Tuesday, February 4, 2014

Libs' ReachTEL: Not Completely Over The Line

Liberal-Commissioned ReachTEL: Lib 48.8 ALP 25.8 Green 15.3 PUP 4.8 Nat 0.7 Other 4.6
Outcome based on this poll if election held now: Probable Liberal Majority Win (13-14 seats)

(Note added 15 Feb: While this poll was interesting at the time in apparently showing the Liberals in a not clearly winning position (adjusted for apparent house effects) I now have reason to doubt the adjustments and to suspect that this poll should in fact be taken on face value or nearly so.)

The first thing to bear in mind when considering the Liberal Party commissioned ReachTEL that was released in the Mercury today is that it is a party-commissioned poll, and all party-commissioned polls should be treated with some caution, since parties only release them when it suits their purposes to do so.  You can bet your proverbial bottom dollar that if this poll had shown the Liberals on 37%, Labor on 35 and PUP and the Greens on 14% each, it would have been rapidly buried under a bush somewhere.  Likewise a poll showing the Liberals on, say, 57%, would have been a risky thing to release, because it might have caused voters who were wary of the Liberals but wanted majority government to believe the latter was in the bag and therefore there was no need to vote Liberal.  Robopolls are less expensive than operator dialling and this has led to a flood of internal polling in elections since they became commonplace.

The present poll is one of at least two and possibly three robopolls that I am aware of having been conducted in recent weeks.  A ReachTEL conducted last Tuesday asked voters their voting intention, how committed to that vote they were, how they voted last election, who their preferred Premier was, whether they thought Labor and the Greens would do another deal, how important majority government was to them, and how they ranked a set of issues including jobs, education, the economy, the environment and the pulp mill.  Results of that poll have not yet been released and I don't know who commissioned it. 

What we have now seen is surprisingly old rope (albeit the freshest data publicly released) - a Liberal Party commissioned poll from two weeks ago, conducted on the night of Monday 20 January.  At that time Labor had kicked the Greens out of Cabinet and the dust had had a few days to settle from the release of the election date, but Parliament had not yet met and passed legislation to protect the probably imaginary Bell Bay pulp mill proposal from legal challenges.   

If you were asking what would be the best set of poll results for the Liberals to release from a strategic viewpoint, the answer would be that those results would show that only the Liberals could win majority government and that they had an excellent chance of doing so, but that the job was not done yet.  It turns out that is more or less exactly what this poll shows. 

The Raw Results

It is not known to me whether the poll included multiple questions, although a robopoll being reported at around the same time did have only one real question.  In any case, the Liberals have released results for a voting intention question, and at this stage, no others.

The sample size of around 2000 is a little smaller than for some of the ReachTELs we have seen lately, but it's still pretty good.  It does make it possible to draw some meaningful conclusions using the electorate samples, but with a high degree of care.  The margin of error for a result of 50% in one of these samples is around 5%; for a 10% result it's about 3%.  However, with so many individual results released at once it would be surprising if one wasn't outside that and a few are bound to be iffy.  It's not likely PUP are really three points more popular in Franklin than anywhere else, and it would be quite unusual if the Liberals' Franklin votes and Bass votes were really so close together.  (Added NB: A regular reader of this site has suggested it is possible the Franklin PUP candidate Debra Thurley is getting a sympathy vote based on the plight of her son, injured footballer Cameron Thurley.)

Compared to the last ReachTEL released in November, and looking at figures with undecided voters reallocated from that poll, the Liberals are down 3.9 points, Labor is up a point, the Greens are down 0.5 points, and others (including PUP) are up 3.5.  In the November poll, Palmer United were not named.

If the above results were repeated at a state election the Liberals would win easily, despite the slight drop from November.  They would gain seats everywhere but Denison, while Labor would lose seats in Bass, Franklin and Lyons.  The final seat would be unclear between Labor and the Greens in Braddon, but I think Labor would most likely get it, for a 14-7-4 result. Given that the Liberals would have a 23-point lead, this is about the expected outcome, whatever the vagaries of given electorates. There would be no joy for PUP or anyone else.  (Concerning the Denison "Other" result, some of these "Other" responses may be ambit claims or from people who think they will vote for Andrew Wilkie.  But even if the 8.3% was real and all went to one candidate, it would be a tough challenge for that candidate to get up on preferences from there.)

