Saturday, February 24, 2018

Tasmania 2018: ReachTEL Confirms Liberals Probably Winning

ReachTEL (after redistributing undecided) Liberal 48 Labor 32.2 Green 12.5 JLN 5.3 (4.0 excluding seats not contested) Others 2.1
ReachTEL has historically skewed against Labor and to the Greens in Tasmanian elections
Most likely breakdown if raw figures are correct 13-9-3 or 13-10-2 (close to 14-8-3 or 14-9-2)
Most likely breakdown based on figures corrected for past skew 13-10-2 (close to 13-11-1 or 12-11-2)
Aggregate of all polling 13-10-2
Current prediction 13-10-2

The first independent poll of the Tasmanian election (see my main guide page) has been released and it confirms growing suspicions over the past few weeks that the Liberals will probably retain an outright majority, although it may well be a majority of one.  Moreover given the historic tendency of ReachTEL to skew against Labor and to the Greens, the poll can be taken as suggesting that the Greens are in serious trouble and that none of their seats are safe.  I am a little sceptical of whether that past pattern will continue.



The poll published by the Mercury today has a total sample size of 3,179 with 641 in Bass, 636 in Braddon, 624 in Denison, 633 in Franklin, 645 in Lyons.  The history of ReachTEL skew in Tasmania aside, a few aspects of the poll create interpretation issues:

1. The pollster has polled support for the Jacqui Lambie Network (JLN) in both Denison and Franklin, where the Network has no candidates.

2. The pollster has provided "Undecided" figures (2.4-5.0% varying by electorate).  However ReachTEL defines "Undecided" differently from other pollsters.  If a voter is undecided when first asked their vote, but on the second attempt names a party, then this voter is treated as "Undecided" (Other pollsters include this voter as a voter who has a preference for a given party).  If a voter doesn't name a party on the second prompt, they are excluded from the survey.  (Newspoll defines this sort of voter as "undecided" but also excludes them from the survey).  The figures ReachTEL supplied to the Mercury would have included the breakdowns of the "Undecided" voters by party.  Without the full figures, I can only distribute them in proportion.

Figures without skew adjustments

With the "Undecided" voters redistributed proportionally, but with no adjustments for the history of ReachTEL skew, the following are the results:


However there is a need to redistribute the Lambie Network votes in Franklin and Denison.  Based loosely on the Senate election preference results I have redistributed them as Labor 55% Liberal 30% Green 15% (and ignored Others since none of the Others in Denison or Franklin will be elected anyway) to get the following:


On these numbers (if no adjustments are applied) two seats would be in realistic doubt:

* In Bass, the Greens would hold a 0.1 quota lead over Labor, but that lead shrinks if Michelle O'Byrne does not cross quota by the time Labor is down to its last two candidates.  My suspicion is that because O'Byrne is by far Labor's most prominent Bass candidate, she would have crossed quota by that stage, saving Andrea Dawkins from the dreaded Ginninderra Effect (sorry, I had to throw that wonky link in).  Assuming that's not an issue, it comes down to whether Labor's advantage on preferences from the Lambie Network - and possibly the Liberals if the vote is not split too evenly between the third and fourth Liberals - is sufficient to account for the Greens' advantage on leakage.  As the minor Labor candidates are eliminated votes will leak out of the ticket, and this is what saved Kim Booth in 2006.  I would think the Greens would probably just hold on these numbers.

* In Braddon, the Liberals are close to holding four seats by the same method that enabled them to do so last time.  Indeed the Braddon figures are quite similar to the 2014 result (Liberal 3.53 quotas Labor 1.39 Palmer United 0.43 Greens 0.42).  However the Liberals only just won in 2014 (by 0.04 of a quota) and had the advantage that Labor's second candidate Brenton Best was on the nose with Greens voters.  Also they were very lucky to get the seat because of an even split between Roger Jaensch and Joan Rylah (the dreaded effect again). In this case I would expect Labor to do better on Greens preferences.  The Lambie Network would not win a seat on these numbers because of leakage and splitting of primary votes between candidates from other parties.

