The Draft Boundaries Would Not Put Labor On The Edge Of Minority

The current round of draft redistributions is complete with the release of the NSW proposal today, following the Victorian and WA proposals two weeks ago.  While the Victorian redistribution led to an outbreak of unsound psephology with false claims that the Kooyong redistribution greatly favoured the Liberal Party (I wrote about this for Crikey), the NSW washup has been pretty sensible, for the first day at least.  One thing I have seen that seems hard to credit is the idea that Kylea Tink, whose seat is proposed to be abolished, would win the now even more marginal Bennelong off two major parties fighting tooth and nail for it.  This is a general article about the impact of the draft changes.  A note that I am not a primary source for redistribution margin estimates, and am here largely relying on the work of Ben Raue, William Bowe and Antony Green for those.

The Victorian draft proposes that part of the boundary of Kooyong expands to take in part of Higgins.  The key issue in the shortlived Frydenberg-comeback debacle was that there's no obvious way to project how an independent would have done if their seat is expanded into an area they didn't previously run in.  One can use the 2022 preference flows from the present Kooyong to distribute votes for Labor and the Greens et al between the Liberals and Monique Ryan (IND) as if Ryan had been running in the new bits, but that means assigning Ryan a primary vote of zero in the new part.  It's saying that voters who would vote 1 Ryan 2 Liberal, for instance, don't exist in the new bit, but we know they do exist in the old bit, or she would not have won the seat.

Ben Raue came up with a method for modelling how an independent might go in an expanded division, which is to assume that differences in the 2PP between the new and old parts of a seat would flow on to the 2CP.  I think this should work well for an urban seat where the proportion of the seat that is new is modest, and on the proviso that the independent still makes the 2CP (which Ryan would easily do).  It now seems to be accepted that while one can make arguments about whether Monique Ryan's position has really improved, the case that the redistribution seriously favours the Liberals in Kooyong was wrong.

Of other matters of suspense in Victoria, the draft redistribution notionally flipped one seat, Menzies, but by small margins (The Tally Room has it as Labor 0.4%, Poll Bludger has 0.7 Antony Green has 0.4.)  This shouldn't really matter as Keith Wolahan is a first-term sitting member and all things equal should get a personal vote, so Labor would only be expecting to take the seat if there was a swing to them.  The draft changes to the micro-close 3CP exclusion point in Macnamara appeared very minor (and any assessment of changes in this seat are rubbery anyway as voters in it vote all over the place geographically), while a more serious change to the Labor vs Green margin in Wills has attracted some attention.  That's not to say Wills would fall on account of Labor's margin coming down to 4.5% or so - just that it would keep Labor on their toes.  Labor's 2PP margin in Chisholm would be significantly cut.  

The WA draft redistribution creates a new seat called Bullwinkel on a margin of about 3.2%.  One thing that has been missed in discussion of this seat is that 70.8% of its voters had a sitting Liberal MP at the 2022 election (most of these were in Hasluck), while most of the rest were Swan voters with no sitting MP.  In theory, there is a loss of personal vote for the Liberals across most of Bullwinkel, and while the seat along with Tangney looks like low-hanging fruit for the expected WA correction, it's not quite the gimme on paper some people seem to think.  The big news for Labor in WA is their margin in Hasluck going into double digits; nothing else including the tightening of Canning is likely to have much significance.  

For NSW I had eyes on all the competitive non-classic seats, not just the teal ones, but absolutely nothing happened in the draft in Richmond and very little in Fowler or Cowper.  Of the teal seats, Wentworth not only becomes safer for Allegra Spender with expansions into red territory west and south but comes down to 51.2% Liberal vs Labor 2PP, closer than it has ever been at an election.  Something like this expansion of Wentworth was a forced move and the draft margins reveal that it is going to be very difficult for the Liberals to defeat Spender in the seat, especially as it is also a "double sophomore" seat where she defeated the previous Liberal incumbent at the last election.  It is so difficult that I'm no longer doing hypotheticals of the form "if the Liberals recover all the teal seats they lost last time".

