Wentworth By-Election: Prospects And Polls

WENTWORTH (Lib vs ALP, 17.8%)
Cause of vacancy: Resignation of Malcolm Turnbull

A by-election for Wentworth has been set down for Saturday 20 October following the resignation from parliament of the former Prime Minister, Malcolm Turnbull.  This post will give an assessment of how this by-election stands in history and will also comment critically on polling. As I start it on 14 Sep I won't be brave enough to predict the winner yet, but I may do so later [edit: as of the final week I am not liking the Liberals' chances of holding off Phelps at all].  This post will be updated with new polls and changed assessments through to polling day.

By-elections often produce big swings against governments, but surely 17.8% is safe?  After all, there has only been one larger swing in federal by-election history (20.1% in Wakefield, 1938).  In fact there are at least two reasons not to take the seat's safety for granted.  These are the historic circumstances of the by-election and the prospect of losing the seat to an independent.  While I've seen some spinning suggesting that a double-digit swing should be viewed as a shocker for the Liberal Party even if they don't lose the seat, I think the circumstances are so unusual that they'd take a win by any margin now.



Wentworth

Wentworth (see Poll Bludger preview) is a Federation seat and as such stands alongside Kooyong with a nearly spotless record of Liberal victory.  It has been won by the endorsed (or an endorsed) Liberal-or-precursor candidate every time but one, the exception being 1929.  Incumbent Walter Marks was one of the rebels who voted with Billy Hughes to bring down the Bruce government.  Marks was not endorsed for his troubles, but beat the endorsed candidate anyway.  The seat was so safe for the conservative side that in 1931 both candidates were endorsed by the United Australia Party, and Marks was defeated by the other UAP candidate. No current federal seat created before 1949 has a completely perfect record of being always won by the endorsed Liberal-or-precursor candidate, though Barker (1903) has always been won by a conservative party of some kind.

Wentworth hasn't always been super-safe.  It almost fell in the 1943 Curtin landslide, and in 1993 then Opposition Leader John Hewson had an unconvincing 55.5% 2PP result.  In 2007 Turnbull (who had wrested the seat from Peter King prior to the 2004 election) was held to 53.9% 2PP.  Overall it was inside 60-40 at every general election between 1984 and 2007.  Margins blew out from 2010 with Turnbull's status as former Opposition Leader and then Prime Minister.

Standard By-Election Swings

As noted in my Canning preview the average 2PP by-election swing against a government in a vacated government seat is 6.3% (with a lot of variation), and it doesn't matter whether the vacancy is caused by a death or a resignation.  This figure ignores the recent Section 44 by-elections where the incumbent recontested.

However, swings are larger when the government is polling poorly on a national basis, and on current polling a swing of 8-9% would be par for the course, all else excepted.  The government's polling may moderate between now and the by-election still five weeks away, so that may change by a point or two.  However it changes, even after considering the usual pattern of by-election swings, the seat would still be rather safe.

Turnbull's Personal Vote

A big question in this by-election is how much personal vote does Malcolm Turnbull take with him.

Personal votes are normally assessed by subtracting Senate votes from House votes and comparing these with results from other sitting members.  Peter Brent has done a lot of work of this kind and now we have a new deluxe treatment from Simon Jackman. Jackman finds that the difference between House and Senate vote for Turnbull is slightly larger than the NSW average (by a point).  He also finds that apparent personal vote declines by about half a point for each increase in the number of candidates (Wentworth had slightly more than normal.)

Looking at the list of seats with the highest personal vote, I think there is a systematic issue with this method of assessing personal vote.  The supposed highest personal votes are in rural seats.  It's well known that rural MPs tend to have higher personal votes and stronger effects on retirement, but there's more to it than that.  In rural seats there are very high Senate votes for One Nation and the Shooters, Fishers and Farmers, and these parties did not contest rural Reps seats much in 2016, but there are no similar parties that do so well in Wentworth.  So I prefer Peter Brent's method that includes adjustment for rural vs non-rural seats and found Turnbull to have the highest personal vote of all MPs in 2013.  However in 2016 this had dropped to a merely above average result, because the Coalition Senate primary increased by 11 points in Wentworth but Turnbull's own primary went slightly down.

