Tuesday, June 30, 2020

Newspoll: More Off-The-Scale Leader Ratings

Newspoll has returned with a second round of the very welcome State Premier approval ratings first seen in late April.  I thought a brief (by my standards) post about the current round of Newspolls was worth putting up overnight as the results are already sparking discussion in Tasmanian politics.

In each case I give the Premier's net rating, followed by the change from April, followed by the satisfied and dissatisfied split.

Gladys Berejiklian (NSW) is on +42 (-4) (68-26)
Daniel Andrews (Vic) is on +40 (-18) (67-27)
Annastacia Palaszczuk (Qld) is on +24 (+8) (59-35)
Mark McGowan (WA) is on +79 (-4) (88-9)
Steven Marshall (SA) is on +52 (+5) (72-20)
Peter Gutwein (Tas) is on +82 (+9) (90-8)

And just for completeness, Scott Morrison (PM) is on +41 (+4) (68-27).

I noted in April that McGowan's satisfaction rating (or "approval rating") of 89% was then not only the highest in Newspoll history, but also the highest I could find in any scientific opinion poll in Australia ever.  (In calling a poll "scientific" I don't necessarily mean that it is accurate, I just mean that it tries to follow proper sampling and scaling procedures and isn't an unweighted opt-in.)  However records are made to be broken, and McGowan's 89% has been broken by Peter Gutwein who, in this sample, hits 90%.  McGowan, however, retains the record for the highest ever net satisfaction rating, with his 83% from last time one point ahead of Gutwein's 82%. 

Of course, with samples of only a few hundred voters per state, these ratings are more than a little bit rubbery, and the ratings of McGowan and Gutwein are not significantly different from each other.  The real margins of error on these ratings are likely to be larger than those quoted by Newspoll (see graphic) but it really doesn't matter whether a given Premier is on +70 or +80 at this point of time.  In contrast, state voting intention samples that were this small would be very unreliable for the purposes of predicting an election, and this is probably why only personal ratings have been sampled.

Andrews' Moderate Loss Of Support

Daniel Andrews is still a very popular Premier, but in this poll he drops from third to sixth in the list of leader ratings, dropping from a net rating of +58 to a less remarkable +40.  Prior to this poll I was especially interested in how much damage Andrews' rating would sustain and to what extent it might be caused by COVID-19 issues as opposed to the Adem Somyurek scandal that has deprived the Premier of three ministers.

The answer seems to be that it is all down to perceptions of Andrews' handling of COVID-19.  Newspoll has very usefully asked a separate question on this subject.  The following graph shows the relationship between Newspoll ratings for the leaders' handling of COVID-19 and their overall ratings, for the April and June polls combined.


The red dots are rating combinations for Andrews and the blue dots are for the other Premiers.  Andrews' April ratings fall exactly on the trendline; his June ratings are the biggest outlier from it.  That is, it appears that Andrews is actually polling much better personally than voter perceptions of his handling of COVID-19 imply - far from being damaged further by the Somyurek scandal, he's more likely being cut slack for his performance as Premier to date.  (This should be treated with a little caution as the only other ratings at that end of the COVID-19 handling axis are for Palaszczuk).

Early Tasmanian Election?

Speculation about an early Tasmanian election has been bubbling along at a low level among Tasmanian politics junkies in recent months, and started firing up immediately when news broke about Gutwein's record rating.  Tasmania is not due for an election until March 2022, and is the one remaining state without "fixed" terms.  The state Labor Opposition has adopted a confused series of positions over COVID-19 and the best strategy it seems to have come up with is to work out what the government is going to do next and then call for it before they do it.  At times, senior members have not seemed to have their ducks in a line in communicating the same messages about COVID-19 (especially on border closures) to the media.

Tasmania has not seen an unforced election as early as the one proposed since 1979 when Doug Lowe (ALP) went a year and a half early in search of a larger majority than the one-seat majority left to him by previous Premier Bill Neilson.  It worked - Lowe won in a relative landslide by Hare-Clark standards - but those were simpler days when major parties only had to beat the other major party.