So, on raw figures, 13-9-3 or 13-10-2.  The individual electorate samples will always be a bit iffy but if the Liberals are really 15.5 points ahead then by historic standards that should be enough for an outright majority.

Adjusted for past ReachTEL skew

ReachTEL has polled Tasmania for the 2013 and 2016 federal elections and the 2014 state election.  At all these elections it overestimated the Greens' performance and underestimated Labor's.  At both federal elections it overestimated the Liberals, but at the 2014 state election its last two polls underestimated them.  On this basis I normally apply adjustments of -2.7 to the Greens and +3.8 to Labor in dealing with ReachTEL Tasmanian polls.  We shouldn't assume that because a pollster has displayed skew in the past it will necessarily do so again, but we also shouldn't assume pollsters necessarily will fix such problems either.

The following numbers are an attempt at interpreting this poll with these historic corrections applied.  The corrections do produce some obvious rubbish results (like a Green vote in Braddon that is plainly too low and ditto for Others across the board) but it is really the position of the three main parties we should be concerned about here.



With the house effect adjustments applied, Labor easily beats the Greens in Bass and the result is 13-10-2.  However, there are a few possibilities that are well within the poll's notional margin of error that we should keep an eye on:

Firstly, on these numbers Labor is within striking range of winning three seats in Denison, at the expense of either the Greens or the Liberals.  This would not be easy, as even if Labor managed to poll three quotas, Labor would still lose votes from leakage from the likely surplus of Scott Bacon and from the preferences of their minor excluded candidates.  Given that there is a fight between Tim Cox, Ella Haddad and Madeleine Ogilvie for the last Labor seat or two seats, someone is going to be excluded with quite a lot of votes, which means a leakage problem.

Secondly if the Green vote was a little worse and one of the major parties better then the Greens would be at risk of losing Franklin, probably to the Liberals, as well.  Again here, the dreaded effect might come into play - if Jacquie Petrusma and Nic Street had a close enough split of votes, then something like 1.7 quotas between them might beat a lone Green on 0.8.

So there are a few black swan scenarios to keep an eye out for here:

1. The Liberals cop a really unlucky distribution of votes, overkilling Bass and Braddon but falling just short in Franklin and Denison, and fail to win a majority.

2. The Liberals do better in Franklin but worse in Denison and win a majority with four threes and a one.

3. The Greens could be wiped out entirely or reduced to one seat.

Probably, none of these things will actually happen, but on this poll they cannot be ruled out, and my home electorate, the People's Republic of Denison (that's the last election where I can call it that), does some funny things sometimes.  It's important not to place too much weight on seat samples that may be aberrant anyway - the overall picture is that the Liberal lead is large enough that they should win outright.  This might be further strengthened by a bandwagon effect because voters wanting to avoid minority government at all costs will be most likely to vote Liberal.

Aggregate of all polling

This is the first public poll since early December last year and has a large sample size but still shouldn't be treated as the final word, especially on the seat breakdowns.  I have included it in my polling aggregate in a way that weights the overall distributions of voting intention in this poll at 80% (since it is so long since the last data) but the seat breakdowns at only 60%.  I've also excluded the JLN from those seats in which the party is not running.  This is the new aggregate:


The Greens figure in Braddon in this aggregate is again too low, and the Others figure in Denison is too high (since the Others in Denison are only a lone Shooter/Fisher/Farmer and two candidates from the comically obscure Tasmanians 4 Tasmania).  The Others figure in Braddon is also too low because of the likelihood that Brenton Best will poll a few points.  However, the numbers in the aggregate point to 13-10-2, which I also posted as my current prediction in the sidebar during this week.

It is possible for the Liberals to win more than 13 seats, especially as their Franklin seat is by no means dead on these figures.

