web-hook system for election night. Their FTP + XML system is very antiquated and been around for years. Time to use an accessible format that developers today widely-use and are familiar with.
I should clarify that the 2PP margins calculated by AEC to get each seat as a Labor vs Coalition count (regardless of the winning party/candidate for each seat) is the best indicator of a seat’s left/right partisan lean.
That would be akin to things like Cook PVI used to measure partisan lean of US states and house districts relative to the nationwide presidential vote.
In a sense the ‘artificial’ 12.5% 2PP/2CP margin for Macnamara would be that seat’s partisan lean (i.e. a safe progressive/left leaning seat)
In fact a useful indicator for Australian elections would be something like a PVI, calculated by deducting the 2PP result for a given seat from the nationwide 2PP.
When applied to Macnamara/Melbourne Ports, the general leftward partisan lean for elections prior to 2016 was only 3-4% above the nationwide average 2PP, whilst it increased to about 6% for 2019 and now to 10% for 2022, clearly indicating the leftward/progressive trend for these affluent inner suburban districts.
My question is what would have happened if Australia used a single vote and voluntary voting besides the fact that the Libs would have the advantage if those two were removed?
I quite like the idea of amending the presentation of the pendulum to include two rows for some electorates – the 2CP, and the nCP in which the winner came closest to elimination (if that count was not the 2CP).
Thanks for the data @Ben, this might be your best post for the 2022 federal election to date. Very informative. I suspected that ONP were very close to making it into the 2CP in Maranoa, thanks for confirming the actual margin.
One stand out division on your map is McPherson on the Gold Coast: low Labor 3CP vote (25%) and relatively decent Greens 3CP vote (~21%). Suggesting that Greens are only around 4% behind taking over Labor on the 3CP. This 3CP contest seems quite similar to Ryan in Brisbane. LNP are only on 53% 3CP. If LNP perform any more poorly and drop below 50%, it is a prime 3CP battleground that Greens might want to consider targeting. Would make sense allocating some resources here if they were also targeting neighbouring Richmond, just over the border. There’s some grassroots anti-development groundswell they could capitalise on in McPherson too. Out of each of the divisions I scanned on your map, this looks like one of the Greens best prospects outside of the other already obvious ones: Macnamara, Richmond, etc.
Conversely, as I expected Moreton looks like it will be quite a struggle on 3CP for the Greens even though in terms of primary-votes, it is their next best division in Queensland (after Griffith, Ryan & Brisbane). Both Labor and LNP have held relatively strong 3CP votes providing a decent buffer against the Greens.
For a bit of fun, I also assessed the 3CP contests from a ONP & UAP prospect perspective:
Similarly to how low primary-votes for Labor benefit the Greens in 3CP contests – low LNP primary-votes generally favour the ONP’s prospects in 3CP contests as seen in some QLD State Election results. One of ONP’s best prospects is Flynn where ONP have managed more than 21% on 3CP with the two majors hovering just under 40% each. In other regional Queensland seats, ONP also managed more than 20% on primary; but in these divisions, LNP achieved much closer to 50% or more on 3CP votes, making them tougher prospects.
Hunter was another good prospect for ONP with an even closer margin between ONP (~18%) and the Nationals (~33%). However, the ONP 3CP in Hunter of 18% was about 4% lower than throughout regional QLD. Labor also performed very strong in Hunter, close to 50% on 3CP. Interesting to note that both Hunter and Flynn are analogous provincial industrial hubs. Expect these two divisions to be strong ONP prospects in times of deep disillusionment within these communities.
UAP is interesting in that they’ve made the 3CP in outer-suburban communities that were flagged as adversely affected by the lock-down restrictions. One stand out in Sydney is Werriwa with UAP on 16% and LP on 36%. I remember very late speculation in the election campaign that Werriwa was an increasing prospect for the Liberal Party based on this sentiment that some observers were tapping into in Western Sydney.
