Monday, December 7, 2020

Tasmania: Secrecy Concerns Or Just Secretive Polling?

EMRS Tasmania (state): Liberal 52 Labor 25 Greens 13 Others 12

If accurate Liberals would increase their majority (14-16 Lib 7-8 ALP 2-3 Green)

 uComms (commissioned by Australia Institute) Tasmania (state): Liberal 50.3 Labor 31.8 Greens 10.7 Ind/Other 7.2

If accurate Liberals would retain majority but probably not increase seat numbers (13 Lib 9-10 ALP 2-3 Green)

New seat aggregate of all polls: Liberal 14 Labor 8 Green 3


"Sir Humprey: How are things at the Campaign for the Freedom of Information, by the way?

Sir Arnold: Sorry, I can't talk about that."

- Yes Minister, Party Games


I doubt I'll ever stop noticing how much The Australia Institute contradicts itself.  The left-wing advocacy think-tank "conducts research on a broad range of economic, social, transparency and environmental issues in order to inform public debate and bring greater accountability to the democratic process."  But when it comes to informing public debate about its own polling, and hence making itself accountable for the research it conducts, its form is indifferent to say the least.  This is at a time when groups like this could do things to help fix the accountability and information problems that continue to blight the Australian polling industry.  It seems TAI is more interested in trying to achieve political change than in truly informing public debate.  And yes, I've said all this before.  

This article was mostly written yesterday when an Australia Institute uComms poll was the only Tasmanian voting intention poll for the last few months, but it's had a very short day in the sun before being joined by a new EMRS.  I've had to do a quick rewrite of an article I was just about to release, a major point of which was to criticise the Australia Institute for again hiding its poll, which made it harder to see that its claims about it didn't stack up.  EMRS is not affected by this criticism; it has published its numbers in the admittedly spartan way that it usually does.  

EMRS

The new EMRS poll is a slightly weaker, but not statistically significantly weaker, result for the Gutwein Liberal Government compared to the August poll.  The Liberals are down two points to 52, Labor up one to 25, the Greens up one to 13, and others up one to 11.  EMRS does not ensure its rounded numbers always add to 100.  Peter Gutwein's lead over Rebecca White in the skewed Better Premier metric is down from a massive 70-23 to a still very large 61-26.  (Better Leader metrics skew to incumbents by about 15 points, so around 20 points of Gutwein's lead there is real.)

If this poll was repeated at an election, we might (ignoring unknown unknowns regarding independents in Clark) get something that looks like this:


In Bass, Labor would lose a seat to either the Liberals or the Greens.  In Braddon, the Liberals might win a seat from Labor, but if the Labor vote split evenly between their candidates and the flow of Greens preferences was good, leakage and a depleted lineup of incumbents could damage the Liberals' chances.  Nothing to see in Clark, the Liberals would regain one from Labor in Franklin, and in Lyons Labor would hold off the Greens because of favourable preferences from Others and leakage within the Greens ticket.  

Australia Institute uComms

uComms presents results in a format without initially redistributing soft voters (which it calls undecided voters, although other pollsters don't), but the tables allow the soft voters to be redistributed.  With this done the Liberals are on 50.3, Labor 31.8, Greens 10.7, Ind/Other 7.2.  These figures are almost identical to the 2018 Tasmanian state election.  If this happened at an election with uniform swings, Labor would be eliminated at the key point in Bass by about 200 votes and lose the seat to the Greens, but that could just as easily not happen.  

The prospect of winning majority by one seat again as implied by the uComms poll is a fragile one for the Liberals, because of the complications in Clark, where it remains unclear what will happen with Speaker Sue Hickey at the 2022 election.  However, this poll should be compared with the mid-term polling recorded by the then Hodgman government in the previous term.  Four years ago the government were at 45.6% according to ReachTEL and 40% according to EMRS, so even the uComms poll is better than those.  

Now You See It Now You Don't?

Now to the curious history of the uComms poll and the dubious claims made about it.  The poll was first reported by The Mercury  in a paywalled article, but only in summary form.  Checking the TAI website I could find no evidence of the results.  However, in yet another Beware of the Leopard moment with TAI's polls I later found the poll PDF still online because it was attached to the bottom of a press release that had been removed from public access but had fortunately clung to life in Google Cache.  Whether the page's removal from public view, its public posting in the first place, both or neither were accidental or deliberate who can say.  

