Wednesday, April 29, 2015

Poll Roundup: Silly Lefties Oppose Senate Reform

2PP Aggregate: 52.9 to ALP (+0.2 since last week, +0.5 in three weeks)
Labor would comfortably win an election "held now"

With the release of a new ReachTEL plus regular offerings from Morgan and Essential, and also a rather striking Morgan leadership phone poll, there's enough new content for another roundup of federal polling.

The Morgan phone poll didn't include a released 2PP figure (they sometimes include them and sometimes don't) so there are three new polls to add.  The ReachTEL taken on Thursday had a rounded 2PP of 54:46 and Essential's was 53:47.  However in both cases the released primaries pointed to the rounding having been in Labor's favour, and I aggregated them at 53.6% and 52.8% respectively.  

Last week's Morgan was 54:46 by last-election preferences; after adjusting for the primaries and the lean to Labor in Morgan's multi-mode series that one went in at 52.3.   The Morgan was unusual in that respondent-allocated preferences only gave an outcome of 53:47 to Coalition.  This was the first time in this whole term that respondent preferences have been a point worse for Labor than the published 2PP; they have been half a point worse four times with the last of those just over a year ago.

In all my aggregate shows a slight, but not yet statistically meaningful, drift back to Labor over the past three weeks.  This is best shown this week on the spiky (non-smoothed) graph of aggregate readings, noting that the ReachTEL has been back-inserted into last week's reading.




Very little should be made of the minor ups and downs over the past eight weeks, which smooth to a nearly flat line.  But at least we have it clear-cut that the recovery by the Abbott Government after the Prince Philip/Queensland election/spill turmoil was limited to undoing the damage that those events caused.  The rapid progress to a really competitive position hoped for after the 51:49 results from Ipsos at the start of March and Newspoll at the end has not yet happened.  Those two polls were slight outliers.

It is now one year since the Coalition's polling blew out by at least three points from a borderline winnable position prior to the 2014 budget.  In that year the Coalition only fleetingly and arguably got back to where it was.  It is now 17 months since the Coalition lost the lead on the smoothed aggregate, but this still does not mean too much predictively, since two of the four governments that trailed for over two years in one term were re-elected.

Reaction to the execution of Myuran Sukumaran and Andrew Chan may well have some effect on polling but it is hard to say what that will be.  I'm hoping that it will be nothing, as it would be sad to see any advantage gained from this event.  I oppose the death penalty in all circumstances, and am happy to discuss my reasons if anyone wants.  That said, and without seeking to equate mere stupidity (whatever its scale) with needlessly killing people, I am very unimpressed by the unhinged and opportunistic attempts to party-politicise the issue at the PM's expense from some sectors of the social-media and celebrity left.  

Leaderships

This is where the real fun is this week, with the Morgan phone poll producing startling results for the leaders of both parties.  The smorgasbord-style preferred Liberal leader poll had Malcolm Turnbull on 38 points leading Julie Bishop on 27 and incumbent PM Tony Abbott on a ludicrous 12, with Scott Morrison and Joe Hockey on 5 apiece, Barnaby Joyce 3, Christopher Pyne 2, Andrew Robb negligible, Someone else 1 and 7 for "Can't say".   The poll also had Abbott losing the lead even among Coalition supporters with Turnbull on 30 leading Abbott and Bishop 25.

But lest Bill Shorten get too excited about that (hard as it is to imagine Bill Shorten getting excited about anything much) he too is trailing a member of his own party with Tanya Plibersek 23 leading Shorten 21 Albanese 13 Swan 10 Bowen 5 Macklin and Burke 4 Someone else 1 and Can't say 19. Worse, Plibersek leads Shorten 30-26 among Labor voters, and Shorten only tops the poll among Coalition voters (of whom 22% support him, compared to 15% for Plibersek and Albanese).

The Morgan press release does say that Shorten's position is secured by the new Labor leadership rules, but the recent deposing of NT Opposition Leader Delia Lawrie shows this aint quite necessarily so.  If enough of a Labor caucus signs up, then the party can still place the leader in a position of having to win a strong majority in a membership ballot over one month, with it being public knowledge at the start that they had lost the confidence of Caucus.  Lawrie chose instead to resign.

Morgan's approval ratings showed Abbott on a fairly benign net rating of -16 (37-53), almost identical to three months ago.  However Shorten was the big mover down, coming out at -14 (34-48), compared to -3 in January.  The poll broke more new ground, showing Abbott as "better Prime Minister", 44-39, which I believe is Abbott's first such lead from any pollster for nearly six months.  It is natural for the incumbent PM to lead in such polling if their party is not too far behind, but in most recent polling this hasn't happened, because Shorten has been much less unpopular than Abbott.  Lately this is starting to change.

