Sunday, April 10, 2016

Poll Roundup and Seat Betting Watch: Newspoll Uber Alles

(Note for Tasmanian readers and anyone else interested: my piece on the Colbeck demotion appears below this one, or click here.)

2PP: 50.6 to Coalition (-0.7 since two weeks ago) - updated from 50.5 following Essential
Closest margin since Malcolm Turnbull became Prime Minister
Seat projection off this 2PP: 80 Coalition, 66 Labor, 4 Others 

After five weeks of reasonably static polling with the Coalition's 2PP at just above 51%, the past week saw another dip that was widely attributed to the government's unsuccessful and apparently half-hearted attempt to convince State Premiers to support state-based income taxes.

Newspoll especially got a lot of attention by finding Labor to be ahead of the Coalition (51:49) for the first time since Malcolm Turnbull took over as PM.  Much reporting - with the ABC alas the worst culprit that I noticed here - treated Newspoll as if it was the only poll in town.  In fact Morgan was the first poll to have the Coalition ahead on 2PP (two weeks earlier but not in the current week), and reputable aggregates all still show the Coalition with a slim 2PP lead.

The attention on Newspoll in the political classes to the exclusion of most others has long been disproportionate but may be becoming even more so.  As evidence that the Turnbull government was in trouble, a single 49:51 result was very flimsy. given especially that:

* Newspoll is, as Antony Green notes here, basically a new poll, albeit one that is asking old questions.  Not only is the new Galaxy-run Newspoll untested at an election, but it is also an amalgam of two polling methods (landline-only robocalling and online panel polling) that both have mediocre reputations.  Perhaps the well-known skills of Galaxy will succeed in maintaining or increasing the brand's accuracy, but we don't know that yet.

* 51:49 to Labor would be a very lineball result because Labor starts at a disadvantage arising from its drubbing at the previous election.  Contrary to the Australian's graphic showing it as a Labor outright win with 78 seats, a 51% 2PP for Labor would be more likely to result in 73 seats (give or take a few) so it is far from clear Labor would win on this percentage.

* The new Newspoll appears to have a slight pro-Labor skew.  My estimate of this has now gone up to 0.7 points.

If we take these points into account, the 2PP lead to Labor in this poll is probably illusory in terms of an election outcome, even before we consider other polls.  While journalists blame politicians for taking Newspoll so seriously, I believe that it's a feedback loop in which each of the media and the political class makes the problem worse.  Newspoll is certainly worthy of more attention than, say, Morgan or Essential, but extent of focus is just silly.  Even tonight the ABC were at it again, declaring Turnbull was "behind in the polls".

Other new polls last week were a 51.5:48.5 result for the Coalition from Morgan (52.5% by last-election preferences) and another 50:50 from Essential.  I have ceased treating Morgan as having any skew (though this may be generous if the previous fortnight's result was just an outlier) but I still have Essential as leaning to Labor based on a large divergence from the aggregate since mid-November.  I aggregated the Morgan at 51.5, the Newspoll at 49.1 and the Essential at 50.7 and my reading as of last week was 50.7 to Coalition.  Each weekend I apply a reset which tends to favour the more recent data, and this knocked the government down again to a starting value for the coming week of 50.5.  There may not be much happening this week with no major poll known to be scheduled (just Essential).  Here's the smoothed tracking graph:

Other current aggregators have recently had: Andrew Catsaras and Phantom Trend 50.4, Poll Bludger 50.8, Mark the Ballot 51.4.  We might be all too harsh on the Coalition should preferencing behaviour change sharply at this election, though the historic record shows that it usually doesn't, not even when people and polling expect it to do so.

A new Metapoll release (which does include data from their own polling on preference allocation, but fortunately this time does not include unstated sampling from their own primary-vote polling) had a polls-only 52.1 to Coalition (or 51.8 with the dubious addition of betting odds).  It is interesting that their modelling now shows an extra seat for the Greens with the Xenophon Group also close to a seat.  Trying to project seats for either off a given primary vote is rather challenging, especially because of the potential of the Green vote to increasingly concentrate in some inner-city areas.

Most governments would be happy to have any kind of polling lead at all twelve weeks from an election, or five or six months from one for that matter.  However this latest dip (which might for all we know be down to random flutterings in Newspoll and not caused by any specific issue) again raises the question of how bad things might get.  Are we still seeing a shakeout from the leadership change from Tony Abbott in which Malcolm Turnbull still carries and can lose residual goodwill?  The closeness of current polling is at least a wake-up call to a government that appears to be struggling to get its economic message in any sort of order.  This may all seem a world away when we get to an election with scare campaigns at maximum volume and hard questions about details of Opposition policy, but just for the moment Labor seems so much more cohesive and focused.


