Thursday, October 27, 2016

Wilderness Society Tasmanian ReachTEL

Yesterday the Tasmanian Government announced a substantial restructure of the government business entity Forestry Tasmania, which recently announced another financial loss.  FT will be renamed "Sustainable Timbers Tasmania" in what might be seen as furthering a pitch for certification and a positive image, but can also be seen as a deliberate trolling riposte to the rampant misuse of the word "sustainable" by environmentalists in recent decades. (I strongly approve of the renaming for the latter reason.)

The business will be downsized (again) and attempts will be made to bring forward logging in land originally reserved under the now defunct "forests peace deal".  The latter change came as a complete and not entirely welcome surprise to some within the industry.  The changes are being seen as making the forests industry a major issue at the 2018 state election. If passed, they will do nothing to end the great Tasmanian sport of Ritual Forest Conflict between pro- and anti-forestry campaigners, and will provide the government with ample fodder for trying to wedge the ALP based on its record when in government. But first we will have to see if the government can get the reopening of deferred land through a now finely balanced Legislative Council, or if this is yet another episode of forestry-related culture-warring for show but not for result.

Yesterday morning the Wilderness Society put out a press release spruiking a poll (PDF link) it had conducted on Monday night.  I must commend the Society for promptly releasing the full results including all questions asked.  One really cannot ask for more from a commissioning agency.  However, it does appear that the Society has gone off half-cocked and been outmaneuvered as a result.  The poll was released on the morning of Resources Minister Guy Barnett's announcement of the planned restructure, which goes further than the poll design seemed to anticipate.  Also, as usual with commissioned issue polls, the issue questions have major design problems, and the interpreter can hence make of the results pretty much anything they like.

Voting Intentions

The press release claims the Hodgman Government has "dropped 10 per cent of its vote in 18 months".  Literally, 10% of the Hodgman Government's vote would be five points, but I assume the release really means to say it has dropped ten points, based on the 41.6% primary as compared to the Liberals' 51.2% at the last election.

In this form, the claim is incorrect, and the reason it is incorrect is that it is standard practice to compute a party's vote share with undecided voters reallocated rather than just listed as "undecided". Polls that ask the undecided voter which way they are leaning and publish those results are especially useful for this purpose.  A voter who is undecided when first prompted but leaning to the Liberals is not a lost vote and would probably in fact vote Liberal if an election was held right now.  

With the undecided redistributed (to form a standard headline figure comparable to the way that Newspoll, for instance, reports its figures) the results are Liberal 44.2, Labor 29.5, Green 14 Other 12.2, a loss of seven points rather than ten.  Moreover, most of the "loss" compared to the election result is to the Others column, where a lot of it would either scatter harmlessly and exhaust or prove to be vote-parking on the day.  So this is actually one of the better figures for the government for a while.

It's difficult to benchmark a Tasmanian state ReachTEL though.  At both the 2013 and 2016 federal elections, Tasmanian ReachTELs were much more Coalition-friendly than the actual result. But at the 2014 Tasmanian state election, the final two ReachTELs both underestimated the government's enormous winning margin.

If we try to project the results by electorate - a difficult task without any recent state electorate breakdowns - the Liberals would probably win either 12 or 13 seats on these numbers.  The critical electorate would (as it usually is in my projections) be Lyons, where there might be a three-party scramble for two seats that would be resolved by the preferences of minor parties and the spread of votes within the Liberal ticket.  At the 2014 election, the Lyons Liberals (Hidding, Barnett and Shelton) had a very even spread of votes and if this continued the party would be well placed to hold three.  But I am not sure this balance within the ticket will continue.  

If Labor indeed polled as poorly as this poll implied, the ten-seat result they have been on for in most recent polling would be in jeopardy.  They might not reclaim a Bass seat from the Greens, and it would also be possible for the fourth seat the Liberals were extremely lucky to win in Braddon last time to be claimed by an Independent rather than the ALP.   

I do not actually believe that Labor is travelling as poorly as this sample implies - it is not consistent with recent strong results for the party in the state at the federal and Legislative Council elections.  A sub-30 reading for Labor, after the removal of undecided voters, seems just too low, and raises the question of whether the same skew seen at the federal election is on show here.  Or it might just be an iffy sample.

Overall, the most likely result if these numbers did show up on polling day would be around 13-9-3.  

