Sunday, August 4, 2019

2019 House of Reps Figures Finalised

Donations welcome!

If you find my coverage useful please consider donating to support the large amount of time I spend working on this site.  Donations can be made by the Paypal button in the sidebar or email me via the address in my profile for my account details.  Please only donate if you are sure you can afford to do so.

The 2019 House of Representatives results have been finalised, a joyous event that tends to arrive unheralded two to three months after every federal election.  Although all the preference throws had been completed and uploaded some time ago, the final figures importantly include the two-party preference flows by party.  Normally I say that this is very useful for assessing the performance of polls.  At this election the polls failed dismally, mainly because of failures on the Coalition and Labor primaries (except for Ipsos which failed on the Greens primary instead of Labor); nonetheless there will be a final review of them here fairly soon.  This article is a general roundup of other matters regarding the House of Reps figures.

Preference Shifting

The final 2PP result is 51.53% to the Coalition and 48.47% to Labor, a 1.16% swing to the Coalition.

There was a very large shift in the preferences of Pauline Hanson's One Nation.  One Nation preferences flowed only 50.47% to Coalition in 2016 but 65.22% to Coalition in 2019 (even more than the 60-40 split believed to have been assumed by Newspoll after considering state election results).  Overall, preferences from parties other than the Greens and One Nation also flowed more strongly to the Coalition by a few points (53.93% compared to 50.79%) but this was caused by the United Australia Party flowing 65.14% to the Coalition.  Excluding the Greens, One Nation and UAP, Others preferences (50.7% to ALP) were 1.5 points stronger for Labor than in 2016.  It is also interesting that Katters Australian Party preferences flowed 14 points more strongly to the Coalition, very similar to the shift for One Nation.

Greens preferences split 82.21% to Labor, up just 0.27 points.  This seems hardly impressive given that 2019 was supposedly a "climate election" (it seems it was so only for Greens and some inner-city voters).  There was clear daylight between the major parties on climate policy despite Green frustrations with Labor ambivalence on Adani, and the current PM wasn't even making the previous PM's futile sympathetic noises on the issue, so one might expect an increase in the flow to at least the 83% of 2013.  However, a closer look by state shows that the flow did increase in six states and territories, most conspicuously weakening by 4.5 points in South Australia where more moderate voters moved to the Greens with the collapse of the NXT/Centre Alliance vote.   The Greens-to-Labor flow also weakened slightly in Queensland.

Overall, the shifts in preferences didn't have an enormous impact on the 2PP, but it does depend on exactly how one measures it.  By the coarsest Coalition/Labor/Greens/all others method, the preference shift helped the Coalition by 0.82 points.  This is the third largest shift since 1980, behind the 1990 and 2013 shifts to Labor.  However, it's still not enormous, and even with no preference shifting the Coalition would have clearly won the election, albeit without Bass and Longman on a seat-by-seat preference flow basis.

Even on a basis normalised for the recent increase in the minor party vote, this is the Coalition's biggest improvement in preference flow since 1974.  Its overall preference share of 40.4% (ignoring three-cornered contest issues) was its best since 2001, when the Greens were only polling 5%.

The loss from the Coalition on three-cornered contests was 0.10%, slightly lower than the last two elections.  However even this was inflated by around 0.02 points by the strange situation in Lyons, where the Liberal candidate was disendorsed with her name still on the ballot paper as a Liberal, resulting in many Liberal voters redirecting their first preference to the Nationals candidate and some of those not preferencing the disendorsed Liberal at all.

So here are some last-election formulae for 2019:

2PP = L-NP+.178*Green+.651*UAP+.652*PHON+.493*Other -0.10

or if removing UAP (as pollsters are bound to do)

2PP = L-NP+.178*Green+.652*PHON+.539*Other-0.10

or putting both in Others:

2PP = L-NP+.178*Green+.562*Other-0.10

or breaking out One Nation and Independent:

2PP = L-NP+.178*Green+.652*PHON+.406*IND+.593*Other-0.10

or breaking out One Nation, UAP and Independent:

2PP = L-NP+.178*Green+.651*UAP+.652*PHON+.406*IND+.553*Other-0.10

After splitting to Labor in some seats in 2016, One Nation preferences split to the Coalition in every seat the party contested, ranging from 54% to the disendorsed Liberal candidate in Lyons up to 74% in Brisbane.  United Australia Party contested every seat and its preferences split to the Coalition in all but four - McMahon (61.1% to Chris Bowen), Watson (58.8% to Tony Burke), Gorton (53.8% to Brendan O'Connor) and Fowler (50.6% to Chris Hayes).  The strongest flow to the Coalition was 79% in Berowra.

