Wednesday, April 20, 2016

Poll Roundup And Seat Betting Watch: Labor Contesting The Lead

2PP Aggregate 50.1 to ALP (+0.7 in a week)
Coalition would still probably win election "held now" (seat estimate 77 Coalition 69 Labor 4 Others)
First ALP lead on my aggregate since Malcolm Turnbull became Prime Minister

In this issue:
A DD Is The Right Thing To Do
This week's polls
Issue Polls
Fishy Polls Of The Week
Seat Betting


In another stirring triumph of tipping skills, a plurality of voters on my sidebar Not-A-Poll have correctly predicted that the Labor Opposition would recapture the 2PP lead on this site in March or April.  35.5% picked this, compared to 34.1% for May or June, 18.2% for not at all, and a dribble for various options that depended on a later election date. It's a trivially small lead, it's not an election winning lead, it's not being replicated by other aggregators yet, and it may not even last long enough to survive on the smoothed tracking, but it's still a big improvement on losing 54:46 just three months ago.

There is some rejoicing and a fair bit of schadenfreude on the left about the direction polling has moved in.  Many lefties seem amused that Malcolm Turnbull pulled a constitutional swifty to beef up his argument for a double-dissolution only to find himself in a position where it might not seem like such a great idea anymore.  With the rejection of the ABCC bill at the second reading the government has no obvious plan B; to welsh on the threatened double dissolution on account of indifferent or even bad polling would just make the PM a laughingstock.

A lot of this is premature based on current polling.  50:50 eleven weeks from an election is not much different predictively to 51:49 fourteen weeks from one.  Incumbent governments polling about 50:50 at this stage are usually re-elected, sometimes comfortably.  (The two regressions I track are giving the government 72% and 79% chances of winning based on historical data, though with expected average 2PPs now down to 50.2 and 50.5, which would  not result in a friendly Senate.)  There's also still the possibility that preference flows will be a lot better for the Coalition than in 2013.

There is no shortage of debate about whether the government would have been better off going to a normal half-Senate election and trying to marginally improve its Senate position.  Antony Green provides a strong case that the outlook in seat terms is if anything slightly worse for the Government at a double dissolution.  On the other hand, even if the Coalition's seat position is weakened slightly, it could well be offset by a more manageable crossbench with fewer parties to negotiate with, especially should the Nick Xenephon Team win multiple seats in SA.  That is assuming the Coalition wins the election at all.

What is missing from all this is the argument that a double dissolution should be welcomed by everyone because it is right, and indeed it ideally should have happened sooner.  The ludicrous nature of the old Senate voting system can be seen in its full force with the blocking of the ABCC bill with the votes of crossbenchers Lazarus, Lambie, Muir and Madigan.  How terrible/incomplete people might think the ABCC bill is or what one thinks of these Senators is beside the point.  At least two and probably three of these Senators would not have been elected under a fair electoral system and only one of them (Muir) still formally represents the party they were elected as a candidate for.  Two were elected as candidates for a party that is now actually polling zero.  To allow the results of the 2013 Senate farce in particular to continue disrupting Australian democracy for another four years in the name of political advantage would have been a severe disappointment.

This week's polls

I noted the 50:50 ReachTEL in an update to last week's edition.  This was followed by a 50:50 Ipsos, a 51:49 to ALP Newspoll, a 51:49 to ALP Morgan (by last-election preferences; 50-50 respondent allocated) and a 50:50 Essential.  Not a lot of variety here! My aggregate treated the ReachTEL at 50.4 for Coalition (they have confirmed this as the decimal value but I was going to use it anyway), the Ipsos at 50.1, the Newspoll at 49.1 and the Morgan at 49.2.  Essential was aggregated at 49.8.  It looks like Essential has ceased favouring Labor compared to other polls, in that other polls have crashed for the Coalition, while Essential (as per normal) hasn't moved.  Therefore I've removed the house effect assumption for Essential.

Of the polls that most favoured the Coalition after Turnbull became PM, one (Morgan) strongly appears to have now ceased doing so.  For those that have been more Labor-friendly, Essential may have changed but Newspoll hasn't. If anything, it has become more ALP-friendly compared to the consensus of other pollsters.  There's a solid case for applying a house effect correction to Newspoll but for the moment I've decided to deal with the issue by ramping up the "global house effect" to 0.2 points to Coalition.  There's a case for going slightly further, in which case Labor's 50.1 on this week's reading might become 50-50 or 50.1 to Coalition.  Which party is in the lead by so little is not hugely relevant for predicting the election when the benchmark for a Labor win is about 50.9.

On current figures, I still have the Coalition on 77 seats (a majority), Labor on 69 and Others on 4.  For those wondering about Nick Xenophon Team, I have done extensive modelling on this and on their current support level (high teens) I don't yet have them winning seats. I was going to explain my modelling approach in this post but the explanation got too long, so I intend to do it in a standalone post soon.

Here's the smoothed tracking graph.  Turnbull has kept the Coalition in front for longer than Tony Abbott did but it would be no surprise now if things got worse for him before they got better.