State Polling Aggregate

I have treated the ReachTEL with some caution because it is a commissioned poll, but the voting intention figures were not the Wilderness Society's reason for releasing the poll, so I am confident they are not prone to "selective release".  I have added the ReachTEL to my state polling aggregate with a weighting of 20% and also another recent Morgan SMS sample - a series I do not much trust -  with a weighting of 3%.  (The Morgan sample - bearing in mind that this series skews heavily to the Greens and against the Liberals - had figures of Liberal 39 Labor 33 Green 16 Others 12, with half of the 12% others being for Jacqui Lambie Network, which is not even known to be contesting state elections.)

The following is my current aggregate of Tasmanian state polling, bearing in mind again that the electorate-by-electorate projections are very rubbery:

In this version of the aggregate, the big uncertainty is whether the Liberals would get an even enough split between their candidates to beat the Greens to the last seat in Lyons, with the assistance of minor-candidate preferencing and leakage from the Greens' minor candidates.  The recent ACT election provided more findings showing how easily this sort of thing can occur, so I have left the seat uncertain, but note that the Greens are also not that far from taking the third Liberal seat away in Bass on these numbers.  Therefore I have left the sidebar aggregate unchanged at 12-10-3 for the time being.

Forestry Questions

The Wilderness Society have correctly noted that only a quarter of their sample chose "undertaking more logging by allowing logging in forests that were set aside for conservation under the Tasmanian Forestry Agreement".  They have sought to portray this as evidence that the voters don't agree with the Government's plan. 

Firstly, the Government has gleefully jumped on the fact that opening up the formerly reserved areas was only part of the plan, and that another part is amply described by the option "maintain current supply and ensure log sale prices cover the cost of production".  The option's reference to "undertaking more logging" is also disputable because the stated rationale for reopening areas is not to increase logging volumes but to ensure that contractual obligations are met.  Finally the question's reference to "forests that were set aside for conservation under the Tasmanian Forestry Agreement" is misleading. It tells one side of the story by failing to note that the TFA (aka the "forests peace deal") is defunct and that those areas are in fact Future Potential Production Forest which, all else being equal, would be available for logging from 2020.  These forests were once set aside for conservation by Labor and the Greens, but their current status is best described as under moratorium.

The basic problem with using this question to argue for public rejection of the government's intended direction for forestry is the assumption that the government can't walk and chew gum.  A voter might agree with elements of one option and also with elements of another and hence be receptive to a plan including elements of both.  Seen this way, question 3 can just as easily be spun as showing majority support for a plan including increasing log prices and reopening the FPPF areas immediately.  Whether this is actually a good or an achievable plan is another question entirely.

Question 4 refers to whether the Government has a clear plan for the industry, as opposed (?) to being incapable of addressing the problems in the industry, or being more interested in attacking Labor and the Greens.  Naturally Labor and Greens supporters almost entirely take one or the other of the negative options while Liberal supporters mostly take the positive option.  Others voters tend on the whole to take a negative view of either the government's intention or its prospects, which is interesting given that Others preferences were quite strong for the government at the last election.  

A striking finding is that voters who are strongly committed to their current voting intention are much more likely to believe the government has a clear plan (48.6%) than those who are uncommitted (21.1%).  This relationship between commitment to vote and support for a particular party's handling of an issue is remarkably strong, even given that left-wing voters are often less committed to a particular party.  It might be inferred from this that some of the cynicism about the government's intentions on this issue taps in to a general disillusionment with Tasmanian politics, that is also evident in the high Others vote even when there are no known prominent Others to vote for.

The biggest problem with Question 4 is that if you were trying to pick the silliest possible time to poll whether the Government had a plan for something, you would probably pick a few days before their plan was announced.  This is exactly when the poll was conducted, so it would be more useful to do the poll, say, next week, rather than on the Monday just gone.

Finally, concerning this article, the usual suspects are directed to the disclosure statement.

Tuesday, October 25, 2016

Poll Roundup: The Turnbull Slide Continues

2PP Aggregate: 51.8 to ALP (-0.1 since last week)
Labor would probably win election "held now" with a small majority

Almost four months since its re-election, the Turnbull coalition government continues to poll ordinarily.  It has still not led in a single poll by anyone, and only now are there some signs that Labor's lead might be stabilising.  With a new bad news cycle hitting the government, oh, about twice a day on average lately, there might even be relief it is not worse ... yet.  An implicit stoush between PM Turnbull and Tony Abbott over the history of rapid-fire shotgun negotiations might have looked like fodder for further movement, but Newspoll suggests it was seen as just beltway stuff.