I'm not really sure why Shorten's numbers are falling, because most explanations that can be offered raise the question of why the same factor didn't have the same effect much earlier.  I wonder if voters are simply bored of him.

ReachTEL's ratings show Abbott continuing to recover from his very bad ratings of early this year, with a net rating (good minus poor) of -27.2.  (ReachTEL calls this "net satisfaction", but isn't someone who rates a leader's performance "satisfactory" satisfied too?) The poll shows Abbott's "Very Poor" rating down a massive 10.3 points in a few weeks, though this has mostly gone to Poor or Satisfactory.  Abbott's "Very Poor" rating (30.6) is in fact the lowest it has been since September.

ReachTEL, however, has shown very little change in Shorten's ratings over a long time now.  His current rating (25 good, 41.1 poor) is still better than the 22 good, 41.3 poor back in October 2014.  It's possible that what is going on here is that those responding "satisfactory" would have given Shorten a positive response to a phone poll six months ago, but would give him a neutral or a negative now.  The other possibility is that the robopoll has been getting the more authentic response all along.

ReachTEL's Liberal leader poll also shows sharp differences with Morgan's, and not for the first time.  ReachTEL also has Turnbull leading, but with 40.8% to Abbott's 28.2%, Bishop's 24.7 and Hockey's 6.6%.    (No other options, including undecided, are permitted).  Abbott also clearly leads among Coalition voters with 51.1% to 23.5% for Turnbull.  There are slight issues created by the two pollsters using different options lists, but as a share of all support for the three strongest candidates, Abbott has 30% in ReachTEL and 16% in Morgan.  He also has 55% of votes for the big three among Coalition voters, compared to 31% in Morgan.  These issues can't be explained away by sample size.

So a question that arises is: is there a shy Abbott supporter effect operating?  Are there voters - Liberal voters specifically - who think Tony Abbott should remain Prime Minister, but who find it embarrassing to admit this to a live human being?

This might be so, but I don't think so.  After all, Abbott's rating as a share of the three contenders in Ipsos in March (23) was similar to the two ReachTELs it was sandwiched between (19 and 25).  I think the spectacularly low ratings Abbott is recording in these Morgan leader polls have more to do with the form of the question.  Morgan asks:

"If you were a Liberal or National Party voter and helping to choose the Coalition Leader for the next Federal Election, who would you prefer?"

Even if the voter is an L-NP voter, it's a messy question form that seems to train the respondent more in the direction of the party's interests than who they would personally like to be leader.  So rather than there being shy Abbott supporters, I think there are Abbott supporters who will say they would like him to stay PM for the moment, but who believe that he will lose the next election, and hence concede they would go to the next poll under someone else - most likely Bishop.

Other Polling

There are the usual polls around showing that voters think the budget will be bad for them and bad for everyone else except big business (and so on).  Such results are predictable and boring.  There are some interesting results in the PUP party breakdowns in the ReachTEL, suggesting that the rump still opting for PUP are rich business types or rugged individualists.

There were some deeply disappointing results in this week's Essential on the question of Senate reform:

I do not really care for this question design.  Some voters might consider the Nationals and Greens to be "small" parties that might be disadvantaged, but the changes would not disadvantage either.  (They are about neutral in terms of likely impact for the Greens' seat representation, and favourable in terms of their ability to influence outcomes between other parties and chances to get a more effective balance of power.)  I would have strongly preferred the term "micro-parties" to be used, or else some indicator of the support levels involved.  The priming of the poll with information about what parties support the scheme is another problem with it - one capable of driving the responses of various respondents in either direction.  Really, these are the sorts of changes that take a paragraph rather than a sentence to explain, and a short question is always going to cause problems.

But all the same, Greens voters have been primed by the information that the changes are supported by their party and yet 49% of them are still opposed, and only 5% strongly approve.  It is possible dud question design is to blame, or that Essential's panellists (who would tend to be heavier than average internet users) have biases not reflective of the general population, or have been brainwashed by Bernard Keane.  But I fear that this is not the case and that what we are dealing with here is simply an outbreak of opportunistic political short-sightedness on the left.