The weakest 2CP flow to Labor involving the Greens was 63.7% in Parkes.

Swing Distribution

Ignoring crossbench contests, the net 2PP swing, if uniform, would have resulted in the Coalition losing only Dunkley and gaining Herbert, Cowan, Longman and Lindsay.  Compared to this, the Coalition won Braddon but lost in Corangamite, Cowan and Gilmore.

For the second election in a row, personal vote effects were swamped by demographics.  The 2016 election saw a number of Coalition sophomores in outer suburban seats lose, and this year it was Labor's turn for sophomore MPs to lose (Braddon, Bass, Longman, Herbert) or suffer above-average swings (Burt, Dobell, Hindmarsh, Macquarie, Paterson, Solomon).  Only in Barton, Cowan, Lyons and Macarthur did Labor MPs who unseated the Coalition in 2016 get small swings, and without the Jessica Whelan candidate malfunction in Lyons Labor would have been very close to losing that one as well.

It may seem that the Coalition underperformed the swing (which is not surprising anyway given the huge swings to it in safe Queensland seats) but this is partly a reflection of Scott Morrison's "goat track to victory".  Apparently believing it was better placed than the national polls said - but not that much better placed - the Coalition set out to focus on defensible marginals and a very small number of possible Labor gains.  Labor-allied groups tried to convince the Coalition that it was losing seats all over the pendulum (including seats like Higgins and Kooyong) on the grounds that if you think the fires are everywhere, you don't know which ones to put out.  Labor also cast its net wide in the belief that it had a highway to victory.  The Coalition campaigned as if the fires did not exist (either because they knew that or because they saw no other hope), and it turned out that most of them didn't.

They Coalition did stunningly well at doing that.  Based on an assumption of a swing that was uniformly distributed but randomly variable, I'd estimated that Labor would need 49.8% 2PP to win more seats than the Coalition.  In fact, based on applying a uniform swing to the election results, they would have needed 51.6% 2PP to win more seats.  Had that been the 2PP, a swing to Labor of two points would have netted them only three seats on above that margin (and a meaningless 2PP win in Warringah) while missing out on nine below it.  Had we known just how well the Coalition was doing in its own marginals (which the average swing figures from seat polling provided no reason for knowing) we would have seen even the incorrect national polling we had as pointing to a merely line-ball contest, not a probable narrow ALP win.

This can be seen starkly on the swing graph below, with best fit lines for Coalition-2PP and Labor-2PP seats.  The colour shows the winning party.  There was only one Labor gain that started as a Coalition-2PP seat (two Labor gains were notionally Labor to begin with).

Labor-2PP seats saw an average swing of around 1.35% to Coalition (slightly above the overall swing) with no differentiation between marginals and safe seats.  But in Coalition-2PP seats the trend was for the swing to Coalition to be larger the more marginal the seat was.  Labor achieved large swings in some Coalition seats, but they were wasted on seats on unassailable margins.  Not only did only one Coalition-2PP seat fall, but very few Coalition seats went anywhere near falling.  This is why the Coalition could still have, remarkably, hung on in minority even without the national polls being wrong at all.  It is rare to see a case where one side is so successful in making the swing connect where it matters and the other side fails so abjectly - and based on the personal vote effects left over from the 2016 election, if one side was going to do this, it should have been Labor!

There is some comparison to New South Wales, where Labor's marginal seat campaigning also failed to deliver the goods. In both cases the Coalition did better in swing terms in its marginals than in its safe seats, and thereby minimised seat losses.  Federally Labor failed to sandbag its own marginals when the swing was such that it needed to, while in NSW Labor sandbagged its marginals brilliantly when it didn't really need to.

