Tuesday, February 19, 2013

EMRS - Liberals locking in support

 Note to Tas or other interested readers: if you haven't already done so please vote in my not-a-poll for Best Tasmanian Premier of the last 30 years on the sidebar.  Looks like the only poll Tasmanian Labor can still be confident of winning!

EMRS: Liberal 55 Labor 23 (-4) Green 18 (+3) Ind 4 (+1)
Interpretation: Liberal 56 Labor 25 Green 15 Ind 4 (Ind figure depending on candidates)
Outcome "if election was held now": Liberal majority win (14-16 seats)

The new EMRS poll is out with a headline rate of Liberal 55 Labor 23 Greens 18 Ind 4.  This is very similar to results polled in late 2011. The Greens have recovered some of the seven points lost last time (a loss which I suspected at the time was caused partly by sample anomalies) but it is still their equal second-worst result in this term of government.  The EMRS headline figure habitually overestimates Green support and the figure from the table with undecideds included (Table 2) tends to provide a much better reading of Green support. 

The very helpful trend tracking on the EMRS website shows that this is the Liberals' equal highest headline reading in this term, matching the 55 in August 2011 and the same in November 2012.  But it is actually better than that for the party because their figure including undecided votes (46 - Table 2) is now at its highest level in EMRS history, compared to the 44 in Aug 2011 and the 43 in Nov 2012.  This suggests the party is increasingly "locking in" the votes it needs to win majority government.  Labor, on the other hand, is if anything shedding votes from the firmer end of its support base, with its Table 2 figure back to a miserable 17%, one point above its all-time low.  The gap of 29 points at the firm end of the parties' support bases is the highest it has been and about twice the size needed for majority government.  While I have assumed in my "interpretation" figure above that the voters who are undecided even after being asked what party they are leaning to will split evenly between the major parties, it is possible it could be even worse than this for Labor.  If the remaining undecided voters break to the Liberals as well then a Liberal vote in the high 50s becomes possible.

While the changes since the last poll are not massive, the apparent fall in Labor support is not at all surprising and close to statistically significant (see below).  In the last few months Tasmania has experienced bushfires that burnt through the town of Dunalley and in all destroyed or seriously damaged over 200 houses.   While there has been praise for the work of the Tasmania Fire Service in fighting and documenting the fires, the government's disaster relief response at the time was widely seen as somewhere between lethargic and invisible, with much of the co-ordination being conducted by volunteers via social media.  There has not been much good news for the government on other fronts either, with the ongoing forestry peace deal process not yet looking much closer to the resolution that a lot of voters probably don't want anyway.  The Legislative Council inquiry into the deal has now heard from so many voices that are deeply critical of the deal and the process that created it, that the LegCo now has plenty of ammunition if it wants to argue a case for just throwing the whole thing out.

The especially telling thing about the government's bushfire response is that governments have often benefited - in the short term - in the wake of natural disasters.  The former Queensland Labor government of Anna Bligh enjoyed a massive surge in the wake of its acclaimed handling of major floods and cyclones, although by election time the bounce had disappeared (as bounces do) and the government received a historic thrashing anyway.  To not even be able to gain a temporary bounce by being perceived as performing usefully in the wake of a natural disaster speaks volumes for the extent to which voters regard the current government as useless.

Modelling EMRS results in terms of seat outcomes is very difficult because of a scarcity of electorate-by-electorate data since the company stopped publishing electorate breakdowns after its May 2012 poll.  In my article Uneven Swing to Liberals in Tasmanian State Electorate Polling I found that what little electorate data is available from 2012 (some of it now a year old) points to the government's vote holding up much better in Franklin than in other electorates.

This is what the current poll looks like if I apply the same assumptions and the same findings about uneven swing as in that article:

This shows that a result of 16-5-4 is, in theory, possible from the current results.  That would, however, be a horribly unlucky distribution for Labor, which would be left just short of second seats in several electorates.  It's more likely the distribution would be slightly different and they would save one or two of the five at-risk seatsa somewhere.  And again, with three of the Liberal wins projected above being relatively narrow, there is still no guarantee that the party wins more than 13 seats, while even 17 seats is possible if the distribution is extremely lucky.   The improvement in the Greens' polling suggests they now have a better chance of limiting their losses to one seat, but with no guarantee of winning more than two.

I also note that the figure for Labor in Bass that this uneven-swing-based projection is producing is starting to look implausibly low.  However, in 1992 in similar political circumstances, Labor polled only 20.8% in Braddon despite having the Premier on their ticket, so it's quite possible Labor will be reduced to the low 20s in one northern electorate or other if it cannot pull its socks up. If you don't believe that a mid-50s vote for the Liberals could lead to them winning four seats in an electorate somewhere, check out the 1992 figures - under the current system, they would have won four in Braddon very easily.

This is, again, just another poll that doesn't tell us the result of an election that is still expected to be over a year away.  But the signs remain extremely bleak for the Labor-Green coalition which has now made no net progress in improving its vote share in a year and a half, while the Liberal opposition's vote share appears to be becoming more secure.  The gaps may close as the election approaches but there is still no evidence to suggest that any result other than a Liberal outright win is even remotely likely.

It should be noted that there are not, as yet, any known significant fourth party or independent candidates for the upcoming election.  Should some emerge it is possible they will have chances to win seats (especially in Denison) and probable that voting intention for them will be greatly underrepresented in EMRS's polling.  However, with Andrew Wilkie still in Canberra and no other proven fourth-party performer known to be running, there is still no basis for assuming any specific fourth-party candidate will be a threat. 

Finally, while preferred premier scores are of little value, some may notice that Will Hodgman's PP score of 46 is actually not a record high and that his score is level with his party's score on Table 2, while Premier Giddings' PP rating of 24 is much higher than her party's standing on the same table (17).  A substantial advantage to incumbents in preferred leader scores is normal and I suggest that these figures are neither good nor bad news for either leader given their party's overall position.

