RCV in the 2022 Australian Election

Last year I wrote about the evidence from Australia on how third parties fare under Ranked Choice Voting (RCV), both the single-winner version, Instant Runoff Voting (IRV), and the proportional multi-winner version, Single Transferable Vote (STV). This year there was another election, and it gives us a lot more data — with some very interesting results. I won’t be repeating much from the previous article, so look there if you’re wondering what a first preference is or where the Liberal party stands politically.

Who did voters like, and who won?

First preferences and seats won in the 151-member House:

• Liberal/National coalition: 35.67% of first preferences, 58 seats won (38.41%)
• Labor: 32.58% of first preferences, 77 seats won (50.99%)
• The Greens: 12.25% of first preferences, 4 seats won (2.65%)
• All independents: 5.29% of first preferences, 10 seats won (6.62%)
• One Nation: 4.96% of first preferences, 0 seats won
• United Australia Party: 4.12% of first preferences, 0 seats won

The Center Alliance and Katter’s Australia Party each won a seat as well; their candidates are better thought of as independents but they aren’t included in that total. There are many other small parties as well, none of which won any seats.

First preferences and seats won for the Senate, where 40 seats were up for grabs:

• Liberal/National coalition: 34.21% of first preferences, 15 seats won (37.5%)
• Labor: 30.09% of first preferences, 15 seats won (37.5%)
• The Greens: 12.66% of first preferences, 6 seats won (15%)
• One Nation: 4.29% of first preferences, 1 seat won (2.5%)
• United Australia Party: 3.46% of first preferences, 1 seat won (2.5%)

An additional two seats were won by candidates who were basically independents. There were many other parties that did not win seats, the largest among which was Legalise Cannabis Australia with 3.33% of first preferences.

How proportional were the outcomes?

In the Senate, which is elected via Single Transferable Vote (the proportional form of RCV), the larger third parties received seats roughly proportional to the votes they received. Not so in the House, which is elected via Instant Runoff Voting (single-winner RCV). The Greens quadrupled their representation from 2019, but they still got less than a quarter of the representation they deserved. The other third parties got completely shut out, despite winning seats in the much-smaller Senate. Talking about proportionality gets difficult when you consider the independents though — more on them later.

Labor felt the Green Squeeze

The Greens won fours seats in the House:

Melbourne was the only electorate with a Green incumbent (Adam Bandt, the party leader). Unlike the previous two elections, this time he faced the Labor candidate in the final round instead of the Liberal, and he won by twenty points. Bandt would have won easily under any voting method.

Brisbane was won with a 7.5% margin in the final round over the Liberal candidate. The penultimate round was much closer:

• Green: 30.09%
• Labor: 28.43% (eliminated)
• Liberal: 41.48%

Supposing the later preferences of Liberal voters in Brisbane were the same as the later preference of Liberal voters in the average electorate where we have data, with 34.24% preferring the Greens to Labor and 65.76% preferring Labor, the Labor candidate would have defeated the Green head-to-head, 55.7% to 44.3%. (Also, supposing the later preferences of Green voters were the same as the national average, the Labor candidate would have defeated the Liberal head-to-head, 54.2% to 45.8%. This is a tad off; the actual numbers are listed in the detailed election results and Labor would have gotten 54.40%.)

Brisbane clearly had a center squeeze — a scenario where a candidate who would beat anyone else head-to-head gets eliminated before the final round. For there not to have been a center squeeze would have required the Green to be ranked above the Labor candidate on 48% of the ballots that went for the Liberal in the penultimate round, and their best performance in this respect anywhere we have data was in Canberra, where the Green got 40.52% of the Liberal vote.

The Greens probably won in Griffith because of a center squeeze too. The penultimate round:

• Green: 36.37%
• Labor: 29.71% (eliminated)
• Liberal: 33.91%

(The Liberal got crushed in the final round.)

If the Liberal voters behaved according to the national average, the Labor candidate would have defeated the Green, 51.0% to 49.0%. The Green would have needed the support of 40.2% of Liberal voters to defeat the Labor candidate head-to-head — well above the national average, but plausible.

The Green victory in Ryan was probably due to a third center squeeze. The penultimate round:

• Green: 33.26%
• Labor: 23.76% (eliminated)
• Liberal: 42.97%

If the Liberal voters behaved according to the national average, the Labor candidate would have defeated the Green, 52.0% to 48.0%. For the Green to have beaten Labor head-to-head would have required the support of 39.0% of Liberal voters.