Preferred Premier

The Preferred Premier indicator has Will Hodgman ahead of Rebecca White 51.8% to 48.2% on the forced choice question.  Hodgman leads with 58.1% in Bass, 62.8% in Braddon and 55.2% in Lyons, but trails with 45.2% in Franklin and 37.2% in Denison.  The Mercury notes that White was preferred by 95.3% of Greens voters and 55.2% of Lambie Network voters.  However on the current poll numbers Greens preferences will only be distributed in any numbers in Braddon.  Everywhere else they will either win with little or any surplus or else finish sixth without being distributed.  Lambie Network preferences will tend to exhaust, especially in Braddon where they have a team of five.

The Mercury states that White has "slipped behind" but Preferred Premier scores cannot be easily compared across different polls.  The numbers do however confirm my feeling that voters who would choose undecided as an option given that option do in fact favour Hodgman.  This is contrary to the normal pattern in which incumbent PMs or Premiers do better if the poll has an undecided option.

Why Is It So?

I have a few comments about the meaning of this poll's results.  Firstly, Labor is losing the fight against the bandwagon effect, which is strong in Tasmania because many voters see avoiding a hung parliament as important.  The Liberals have churned out internal polling data showing their own party ahead if not clearly winning and Labor has not effectively countered this narrative.  Labor has claimed its internal polling shows that Labor can win a majority (and were out doing so again today) but this is especially unconvincing when they can't provide evidence that it is true.

The second thing is that this is playing out as a north-vs-south election and the poker machine issue, as well as Rebecca White being a relatively left-wing leader, is having an impact there.  On my corrected numbers the swing back to Labor on primary votes is highest in Denison and Franklin (11-12 points compared to eight points elsewhere) while the swing against the Liberals is reasonably even across the board.  This contradicts previous data which suggested that since the northern seats had swung the most in 2014, they would also swing back the most this time.  It would seem that the pokies issue plays strongly in both parts of Denison (as an issue of middle-class concern in the Hobart part and by direct impact in the Glenorchy part, in both cases amplified by the influence of Andrew Wilkie) but that outside the south the issue is having less traction.

Reactions

On TV news bulletins, ALP figures continued to say that their internal polls showed they could win majority government, but again no evidence was provided.  State secretary Stuart Benson did refer to
the scale of Labor's campaign in terms of the numbers of conversations with voters.  I tend to agree with him that conversations are more effective ways of winning votes than billboards, but voters still have to be receptive to what is being said.

Richard Herr on Southern Cross news suggested that the Greens could have a chance of holding Lyons on the back of anti-major-party votes from the Jacqui Lambie Network.  However, as the Senate election preference flows showed, Lambie Network voters do not like the Greens and are more likely to preference both major parties, as Palmer United voters also did at the last state election.

The Mercury's editorial makes much of the possibility of Labor winning three seats in Denison.  However its argument (that the Liberals are "stranded just shy of two quotas in Denison" and might hence lose if all the independent/undecided voters swing to Labor over pokies) not only represents an unrealistic scenario but also ignores Labor's greater exposure to leakage from surpluses (and probably also excluded candidates). The more realistic arguments for three Labor in Denison would be either sample error or pollster house effects.  The editorial then goes on to say that Rebecca White could easily become Premier while keeping her agreement to not deal with the Greens, simply by having the Greens support Labor on confidence and supply.  This ignores the fact that White has not just said she will not deal with the Greens, but has said that she will not govern in minority at all, so she would still have to break a commitment to govern in that case.