This article will probably be updated with comments on media coverage and political responses.


Update: I often take media to task for reporting results as significant when they are actually within the margin of error of a previous result, so it's a bit naughty of me to have to now comment on a case where a result is described as being within the MOE when it actually technically isn't!  As you can see if you play around on the fabulous Poll Cruncher a change of 3 points for a sample of 1000 is within MOE if the base result is 50.  But with a smaller base result, the MOE drops.  So for a previous result of 20 for Labor (Table 2), a drop of 3 points to 17 is outside the MOE for that result, which was about 2.5 points.  This is a rather tricky concept to grasp, and EMRS should give 3.1% as the maximum margin of error in their reporting.

Note also that a change outside the margin of error of a previous poll is not necessarily quite "statistically significant", as you can see if you enter some of the values in this case (20% from 1000 and 17% from 1000).

Update 2:  Apparently it is very important to the Greens to respond to every EMRS poll almost as quickly as I do, so much so that with leader and poll junkie Nick McKim out of the state, Tim Morris has been given a go, and hath said:

"Pleasingly this poll suggests steady support for the Greens over the past couple of years with fluctuations within the margin of error for this poll."

Now, it's true that most of the Greens' headline results since May 2011 fall within MOE of an average of 19.5, or near enough to it, and one of the ones that fell outside that range was the last one, which was probably a bad sample.  However given that that is still below the 2010 result and that EMRS has habitually overstated the Greens' vote, I'm not entirely sure how pleased they should be.  Probably some swing against them next time was always inevitable, and to be polling figures that give them a fair chance of holding four seats is something they would see as better than it could have been.

Labor Deputy Premier Bryan Green has issued a quite amusing press release trolling the Liberals.  I have long wondered whether Green's heart is still in it given the uncomfortable political position he has found himself in in this term, but at least he still has the spirit for a stir.  I think that if the Liberals had intended to influence the outcome of this poll with their advertising they would have been quite pleased with this result, although there was no gain on the headline figure. That said, I don't believe their advertising had much to do with it.

Update 3:  Not specifically connected with this poll, Bryan Green has just responded  to a recent call by perennial union voice Kevin Harkins for him and  Michael Polley to stand aside.  The basis of Harkins' call is that Labor's fresh young talent, elected in 2010 (Rebecca White, Brian Wightman, David O'Byrne and Scott Bacon) could be wiped out in 2014 if the oldies recontest, presumably because they would hog the seats.  But a seat-by-seat comparison shows this doesn't really stand up to scrutiny in the case of Green (who would be a threat only to career backbencher Brenton Best if Labor lost one in Braddon), nor in the case of Bacon (whose only fellow incumbent is Graeme Sturges who has already lost his seat once and would do so again if Labor was reduced to one).  O'Byrne's ticketmate is the Premier (who is hardly dead wood at 40), and Wightman's is Michelle O'Byrne (who wasn't mentioned in the Harkins call).  That reduces the scope of the Harkins call to Polley vs White in Lyons, where there is a risk that Labor might lose White if they were to be reduced to one.  So while it has merit in one case, for the most part Harkins' claim misses the mark, and should probably be interpreted in the context of the brief leadership cold war he conducted on behalf of David O'Byrne.  O'Byrne was not publicly involved in the rumoured challenge, which never actually came to a public vote, but Polley intervened in Giddings' favour to pour cold water on media speculation.

 Spammy PS: If interested in the upcoming Tasmanian Legislative Council elections, you may wish to check out my challengers guide, which is updated as new candidates emerge and is now my most visited Tasmanian article!


  1. Is it reasonable to say that this at suggests a 4:1:1 outcome for the federal senate elections later this year? 55 is not far from the 57.1% required for 4 quotas.

    Is there data comparing federal and state intentions in Tasmania?

  2. Tasmania often votes quite differently in state and federal elections. There is very limited evidence on federal voting intention for Tasmania, and what there is relates to the House of Reps, with the Senate being almost impossible to poll accurately. The suggested swing against Labor for the House is almost as large as the swing in state polling, but that is off a much lower base for the Liberals than the state vote, and that makes all the difference in terms of the chance of 4-1-1.

    In 2010 the Liberals polled 33% in the Senate for Tasmania, Labor 41.4%, the Greens 20.3%. Liberal feeder parties polled about 3%.

    For the Reps the figures were 33.6% Liberal, 44% Labor, 13.8% Green, 5.9% others (including Wilkie).

    Four Senate seats probably requires a swing of 21% from ALP+Greens to Liberals. What little federal polling there is has pointed to a swing of 10-14% with the recent Bass ReachTEL giving 17% for that electorate (though I think it is really more like 15% as I think some Labor votes are hiding in the Others pile).

    I think 3 Lib 2 ALP 1 Green is an extremely likely Tasmanian Senate outcome at this stage. 4-1-1 is very unlikely but if things get really nasty for the non-Lib forces it can't be ruled out entirely.

    In this I'm assuming Labor and the Greens preference each other ahead of the Liberals as per normal.

  3. 3:2:1 is certainly more likely; barring, as you say, further disasters from the Alliance. I expect a reluctance to hand senate control to the Liberals still persists despite the huge swing towards them in the lower house.

    As to Kevin Harkins comments, one of the beauties of the Hare-Clarke system is that the people get to choose at what point the old gives way to the new. There's no point not running the older candidates. When time is up, its made clear at the election.

    The problems with our system here are by and large not with the lower house, which I would suggest fulfils its role of representation of the diverse peoples of the state better than any other system in the country. It requires a fourth party to provide stability; one will inherently appear in due course.


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.