Thursday, March 13, 2014

Newspoll To Deliver Final Nail

Newspoll Tasmania state (first 1000 respondents) Liberals 53%
Outcome if correct: Liberal majority government (14+ seats)

In the face of over three years of high-volume polling pointing in every single case to Liberal majority government at the upcoming Tasmanian state election, Tasmanian Labor has been able to hold out one small hope.  That hope is that the polls are wrong.

That hope has been not without sustenance.  All the polls in Tasmania in that time have been conducted by EMRS and ReachTEL.  EMRS has a clear track record of underestimating the Labor vote and overestimating the Greens, and greatly misread the 2006 election until very close to the poll.  ReachTEL is a new player, untested at this level, and which overstated the Liberal vote in Tasmanian seats at the federal election. That said, there is strong evidence (see ReachTEL house effect section) that this was due to a method issue specific to the options list in polls commissioned by newspapers, and since fixed.

For some time it has been possible for Labor to ask, what if those 48-24 leads shown in polling were really, say, 44-29?  Might not the Liberals then get only 11-12 seats, and the battered regime stagger on in a still further depleted and inevitably chaotic form for at least some part of another four years?

It looks like that little candle will be snuffed out tomorrow.  The Australian has reported that the half-time score in Newspoll, based on the first 1000 respondents, is 53% to the Liberals.  No other interim figures are available, just some very rubbery seat projections.

Of course, that sounds a little high compared to other polling results seen lately, and given that a sample of 1000 has a margin of error of 3%, it wouldn't be too surprising if it came down to, say, 51.  But it might not.  And even if it does, it's overwhelmingly likely that the full Newspoll sample will be a result at least as good for the Liberals as the run of recent combined EMRS and ReachTEL surveys.

It is well worth examining the form guide of Newspoll in Tasmanian state elections at this point, and there is a lot of it to examine.  This table presents the results of pre-election Newspolls compared to the actual results:

(I count the Green Independents as Others in 1989 for comparability purposes, as the Newspoll that year did not explicitly name them as an option. Average figures for Greens include only the years they were listed as an option.)

As polling at state level goes, this is an excellent record.  Newspoll's results display negligible skew with an average error for major party votes of around 1.5 points.  It's not completely perfect - note that at the last three elections the Liberal vote has been underestimated by 3.4, 2.8 and 2.5 points, and that the Green vote was overestimated in the previous five elections including a 3.9 point overestimate in 2010.  (The latter was the result of a methods change that came in after the 2007 federal election.)  But it's not looking even remotely likely that Newspoll are about to get the Liberal vote too high by, say, five points.

And assuming the final campaign Newspoll for Tasmania is indeed 50-ish for the Liberals or higher, it is hard to see what argument can still be offered that the result will be anything but a Liberal majority win.  The argument that there's a real chance of a hung parliament is running out of legs to stand on.  It can't be discounted entirely (in the absence of absolute advance proof) but there is no credible line of polling-based argument to support it.  Doubt about the outcome is drifting from the level "credible" (however small the chance of something else) towards the level "token".  No surprise the bookies are only offering $1.02 on a Liberal win.

The Australian reports electorate breakdowns as indicating at least 13 seats for the Liberals, 6 for  Labor, 4 for the Greens with PUP struggling to get a seat anywhere.  In the absence of precise figures, this should be largely ignored for now.  The seat breakdowns of 200 electors are too small for meaningful interpretation, even assuming that they have been interpreted correctly.  However while my forecast currently gives the fifth seat in Braddon to PUP, it does not do so by much, and a disappointing result in the Newspoll could remove it.   On account of Newspoll's track record, I intend to weight Newspoll heavily in tomorrow's aggregate, provided that PUP are explicitly named as an option in the poll.  A concern I have with the PUP vote in Braddon after visiting the electorate is that it may be dispersed between their different candidates rather than concentrated enough with Kevin Morgan, leaving the party at serious risk of leakage.

A fresh article covering the final Newspoll result should be posted early tomorrow.


  1. Good analysis, but I believe there's a pretty serious error in your second last paragraph: "13 seats for Labor, 6 for the Liberals..."

    Shouldn't it be the other way around?

    1. Typo and fixed, ta. Hopefully none of the 29 people who saw it before I fixed it were fooled.

  2. Come on Kevin you know as well as I do that Newspoll is horribly biased in a pro Liberal manner. What's Morgan Gallup or Nielson say?

    1. Er, is this "Newspoll" you refer to the same horribly pro-Liberal-biased poll that had Labor up 54-46 federally a fortnight ago, or was that a joke?

      Morgan said recently that Tasmanian *federal* voting intention was 52.5-47.5 2PP to the Coalition with a PUP vote of 11.6%, but that is off a very tiny sample, and if that's the federal picture the state one tends to be much stronger for the Libs. Morgan haven't polled at state level here for many years (alas, their state polling was rather good when they did, much lower undecided rate than EMRS for one thing) and I don't remember Nielsen ever doing so.

  3. Interesting analysis, however, knowing that much of Tasmania's voting is personality based and given that there is a large "undecided" component in any poll that includes that category (I can't see "undecided" in any of the polls you quote above), I wonder if these are factors that need to be taken into account. Eg, hard to believe that Petrusma will get elected over Giddings and O'Byrne, or that a third liberal will get elected over O'Byrne and Wightman - both of these need to occur for Libs to get 14, and if neither come to pass they end up with 12

    1. The large "undecided" component thing is a media myth created by nothing but the irregular methods of one poll and bad media reporting of its results. EMRS used to release a headline table that showed results at the stage of polling where they just asked people who they were voting for, and didn't prod voters who weren't sure.

      Undecided rates of 20-30% were common in those tables, but pollsters in other states do not count these voters as undecided if they can be prodded into naming a party. Basically, if someone says "I'm not really sure yet but I'll probably vote Labor" their chance of actually voting Labor is very high.

      When those EMRS undecided rates are looked at on a scale comparable to other polls, they often drop down to about 10-14%, despite which some media outlets create confusion by wrongly reporting the higher figure. But even 10-14% is much higher than what other polls that poll state voting intention are getting. My suspicion is that a high undecided rate results from pressure placed on EMRS staff to complete interviews quickly.

      When ReachTEL were including the undecided option, they were getting undecided figures of only 5.6% (August) and 6.6% (November) - not high at all and much the same as for federal Newspolls a similar time out from an election. And when Morgan used to poll Tasmania they never got high undecided rates, even when EMRS were getting them. There is nothing that is actually unusual about undecided voter rates in Tasmanian state elections.

      In Franklin I don't think Petrusma will get a primary vote like that of Giddings or O'Byrne but I do think Will Hodgman will get an enormous surplus again and that his excess has to go somewhere. Likewise with Bass the minor Libs might not have the primary-getting profile of O'Byrne and Wightman, but Ferguson and Gutwein between them will have a lot of slack to pass on to ticket-mates, and I also think Libs 3,4,5 should get more primaries combined than ALP 3,4,5 who are comparatively low-profile. There will be losses due to leakage in all this but if the primary vote differences between the tickets are large enough it will not matter.

    2. Here's the "undecided" rate for the Newspoll: 3% uncommitted and 1% refused.