Image for Swing voters mistrust Labor-Green coalition

EMRS May 2010: Liberal 38, Labor 23, Green 24, Ind 3 Undecided 12
A Possible Interpretation: Liberal 42 (+3 since election) Labor 32 (-5) Green 24 (+2 since election) Ind 2 (=).

Better late than never, I hope.  Your columnist has spent the past couple of weeks on fieldwork in often-remote portions of the northern Queensland electorates of Leichhardt and Kennedy, leaving the day the last EMRS poll was released and returning a few days ago.  The first EMRS poll since the election appears to be a very useful one, and we should be glad it has been released.  However, as usual, working out what is really going on is not an easy task.  The headline figures ( available HERE, PDF Download ) show the Liberals on 38 points, Labor 23, Greens 24, Ind 3 with only 12 points undecided.  The figures with those “leaning” to a party redistributed, a set of figures more consistent with results of other pollsters, are 40-25-25-4-7.  It is pleasing to note that respondents were distributed evenly between the five electorates.

We should not read too much into these raw figures as a sign of where the parties really stand.  In the February EMRS poll Labor supposedly led the Greens by only one point on the headline figures, and in the March poll taken very shortly before the 2010 state election, Labor supposedly trailed the Greens by a point. With “leaning” voters included the gap was supposed to be 3-4 points. Yet at the election itself, despite a poor campaign, Labor outperformed the Greens by 15 points.  Likewise, the Liberal-Labor gap, polled by EMRS at 7 and 8 points on the headline rate in its last two pre-election polls and 5 and 7 points on the rate including “leaning” voters, but the eventual gap was just two points.  The ultimate opinion poll – the election – showed again, as in 2006, that EMRS are just not capable of getting a handle on the views of the soft end of Labor support.  By comparison, the one and only Newspoll of the campaign also overcooked the Greens’ vote significantly, but got the gap between the majors extremely close to correct. 

Therefore, any reporting of this poll that claims that it shows the Greens to now be as popular as Labor is rubbish.  We know from experience that at an election, the Greens typically get no more than their EMRS baseline rate, and that the “undecided” voters plus those who claim they are leaning Green tend to vote for the major parties and in particular for Labor.  In the 2006 election the break of the EMRS-undecideds (headline rate) was something like 85:15 to Labor, while in 2010 it was more like 65:35. So if an election was held now, it is very likely Labor would again outpoll the Greens by several points.  My estimate is about eight.

What has changed radically from the February and March polls to the present one is that the EMRS headline undecided figure has dropped from absurdly high figures of 23% and 26% to a mere 12%, by far the lowest figure ever recorded by EMRS except in the runup to the 2002 and 2006 elections.  On the figures with “leaning” voters included, however, the “undecided” result is not much different to the March 2010 figure (7% compared with 8%).  It is not known whether EMRS have made any method changes that allow them to better capture the preferences of undecided voters, but I am not aware of any evidence (beyond the low figure itself) that they have.  What may well have happened is that the Liberals are now gaining clear support from most of those who were leaning towards them before the election and voted for them then (about six points), and those who decided to vote for them late in the piece having previously flagged themselves as undecided or perhaps even leaning Green (about four points).  They may well have gained a little more beyond that, because it is hard to believe that none of the remaining points of “undecided” would vote Liberal, although it is possible that all those who are going to express a pro-Liberal preference are now doing so and that the Libs have only really gained in strength of support, not significantly in amount of it.

What I think is going on here is that swinging major party voters are wary of the new Labor-Green coalition.  As noted before, trends over the past 20 years suggest that about a fifth to a quarter of Tasmanians swing fairly readily from one major party to the other and back again in order to avoid hung parliaments, only really voting according to their ideological leanings at elections like 1996 and 2010 where no party seems to have any real chance of a majority.  At the 2010 election, both parties made similar noises indicating a reluctance to deal with the Greens, but the Liberals stood by those noises while Labor claimed the result showed a need to adjust to new realities.  Majoritarian swing-voters who voted Liberal would feel that the formation of a coalition between Labor and the Greens has confirmed their decision as the right one and strengthened their desire to elect a majority Lib government in 2014 (or 2013, 2012, 2011 …).  Majoritarian swing-voters who voted Labor will be decidedly nervous about what is happening, both in terms of the potential for instability and in terms of the risk of too much of the Greens’ agenda being adopted.  Most of these voters will now be claiming to be undecided and some may even now be supporting the Liberals.  The burden of proof will be on Labor to deliver a government that is stable and prosperous if it wants to retain the faith of those who voted for it in 2010.  If it fails, the Liberals will win a majority at the next election.

