Greens at the Gates?

Introduction

Recently, the leader of the Australian Greens, Adam Bandt, has claimed that Australia is likely headed to a hung parliament (where no party or formal coalition has an outright majority of seats and therefore negotiations have to occur to form a government), declaring that:

“The maths just says we are heading towards a power-sharing parliament; (there) is a swing against the government…It shows Morrison being pushed out of majority government but not enough for Labor to win in its own right.”

Adam Bandt, The Australian

In response, the leader of the Labor Party, Anthony Albanese, has rejected the idea of a power-sharing deal with the Greens. Leaving aside the issue of whether or not “the maths just says” that a hung parliament is likely (as per Dr Bonham, it doesn’t, but I will have more to say on that later), I first want to examine the rationale that most people give for why Labor would reject the idea of forming a government with the Greens.

Namely, the idea that a formal coalition between Labor and the Greens would be “electoral poison” or “disastrous” or (insert dramatic synonym for bad here) for Labor’s electoral prospects. Proponents of this view tend to bring up disastrous Labor-Green governments who proceeded to lose the next election such as the 2010 Labor minority federal government and the 2010 Labor-Green government in the state of Tasmania. In response, proponents of a Labor-Green coalition tend to bring up the successful Labor-Green coalition government which has governed the Australian Capital Territory (ACT) for over a decade.

Is it true that Labor-Green coalitions tend to be bad for Labor? Well, below are all the Green-supported governments which have been formed in Australia along with the swing to/against the major party and the swing to/against the Greens at the following election:

Partnering with the Greens has (so far) usually cost both parties votes

The first thing to note here is that the sample size is tiny (there have only been seven governments in Australia formed with the support of the Greens) and very heavily dominated by the two jurisdictions which use proportional representation in their chamber of government (Tasmania and the ACT). Hence, a lot of caution should be taken in extrapolating exact results from this dataset to the broader nation; we’re looking for trends here, not exact figures.

Still, for a general trend, the idea that major parties tend to go backwards when they partner with the Greens to form government seems pretty strong. Six of the seven governments above saw swings against them at the next election, with the only government to have avoided that fate (the Stanhorpe/Gallagher Labor-Green government in the ACT) seeming to have done so at the cost of its Green partner (as well as falling behind the Liberals on votes). Similarly, of the seven times the Greens have supported the formation of a minority or coalition government, they went backwards on votes six times, with the sole exception being the 2020 ACT general election.

Again – the sample size here is tiny, and none of this should be taken as conclusive; it is entirely possible that we will someday get a highly popular Labor-Greens coalition government which is re-elected with swings towards both parties. At the same time, though it may not be conclusive, an average swing of 3.3% against the major party (-5.8%, if you exclude the ACT) with a negative swing in six of seven elections is highly suggestive of the idea that a major party will probably lose some votes by partnering with the Greens to form government.

(Or, to paraphrase xkcd: It’s not conclusive, but it does waggle its eyebrows suggestively and mouth “Look at that trend”)

Another point to note is that historically, Tasmania and the ACT have been two of the most Green-friendly jurisdictions both federally and at the state/territorial level. (In fact, while the Green vote in Tasmania used to be 2 – 8% higher than the nation from 2001 to 2010, ever since the 2010 election where the Greens helped Labor to remain in power, the Green vote in Tasmania has been slightly lower than the nationwide Green vote, an interesting data point in the coalitions-hurt-Greens discussion) If Labor still suffers an erosion of its vote share after the formation of Labor-Green coalition governments in the Apple Isle and Canberra, can federal Labor expect to hold its vote share in less Green-friendly states such as NSW or Queensland if it goes into another Green-backed coalition?

Counterpoint: maybe minority governments of any stripe would have fallen apart anyway, Greens or no Greens. After all, in many cases (outside of Tassie and the ACT) Australian minority governments are produced when an incumbent government nearly gets tossed out but barely manages to hang onto enough seats to negotiate their way into power (e.g. federal 2010, SA 2014, SA 1989). In a few other cases they can occur when an opposition manages to win enough seats to cobble together a government without having won the popular vote (e.g. SA 2002) – neither sound like good strategies for a stable, popular government! So what does the average swing for a newly minted minority government of any stripe look like?