Non-classic seats

There were fifteen (down two) seats where the final pairing wasn't Coalition vs Labor at this election:

ALP vs Green (3): Grayndler, Cooper, Wills
Coalition vs Green (2): Kooyong, Melbourne
Coalition vs IND (6): Indi, Cowper, Farrer, New England, Wentworth, Warringah
Labor vs IND (1): Clark
Coalition vs Centre Alliance (1): Mayo
Coalition vs KAP (1): Kennedy
Coalition vs One Nation (1): Maranoa

All of these except Kooyong and Wentworth were also non-classics last time, though Warringah was Coalition vs Green.  Melbourne repeated its 2016 distinction of uniquely having the 2PP winner not finish in the top two.

As explained in 2013 it is interesting to derive Labor-vs-Coalition 2PP splits for those voters who preferred the "non-classic" candidate to the majors, as this reveals what sort of voters a potential crossbencher might be beholden to.  Sometimes this can be done exactly for the voters who put the non-classic candidate first, rather than just for those who put them above the majors.  Here's a table showing 2PP preference flows from the non-classic contender to Labor in the non-classic seats:

(Notes: ???? - No split calculable for Indi 2016 because Labor finished fourth; @sorceror43 estimated 72%.  * - different non-classic contender in 2016 (McGowan in Indi, Windsor New England, Greens in Warringah, # - previous Wentworth figure is from 2018 by-election not 2016)

(NB Farrer was mistakenly omitted from the table.  For Farrer the 3CP flow to Labor from Kevin Mack (IND) was 45.9%.  77.4% of Mack's 3CP votes were his own.  The 2016 comparison is not relevant as the Greens finished 3rd in 2016 when the seat was a classic contest.)

This table shows that the preferences of voters for the main non-classic contender became more polarised in some seats, such as Kennedy and Maranoa on the Coalition side, but also Cooper (largely because of no David Feeney), Cowper, Kooyong, Mayo, Melbourne (despite the Labor candidate getting disendorsed), Wentworth (cf the by-election) and Wills all on the Labor side.

There were no Liberal vs National 2CPs this election, but there was very nearly one in Mallee, where the Liberals were beaten to second by Labor by 338 votes, after Labor had got into the top three by 248 votes over an independent, who himself had got into the top four over another independent by 65 votes.   It's very likely the Nationals' Anne Webster would have beaten the Liberals' Serge Petrovich anyway, as she had been outperforming him on the exclusions up to the final one, and Petrovich would have needed over 72% of preferences coming from Labor (nearly half of which were not 1 Labor votes).

Strongest preference flows

The following are the strongest preference flows I could find evidence of, whether those preferences were distributed or not:

1. 93.2% Tim Hollo (Greens) to Alicia Payne (Labor), Canberra
2. 92.6 David Risstrom (Greens) to Ged Kearney (Labor), Cooper
3. 90.7 Adam Pulford (Greens) to Peter Khalil (Labor), Wills
4. 90.3 Jim Casey (Greens) to Anthony Albanese (Labor), Grayndler
5. 89.5 Serge Petrovich (Liberal) to Anne Webster (National), Mallee
6. 89.2 Robert Holian (Greens) to Lisa Chesters (Labor), Bendigo
7. 89.1 Kingsley Liu (Greens) to Susan Templeman (Labor), Macquarie
8. 89.0 Jamie van Burgel (Australian Christians) to Andrew Hastie (Liberal), Canning
9. 88.7 Gary Whisson (Greens) to Brian Mitchell (Labor), Lyons
10. 88.6 Matthew Thompson (Greens) to Tanya Plibersek (Labor), Sydney

Cases shown in bold are estimated figures based on the 3CP results where no 2PP flow for the candidate exists.  They are likely to be slight underestimates.  Probably, the Cooper flow was actually even stronger than the Canberra one, and I suspect these are the two strongest Greens to Labor flows of all time.

Unexpectedly, the strongest preference flow to a losing candidate was 87.6% from Matt Donnelly (Liberal Democrats) to Ed Cocks (Liberal), Bean.  And no, Donnelly didn't get the donkey vote.