For the Liberals this poll would come as a substantial relief.  They seemed outmanoeuvred and foolish in the processes leading to the formation of the new Bartlett government, but this poll shows that their support has at least firmed and perhaps increased since then.  Problems will arise for them if the new government is productive enough that any criticism of it can be met with “well, you could have been here and all your fearmongering about the Greens and hung parliaments was nonsense” but the initial response to their refusal to deal with the Greens has been positive. 

For the Greens, the poll is interesting because it shows that the kind of “sellout” backlash seen from some Greens voters on this site is a fringe view even by Green standards and that the mainstream of Greens voters have great goodwill towards both Nick McKim and his decision to lead his party into coalition.  Green voters appear to be pragmatic about this choice -  they think that in principle it was the right thing to do, and they are willing to wait and see what it delivers.  But if after a while it turns out that the answer is not much, or if the curious nature of the partly-Government partly-crossbench Green lineup generates internal tensions, that goodwill could well be tested.  It is extraordinarily lucky for the coalition that in the first few months of its existence, the issue that most threatened to divide it (forestry) has receded for the time being with forestry companies and agencies experiencing collapses, financial losses, and board transfers, so that the Bartlett government is free to take a less than gung-ho position on forestry matters without the backlash it would normally receive.  But how long this lasts is anyone’s guess, especially if the “round table” process is unsuccessful.

Not too much can be read into the very small seat-by-seat breakdowns provided by EMRS given that this is the first poll after an election and given that these breakdowns distribute the undecided voters proportionally, a practice that has been demonstrated over and over again not to work because undecided voters don’t vote Green.  However there are some significant things going on here.  The first one is that there is more than a blip on the radar for Andrew Wilkie now.  While an 11-point result does not prove his support has increased since the election (a 200-vote sample has a margin of error of about seven points), it at least shows that at the moment, he is sufficiently on people’s minds to be mentioned as an alternative.  The second notable issue is that comparing the EMRS figures to the election, the Liberals have gained most in the most rural and traditional seats, Braddon and Lyons, where distrust of a coalition with the Greens would be naturally strongest.  The Greens are up more or less in proportion to their election result in each electorate, which suggests a small across the board increase amplified by the incorrect practice of redistributing undecideds evenly.  In Lyons (the one rural electorate without a low Green vote) Labor is supposed to have dropped 17 points since the election, but four are to the non-existent independents, while another 2 points or so are EMRS’s “house bias” to the Greens.

It is well worth looking at the last election results to see where the early battlelines are drawn in terms of the next election.  The Liberals need to gain three seats to form government.  They win Braddon on about a 1.8% swing from Labor or a smaller swing from the Greens (if the swing from Labor to Liberal is much larger, they may instead win it from Labor with the Greens retaining).  Their other realistic targets are Bass, where about a 4 to 4.5% swing is needed depending on leakage and Green preferences, and Franklin, which is a little bit tricky because a swing just over 3% can knock Labor down to one seat, but may give the Greens a second (if they run a better #2 candidate!) instead of the Liberals a third.  To avoid this the Liberals would need a bit more, so most likely the target for the Liberals is a 4.5% swing in the right seats for victory.  On this poll, an election held right now would deliver them an eleventh seat in Braddon and make them the largest party, and might give them three in Bass if Labor also dropped votes to the Greens, but would not be likely to give them the third in Franklin.  However, with three seats the Liberals need to win essentially “marginal”, no-one should assume a hung parliament to be a permanent feature of the new state of Tasmanian politics.  If things go wrong for this government, the Liberals are very much within striking distance.

