NSW 2023 Lower House Preview: Is Dom Doomed?

It's an almost annual tradition on this site to release something every Christmas Day. Click the Xmas tag for previous examples. As there is considerable interest in the NSW election I've decided, as in 2019, to go with the NSW leadup.  It helps that there has been a larger volume of lead-up polling in the last few months of 2022 than in 2018. Hopefully this continues during February and the early to mid campaign.

State and territory conservative governments haven't had a great run of it in elections in the last several years.  Of those that were either in power when Tony Abbott's Coalition won the 2013 federal election or came to power not long after, four (Queensland, Victoria, NT and South Australia) were kicked out after a single term and one (Western Australia) lasted two terms before being drubbed (with the remains of it obliterated four years later).  We are left with the almost twelve year old Perrottet-led Coalition in NSW, on to its fourth Premier, and the almost nine year old Rockliff Liberal Government in Tasmania, on to its third.  Both benefited at their most recent re-elections from weak Labor opposition, but the current NSW Opposition Leader Chris Minns seems to be a sharper customer so far than his predecessors.  

Meanwhile while Dominic Perrottet has so far sought to lead with surprising energy, enthusiasm and a sense of the problems that his government faces, the NSW Liberal Party continues to be blighted by preselection-related and factional infighting.  Candidate chaos is not necessarily an obstacle to victory (the Morrison government won in 2019 despite having to drop or disendorse 11 candidates along the way) but headlines such as the Premier's recent attempt to improve gender equity in the upper house being overridden are not helpful.  

In 2019 what was then Gladys Berejiklian's second-term Coalition government seemed in danger of at least losing its majority but clung to a three-seat majority in an election where hardly any seats changed hands.  The Coalition won 48 seats, Labor 36, and Greens, independents and the Shooters, Fishers and Farmers three each.  

A lot has changed since then.  All the Shooters became independents, as did Liberals John Sidoti (ICAC investigation, now suspended from parliament) and Gareth Ward (criminal charges, also suspended).  Liberal Andrew Constance left Bega prior to a very narrowly failed attempt to win the federal seat of Gilmore, and Labor's Michael Holland snapped it up at a by-election.  Labor's Tania Mihailuk became another independent after unproven corruption allegations she made under privilege against preselected upper house candidate Khal Asfour were not appreciated by the party.  Mihailuk herself is facing accusations of bullying, which she denies, and is also involved in a two-in-one-doesn't-go situation with MP for Lakemba Jihad Dib, who will contest the redrawn Bankstown.  The starting line for the election in terms of currently occupied or notional new seats is Coalition 45 seats Labor 36 Greens 3 and independents 9 (two currently suspended) but Sidoti is retiring at the election and Ward will either not contest or presumably be heavily defeated.  (NB since writing this article I've had feedback suggesting Ward can be competitive despite the charges as he has been a very popular local MP in the past and the Liberals have no candidate as yet.)

Despite ending the term narrowly in minority, Dominic Perrottet's government has not been at any actual risk of falling during the current term.

General state election history and this election

I have found on this site that the two biggest killers of state governments are age and federal drag.  State governments are boosted if they are of the opposite party to the federal government of the day. This may be partly because voters like having different parties in charge at different levels (thinking same-party governments will be too compliant with Canberra's wishes) but it also seems that voters use state elections to give federal governments of the same party a kicking unless those federal governments are very popular, and that more and more voters do this the longer the state government persists.

Fortunately for Perrottet's regime, the Morrison federal government was booted in May this year, so federal drag no longer affects it.  However time is not its friend; governments that have been around this long are likely to lose seats compared to the previous election (eg SA 2014) even when they are not federally dragged.  

I take the modern condition of Australian state politics as starting from 1989, the end of the last of the ageless state governments of the second half of the 20th century.  Updated for Victoria my current simple starting regression for the fate of state governments (ignoring polling) is:

government seat share change = -0.042-0.0136*age+0.153*different +/- .104

where "different" is 1 if the federal and state governments are different parties and 0 if they're the same.  The Perrottet Government is so old that the regression predicts it to probably lose seat share even with federal drag on its side, with a median projection of -.052 of the parliament (5 seats) and a 50% chance of a result between a 15 seat loss and a 5 seat gain.  

It also may not help that the Albanese federal government continues to cruise in the polls. Historically there's been no evidence that the popularity of an opposite-party federal government matters, but it would be hardly surprising if it did.  (The popularity of same-party federal governments matters a lot, as seen in Victoria, so I use a more complex formula in those cases).  

What are the targets for victory?