Some other aggregates: Andrew Catsaras 50-50, BludgerTrack and Mark the Ballot 50.1 to Coalition,  Luke Mansillo 50.4, Phantom Trend 50.6 to Coalition, Metapoll 50.9 to Coalition (I'm quoting their betting-odds-free figure but note that they are using respondent preferences) and I'll add others as I come across them.  Metapoll are now giving the Greens two seats, but I believe they have the Green vote too high at 11.6%. The polls that have it above 11 were both untested in their current forms at the previous election (at which even the more conservative polls overstated it).

An otherwise problematic World Wildlife Fund commissioned ReachTEL (more on that later) had a very interesting finding: Coalition and Labor supporters said they were equally likely to change their minds on which way they would vote, with almost 80% of each saying they would definitely not change their vote, compared to 62% for the Greens.  Normally, the Labor vote is softer than the Coalition's.


The most revealing leadership result this week was from ReachTEL.  Turnbull's plunge to a -11.1 net rating was no surprise after the previous (-10) Newspoll but what is interesting is that only 25.5% of respondents rated Turnbull's performance Good or Very Good.  While Turnbull's net ReachTEL rating is higher than any recorded by Abbott in 2014 or 2015, his combined Good/Very Good score is worse than any polled by Abbott in Abbott's last five months in office.  Compared with Abbott in August 2015 (a few weeks before he was dumped) Labor voters are more likely to think Turnbull is doing well (5.8% compared to 2.6%), Green voters are much more likely to think so (13.6% compared to 2.8%) but Coalition voters are less loyal to Turnbull now than they were to Abbott in the circling-the-drain phase of his leadership.  58.3% of Coalition voters thought late-period Abbott was performing well or very well, but only 47.6% of them now think this of Turnbull.  This is not because they're all on the fence either - at least as many think he is doing badly.

That finding probably won't be reproduced by other pollsters because of the different scales used.  What the result reveals is that the "satisfied" or "approve" responses from Coalition supporters to other polls will include many voters who will only rate Turnbull's performance as "satisfactory" when given the options of "good" or "very good".  The base suffered with Abbott because they liked what he was saying, but they expect Turnbull to dominate the polls.

Turnbull polled a net -13 from Newspoll (-3, 36-49), meaning he has now lost 51 netsat points in five months, only four points shy of Keating's record for this or any similar timescale.  He's also lost 24 points from his "satisfied" rating in the same time (Keating lost 26).  Ipsos' approval ratings have long been kinder than Newspoll's satisfaction ratings but even so Turnbull is down ten points in a month there to a net +13 (51-38).

All polls record a closing of the preferred/better Prime Minister gap: Newspoll by two points to 19 (47-28), Ipsos by 12 points in a month to 27 (54-27) and ReachTEL by 3.2 to 16.8 (58.4-41.6 forced choice).  All these are good leads to Turnbull in the context of such close 2PP polling and his own indifferent ratings, but the explanation is that Bill Shorten still struggles.  There is some improvement on average across all polls there with Shorten up 8.2 points to net -19 in ReachTEL (his best since May last year), stable on net -21 at Newspoll, and down three to -22 at Ipsos.

Ipsos found that Turnbull keeps flogging Shorten on attribute polls, winning nine of 11 categories with an average lead of 14 points.  However he has dropped an average of 14 points on the listed attributes since October, while Shorten's ratings have barely moved.

Issue Polls

Bank-bashing is an Australian national pasttime.  Although Australians complain about banks a lot, they are famously reluctant to exert consumer pressure on them by switching.   Banks exploit this by charging high fees, and are hence easy targets, irrespective of more specific problems.  It's little wonder that Labor's call for a Royal Commission into banking is a big winner at the moment, with the seal of approval from Ipsos respondents (65-26), Essential (59-15), and ReachTEL (54.1-18.3).  Essential does another of its A/B sample experiments and finds that respondents who are given the other side of the story (attributed to Turnbull and Shorten) are a little less enthused about the exercise, though that could be because the text gives an argument against but no arguments for.  Coalition voters seem most easily pulled into line here.  A slight question design difference is also worth noting here: ReachTEL asked whether respondents thought Malcolm Turnbull should call a royal commission, as opposed to just asking if there should be one.

The late-release questions from last month's ReachTEL show the economy rates way above a list of other issues as the most important issue of the election, with 38.7% to just 16.3% for health and 12.2% for education.  All may not be quite as it seems: economy's frequent competitor "jobs" wasn't in the readout, while "housing affordability" may have split the left-wing vote.

Newspoll (PDF download) found "The Turnbull led Coalition" preferred 45-31 to "The Shorten led Labor Party" on "most likely to spend responsibly and manage government debt".  This is an unremarkable result since the economy is a traditional Coalition strength area.  All the same, it is interesting that Green voters broke as weakly as 41-26 in Labor's favour, assuming that figure is accurate (the breakdowns imply that Others voters were nearly all uncommitted).  Another Newspoll program finds that respondents prefer to reduce spending to pay down government debt (39%) or to reduce taxes (26%) rather than increasing spending "on government programs" (23%).  However I don't trust the design of this question. Had the spending increase been worded as "on health and education" rather than "on government programs" there would have been a very different result from non-Coalition voters.

