New England Washup and Bennelong By-Election Preview

I've decided to combine some post-result comments about New England with an overdue preview of the Bennelong by-election, which will be updated for any further polls that are published.  It's really not possible to talk about Bennelong now without talking about the former and whether anything seen in New England does or does not apply.

New England: Barnaby Bolts Back In

While Barnaby Joyce's re-election was always extremely likely given the lack of serious opposition, the scale of it surprised me.  Currently Joyce has 64.6% of the primary vote, a 12.3% swing to him, and a 73.6% two-party vote against Labor, a 7.2% swing to Joyce.  It seems that Labor are second, although in theory independent Rob Taber might overhaul the 4.4-point gap on the 17.2% of minor-candidate preferences.  Even if he does, his two-candidate result against Joyce won't be much better than Labor's.

My pre-election expectation was that Labor would not get much 2PP swing and that there would probably be a modest 2CP swing away from the result achieved by Tony Windsor in 2016, but that the gaggle of candidates running against Joyce might be able to take him to preferences and delay the result.  Instead Joyce has picked up a primary vote swing compared even to his 2013 result, when Tony Windsor wasn't on the ballot paper.



The main post hoc response to this result was to say that it is unsurprising given the nature of the campaign.  Once it was known that Tony Windsor was not running, and with Shooters, Fishers and Farmers and One Nation not interested either, it was generally accepted that the outcome was a foregone conclusion.  Joyce kept spending and promising money but other major players lost interest and devoted their efforts to Bennelong instead.  On this construction, the result in 2016 was abnormally close because Tony Windsor's presence in the contest created a perception that the seat was in play and resulted in major attacks against Joyce, suppressing both his primary and his two-party-preferred vote.

I am not convinced this is the whole story.  In 2013 there was also not a serious contest, but although the Coalition was riding high at that election Joyce did not do nearly so well.  Granted, at that stage he was not as prominent on the national stage as party leader.  I did question whether the idea of a sympathy vote for a member who is disqualified on a technicality would still hold up this long after the original election, but the result suggests at least a possibility that it has.

One further explanation we can scratch is that the government's acceptance of a banking Royal Commission spurred a swing to Joyce.  If this was the case there would be a much weaker swing in pre-poll voting, but this is not the case so far.

The result has been hailed as a record swing to an incumbent Government, but is not.  There are clear-cut cases of larger swings in Opposition-held seats - the Nationalists recorded a 12.6% swing in Labor-held West Sydney in 1921, but still lost, as did the Liberals in recording a 13.4% swing in Australian Capital Territory in 1970.  Also in the strange case of McPherson 1981 a 2PP swing of 16.2% in favour of the Fraser government is claimed (Labor finished third as the seat had become a three-cornered contest. An independent had directed preferences against the government in 1980).

Bennelong (Lib, 9.7%)

Assessment: Narrow Liberal retain favoured based on historic patterns and polling, but evidence needs to be treated with caution.

Bennelong was historically a safe Liberal seat but boundary and ethnic changes over time have made it less so, culminating in the 2007 defeat of Prime Minister John Howard by Labor's Maxine McKew on a swing that was basically the same as the national swing and similar to that in surrounding seats.

John Alexander, a formerly top-ten ranked tennis player and later commentator, recovered the seat in 2010 and has outperformed the national swing at all three elections he has contested, even recording a 2% swing in his favour in 2016 despite the national 3.1% swing to Labor.  On Alexander's watch, Bennelong has therefore gone from 1.3 points more favourable to the Coalition than the national average when McKew won to 9.3 points more favourable now.  However, a swing to the Coalition (worth about three points two-party preferred) was also seen in the 2016 Senate results in Bennelong, so Alexander appears to have benefited from the Turnbull government's appeal to inner-city voters in the 2016 election.  If that appeal has since worn off more than the government's standing elsewhere then it is likely Alexander's margin from the last election is a bit inflated as a starting point.

Once factoring in current polling, the expected swing for a typical by-election would be quite close to the 9.7% target.  See my findings on this pre-Canning last year, and also William Bowe's (which include state elections).  From my own findings, an 8.2% swing against the government would be expected for a generic by-election, while William has it even closer to the score required.  However, this is not a typical by-election.  Irrespective of the sympathy-vote angle that may have been present in the New England results, one difference that is certainly significant is that when an MP is disqualified and recontests, their personal vote advantage is not lost.  On that basis, Bennelong should not be expected to change hands.