Turnbull's popularity plunge in Newspoll grows ever more spectacular.  A new worst -10 (38:48) rating this past week means he has now lost 48 netsat points in four and a half months.  Paul Keating alone is still ahead of him on that timescale (with 55 points lost in that time) but 55 points is the most Keating ever lost.  If Turnbull loses another eight points soon he will set an all-time record for the most netsat points lost in less than eight months, if not longer.  Turnbull's lead as perceived better PM was also cut by 10, to 21 points (48:27), the first time fewer than 50% of respondents have preferred him to Shorten.  A steady supply of free kicks for Bill Shorten sees the Opposition Leader's personal rating climbing to -21 (32-53), his best since last August.

There was apparently a plan to throw Shorten under the bus if necessary but there now seems little doubt that he will lead the party to the next election, especially if it is held on July 2.  I've closed my Not-A-Poll on the ALP leadership and here are the final results:

In the end Shorten's margin and overall vote share are unflattering, but many of the votes against him are old rope and in recent months he has been pulling away from the field.

Newspoll had a rare (for it) poll of preferred Liberal leader between Turnbull, Julie Bishop, Morrison and Abbott, and also on whether Abbott should have been removed.  On the latter question, while the margin for removing Abbott has narrowed from 62:27 to 57:31, the narrowing comes entirely from non-Coalition voters, who initially very strongly backed the removal of Abbott but are now adapting to the normal pattern of disliking whoever the leader is at the time.

Essential has a preferred Treasurer poll with incumbent Scott Morrison leading Labor's Chris Bowen 26:23, a four-point improvement for Bowen since January while Morrison's figures have changed little since he took the job.

Newspoll found massive opposition (19:58) to the state income-tax plan, an unsurprising result given that the thought of potentially having to fill out two tax returns is enough to scare many voters off it before they even think about whether it would cause tax rises or whether services might be worse.  Essential managed to get a 34:34 split but its wording did stress that federal income tax would fall.


Two new commissioned seatpolls have been released, one concerning the seat of Indi.  The ReachTEL commissioned by The Australia Institute found Cathy McGowan to be leading former incumbent Sophie Mirabella 37.3% to 26.8% on primary votes.  This was widely reported as a close result: a small lead, close because of the 10.3% National vote, and so on. Even the incumbent said the result showed she could lose.

In fact on these primaries based on preference flows in Indi last time, McGowan would win massively.  Assuming even a 15% flow from the Nationals she would win 58:42, but a substantially higher flow and a margin approaching 60:40 is entirely plausibe. Some people forget that McGowan trailed on the primary vote by 13.5 points and won the seat on some of the strongest preference flows in the entire country.  I don't have that much trouble believing that McGowan could retain the seat this easily or nearly so: rural independents once elected often become very popular, and preselecting Mirabella for another shot was such an obvious case of ignoring the electorate's Mirabella-specific message that the only possible benefit of it could be that a heavy loss will shut her up.

The other hails from Mackellar.  In the midst of Liberal preselection aggro, a MediaReach poll shows that Bronwyn Bishop would be thumped 54:21 on primary votes by independent Dick Smith, a margin somewhat difficult to credit (though the finding that he would defeat her is not).  MediaReach is a robopoll and I can find no evidence that it has been tested at an election under this name, though it may have been tested under a previous identity.  Anyway the lack of release of full details of this poll including question wordings is a caution-flag on this one.

Other Polling

There is a vast amount of issues polling out there and as usual I will just select items of most interest.

The Australia Institute has been churning out commissioned polling constantly in its quest to convince newsmakers that we're really all lefties deep down. The findings are mostly unsurprising.  Rural voters would support more ABC funding if they got more rural news, film at 11.

The TAI poll on whether to reintroduce the ABCC or to have a broader anti-corruption commission is really a bit of a "pony poll" (everyone loves the idea of a national ICAC if they haven't been told what it will cost or that it will probably sooner or later exceed its powers.)  However the wording also uses a leading tactic that the government says blah and that "others" have a better idea (the "others" are not named, so that only one side of the equation is politicised.)

One of the TAI's more interesting recent polls is this one on awareness of different ministers.  Some of their recent polls including that one have been using Research Now instead of ReachTEL.  Research Now have a history of using online panel polling (a la Essential) and while I've seen claims that they are a robopoll, I can't find any evidence of this.  From what I can determine, while Research Now has been about for a while, it has not released any published voting-intention polling that has been tested at an election.

