Tuesday, November 15, 2016

New Tasmanian State Poll: Rolling Comments

ReachTEL Liberal 45.6 Labor 30.9 Green 15.1 Other 8.4 (Undecided removed)
On raw numbers, it is a tossup whether Liberals would just retain majority (c. 13-10-2) or just lose it (c. 12-10-3)
Adjusted for house effects, Liberals would be likely to retain majority (13-10-2)

Introduction (15 Nov)

A new Tasmanian ReachTEL poll of state voting intention, with a large sample size, is about to be released by The Mercury.  As noted in the teaser, the poll points to the probable loss of two or three seats if the government were to face the music today.  A loss of two seats would leave the government with a majority of one while a loss of three would create a hung parliament.  This will be familiar territory for those who have followed my state polling coverage in the term, as my aggregated polling model has pointed to either 12 or 13 Liberal seats for a long time now.

Polling in Tasmania was inaccurate at the federal election, and has shown large pro-Liberal house effects at both the last two federal elections (but not at the 2014 state election) so there will always be some room for doubt about it.  Another important factor is the potential for a tactical bandwagon effect as seen in the 2006 election - if one party has a realistic chance of majority government and the other does not, voters may gravitate to that party.  While polling says the government's majority would be touch and go if an election were "held now", that assessment has very little predictive value because of this kind of strategic voting.

I have seen some of the results for the purposes of expert comment, but a recent business model change by The Mercury means I will be going about posting analysis of this poll in a slightly different manner.  As I understand it, many of the results will be released behind a paywall in the first instance and then published in the print edition the next day.  I won't be posting analysis of the results until they have appeared in the print edition, at least in basic form, or been freely reported.

As results are revealed over the next several days, in-depth coverage will be posted below.

Voting Intention

Below are the voting intention results, with undecided voters redistributed and converted into quotas and my estimate of seats.

Three electorates require some sort of comment.  The situation in Bass would be not so different to that in 2006 where Kim Booth had to catch up from notionally behind, but Labor would be less exposed to leakage than in that case (one full quota rather than two) and Andrea Dawkins' profile is not as high as that of Booth.  So I think Labor would win the seat.  In Franklin, Labor is ahead in the race for the final seat, but the Liberals are not that far out of the picture.  If Labor's vote suffers because Lara Giddings retires as speculated (Giddings has said she will make an announcement later) and leakage from Will Hodgman's surplus is not too bad then the Liberals are competitive.

The key seat is as ever, Lyons, where the Liberals' ability to win three seats would depend on the flow of preferences from minor candidates (which in 2014 was strong) and the alignment of votes within the ticket.  At the last election the Liberals' three winning candidates had roughly the same vote, which is the best possible scenario, but I expect that next time Guy Barnett will have a greater share of the Lyons Liberal vote.   On the poll figures the Liberal vs Green final seat in Lyons is a tossup (there's also a faint chance they would both win at Labor's expense.)

So, probably either 12-10-3 or 13-10-2.

Corrected Results

I have some concerns, however, about the tendency of ReachTEL polling in the state to underestimate the ALP and overestimate the Greens.  At the three elections polled by ReachTEL (2013 federal, 2014 state, 2016 federal) the final public ReachTEL polls have overestimated the Greens by 2.2, 4.6 and 1.3 points, and underestimated Labor by 4.1, 3.7 and 3.7.   Furthermore the overestimation of the Greens vote has been most significant in Franklin and Lyons.  The evidence on the Liberal vote has been mixed - overestimating it at both federal elections but having it too low at the state election.

The following is an example of how this poll might be adjusted to account for these apparent house effects:

Adjusting Labor's vote upwards means Labor easily wins two seats in each electorate.  The only seat that requires explanation is the Liberals vs the Greens in Lyons.  The Greens would suffer from leakage because they have no incumbent and the flow of preferences in Lyons tends to be bad for them.   If one Liberal achieves quota easily, or if none does, then the Liberals should be able to stay ahead of the lead Green by splitting their vote across multiple candidates.  The worst case for them would be two members in with easy quotas leaving a one-on-one race with the Greens, but even so if the preference flow from the 2014 election is repeated the Liberals would win.  So the most likely result is 13-10-2.

I believe the adjusted results are more realistic, with the possibility that they could be generous to the Government.  Labor dominated the federal election in the state and won a Legislative Council seat, unseating an incumbent.  The idea that Labor is only on 31% at state level after polling 38% in the federal election (where its Denison vote is gouged by Andrew Wilkie) does not, to me, seem all that plausible.

I note that this poll has a much weaker Ind/Others vote than the recent Wilderness Society ReachTEL.  Otherwise its results are pretty similar.

I am working on a revised state polling aggregate including this poll and will post it soon.

Revised Aggregate

In the current aggregate I have included the adjusted ReachTEL figures above, weighted at 40%.  I've also back-applied some corrections to the Wilderness Society poll.

The Lyons figures in the aggregate are very similar to the adjusted read of the current poll above and the Liberals would probably win the seat, depending on the breakdown between their candidates.  On this basis my aggregate goes to 13-10-2.

There are a few key features in these aggregated results:

1.  The swing is from the Liberals to Labor and to a lesser extent Ind/Others.   Once the tendency for polls to exaggerate the Green vote is considered, the Greens do not appear to be gaining support.