As usual I have established a 2PP conditional probability model that factors in seat-specific effects such as personal votes.  I expect I'll post some outputs of this during the campaign. I treat Heathcote as a Liberal seat with a margin of -1.7% rather than a Labor seat, and I also treat Bega as a Liberal seat based on the previous election, but apply a hefty penalty (currently 6% though I am going to fine-tune this later) because it is a "disrupted seat".  Disrupted seats won by other parties at mid-term by-elections tend to produce results somewhere between the last election result and the by-election result.  

The 2PP model includes Kiama, Bankstown and Drummoyne as major party seats but does not include the nine seats won by the crossbench at the last election.  On the assumption that there are no gains or losses from those nine seats (discussed further below) I get the following:

* On average the Coalition needs 51.8% 2PP (-0.2%) for a better than even chance of a majority.  Almost any swing to Labor, all else being equal, means the Coalition is likely to lose its majority.  This results from personal vote effects (including leadership change) for Labor in close seats, but the Coalition having some very close, disrupted (Bega) or adversely redistributed seats that lack any personal vote protection.

* On average the Coalition needs 49% 2PP (-3.0%) for a median result with more seats than Labor, while Labor needs the Coalition to fall to 48.2% (-3.8% and 51.8% to Labor) for a median result where it has more seats.

* On average the Coalition needs to fall to 45.6% 2PP (-6.4%, 54.4% to Labor) before Labor has a more than even chance of a majority. This results from a shortage of marginal Coalition seats.

What is noticeable here is that the range of swings projected to produce a hung parliament is remarkably wide - any swing to Labor that isn't very large appears a good chance to do it.  But hung parliaments have a track record of not happening lately (fifteen state and federal majorities in a row, with twelve of the campaigns featuring various levels of media speculation about a hung parliament being likely.)  So that still needs to be treated with caution.  

These estimates assume swings vary by personal vote and by effectively random factors, but have not included any possible systematic regional variation or one side outperforming the other in marginal seats compared to what would be expected for their 2PP.  In NSW 2019, Victoria 2022 (a remarkable case) and the 2019 federal election, the incumbent governments did especially well at sandbagging their own marginal seats, while in the 2022 federal election the Labor opposition did this.   If time permits, sometime before the state election I will run a model that assumes that the sort of realignment seen in Victoria (with Labor doing well in inner city seats and the Coalition holding up better in outer suburbs) also applies here and see what difference that makes.  

Crossbench issues

While the above model assumes the crossbench will stay the same (excluding mid-term defections) it's far from certain this will be so.  Two of the ex-Shooters independents (Helen Dalton in Murray and Roy Butler in the vast seat of Barwon) are on relatively small margins and the Nationals will be hoping the turmoil in the Shooters, if nothing else, helps them to recover both those seats.  

The most obvious downside risk for the Coalition is losses to teal independents.  Although teals wiped out in the Victorian election, NSW provides two ingredients important to their federal success: seats where Labor is completely uncompetitive, and a Liberal government to protest against. However, optional preferencing makes it harder for independents to win from well behind compared to federally.  A good illustration of this was the Willoughby by-election, at which independent Larissa Penn fell well short after preferences off an only slightly larger primary vote gap to that which Kylea Tink easily overturned in the federal contest for North Sydney.  Teal campaigns will also be subject to NSW disclosure and spending requirements.

Identified tealoid candidates include Joeline Hackman (Manly), Jacqui Scruby (Pittwater), Karen Freyer (Vaucluse), Victoria Davidson (Lane Cove), and Helen Conway (North Shore).  Wakehurst may also attract a tral candidate and Simon Holmes à Court
 has said there is no limit to the number of teals Climate 200 might support.  (This list will be edited to add more as announced.)

NSW also has a long tradition of rural independent success.  Dubbo was a close 2CP vs independent last time (2.0%) and there's been a recent curve-ball with the defection of federal Calare MHR Andrew Gee to the crossbench.  I am not sure the latter will actually hurt the Nationals.  Another seat that was vaguely close was Wollondilly (5.5%, 6.0% post-redistribution according to Antony Green).  It, like Dubbo, was a vacancy in 2019, but it is also interesting in view of a high One Nation vote and a recent trend of One Nation voters preferring indies over major parties.

There is a trend at recent elections for the media to hype pretty much any independent who is running without requiring any empirical evidence that the candidate has any prospects of success.  This was especially apparent in Victoria, where no fewer than five much vaunted independents failed to make double figures. 