Essential have their usual party attributes polls (ALP, Liberal, comparison) with some slightly surprising findings.  It's not clear what the ALP have done to be nine points less moderate than when the survey was last taken, or less moderate even than the Greens-beholden Gillard government. Nor is it clear why Labor scores four points worse for having good policies at a time when it is actually releasing some rather than having none at all.  Looking at the overall tendency for both parties to be down on both positive and negative attributes, perhaps this was an especially "meh"-prone bunch of respondents but that doesn't explain those indicators that are up.  Anyway the Liberal Party scores poorly on the comparison, winning only on its leadership team, vision for the future, clarity of message (??) and having good policies, while the rest are Labor wins except one tie.

Essential's tax cuts question has been fairly criticised in some circles for implying that either company or personal income tax cuts must be good for the economy (there is no option for saying neither would be good but one would be worse than the other).  Naturally, voters more often want income tax cuts and their strength of desire varies by party in a predictable fashion.

Essential finds pretty strong approval (44:23) for calling a double dissolution to break the ABCC deadlock should it persist.  Only Others voters say no, presumably in the belief that micro-parties will be eliminated from the Senate.


The usual disclaimer: the purpose of this section is to monitor the effectiveness of betting odds, and not to encourage either betting nor the belief that betting markets are strongly predictive.

There has not been a great amount of change in seat betting, possibly because not a lot of money is going into these markets yet and polling is relatively static.  Based on my usual formula of counting a seat as "close" if two or more parties are at $3 or less, this is the lineup of expected changing or close seats at Sportsbet:

Loss (Coalition to Labor): Barton*
Close Loss (Coalition to Labor): Solomon, Petrie, Capricornia, Hindmarsh, Lyons
Tie (Coalition-held): Paterson*
Close Loss (Labor to Coalition): McEwen
Loss (PUP to Coalition): Fairfax

Coalition Close Holds: Eden-Monaro, Dobell*, New England (vs IND), Brisbane, Bass, Braddon, Cowan, Burt
Labor Close Holds: Parramatta, Moreton

The only changes here from last time are that Dobell is no longer a loss in the markets while Brisbane and Bass are additions to the close seats list.  So that's 86 Coalition, 59 Labor, 4 crossbench and a tie, but the Coalition has a few more shaky seats now.  Also while the crossbench is only favourite in four, there are many in which some non-major party candidate is at single figure odds.  Some further notable examples to those mentioned before are Melbourne Ports (where the Greens are at $5 although Michael Danby has said he personally preferences the Liberals) and Mackellar (where Independent is at $6 because if Bronwyn Bishop is reselected an indie could very well beat her.)

Labor's market favouritism in Lyons has started to make some kind of sense with Malcolm Turnbull more or less ignoring the state in his state tax push; there may also be some feeling among punters that Tasmania's energy crisis could worsen.  An unfriendly Morgan sample this week saw the BludgerTrack swing in Tasmania change to 1.9% against the Coalition, which could be enough for the Coalition to lose a seat somewhere, despite sophomore effects.  While I have said the Coalition incumbents in Tasmania were well placed to retain, a lot of caution is required if Coalition strategy is to place the state at risk to shore up WA.

The Sportsbet 2PP market is crediting the Coalition with about 52.5%, and the "line" market has the Coalition winning by 18 or 19 seats (about 82-83 for Coalition).  The William Hill exact seats market seems much more bearish (I take it as implying only about 77 Coalition seats now) and the Sportsbet seat totals markets aren't online anymore.

I will update this piece with new polls this week, and expect there to be many polls in the week following.

Update: Essential (12 April): Today's Essential came out at 50:50 again, which I aggregated at 51 to Coalition.  This brought the Coalition an extra tenth of a point on the aggregate, but 50.6% is still their worst value since Turnbull took over.  Essential's relatively mild leader ratings have Turnbull on a netsat of zero (39:39, down ten in a month), Shorten on -14 (30:44; +6) and Turnbull preferred over Shorten 44:22 (compared to 48:19).

Other findings include no real change since March on the ABCC, which 34% of voters consider important and 41% not important (with predictable partisan skew).  Support for a double-dissolution should the Senate fail to pass the ABCC bills is up from 34-22 to 39-24.

ReachTEL Update (15 April): I was going to leave the ReachTEL til next week but just a quick note that Seven News is greatly overplaying the extent to which the new ReachTEL points to a hung parliament.  It's 50:50, rounded down from about 50.4 to Coalition, which would equate to about 79-80 Coalition seats.  51:49 to Labor (as in the most recent Newspoll) would be a likely hung parliament result and anything more than that to Labor would be a likely outright Labor win.  The ReachTEL hasn't actually moved my aggregate at all - it ends the week at 50.6 but the new week will start at 50.4.

No comments:

Post a Comment