This government, like Lowe's, would have some valid reasons to sneak off to an early election.  The Lower House is not entirely stable, though the Government hasn't lost a vote since ex-Labor independent Madeleine Ogilvie rejoined the house on a recount and promptly started voting mostly with the Liberals.  Under normal circumstances it will only lose votes if all of Labor, the Greens, Ogilvie and its own periodically renegade Speaker Sue Hickey vote together, something they are yet to do.  An early election might create clear air while the government grapples with COVID-19 economic challenges over the next few years, and would give the government a clear mandate to spend massively on state-building programs.   What it won't do is get rid of the Legislative Council where the left has a majority.

However, an early election would carry risk.  Voters sometimes punish governments that go early unnecessarily, and might prefer to just relax and not think about politics too much at the moment.  Majorities are so difficult to guarantee in Hare-Clark and a bad bounce of the ball might result in the government being suddenly out of office (as was Robin Gray's government despite polling 47% of the vote in 1989).  A sudden resurgence of COVID-19 as is occurring in Victoria could also be a disaster on the campaign trail, especially if the point was rammed home that voters were being exposed to unnecessary risks.

The government has the good fortune that on August 1 there is a Legislative Council election for the seat of Rosevears where it is running a candidate.  However LegCo elections are challenging to interpret at the best of times, and in this case the Liberal candidate is a household name who was widely expected to romp in anyway.  A very high vote indeed for Jo Palmer (an outright majority or nearly so) would be needed before one would take the election as a pointer to Gutwein's popularity converting to votes.  It will be interesting to see if any more Tasmanian polling comes out soon.

Federal Newspoll

This week's federal Newspoll saw very little change, with the 2PP unchanged at 51-49 to Coalition, Scott Morrison up four points on net satisfaction to a personal high of +41, Anthony Albanese down one to +2, and Morrison's lead on the skewed (to incumbents) Better PM metric up two to 32 points (58-26).  As my recent technical article notes, neither Better PM nor PM approval ratings seem to add any predictive value at election time, so they are better seen as figures that help explain voting intention, except for times like this when they don't.

Anyone looking for a signal of a national mood that might influence the Eden-Monaro by-election (in which pre-day voting is well underway) would have been sorely disappointed.  The main takeaway is that to this point nothing has dented the PM's approval in what seems to be a good time to be in government and a bad time to be in opposition at any level.

I am keeping an eye on the volatility (or lack thereof) of the new version of Newspoll since it commenced late last year.  The version running from 2015 to 2019 was especially "under-dispersed" (more prone to repeat similar values from poll to poll than would be expected for its sample size).  It is far too early to draw robust conclusions about the new version, but should the pattern of very little poll to poll 2PP change continue for long I will be exploring this issue in detail.

It's Time To Go, Simon ...

It is really past time for The Australian to find (or return to) someone who can write about polls in a mathematically informed fashion.  Simon Benson's output is embarrasing, it's just wrong for Australia.  In the latest howler, he claimed Scott Morrison was nearly twice as popular as Annastacia Palaszczuk in Queensland because his net rating was twice as high.  As William Bowe pointed out such a comparison would run into problems if the leaders had netsats hovering around zero.  Indeed if Morrison had a netsat of +5 and Palaszczuk's was -10, would Benson report that Morrison was minus half as popular as her?

The day before, he continued to compare Morrison's approval ratings to early-period Kevin Rudd's, while ignoring the methods change in late 2019 that has meant they are no longer comparable.  (Using positive satisfaction as the yardstick, Morrison is the most popular Premier over a two-month period since the early months of Rudd, but using the better yardstick of net satisfaction, Rudd was up there in early 2009 too.)

Really if the Oz can't do better than Benson it would be better off just publishing the figures and outsourcing the commentary to members of the blogging psephosphere.  Those of us with fewer  moral scruples than the others would do the same thing much better and probably more cheaply.  After all, we understand Newspoll because we don't own it.

PS: Morgan

There was also a Morgan poll but Morgan's data-reporting practices are so horrible I'm finding it difficult to regard them as useful data.  Twitter thread here.  

