Wednesday, November 21, 2012

EMRS - Libs on course for massive win

EMRS: Lib 55 Labor 27 Green 15 Other 3
Interpretation: Lib 56 Labor 28 Green 13 Other 3
Result if election was held now: Comfortable Liberal Majority (c. 14-16 seats)

Tasmanian pollster EMRS has released its quarterly poll of state voting intentions, available here. (PDF Download).  They have also, very nicely, put up a trend graph since the last election here.  The trend graph shows that following the formation of a Labor-Green coalition after the March 2010 state election produced a 10-10-5 result, there was soon a blowout in favour of the Liberals, who have been in a majority-winning position since at least the start of 2011. (Typically a major party with a statewide lead of more than a quota (c. 16.7%) over the other should be expected to win a majority no matter what the Green vote; winning with a lead of 12% is quite often possible if the Green vote is low.)  Note that since the EMRS figure historically favours the Greens but underrepresents Labor, on all data points on that graph the Green figure should be assumed to be a bit high (typically by 3-4 points) and the Labor figure a bit low by a similar amount. 

EMRS polls must always be treated with some degree of caution because the company traditionally polls notoriously high undecided rates, and also because there has been a report that in one instance last year the pollster asked other questions before voting intention, a practice that can lead to skewed results.  However, I have not yet had further reports of the latter practice, and these latest results are broadly consistent both with a recent ReachTel poll (albeit, one commissioned by the Liberals) and furthermore with another recent internal poll that I have seen that has no whiff of potential pro-Liberal bias.  I have no reason to doubt that the picture painted by EMRS is broadly reliable. 

The headline figures show the Liberals on 55 points (up 6 since August 2012), Labor on 27 (steady) and the Greens on 15 points (down 7).  Both the rise in the Liberal vote and the fall in the Green vote are statistically significant by some margin.  The Liberal vote is at its equal highest level since the suspect poll in August 2011 and the Green vote is at its lowest in this term.  The gap between Liberal and Labor, 28 points, is at its third highest level, and almost back to the levels of August 2011 (33) and November 2011 (31).  The figures including those who are undecided but leaning to a party are Liberal 46, Labor 23, Green 13, Independent 2.

EMRS have not recently been in the habit of publishing electorate by electorate breakdowns, and a breakdown of this survey would be of limited use anyway given how different it is to the most recent survey.  However, a result of this magnitude would almost certainly win the Liberals three seats in every electorate except Denison.  It is possible it would win them a fourth seat in Bass, Braddon or both, or a third seat in Denison.  It is deeply unlikely that all these things would happen even with 55% of the vote, so 14-16 seats is the most likely projection.  If the Liberals were very unlucky, such a vote might conceivably net them only two seats in Franklin and hence only 13 overall, but this would be the absolute worst case scenario on these figures.  (Note that Denison is impossible to predict on any polling as it is not yet known whether there will be a significant independent running.)

For the Greens this poll is spectacularly awful, though it is plausible that the crash is exaggerated by random sampling variation and that they might poll, say, 16, if an election was held now.   It doesn't seem right to me that their vote should have held up not too badly for so long and suddenly crashed 7 points in this polling cycle.  If the results of this poll were repeated at an election the Greens might be cut back to just two seats in Denison and Franklin, but might win more if they got lucky on the distribution of the votes of other parties.  Even allowing for the possibility that sample error is against them, it would be extremely difficult for them to retain all five seats.  Something like 15-7-3 would be plausible if this sample was repeated at an election, perhaps with an Independent taking one of the major party seats if there was a suitable candidate running.

Considering the EMRS sample in comparison to the Liberal-commissioned ReachTEL from October, the Liberals are better by 3.5 points, Labor by 4.1, the Greens worse by 2.7 and Others worse by 5.1.  I suspect that the ReachTEL "Others" score includes some voters who would have been recorded as undecided by EMRS rather than specifying Others, and that the true likely support level for independents (if there are quality independent candidates) is somewhere between the two polls.  That the Green vote is worse in EMRS even though EMRS tends to favour the Greens could suggest one of a few things (i) that ReachTEL's results are also friendly to the Greens and their vote was already falling in October (ii) that the Green vote has fallen since the ReachTEL poll (iii) that the EMRS sample is unduly harsh for the Greens through sample error.

It is always possible, of course, that the EMRS sample is rogue (ie well outside the margin of error) but my observation is that the term "rogue" is too frequently thrown about for polls that turn out to be well within the MOE (if a little generous to someone or other), and that it is better to assume such poll results are within the MOE until proven otherwise.

(A note for non-TT readers who have not been listening to my same broken record on this for the last several years: EMRS distributes their undecided vote proportionally, but at elections the Greens tend to perform no better than their result before this distribution.  This suggests that undecided respondents generally vote for the major parties and not the Greens.)

It is true, of course, that the election is still over a year away and a lot can happen between now and then.  But I greatly doubt the Labor-Green coalition has the capacity to turn its situation around to an extent that would actually deprive the Liberals of majority government.  The person who I think has the most capacity to produce such a result is Tony Abbott, and to do that he has to become Prime Minister first and immediately annoy a lot of Tasmanians.  It remains the case that the most likely question of the next state election is whether the Liberals win a big majority or a small one.  

I will edit in follow-ups to this post if there are any media reports worth commenting on.  The three biggest traditional media goofs in reporting EMRS results are:

(i) making a big deal of the proportion of voters who are "undecided" as if this indicates there is great uncertainty in the electorate.  Actually this results from a combination of the way EMRS presents its data (publishing a figure that treats voters leaning to a party as "undecided" - other pollsters generally don't do this) and the company's difficulties in extracting a meaningful response from voters.  Other pollsters polling Tasmania in the past have typically not had the same problem to anything like the same extent.