In two-party terms the ABC's graph shows a huge leap in Wentworth 2PP results for the Coalition relative to the state average for the last three elections.  However this could also be attributed to Labor decisions to focus resources on more winnable seats.

It's difficult to decide how much should be added to the expected 2PP swing based on the loss of Turnbull's personal vote.  Any estimate from hardly anything to perhaps even double figures could be justified by reference to some item of data.  If we go somewhere in the middle and say five points or so, even that makes the 2PP vote close and a potential two-candidate count against an independent difficult.

The Circumstances, and Wentworth as potential indie territory

I think that treating this by-election as a routine retirement of a popular MP is misleading.  It is very unusual to have a sitting PM rolled by their own party in circumstances that have been impossible for the public to make sense of.  The prospect of voters venting against the ex-PM's party on account of the rolling of their own local member must be a strong one, even with Malcolm Turnbull stating he will support preselected candidate Dave Sharma.

There is really only one federal precedent of any kind for this contest.  In 1992 Bob Hawke was removed as Prime Minister by his party and his seat of Wills (which had previously always been won comfortably by Labor) was won in the by-election by popular local football coach, the independent Phil Cleary.  The "false swing" from Labor's 1990 Wills result to their result against Cleary was 23.6%. However, Hawke's rolling was much more justifiable to the public - he had obviously passed his use-by date and Labor polling had repeatedly collapsed to disastrous levels.  Probably, Cleary was a much more charismatic figure than anyone likely to run against Sharma, but this shows the sort of danger that independents pose - especially as Labor was actually coming off a low base in Wills following a swing against Hawke in 1990.

Wentworth also has a lot of the characteristics that make it a potential independent win in these circumstances.  In 2016 Labor polled 17.8% and the Greens 14.9%.  Provided the Liberal primary is knocked down to anything below about 46%, these Labor and Green primaries make excellent stepladders for an indie to win from third or perhaps even fourth on primaries.  Both Labor and Green primaries will flow strongly to independents.

Candidate Factors

The Liberal Party made a mess of its preselection for the seat, with PM Morrison raising expectations that a woman should be preselected, but failing to push this strongly far enough in advance.  Had there been a clear and open party policy of preferring a female candidate, a stronger field of candidates might have applied - as it was the initially favoured female candidate Katherine O'Regan was criticised as not notable enough for such a blue-ribbon seat (it turns out she is also a former Labor member.)  The preselected candidate David Sharma has strong credentials, is a diversity candidate on account of his Indian heritage, and is a former Ambassador to Israel running in a seat with an even higher Jewish-voter percentage than Macnamara (Melbourne Ports).  He is also backed by Malcolm Turnbull and John Howard.  However there may be some disappointment that a male candidate has won preselection, and it is also perhaps significant that Sharma doesn't live in the electorate.

The by-election will attract a large field, with nine candidates already confirmed.  The most speculated about independent (edit: now confirmed) is Kerryn Phelps AM. Phelps is a former Australian Medical Association President, a medical professor and a former Deputy Lord Mayor of Sydney.  She is also a convert to Judaism.  She has been recently prominently involved in LGBTI issues including same-sex marriage, and also in debate about My Health Record.  Phelps has a Twitter following of over 20,000 and has recently retweeted quite a bit of left-wing and independent material criticising the Liberals.  I also cannot resist mentioning that her complementary medicine involvements won her the 2008 Bent Spoon award from the Australian Skeptics!   Phelps, a self-described centrist, was a member of the Clover Moore team on Sydney Council, but fell out with Moore, with each side blaming the other.

It's not impossible in such extreme circumstances that the 2PP swing would even be large enough for Labor's Tim Murray to win the 2PP.  However, as we saw with the Wagga Wagga state by-election, winning the 2PP does not guarantee a Labor victory.  Deserting Liberal voters will probably be more attracted to an independent if there is one with sufficient profile.  Phelps, if she runs, would seem to be a major threat who will not be easily countered.  On the other hand, Murray, an economist (background here) has received some very interesting endorsements.