North Sydney is abolished in the draft and distributed between Bradfield, Bennelong and Warringah.  Because independent Nicolette Boele made the 2PP in Bradfield last time, it is possible to make a notional combined Liberal vs IND 2CP for the new Bradfield based on Kylea Tink's 2CP in North Sydney and Boele's in Bradfield.  This sees Paul Fletcher's notional margin cut from 4.2% to 2.5% (Raue).  However, there's the question of who is the candidate when Boele has already been campaigning for some time, and whether one or the other will agree to stand aside or run somewhere else.  Tink gets a personal vote in the North Sydney part, and it's common for independents to get outsize boosts at the end of their first term.  But the North Sydney part is a minority of the seat.  In theory Boele and Tink could both run, but the challenges there would include generating a sufficient preference flow between the two for one to win without looking like a party, and also convincing potential strategic voters that they now needed to vote for two independents - it would be a weird contest.  

The news has led to speculation Matt Kean could seek to have preselection re-opened to challenge Paul Fletcher on the grounds that he would supposedly be a better candidate to beat Kylea Tink.  Looking at the 2022 results I actually see little difference between Boele's performance and Tink's beyond Bradfield being a harder seat to win.  The primary vote swings against the Liberals were similar in both seats.  Tink got a greater share of preferences (75.4% to 73.0%) than Boele did, and this was mostly caused by Labor preferences, but could have been just because the Labor candidate in Bradfield was male.  (A note that while the gap between Tink's 2CP and Labor's 2PP looks much higher than the gap in Bradfield, this is partly because of an apparent uncorrected transposition in the 2PP figures for the Willoughby PPVC booth.)

I hinted above that I don't think anything of the idea that Tink will win Bennelong.  Parts of the new Bennelong are  tealy enough but the overall primary votes on Ben's estimates start at Labor 32.1 Liberal 40.7 Green 10.3 Tink 8.2.  Of course Tink would take some votes from the majors but any gain that doesn't come from building a personal vote has to come from the 2022 part of the seat.  Overall the 2022 seat didn't show any great signs of teal-compatability (for starters it might be too multicultural) - including having a below-average Green vote by national standards.  On a 2PP basis the seat is being estimated as very marginally Liberal, presaging a massive two-party fight - how is an independent supposed to get either major party's primary low enough to make it into the final two and not lose the 2CP contest in the process?   Teals trying to win marginal seats didn't work in the Victorian election and I don't see how it works here.  

On the 2PP front, I mentioned that Bennelong has notionally flipped (again think little of it by itself since Jerome Laxale should get a personal vote bonus).  The Liberal margin in Hughes is halved (though they get a new personal vote there in turn), and Labor hasn't got that much to crow about (the Wentworth 2PP boost being most likely wasted).

What would this all mean?

There was a widespread perception going into the redistribution process that the draft boundaries would make it harder for Labor to win a majority.  If nothing else, the reduction from 151 seats to 150 seats would mean that Labor had one fewer seat to get a majority out of (a majority is 76 seats in either case).  But on balance I find the draft has done Labor's prospects of majority no harm whatsoever and may have even helped them, primarily by getting rid of North Sydney which they were probably not winning anyway.  

Thinking of seats in terms of who occupies them (or would notionally hold them if new) the net impact of the drafts would be that the crossbench loses one seat (North Sydney), Labor loses one seat (Higgins) and Labor gains a seat (Bullwinkel), and goes into the election as the holder or notional holder of 78 (including Aston picked up at a by-election).  If including notional changes, there's a swap of Menzies for Bennelong, if one agrees that Bennelong is notionally Liberal (it's so close that estimates may vary).

Because Bennelong moves to notional Liberal status while Menzies is only very narrowly marginal Labor it appears on notional swing that Labor moves very close to loss of majority (using the 0.4% estimate for Menzies) in the draft boundaries, at least if Aston is still treated as a Liberal seat.  But the appearance based on uniform swing is deceptive.

My method of modelling targets for minority or majority is to use a conditional swing based probability model of the current major party seats that allows for the fact that swings are never uniform, and that also takes in personal vote factors.  I have run a quick version of this with the following settings: +1% for first-term incumbent, an extra 0.5% for "double sophomore", -1% for a known retirement, deselection or defection, 2% for a change in whether the incumbent is the party leader, and -0.5% for a mid-term change of member in a by-election without a party change.  