2. The Liberals can realistically win the election outright with a primary vote lead of only ten points.  The relationship between the primary gap between the majors and whether or not a majority is won has a lot to do with the Green vote.  In 1989 the Green vote was high and the Liberals failed to win a majority with a 12.2 lead (this was with 35 seats, but might well have also happened with 25).  In 1998 Labor won 14/25 seats with a primary vote lead of just 6.7 points.

Other poll results will be released in coming days.

Light Rail (17 Nov)

Today's release involves the question "Do you think that Hobart needs a light rail system?"  This gets the thumbs up from 45.3% statewide wioth 30.5% against.  In Denison there is 62.2-21.9 support while Bass (which contains Launceston, Hobart's eternal rival) was the only electorate to give a narrow thumbs down (30.2-39.4).

The Denison results are not surprising because traffic congestion around the city, while trivial compared to major mainland cities, has worsened substantially in recent years.

All the same I think a question of this kind is a bit of a pony poll.  It measures in-principle support in something that sounds yummy without exploring attitudes to cost.

Leaderships (19 Nov)

I don't have time to write much about this now and will be offline a lot of the time until Sunday night, but today's print edition should include some rather dire leadership polling for Labor leader Bryan Green.  Firstly he trails Premier Will Hodgman as preferred Premier by a nearly 60:40 margin including in every electorate and among every surveyed demographic.  This is a very large gap for a ReachTEL poll, in which the absence of an undecided option avoids the normal skewing of preferred leader polls.

Secondly Green is in fourth place as preferred Labor leader behind Rebecca White, Scott Bacon and Lara Giddings.  More comments on this later.

More On Leaderships (22 Nov)

Some detailed comments on the leadership polling.  Firstly here are the tables:

Rebecca White leads in every electorate and every demographic and party breakdown.  A useful comparison is with the June 2015 ReachTEL which also asked about the Labor leadership.  (Note however that the question has changed slightly from best leader at the next election to best leader now.) Despite the addition of two new female candidates and the removal of one male one, White's lead has only increased.  The pattern of Green leading neither his own electorate nor his own party was already present in 2015 but has become more clearcut.  It is not unusual for incumbent leaders to be opposed by voters for the opposing party, but when they are only third among voters for their own party there is a problem.  In this case fewer than 19% of Labor voters pick the incumbent.

Sally Glaetzer's excellent in-depth piece on Green's leadership adds another issue - that Green's connection with his own electorate (Braddon) has been reduced as he has moved to Mt Stuart in the electorate of Denison.  Green's troubled past including the TCC scandal is also cited as a factor undermining his leadership.  I don't think any of this would matter much if Green was seen as offering something distinctive and positive - if he was extremely charismatic or dynamic for example.  But without such a positive, what is Labor to do?

White is the clear public favourite but she is currently on maternity leave.  Whether she might even be interested in the leadership, say, mid next-year with a nine-month lead-in to the election, is unknown. In the past there has been a view - perhaps founded too much in conventional politics in a world where voters are seeking fresh approaches - that White is not leadership material and that Bacon is the heir apparent.  Glaetzer's article notes that although Labor has had two and a half years to prepare for a transition, Bacon is still considered "not ready" by several party members.

The better-Premier polling also shows that Labor has a leadership problem.  Will Hodgman leads Green 59.8-40.2.  This would be no big deal in a Newspoll-type poll but for a forced-answer ReachTEL this is a massive lead.  Hodgman is preferred by 59.1% in Bass, 65.1% (!) in Green's electorate of Braddon, 52.2% Denison, 61.6% Franklin, 61.0% Lyons.

No-one should assume Bryan Green is unelectable.  We have seen so many "unelectable" leaders win elections that it should be clear now that any Opposition leader can win if the government does enough wrong.  But there have also been enough past cases in Tasmania where poor leader choice has led to bad results (for instance the Liberals' disastrous 2002 election) that it should also be clear leadership matters.  We'll have to see how Labor resolves this quandary through next year.


  1. Realistically, in Tasmania, do you think, say, a 12-10-3 result would ultimately eventuate in a Liberal minority government, a Labor minority government or a Labor-Green government? (Liberal-Green sounds unlikely)

  2. The nearest comparison was 1996. In that case Labor had a policy of governing in majority or not at all. The Liberals as the incumbent government had to stay in office in minority with very minimal support from the Greens, which was a turnbulent experience for all concerned.

    So far in this term Labor has adopted a policy to not have the Greens as cabinet members and to consult with the broader party on a minority government arrangement. However they have not entirely ruled out accepting minority government. Greens supporters would strongly expect them to support Labor over the Liberals given a chance. So I suspect such a result would lead to a Labor minority government.

  3. How is Denison looking for an independent if Wilkie puts a candidate up?

    1. Not ideal because current polling suggests there would not be a lot of spare preferences floating around. So that candidate would currently need to pull votes from all over the place and get something well into double figures as a percentage in their own right. The same applies in Braddon which is the other obvious target for an indie. In the case of Denison, Wilkie's own following is so massive that if he can find a solid high-profile candidate they might be able to pull that off.


The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.