Thursday, June 25, 2020

White Goes First, Right Goes Beatup: The ABC Did Not Attempt To Cancel Chess

In recent days I've been involved in a media and social media flurry sparked by the ABC's decision to explore the subject of whether White moving first in a game of chess was in any way connected to race issues.  This claim was once most commonly seen as a spoof of anti-racism campaigns, but these days, a small number of people seem to be actually fearing chess might be symbolically racist.

I appeared on ABC radio and gave an interview that outlined that there is no evidence this is the case.  The host did not try to argue that there was, just mentioned that people on social media have held concerns about the issue.  The mere existence of that interview has triggered a massive backlash from right-wing culture warriors, which had already started before the interview aired.  The thing is, it is unclear that the enemy they're tilting at exists!  The ABC may be guilty of filling up its programs with offbeat fluff on the slender pretext of a few tweets, but that does not mean it was trying to have chess cancelled.

Saturday, June 13, 2020

Legislative Council Voting Patterns 2016-2020

Advance summary:

1. This article presents a revised analysis of voting patterns in the Legislative Council (the upper house of Tasmanian Parliament) based on contested divisions involving the current MLCs in the last four years.

2. Although there is a degree of independence in all Legislative Council voting (except among caucusing party MLCs), the Council continues to have a clearly defined "left wing" consisting of the four Labor Party MLCs, and independents Mike Gaffney, Ruth Forrest, Kerry Finch, Rob Valentine and Meg Webb.

3. The two Liberal MLCs and independents Ivan Dean and Robert Armstrong belong to a similarly clearly defined "right" cluster.  Independents Tania Rattray and Rosemary Armitage do not belong to any cluster but currently side somewhat more with the right cluster than the left cluster.  

4. A possible left-to-right sort of the Council is Webb, Valentine, Forrest, the four Labor MLCs (Farrell, Lovell, Siejka and Willie in no particular order), Gaffney, Finch, Armitage, Rattray, the two Liberal MLCs (Hiscutt and Howlett in no particular order), Armstrong, Dean.  However Webb's placement is unreliable because of limited evidence.  

5. Going into the 2020 elections, the left holds an absolute majority in the Legislative Council, normally meaning that the government needs the support of Labor or at least two left independents to win votes.  This will remain the case, the question being the size of that majority.

---------------------------------------------------------------------------------------------------------------

There has been excellent news for election junkies and for Tasmanian democracy more generally, with the announcement that the 2020 Legislative Council elections for the divisions of Huon and Rosevears will be held on August 1.  These elections were originally scheduled for the start of May, then postponed til the end of May, then postponed to June, July or August and then the subject of legislation that could in theory have seen them held next year.  Fortunately COVID-19 has apparently been kickbooted (someone please tell me where I found that word) from the state as of June 12 and it is great to have these elections back on track.

My original guides for these seats can be found here:

Huon
Rosevears

and in the next few weeks I will decide whether to relaunch them or just keep rewriting them to clean up the mess as best I can.  Huon sees the first seat defence by conservative independent and former local mayor Robert Armstrong, while Rosevears is a vacancy caused by the retirement of three-term left-wing independent Kerry Finch.  The government has high hopes of improving its situation by winning Rosevears, where its candidate is a very high profile newsreader, Jo Palmer.  As this article shows, the government's current situation upstairs is difficult.  COVID-19 has seen a great reduction in contested votes and politics as normal, but they'll be back, quite probably very soon.

In most years as a curtain-raiser to the LegCo elections, I put out an article about voting behaviour in the chamber over the last four years.  Last year I did not do this because there had been very few contested votes since the previous year, and the first half of last year was insanely busy.  The most recent prior version was the 2018 version (which links back to earlier articles).  As of 2018 the chamber had eight left-wing MLCs (including four Labor), four right-wing MLCs (including one Liberal), two independents who were centrist to centre-right, and a President who was notionally fairly conservative but never voted.

Since 2018 we've seen one right-wing independent retire and be effectively replaced by a Liberal, and the incumbent President retire and be replaced by an independent who had run a left-wing campaign.  The latter setback for the Liberals was in theory offset by Labor's Craig Farrell taking over as President, but Farrell has in fact said he will use his casting vote along party lines if necessary.  (So far this has yet to be tested.)  So the question is, have changes in the voting behaviour of the existing MLCs helped the government with an upper chamber seemingly stacked against it?  The answer is, not that much.