For the Aston disruption I've used my usual method for a "disrupted" seat that changed hands at a by-election (which is to average the previous election and the by-election 2PPs) and then adjusted that for the redistribution.   (I should caution that there's very limited data for disrupted seats involving cases where governments win seats from oppositions rather than the other way round - my previous analysis was of the reverse - but the few state level cases I've checked suggest that when this occurs the seat often proves "sticky" for the new incumbent.  Benalla 2002 was one counter-example.)  An alternative for Aston would be to treat it as a standard personal vote gain/loss situation; the difference between these two treatments is fairly small (about 1.6%).

I find after all this that Labor's position as concerns retaining a majority looks better than the uniform swing model suggests, for these reasons:

1. Labor has gained expected personal vote bonuses (assuming its new sitting members are any good) in nine seats at or below 8% including four "double sophomores", their only likely personal vote negative to date being the mid-term change of member in Dunkley.  On the other hand in Coalition seats at or below 8% there are six personal vote bonuses but so far five retirements or deselections.  If there is little or no overall swing then these bonuses help Labor considerably.

2. There is an asymmetry in margins at certain scales.  For instance the Coalition has eight seats on less than 2% while Labor has four (plus by my hybrid method Aston).  Because swings are variable this places the Coalition at more risk of losing seats if there is not much net swing.

3. The disruption of Aston via Labor winning it mid-term means it's just not reliable to treat it as still at its 2022 margin. 

All these things are also true (2 to a slightly lesser extent) of the current boundaries, but I have not bothered modelling those since they certainly won't be the boundaries for the next election unless there is a very early election call with mini-redistributions.  

All this means that supposing the draft is adopted and there is no 2PP swing at all, my model then projects Labor to on average win the 2PP in 79 of the 135 seats in the model to the Coalition's 56, meaning Labor gains an extra seat on top of its position including Aston.  This model could of course be undone by regional patterns or by one side outperforming the other in terms of swings in the marginals, but it's quite strong evidence that Labor is in better shape than the pendulum suggests.  

My model assumes that there will be no net change in crossbench seats, and it's always possible that Labor could, for instance, drop a few seats to the Greens and not recover Fowler.  But if there is no change involving the crossbench, then Labor can on average drop as low as 51.07% 2PP before their chance of a majority falls below 50% - that's more or less exactly where I have them now.  There is a widespread view that the next election will inevitably be a hung parliament and while that is an obviously very possible outcome, a lot of the talk about it at the moment is way overconfident.  Polls at this stage are not very predictive of the eventual 2PP at an election, at least not if it's next year.

Here are some other rough targets I have estimated:

* For a plurality of seats with no changes to the crossbench (beyond the abolition of North Sydney and the Coalition recovering Monash and Calare from defectors) the Coalition needs on average 51.3% 2PP

* For a majority of seats with no crossbench changes as defined above, the Coalition needs 53.4%

* If the crossbench changes only via the Coalition recovering every marginal seat the crossbench holds against it (that's 4 teal and 2 Green seats) the Coalition needs 50.6% 2PP for a plurality and 51.9% for a majority

* If Labor wins every crossbench seat that it lost marginally based on 2CP, 3CP or both (that's Ryan, Brisbane. Griffith and Fowler) and the Coalition recovers nothing then Labor needs 50.0% for a majority and 48.2% for a plurality.

* In the absence of any other changes involving the crossbench, Labor can lose all of Macnamara, Richmond and Wills to the Greens and still have a just over 50% chance of majority if there is zero 2PP swing.

The draft is only a draft and sometimes major changes are made before the process finishes.  Also I am still assuming here that the process will finish and an election won't be called before mid-October.  But I think it is useful to consider the draft in preference to the current margins in thinking about what current 2PP and primary vote polling levels might actually mean.  

I should also note that there's a further redistribution coming that might get done in time for the election if it is in 2025, depending on when.  That's the Northern Territory where Solomon must take voters from Lingiari, most likely in Palmerston.  That's likely to make Lingiari less close and Solomon closer, and this will benefit Labor's prospects slightly in my model if it happens in time.  

Of course, my model makes no allowance for geographic factors.  If there is the expected large correction in Western Australia then Labor will lose Tangney and Bullwinkel even if the national swing is not that high.  But what goes up in one place often comes down somewhere else; for a given 2PP a larger swing back in WA would mean the swing elsewhere is smaller, boosting Labor's chances of retaining the marginals it occupies elsewhere or picking up some Liberal marginals.