Update (25 Feb): Issues Questions

Today's Sunday Tasmanian contains issues questions where voters are asked which party is better placed to manage various things and given the options of Liberal, Labor, Greens or JLN:

Economy - Liberal 53.2 Labor 35.2 Greens 6.7 JLN 5.0
Health - Liberal 48.1 Labor 40.3 Greens 7.0 JLN 4.6
Public education - Liberal 48.2 Labor 39.4 Greens 7.9 JLN 4.5

The report says that "voters back the Liberal Party over Labor to manage the economy, health and education [..]".  However, this claim is only clearly supported in the case of the economy, on which the Liberals received over 50% of support.  For the other questions, it is highly likely that those preferring the Greens would strongly back Labor over Liberal, and probable that those preferring JLN would do so to some degree.  I suspect the Liberals would be just ahead on a two-party basis but it would be very close.  I am not greatly fond of the inclusion of minor parties in questions like this, as it is well established that outside of certain pet items (such as the environment or climate change for the Greens), minor parties are not well regarded as able to manage issues, and their inclusion distorts the gap between the major parties.

Nonetheless, health and education are normally Labor Party strength areas and the party should be doing better on these questions.  My suspicion is that the focus on pokies as the key conflict issue on the election has made it more difficult for Labor to sell its position on health.

Most Important Issues (26 Feb)

Today's Mercury includes the question "When voting at the upcoming Tasmanian state election, which of the following issues will influence your vote most?" (my bolding)  Health leads on 31.9% followed by jobs and the economy 30.2% Pokies reform 14.3% Education 9.7% Cost of living 7.9% Energy security 3.3% and fish farms 2.8%.  In Denison, pokies reform tops the list with 26.1% to 24.9% for health and 20.6% for jobs and the economy; in all other electorates the big two are first and second in various orders.

The poll also finds 57.1% to 42.9% support for Labor's poker machines policy over the Liberals, rising to 70.9% in Denison, with only Braddon (48.1%) not preferring Labor's policy.

However the reporting of this poll says "Pokies reform has been touted as the biggest issue facing Tasmanians as they head to the polls on Saturday, but exclusive polling indicates it will influence just 14.3 per cent of the statewide vote."  Oh no it doesn't!  Voters can be influenced by more than one issue at a time!  (There would also be voters who are influenced by none of the issues mentioned, because their vote is already locked in for some other reason, and who just pick one arbitrarily to get through the question).  Unfortunately, the reporting is riddled with other incorrect assertions about the meaning of the results.

I think it is very interesting that pokies reform has scored so highly in the Denison sample, but it is important to be aware that people picking that option will include some voters who are opposed to Labor's policy and think opposing it is the most important issue.

I will be covering the election on election night but don't know if it will be here or elsewhere; a link will be posted if the latter.  I expect to be offline most of Sunday March 4 and unavailable for in-person interviews (phone may or may not work) unless somebody pays me to remain in Hobart.  

25 comments:

  1. Would the G effect come into play for Ogilvie/Cox/Haddad?

    ReplyDelete
    Replies
    1. Not that likely to on current numbers because the target lone candidate (in this case O'Connor or, say, Archer) would be very close to a quota, so the most likely way Labor would win would be by starting with, say, 3.1 quotas and then keeping losses on leakage down and getting three in with quota. There's more room for it to work when the target candidate is significantly below a quota.

      Delete
  2. Thanks! Be interested in your view on the contest between Broad, Best and Dow too.

    The 2nd para of your Preferred Premier section is missing a certain je ne sais quoi at the end ;-)

    ReplyDelete
    Replies
    1. Others being only on a few percent in this Braddon sample I don't think Best has a hope. I'm not aware of him running a big campaign and suggest he is mainly running to spite his former party for chasing him over fees.

      Delete
    2. Indeed, just wondered if he might successfully drag votes away from ALP to an extent that it hurts rather than annoys them.

      Delete
    3. I think he might well cause a small amount of pain. Not just with people switching to voting for him, also with people being turned off voting Labor by his attacks.

      Delete
  3. Possible alp 3 in Denison and 1 in Braddon? Seems strange occurring side by side

    ReplyDelete
    Replies
    1. I don't think those will both happen together. 1 ALP in Braddon is possible in the raw results but not the adjusted results, 3 in Denison is possible in the adjusted results but not the raw results.