The TAI presser makes a big deal of the poll supposedly showing "how quickly voting intention can change",  "a dip in the Government’s popularity and a rise in support for the Labor Party" (since the August EMRS), and all this is implied to be linked to the lack of recent progress on electoral donations laws.  TAI's supposed informing of public debate doesn't mention that the last poll by the same pollster, late last year, was 11.3 points worse for the government than this one.  

The idea that the uComms shows a change in voting intention from the three-month-old previous EMRS has already been undermined by the new EMRS, which on a like-for-like basis finds a statistically insignificant change.  But in any case:

1. EMRS has undergone major methods changes since the 2018 state election while uComms, as distinct from ReachTEL, has never polled at a Tasmanian state election.  As such neither are polls for which a clear house effect baseline has been established and so any difference between them could easily be a result of unknown house effects in either poll rather than any actual change in voting intention.  

2. Indeed, both uComms polls for Tasmania so far (this one and Nov 2019) have had the Liberals on a significantly lower vote than the nearest EMRS polls (EMRS has released results for November and December 2019).  This is the only aspect that has been consistent in the differences between the two across these polls.  Last year's uComms had the Greens and Ind/Others higher than EMRS did at the time, and Labor lower.

3. The actual difference between the Liberals' results in the two polls is 3.7%.  It looks larger when the Liberal uComms result is given as 47.7%,  but that is not comparing like with like as "undecided" respondents have not been redistributed.  The difference would be statistically significant in theory if we treated the two polls as the same, but not by much, and perhaps not if we used a real measure of margin of error rather than a spurious one that treats polls as random and unweighted.  

4. If we were inclined to treat the polls as like for like comparisons, we could consider that EMRS underestimated Labor in 2010 and 2014, but underestimated the Liberals in 2018.  uComms polls elsewhere don't seem to have any consistent pattern of skew so far.  

So this doesn't show how quickly voting intention can change, and doesn't even clearly show that it has changed at all.  However, it is interesting that Labor's result is much better in the uComms poll, which seems less prone to overestimate Others than EMRS.  EMRS also has more of a history of overestimating the Greens, though their recent methods changes seem to have addressed this.

Overall, the Australia Institute are claiming that there is a major change in voting intention that can be attributed to the donations disclosure issue because they want to believe that voters care a lot about that issue and that it will damage the government if the government does not act on it quickly.  In the context of the COVID-19 pandemic and the economic issues it raises, it is more likely that donations reform is currently a bubble issue that is salient only to political junkies, most of whom have set opinions anyway.    Especially with the new EMRS (which was taken sooner after one of the recent waves of publicity over donations than the uComms) it looks more likely that the differences between these polls, especially in the Labor vote, relate largely to "house effects" of the two pollsters.  

Young Conservatives?

The uComms poll also has a strange feature that has been seen in many (but not all) recent uComms breakdowns: the 18-34 year old age group appears to be very conservative:

This (and it's not just Tasmania) flies in the face of everything we get in post-election surveys, and strongly suggests that the 18-34 year old age group as contacted by uComms via robopolling is unrepresentative.  (It's a problem with robopolling generally.)  This is presumably cancelled out to some degree by some other group, as uComms is not showing a general overall skew in its polling and has had good results so far.  But it's a bit hard to take uComms completely seriously when we keep seeing this over and over.

Issues Questions

The issues questions in the uComms poll vaguely concern political donations and government transparency.  Political donations laws have been a hot potato from time to time especially since the last state election, at which there was a large third-party advertising spend by interests opposed to Labor's then-proposed poker machine gambling restrictions.   Progress on Electoral Act reform (affecting not just donations but also other areas, eg requirements on naming candidates without their consent) has stalled and it is unclear when/whether the government intends to continue the process.  However there is still plenty of time for it to do so at some stage during 2021.

The first question is:

Thinking now of transparency and good governance in Tasmania

Do you agree or disagree with the following statement?

"Reform in Tasmania is needed to tackle perceived or actual corruption."

Firstly this is an agree/disagree question, and polling of this nature is problematic generally because of acquiescence bias.  But it's also an extremely vague question, because it doesn't specify any level on which the respondent thinks "perceived or actual corruption" applies, nor has the respondent been asked to think of transparency and good governance at any specified level.  

The second question is:

Following the 2018 election, Former Premier Will Hodgman committed to political donations reform. The state Government is yet to legislate donations reform. Should the Tasmanian government prioritise donations reform so new rules are in place before the next state election?

