Rockingham By-Election 2023: How Big A Swing Is Too Big?

Note for Tas audiences: Cassy O'Connor recount is covered here

-----------------------------------------------

Rockingham (WA) (ALP 37.7%)

By-election for resignation of Mark McGowan (ALP), MP since 1996

CALLED 7:40 pm Magenta Marshall (ALP) retains, will probably just be taken to preferences

Second place after preferences TBD between Hudson (Lib) and Edwards (IND) (edit: Edwards was second.)

Live Comments

Sunday: It is notable that Labor's 2PP at a by-election with the seat vacant will exceed its 2PP in this seat at three elections that Labor won and where its candidate in this seat was an incumbent.  Also the 2PP swing is about the same size as Balga (1988) despite the baseline state 2PP being over 15 points higher.  The by-election is consistent with Labor still being a long way in front statewide.

With the gap between Hudson and Edwards only 1.7%, I note that William Bowe projects (i) that Edwards will overtake Hudson on preferences (ii) that the Labor-Edwards 2CP will finish around 61-39.

9:44 The 2PP swing snuck up to 22.5% but I'd say this is only just acceptable for the Liberal Party and still consistent with them being heavy underdogs statewide.  

8:40 Not sure if we're getting a 2PP count on the remaining c. 5300 votes that have been counted to primaries tonight.  The 2PP swing currently sits at just below 22%; again, nothing to see there.  If the poll released this week were true and this seat were representative we would be seeing much, much more than this.

7:50 A partial 2PP count has appeared with a swing of only about 15% but I don't know which booths that's from and it might be a tad low.  Also of note here is Legalise Cannabis beating the Greens again (as they did in Fadden) but the LC candidate is a local councillor.  In 2021 in the Legislative Council election the Greens outpolled them 5.59-3.66.

7:40 With the addition of significant postals and prepolls Marshall has obviously won.

7:30 The primary vote swing will be massive (currently about 31%) but because those votes are spraying, the 2PP swing's not so big (it might be currently 21%).  In terms of this as a sign that the Liberals are back in the game, it's pretty borderline so far.  

7:12 Across the board swings of 30-35% so far, so while Labor might or might not win on primaries there is still no problem for them here as everyone else is way back on 15% or less.  Edwards is now challenging for second on the primary vote and while her position might not survive postals it is also likely she could pass the Liberals on preferences, so it is not yet clear this is a classic contest.

7:00 A couple more booths with no concern for Labor at all on primaries result-wise but also a big spread of primaries with a high Legalise Cannabis vote so far.

6:53 The first booth is in with the small matter of a 32.2% primary vote swing against Labor - which if continued is harmless in terms of the result but they wouldn't want it to ramp up another 5% or so.  In this booth Edwards has only 10% which is not especially promising though her support may vary across the seat.  

6:00 Polls have closed. William Bowe has noted that the WAEC doesn't publish it's indicative 2CP splits until it's confident it has the right top two.

Some live comments (scrolling to top) will be posted during the count, and post-count commentary if the seat remains in doubt.  I do not intend this to be as detailed as the Fadden by-election but refresh every 20 mins or so from about 6:45 pm WA time onwards and there may be something!  Note that there will be detailed coverage at Poll Bludger.

---------------------------------

Intro

I wasn't intending to cover what looked like one we starling-staters really shouldn't stay up for but tomorrow's WA state Rockingham by-election count has developed some signs of life with the publication of an Utting Research "robopoll" purporting to find the Liberal Party (all two of them) ahead of Labor statewide, 54-46.  That would be a 24% swing from the 2021 election landslide, with the cited causal factors including the retirement of the political legend that is Mark McGowan, the demise of COVID politics and backlash against the state's Aboriginal Cultural Heritage Act.  

Before we get too far ahead of things here, Utting Research was not the most reliable of pollsters at the 2022 federal election.  The poll produced results in March that were a pretty good portent of what was to come, but the West Australian's article talks those up without mentioning a final sample in May that had things much closer only for them not to be, underestimating the average swing to Labor across four seats (Hasluck, Tangney, Pearce and Swan) by just over 6% and having the Liberals incorrectly ahead in two (all four were lost).  The poll having a Voice finding with a Yes vote 10 points below any other WA sample is also a reason for caution here.  It has also been hard to find sufficient details of this polling's methods and even whether it is a voice robopoll or an SMS robopoll (a copy of an SMS robopoll with a compatible script from the same time has been seen, and it doesn't look like Morgan).

But obviously, the WA government is not nearly as popular now as at the 2021 election.  That could be true and it could still be very popular indeed.  In the circumstances a massive swing is likely to be hyped by the media as the beginning of the end for the Cook government, but the swing can be very large indeed and still be no such thing.  Or then again, it could be larger.

Labor operatives are doing the usual routine of saying that the seat will go down to the wire.  As noticed in past coverage Labor operatives are a bunch of Eeyores who like to use this sort of pessimism as a form of advance damage control and then be pleasantly surprised afterwards.  The media have shown that they will fall for it every time.  

Much has been made of a change to Labor's how-to-vote card which has overall made it more like a donkey vote (pun unintended) with a bit of spite towards the main independent opponent (who has also recommended putting Labor way down) by needlessly putting her below the Australian Christians.  Labor will not be excluded here barring a primary vote swing of over 50%; the only advantage of the changed card would be a lower risk of informal voting which might save them a few primary votes if the seat is close.  (Even this is less of a factor in WA than elsewhere because of WA's savings provision for votes with skipped numbers.)