Minority governments tend to go backwards at their next election, but not as much as Green-supported ones do

While the average minority government has seen a drop-off in support regardless of whether it was Green-supported, it’s somewhat smaller than the average Green-supported minority government. Specifically, for an apples-to-apples comparison (and excluding the weirdness of Liberal/National coalition-formation which is unlikely to be repeated federally), governments backed by the Greens have suffered a 1.3% bigger swing against them on primary votes compared to governments backed by other groupings/independents; with the difference being somewhat larger (2.3%) when results from the ACT are excluded.

(For comparison’s sake, the decline in the Labor primary at the 2013 federal election was 4.6%)

Again, I should stress that this is a small sample, and the data provided here is not conclusive by any means, but it does seem to suggest that Green-backed government can be a turn-off to a segment of the electorate. (In particular, I suspect a thorough analysis of polling in Tasmania would reflect this) This may someday change – young voters in particular seem to be more comfortable with voting Green than their older counterparts and might well be more supportive of a Green-backed government – but for now it does seem like the conventional wisdom of Green-backed government being a/an (insert dramatic synonym for bad thing here) for Labor seems to be broadly correct.

Does “the maths just say” that we’re headed for a hung parliament?

Normally, this would be the focus of a piece of mine tearing this assertion apart, but thankfully, Dr Bonham has done a large chunk of the work on that particular claim. There are just a few things I’d like to add to that piece.

(Apart from, of course: what the hell do you mean ‘the maths just says’? Unless you have mathematical proofs, math doesn’t bloody talk to you; the data you have is at best a whisper in a hurricane. Please stop claiming mathematical certainty for your Monday Motivation propaganda, and stop unfairly damaging the credibility of scientific industries in the process.

And this goes out to all politicians, not just Adam Bandt.)

Let’s take its broad assertion as correct. That, if replicated at an election (a very big if), Labor’s 2-party-preferred (2pp) polling of about 50 – 52% would result in a hung parliament. Does that mean that a hung parliament will happen or is even more likely than not?

No.

Polls are not point predictions. They are probability distributions. I’m planning to write more on this at some point, but basically, when a poll comes out with a 2-party-preferred estimate of, say, Labor 51%, that does not mean that Labor will win 51% of the vote. It also does not mean that (with a typical theoretical margin of error of about 3%) Labor’s vote will be between 48% and 54% and any value between those two is equally probable.

It means that the outcome with the highest chance of happening is 51%, with a slightly lower but nearly as high chance of Labor winning 50% or 52%, and a lower-still chance of Labor winning 49% or 53%, and a yet-lower chance of Labor winning 48% or 54%, and a small (but not zero – it’s about 1-in-20) chance of Labor either winning more than 54% or less than 48%.

In other words – polling is a bell curve. Allow me to demonstrate with a diagram.

Let’s say that if Labor wins more than 51.5% on the 2pp, they will probably form a majority government. (The largest 2pp which didn’t result in a government was 50.98%, so 51.5% is a fairly reasonable upper bound.) Let’s also say that if Labor loses the 2pp – that is, if the Coalition wins more than 50% of the 2pp, the Coalition will probably hold onto majority government due to marginal sandbagging and candidate effects and whatnot. Federal governments usually survive adverse swings, with federal governments having held onto power with losing 2pps up till 49.02%, so this serves as a lower bound.

Now, let’s look at the polls. I’m going to take the latest Morgan, because its 2pp estimate of Labor 50.5% is clearly within the “hung parliament” range. The average error on the 2pp estimate for polls taken within the last week of the election is about ± 1.6% (margin of error of about ± 4%); so even if this Morgan had come out right before the next election, the distribution of outcomes would look something like this:

The area marked in red is where Labor’s 2pp is greater than 51.5%, while the area marked in blue is where the Coalition’s 2pp is greater than 50% (and the purple area is our “hung parliament” region). The purple bars are the highest, so they’re the most likely outcome, right?