Perhaps recognising this, Labor have preselected a very high-profile candidate, the former NSW Premier Kristina Keneally.  Keneally's profile might make her perform unusually well - which may be necessary to get near winning the seat - but the controversial nature of her Premiership also makes her a high-risk selection and easy target with the possibility of a bad result.

The election has seen two gaffes and one bout of unwanted publicity for the Alexander camp.  Firstly in an image of his team supposedly making phone calls to voters, the phones were not connected.  Secondly, video surfaced of him telling a dirty joke with rape and racial references at a private function in the 1990s.  Thirdly, two days out the SMH reported that Alexander had failed to declare rental income from a country estate.

The Keneally campaign has come under scrutiny with claims about Medicare office closures and waiting times being strongly challenged by opponents, and Keneally's claims about Catholic education being vigorously attacked by journalist Samantha Maiden.  Keneally has also faced many questions about the Obeid era and her campaign has, in general, been a fact-checker's paradise.

Another major campaign focus has been the electorate's large Chinese vote - 13% born in China, 21% with Chinese ancestry.  The by-election has been ethnically charged because of controversy about Senator Sam Dastyari's actions at the behest of a Chinese donor, which have forced him to announce his resignation, and also because of broader issues about foreign donations and influence in Australian politics.  However there have been claims of "China-phobia" against the Liberals as a result of the way they have pursued Dastyari and the disclosure .   I share the view that treating Chinese and for that matter other Asian voters as a homogeneous mass who are going to care desperately about the mud-slinging in this row is silly, but some will.  To the extent that Alexander has padded his margin by appealing to Chinese communities, he must be at some risk of losing some of that buffer.  There are also some Chinese groups that have been specifically active in the by-election, lobbying against the Coalition, and this makes things quite unpredictable.

A sleeper issue is Alexander's age and future intentions for the seat.  At 66 there's speculation he won't run again, although it is unconfirmed by the candidate.  Voters may be wondering about his commitment level if this is his final term.

There is reason for caution about whatever drove the New England result applying in Bennelong.  The New England result shouldn't be taken as a sign that the government is in good electoral health, as it is possible that the National Party is travelling OK and the Liberal Party is not.  Also, there is the obvious difference in campaign intensity to consider.  Nonetheless, even completely discounting anything we saw in New England, and completely ignoring sympathy-vote issues, historic evidence says the seat should probably be retained.  The question is whether the issues specific to this intense and rather singular campaign are enough to overturn what should be a buffer of a few points for the incumbent.

Among the other candidates, the greatest interest in the election is the comparative performance of the Australian Conservatives and Christian Democrats.  The Greens are also running but will not get much attention, and the remaining seven out of twelve are cluttering up the ballot paper and are not likely to get their deposits back.  Alexander has the benefit of the donkey vote over Keneally; however this is only worth a few tenths of a point these days.

Overall, my view of this campaign is that both sides have made impressive efforts to lose.  However, there can only be one loser.

Keneally's Past Ratings

Keneally is associated with the infamously bad Labor state government that was turfed with a massive swing against it at the 2011 New South Wales election.  There have been two myths about this - firstly that Keneally herself was not unpopular and secondly that Keneally's popularity was only dragged down by her own government.

The following were Keneally's Newspoll net satisfaction ratings as NSW Premier, in order: +15, +16, +10, -5, -12, -14, -27, -24, -28.  The last three readings came in early 2011 after Keneally's controversial decision to have Parliament prorogued, which was seen as avoiding scrutiny.

Other leaders whose doomed state governments were smashed have often, but not always, polled bad personal ratings in the leadup to that.  Some have polled somewhat worse, eg Anna Bligh at -43 and Lara Giddings at -37.  One notable exception: Carmen Lawrence was still +14 a few months prior to her party's thrashing in WA in 1993, and although there doesn't seem to be an election-eve Newspoll for her, she was fantastically popular as Opposition Leader after Labor's loss.  Another: Joan Kirner (Vic) had slipped as low as -31 but had recovered to only -8 by the time of her party's defeat in 1992. These exceptions show that being at the helm of a doomed state government does not necessarily translate to bad personal ratings.