(Interesting note: In Ryan, the Green candidate did slightly better against the Liberal than the Labor candidate would have. The Green got 52.65% in the final round, but the Labor candidate would have only received 52.42% against the Liberal.)

The Greens didn’t quite win in Macnamara, but it was very close. The penultimate round:

• Green: 32.84%
• Labor: 33.48%
• Liberal: 33.67%

If merely a third of a percent of the electorate had voted for the Greens instead of Labor, the Greens would have won this seat even though they’d get absolutely trounced head-to-head against Labor. Richmond is another electorate that the Greens came close to winning via a center squeeze, though not as close as in Macnamara.

Bottom line: The Greens won four seats, but probably would have lost head-to-head against the Labor candidate in all but one of those races.

The teal independents: a centrist success

One of the most notable differences from 2019 was the remarkable success of the teal candidates. (Why teal? Teal is a color between blue and green, where blue is the color of the center-right Liberal party and you can probably guess which party is characterized by the color green.) The teal-aligned candidates won 10 seats in the House, including four won by incumbents. (Three of those incumbents were independents; the other was Rebekha Sharkie, the only candidate in the Nick Xenophone Team who won a seat in the House back in 2016.) I have compiled the results for teal candidates in this spreadsheet; I identified the teal candidates based on this table and copied the relevant data from the official results.

By all indications, the teals are seen as centrists. All but three of their 20 House candidates ran in seats where the Liberals were preferred over Labor, and they won 8 of these 17 races. About 77% of Labor voters preferred the teal to the Liberals, and the only race in which a teal candidate faced off against Labor in the final round showed 74.96% of Liberal voters preferring the teal.

Since the teals were centrists, were they the victims of any center squeezes? For the most part, they couldn’t have been. The teal candidate reached the final round in 14 of the 20 House races with a teal candidate. Of the 4 races where the teal candidate finished third, the Liberal candidate had so much support in the penultimate round that there’s no way the liberal wasn’t the Condorcet winner. In the last two races, the teal candidate finished fourth — behind Labor, the Liberals, and the Greens. In Casey, the teal candidate got 15.02% of the vote in the round she was eliminated, and in Boothby, she got a mere 9.36%. However, there is a real possibility of a center squeeze in each of these final two elections.

The central challenge in determining whether a center squeeze occurred in an Australian election is that Australia does not provide a pairwise comparison matrix that tells us the voters’ actual preferences. My fallback is to assume that the ratio of (for example) Labor>teal>Liberal voters to Labor>Liberal>teal voters is the same as the average ratio in electorates where we do have these data. To handle cases with four candidates, I pretend that the Green candidate is eliminated first, and distribute their votes among the other three candidates in accordance with the average in other races with the same parties (and a teal independent) in the top 4, and do likewise with the the Liberal or Labor candidate in the imagined penultimate round that includes the teal candidate. See the spreadsheet for the calculations.

Using this methodology yields these estimates:

• In Casey (where the Liberal actually won), the teal would have lost to the Liberal head-to-head, 48.2% to 51.8%, and beaten the Labor candidate 56.6% to 45.4%.
• In Boothby (where Labor actually won), the teal would have lost to the Liberal head-to-head, 48.9% to 51.1%, and beaten the Labor candidate 50.2% to 49.8%.

These estimates have substantial uncertainty; a particular weakness is that I am inferring the preferences for Liberal voters between teal and Labor candidates from a single election (in Clark) in which the teal was the incumbent. But if we take these numbers at face value, as we shouldn’t, we find that IRV elected the Condorcet winner in Casey and that there was a Condorcet cycle in Boothby. But both elections are close, and either could have been a center squeeze. I rather doubt that there was actually a Condorcet cycle in Boothby; the band of preferences in which the teal candidate would have beaten Labor while losing to the Liberal is fairly narrow. However, I am confident that the teals should have appeared competitive in these races. A quick glance at the IRV results (especially with the focus on first preferences) makes it looks like the teals were irrelevant in Casey and Boothby, but in each race, the teal candidate came at least close to being the Condorcet winner.

My best guess is IRV elected the Condorcet winner in every election involving a teal candidate. The teals were weird. They achieved a great deal of electoral success — they won fully half of the seats in the House where they had a candidate — but they fielded remarkably few candidates. A big reason why we didn’t see clear center squeezes is that voters on the Left tended to rank the teals higher than Labor; the teal had over twice as many first preferences as labor in several electorates. Another factor is that races in which the teal candidate finished third occurred in the teal-contested electorates where the Liberals were the most popular, having an outright majority (or very close) even before the teal candidate was eliminated and defeating Labor by over 13 points in every single one of them. Had the teals been a true national party, fielding candidates and campaigning in almost every electorate, I suspect they would have fallen prey to a great many center squeezes.

