Tuesday, August 18, 2015

EMRS: Surprise Slide For Liberals

EMRS: Liberal 40 Labor 29 Green 21 Ind 9
Interpretation (provisional): Liberal 42 Labor 33 Green 18 Other 7
Seat distribution based on poll: Hung parliament (Liberal 11-12 Labor 9-10 Green 4)
Aggregate of all recent state polling: Liberal majority (13-9-3) slightly more likely than not

A new EMRS poll of state voting intentions has been released (also see the tracker) showing a surprising six-point slide for the Hodgman Liberal government from the May sample.  The poll result is surprising mainly because, unlike last year's indifferent polling, the government does not seem to be in any particular trouble.  It is also against the run of play given a very strong result for the government in the large ReachTEL in June and also strengthening results in the small Morgan state samples.  Therefore, the result should be treated with a fair bit of caution for now.

The poll features a surge in voters stating they will vote "Independent" to an EMRS record nine points, and also a very high Green vote.  Overestimates of both these categories have long been a feature of this pollster's results, but I am wondering whether the pollster has become even more prone to overestimate Greens and Others votes, as the old landline-only Newspoll did in its final months.  That said, the Greens are presently on a national surge and this may be rubbing off in Tasmanian polling.

The poll does not show any lift in support for Labor at the government's expense, and suggests that soft supporters of the government may be parking their vote with third parties, something which could well be driven partly by federal as well as state factors.  However there is absolutely no reason to believe that federal factors are responsible for all or even most of the change.

I am also wondering whether EMRS's sampling is fully random across the whole population. I'm wondering that mainly because I was again included in the sample, the second time I've been sampled in the six polls since the election of this government, and having also been polled recently by EMRS commissioned polls.  With the increasing non-connection rates for landline polling making it much more expensive it would be tempting to use a partly or entirely panel-style method in which preference is given to numbers with known successful connections.  EMRS do not explicitly state that their sampling is random across the entire population.

Voting Intention

The raw voting intentions after excluding undecided voters are Liberal 40 (-6) Labor 29 (=) Green 21 (+3) Ind 9 (+2); rounding prevents totals adding to 100.  This is the worst result for the Liberals in this term in office, and indeed their worst since August 2010.  As already noted it's the highest for "Independent" ever.  It's the equal highest for the Greens in this term; they were last higher in August 2012.

As noted in previous EMRS articles I apply various changes to EMRS raw figures based on consistent experience at past elections that the pollster overstates the third-party vote at the expense of the majors and especially Labor.  In some cases this will turn a poll that points to a hung parliament into one that points to a Liberal majority, but not this time.  Based on my interpretation methods and what data we have on electorate-by-electorate sampling, here is a possible statewide model for this poll:

The Liberals as usual lose their fourth seat in Braddon and their third in Franklin but this time they also lose the crucial third seat in Lyons to the Greens, and possibly their third in Bass as well.

I repeat the usual caution that many Tasmanian voters vote strategically to avoid hung parliaments, except when they cannot work out how.  So if the Liberals went into a campaign leading in polling by eleven points and clearly the only party capable of winning outright, I would expect their lead to increase during the campaign.

State aggregate

We shouldn't read too much into just one poll and that's why I've been trying to aggregate Tasmanian polling (a difficult task because polls are infrequent and not very reliable, every pollster has substantial house effects, and in all cases it's possible that the house effects change from time to time).

My aggregate was last updated for the small Morgan sample last week, which showed the Liberals on 44, Labor 31, Green 20, others 5.  That was a strong result for the Liberals given that the Morgan series has at times had them behind, and has seemed to be skewed against them by four or five points. The new EMRS suggests the Morgan Tasmanian samples might be becoming less skewed (my new estimate is that they skew to Labor and the Greens by two points each, and against the Liberals by 3.5).

Given that there was a ReachTEL in the last few months and a Morgan very recently I have weighted this EMRS at 35%.  Here is the new aggregate:

Braddon and Franklin, you know the drill.  In this version of the aggregate the critical third seat in Lyons is very close to being lost to the Greens.  I am continuing to treat that with caution because of unfavourable third-party preferences and the Greens lacking an incumbent, so although that seat is precarious for the time being the Liberals retain it.  The one change to the aggregate is in Bass, where although the Greens are projected to trail Labor slightly on quota, Labor has only one incumbent, so I am now giving the Greens the benefit of a great amount of doubt to hold on in Bass.  Therefore the aggregate changes from 13-10-2 to 13-9-3, and although one poll has pointed to a hung parliament, my aggregate still shows a Liberal majority as slightly the more likely outcome if an election was "held now".


The poll shows Hodgman (49) leading Bryan Green (21) as preferred premier with 30%, the highest since this two-leader series started last year, undecided.  This is Green's worst result so far and is further evidence that this poll is hardly showing firm enthusiasm for Labor.

I will update with any notable public responses to this poll I come across, noting that I am trying (and probably failing) to write two other articles today!


  1. Is there anything to suggest that the Greens could win a second seat in Denison? If not, at what percentage of the vote could it be suggestible? At the moment would a result in Denison most likely be 3 Labor, 1 Lib and 1 Green

    1. Denison would still be 2-2-1 on this sample after adjustment for EMRS's house effect in favour of the Greens. About another 3.5 points off the Liberals and to the Greens would make Denison go 2-1-2 all else being equal.