But Perhaps It's Not That Simple

Aside from the poll being a party-commissioned poll, the main reason for caution about this poll is that at the federal election, the actual results showed quite consistent differences with the ReachTEL polls in August (and in the case of Bass, September).  For the state as a whole, these differences were seen between the August federal polling and the September federal outcomes:

The rise of PUP from nowhere from the federal poll can be explained by a low-profile and new party splashing out with a lot of money close to the election.  It's possible PUP will build strength close to the state election too, but on the surface it seems likely it would be a less dramatic shift.  The decline in the Green vote in the last few weeks would seem to have more to do with the anti-major-party vote being scooped by PUP, and the PUP factor might also affect the major party results.  However, as noted in my article about modelling state results off the federal election, it can also be argued that voting intention shifted from Labor between the taking of the August ReachTEL and the election, and that all else being equal the polled result for Labor in August should have been above the election result. 

All the above considered, here is what the current poll might imply if a similar pattern occurs as in the federal election.  It may be that it will not and that the state polling will turn out to be more or less perfect, but the possibility should be examined.  For these purposes I have applied the following adjustments: reduce Liberals by 4, increase Labor by 4, reduce Green by 1, increase PUP by 1.5 except in Franklin, reduce Other by 0.5 (based on the combined fate of Others and Wilkie), and distribute half the Nationals vote to the Liberals and the other half to Others (since their preferences will very likely favour the Liberals.)

That produces the following:

In this modelled version, the Liberals' lead over Labor is cut to 15.4%.  In Bass and Franklin the Liberals drop substantially below 50%.  Braddon, Denison and Lyons are unaffected but the Liberals' third seats in Franklin and Bass come under the spotlight.

The Liberals would only have to win three in one of these to win a majority.  But on the modelled figures, in Franklin the Liberals would have a big problem in the form of leakage from Will Hodgman's surplus.  In 2010, 6.8% of those voting 1 Hodgman then voted 2 for a non-Liberal candidate.  Most of these leaks went to Labor.  Hodgman's surplus improved Labor's position compared to the Liberals by 0.11 quotas.  It wouldn't be that high this time since Petrusma and Harriss are incumbents (one from each house of parliament).  But I wouldn't like the Libs' chances of winning three in Franklin if their primary lead over Labor was a mere 0.06 quotas (about 600 votes).

In the division of Bass, the modelled figure seemingly shows Kim Booth (GRN) well behind the Liberals, with 0.74 quotas vs 2.83.  (I have assumed Labor would win two on these figures given they have two incumbents.)  But in 2006, the Greens had 0.81 quotas to Labor's 2.98 quotas, yet Booth retained his seat by 136 votes over the third ALP candidate in one of the more famous Hare-Clark results.  It is true that in 2006 Booth had help from Les Rochester (IND) and that for that reason the gap was really more like 0.11 quotas than 0.17.  It is also true that Booth had the benefit of Michelle O'Byrne overkilling her quota in a way that I doubt Michael Ferguson and Peter Gutwein are going to do.  But the scenario is otherwise similar in that the Liberals do not have a clear frontrunner for their third seat and could struggle with leakage when their minor candidates are cut.  I would favour the Liberals in this instance, but there's not that much in it. (Note added: A factor in the Liberals' favour here that strengthens their position somewhat is the possibility of PUP preferences being unfriendly to the Greens.  See comments section.)

Whatever views one might hold about the fates of Franklin or Bass in these precise scenarios, the poll taken alone at least puts both close enough to not being picked up that it can't be said to show that either is a done deal. 

How Much Notice Should We Take Of This?