For Labor, the path back to majority government is to reclaim three of the four seats lost last election.  But the seat by seat picture is far more difficult for Labor than the Liberals – Bass (where they didn’t lose any seats anyhow) and Franklin look like mission impossible, and they need about a 6% straight swing in Lyons and about a 7% straight swing in Braddon.  That leaves Denison, which is very complicated because of the Wilkie factor, but Labor would be lucky to win it back without picking up at least 10 points from somewhere.  All up, it will be extremely difficult for Labor to win majority government at the next election even if the new government is a resounding success.

Mainstream and political reaction to the latest poll was for the most part over the top.  ABC Online claims a dramatic crash in Labor support since the election from 37 points to 23 but fails to note that EMRS always understates Labor support, especially in its headline figures in which voters “leaning” to a party are wrongly classified as undecided.  The article also notes that the Liberals believe they would win outright if an election was held now, but in fact this is fairly unlikely.  Angus Livingstone in The Examiner suggests Labor could win just seven seats if an election was held now.  In fact if you take the EMRS seat by seat breakdowns as gospel, Labor would win just a miserable five (!) and the Greens would become the Opposition to a Liberal majority government.  However these breakdowns are not only way too small to be meaningful but are notorious for overstating the Green vote, sometimes by double-figure margins.  More likely outcomes of an election held right now would be 9-12-4, 9-11-5 or 8-12-5.  Livingstone also notes “Mr Hodgman’s decision not to deal with the Greens played especially well in Bass and Braddon, and even saw him outperform Mr Bartlett in his home electorate of Denison.”  But in fact Bartlett was last ahead of Hodgman in Denison on the predictively meaningless preferred-premier score in May 2009, and was nine points behind in August and twelve points behind in November compared to seven now, so there is no evidence that Hodgman’s decision has the slightest thing to do with it.

Patrick Caruana in the SMH makes the really odd comment that “While Labor’s support has remained steady since the election, it is 10 points behind where the party sat in May last year, and some 15 points behind the Liberal Party.”  What this apparent mathematical absurdity really means is that Labor’s headline support rate of 23 is the same now as it was in the EMRS February poll (which was in fact the second-last EMRS poll taken before the election – the March one is not included in the EMRS statistics because it was an Examiner exclusive).  But in fact that headline rate was less than two-thirds of Labor’s actual support at the election, and the headline rate being the same now does not prove Labor support is steady either since the election or since February 2010.  The reason for this is that the number of 2010-election Labor voters not actually captured in the February headline Labor rate is so large that it now slightly exceeds the total “undecided” vote, so perhaps at least some 2010-election Labor voters now support somebody else (or perhaps it is just sample error).  This just serves to underline how awful the EMRS polling in February and March 2010 was as a guide to percentage support, and how badly the commentariat is served by not having more accurate regular polling.

Sue Neales portrays the poll as a “savage blow” to the Bartlett Government, claiming a “10 per cent plunge” in Labor’s vote (on the basis of the figures with “leaning” voters included and undecideds redistributed) but again, those figures ignore the proven fact that EMRS-undecided voters tend not to vote Green, and tend disproportionately to vote Labor rather than Liberal.  The claimed 10 per cent plunge is less silly than the 14% plunge claimed by those working off the headline rate, but is still not credible.  Neales also repeats the common error of confusing Bartlett’s preferred premier score with his “popularity” and “mass appeal” (which have been measured only once so far, by Newspoll before the last election).  To repeat a point I have had to make far too many times before, since “preferred premier” simply reflects which of the three leaders a given voter likes the most, some voters will prefer a given candidate without liking any leader, while others might prefer one leader while approving of all three.  We can see the discrepancy between popularity and preferred leader at federal level, where Rudd continues to enjoy large leads as Preferred Prime Minister over Tony Abbott, but in fact neither leader is currently popular, and both have disapproval ratings higher than their approval ratings.  Rudd leads on 2PP because so many voters who are disappointed with him consider that Abbott is worse.

Neales repeats the error, referring to a “high approval rating for Mr Hodgman” just because he is ahead on preferred-premier over Bartlett and McKim.  I suspect that Hodgman would indeed have a high approval rating, but so for that matter would McKim, if the pre-election Newspoll is anything to go by.

