Image for Demolition row

UNLESS EMRS polling has malfunctioned completely (and done so twice in a row) the Bartlett Labor Government is set for a very serious trashing at the state election next Saturday.  It will do well to keep the swing below double figures, is struggling to hold any of its four most vulnerable seats, and could even do so badly that others come into play.  The main thing in the way of a Liberal majority government at this stage of polling (bearing in mind that the poll was taken last Tuesday and Wednesday) is the very even Liberal/Green split in the votes that are leaving the Government.  One way of looking at it is to see the pre-2006 Labor government of Bacon and Lennon as a broadly centrist regime capturing both centre-left and centre-right votes.  With the many failures and scandals of the last four years, and deliberate repositioning of both the other parties towards the centre, both wings seem to have fallen off the Labor aeroplane at once.

I was very sceptical of this pattern when it first appeared in the February poll, thinking that the poll could well have overcooked the Green vote as polls tend to do close to elections, and bearing in mind the swing back to Labor that had occurred at the previous two.  However, the March poll is completely consistent with the February poll, leaving systematic polling error as the only possible alternative – and EMRS’s polls taken within the last two weeks of a campaign were pretty close in 2002 and 2006. Labor has completely failed to improve its position in the mid-period of the campaign in the way it did in 2006, most likely because it has been unable to sell itself as having any chance of winning majority government. 

The figures were released in the Sunday Examiner and include electorate by electorate breakdowns of intended candidate choice.  Much is made of these in the reporting but it must be kept in mind that over half the voters did not indicate a preferred candidate.  Furthermore, the margin of error of such results is very high because of small sample size, and finally, EMRS’ method of adding #1 and #2 votes together is rather strange given that not all #2 votes reach the candidate at all, and if they do they are often not at full value.  Sometimes these candidate samples can be useful in detecting that a candidate is already doing well, but they are not very reliable as indicators that a candidate is doing badly.  And for the samples to reliably indicate a candidate is doing well, they generally need to show that candidate well into double figures.  Two and three percent results translate to 4-6 responses and are meaningless.

The headline figure of the current poll shows Labor 21 Liberal 29 Green 22 and a ludicrous 26 “undecided”.  As noted before most pollsters would not class these undecideds as such, and the idea that more than one in four voters really wouldn’t have the foggiest at this late stage is just silly.  With the ummers and aaahhhers allocated to their preferred party, the genuine undecided rate of 8 is high but not that high, and nor should it be sky-high at this stage.  The results comparable to other states’ polls – Labor 30 Liberal 35 Green 26 Other 2 Undecided 8 – show Labor 14 points worse than at the same stage in 2006, the Liberals seven points better and the Greens eight points better.  If trends from the last poll to the election are similar to last time, Labor will poll in the mid 30s, Liberals around 40 and Greens well into the 20s (I am a little nervous about my past assertions that somewhere around 23 was the upper limit for the Greens this time!) But bad as that is, there is the possibility that the late swing back to the government will not appear, and that Labor will do even worse.  Either there is something very wrong with the polling or else, barring a late chicken-out by those considering voting Green in particular, we are on for a spectacular, strange and nationally historic result.

It has been my habit to pool consecutive EMRS polls when they return similar results, in order to try to make some sense of the small samples.  There is no significant difference between the February and March polls on a statewide basis so I am doing this again, and weighting slightly in favour of the March sample.  Indeed, the claim by Tony McCall that the differences between the February and March breakdowns demonstrate “volatility” is completely incorrect.  That there have been cases where a party is up 7 points or down 5 in given electorates between these two polls means nothing given that the margin of error of such small samples is about 7 points and that many of them are being considered at once.  If anything, if the February and March polls are both accurate, then far from demonstrating voter volatility, they demonstrate a remarkable lack of it given the turbulence seen in polling results through this government’s tenure.  If that is true, that is another very bad sign for Labor.

On to the batched results and some comments based on them (bearing in mind that these depend on the dubious assumption that the relatively few still completely undecided voters will split proportionally).  In Bass (30-41-26-2) the Liberals have more than half a quota lead over Labor but the result would be 2-2-1.  Ferguson and Gutwein are in my view assured of election for the Liberals and it is too early to say who might take the second Labor seat (provided there is one). 

