Saturday, November 12, 2022

Victorian Lower House: Not All That Much Narrowing Here

POLLING AGGREGATE 55.8 TO ALP (not necessarily accurate)
Labor currently appears overwhelmingly likely to win, most likely in majority with low but realistic chance of hung parliament

A roundup of Victorian lower house polling themes is about a week overdue, but before I start this one, a big shout out especially to Ben Raue at The Tally Room for his coverage of this election.  While I've been appallingly busy (not to mention busily appalled) Ben has issued a remarkable total of twelve articles about the Victorian election since October 29, a rate of just about an article per day!

As of my last article on Victorian lower house voting intention back in late August, Daniel Andrews' Labor government was polling far better than a government of its age should be doing, now that it no longer has the Morrison Coalition government to rail against.  I expected that things would get a little closer and that by this time we'd probably no longer be seeing 56-44s, but thus far only Newspoll has been consistent with that expectation.  And even then, Newspoll was still pretty lopsided.  



Newspoll

The Newspoll taken 31 Oct-3 Nov found Labor leading 54-46 (Coalition 37 Labor 37 Greens 13 others 13).  This was the first public poll to show Labor any less than 55-45 ahead since mid-2021.  It was enough to bring out a few Hung Parliament Club types in media commentary, but in fact on such numbers Labor would almost certainly win a majority.  My 2PP model has Labor winning the 2PP in a median 52 seats (not counting the five seats with recontesting crossbenchers), and while one might expect the Greens to take one or two of those and independents a few, a lot would have to go wrong for Labor not to get 45 wins.  A key point here is that if the 2PP is only 54-46, then the Coalition is probably winning the 2PP in its marginals and also the Hawthorns and Bayswaters of this election, so the possibility of Labor missing out to an in those doesn't knock much of a dent in Labor's expected seat tally anyway.  

Of the seats Labor would be otherwise favoured to win, it's mainly the western triad of Melton, Werribee and Point Cook where indies are a known threat, but how much of a threat is not that clear.  Ian Birchall was actually very close to winning Melton last time (despite finishing third), but that was a vacancy.  For the other two, there's a perception the government has been crossed off a lot of Christmas card lists, but not much public data basis.  

The Newspoll had Andrews on a solid net satisfaction rating of +7 and Matthew Guy on an unimpressive net -20.  However, continuing the trend of Andrews underwhelming on Better Premier (perhaps because those who don't like Andrews tend to hate him), Andrews' lead over Guy is a modest 51-32.

Other Polls

An interesting entry was the Victorian debut of Freshwater Strategy, a relatively new firm involving experienced ex-C|T Group pollster Michael Turner, who was house pollster for the Morrison Government at the last two elections.  This online poll taken 3-6 Nov had a 56-44 Labor lead off primaries of Labor 37 Coalition 34 Greens 14 and others 15.  Freshwater Strategy is not at this stage a member of the Australian Polling Council and the details about the poll available online are limited (eg data are "weighted" but it doesn't say by what) - their site does offer more detailed reporting by email but having to email to get details of polling is inefficient and pollsters should release such data openly.  On net favourability (which is a different thing to net satisfaction, because voters may dislike a politician yet still think they are performing well) the gap between Andrews (net -9) and Guy (net -16) was surprisingly small, but Freshwater found a slightly higher preferred Premier lead than Newspoll (50-28).  Whether using the word "preferred" or "better" in such questions impacts the results at all would be interesting to test.

Morgan has recently released an SMS poll taken Nov 9-10 with a supposed 57-43 lead to Labor, but it is not stated on what basis the 2PP was calculated and my last-election numbers have it as 59.5-40.5 to the ALP.  I distrust SMS polling as it has a poor track record and is likely to be affected by motivated response.  The primary votes are Labor 40 Coalition 29 Greens 11.5 and Others 19.5 (broken down into Teal Independent 4.5 DHJP 1 UAP 0.5 other parties and other independents 13.5).  It since turns out that UAP are not contesting any lower house seats and DHJP will not much trouble the scorers in the ten they are running in.  I'm assuming "Teal independent" was on the readout everywhere and is hence overstated (since there only seem to be about nine of them, though several of those will poll significantly).  Morgan also back-released a multi-mode result of 60.5-39.5 to Labor but with no further details; this is from sampling conducted through October.  In the SMS sample Andrews had a netsat of +17 (58.5-41.5 on forced choice, but I suspect non-responders who responded to voting intention aren't removed from the latter) and a thumping forced-choice lead as better Premier (65.5-34.5). 

It's a while ago now but the last entry from Resolve Strategic (Oct 20-24) had Labor up 59-41 off primary votes of Labor 38 Coalition 31 Greens 12 Independent (presumably on readout everywhere) 12 others 6.  My estimate off the primary votes was 58-42.  Resolve has been producing some very strong results for Labor at federal level and both in NSW and Victoria recently, something that was not apparent prior to the federal election.  Resolve had Andrews ahead 49-29 on the skewed preferred Premier metric.