      Delete
  4. Kevin - in your analysis you are not giving GRN any hope in Lyons? Would polling fail to pick up pockets of Green support in Lyons. There is some disaffection on the East Coast regarding planning laws and fish farming. What primary vote would the GRN candidate need to be in the game?

    ReplyDelete
    Replies
    1. Polling could fail to pick up pockets of any party's support anywhere, but in the past, if anything, Lyons has been the seat most prone to the Green vote being overstated. Even assuming ReachTEL has fixed the issues that have caused their Tasmanian polls to skew to the Greens in the past, the Greens are further behind than they appear because the JLN and Shooters preferences will be unhelpful. It's still within sampling error that the Greens could get up in Lyons but the poll doesn't provide any support for it. On these figures with the JLN vote down the Greens would need about 14% for a decent shot at winning, but if the Liberal vote is actually lower than shown then it could be possible to win with 12-13%.

      Delete
  5. Kevin, what are your thoughts on the campaign so far? I agree with your Twitter comments that the left (especially the Greens) have not been campaigning as well as the Liberals. I’m surprised that after Labor’s great Leg Co results over the last couple of years (and the Liberals’ shocking campaign in Pembroke) that Labor seems to have dropped the ball a bit after a good start, and the Liberals seem to have campaigned fairly well.

    One thing I was wondering about having a look at the numbers - is it possible we will get a female majority House? It seems close at this point, and if it happens we will only be the second jurisdiction to do so.

    ReplyDelete
    Replies
    1. It is possible we'll get a female majority House but my current estimate is that it will be a little short of that. Denison will almost certainly be 3 females (4 is possible, but requires Haddad beating Cox). Franklin will probably be 2 (perhaps fewer), Lyons looks like 2, Braddon looks like 1-2, Bass looks like 2-3.

      My perception so far is the Liberals caught everyone napping from mid-January on and have campaigned with a lot of energy and not just a lot of money. When they have been making claims and pushing issues in the media, Labor has not been countering them all that effectively. That applies especially to the issues of pokies and majority government.

      I also think that Labor have not sold their pokies policy convincingly, in part because they have not realised that it is not just an issue of evidence but also a philosophical issue. I think they should have gone for an approach around better regulation of pokies, not their complete elimination from non-casino venues. (I will change my mind on this if there is a hung parliament because Labor wins three in Denison.)

      Delete
    2. Thanks for your insights and I broadly agree - Labor has been quite “courageous” but dare I say hasn’t presented it well.

      I’m interested in your thoughts on the Standen-Midson contest in Franklin and Ogilvie-Haddad-Cox fight in Denison. Midson is obviously running very hard, with even a Shorten endorsement (however value that may be), but I wonder if he will be perceived as too similar to O’Byrne, while Standen has a slightly different profile, which may be helpful if Greens voters are looking to go somewhere else.

      In Denison, I’ve been surprised that all of Labor’s candidates have been running serious (ish) campaigns. Obviously Bacon will one seat easily, and I don’t think Sherlock can get enough without strong party backing, so the contest for the other seat (or two) comes down to Ogilvie, Cox and Haddad. I haven’t seen Cox campaign much at all, though obviously his name recognition might outweigh that entirely. Haddad on the other hand has been campaigning non-stop, and has strong union backing. I do wonder if Ogilvie can beat her baggage to win over Haddad (and Cox), she’s obviously campaigning hard to do so.

      In an election where the actual party battle has seemingly fizzled out (and the question of majority government perhaps already answered), I've found these intra-party contests to be quite an interesting dynamic - an added bonus of Hare-Clark, and one which I'm sure the Labor leader appreciates very much.

      Would be interested to hear your thoughts.

      Delete
    3. I have a similar reservation about O'Byrne and Midson competing for the same vote. I suspect that concern only goes so far considering the effort level of the Midson campaign but it will be interesting to see how they go compared to Standen anyway. Standen is indeed a natural fit for any fleeing Greens voters.

