THE Mercury yesterday issued a TasPoll and headed its report “Libs on the move”, forecasting an 11-9-5 result for the state election.

The poll report also gives a seat-by-seat breakdown of the poll findings. In this latter respect The Mercury’s poll reporting is superior to The Examiner’s, since The Examiner (as far as I can determine) failed to release a full breakdown of its latest EMRS results. In all other respects, The Mercury’s report is a shocker. There is no mention of the number of electors surveyed, no caution whatsoever about the poll being subject to sample size issues, simply an assumption that the poll results are gospel truth about the intentions of Tasmanian voters. This is even though some of the results are clearly not credible. For instance the TasPoll’s results for Braddon show the Greens on 18% with 20% either undecided or stating that they will vote for “other” (even though there may be no “other” to vote for and most of that 8% wouldn’t actually vote for them anyway). That suggests a Green result in Braddon of 18/80 or 22.5%, a figure so high that no experienced observer would take it seriously.

The Mercury also claims that on the basis of its polling “Labour [sic] will clearly win 10 seats, the Liberals seven and the Greens five at the election” and that only three seats are in doubt. This claim is absolute rubbish — as previously discussed, with a margin of error of close to ten percent, very few seats can be forecast based on 200 vote samples. Indeed only seven Labor seats, five Liberal seats and no Green seats can be individually marked down as secure using this poll’s seat-by-seat results alone. The statewide figure would lead to more seats being added with confidence to each party’s total, but there is no way the Greens can be sensibly said to have five seats locked in on the basis of this poll.

The interpretation of the figures directly attributed to Richard Herr is absolutely correct — if the results given by the TasPoll were repeated in a state election the result would indeed be 11-9-5 with 3-1-1 in Denison and 2-2-1 everywhere else. Only Lyons would be reasonably close. (The Mercury says Herr calls the last Lyons seat the “fifth-most-uncertain”. I suspect he said, or meant, fifth and most uncertain.) This follows an EMRS poll with seat-by-seat breakdowns that add up to 11-8-5 with 1 unclear. Also, the overall support levels shown by each poll are very similar. EMRS shows 34-29-19, or 41.5-35.4-23.1 in three party terms. TasPoll shows 37.5-29.6-19.2 or 43.5-34.3-22.2 in three-party terms. It is interesting that The Mercury makes so much of the supposed swing to the Liberals when the EMRS results put the Libs in a slightly better position. On the whole this was a rather poor piece of opinion poll reporting by The Mercury and I hope to see better acknowledgment of sample size issues from them in the future. The Examiner has, on the whole, been pretty good in this regard.

Merged sample

I have deduced that the TasPoll sample is about 200 votes per electorate based on (i) this being the same sample as the previous TasPoll (ii) the percentage figures per electorate always ending in either .0 or .5. Since the EMRS and TasPoll sample sizes per electorate are so pitifully small, I thought it would be a useful exercise to merge them to obtain a single sample with a sample size of 400 per electorate (although this figure comes down to around 330 for each electorate bar Denison once “other” and “don’t know” votes are removed). I give here the results of this merged sample and a distribution that would most probably produce for each seat. The margin of error is around 5.5% per electorate:

Braddon ALP 43.0, Lib 40.5, Green 16.5 Result 2-2-1.
Bass ALP 40.0, Lib 40.5, Green 19.5 Result 2-2-1
Lyons ALP 48.5, Lib 31.8, Green 19.7. Result 2-1-1 with 1 unclear (very close – ALP marginally favoured over Libs)
Denison ALP 41.6, Lib 27.6. Green 30.8 Result 2-1-1 with 1 unclear (Greens favoured over either major party)
Franklin ALP 40.5 Lib 34.1 Green 25.3 Result 2-2-1.

That gives a total of 10 Labor, 8 Liberal, 5 Green and two undecided, one of which cannot be won by the Greens. (A few corners have been cut in the handing of “Other” and “undecided” votes by simply redistributing them, but this will only affect the above results by fractions of a percent). However within the margin of error of the combined sample only 10 Labor, seven Liberal and three Green seats are assured (in the case of Bass, the Greens could drop below a quota within the margin of error, but neither major party could get close enough to three to have a real shot of beating them.)

On the basis of this combined sample a Labor majority is still possible, because all the existing seats are winnable within the margin of error depending on the results from other parties. However, Labor is up against it in most of these seats.

My near-writing-off of Labor’s third seat in Denison a few days ago was probably premature, though the small sample size means TasPoll’s sample to the contrary alone should not be taken that seriously. My thinking now is this: if Labor can poll reasonably in Denison (43% and up on a three-party basis) then it can win three provided that its second and third candidates are fairly close together and neither other party is too close to two quotas. Since Peg Putt and Michael Hodgman should be elected easily (with significant leakage in Putt’s case, most of it to Labor), at some point there will be only one Green and one Liberal left in the hunt. If Labor has two candidates left and one elected (or better still, three candidates all close to a quota, although this is not likely) it may be able to eliminate one of the other parties and then get its third seat on that party’s preferences, since both Green and Liberal preferences favour Labor slightly.

Kevin Bonham was too disorganised to get money on a Labor majority at the ludicrously long odds initially offered by Centrebet.
Authorised by Kevin Bonham, 410 Macquarie St, South Hobart, Tas 7004.