I also mention that while Redbridge has not yet released any statewide polling, Redbridge director Kos Samaras has rejected the polls that have Labor in the very high 50s and has suggested that Newspoll is closer to the mark with the majors perhaps around the mid-30s.  

A simple time-weighted aggregate of these polls (using the 5/3/2/1 formula by week that I've used in past federal aggregation) gives a 2PP estimate virtually unchanged from 2018 at 57.4% to Labor.  With adjustments for quality and to downweight 2PP-only polls and polls with small sample sizes or a lack of track record this comes down to 56.2, which would see hardly any government seat total change though some narrowing from here to polling day should be expected.  One might apply further adjustments for house effects but that is messy, especially after the 2018 Victorian election where every poll underestimated Labor (all bar one ReachTEL by a lot).  

The outlook

If these polls are collectively anywhere near accurate then the prospects for the Coalition are grim.  A degree of narrowing should be expected, but not necessarily very much given how close to election day we are.  Some combination of narrowing and polling error might yet drag the 2PP into the low 50s range where a hung parliament becomes a serious possibility, but it's far-fetched for the 2PP to make it all the way back to 50-50, and even if it gets that far the Coalition still probably lose, care of an unfriendly pendulum and risks of dropping seats where they win the 2PP to teals and other independents.  

Australian Election Forecasts has a Victorian model up which has the Andrews' government almost certain to win in some shape, with about a 15% chance of anything other than a Labor majority.  

There have been some suggestions that recent Victorian and NSW polling shows that federal drag is dead.  As discussed in the previous article, a simplified federal drag model that ignores federal government polling would predict that this government probably loses.  But while the honeymoon bounce for the federal government has come down a little and reaction to its Budget was mostly pessimistic, it still has large leads (55-45 in the most recent Newspoll.)  Plugging that into my historic federal drag model with a federal polling adjustment, the projection is 48 Labor seats (consistent with a 2PP of, say, 52.7 after considering the crossbench risk issues).  It might be argued that Labor is at more risk than the average government in the model based on competition from Greens and teals, but on the other hand Labor also gained seats in the redistribution, which the projection ignores.  

The Newspoll at least is consistent enough with what federal drag theory expects at this stage of the federal government's career, and the federal drag model has a pretty wide margin of error anyway.  If, on the other hand, Morgan is correct, then we could be headed for the very unusual sight of a second term federally-dragged state government getting a 2PP swing to it.  Unusual but not unprecedented; the 2021 Tasmanian election saw a similar result and also involved a shambolic opposition.  But that was a result that appeared influenced by COVID management.

Polling error in any election is always possible, and the 2018 Victorian election was one of the more serious cases (only the side the polls said was winning easily won very easily, so nobody cared).  The thing about this, however, is that the current polls are not clustered.  A three or four point error on the average would put the result somewhere in the range 52-60.  At the moment if Newspoll was four points too high for Labor then other polls would be ten points too high, and that level of error would be more or less unprecedented in Australia.

What might be the independent vote?

Independents polled just over 6% in 2018.  They contested 53 divisions with five seats attracting four or more (Melton had six).  At this election independents are running in 66 divisions (three-quarters of the total).  However their increased interest is more than matched by micro-parties, which are running a combined 350 candidates (nearly twice as many as last time); the micro-party inflation is headlined by the Lyle Shelton/Peter Bain "Family First Victoria" party running in every division.  The Animal Justice Party running in ever seat instead of just half the seats is also notable here.

Micro-parties polled 5.17% in 2018, the AJP's 1.82% being by far the biggest contributor.  If the micro-party vote is unchanged then that would put Newspoll on a statewide independent vote of about 8%, Freshwater at 9% and Morgan at about 14.5%, alongside Resolve's 12%.  Resolve and Morgan would have independent on the readout everywhere.  With only a quarter of seats uncontested by independents that might not make a big difference, but in many of these cases "contested" will be a loose concept.

Assuming there is an independent vote of, say, 9%, what will that do?  Because Independent support is so concentrated in particular divisions it is often hard to predict how many seats Independents will win.  But there is a general relationship in single-seat elections between the overall Independent vote and seat wins.  Here is a graph showing that relationship for all compulsory preferencing state and federal elections from 2004 onwards/  I've picked 2004 as a start point because of the rise of the Greens at that election, which has changed both the preferencing environment and the options for independents to pick up loose anti-major-party votes.


The red dots are Victoria, which has underperformed in this period so far when it comes to vote share to seat conversion for independents.  The most striking thing on this graph is that the ten-independent result at the 2022 federal election is actually not an outlier - based on the total independent vote share, it was only a little bit better than average!  