In Parliament during the debate about the pulp mill amendments, Peter Gutwein (Lib, Bass) repeatedly accused Labor and the Greens of running a "12-9-4 strategy".  The idea was presumably that there was a Labor-Green plan to cling onto government by holding the line at the loss of the Braddon Green seat to the Liberals, and the loss of only one of Lyons and Bass.  The claim that there was such a strategy involving Labor and the Greens was utterly unsubstantiated at the time and it did surprise me that a claim of a strategy involving a given number of seats was being pushed so hard.

This particular poll (which would have been known within the Liberal Party before Gutwein's comments) does indicate 12-9-4 as a realistic chance on the above reinterpretation only, although 13-9-3 would seem more likely.  It should, however, be seen against the backdrop of years of results pointing to easy Liberal majority wins.  A single 400 vote sample would be not enough evidence to overturn all previous indications that the Liberals are winning three in Bass, even if that poll was commissioned by a neutral source rather than a party.  It would be interesting to know why this poll took more than two weeks to see the light of day and whether the Liberals have any fresher polling they have not released.  But the more mundane and natural suggestion is simply that the Liberals have waited until other issues have quietened down before releasing this polling.

Lastly there is the question of just what a 12-9-4 parliament, if it did somehow happen, would mean.  Labor would only be able to form government in such a situation with the support of all nine Labor and all four Green MHAs.  In the last parliament, which had a balance of 10-10-5, Kim Booth did not support the deal done with Labor, while Brenton Best eventually started supporting motions of no-confidence in Nick McKim.  There would be no room for such antics if the government perpetually hung by the thread of a single seat, and it would be very interesting to see how such a balance would actually play out.  It is no foregone conclusion that another Labor-Green government would form.  Something tells me Best wouldn't settle for another term on the backbench.

Given the amount of polling data already available in the state and given that this ReachTEL is a commissioned poll, I'm not sure whether it's a good idea to include it in my state polling aggregate (of which the last version appeared in my last EMRS article.)  But even if it is included and weighted at 33%,  the aggregate still points to 13-8-4.  I am a little sceptical of the continued strength of the Greens in the aggregate given the disastrous results they have returned in so many elections lately, and with Bass and Lyons perpetually iffy for them and even Franklin now and then on the edges of the loseable zone.  However, their isolation from the major parties on the pulp mill issue is unlikely to have harmed their chances of avoiding the worst of the damage.

In the normal scheme of things I would expect at least one EMRS poll before the election, if not two, and it would be very surprising not to see further ReachTELs or other robopolls from various sources.  (Indeed, I have heard that EMRS are in the field already, though whether for a public poll or an internal poll is not known.)

PS (added 9 Feb): A reader has mentioned something possibly significant about the timing of this poll in interpreting its Denison results: the poll was taken just before Barbara Etter (PUP) announced that she was running.   Thus it may underestimate the Denison PUP vote.

UPDATE (15 Feb): Following the more recent ReachTEL commissioned by the Mercury, I have a different theory about this poll - that methods changes may have largely fixed the ReachTEL house effect.  If that is so then the remodelling in the "But Perhaps It's Not That Simple" section becomes void, and the poll does show a clear Liberal majority win, with something like 14-7-4.


  1. I'm really surprised at the result for Franklin. Ignoring your adjustments, 49% Liberal vote in Franklin with 8.9% PUP (which can be taken to split 50/50/0) suggests 53-54% after splitting it up, at which point 3 seats is likely rather than possible.

    How long since Franklin had three Liberal seats? Braddon has slipped back to a solid 3-2-0, after edging towards 4-1-0 last year, but I would have expected 4 seats there before three seats in Franklin. 12-9-4 as a strategy is interesting, but a bad fall of preferences for the Greens from the numbers above and the result ends up 14-9-2.

    Are the Greens making a concerted push for a second rep in Denison?

    Saw my first Tim Morris sign on the weekend, out the back of nowhere - actually two on the same building. But that's it. The paucity of Labor and Green campaigning at this stage is intriguing; I wonder what the differential between a short, concentrated campaign and an extended campaign is in terms of voting patterns/outcomes. Also how much of a impact signage has relative to other forms of marketing. If nothing else, it leaves the field _looking_ more like a Liberal primary than a general election.