UAP’s most interesting standouts however are some of the seats in Melbourne: Holt & Bruce in the East with ~16% UAP & ~33% Liberal Party. And finally, in the West – Gorton, where UAP achieved 18% and LP just under 31%. Both Melbourne’s outer-West and outer-East will be interesting to watch at the Victorian State Election assuming the same level of resentment sticks around.
@Yoh An, you could produce a PVI from the dataset I used for this blog post where I looked at each seat’s 2PP rank.
Agree Ben, the rank is probably a better measure of left/right partisan lean. For the example of Macnamara/Melbourne Ports, it is consistently left leaning apart from 2016 where these small ‘l’ liberal voters may have preferred a moderate/centrist leader and local Liberal candidate in Malcolm Turnbull and Owen Guest respectively, thus narrowing the left/Labor lean of this seat.
I like Nicholas’ idea of the two column pendulum: the 2CP margin and the nCP (closest to elimination) margin, where it differs from the 2CP.
I think something like that will be increasingly important, as it’s also relevant in seats like Brisbane & Ryan where even though the difference between 2CP and 3CP isn’t the difference between “safe” and “marginal” in those seats, the Greens v Labor 3CP margins were still both considerably closer than the Greens v Liberal 2CP margins, meaning mathematically Labor are still closer to winning those seats off the Greens than the Liberals.
Yoh An, I also agree that the left/right lean as compared to the national 2PP would be useful to see more commonly used as well. I think Ben does a good job of presenting that on his seat analyses on this site but it should be more common, because as you say it’s a great way to judge how a seat is trending over time.
My question is what does the 2pp mean in seats where one party may not win eg in Mayo where the alp outpolled the liberals on the 2pp even though they did not win the seat .ca did and in a normal contest would not win.
Hey Ben,
When I click on the ‘download here’ link. It takes me to the AEC downloads site. Not to your spreadsheet.
Whoops, sorry Jake. Fixed now.
Some of the above points about being counted in the 2PP but not being close to winning in reality make sense.
For example, in the seat of Melbourne, the 2CP was Greens vs Liberals at a few elections but hypothetically if the Greens were not running, Labor would win for sure. This is because of tactical voting. The same happened at the 2011 NSW state election in Balmain (an affluent, inner-city area that used to be solid Labor) when the Liberals came first in the primary vote but the Greens won the seat and the 2CP was Greens vs Liberals.
In Griffith, the Greens vs Liberals contest makes sense whilst the story is a bit different in Brisbane. Unlike Melbourne and Macnamara, Brisbane had a real three-way contest.
Party 3CP % Prim %
LNP 41.48% 37.71%
GRN 30.09% 27.24%
ALP 28.43% 27.25%
Let’s say that the LNP vote at the next election, without the incumbency advantage. goes down 2% and the ALP vote goes up 2% in the 3CP but the Greens vote is the same. The ALP would obviously win.
Greens should target LNP held Casey and Sturt (in that order), and ALP held Moreton and Perth, plus 2019 targets Richmond, Macnamara and Higgins. There’s also the LNP preference set – Cooper, Wills, Canberra, with 2nd tier Cunningham, Newcastle, Fraser (Grayndler if Albo loses job, Sydney if Plibersek is a terrible environment minister).
I suspect a small ALP to LNP swing, but Greens can make a lot out of a “disappointing ALP government” election, even enough to build momentum in the LNP held marginal targets.
Gold and Sunshine coasts are places Greens can come 2nd, but hard to see LNP taking a big enough hit there to be at risk in opposition (especially with leader Dutton). Those local group also aren’t as organised as Brisbane based Greens – I think there’s a story to why Noosa isn’t a Green hotspot any more that isn’t just the rise of Sandy Bolton.
Something that would be interesting to find out is if you took the existing ballots in the really close 3CP races and used a Condorcet method of counting them rather than Instant Runoff if any of the results would have been different. I could see that perhaps being the case in Macnamara, Griffith or Brisbane in particular.
Comments are closed.