Would Defeat Cause An Election?

Prime Minister Scott Morrison (it still feels odd writing that) has ruled out an immediate election if Wentworth is lost.  Defeat in Wentworth would leave the Coalition on 74 seats plus "government crossbencher" Kevin Hogan, with 75 seats for Labor plus the unambiguous crossbench.  With Bob Katter and Cathy McGowan as well as Hogan giving Morrison confidence it's highly unlikely a move to bring down the government would have the numbers without someone crossing the floor or going AWOL.  Even if the votes for a no-confidence motion did somehow reach 75-74 on the floor this might be resolved (if detected in advance) by the old party trick of the Speaker resigning.  There would never be the numbers to suspend standing orders to bring on a vote without notice.

The more interesting question is whether the government, if it lost Wentworth, would struggle to get bills through the Parliament, since it would need at least one crossbench vote apart from Hogan's to do so.  It was the temporary loss of numbers that caused the government to cave in over the banking royal commission last year.  However the government is in such a poor shape to fight an election at the moment that it would be better off carrying on in a semi-impotent state than going to the polls right away.

Wentworth Polls

Wentworth being such a rich seat, we have even had the novelty of seeing preselection candidates trot out their own commissioned polling, but the polling that's been seen so far is mostly silly. Aside from the general issues with seat polling and the specific issues with the polls below, it's also notable that the recent Super Saturday seat polls badly underestimated the vote for tailenders in large fields.

1. TAI ReachTEL: The Australia Institute released a commissioned poll with a 50-50 2PP in a headline question without naming parties.  The poll usefully showed that naming candidates makes a big difference - when Alex Greenwich and Kerryn Phelps were named as independent candidates alongside a mix that included Sharma, the combined Independent/Other vote rose from 7.3% (undecided redistributed) to over 36%.  However this is not comparing like with like as the headline question concerned a general election while the second question specified a by-election.  I'm also slightly concerned that highlighting "Your local MP, former Prime Minister Malcolm Turnbull, is resigning" could have skewed the results compared to just asking which of the following candidates the voter would vote for in the by-election.  Also, Greenwich is not actually running, so it is difficult to use the figures to assess the prospects for an independent win, beyond that they exist.  Yet another reservation is that the poll called the Liberal candidate "David Sharma" when he uses the less formal "Dave Sharma" (which is also closer to his birth name anyway.) Finally, any polling of Sharma would be standing-start polling before he has had the chance to use the campaign to build profile.  He would need to build a lot from these figures to survive on preferences.

2. Bragg poll (pollster unknown): Preselection contender Andrew Bragg (who later withdrew) issued polling from an unknown pollster with a sample size of 1000 and primary votes of Liberal 39 Labor 25 Phelps 20 leaving 16 for Greens and others, but the Liberal primary said to increase to 43 if the Liberal candidate was female.  This was widely seen as justifying his decision to withdraw given he is said to be interested in a Senate spot.  There is too little detail on this polling to comment at length but on those primaries Phelps would be a chance to win from third place, depending on the split of Green preferences between her and Labor. If Phelps didn't make the final two, the Liberals might just hang on in this scenario, depending on Phelps' preferences.

3. King Sexton poll: Preselection contender Peter King (who finished third) produced a uniquely weird poll by Sexton Marketing, a sometime past commissioned pollster for Liberal internals and social conservatives.  King's poll, with an inadequate sample size of 400 (this is effectively about 70 by seat poll standards!) had a 47-53 2PP result if King wasn't the candidate and a 50-50 result if he was.  The weirdest thing about this poll is that choosing King as the candidate supposedly caused the Others (this excludes Independent) vote to increase from a ludicrously low 1% to 10%.  The increase would be entirely from female voters, which would make sense if one assumed that they didn't want a male Liberal candidate.  Except that the big drop in female votes in that scenario came not from King but from Labor!  So one has to imagine that there would be a big bunch of female Labor voters who would flock to Peter King but an equally big bunch of female Liberal voters who would flock from him specifically to Others, while male voters displayed little gender effect.  It's much easier to imagine this is just a silly poll.