The main cause of this problem is that the current crossbench has severely frustrated the Coalition Government.  Some left-wing voters like this, and think the Coalition would have it easier under a Senate system that got rid of the micros.  Whether this is actually so is not so clear - the primary votes recorded at the last two Senate elections would probably have given Labor and the Greens half the Senate combined (and hence a blocking majority) under the proposed system, but on the other hand right-wing micros could well have adapted to the system by merging, and hence achieved a slightly better outcome for the right.  (Strategic running of separate Liberal and National tickets might also have worked in some cases).  Most likely we'd have had a crossbench of something like four or five, with Labor and the Greens needing one or two of them to block things.

If the current system survives, left-wing voters happy about the micro-parties because the micros have held the Coalition to ransom might come to regret that stance when it is the ALP in power and trying to get progressive legislation through two social-issue reactionaries, an anti-Islamic hysteric, an implacable ideological libertarian, a pawn of Clive Palmer and so on - to say nothing of what the next half-Senate election might add to this random gaggle.

They should also reflect on what we narrowly avoided and might well get in future.  As already noted, Bob Day is bad enough, but in Tasmania, we missed the appalling spectacle of anti-gay extremist Peter Madden being elected to the Senate on Labor and Green preferences, by just 821 votes.  (Instead, we ended up "just" getting Lambie.)

I am no fan of the "big three" parties and I usually vote for somebody else.  But there are far worse things out there than Liberal, Labor and Green, and I'd sooner make it harder for good micro-parties to win than keep it too easy for the worst of them to flourish.  And, ultimately, no-one should have the political advantage of their own type of party at the top of their concerns list here.

This should be about ensuring that those with negligible support do not get elected through a broken system that makes it too hard for voters who might want to control their preferences to do so.  If that makes it slightly easier or harder for governments of a particular stripe to pass legislation for which they can be judged at the next election, or means we get more boring pollies and less who seem to have escaped from a philosophy-class reality-show, then so be it.  That is for the voters to decide.

In a vaguely hopeful sign for voter intelligence, 77% supported the government's very lenient recent decision to merely withhold tax and childcare benefits from parents who endanger the health of their own children and other children by refusing to vaccinate. Even 12% opposed is far too high, and again, alas, the left has more to blame for here.

2 comments:

  1. Kev

    It intrigues me a little that of the three "main" on-line bloggers who attempt a poll aggregation, using the same data (I understand MtB ignores Essential) show quite a bit of variation. As I write this, I believe the latest ALP 2PP aggregates are:

    PB 52.4
    KB 52.9
    MtB 53.5

    Or maybe it's not surprising? I know you each have your own methods and allowances for house bias. I’d be interested to know why there’s a variation and the differences in the methodologies (that a layperson like me can understand).

    Peter Tucker

    ReplyDelete
  2. Differences between MtB and PB/KB are mainly explained by different treatments of house effects. MtB sets his model so that the house effects of all the different pollsters are assumed to sum to zero at any given time. PB/KB do not use this assumption. We both believe that Morgan is skewed to Labor, that Ipsos is probably skewed to Coalition but not by as much as Morgan is skewed to Labor, and that the rest are about neutral. As a result PB and KB will generally show a less-Labor friendly result than MtB.

    On average the readings of PB and KB have been almost exactly the same since the last election, but from time to time they differ, and for the last three weeks mine has been about 0.5 better for Labor.

    There are a number of minor differences between our methods that can cause these differences from time to time:

    * PB uses the primary-vote data released by pollsters and ignores the 2PP released by the pollster. My method uses both the released 2PP and the released primary votes. So say for instance the primaries for a Newspoll as released give a 2PP of 53.0 if you plug them into a formula from the last election, but the 2PP Newspoll releases is 54. This could be caused by the way the released primaries and the released 2PPs have been rounded to the nearest whole number, or it could be caused by (for instance) the Greens doing very well in a state where their voters more strongly preferenced Labor. In this case PB will effectively treat the poll as a 53.0 but I will treat it as 53.6.

    * I assume Essential has no house effect, but I downweight it. PB assumes it has a house effect that varies from time to time.

    * PB's maths are more complex and dynamic than mine - mine tends to retain older data slightly longer and varies by slightly less on average from reading to reading.

    * We have different quality weightings for different pollsters, and also slightly different estimates of their house effects.

    * PB takes more account of sample size but I treat all polls with a sample size of at least 950 respondents as being on the same level.

    There's also a fourth one, Phantom Trend, which currently has the ALP at only 51.89. I believe the cause there is different house effect assumptions based on benchmarking off just federal elections, whereas William and I benchmark our house effects off federal and state elections. (Actually mine more or less copies William's benchmarking.)

    ReplyDelete