      My suspicion with Ogilvie is that she is short on support within the party and while she may beat one of Cox and Haddad, it's a tall order to defeat both of them.

      Assuming Labor doesn't win, these intra-party fights have a great impact on the future shape of the party including its future leadership - so they are indeed very interesting.

      Delete
  6. The Mercury does seem to be intentionally engaging in some funny games to prop up the image of the Liberal party. Some may suggest they are covering the campaign as if they were their propoganda unit. I made the decision to cancel my subscription because that's just not what a major newspaper should be doing.

    What sort of vote spread in Denison would put Cassy O Connor at risk of losing her seat? Peg Putt got 13.5% in Denison in 1998 but held on, would it be around that figure?

    ReplyDelete
    Replies
    1. In 1998 Putt started only 0.1% ahead of Labor's third candidate (based on party quotas) but ended up winning by 2.4%. Factors in this included:

      1. leakage from Jim Bacon's surplus (1% gained on the spot)
      2. favourable Democrat preferences
      3. Labor running seven candidates, increasing exhaust
      4. the four minor Greens candidates polling hardly any votes, meaning that Putt was much less exposed to within ticket leakage.

      1 will probably still apply this time but from a different Bacon. 2, 3 and 4 will not apply (4 won't as Helen Burnet tends to poll a good vote in her own right).

      The danger for the Greens is that the quota split is more or less 3 Labor 2 Liberal 1 Green but they are a little bit short. On this basis there is a risk of losing with a total Green vote of anything below 16% though she would have to be very unlucky with the breakdown of the other parties to lose with that.

      Delete
    2. Matt, I just cancelled my newspaper subscription too for the same reason!

      Delete
    3. I think a reasonable number of people have. When I spoke to the call centre operator I asked if there had been an increase in cancellations recently and she said " there has been a few ". It's probably a lot. Dumbed down shock jock reporting appears to be the directive from the new editor. Incredibly sad I have been reading the paper since I was a kid! Those page 2 and page 3 spreads from the editor about The Mercury always fighting for truth and the "we're for you" slogan is simply damage control. My subscription ends on March 16 until that time I can use or maybe waste some of my spare time on the circus, which is their online comments. For quality local reporting I will always have the ABC, and Kevin Bonham.

      Delete
    4. I understand that a fair number of people have cancelled subscriptions over the Mercury's opposition to the cable car. It's a niche issue but taps into some deeply held values for those it does resonate with.

      Delete
  7. Is it usual in Tasmania for parties not to recommend preferences beyond their own candidates? How common is exhaustion?

    ReplyDelete
    Replies
    1. Completely usual - I cannot actually remember a party recommending preferences outside itself, though it has probably happened sometime. A restraint on it is that a party needs the written permission of any other candidate to name them.

      Recently exhaust rates when a party running 5+ candidates has had all its candidates elected or excluded have been around 35-40% for the Greens and 50-65% for other parties.

      Delete
    2. Also, recommending preferences to other parties is ineffectual because how-to-vote cards at polling stations are banned.

      Delete
  8. Kevin, would love to see any 'size of parliament' ponders from this poll, and/or the actual election result. The 2013 article looked at the 5 seat verses 7 seat effect.

    From some pretty simplistic modelling of the last few elections I suspect the 6 seat option would be least likely to create hung parliaments (which may or may not be a good thing).

    Be interested in your thoughts (if), as, and when.

    ReplyDelete
    Replies
    1. On the raw poll numbers, 18-13-4 or maybe 19-12-4. Adjusted for skew, 18-13-4 or 18-14-3. So it is the same in terms of the Liberals probably getting a one-seat majority, but probably a little better for the Greens.

      Delete
    2. I will also calculate these figures for the election result some time after the election, possibly after all votes are known.

      Delete