Most of our polling still comes just from Essential and Newspoll though this week a third player, Morgan, dipped a very tentative toe in the water with one of its irregular phone polls.  The only voting intention result released was 55:45 to ALP, easily Labor's biggest lead from anyone so far.  Given the small sample size and the lack of adequate documentation of 2PP method or primary votes, this only counted for 10% of a normal poll in my aggregate, but even that made some difference.

The last two Newspolls have been both 52:48 to Labor.  The last few Essentials were 52, 53 and 52, but the last 52 looks like it was very close to being a 53.  I aggregated the Newspolls at 51.9 and 52, and the Essentials currently in the frame (which is this weeks and the week before lasts) slotted in at 52.2 and 52.4.  (Last week's was 53.4, and it will count towards next week's reading - see the methods page for how this works.)  I should note that The Australian helpfully published finer breakdowns of the Newspoll votes for One Nation and NXT, so I have back-calculated those Newspolls for which that data is available, though in most cases that made no different.

While I have the Coalition up from 48.1 to 48.2 on the single-week reading (assuming that there aren't any more polls this week), the smoothed aggregate graph is a one-way street.  It's extremely dull viewing except for ardent ALP supporters, and it looks like this:

Still, early days ...

Turnbull Is Less Popular Than Trump?

One Ben Cantwell on Twitter has just picked up the amusing statistic that Malcolm Turnbull's Newspoll netsat of -28 (29-57) is worse than Donald Trump's Real Clear Politics net favourability aggregate (currently at -25.2).  This isn't really comparing oranges with oranges, firstly since favourability is a lower bar than satisfaction, and secondly since Newspoll is among the harsher measures of Australian leader popularity. (In BludgerTrack, Turnbull's aggregated netsat across all polls is only -16.4, though it is bound to take a hit when this latest Newspoll is added.) You can find polls where Trump is doing much worse than -28, but still, the comparison is a surprising one.   Very few recent polls have not had Trump's favourability higher than Turnbull's current Newspoll satisfaction rating of a feeble 29%.

That 29% is more notable because it is lower than the 30% satisfaction rating Tony Abbott had when he was dumped (although Abbott's net rating, -33, was worse than Turnbull's present -28).  Indeed Abbott only polled three sub-29% satisfaction ratings in his whole career as PM.  These came in a row in early 2015 following the Prince Philip knighthood, the Queensland election defeat and the spill motion against Abbott's leadership.  For comparison, 27% were satisfied with Bob Hawke just before he was finally removed by his own party, 36% with Kevin Rudd, and 28% with Julia Gillard.

Bizarrely, we are starting to see some hemi-demi-semi-serious discussion about bringing back Tony Abbott!  Let's not forget that it was only by listening to the electorate for once and throwing Abbott under the bus that the Coalition got itself re-elected at all.  An Abbott revival would be history repeating as farce and the ultimate debasement of the idea that party leadership at election time means anything at all.  Abbott is not Kevin Rudd the second; voters never understood why Rudd was removed in the first place and so the hankering for his return was natural as a matter of fairness if nothing else.

The Silver Lining

There is something interesting developing in the relationship between Turnbull's netsat and the Coalition's 2PP.  Falls in Turnbull's netsat hurt the Coalition much less than the same falls for any previous Prime Minister of the Newspoll era.  On average, Turnbull has to lose 15.7 netsat points for the Coalition to drop one 2PP point.  This compares with 7.0 netsat points for Abbott, 6.2 for Gillard, 8.6 for Rudd (first time), 6.4 for Howard, 5.8 for Keating and 6.0 for Hawke.  The average 2PP reading for a party whose leader is rating this badly is a mere 46.3%.

There are a few possibilities, and we'll probably need to wait til ReachTEL comes back to find out which one is true.  One is that a lot of people are just mildly dissatisfied with Turnbull, with very few strongly disliking him.  A second is that a lot of people really dislike Turnbull, but do so from the Right and therefore aren't open to voting Labor.  A third is that a lot of people are really fed up with the PM, and would in theory be open to voting Labor, except they are even more fed up with Labor.  I think it's more the first two than the third, as the statistic I call the "disconnect index" (the minimum proportion of voters who must be preferring a party while being dissatisfied with its leader) is at modest levels by the standards of the last few years, though it is starting to creep up.