(ii) treating changes in party standing that are not near statistically significant as meaningful.  However, there is hopefully not much risk of that one this time!

(iii) referring to Preferred Premier scores as indicators of a leader's "popularity".  They are not; they simply indicate which of the leaders voters would prefer as Premier.  Depending on the popularity of the other leaders and the mood of the time, a leader might poll a good Preferred Premier score while being unpopular (because the other leaders are considered even worse), or a bad PP score while being popular (because the other leaders are considered better).  Indeed while Gillard and Abbott are unpopular (based on net approval ratings), from time to time one will lead the other by a large amount on Preferred Prime Minister.  True popularity ratings are rarely sampled in Tasmania.



Numerous things have been said about this poll by politicians.

For instance, Premier Lara Giddings, as well as taking a sneaky potshot at the Liberals over their curious acceptance of tobacco company donations (probably not really a major factor in their ability to afford advertising, which in turn may not be a major factor in their current polling) has claimed that  "Will Hodgman's own popularity has come down".

For reasons mentioned in (ii) above, Preferred Premier scores do not necessarily indicate "popularity" (and it was good to hear EMRS's Samuel Paske make this point very clearly on 7ZR this afternoon).  For what it is worth, Will Hodgman's preferred premier score of 47 in this poll is his third highest to date from EMRS, behind only the 52 in August 2011 and the 48 in November that year.  So there is only one poll showing him significantly above his current level as Preferred Premier.

It is possible Giddings has access to other data showing Hodgman to be clearly less popular, but if so, it isn't in the public domain.  Generally Hodgman's PP score has been running an average of 2.5 points ahead of his party's base score (with voters leaning to a party redistributed) and an average of 6 points behind his party's headline figure (scarely surprising as PP does not redistribute undecideds.)  This poll shows him 1.5-2 points short of those marks, but not significantly.

Greens Leader Nick McKim has claimed that:

"[The Greens] don’t place a great deal of emphasis on the result of political opinion polls, regardless of whether we’re going up or down"


The Greens were up in the polls a similar amount five months ago, and we went down this time around so what this data really highlights is the inherent volatility of political opinion surveys.”

Now I found this rather amusing, because Nick is a fellow life member of the brotherhood of incurable opinion poll junkies.   He is frequently in a rush to issue a press release when EMRS hits the street and tell us what the poll says about the Greens' performance (while of course, restating that the Greens are not a poll-driven party - which I tend to largely agree with).  For example in May 2012 after the Greens' headline figure increased by 5 points, McKim said:

"The Tasmanian Greens today said the party’s rise in support in the latest EMRS state voting intentions poll reflected a growing awareness of the Greens’ constructive role in Tasmanian politics."

In Feb 2011 after the party recorded two good results in a row he said:

"This poll reflects the fact that Greens have and will continue to work hard to deliver progressive policy outcomes, and contribute constructively to a workable and stable government."

and also:

"The standout from this poll for me is that the Greens vote continues to rise.  I am way more interested in the Greens’ popularity than my own, and anyone who places more importance on their own individual performance rating over that o their party should not be in the position of Leader."

(my bold)

A drop in support for the Liberals in Feb 2012 was responded to thus:

"Perhaps the Tasmanian public are beginning to grow tired of their negativity."

There are so many more such examples.  But when the Greens finally record a result that is really bad, there is no attempt to explain what it might indicate in policy or issue terms at all, and instead it is "inherent volatility".  

Which it probably isn't, since unless this is an outright rogue, the difference is way too large to be explained by volatility alone or even largely, and is proportionally one of the largest changes in a party's polled support in EMRS's history, if not the largest.  Three months is a long time between polls and opinions do change over such a period sometimes.  The average change in the Green vote between polls for the previous ten was just 2.4 points.  Of those ten changes, seven were statistically insignificant, two were barely significant, and the May 2012 rise was the only one that clearly meant anything much by itself.  So that sort of change (five points on the headline) was atypical and this one (seven) even more so.

It will be interesting to see how much the Greens will bounce back next time.  I suspect they will increase their vote in February.   If they are back in the low 20s then this sample was probably just a dud, but any result below 20 will be confirmation they are struggling.

I have not yet seen any official Liberal Party response to this poll.  But when you are leading by this much, it is so much easier to let the numbers do the talking - and so easy to only say things that are actually true if you do need to comment at all.

Update 2

A tad disappointing to see ABC TV news  report the Table 2 figures as the results of the poll when they are not the headline figure and not a form of results that would be reported at all by other pollsters.  (Other pollsters who do report an undecided figure report something equivalent to Table 3.)

Also disappointing, but even less surprising, is that the things Liberal Leader Will Hodgman has said in response to this poll thus far are much more boring than the comments of the other two.  He has declared that "the clear message is that the Labor-Green experiment is failing Tasmanians" (I'd say it is that the voters think it is) and also pressed the usual message about only the Liberals being able to deliver "the strong, stable majority Government that Tasmania needs." Of course, if the Libs were really so obsessed with giving Tasmania majority government, they could have just campaigned as badly in 2010 as they did in 2002 when they did a stellar job of ensuring such an outcome! 

Thanks to Zoe Edwards (@zedwoods) for tips that have assisted me to comment on this poll release quickly.


Note to TT posters: As already noted I have left Tasmanian Times as a poster and commenter although I still encourage my work to be linked from TT and selected excerpts to be published there.  If you want to ask me questions and get a response you will generally need to do so here, although if I find a TT comment especially interesting I may write an article in response to it.  You will need to have (or to create) a Google/Blogger account, LiveJournal profile, WordPress account, AIM profile, TypePad account or OpenID to comment.  If you find this impossible, you can email me your question or comment (link to email here) with a clear statement of permission to publish it.

No comments:

Post a Comment

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.