There are three obvious systematic sources of swing against Labor here:

1. It is a government vacancy  

Government vacancies on average result in substantial 2PP swings to Oppositions.  The reasons for this include (i) the loss of the outgoing government MP's personal vote (ii) the fact that governments are often less popular between elections than they were at an election that they won (iii) the tendency of voters to use by-elections to let off steam and protest about local or passing issues.  After cutting numerous corners involving seats not thrown to preferences and seats where one major party ran but failed to make the final two, I estimate the average 2PP swing against WA Governments in their own seats in the past 50 years at about 5.9%.  (For the short and very cherrypicked list of Opposition seat by-elections in this time I get an average 2PP swing of 0.4% to government.)

That average is influenced by two outliers.  In the 1988 Balga by-election the departure of Premier Brian Burke saw a 23.1% swing against Labor (on its way to a scraped victory at the 1989 general election with a 6.5% 2PP swing against it).  Also in the 1991 Geraldton by-election Labor lost a seat they held so heavily that they failed to make the final two, with a 31.1% swing against on primary votes and probably something similar on 2PP (but no-one was counting.)

2. The loss of a massive personal vote

The following graph shows the lean of Rockingham to Labor compared with the state 2PP since the seat's creation.  It was held from 1974 to 1996 by Michael Barnett and since then by McGowan.

Various redistributions have affected the margin but not greatly (at least since 1980) so far as I can tell.  

Once the vacant seat had been won by Barnett it had an average lean with either Barnett or McGowan it had an average lean of +11.5% for elections up to 2008.  During the 2008-13 term McGowan became leader and since then the seat has averaged +18.8, suggesting that making McGowan leader has given his party a boost of around 7% in the seat.  There is a suggestion it could be even larger than that from the Labor primary running 14.7% higher in the lower house than the upper house in this district (compared to no difference statewide, having an incumbent would be worth a few percent on average by itself.)

(It would be interesting to compare Balcatta/Balga but it is impossible in view of Brian Burke holding the seat for virtually its whole existence).

3. Comedown from a landslide

Labor won the 2021 election with a statewide 2PP of 69.7%.  Whatever its current support in the electorate is, it is obviously not that.  

How to set expectations

Let's assume for a moment that this finishes as a 2PP classic Labor-Liberal contest, which it may well not.  In that case, it's important not to double-count the damage here.  The loss of McGowan's leadership boost is itself likely to be worth 7%, and the average swing in a government by-election is almost 6%, but that average is also skewed upwards by the departure of other heavy hitters like Brian Burke.  Likewise, some part of the swing against the government since 2021 is already factored in to the average swing for government vacancies, because governments would normally be a few percent below their election-winning form at the time a by-election was held.

The leader boost factor and the average by-election swing factor alone suggest at minimum 11% should go straight out the door, and that anything below 15% is great for the ALP.  Suppose the swing was double that (22%), even that I think would be consistent with a government that was still way ahead of its opposition.  (It is hard to say because normally the relationship between polling swings and by-election swings at least federally is shallower than 1% lost for every 1% in polling, but I think in this extreme case it might not be).  If the seat finishes as a 2PP contest, or if Labor wins and a 2PP is at least calculable, I would think the swing must be above that to even provide reasonable evidence that the Liberals have returned to the hunt.  

Even if the swing were, say, 28%, one could still make the case that this is 15% off the state swing and 13% off McGowan factors, general anti-government mood and candidate controversy (Magenta Marshall has a recent drink driving offence that was revealed two weeks before election day) and that it is still consistent with a modest state government lead (but one that would spark significant backbencher panic levels).  Allowing for the variability of by-elections maybe even 30% 2PP is defensible.  After all, suppose McGowan had decided to drive his Corolla back to the east coast in 2006, would Labor have expected to do better than 57-43 in a by-election then?  It seems unlikely. 

Indie Factor!

A lot of interest has been added here by the candidacy of Hayley Edwards, Deputy Mayor of Rockingham and a former Labor member who appears to have thrown in her membership after being overlooked in favour of Marshall, a staffer with a pretty thin non-political CV.  (Edwards has been a military medic, a paramedic and a fitness franchisee).  Edwards' municipal form isn't the big thing it would be in some other states (she topped her ward with 1487 votes and was then elected Deputy around the table) but seems to be prominent enough to potentially make the top two, in which case she could expect strong preference flows.  (A massive primary vote swing against Labor, something like 40%, would still be needed to make the seat loseable).  With the Liberals running a very young candidate whose name would be worth a lot more free football votes in Tasmania than here, Edwards is more likely to be the threat if the seat is actually at any risk of falling.  That is not to say she is necessarily more likely than the Liberals' Peter Hudson to make the top two, just that the preference flows to her will be better if she does.  

(The West Australian conducted an exit straw poll with a trivially small sample size but (i) their analysis of it assumed that all voters would follow HTV cards, which most don't and (ii) exit polls of prepoll are skewed against Labor because Labor voters are more likely to vote on the day.)

In the event that Labor wins and Edwards is second (or even that Edwards wins), the 2CP "swing" from 2021 is meaningless.  Swings should only compare changes involving the same pair of party opponents.  It is often the case, both in by-elections and general elections, that an independent is more competitive on 2CP than the other major party.