Image for New poll cannot be trusted

THE FEBRUARY 2010 EMRS poll (HERE) is among the most awaited in Tasmanian polling history.  Yet the results of this poll serve mainly to confuse the picture, and to further highlight the problems with having the state served by only one somewhat quirky polling source, which only releases results every three months.  If this poll is no more problematic than any other issued by EMRS, then it shows that Labor’s recovery in the last three months has been too little too late to retain majority government.  In that case, a hung parliament with roughly even representation for both majors is the most likely scenario, although a big blowout in favour of the Liberals remains an outside chance.  If, on the other hand, the poll is simply a statistical rogue or methodically aberrant, then it tells us little or nothing, and we are stuck with four-month-old data and subjective impressions as our best guides to what may happen.

The headline result bizarrely shows the incumbent Labor government only one point ahead of the Greens, with figures of 30 Liberal, 23 Labor, 22 Green, 2 Independent and a massive 23 undecided.  As previously pointed out, this headline result is not really comparable to those issued by other opinion pollsters, and the “%Support + % Leaning” figure in table 2 is where the action really is.  That figure shows a 34-27 lead to the Liberals with the Greens supposedly on 24, Independents 3 and 13% undecided.  The gap between the major parties has narrowed by four points, which may sound like a substantial recovery, but is not.  The main cause of the narrowing is supposed drift from both majors to the Greens, and as a proportion of all major party votes, the swing to Labor since the last poll is less than three points. 

Assuming the poll has only the same problems as the polls by the same company in 2006, then Labor’s position relative to the Liberals is about seven points of swing worse.  In February 2006 Labor led the Liberals by seven points on the EMRS headline rate, and now they trail by seven.  In March 2006 Labor outpolled the Liberals by 17.5 points at the election.  If the difference between this poll and the election is the same as in 2006 then Labor may yet outpoll the Liberals by 3-4 points this time, consistent with Peter Tucker’s suggestion (HERE) of a 4-8% swing.  However, the overall reliability (or not) of this poll aside, the 2006 election had a different dynamic because Labor clearly had a chance of majority government (even if many pundits wrongly thought that chance was very slim) while the Liberals obviously had none.  Also Will Hodgman is a more competitive leader than Rene Hidding.  There is therefore some potential for the Liberals to get a bigger swing depending on the campaign and how sick of Labor the electorate really is.

The most surprising feature of this poll, however, is that the Green vote is supposed to have risen five points in the last three months to new EMRS records of 22 (headline) and 24 (including “leaning”).  This claimed rise has really come out of nowhere, since the Greens have been more or less flat around 17 points (headline) in EMRS polls for the past four and a half years, except for a brief dip to 15 and then 13 in early 2009.  There is also nothing novel in the political landscape driving it (the significant part of the Australian Story “toxic trees” claims – which will not necessarily have that much electoral impact anyway though the St Helens booths will be well worth watching on polling night - broke after the polling period finished.) It is true that the Greens have not thus far made any significant campaigning errors, and their marketing pitch “Ready to Deliver” may be sufficiently punchy to those unconcerned about the murky waters of “deliver what and how?”  Another possibility is that the Liberals have genuinely gone backwards as a result of internal tensions and negative campaigning by Labor, but Labor themselves are so on the nose that the Greens are picking up all the slops.  Yet another is that the Greens are getting a relatively free ride so far as the major parties do their best to tear strips off each other. 

All in all though, I simply don’t believe the Green vote will be 24, let alone the 27 with undecided votes redistributed, come polling day.  In 2006 various polls in the leadup had the Green vote between 18 and 23 points (some of these with quite high undecided figures), and all of them overestimated the final Green result of 16.6%.  I’ve suggested before that to convert an EMRS poll into a likely real election outcome it is necessary to take two points off the Green vote and give all the undecided vote to the major parties (in the past, typically mostly to the government).  On that basis, if the poll is representative, the Greens might be on for, say, 20% - which would be a brilliant result for them but still below what this poll is suggesting.  The other possibility is that the Greens are not really doing that well at all, either because the poll is a statistical rogue (outlier) or because the uncertainty of the election is making far more voters than normal say they will vote Green, when in practice they won’t do any such thing.  Such are the difficulties of knowing what to make of a single poll issued after a three-month gap lacking other polls to compare it to.

I do not want to go into too much detail about EMRS’s electorate by electorate breakdowns because it is just not possible for anyone other than a flying pig to take this particular set of 200-vote samples seriously.  Even accounting for the margin of error, we are supposed to believe that in Denison the Greens are strongly poised to outpoll Labor (perhaps even by a quota) having been outpolled by Labor by 20 points in 2006.  On the figures as given, first-term Green MHA Cassy O’Connor is supposed to easily outpoll not just the sitting Premier but his entire team including two other incumbents.  This won’t happen.  The Greens are supposed to more than double their vote in Braddon - again, this isn’t credible.  It is notable that the biggest swing from the Liberals back to Labor compared to the November poll is in Denison, which fits as the Liberals have had candidate issues there, but given the absurdity of electorate breakdowns with undecideds redistributed evenly, it’s hard to place any great trust in even that.  If EMRS’s seat-by-seat breakdowns were repeated at the election the new House would comprise nine Labor, ten Liberal and six Greens, but pondering what would happen if two plus two equalled five could be a more productive use of your time.  Incidentally, even adjusting the seat-by-seat breakdowns in the manner I suggested in the paragraph above, the Green vote in Denison still comes out well into the 30s, and assuming the 6% for Independents is mostly for Andrew Wilkie (not necessarily a safe assumption), that still seems suspiciously high. 

On to a brief roundup of the early online reaction.  Fellow psephologist William Bowe (HERE) “personally wouldn’t put money on” the result being reflected at the election.  Antony Green (HERE)  presents an accurate summary of what the poll means if it itself is accurate without extensively canvassing the possibilities otherwise, apart from noting that the Greens don’t tend to pick up the undecided vote.  As noted above I am not convinced the Greens really start from a base of 22.  Peter Tucker has coverage up (HERE) pointing out how the way polling patterns are going in the leadup to the 2010 election is very different to the previous two.

There is plenty of mainstream coverage up already, much of which reports the poll’s results uncritically and concludes that Labor is gone and a hung parliament is more or less certain.  For instance, the SMH report (HERE) incorrectly reads the seat-by-seat breakdowns as 8-11-6 (actually even on those improbable figures Labor is near or over two quotas in every seat but Denison while the Libs are nowhere near three in any). 

It is reasonable to conclude that unless this poll is wrong in ways that past EMRS polls have not been, or there is a major campaign disaster for one of the other parties, that Labor will not retain majority government.  But it is far too early to write off the possibility of Labor retaining government as the largest party in minority or at least containing the damage to the messiest possible scenario, 10-10-5.  Also, given that Labor led by seven points at the same stage last time and went on to win outright easily, there is still a slim possibility that the Liberals, leading by the same amount now, could do the same thing.  I really doubt it though.  The undecided vote in elections generally tends to either go back to the government or else break evenly, EMRS polls have tended to represent an extreme case of the former, and the momentum the Liberals had late last year appears for now to have peaked.  Furthermore, it will be hard for the Liberals to sell themselves as a party capable of winning majority government after this result.

Hopefully there will be more polls released by various sources between now and polling day as it is really very difficult to know quite what to make of this one!