Methods

I've continued to use the same methods as in the 2018 article, but a few comments are in order.  Firstly I've continued to treat the four Labor MLCs as one entity.  While, in theory, a conscience vote could arise where they might not all vote together, that so far hasn't happened.  I can also now treat the two Liberals the same way.  When there were previously two Liberals in the Council, at times they would vote differently, but this has yet to occur since Leonie Hiscutt was joined by Jane Howlett.

Instead of weighting the figures by agreement with different Labor or Liberal MLCs, which would create a bias towards recent data, I've simply extracted a "Labor" position from Josh Willie's votes, or where Willie was absent, Craig Farrell's.  For the Liberal side I've used Leonie Hiscutt's votes since Hiscutt has not missed any division attended by Howlett.

I again use only the last four years of data and I only use recorded divisions with at least two votes (including pairs for absence) on either side. Where an issue generates multiple divisions on the same day, I ignore any of the divisions that are repeats of the earlier ones.  Where an issue generates identical divisions across different days, I count both.  However, in this term, there was one issue (marriage law amendments relating to transgender issues) that accounted for 12 of the 88 contested divisions, including six different division patterns on the one day.  All these different patterns had the government plus Ivan Dean on one side and Labor plus all the left indies who were present on the other.  The differences between the votes came from various combinations of Robert Armstrong, Rosemary Armitage and Tania Rattray, and also from absences.

I've added 36 contested divisions since the last report.  Of these, the marriage law amendments issue was the only one to generate more than two votes.  The rest were quite eclectic and did not include a lot of obvious partisan culture-war issues.  I do not treat substantive and procedural issues separately as there are relatively few purely procedural votes and I believe that frequently the reasons for procedural motions are political anyway.

Agreement matrix and left-right sort

This chart shows how often the Legislative Councillors agree with each other on contested votes.  For instance, the chart shows that Forrest and Rattray currently agree 46% of the time.  As usual I've highlighted agreement scores over three-quarters in red and dark blue, and scores close to that mark are highlighted in orange and pale blue.  Given that I've removed all the 100% agreement scores for party MPs with their own party, the highest remaining agreement scores are 85% for Forrest/Webb, 85% for Armstrong/Dean, 82% for Forrest/Labor and 81% for Gaffney/Finch.  The lowest are 13% for Webb/Armstrong, 18% for Valentine/Armstrong, 19% for Webb/Liberal and 20% for Gaffney/Dean.  However, Meg Webb has only attended 16 contested votes, 4 of which were on the same issue, so there is not really that much data for her yet and these results may be unreliable.  (Indeed Armstrong was absent on a vote where Webb sided with Dean and Rattray, so might well have voted with her on that had he been present.)

The "score" figure is an indicator of how strongly each MLC falls on one side of the left-right divide or the other, red/orange for left and light/dark blue for right.  The score figure is the average agreement with whichever cluster the member agrees with most (counting Labor and Liberal each as one) divided by the average agreement for the other.  The higher, the stronger the pattern. 


(Note and see comments: "strongly" doesn't mean extreme; it just means there is clear evidence that an MLC is very much on one side rather than the other of this chamber, which seems to run from slightly to the left of Labor to slightly to the right of the Liberals, with no far-left or far-right elements.)

The matrix shows two obvious clusters.  The left cluster consists of Labor and five independents - most pairs from this group agree with each other on more than 75% of contested divisions, and the rest are close.  The right cluster consists of the Liberals, Dean and Armstrong.  Armitage doesn't have a strong tendency to agree or disagree with anyone, and Rattray's highest agreement score is 72% with Armstrong.  Both Armitage and Rattray in the last four years have voted substantially more often with the right cluster than the left cluster, which hasn't always been the case in the past.

As noted above there really isn't enough data to place Webb reliably yet so her ranking to the left of Valentine and Forrest should be treated with a lot of caution.  Nonetheless I expected that Webb's voting pattern would be similar to Forrest.  Gaffney and Finch, while clearly part of the left cluster, are more likely to agree with the conservatives than the other seven left MLCs are.  On the right side, both Armstrong and Dean place slightly to the right of the Liberal Party.