How did the teals do in the Senate? For the most part, they didn’t try. They won a seat in the Australian Capital Territory — one of the two two-winner Senate races, and therefore a race that is usually one of the least hospitable to third parties. They fielded a candidate in only one 6-winner race, who lost. Given the lack of competition in the Senate, it’s mostly incorrect to think of the teals as a third party, even though approximating them as such is useful for identifying center squeezes.

Strategy and the Teals

One big question that I cannot answer: To what extent were the teals helped by strategic voting? To quote an actual Aussie:

[I]f someone’s sole concern is beating one of the major parties, then the correct strategic move is to preference the indie above both major parties, putting the party they want to lose last. If the indie ends up being uncompetitive, it doesn’t matter anyway as their preference flows to the preferred major party at full value.

That is to say, a voter who wants to oppose the Liberals is best off ranking a teal independent above Labor, even if that voter prefers the Labor candidate over the teal. Strategic voting prevents center squeezes, and if many Labor supporters voted strategically that would explain why the teal independents got so many more first preferences in the electorates they won than Labor did. I’m very skeptical of this explanation though; in my (relatively brief) search of news coverage for these races, I found no mentions of strategic voting, which suggests that it isn’t widespread. Furthermore, strategic voting under IRV is far less intuitive than it is under plurality, so I doubt that more than 10% of voters would employ it without extremely visible direction.

What did matter, I think, was campaign spending. According to the UQ Election Ad Data Dashboard, in seats that were either won by a teal insurgent or which were hotly contested by a teal candidate, there was heavy spending on the teal candidate but little or nothing was spent by Labor. But in Boothby, Labor outspent the teal candidate by at least a 5-to-1 margin. (I didn’t see data for Casey.) The teals were only successful in electorates that Labor didn’t even contest.

Basically, center squeezes with the teals weren’t prevented by strategic voting so much as by strategic campaigning. This had a perverse consequence: Labor was the most interested in spending money in electorates that were evenly split between Labor and Liberal voters, and this spending meant that teal candidates (aside from the incumbents) could only be competitive in electorates that favored the Liberals. The centrists couldn’t win in the electorates closest to the political center.

Ultimately, I think Boothby (probably along with Casey) was just the tip of the iceberg. If Labor, the Liberals, and the teal candidate had equal funding, the teal candidate probably would have been the Condorcet winner but would have had a serious risk of being eliminated before the final round. But the funding wasn’t equal, so we can’t infer a center squeeze from the election results. I suspect that this is equally true in electorates that were hotly contested between Labor and the Liberals and which didn’t feature a teal candidate: a teal-aligned candidate could have entered the race, but didn’t as a result of the funding situation. Center squeezes could have been widespread, but were preempted by the strategic allocation of campaign money.

Conclusions

In my previous Australia post I listed four conclusions. I think they mostly hold up, but two of them are worth revisiting:

Center squeezes are less common than you might expect. The simplest and most intuitive model of politics is that candidates are on a one-dimensional Left-Right spectrum; in this model center squeezes are basically ubiquitous. The Australian data show that this model is largely wrong; instead, it’s extremely rare to have over 90% of a minor party’s supporters prefer one major party to another, and similarly with regards to preferences of major party supporters between a minor party and the opposing major party. Intuitively, if 45% of voters prefer a leftist, 45% prefer a rightist, and 10% prefer a centrist, we might expect the centrist to be the Condorcet winner. Wrong. In practice, enough left-wing voters will prefer the rightist (and vice versa) that the centrist should never win.

This has held up in the sense that none of the results from the 2022 election contradict it, but center squeezes still played a big role. I’d now say that center squeezes are uncommon in elections that look like two-way races. Center squeezes seem quite common in Australian elections with three or more truly viable candidates.

The idea that other single-winner voting methods would break the two-party system whereas IRV wouldn’t is dubious. STAR and Approval Voting yield far more visibility for minor parties than IRV, but this isn’t as much of an issue in Australia where minor parties are represented in the Senate. Maybe third-party visibility is important in Australia insofar as the visibility gained in the Senate allows the Greens to get 10% of the vote instead of something like 3%, and it’s likely that using STAR or Approval instead of IRV could yield a similar effect. But the difference between 3% and 10% support is pretty slim in single-winner elections. Either way, you’ll lose almost everywhere.

