Saturday, August 27, 2022

Victorian Lower House 2022: Labor Well On Track Despite Federal Drag

Yesterday saw the release of the first Newspoll for Victorian state voting intention this year.  This is the first new public polling voting intention data for the state for over four months with the exception of a couple of Morgan SMS offerings, on which I can't place a lot of weight.  The release of a Newspoll with about three months til the election is a good time to launch my coverage of the Victorian lower house election, having not covered the Victorian lower house contest since the last Newspoll late last year.

The new Newspoll estimates Labor leading 56-44 (an effective 1.6% swing from the last election after accounting for the Liberals not contesting Richmond in 2018) off primary votes of Labor 41 Coalition 36 Greens 13 others 10.  These results are not wildly different to the 2018 election, with Labor down 1.9%, the Greens up 2.3% and the Coalition up negligibly.  For all the hype from some on the right about how hated he supposedly is, Daniel Andrews polls a rather strong +13 net rating (54-41) with the artist now known as Matt Guy on a not so flashy -17 (32-49).  A modest Better Premier lead for Andrews given the other scores and the way preferred leader scores skew to incumbents (51-34) does provide some support for the idea that Andrews is a polarising figure and the minority who dislike him tend to do so strongly.

Seat Model

What might this poll look like if realised on election day?  By uniform swing Labor would drop four of its 58 notional seats to the Liberals and Northcote (just) to the Greens, coming out just two down on its 2018 election result, but the uniform swing model is if anything potentially unkind to Labor's chances vs the Liberals in the 2PP seats.  There are two major reasons for this: firstly, personal vote effects from the 2018 massacre will make it hard for the Liberals to pick off Labor marginals, and secondly the Liberals have a lot of ultramarginals that are at risk from random variation unless they can get a more substantial swing.  They have nine seats on 1.5% or less (including occupied notional Labor seats) to Labor's four.  

I have run my usual conditional probability seat model to try to project how many 2PP wins there should be for each party assuming that swings in given seats are determined only by effectively random variation and personal vote effects.  For these purposes I have set the standard distribution of seat swings at 4%, the values of retirements and sophomore effects at 1% each, and the value of the "surge" effect (recontesting sitting member defeated opposing recontesting sitting member at the previous election) at 0.5%.  I have looked at the composition of those remotely close new seats that have been produced in the redistribution to see if I should make special adjustments for those, but so far the only adjustment I have made has been for Pakenham.  Pakenham consists of parts of Bass and Gembrook which both had sitting Liberal members in 2018.  However the Liberal member for Bass was defeated and the Liberal member for Gembrook is contesting Berwick, so despite being Pakenham being a vacant notionally Labor seat I have coded it with a 1 point bonus to Labor.  I have also adjusted Yan Yean by 5% to Liberals because the candidate was disendorsed in 2018.  As usual I adjust each side's projected result in each of its own seats so that its net gain or loss from personal vote effects gets cancelled out.  Five seats occupied by recontesting crossbenchers who were elected as such are excluded from the model.

There is a lot of trendy stuff to the effect that such 2PP models aren't useful anymore but at the federal election they worked fine in forecasting Labor's seat tally (not so well for the Coalition but most of the suspense was around whether Labor could get a majority or not given the polling anyway).   A similar point applies to Victoria: it would not surprise to see both sides drop seats to new crossbenchers, but the number should be modest in the context of the election overall, and in Labor's case any crossbench losses in seats where they win the 2PP will probably be a modest number of Greens and the odd local uprising.  Also, some Green gains may be possible to project off polling.   

Here's a sample partial output for a 2PP of 56-44.  The columns are in order: current margin, projected margin if the poll 2PP is correct, projected probability of that side winning the 2PP in the seat if the state 2PP is correct:

Any of the individual probabilities may turn out to be nonsense, for instance if there are great regional imbalances in swing (which could well be the case with the outer western/northern Melbourne dynamic seen in the federal election.)  But usually what goes up in one set of seats goes down somewhere else: if Labor is being crushed in outskirt seats and losing a few rogue seats on big margins, it may also be surging in the inner cities.  