4. Getup! ReachTEL: The GetUp! ReachTEL was taken September 17, before Kerryn Phelps' announcement that she would preference the Liberals.  The poll has a 52-48 2PP to Sharma but this should be ignored as on those numbers the real opponent is Phelps, who would win easily on those numbers.  The primaries as calculated by @GhostWhoVotes after reallocating leaning voters are "LIB 39.3 (-23.0 since election) Phelps IND 22.5 (+22.5) ALP 17.4 (-0.3) GRN 12.6 (-2.3) Heath IND 4.4 (+4.4) Other 3.9 (-1.2)." I strongly suspect the vote for Other will be higher. More on this poll later maybe as time permits.

5. Licia Heath uComms: A uComms poll conducted using ReachTEL's platform and commissioned by Clover Moore backed independent Licia Heath claims a 51-49 2PP result but presumably that's Liberal vs Labor.  This is off primaries of Liberal 40.6 Labor 19.5 Phelps 16.9 Heath 9.4 Greens 6.2 Others an absurdly low 1.8 "Undecided" (soft voters) 5.6.  After redistributing the undecided this poll is more bullish for the majors than other offerings.  The poll also includes a more likely/less likely question concerning Phelps' decision to preference the Liberals but this style of polling is predictively useless and should be ignored. On these numbers it would be challenging for Phelps to get up from third because of the preference flow required but the possibility should be considered rather than just reporting a 51-49 2PP as indicating anything after the sorts of seat poll failures we have seen lately.  The poll does raise the question of whether the Liberals are benefiting, as Tony Abbott did in Warringah, from being up against a circus of opponents fighting against each other.

Spurious Push-Poll Allegations: On October 13 (took their time!) Phelps campaign sources alleged this poll was a "push-poll".  However their case reduces to that a subject thought the poll was a push-poll because it asked if various factors might change their vote, and that a Phelps campaign strategist said "If that's not push polling, I do not know what is".  This does not make the poll a push-poll at all.  Push-polls involve the use of very large "sample" sizes to convey information under the guise of conducting a poll, and the information is then thrown away.  This poll used a typical sample size and gathered voting intention data that were published - albeit incompletely.  That the poll also conducted some scientifically invalid message-testing nonsense which was then reported in a way damaging to Phelps is, alas, not unusual.  The poll also reports someone who lived outside the electorate being polled, but this has been a common issue with ReachTEL seat polls, and it is as yet unclear whether it occurs to a significant enough extent to affect results.

The big question about this poll is: in what order were the questions asked.  The poll showed the Heath vote much higher than any other poll has, and was commissioned by Heath.  We need to know for sure that voting intention was the first question asked as if it was not, the other questions may well have contaminated it.  Ideally the full wording of all questions asked in the order asked should be published and the media (in this case the Guardian) should refuse to report on these polls unless this is done.

6. Refugee Council ReachTEL: This ReachTEL has a 53-47 2CP to Phelps if she makes the final two, but a 50-50 2PP Liberal vs Labor.  This is off primaries that, after redistributing undecided, William Bowe has at Sharma  (Lib) 39.9%, Murray (ALP) 25.0%, Phelps 17.3%, Greens 9.1%, Heath 3.6% leaving 5.1 for others.  On those numbers, Phelps would not overtake Murray so it would be a Liberal-Labor contest.

7. Voter Choice (12 Oct): A Voter Choice poll was reported by The Australian with Phelps ahead 55.4-44.6 but oddly Murray ahead by even more (55.7 - in the event of him making the final two).  Voter Choice is a relatively new and at this stage very experimental opt-in panel-type pollster.  The full report on the Voter Choice website displays some impressive detail (including a mock preference distribution based on a respondent ballot of the main contenders) but also some significant red flags.  Especially the pollster's candid admission that she didn't like the results (for what reason?) so she added some extra re-weightings post hoc:

"So I started with the usual weighting by age, gender and 2016 vote. Numbers still looked wrong. Added weighting by suburb – remarkable difference… but still not quite there. [..] I’ve never considered using the volatility score as a weight before, but I tried it – and those numbers I liked – so that is what I’m presenting here."