Neales makes the following observation on the interpretation of the poll:
“Liberal insiders [why these needlessly anonymous sources again? - KB] yesterday claimed that the poll, if converted into seats in the House of Assembly at an election, would have resulted in the Liberals seizing 13 of the 25 Lower House seats up for grabs, leaving just six each for Labor and the Greens.
But political commentator Richard Herr disputed this claim, pointing out that under the Hare-Clark voting system it was just as feasible for the 10-10-5 vote of March 20 to be replicated under the results of this new poll.”

Herr is right to be wary of the Liberals’ confidence that this was a majority-result poll.  It would probably result in some gain of seats to the Liberals, but another 10-10-5 is indeed still feasible on these figures if one assumes that all the poll shows is the Liberals solidifying support they received anyway, plus a few points for margin of error.  This requires assuming that those indicating themselves as “undecided” or even leaning Liberal or Green in this poll would virtually all vote Labor.

Despite the approval-vs-preferred-premier howler, the Neales piece is the meatiest and most interesting of the mainstream discussions of the poll.  In particular, Neales points to a six-point dip in Will Hodgman’s preferred-premier score in the February 2010 poll and attributes this to “the widespread view that the Liberal Party ran a poor election campaign.”  In my view, the Liberal campaign was merely lacklustre and nowhere near as dire as Labor’s, but the movement of preferred-premier support from Will Hodgman mainly to the “none of the above”/undecided pile in the leadup to the campaign, then back again after the election, suggests that the party and its leader lost momentum in early 2010 and could have otherwise caused Labor even more damage than they did.  Had this been a result of successes or backflips by Bartlett, it would have been reflected in improved preferred-premier ratings for him, but this was not the case. Negative campaigning by Labor? Perhaps.

Lastly, I would like to make some early comment about federal election prospects in the seats of Denison and Bass.  A very common response to the 2010 Denison state results and to recent federal polling for the Greens has been to suggest that the Greens would have a red-hot chance to win Denison federally.  With a low-profile Labor candidate being endorsed as Duncan Kerr’s replacement there must be some potential for the Labor vote to crash.  But to win Denison federally the Greens would have to at least come second ahead of the Liberals, in order to win on Liberal preferences.  The state result provides no reason to expect such an outcome – firstly because the Green vote usually lags at federal level compared with state, and secondly because even if the state Denison result was replicated at federal level, the Greens would not win anyway.  The idea that they would rests on adding the Green and Wilkie votes together to form a “progressive” vote that exceeds that of the Liberals, but in all likelihood the Wilkie voters came nearly as much from the Liberals as the Greens.  There is a remote possibility of the Greens winning Denison if the Liberals make a truly massive effort to throw them second place, but the chance of it occurring seems no more than a few percent right now.

In Bass, an EMRS poll showing 38% primary support for the Liberals’ Steve Titmus, 35% support for Labor’s Geoff Lyons, and 16% for the Greens who have yet to announce their candidate, was released in May, resulting in a warning from the Greens that the party “already had the votes, in Bass, to sway the federal election result.”  Oooh no, you mean Bass might go to preferences?  Hardly the shock horror scoop of the year.  In fact, given that figures at the 2010 election were 43.5-37.2-15.3 and that undecided voters don’t vote Green, this poll doesn’t tell us anything we did not already know about likely Green support in this electorate, hardly confirms a “quantum change” towards the party as claimed by Christine Milne (unless by “quantum” she means small) and if anything is mildly positive for Labor.  Lyons’ attempt to claim underdog status is none too convincing given that the poll shows him just three points behind on primaries and Jodie Campbell won the seat from 6.3 points behind on primaries last time, while Titmus’s claim that the poll shows disappointment with Labor is implausible for exactly the same reason.  So a seemingly unremarkable and inconclusive marginal-seat poll has succeeded in generating nonsense responses from all three parties at once! 

As noted before, Green threats concerning preference direction are usually pretty toothless given that Green voters in Tasmania overwhelmingly preference Labor at federal level no matter what reasons are advanced concerning why they will supposedly stop doing so, and the party’s preferencing decisions rarely move even a tenth of its voters to change the way they vote.  Green preferences flowed nearly three to one to Labor over the Liberals in Bass in 2007.  If the Greens as a party want to really be players in this seat they need to outline what issues they will actually direct preferences over – and how they will stop Labor from laughing if they threaten to tell Green voters they should preference a Coalition led by Tony Abbott.