As concerns the Greens, one seat of interest is Lismore.  This was won by former federal MP Janelle Saffin in 2019 but Saffin only beat the Greens on 3CP by 361 votes.  According to Antony Green the primary votes shift in the Greens' favour compared to Labor by 1.0% - this is actually enough after preferences to make Lismore notionally National, as the Greens had worse preference flows.  Saffin as a first-term incumbent would be expected to get an incumbency boost, but Saffin already had a massive profile in the region and given the swings seen in the federal contest for Richmond there is a case for Labor to be nervous.  The other reasonably close seat is Ballina where on 2019 figures a 3.9% 3CP Greens to Labor shift would have seen Labor make the final two; I suspect that such a swing is unlikely.  There is some interest in Balmain as the first ever case where a Green incumbent in a single-seat state or federal division has retired; the 10% buffer suggests the Greens should not have trouble holding it but the margin will be worth a look.

A challenge at this election is predicting what various crossbenchers might do if there was a hung parliament.  I would expect the Greens to support a Labor government irrespective of seat totals.  Independents Alex Greenwich and Greg Piper seem more left than right, but might in theory support (or at least not vote out) a Coalition government especially if the alternative was unworkable and the Coalition was near a majority.  Joe McGirr is socially conservative but his positions on other issues suggest he isn't a lock to support the Coalition.  The ex-Shooters independents would be safer electorally backing the Coalition but might also be motivated by which side could deliver more on issues of concern.  It's also important to bear in mind that crossbenchers don't have to make a choice either way; in theory a minority government can persist without guarantees of supply and confidence if there is not the will to vote it out.

Recent polling

There have been rather a lot of polls in the last four months.  In brief summary:

* Multi-mode Morgan samples in September, October and November with 2PPs of 53, 57 and 52 to Labor respectively (major party primaries 34-34, 36.5-32 to Labor and 37-35 to Coalition).

* Newspoll in September with 54-46 to Labor (major party primaries 40-35 to Labor)

* Freshwater Strategy in October with 54-46 to Labor (major party primaries 37-36 to Labor; this 2PP seems a little generous to me, my last-election estimate for the published primaries is 52.8)

* Resolve Strategic polls in September and October. The first gave Labor a 43-30 primary vote lead and I estimated this as a 60-40 2PP (an obvious outlier) and the second had a 38-35 primary vote lead, which I estimated at 54.5 2PP for Labor.

* Incomplete details of a small Essential sample in September with the Coalition ahead 36.4-32 on what I assume to have been raw primary vote; a very rubbery 2PP estimate off the incomplete numbers reported is 50.5 to Coalition.

A simple average of all the above is about 54-46 to Labor, which if repeated would only give Labor a close to even chance of a majority on my numbers.  There was also, unfortunately, a mid-November report of a Nov 8-10 unnamed industry group poll by an unnamed pollster using unstated methods, which the Australian claimed to have been "leaked" (meaning that the sponsor gave the poll to them for free).  This had the same primary vote lead for Labor (40-35) as the September Newspoll.  This level of opaque poll reporting by The Australian is completely unsatisfactory and I see no reason why such behaviour should be allowed.

Both "teal independents" and One Nation are scoring highly in polls that ask about them. The Greens are currently running at a slight increase on 2019. However, it is not clear One Nation are running in any significant number of lower house seats as yet, and teals will only run in a small minority. There may be lots of spare votes around here, but polls are also prone to underestimate the votes that obscure micro-parties get between them when a lot of them run.

Overall, for the time being the Coalition on aggregate is in a losing or at best struggling position and it will probably need substantial improvement (perhaps on the order of 3-4% 2PP) over the next few months if the government is to survive in any form. There are, however, plenty of precedents for such recoveries. The Berejiklian government was behind about 49-51 in late 2018 and finished up winning 52-48. The Palaszczuk Labor government in Queensland was down about 48.5-51.5 in mid-2020 and won 53.2-46.8 in October. (Both these races had a lot less polling at the same stage.)

The only seat-specific polling I have seen was some proof-of-concept style Redbridge polling in possible teal target seats where respondents were primed with statements about federal teal independents then asked how they would vote if a similar candidate ran in their state division.  This was taken in late September and early October.  These polls were flawed in that the preferencing question did not allow the respondent the option of exhausting their preferences.  The track record from the federal election was that even the "prompted" version of these polls tended to have the Coalition primary too low, but did not necessarily overestimate the teal independent.  The numbers in the poll suggest that the right teal campaign might win Manly, North Shore, Pittwater and would be at least competitive in Wakehurst and Oxley, with Lane Cove the only seat where support appeared too low.  

Leadership polling continues to show pretty good results for both leaders - for instance Perrottet was net +6 and Minns net +15 in Newspoll; they were net +2 and net +11 in Freshwater.  They are more or less swapping the lead as Better Premier, which is not greatly surprising given that that indicator skews to incumbents but the incumbent's party is so far behind.  Not much should be read into this as there is plenty of history of Premiers who take over near the end of a government's life polling well.

That's all for this preview unless I think of anything else to add. Secular season's greetings and Happy New Year to all readers.