    2. I suppose that's the origin of that "True Liberal Values" tactic from the Greens in Hobart.

    3. Well over time in Denison, the fifth seat will come down to a 3rd ALP - 2nd Lib - 2nd Green tussle, on current polling in would suggest 2 ALP - 2 Lib - 1 Green, but if the polls continue to slide for the Libs i believe by the next election state election 3 ALP - 1 Lib - 1 Green is quite plausible. Also the perception from politics in Melbourne is that Greens in Tassie tend to be wet libs rather then the Labor dissenters in Victoria and the hardcore trots in NSW.

    4. My perception is that core Greens voters in Tas are just greens - they are mainly driven by environmental concerns, rather than refugees from any other movement in particular. "Wet libs" will readily vote for Green-ish independents (eg Wilkie) but seem very reluctant to support the actual party.

    5. As well that is another variable to add in Denison, especially in state election. Wilkie does have quite a large following and as to where his vote transfers could be anywhere, to all three parties mostly Labor and Greens predominately. If Wilkie did back independent candidates in the vain of Xenophon it wouldn't be crazy to believe that 1 or maybe 2 in Denison is elected or even 1 in Franklin as a lot of the booths are in the proximity of Hobart.

    6. Wilkie has hitherto avoided backing candidates at state level, and denounced an independent who stressed similarities with him last campaign. However, he has backed candidates for Glenorchy Council. Because the Greens tend to poll over a quota in Denison there are always loose votes going around for an independent who can get a big enough primary vote. It didn't happen last time but sometime it may happen.

  2. Would this be a good time to hand over the ALP leadership to Rebecca White? The advantage the Libs have is that many voters will want to avoid a hung parliament so if the polls are showing them on the same sort of figures as this close to the next election there would be a drift to the Liberal Party.

  3. Too early to change leaders in my view.

  4. I notice at the 2002 state election the Greens returned 3 new people to parliament: Mckim in Franklin ( party swing of +10% ), Moriss in Lyons ( + 7.1% ) and Booth in Bass ( +7.4% ). The statewide swing to the Greens was +7.95%. Are you using these figures to determine what sort of vote will be required for Lyons and Bass to be recaptured by the Greens based on the statewide polling results? You keep saying you are reluctant to give away more seats to them. As shown by Mckim in 2002 a new face can outperform a general statewide swing to the party, and I would think there would be a lot of variables involved. I'm just wondering what your formula is.

  5. I don't have a formula as such. Neither seat is about any specific percentage, beyond that if the Greens get a quota (16.67%) in their own right in a seat then they just about certainly win. Bass is currently about competition with the second Labor candidate, though if Labor's polling improves it could become about competition with the third Liberal. Lyons is probably about competition with the third Liberal, though 3-1-1 isn't impossible either.

    Taking into account incumbency issues and preference flows, I think the Greens can beat Labor in Bass from slightly behind Labor's remainder after their first quota (as Kim Booth did in 2006), but I think the Greens need to be well ahead of the remainder of whatever party they are competing with in Lyons to win. Indeed even with incumbency benefits, starting roughly level in 2014 did Tim Morris no good at all.

    The scale McKim's win in 2002 was probably partly down to being a new and different "kind" of Greens candidate at the time; such candidates don't exactly grow on trees. However, there were other causes, including:

    * The weakness of the Liberal ticket which had only one recontesting incumbent, who had been an MHA for less than a year.
    * The swing to the Greens was from 1998 when the incumbent was Mike Foley, who was never all that high profile.
    * There was a fair amount of demographic shift to the Greens in Franklin through the 2000s mainly through the "treechanging" boom in the Channel area.

  6. Any thoughts on the latest state breakdown from Morgan? samples sizes are pretty small, but it seems to at least back up this EMRS poll that the libs are sitting around 40% on Primary.

  7. Even aside from the small sample sizes I don't take Morgan's state polling all that seriously - in NSW and Victorian elections it was far too volatile and in Tasmania it seems to have serious house effects against the Liberals and to Labor and the Greens. (My current estimate is that it is skewed against the Liberals by 3.5 points; earlier that estimate was as high as 5).

    But it is slightly interesting that the sample has the gap between the major parties at only 2.5 points when in the previous months Morgan had the gap at 5.5 then 9.5 then 13. Even if house effects mean that that 2.5 point gap in this sample should be treated as at least 8 points, it does provide some very weak supporting evidence for the EMRS finding that the gap between the majors has narrowed again.

  8. Something else to add re the Morgan: it is the first post-Mantach data. While I wouldn't expect the Mantach saga would have a vast impact on voting intention in Tasmania, Morgan polling often seems to be very reactive to news-cycle events.

  9. Tasmania is too difficult to predict, as people try to game their vote to avoid a minority government. This and a relatively unpopular Labor lead means it is likely even at polling in low 40%, they would scrap a majority. I am not sure why they are polling this low though, I am no Liberal support and have some personal circumstances to not to support them, but holistically i would not have said that there has been anything too negative that would drop their polling numbers. Be nice to see another ReachTel poll here.