Of course something should be done about Tasmania's overly lax donations disclosure requirements.  But this question on this issue is a classic example of what I call a skew-poll, since the question presents arguments for political donations reform (a reminder of the 2018 election, and the fact that Will Hodgman committed to reform) and no argument against.  Another problem with the wording is that it implies that the government must prioritise donations reform (which might be taken as meaning that it should make it the highest priority, or move it up a list of priorities) in order to ensure it happens in time, but that isn't necessarily true.  A further problem is that the Government does not alone determine whether or not donations reform occurred - as the other side of politics found, the Upper House gets a say.  

The third question is:

Proposed reforms to Right to Information/freedom of information laws would require the government to publicly release detailed information on applications that were refused and why, and publicly release more information even if not subject to Right to Information request. Would you support or oppose these reforms?

As well as being another agree/disagree type exercise this one is a pony poll - it's a specialised issue that won't be uppermost in the respondent's mind, but it sounds nice on paper so a respondent unfamiliar with any surrounding debate that might exist is likely to agree without much thought.  It's also another vague question: "and publicly release more information even if not subject to Right to Information request" - more information about what?

Aggregate!

With two polls out at once, it's time to relaunch my Tasmanian voting intentions aggregate.  For this relaunch I am weighting the new EMRS at 45%.  The new uComms is weighted at 35% (it is smaller than the EMRS but slightly fresher, but it should also be marked down because it is a commissioned poll that the commissioning source has used to attempt to make political points.) The previous EMRS which is now three months old is weighted at 15% and the one before which is now six months old is 5%. For now I make no assumptions concerning house effects, though the way I have weighted the polls takes more of a lead from EMRS because we have more data from them.

Here's what I get:


In this aggregate the Liberals have 14 likely seats.  There are two unclear seats.  On my analysis the Greens would recover Bass on Labor preferences on these numbers (Labor's vote would be too unevenly split to get two, or would drop if Michelle O'Byrne retired).  However, this is not clear cut as the Greens lack an incumbent and the Liberals have a very strong lineup here, so the Greens would need a good candidate and a lot of profile building to make it happen.   The Liberals would in theory be in the mix for four in Braddon if they could get an even split between their third and fourth candidates, but with two incumbents and one part-term incumbent I am sceptical that that would be the case.  (Or if an incumbent retired their overall vote would weaken.)  On balance I think this seat would therefore more likely go to Labor.  Note also that the Liberals' third Franklin seat (where they have a weakish lineup against two Labor incumbents) is not all that safe on these numbers, but chances are if they underperformed there, they would overperform somewhere else. 

On this basis my new seat projection based on all available polling is that an election held now would return something like 14 Liberal 8 Labor 3 Green.  However this assumes the polling is broadly accurate, and in any case we have seen that in election campaigns, voting intention here sometimes really changes quickly!

If the polls are accurate, the Liberals are probably riding slightly above their 2018 election level at the moment, but that is a result of a blowout caused by COVID-19 and it remains to be seen whether the glow from the government's largely successful handling of COVID will persist as long as early 2022.  It might be seen as not much of a pandemic boost, but Tasmanian governments have a long history of mid-term polling slumps that are usually recoverable.  A COVID bounce can only be measured off a polling baseline, not an election one (indeed one might argue the 2018 election was itself an inflated baseline) and the government's baseline at the end of last year was around 8 points lower (on a similar aggregated basis) than it is now.  

3 comments:

  1. Apart from the young conservative result found by the uComms poll, they also have the male demographic skewing left relative to the sample (much higher Green vote).

    All over the world, polling and research in similar countries such as the US, UK, and here in Australia all show that when there is a gender gap, female voters tend to tilt left while male voters tilt right (relative to the sample). While I appreciate their willingness to put out an outlier, if this uComms poll was correct, that would make Tassie voters quite exceptional, gender-wise.

    ReplyDelete
  2. Hi Kevin, love your work. I just got off the phone from a Nick McKim automated poll (can't remember who did it). It seemed that every issues question followed a similiar pattern: about three Green options, one Labor/Green option, and one Liberal option. Expect a poll-shaped object trumpeting widespread love for the Greens and their views to come out in the next couple of weeks.

    ReplyDelete
  3. These polls demonstrate why the Libs don't lose a wink of sleep about the stuff in the Mercury about secrecy or environmental issues with Salmon Farms. People simply don't care.

    ReplyDelete