The surge in micro-party candidates could be a dampener on the independent vote, but it's hard to say how much.  Supposing that independents do in fact get 8%, then the linear relationship above puts them on for about eight seats on average.  Even if the past poor Victorian performance of independents in seat terms is assumed to be predictive, at least five independent seats would still seem likely.  

Greens

The Greens are polling rather well at present, following on from their strong performance in the federal election and with a thus far much less troubled Victorian campaign than in 2018.  There was a tendency of pollsters to overestimate the party in the past (notably in Victoria 2014 where all seventeen published polls did so) but this seems to be less of a thing nowadays, perhaps because live phone polling is pretty much extinct.  The Liberals will recommend preferences to the Greens.  The Liberal vote was hollowed out in Northcote to a little over 10% in 2018, and wouldn't have been much higher in Richmond had the party contested it, but even a 3-4% boost in 2CP from a changed Liberal recommendation could be huge for the Greens in Richmond and Northcote.  In Albert Park the Greens' biggest challenge is still to get over one of the majors in the first place.  

As I write someone from Liberal HQ has been quoted as saying that "As a result of this decision, Labor will lose up to three seats to the Greens, boosting our election chances", but at best any such boost is insignificant.  The Greens are bound to back Labor in any minority situation because their voters who heavily preference Labor will desert them in droves if they do not.  There will be Greens elected no matter what (at least if the Greens are winning any of Northcote, Richmond or Albert Park then they must be winning Melbourne too) and if Labor needs both Greens and Independents then the breakdown of the Labor-Green seats between Labor and the Greens is irrelevant.  Just maybe (i) the Labor vs Green seats could be the difference between Independents holding the balance of power alone and sharing the balance with the Greens or (ii) some independent might genuinely base a decision on which major party has the most seats, but at least as likely as (ii) would be unicorns on the ballot paper.  A more coherent argument would have been that the decision would force Labor to work harder to hold its seats, but at least Northcote and Richmond were hard work anyway.

The interesting question is whether this opens up a really serious chance for the Greens anywhere else where they might get into second. I looked at all of Williamstown, Footscray, Essendon, Preston and Pascoe Vale and in most cases the Greens seemed to need 10-15% 2CP swings from Labor to win above what they got out of the Liberal decision.  

Pascoe Vale seems closer however. The independent campaign of Oscar Yildiz last time messes up the baseline (though Yildiz' votes seem to have been mostly major party votes); I estimate a non-Yildiz baseline primary of something like Labor 43.2 Liberal 18 Green 21.9 after accounting for the Greens-friendly redistribution, and a 2CP of about 57-43 to Labor after adjusting for the Liberal preferencing decision.  

The decision also helps the Greens in Prahran because it means that if they are over either major party after preferences they win.

I will be posting live coverage on the night of the election and, as usual, what I hope to be extensive postcount coverage thereafter.  I may add some more items to this one later.  

Update 4 Nov: Redbridge

A Redbridge poll for the Herald-Sun has been released.  The raw primaries are ALP 33.9 L-NP 33.6 Green 12.3 IND 7 others 2.9 with a 2PP of 53.5 to Labor.  The report gives primaries with undecided redistributed as 38-38-14-8-3 but also gives a breakdown of the leanings of the undecided voters.  Following the (standard) method of redistributing leaning voters to the party they are leaning to, I instead get the primaries as ALP 36.7 L-NP 35.5 Green 13.2 IND 8.5 others 6.0.  In either case, I get the last-election 2PP as 55-45 not 53.5-46.5 and at this stage I am not sure where the latter's coming from.  While this poll is being used to justify more hung parliament speculation, I estimate the seat distribution if it is accurate at 50-26-5-7 (5 Green 7 IND). 

The poll moves my aggregate down from 56.2 to 55.8 (not a prediction).  

The poll has an issues question of the sort the Herald-Sun always loves - voters are asked if they agree there is a health crisis in Victoria and then asked which party would address the health crisis better, whether they would support more hospitals, and whether funds should be diverted from planned rail upgrades to health.  Asking voters if they want more hospitals is a pony poll (they always will) but the priming of respondents with the idea that there's a health crisis potentially distorts all the other health questions and I suspect results would be very different without this leading question.

1 comment:

  1. My prediction.. 2018 and 2019 votes on 2pp basis are similar with maybe a slight swing to Labor. LIbs cannot win.. seats like Carrum and Frankston will have 10% plus margins... real fight is over seats on the range of 2 % either way.
    These seats will favour Labor. Esp those in country areas where demographic change is favourable t9 Labor. Libs will lose massively maybe keep 4 to 5 seats. Another 13 seats green indep and nats.
    So Labor will have maybe another 10 seats or more... more nats then liberals possible

    ReplyDelete

The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.