In the past I have posted a 2D principal components analysis graph of the above but this time I made it then decided that it didn't actually say anything, so sorry for those looking for some image content, but I haven't posted it.  The left-right axis explains a massive 82% of the variation in voting patterns.  The second axis just picks up the most unaligned voter it can find (Armitage).   Possibly voting agreement percentages wouldn't show it anyway, but there's no evidence of there being a particular kind of commonly-occurring issue that breaks the left-right stereotypes in a particular way,  Rather there are a lot of different issues that do so, each in their own individual way.

The presence of nine left-wing MLCs means that the government doesn't get its way on contentious matters very often.  Since Labor won its fourth seat the government has prevailed on just 13 out of 42 contested divisions.  Even after removing the marriage law amendments, it is still only striking at 43%.  The Labor block of four makes Labor very powerful in the Council, because cases of the government prevailing with Labor voting against it are so rare.  Since Labor won its fourth seat the government has won divisions with Labor on the other side just four times:

* A division on sentencing in 2017, supported by all the independents then in the Council except Valentine and Forrest.
* A vote on bikie insignia in 2018, supported by all the independents then in the Council except Gaffney and Valentine.
* A vote on a work cover amendment in 2019, supported by the right independents and Gaffney with Forrest absent
* A vote on a burial and cremation bill in 2019, supported by the right independents and Finch and Webb.

Slim pickings indeed, but the good news for the government is that, unlike the independents, Labor can sometimes be wedged into supporting things.

Because the left holds nine seats in total, even a government win in Rosevears and an Armstrong retain in Huon would still leave the left with an 8-7 majority.  However that would improve the government's prospects, because it would give it seven votes on the floor whenever it had the support of all the right-wing and centre-right independents.  That would then enable it to win votes if just one left independent voted with it, or even if one of the left MLCs was absent and not paired.  In the latter scenario, the government would win 7-6 on the floor, and President Farrell's potential willingness to vote against it would be irrelevant as the casting vote is used only to break ties, not to create them.  While the current elections are significant in terms of how difficult working with the Council is for the current government, they will not change the fact that it is difficult.

There is one more Legislative Council session before the election and I may update this analysis after it if anything changes.


Sunday, June 7, 2020

Unpopular State Premiers Still Have Dire Historic Fates

It's been over a month since I posted a new article on this page, though updates to previous articles have continued, especially Eden-Monaro.  I have some vague idea where that time went (a number of distractions from psephology lately) but there hasn't been a huge amount going on lately and I tend not to write just for the sake of having something up.  There will always be something new here eventually!

This article is another piece where I update a previously published article from some time ago and see whether the pattern described in it is still holding up.  Today's target for an update is Unpopular State Premiers Have Dire Historic Fates, from 2013.  This article was inspired by a bad Newspoll for then Western Australian Premier Colin Barnett.  Barnett had been re-elected with a 57.3% 2PP nine months earlier so it probably seemed adventurous to see a single Newspoll still showing his government in a narrow lead as the first of the circling vultures.  But it was - Barnett survived a leadership challenge in 2016 but was dumped by the voters in 2017 with an enormous 12.8% swing.

He wasn't alone.  Since I released the original article, Campbell Newman was dumped by voters with a massive swing, as was Lara Giddings. Jay Weatherill also lost (albeit with a 2PP swing to him) and Mike Baird, who had been very popular in his first term, became somewhat unpopular in his second and resigned.  The four election defeats for unpopular Premiers helped beef up the evidence that it is the voters, and not just the parties, who tend to show them the door.  In the same time, Premiers who had not polled such bad ratings in their terms were re-elected twice in NSW and once each in Queensland, SA, Victoria and Tasmania, with Victoria's Dennis Napthine (worst netsat -4) the sole casualty to not poll a bad rating.  The chart below (click for larger clearer version) shows the fates of every state Premier who has polled a netsat worse than -10 in Newspoll history (which starts in 1985).  Premiers are sorted by the worst netsat they polled during the term.