Meanwhile, Bill Shorten is on a net -15 (36-51), his ratings having barely moved since June, and Turnbull is rated "better PM" 42-32, his closest lead over Shorten so far.  Last fortnight's better PM reading (45-30 to Turnbull) was also interesting, because only once had a PM with such a lousy netsat held a better PM lead so large.  Paul Keating (then -27) had a 24-point better PM lead over Alexander Downer in December 1994.

Essential actually had Turnbull rebounding to a mere -3 net approval this month (38-41, showing how different Essential and Newspoll scales are, or how different the Essential voter panel is) with Shorten the same (37-40), and Turnbull ahead 41-28 as better PM.

Other Polling

As usual, there are dozens of issue questions over at Essential, and those interested can read them all over there.  Here are some that I found interesting:

* Labor, Coalition and Others voters rate "ISIS and Islamic radicalisation" the biggest threat to global peace and security in a field also including global warming, Russian aggression, inequality, overpopulation and Trump.  For Greens voters, ISIS and the Islamists come in third behind Trump and other forms of global hot air.

* While many voters have no view, those who do have a view support reestablishing the ABCC, 36-16.

* The Turnbull Government has a net -27 confidence rating for its ability to get stuff done "given the Senate".  The question design does seem prone to invite agreement compared to, say, one that also mentioned that governments can get some things done by gazetting them without Senate approval.

* Australia is quaking in its boots about the small risk of Donald Trump becoming President.  This is not unusual; previous polls (including pre-Trump) have found that Australian Liberal voters are about as pro-Democrat as Australian ALP voters.

* 61% of the panel thinks multiculturalism has been a net positive for Australia; 23% disagree. Predictably Others are most likely to whine.  24% think a migrant family should be rejectable purely based on their religion, and 56% disagree.  (The question "which religion" might have been an interesting followup to see if the concern was purely about Islam).

* Essential again found that voters overestimate how many Muslims there are in Australia, though awareness of the real number has improved.  Information about the actual number makes little difference to concern levels, which drop from 53-42 to 47-47 post priming.

There has been a general run of rather xenophobic results from the Essential panel, and the Morgan phone poll provides quite a different view.  However, the difference is probably caused directly by the polling method.  People are much less likely to express xenophobic sentiments in a live phone poll because they do not want the interviewer to think they're a racist.

Those Fickle Millennials

It was only about six weeks ago that I realised "millennials" are more or less the same people that not so long ago were called "Generation Y".  This probably shows about how much attention I pay to generational stereotyping and how little regard I hold it in.  Anyway, the relevance of said cohort to this website can be found in an AFR article which referred to the difficulty of connecting with millenials politically.   The claim is that those born between the early 80s and the very late 90s have no brand loyalty and churn through products, and that the same thing applies to how they interact with politics.

I have a different view, one that is closer to that of Labor's George Wright than of anyone else quoted in the article.  You don't get the votes of younger voters if you don't appeal to them, and all generations are like that. The millennial cohort - together with various others - discards political products so quickly not because they are compulsive churners, but because the political products are crap.

Saturday, October 15, 2016

ACT Election Live and Post-Count Comments

Labor minority government returned - ALP 12 Liberal 11 Green 2.  
This was also the interim outcome

Postcount Comments

(NB Through the week I will be posting comments here only in the evening, because of work commitments)

What happens now is that votes are gradually entered into the computer and a series of interim preference distributions are released, leading up to the final "button press" which is expected to happen next Saturday.  As we get closer to the full count these should become more accurate as an indication of what will happen, but for the time being it is useful to keep an eye on differences between the current primary count (which is based on more votes) and the primary count for the interim distributions, and also to bear in mind that the counting of postal votes may favour the Liberals.

Saturday: It's over!  In Brindabella Nicole Lawder (Lib) has won the final seat from Labor's Angie Drake by 553 votes.  Drake outlasted Steven Bailey by 307 votes at the second-last exclusion, but had Bailey survived that exclusion he would have lost anyway, probably by more than Drake did.  In Ginninderra the Ginninderra Effect has indeed struck again - Labor with 2.74 quotas at the final exclusion have won three seats while the Greens with 0.78 at the same point have missed out.  This is because two of the Labor candidates had a very evenly split share of their party's votes, and indeed it wasn't all that close between them and the Greens (701 votes for Tara Cheyne over Esguerra with Gordon Ramsay actually ahead of Cheyne at that point).  In Murrumbidgee the Greens are up by 800.  The other two electorates were decided by thousands.