In Braddon (35-42-20-2) the split would also be 2-2-1 with a significant possibility of Adam Brooks tipping out Brett Whiteley (I base this on my perception of the campaign; the EMRS samples say a similar thing but are too small to be conclusive).  But it is rather hard to believe that the Green vote in this seat will really double even with two polls showing something of that sort, so while things may be looking especially good for Paul O’Halloran at the moment, the Greens should only believe they have a quota in this seat if they see it on the board rather than in opinion polls.  A particular concern for the Greens would be if their vote softened slightly with the undecided vote going to the Liberals.  2-3-0 remains a threat in this electorate, especially if Brooks and Whiteley poll substantial primaries.

In Denison the merged-sample breakdown is bizarre: 27-30-38-5 (the 5 do appear to be mostly Wilkie voters).  On these figures the Greens would get two seats easily, leaving Labor to attempt to scramble two on the surplus of the Greens and the preferences of Andrew Wilkie. (I think they would probably get there as aside from Matthew Groom I suspect the Liberal vote is going to scatter.) I do not believe the Greens will really poll anything like 38 in Denison but even if they fall several points short, the way things are shaping up between the other parties there is a real chance of there being enough slops for them to pick up a second seat.  The electorate sample clearly demonstrates substantial name recognition for Greens #2 Helen Burnet.  Threats to David Bartlett’s seat are being a little overstated on the current figures but Singh and Sturges should be very concerned about Scott Bacon potentially cutting the number of seats they are competing for from one to zero.  2-1-2 is no longer just an outside chance in Denison now, but a very realistic possibility.

In Franklin, 29-39-30-2 sounds (and given the 2006 result, is) terrible for Labor with the Greens seemingly fighting with them for a second seat, and the possibility being floated of Labor only getting one and Lara Giddings losing.  But even if those sorts of figures do occur (or even if Labor is three points behind as in EMRS’s March sample) reading the result off the raw figures is misguided.  Firstly Liberal preferences may well favour Labor (especially name-recognition leakage from Will Hodgman, which will not help the Greens since Nick McKim will have a quota) and secondly the Greens lack sufficient profile for their support candidates and have probably made an error in picking Adam Burling (known as an activist and hence likely to be seen as too extreme) as the candidate to build as number 2.  This is especially obvious in the EMRS samples which show a mere six of the 38 voters who named Nick McKim as number 1, able to name another Green as their preferred number 2.  Thus, my reading of the combined sample (if anywhere near accurate) is most likely still 2-2-1.  I do not take for granted that Jacquie Petrusma will be the second Liberal in should they get two; with his strong Eastern Shore connections Tony Mulder should not yet be discounted despite his campaigning hot water this week.

Lastly there is Lyons (39-40-21-1).  Those figures would lead to a straightforward 2-2-1.  But which two is an interesting question in the cases of both major parties – it is far from clear who would be the second successful Liberal alongside Rene Hidding.  Also, it is interesting to see David Llewellyn barely raise a blip in the EMRS sample, because while that does not necessarily mean a thing, I have often noticed that when politicians are seen to be unsure about serving out their next term, the voters will sometimes make that decision for them.  Llewellyn has stated he will serve the full term but whether anyone in voterland believes him is another question. 

Adding up my own breakdowns based on the merged samples electorate by electorate, the total rather strangely comes to 10-9-6 – Labor could win more seats despite being significantly behind in at least three electorates.  But Labor’s 10 there includes two that are rather shaky, while the Liberals’ nine are rock solid in those samples and they have realistic chances of at least two more.  The Greens are solid on six in the merged samples with some chance of another, but the Green seat tally has often been overprojected in the past and no-one should get carried away about it in advance this time in case the (supposedly) Green vote yet again softens.

It remains to be seen whether the usual late scare campaigns against the Greens have much effect.  My perception is that Labor has done its best to scare voters off the Liberals, but is so much on the nose itself that many of those thus scared off have declared a pox on both houses and at least considered voting Green.  The dramatic intervention of ex-Premiers Gray, Lennon, Rundle and Field is problematic in the sense that while they can tell voters to vote for majority government, it is not at all clear how the voters should go about doing that even if they want to.  It may knock some soft edges off the polled Green vote, but in terms of the net outcome, that doesn’t seem likely to matter. 

I will try to post a final prediction of where I think the seat tally will end up a few days out with the benefit of any other polls that may or may not emerge in the final week.  It looks likely to be a very odd election night indeed.

(Kevin Bonham apologises for the lack of hyperlinks and other such goodies in this article as he is posting from remote areas with limited internet access on fieldwork – oh yes, there are places with more snails than Adam Brooks signs.)