Winning from behind and on minor party preferences

Twelve seats (down four from 2016) were won by the candidate who finished second on primaries.  Labor won ten of these on Greens preferences: Macquarie, Lilley, Cowan, Corangamite, Moreton, Griffith, Dunkley, Richmond, Perth and Macnamara.  Crossbenchers won two, Indi and Mayo.

In Macquarie, Labor needed 87.0% of Green preferences to win (already above the national average) but managed to beat that, getting 89.1%.  However, Macquarie has a history of strong Greens to Labor flow, and in 2016 the Greens to Labor flow there had been the second highest in the country at 90.3%.

Aside from Macquarie, Labor did not need an above-average share of Greens preferences to win any of the other nine listed, but beat the average in all of them except Cowan anyway.  Lilley (needed 79.5, got 84.1) and Cowan (needed 71.4, got 79.7) were two besides Macquarie that could conceivably have been lost had the Greens preferenced the Coalition on their how-to-vote cards.  In all of Corangamite, Moreton and Griffith Labor got a share of Greens preferences around 11-12 points higher than needed (probably beyond the reach of how-to-vote card impact) and in Richmond, Perth, Dunkley and Macnamara the flow would have had to be more than 20 points weaker for Labor to lose.  So the Greens' decision to preference Labor may have saved as many as three Labor seats at this election.

Macquarie was so close that it was also won on Animal Justice Party preferences.  Labor needed 59.9% of these to win the seat and got 65.1%.  The national average from this party was 61.8%.

One seat (Bass) was won by the primary vote leader but critically depending on United Australia Party preferences. The split of UAP preferences to Liberal in Bass was 61.2%, and had it been below 52.8%, Labor would have retained the seat.  The flow of Greens preferences to Labor also weakened slightly compared to 2016 in Bass, but this only contributed to the loss rather than causing it.  In Chisholm, the Liberals needed 38% of UAP preferences (a share exceeded in every seat in Australia) and got 74%.  A slightly stronger flow of UAP preferences in Macquarie would have sunk Labor there - the LNP needed 71.2% but got 66.4% (still above the national average).

For all the noise about One Nation preferences, the LNP won the Queensland seats it won so massively that it didn't need One Nation's help.  There was not a single seat the Coalition won where it would have been behind with an even split from One Nation.  As a result, the shift in One Nation preferences blew out the two-party preferred vote by 0.45 points, but didn't change the outcome in a single seat.   In Longman, the LNP needed 41.3% of One Nation preferences and got 66.1% (compared to just 43.5% in 2016).  In Blair, Labor could have plausibly lost to One Nation preferences but didn't; they needed 29.8% and got 37.0%.

One seat (Chisholm) saw the Coalition rely on a combination of right-wing preferences to win after leading on primary votes.  Had all of the preferences of the UAP, Rise Up Australia and Ian Dobby (IND) split 50-50, Labor would have won the seat.  Dobby's platform was not unambiguously right-wing and the 61% to Liberal split on his preferences was mostly because of the donkey vote.  No other seat had this attribute though Longman came very close.

In Mayo, Rebekha Sharkie would have won on either Labor or Green preferences even if the other party had split 50-50.  This wasn't quite the case for Helen Haines in Indi, but her victory was largely down to Labor preferences (of which she needed 70% and got 81.6%).

Addendum: The Australian notes three seats where Labor did not trail on primaries but still needed Greens preferences to win after falling behind during the preference distribution.  In Eden-Monaro there was a significant Nationals vote, meaning that Labor's primary vote lead was artificial.  Labor ended up needing 77.3% of Greens preferences and getting 87.0%, so that's another one where Greens how-to-vote cards could have swung the seat.  In Dobell and Solomon Labor dropped behind on adverse right-wing minor party preferences but needed only low shares of Green preferences (61.3% and 53.9% respectively, in both cases exceeded by more than 20 points) to get home.  The Australian's method omits two further seats, Blair and Gilmore, where Labor never trailed during the count but would have lost with a 50-50 split of Greens preferences.  So in total there are 15 seats for which this is true.

Other sections may be added to this piece if I notice anything else worth adding, or if there are any interesting requests that are practical to answer here.

No comments:

Post a Comment

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.