Nope. In fact, even with a 2pp pretty much smack dab in a dubious “hung parliament” range, a hung parliament is still the least likely of the three outcomes. It is actually much more likely that either side – Labor or the Coalition – win an outright majority than it is that a hung parliament occurs, simply because the blue and red areas combine for a heck of a lot more than the purple. Even with our highly favourable assumptions, for purple (hung parliament) to be the most likely option, the polls have to be right in the middle of “hung parliament range” and very, very accurate (about twice as accurate as they’ve been historically). In other words, even if a poll’s 2pp estimate lands in “hung parliament region”, due to the nature of polling and probability distributions, that does not necessarily imply that a hung parliament is the most likely outcome.

That’s not to mention other issues. I alluded to this before, but there can be a lot of uncertainty around how many seats a particular two-party swing will deliver; governments can hold on against adverse swings like the Howard government did in 1998, holding onto 80 seats when a uniform swing would have cut them down to 68 and cost them government. On the other end, when Howard was tossed out in 2007, Rudd’s Labor won 3 seats more than they would have on a uniform swing; while at the next election, a uniform swing of 2.6% to the Coalition should have left Labor with just 68 seats (accounting for Melbourne and Denison being won by Adam Bandt and Andrew Wilkie respectively) but Labor instead managed to hold 72 and negotiate its way into government.

All of this to say that at every federal election, there is a lot of variation in swings between each electorate (on average, the difference between a random electorate’s 2pp swing and the national 2pp swing is about ± 2.4%), and a variety of factors such as popular local candidates, demographics, campaign issues etc can remake the map to save a government or amplify a swing against it. Uniform swing pendulums are useful for rough estimates of how many seats a party might win with a given swing, but swings are never uniform in real life and claiming that uniform swing pendulum shows that X would not win enough seats for majority is dangerously overconfident at best.

One final issue is with the idea of a Labor/Green power-sharing parliament specifically. Assuming the Coalition regains Craig Kelly’s seat of Hughes, the crossbench will consist of 4 MPs in Coalition-leaning seats (Indi, Warringah, Mayo and Kennedy) and 2 MPs in Labor-leaning seats (Melbourne and Clark). While obviously the 2-party lean of a seat doesn’t necessarily translate into that crossbencher’s support, Zali Steggall (Warringah) and Rebekha Sharkie (Mayo) have previously said they would back a Coalition government while Katter (Kennedy) has previously backed the Coalition in a hung parliament.

This means that, for a “power-sharing parliament” the likes of which Mr Bandt describes, Labor must win a minimum of 73 seats – it simply isn’t enough to reduce the Coalition to 75 or 74 seats as they will probably be able to get back into power with support from independents in conservative seats. It also makes a Labor-Green parliament that much less likely; if Labor wins fewer than 73 seats, the Coalition could get back in, whilst if Labor wins 75 seats, they can opt to negotiate with any of the independents and sideline the Greens (and if Labor wins 76 or more, it’s a majority). In other words, there’s a very tiny band of Labor seats – either 73 or 74 – where Labor would both be in a position to form government as well as probably have to seek Green support to form government, making the specific type of hung parliament (a Labor/Green parliament) even less likely than hung parliaments in general.

(while the Greens could gain Labor seats, the Coalition too could gain crossbench seats [Indi (very close in 2019) and Warringah (no Tony Abbott factor) are probably key targets, Mayo might be too] to stay in majority despite losing Labor vs Coalition seats)

So does that mean that a hung parliament isn’t going to happen?

Well…no. It could happen, but it’s just not very likely in general.

We’re still (probably) at least a few months out from a federal election, and there is a lot of uncertainty regarding what the final vote will end up looking like. Maybe Labor does something unpopular and the 2pp blows out to 53-47 against them – in which case the hung parliament question will pretty much become moot. Maybe the Morrison government will experience a scandal which actually affects less-politically-engaged voters’ opinions of them, and the 2pp lead expands to 54-46 in Labor’s favour; in which case Labor will almost certainly win some kind of majority, with the only question being where they win it.

Given we don’t even know when the next election will be held, and in what conditions it will be held in (e.g. Lockdown? Booming pandemic recovery? Bushfires?), it’s pretty foolish to claim that current polling suggests that a hung parliament is the most likely outcome. As I demonstrate above, even if the final polling prior to the election demonstrates a close 2-party-preferred, that is still no guarantee of a hung parliament due to uncertainty in polling and non-uniform swings between electorates.