Overall in this model for a 2PP of 56-44 I estimate Labor wins the 2PP in 56.3 seats vs 26.7 for the Coalition.  I would give the Greens at least one new seat on balance off the primary votes and there are the usual unknowns with teal and other independents and the Greens' ability to intensify their vote where it matters, but there is no doubt that a lopsided 2PP to Labor would be another clear Labor majority.   As a sign of just how hard the pendulum and personal vote landscape is for the Liberals, I have Labor with a 50% chance of a majority if it gets 51% 2PP and drops one new seat to the Greens, while I have the Coalition needing to get 52.2% for a 50% chance of a majority if it drops nothing to the crossbench at all.  All else being equal the swing the Coalition needs to form any kind of government here is likely to be something like 9%, or more.  A daunting task.  

But how plausible is it that Labor could have such a big win on election day?

This Election And Federal Drag

Federal drag (the impact on state election results of who is in power federally) is one of the main themes of this site's coverage of state elections.  All else being equal, it's a huge disadvantage to be the same party as the federal government, whether because voters doubt that a party of the same stripe will stand up to the feds properly or whether because they just like using state elections to send the feds a message.  

After adding in data since my last article on this issue, the regression continues to suggest an eight year old state government that is federally dragged should be cruising for a whacking:

Expected seat share change = -.0384-.0156*Age+.1624*Different +/- 0.104

(where different = 1 if the parties in power at the two levels are differnent, zero otherwise)

This suggests that on average the Andrews Government would expect a loss of about .163 of the parliament, or 14 seats - enough for about a two in three chance of it going into minority if not out of office entirely, a model that wildly disagrees with, well, everything else out there.  But it is not that simple, because the fate of same party governments is also affected by how popular the federal government is, and at the moment the federal government has a major polling lead.  

That's something we've not seen for a while if it lasts til election day.  The last same-party governments to go to the voters with the federal government clearly ahead in the polls were those of South Australia and Tasmania in March 2010 (both lost seats but survived).  Since then eleven same-party governments have gone to the voters with the federal government behind, with seven of these losing office and three of the other four losing seats while the fourth broke even.  

For same-party governments there is a relationship between federal drag and the strength of the federal government's polling:

Expected seat share change = -.025 - .0162*Age + .0151*Polling margin +/- 0.104

(where "Polling margin" is expressed as government 2PP - 50, ie 54-46 = 4).

At present the federal government might be ahead, say, 57-43.  So on that basis the expected seat share loss drops to .049 of the parliament (c. 4 seats) but the formula is also blind to redistributions and pendulums that favour one side much more than they did at the previous election.  On this basis, it is not that unusual (without even considering special state factors) that Labor gets a 56-44 state poll at this time.   However if the federal honeymoon is boosting Labor's polling at state level then the latter may recede as the former does - for this reason I'd be fairly surprised to still be seeing 56-44s in November.  Federal drag is just one prior - one needs to also weight in polling and other factors - but if the federal government continues to poll well that makes it much harder for the Coalition to win in Victoria.

Is it possible to lose from here?

The unfriendly nature of the pendulum/personal vote landscapes means that the Coalition is very unlikely to be able to win by, say, bringing Labor down to 50.5 and getting lucky with the seat distribution.  The swing back from the current Newspoll to victory would be something like 8% in three months.  

There is some precedent for Victorian governments losing large leads.  In the July-August 1999 Newspoll, Jeff Kennett had a 50-40 primary vote lead (about 54-46 two party preferred) but in mid-September his government was dumped by a whisker by Labor led by Steve Bracks.  In July-August 2010 Labor under John Brumby led 55-45 but in late November Labor lost, again very narrowly, to the Coalition under Ted Baillieu.  Both these governments were also federally dragged, and in neither case was the federal government all that unpopular.  On the other hand, these were the days of the older bouncier version of Newspoll, and Baillieu and Bracks were far better regarded Opposition Leaders than Guy, for what that's worth.  