Firstly, this means is that what's being reported isn't a method the pollster has tested against previous election data, but rather an untested hypothesis about what might work.  Secondly "by adding volatility as a weight, it discounts those who may change their vote to, say, half of someone that has completely committed to who they are going to vote for. "  There are very good reasons to be very cautious about doing this - voters for certain parties (especially Labor) are more likely to report they are iffy about their votes than others.  (Weighting by past vote is also risky because voters will sometimes misreport it).  The opt-in nature (a la Vote Compass) also requires care - probably no amount of weighting can entirely undo risks with such methods.

The poll's preference distributions have Phelps almost caught by Murray off Green and Heath preferences, but that is based on a preference flow (94% Greens to Labor vs Phelps and Liberal combined) that is certainly nonsense.  Overall the poll finds Phelps getting 84% of preferences vs Sharma, which might just be believable, but Murray getting 87%, which isn't.

8. Liberal Internal: Internal polls are unreliable since parties tend to strategically release them when it suits their purposes to do so, and not otherwise.  In this case, probably in the interests of expectation management while maintaining some hope of victory, the Liberals are (13 Oct) claiming to be a fraction of a point behind Phelps, who is now "likely" in their view to finish second. 

9. Liberal Internal (17 Oct): The Australian reported a "leaked" (which usually but not always means it was deliberately given to the press by the party) internal poll in which the party is 45-55 behind Phelps.  Primary vote details are confined to:

"It is understood that the latest polling had Mr Sharma polling in the mid-30s on primary votes and Dr Phelps well into the 20s, clearly ahead of Labor."

 A curiously high 75% are supposed to believe the seat will be held by the Liberals.  The unnamed source said the party was being "killed by the expectation" that they would hold the seat, implying that they were struggling to hold off protest voters who wanted to send a message without affecting the result.

10. Greenpeace ReachTEL: The Guardian reports a Greenpeace ReachTEL on Monday that had the Liberal primary in meltdown on 32.7 from Phelps 25.8 Murray 21.6 Wy Kanak 9.1 Heath 5.6 Undecided 2 leaving a manifestly inadequate 3.2 for the rest.  The poll also asked voters to self-report their 2016 vote, which matched the Liberal primary (I am sceptical when I see this because voters do not always accurately self-report).  If these numbers were accurate Murray would have some chance of getting into second and winning on preferences, because he would need only 62% of preferences, which is quite achievable.  However there has since been some move towards supporting Phelps based on strategic considerations.  Again, this poll is commissioned and the sample size is not large, so treat with some caution, but it would have to be a long way wrong for the Liberals to win.  No 2PP was published with the poll initially but William Bowe reports Phelps would win by a ridiculous 62.4-37.6 after preferences.

11. More Liberal Internals: The report on the Daily Telegraph has so far eluded my paywall-evasion skillz but apparently says that tracking has them on a 39% primary and that it has had them as low as 41-59 2PP.  As William Bowe notes, these tracking polls are wildly unreliable and cherrypicking single-day figures from them as evidence of anything cannot be recommended.  (As we saw in Braddon.)

Authorisation

The Australian's article on internal polling (item 9 above) also says Phelps has been accused of failing to authorise social media messages as now required by the Electoral Act.  However the requirements are weak and it may be that having her name and city in her profile, which she does through the mention that she is a "City of Sydney" councillor, is sufficient.  It is not clear that any specific form of words such as "authorised by" is required so long as correct "authorisation particulars" are included.