How To Best Use Your Vote In ACT (and Tasmanian) Elections

This is a quick note because there are some incorrect myths doing the rounds about how people should vote in the ACT election today.  I was asked to write this piece because there have been some claims that voters who vote for independents should refuse to preference anyone but independents.  A claim to this effect was recently published in the Canberra Times in an otherwise generally good article.  It happens that the same advice I'm giving here for the ACT also applies to Tasmanian lower house elections, because both use the Hare-Clark system, albeit with some minor differences.

My basic recommendation for both ACT and Tasmanian elections is simple: If you want to make your vote as powerful as possible, number every square.  Don't even leave the final box blank.  If you don't feel you can number all the squares, you should number as many as you can.

This is the key incorrect claim from the Canberra Times article:

Tuesday, October 4, 2016

Recent NSW State Polling

The purposes of this post are: firstly, to add some context to the dramatic findings of a recent NSW state Newspoll, and secondly, to link to the full results of some late August NSW state union polling that may be of some interest to somebody out there.  I have actually had the latter for a few weeks but because I have been overseas and then very busy it has taken this long to do anything with it.

Newspoll record?  Well, not really ...

The recent Newspoll has found that Mike Baird's status as Australia's most popular Premier by far has come to a sudden halt after nearly two years of stellar ratings.  This finding was also foreshadowed by a slightly earlier Fairfax ReachTEL which showed Luke Foley ahead of Baird as Preferred Premier.   The latter result, while still striking, was not quite the sensation it appeared to be: the forced-choice method used by ReachTEL does not advantage incumbents in the way that the method used by Newspoll, Galaxy, Ipsos and Essential does.  So an Opposition Leader being preferred Premier in a ReachTEL poll usually just means the two-party race is close and the Premier is a little bit under the weather.                                    

The Newspoll found Mike Baird's net satisfaction rating has crashed from +39 (61-22) in January to -7 (39-46) now.   The Australian listed this at the top of a list of the greatest netsat falls by a Premier in Newspoll history.  In a poll-to-poll sense this is true, but it is a misleading statistic as the intervals between two consecutive Newspolls in a given state historically have been anything from a few days in some cases to four years in others.  In this case, eight and a half months is an unusually long interval, especially when the old Newspoll often routinely polled at intervals of two months.

Wednesday, September 28, 2016

Poll Roundup: Bad Start For The Second Turnbull Government

It's early days, but here we go again ...
2PP Aggregate: 51.5 to Labor (+0.6)
Labor has led since aggregate resumed

Welcome back to Poll Roundup, a series which looks at the aggregated state of the federal polls, leadership ratings and polling snippets of interest.  Nearly three months after a narrow victory in the 2016 federal election, the Turnbull government's progress has now been measured in three Newspolls and eleven weekly Essential Research readings.  It's not a great start for the returned regime.  It hasn't been ahead in even one of those, and is now clearly behind.  Indeed, when the primaries from the early polls are converted using 2016 preferences, a case can be made that Labor has been leading in the lot.

If we look at the two-party-preferred votes for the newly returned government, the Newspoll sequence of 50-50-48 is exactly the same as what the Gillard government received in 2010, in its first three polls after an even narrower escape, on its way to three years of generally wretched polling and, eventually, a heavy defeat.

Saturday, September 24, 2016

Field Guide To Opinion Pollsters: 45th Parliament Edition

Just before the 2013 election I posted a Field Guide to Opinion Pollsters, which has become one of the more enduringly accessed pieces on this site.  However, over time parts of its content have become dated or specific to that election, and with more and more pollsters emerging as others disappear, the thing has got too long.  I've decided therefore from now that I will post a new edition shortly into the life of each parliament, editing it through that parliament as the need arises.  Pollsters not expected to be active in the life of the current parliament will be removed, but the old edition text will remain on the previous page.

There are a lot of polls about in Australia these days.  But how do they all work, which ones have runs on the board and which ones can you trust the most? This article describes what is known about each pollster and its strengths and weaknesses and includes coverage of general polling issues.

The gold standard for success for an opinion pollster is seen to be that its polls at election time get the result as close to right as possible.  However, some pollsters are little-tested against actual elections, and getting a specific election right is a combination of skill and luck.  In elections where there is a swing on the last day or two of the campaign, a pollster that is actually not polling correctly may have its errors cancelled out by the swing, and hence record a lucky hit.  There is more to being a good pollster than just getting it right at election time - a good pollster should also provide useful data between elections and do so using well-designed questions that are easy to interpret.  And a pollster should also present their data in a way that makes sense and isn't misleading or confusing.