Bennelong Polls

Two seat polls were published early in the campaign.  Seat polls in Australia are rather unreliable.

Galaxy found a 50-50 result with Alexander holding a 42-39 primary vote lead.  (That would not seem enough except that Bennelong has a very high Christian Democrat vote, 6% at the last election, and the CDP preferences are conservative).  ReachTEL had a respondent-allocated 53-47 to Alexander off raw primaries of LIB 41.6 ALP 34.5 GRN 5.9 ON 5.4 CDP 1.6 CON 1.4 Other 1.2 Undecided 8.3.  With the "undecided" splitting 33-27 to Alexander, that's 44.3-36.7 between the two leading candidates.  The religious-party scores look low and we now know there is actually no One Nation candidate.  (Thanks to the well informed reader who sent me the full primaries.)

The ReachTEL found favourable ratings for both Alexander (good 51.2 poor 15) and Keneally (good 41.6 poor 28.1).   A national YouGov found mildly favourable impressions of both Alexander (40-28) and Keneally (39-29) but this is of very little use.

In the second-last week of the campaign both sides agreed Labor was "behind" and Liberal internal polling was said to show about a 54-46 lead for Alexander.

On the last weekend, Newspoll doubled down on the earlier Galaxy result with a 50-50 2PP off primaries of Liberal and Labor 39, Green 9, Australian Conservatives 7, Christian Democrats 2 (others 4).  This close poll, contradicting the agreed narrative that the Liberals were ahead, is likely to reshape campaign coverage.  The result also highlighted a point of interest: the contest has big implications for the Christian Democrats, who might have to seriously explore merging with the Conservatives if they are thumped in this one.  The Conservatives result might be exaggerated by them being named in the readout, but if that is true it is also true for the Christian Democrats, and coming off over 6% last time they certainly don't want to go there.  Another point about this poll is that the sample size is small.

Overnight on the final Wednesday, a 53-47 Fairfax ReachTEL was reported.  So we get the same pattern we saw in the WA and Queensland state elections where ReachTEL is much better for the Coalition through the campaign (in both cases, ReachTEL converged at the end.)  However this time the reasons are probably different. Primaries were Liberal 40.4 Labor 35.7 Green 7 One Nation (not running and should have been removed) 2.6 Conservatives 6.2 Christian Democrats 2.3 other 3.4 undecided 2.4, so add about a point to each major for the undecideds.

A report on the Fairfax poll had 23.5% of voters saying they were more likely to vote Labor based on news of Sam Dastyari's Chinese connections and 28.4% less likely, but this polling format is a waste of time.  Polls in this format routinely generate 30% effect sizes even when the issues canvassed are trivial or even fake.  Even if voters respond honestly and accurately (which many won't) the poll doesn't say how much more likely.  This sort of poll creates a fake environment in which the voter is expected to consider their vote and a single issue in isolation - that's not how most voters decide their vote.

Finally on the day of the election we had a Galaxy with a 51-49 lead to Alexander off primaries of 40% Liberal, 38% Labor, 8% Green, 7% AC, 3% CDP, 4% others.  Again the sample size was only just over 500.

A special concern I have about polling for this electorate is the high proportion of voters with English as their non-primary language.  These voters may be difficult to poll and yet their votes could have a major impact.

I will have live comments on the Bennelong count on Dec 16.  See also the Poll Bludger guide.

Grading The Result

Sean Kelly has an article here on the difficulty of grading the Bennelong result, especially if the Coalition wins.  He suggests that the result is only genuinely dramatic if Alexander gets such a scare that the result is unclear on the night, or if Alexander wins with a margin similar to last time.

Really the difficulty with marking this one is that we don't know what the sympathy vote for the disqualified MP is worth as there are so few precedents and their circumstances differ.  Here's my scorecard for 2PP results:

Below 50 (Loss by any margin): Self-evident disaster
50-51.9: Poor
52-55.9: Inconclusive; can be spun any way you like
56-57.9: Good
58+: Excellent, and Labor has questions to answer.