If a strict "authorised by" formula is required then on the same basis, several other candidates including: Murray, Wy Kanak, Vithoulkas etc would also be in strife just based on Twitter alone.  But I see no evidence this is the case.  Wy Kanak might have an issue anyway as his Twitter profile doesn't contain authorisation particulars but it does contain a link to his Facebook page (that might be enough, though not on a literal reading of the link above.)

Betting

Betting odds are not reliably predictive but it's worth keeping an eye on this to see how much it jumps around.  As of 14 Sep, Sportsbet Liberal 1.65 Labor 3.00 Independent 5.00 Greens 24.00 and ignore the rest.  Ladbrokes similar. The market would be pricing in that Phelps has not yet confirmed she will stand, so if she does there may be some movement there.

As of 17 Sep, Sportsbet odds with candidates listed have Sharma 1.60 Phelps 2.40 Murray 7.00 Vithoulkas (IND) 14.00, Wy Kanak (Green) 18.00 Heath (IND) 31.00 (etc).  IND briefly blew out to 6.00 before these went up.  Betting odds are not reliably predictive.

As of 21 Sep, Sharma 1.45 Phelps 2.65

As of 2 Oct, Sharma 1.33 Phelps 3.50 Murray 7.50 Gunning (Liberal Democrat) 12.00 Vithoulkas 16 Wy Kanak 21 Heath 31 etc.

As of 15 Oct, Sharma 1.80 Phelps 2.40 Murray 8, Heath 16, Wy Kanak 21, Gunning 31 (etc).

As of 16 Oct, a further plunge: Sharma 1.88 Phelps 1.92 Murray 12, Heath 16, Wy Kanak 21, Gunning 31 (etc)

As of 17 Oct, it has flipped: Phelps 1.88 Sharma 1.92 etc.  And within hours, Phelps 1.75 Sharma 2.00 and the rest are all blowing out.  At 2 pm 1.65-2.15.

18 Oct Phelps 1.55 Sharma 2.40 Murray 14.  There's been a fair amount of comment on pseph websites suggesting that Murray's chances are better than that but so far the betting hasn't been reflecting it. By the end of the day Phelps 1.40 Sharma 2.75.

Through 19 Oct it bobbed about a lot with Sharma at times as long as 3.50 and at times as short as 2.50.  There is probably some profit-taking by punters who earlier backed IND at long odds on other sites in the mix.

Election day 1 pm Phelps 1.32 Sharma 3.20.

Phelps Preferences Sharma

On 21 Sep, Kerryn Phelps announced she will preference the Liberal Party on her how-to-vote cards.  This only makes a difference in terms of the mechanics of the count if Phelps is eliminated.  If Phelps makes the final two her preferences won't be distributed.

In terms of perceptions though, this is firstly aimed at discouraging Liberal voters from seeing Phelps as a Labor/Green stooge candidate.  Secondly it helps create a perception that Labor cannot win the seat (though whether those voting for Phelps would actually follow her card is another matter).  This perception can easily become a self-fulfilling prophecy.  If Labor want to force the Liberals to waste resources defending the seat (and possibly fail in the process) their best strategy is to not run too hard and preference Phelps.  Liberal supporters may have an incentive to strategically vote Labor in order to try to keep Phelps out of the top two, but relatively few voters vote like this.

Feedback on the move has been very mixed with some political observers saying it is very smart and others saying it will drive potential Phelps voters back to Labor and makes Phelps look like just another politician. My observation is that while I understand the strategy, voters like authenticity and this doesn't look like it - especially not when she encouraged voters to put the Liberals last last week and continually attacks the Government on Twitter.

It will be worth keeping an eye on any preferencing moves from fellow independent Licia Heath, who is backed by Clover Moore and Alex Greenwich. (Update: Heath is running an open ticket.)

The Greens are preferencing Labor, for which they were slammed by Joe Hilderbrand.  However relatively few Greens voters (rarely more than 5%) follow the party's how-to-vote card religiously, so this decision is only likely to make a difference if the margin between Labor and Phelps at the crunch point is very close (within 1 point or so).  At least some Greens voters will preference Phelps ahead of Labor no matter what the card says.