The 2022 election has done nothing to refute this. In fact, the 2022 election suggests the opposite possibility: might IRV be better for third parties than voting methods like STAR and Approval which aren’t vulnerable to the center squeeze? After all, three of the four Green victories in the House were most likely due to center squeezes, so the Greens probably would have won fewer seats under STAR or Approval.

I think that IRV is better than STAR or Approval for extremist third parties provided they receive a lot of visibility from something like Australia’s Senate. The qualifiers are important. The advantages of IRV for the Greens are disadvantages for centrists like the teals. And from the perspective of American politics, would having additional extremist parties actually make things better off? I think the two-party system is a big problem, but the additional parties I see as being the most helpful would either be in the center or would defy the Left-Right spectrum altogether. Visibility seems important too; I doubt the Greens would have had anywhere near this much success without the Senate keeping them on the radar.

My takeaways from the 2022 election:

1. IRV’s pro-extremist bias is important. Three-quarters of the Green party’s wins in the House were probably due to the center squeeze, and a brief glance at the official results makes the teals look like they were irrelevant in the electorates that were seriously contested by Labor, even though the teal candidates would have been at least competitive head-to-head against anyone for those seats. Had Approval, STAR, or a Condorcet method been used, both the results and the campaign strategies would have looked very different.
2. Money matters. When the teal gets five times more first preferences than Labor in Kooyong (where the teals outspent Labor massively), but Labor got five times more first preferences in Boothby (where Labor outspent the teals massively), it really looks like money makes a difference. I don’t understand how money is important, but I think this is a critical ingredient to understanding what makes independents and third parties succeed or fail.
3. Due to a lack of serious campaigning by three or more candidates, most center squeezes don’t materialize. From what I can tell, the Greens targetted a handful of House seats while putting essentially no resources behind any of their House candidates elsewhere. Similarly, the teals only fielded 20 House candidates, despite their high popularity where they ran, and I’m unsure if any seats were hotly contested by both the teals and Labor. The result is that the vast majority of races were effectively just two-way races. It’s easy to say something like, “There were center squeezes in 2% of races (3 out of 151) and 2% is a small number, so center squeezes are rare and not very important.” But this doesn’t tell us what we care about. At least in Australia, center squeezes are common in races with more than two serious candidates. It’s races with more than two serious candidates that are rare.
4. Single-winner elections still can’t deliver proportional results. The Greens got their best showing yet in the House this year, but they still won a fraction of seats that was far smaller than their share of the vote.
5. When IRV struggles, it still beats Plurality. I haven’t bothered analyzing how many races would have been decided by the spoiler effect had Plurality been used instead of IRV, but it’s pretty high. And of the three probable center squeezes, the Liberals (the Condorcet losers out of the top three parties) would have won two of them had they been conducted using Plurality. The 2022 Australian elections are the clearest demonstration of the flaws of IRV I’ve seen yet, but they are an even clearer demonstration of IRV’s superiority over Plurality.

Appendix: List of electorates in which a Green finished in the top two:

(Note: I’m listing the percentages based on preferences in the penultimate round, not on first preferences)

• Grayndler: Greens lost to Labor in the final round. Greens won 39.55% of the Liberal vote, Labor won 60.45%
• Sydney: Greens lost to Labor in the final round. Greens won 36.04% of the Liberal vote, Labor won 63.94%
• Cooper: Greens lost to Labor in the final round. Greens won 32.68% of the Liberal vote, Labor won 67.32%
• Melbourne: Greens beat Labor in the final round. Greens won 29.97% of the Liberal vote, Labor won 70.03%
• Wills: Greens lost to Labor in the final round. Greens won 26.66% of the Liberal vote, Labor won 73.34%
• Brisbane: Greens beat Liberals in the final round. Greens won 83.15% of the Labor vote, Liberals won 16.85%
• Griffith: Greens beat Liberals in the final round. Greens won 81.05% of the Labor vote, Liberals won 18.95%
• Ryan: Greens beat Liberals in the final round. Greens won 81.56% of the Labor vote, Liberals won 18.44%
• Canberra: Greens lost to Labor in the final round. Greens won 40.52% of the Liberal vote, Labor won 59.48%

On average, the Greens were ranked above Labor by 34.24% of voters who preferred the Liberals, and the Greens were ranked above the Liberals by 81.92% of voters who preferred Labor. Nationwide, 85.66% of voters who had a Green candidate as their first choice ranked Labor above the Liberals.