ReachTEL last month found a slightly negative report card (29.3% positive, 38% negative) for the eight Senate non-Greens crossbenchers, even though the question incorrectly called them all "Independent".  Coalition supporters were very scathing about the crossbench while other party supporters were positive.  The question might be the first one I have seen to include both "undecided" and "don't know" as options, which certainly helps to separate those who are sure they're not sure from those who don't know if they do know or not!  (Others voters were most prone to the ultra-dither category here.)

Fishy Polls Of The Week

Wirrah Award For Fishy Polling (image source)

A closely fought contest for the prize of the week's most dubious commissioned poll.

Some readers might be expecting an easy win for the Greens' poll of Melbourne Ports.  This poll's wording (as always, doubtless at the behest of the sponsor and not the pollster) included the questions:

Michael Danby has stated that he will preference the Liberal Party ahead of The Greens, do you support or oppose his decison?


Does Michael Danby's decision to preference the Liberal Party ahead of The Greens make you more or less likely to vote for him?

Now, it is literally true that Michael Danby has said he will preference the Liberals ahead of the Greens - on his own ballot paper.  But when voters hear questions like this they are very likely to be misled into thinking Danby has caused or even could alone cause how-to-vote cards in his seat to show a recommended preference to the Liberals, something far more significant and which actually hasn't happened.  They may even think he has sent preferences to the Liberals in the sort of way that used to be possible under the old Senate system but is now thankfully banned (High Court willing).  The result is a skew-poll based around a strawman: the Greens get press attention for apparent voter outrage at a decision that doesn't actually exist.

This is all then topped off by an especially silly claim by Adam Bandt:

"In the three horse race in Melbourne Ports, I expect progressive Labor voters will shift to the Greens' Steph Hodgins-May rather than risk their preferences electing the Liberals"

Actually, any Labor voter in Melbourne Ports who wants to eliminate any risk of their preferences electing the Liberals can achieve this by just putting the Liberals last.  Even if they then put the Greens second-last, their own preference will never elect the Liberals.  Bandt should know better, he should hope that progressive voters know better, and even if he thinks that they don't, he should refrain from making rubbish claims about the preferential system in a quest for votes. There is quite enough of this sort of nonsense from the majors.

But no, there was another contender!  I continually link to a Yes Minister skit in which Sir Humphrey shows how you can train a respondent to give whatever answer you want by bombarding the respondent with the right line of questions.  Here's one that actually sort-of does it.  After asking respondents what they think about subsidising fossil fuels, the fuel costs of mining companies, the fuel costs of airlines, the fuel costs of farmers, and the production of oil and gas, the World Wildlife Fund poll then at question eight asks which of a bunch of revenue measures voters would prefer in the upcoming Budget.  Surprise, one of them is ending fossil fuels subsidies, and would you believe it, it actually wins!

At least WWF didn't throw away the other questions and publish the last (probably because they thought the remaining results were pretty juicy), but their online article did lead with the completely artificial question eight finding.  A series of lead-in questions making the case for any of the other options would have led to a totally different result.

Betting Watch

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.

The headline rates as of Friday had the Coalition between a 69% and 72% implied chance of winning depending on the market.

Checking the close seat list (defined as more than two parties at $3 or less) Paterson has moved from a tie to the ALP column as the incumbent is retiring.

Loss (Coalition to Labor): Barton*
Close Loss (Coalition to Labor): Solomon, Petrie, Capricornia, Hindmarsh, Lyons, 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

That's the only change so that's 85 Coalition, 61 Labor, 4 crossbench based on individual favourites.  However a market on "number of lower house crossbenchers" now has five equal favourite with four for the first time.

The Sportsbet 2PP market is still crediting the Coalition with about 52.5%, and the "line" market still has the Coalition winning by 18 or 19 seats (about 82-83 for Coalition).  The William Hill exact seats market however is down to an implied 76 seats for the Coalition now, so there are some very different expectations on the different markets.


  1. For Tasmania, early prediction is probably similar to these seats, Lyons to Labor, Bass and Braddon contested but ultimate Liberal hold. Senate goes 5-4-2-1 (Liberals, Labor, Greens, Lambie) although Greens could pick up a third from either Liberals or Lambie, particularly if Turnball keeps shunning Tasmania.

  2. I'll be really surprised if the Greens get three in Tas though I don't rule it out completely. They would have done so in 2010 had that been a double dissolution but that was the year they got 21.6% at state level. I am not convinced their base vote has rebuilt to anything like the 2010 level yet (polling that always overstates their vote here notwithstanding.)

    Lyons I was rather confident of a Liberal hold until recently (off sophomore effect and the apparent lack of swing in Tas federal polling samples) but the state income tax charade has made me less so.