Crossbench contenders

There has been a lot of hype about potential new crossbenchers this election following the teal and Green successes federally, but also in the expectation that something might rise up on the right to win seats where the Liberal Party cannot.  The hopes on the right seem so far to be more wishful thinking than anything tangible.  The Victorians Party which was the subject of all kinds of overpuffed nonsense by the Herald-Sun has ended up not even running at all, and while the argument has now shifted to so-called "purple independents", it seems hard to find high-profile contenders who fit the bill and are decared candidates.  The declared independents so far are generally teals or obscure.  

Candidates being referred to as teals or connected to the teal wave are so far contesting the vacant Labor seat of Albert Park and the Liberal seats of Caulfield, Mornington, Kew, Sandringham and South-West Coast.  (Brighton is also expected but does not appear clearly confirmed - UPDATE confirmed on 31 August).  Many of these campaigns lack the funding, polish and experience of the bigger federal teal campaigns, but this may also be true of their opposition.  Nonetheless, cases like Nomi Kaltmann in Caulfield running as a teal independent months after contesting Labor officebearer positions seem to be pushing the envelope.  I'm interested in reports of other INDs (teal or otherwise) who show signs of being competitive.

Albert Park is also a Greens target seat, and hence with a good supply of preference "stepladders" if an independent does get on a roll there.  The most serious Greens chances are obviously Northcote and Richmond, with anything else needing a large swing.  Richmond (ALP 5.8% vs Green) is less safe than it looks on paper because of the retirement of Richard Wynne and also the fact that Kathleen Maltzahn (of controversial gender and sex worker views) is no longer the Green candidate.  

There are also rural non-incumbent indies, the most significant of which probably being Jacqui Hawkins in Benambra, which Bill Tilley retained against her 52.45-47.55 in 2018.

Most of the existing crossbenchers are in theory on tenterhooks, but whether there is actual risk we'll have to see.  Firstly Ali Cupper in Mildura has had her seat redistributed to be notionally National.  Sophomore effect is alone enough to cancel out the Nationals' notional lead, and at the federal election there was no appetite for throwing out recontesting independents, but it will be interesting to see how Cupper goes after a somewhat odd first term including briefly forming a coalition with Reason's Fiona Patten.  

In Prahran, the Greens' Sam Hibbins has won from third twice in a row but now faces two new threats. The first is the redistribution, which may have pushed him into a notional third after preferences though that includes Labor votes from now-vacant Albert Park so may not be reliable. The second is the risk that the Liberals fall to third and their preferences elect Labor.   The Greens' other two seats of Melbourne and Brunswick are also on tight margins though at this stage it doesn't look like Labor is building vote share in a way that would endanger them.  

This concludes my first piece for this year on a Victorian lower house contest that I'm quite looking forward to.  I am not so looking forward to the upper house, as the Victorian parliament has disgracefully failed to fix the Group Ticket Voting system used for said body despite fully one quarter of MLCs being not elected on merit in 2018.  All I can say is that everyone, whoever they vote for, should vote below the line in the upper house.  The more boxes a voter can number the better, but even a 1-5 and stop is better than letting parties send preferences goodness knows where.

1 comment:

  1. I suspect that in terms of Commonwealth drag, given the newness and popularity of the Commonwealth Government, this election will be more like 1996 than 1999, 2010 or 2014.

    If the swing distributions are like the Commonwealth Election, the swings will be concentrated in mostly safe ALP seats in the western and northern suburbs and the more marginal seat including southeastern suburbs, so Coalition gains could be concentrated in the in the southeastern suburbs. Werribee could be a gain for Garra, if he runs in Werribee again (although he seems to be in Werribee South, which was redistributed into the new Point Cook that the Liberals have made noises about targeting).

    If the Coalition does regain Shepparton, the controversial school merger (which was in the Shepparton part of the seat, which was Sheed`s stronger area) is likely to have been a major cause thereof.

    Melton might be winnable for the Liberals or an independent, or it may be that the ALP vote was artificially deflated last time by the high informal vote (10%), due to the large number of candidates.

    The ALP versus Green margin in Footscray last time was likely inflated by the Greens` candidate issues there and the subsequent redistribution has been favourable to the Greens. The Greens also did well in the overlapping areas in the Commonwealth election. I expect the ALP versus Greens margin (which the VEC did not actually count, as far as I know) to shrink significantly.