    1. Yes I just don't fully trust that high Franklin result. Franklin and Bass should not be so close especially with that high PUP vote in Franklin. Needs corroboration from other polling.

      I saw some Labor signage in Denison and southern Lyons on the weekend - Bacon, Carnes, Gordon, Llewellyn, Clark. But nothing like the carpeting of Groom and Archer in Denison and Shelton especially in even southern Lyons (although he hails from the north of the electorate).

      It is true that the distribution of PUP preferences especially could be a factor in whether or not the Greens hold either Bass or Lyons. Thus if it is assumed PUP preferences go more to the majors than the Greens, then 14-9-2 (or even 13-10-2!) is realistic even on my adjusted numbers. In both Bass and Lyons, PUP are not running full slates, which may reduce the propensity of their preferences to exhaust. (Likely in Franklin and Braddon at least half the PUP voters will stop at five.)

      In the federal election in Bass, PUP preferences split 44% Liberal, 43% ALP, 13% Green and in Lyons, 44% Liberal, 41% ALP, 14.5% Green. In those cases however they were splitting to fairly poorly known Green candidates who were running stock-standard campaigns; Booth and Morris would likely do better but it isn't clear how much. I'd normally expect Morris to hold anyway but if his final opponent is either Llewellyn or White then he won't have the sort of profile advantage Booth had over his little-known final opponent in 2006. Even the loss of, say, 0.03 of a quota due to unfriendly PUP preferences would make Booth's task much harder on my figures.

      The Greens seem to be making a semi-serious attempt to get Bill Harvey up in Denison - he is very prominently pushed on their otherwise disorganised website - but I reckon they're dreaming. As normal. :)

    2. One has to dream from time to time : )

      First Brian Wightman sign today in Bass.

      Unless there is a distinctly different distribution for the PUP preferences than the first preference results, exhaustion shouldn't make much difference.

    3. The difference exhaustion makes is that my usual assumption with fourth parties and indies collectively is that their preferences will splatter more or less equally between ALP, Lib and Green. For instance the preferences of the five indies in Bass 2010 went 344 Green, 323 Lib, 307 ALP. So that means the "ind/other" column can be pretty much ignored in extrapolating the result, unless you get a situation like Bass 2006 with Rochester who was clearly attracting lots of green votes.

      I was doing the same with PUP on the grounds that they take more or less equally from ALP and Liberal and seem to compete for the anti-major-party vote with Greens, but the federal figures suggest they don't give preferences back to the Greens and so that assumption needs to be reconsidered. That being the case, the less the PUP vote exhausts, the more damaging it will be. Though if a PUP voter wants to stop at 5, it might be their preference is not that likely to reach any of the candidates competing for the key spots anyway.

  2. I would have thought that a 1:1:1 distribution of preferences would have had a more distorting effect on the count (2:2:1 normally, say 3:2:1 this time) than a 4:4:1 preference set like PUP recorded.

  3. It does in terms of the proportional vote shares of the parties, in that the ratio of the major party vote to the Green vote goes down. However because preferences from 4th parties are normally more prone to splatter than to follow the ratio of primaries, I usually think about these things in terms of gaps rather than ratios. For instance in my reworking of the poll above, 2.83 quotas vs 0.74 quotas - if this is taken as a ratio then the party with 2.83 quotas wins easily. But in practice the Liberals just don't get 3.8 times as many preferences from minor candidates of any stripe as the Greens do, and 0.09 quotas is a relatively small and competitive absolute gap. (That is, so long as the Green can avoid getting excluded as a result of the even distribution of the opposing candidates' individual votes - which won't be a problem with the Greens against the Liberals in Bass because Ferguson and Gutwein will dominate the Liberal tally.)

    I think a part of what's going on with these 1:1:1-ish splits is that there is an anti-major-party sentiment that causes voters for 4th parties to be be more likely to preference the Greens than otherwise. Another part is that often in such a contest for preferences some of the big guns for the major parties have been elected with quota, while the highest-profile Green will still be in the mix.


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.