EMRS Aug 2011: Liberal 44 Labor 16 Green 14 Ind 4 Undecided 22
Interpretation withheld – poll provisionally rejected

This article was to begin as follows:

If a state election was held right now, the Liberal Party would win outright with as large majority as is likely to be seen under the Hare-Clark system. The previous two were terrible enough but the latest EMRS poll shows a six point swing to the Liberals compared to the May polll results, and the question is not whether the Liberals are in majority territory now, but how much they would win by if an election was to be held. Labor could halve the Liberals’ current lead and still lose an election outright.

But I am not sure whether we can really conclude that too strongly, because I have very limited faith in the reliability of EMRS as a pollster – even taking into account the foibles we already knew about many times over such as the large undecided rates and overestimation of the Green vote. A post by switched-on young Labor scrutineer (and 2010 Wightman campaign staffer) Adam Clarke to pollbludger reports questions he was asked in the poll and states that voting intention was not the first significant question asked. Instead by the time the respondent got around to answering questions about voting intention they had already expressed views about their expectations of standard of living, their expectations of their own employment status and their expectations of the state economy, as well as listing social and economic issues of importance.

This in my view, if Clarke’s account is accurate (and despite his party background I don’t doubt him) is reason to consider this poll’s findings to be suspect and to reject this poll and all future EMRS polls until it is publicly clear that voting intention will always be the first significant question asked. If respondents are asked to think about general state economic issues at a time of economic gloom before being asked their voting preference, they are more likely to focus on the (many) negative perceptions of the current Government and prefer the Liberal Opposition. Likewise if the first three questions canvassed the respondent’s opinions on rainforests, whales and the virtues of organic chai before asking about voting intention you can bet the Green vote would be through the roof. In contrast I was polled by Newspoll in the last month and voting intention was the first question asked, as it is for Essential Report and other credible polling sources.

I have been railing for years against activist-group-commissioned polls that somehow seem to always slip in a potentially distorting preamble or a preliminary question that might somehow “train” the respondent’s mind (as parodied in this classic Yes Minister snippet, but even now it still goes on abundantly). While I am certain it was not deliberate in EMRS’s case (can’t say the same for the activist-commissioned examples) it is still not good enough.

The rest of this article is what I originally wrote before I found out about the method issues. The comments about comparisons with the previous poll still hang together, but those about projected outright results need to be treated with caution since the poll’s method defects may mean it is slightly exaggerating the government’s plight:

Again the (misleadingly low, but even the true figure is likely to be dreadful) EMRS headline rate is the worst the Government has ever recorded, and the Liberal lead is up by nine points from the previous poll, which already represented their record to that date. In the last poll, the Liberal vote exceeded Labor and Green combined for the first time, and now it exceeds it by fourteen points. The preferred-premier index (not the most useful indicator) shows Will Hodgman gaining a massive ten points in three months while Lara Giddings is only just above the figure at which Paul Lennon decided to call it a day. This poll is significantly worse for Labor than the one before it, which means it is significantly worse than merely terrible.

At times like this it is always important to remember that Tasmania is very sparsely polled, and with only one poll each three months there is no way to be sure a poll like this is accurate. But even imagining that this is a “rogue poll” and giving Labor the benefit of more than the margin of error, it still doesn’t help. And polling error can go in either direction; just as we should consider the possibility that the picture is really not so bad, we need to consider the possibility that it is worse – or that it will still get worse.

Normally I batch two consecutive EMRS polls to get an idea of likely electorate-by-electorate results. I have not done that this time because the August samples are significantly different from the May samples and therefore any such exercise will understate just what a bad result this is. However even using such weighted samples the Liberals clearly get their three in the electorates of Bass, Braddon, Franklin and Lyons and probably do so in Denison, although they are shy of their third quota there and it would be quite likely their third seat would be snaffled by Andrew Wilkie if he were to lose his federal seat before the State election and contest the latter. But as a worst case for the Opposition, the current poll gives them 14 seats out of 25. Even more remarkably, the Braddon sample in the current poll, if reproduced at an election, would probably win the Liberals four (!) seats with Labor and the Greens each losing one. However, the sample size is too small to treat that as a dependable outcome unless it is confirmed in a subsequent poll. 3-2-0 is more realistic for the time being, so we could take this poll as pointing to a 15-6-4 or 14-6-4-1 win for the Liberals. With only one Labor seat on the table in Franklin, it is an open question whether Giddings would retain or be defeated by David O’Byrne.

Of course, it should be a long way to the next election, and furthermore this daddy of all shockers comes following a particularly putrid polling cycle for the Government. For the last three months it has rarely been able to get forestry (both the “peace deal” process and the pulp mill) off the agenda, and the Government is caught in the middle on these issues with its actions impressing nobody. To opponents of the industry the Government is paying off an unprofitable pursuit to stop doing what it shouldn’t be doing (and is commercially choosing not to do) anyway, and continuing to throw lifelines to a shambolic, clearly dead pulp mill fantasy run by a moribund company whose share price is trending towards zlotys. (Peter Henning has done a nice job of forking the government’s illogic on the issue of pulp mill permits – if nobody knows for sure what “substantial commencement” really is then why was legislation containing such a term enacted in the first place?) To industry supporters the Government is supporting a shoddy cheap fix that buys out the industry for pathetic compensation in a gloomy employment marketplace without even delivering lasting peace, following political processes that have no credibility and are driven by implied balance-of-power thugging by the Greens both state and federal, assisted by their pet ENGOs and the endless ability of the environment movement to hedge by having some components inside the tent and some out of it. When the government is copping such justified critiques from all sides of the argument it is no surprise that its standing is in tatters.

Meanwhile the Green vote is also now falling, with disenchanted hardline supporters starting to sniff around for viable fourth parties, and soft Green voters who in 2010 picked the Greens as a “neither of the above” starting to see the Greens as part of Government and just another major party to be rejected. What is significant here is that some part of that soft Green vote seems to be deciding that chucking out this Government is a higher priority than giving the thumbs down to both majors as in 2010, and some of these votes, instead of looking for their local Andrew Wilkie, are going direct to the Liberal Party.

There is a danger here that we could end up with a NSW-like situation where the electorate decides it is going to evict the government and becomes so fixed in that resolve that the government is unable to recover to even a respectable loss and a hiding becomes inevitable.

It would be very nice to be able to draw those sorts of conclusions and make such comments more firmly on the basis of a measurement of public opinion I was confident of. But in this case I cannot and I must also add a similar degree of caution to my views in the previous article as well, and ditto for all recent articles dealing with EMRS polling.

For those who really want to know what I think this poll would translate to aside from the question order method problem, it would be something like Liberal 55 Labor 26 Green 15 Other 4. But with a cloud over the poll’s methods, it is tempting to shave at least a couple of points and perhaps more from that massive Liberal result.

EMRS are not currently up to scratch, have never been particularly good, and it is time for Tasmania to get a new pollster. Failing that, you will not be hearing much more (except complaints!) about state opinion polls from this here column.

First published: 2011-08-31 07:22 PM