Sunday, July 12, 2015

ReachTEL: Liberals Narrowly Holding In Bass And Lyons

ReachTEL: Bass Liberal 51% 2PP (-3 since election) Lyons Liberal 53.2% 2PP (+2) (2PP estimates mine)

Analysis: Poll consistent with other polls showing small net swing to Labor in Tasmania, but not clearly sufficient for seat gains

The Examiner has commissioned a ReachTEL of the marginal Tasmanian federal seats of Bass (LIB, 4%) and Lyons (LIB, 1.2%) (see report here).  At this time the full data are not available online but they are present in the paper edition, albeit in a slightly garbled form.  The format of the polling is also unusual and requires some further discussion.  The poll has been reported as showing that Andrew Nikolic (Bass) and Eric Hutchinson (Lyons) would have been "comfortably returned" in an election held on Thursday.  My analysis agrees with the "returned" but disagrees about the "comfortably".

Normally ReachTEL seat polls either force the voting intention question, such that if the voter is undecided and does not pick a party they are excluded from the poll, or else they don't force it, such that if the voter says undecided then that response is excluded from the results.  In this case they have given the Undecided voter a second bite of the cherry, asking them "Being undecided to which of the following do you have even a slight leaning?"  The options then given are Liberal, Labor, The Greens and Other/Independent (I'm assuming in that order) though the Examiner's presentation has stuffed up slightly by omitting the Liberal option from Bass and including the Labor option twice and no Liberal option in Lyons.  As best I can tell, this is much the same as a poll with the hard undecided excluded, except that the voting intention question has been asked in two stages rather than one.

The normal standard in presentation of polling results is to present the figure with all leaning voters allocated to the party they are leaning to as the headline figure.  Many pollsters do not even publish the figures prior to prompting, though they are useful data.  In this case the use of the pre-prompting figures as the headline in the report, without any recalculation, has given the Liberals a more impressive-looking lead than they actually have.

This is what I think the poll would look like with undecided voters redistributed, and I have also included a 2PP estimate using the last election figures for these specific electorates.  (There is some uncertainty with the Others vote because PUP support has collapsed, but the Other/Independent votes in this poll are probably higher than would actually be recorded at an election anyway):

If repeated on election day these results would deliver a 3% 2PP swing to Labor in Bass and a 2% swing to the Liberals in Lyons.  It is quite unlikely that the seats would actually swing that much in opposite directions so a more likely reading is that the differences between the two mostly reflect margin of error movements in samples of just under 600 voters per electorate, and that overall the poll is showing a small swing to Labor across both electorates, unlikely to be sufficient to win either seat.  This is consistent with the federal Bludger Track readings for Tasmania, which currently have a statewide swing to Labor of 1.1%.  The other caution to be applied to the closeness of the Bass result is that the Labor vote is soft in this sample (with a large slice being "leaning" undecided voters), so Labor might not really be as competitive as 49:51.

The Greens' results in this poll are modest, but not as bad as the 7.9% in Bass and 8.3% in Lyons they polled in 2013.  Polling generally tends to slightly overestimate the Greens vote so the poll should only be taken as showing some improvement and not necessarily the full 3-4 points implied.

Media-commissioned ReachTEL polls predicted the winners of all five Tasmanian electorates at the last election, but did so with a lean to the Liberals of about five points.  This was probably because those polls all allowed a hard Undecided option.  In Tasmanian state ReachTELs the Liberal vote dropped sharply once this was removed, and federal ReachTELs without an undecided option do not generally show much skew either way.  In this particular poll, the difference between the prompted and unprompted figures is about three points 2PP for Bass and very little for Lyons.  On this basis it's still possible that the true state of play is closer than the poll suggests and that the seats would be very close.

An unusual aspect of this poll is a massive gender gap in the Lyons samples with the Liberals polling 52.7% among males and 35.0% among females, with Labor polling 26.2% and 35.0% respectively.  Each gender has a sample size of around 300, but the normal margin of error of about 5.5% is increased by the use of scaling.  Also with so many breakdowns published at once there is an increased chance of some breakdown being statistically "rogue" and eye-catching but not necessarily real, so I wouldn't take the size of this difference seriously.  Likewise it's unlikely that only 3.3% of 18-34 year olds in Lyons really vote Green.

Other recent Tasmanian federal polling (commissioned by lobby groups) was reported in Would Wood Waste Waste The Seat of Franklin? The AFPA poll, in a format comparable to the "firm" scores for this poll, had the Liberals in Bass on 46.7%, Labor on 31.1% and the Greens on 9.8%.  For Lyons, 44.4%, 31.5% and 10.9%.  On a 2PP basis, the AFPA poll would have had the Liberals about a point better than this one and in Lyons there would have been little difference.  The similarity between the AFPA voting intention figures and The Examiner's further supports my initial view that the AFPA poll voting intention figures can be trusted.

As I continue to note, seat polling is an enterprise in crisis.  When seat polls taken weeks out from elections have as little predictive value as recent elections have suggested, the predictive value of any given seat poll taken over a year out and showing a close result is basically zero.  However, this general picture of Labor stuggling to turn a national swing currently running at 6% into any seat gains in the three Tasmanian marginals on 1.2, 2.6 and 4.0%, is potentially a big deal should there be a close election.  And it's not just ReachTEL either - the last five fortnights of Morgan samples have had the swing to Labor at only 2.3%, little if any of which remains after accounting for state and national house effects.  At this extremely early stage, there is no credible evidence that any of the five Tasmanian federal seats are going anywhere.

Tomorrow The Examiner will report on polling of same-sex marriage and issue importance and I will add some comments here.  I expect the SSM results to be underwhelming: Bass and Lyons are quite socially conservative electorates and robopolling generally shows weaker results for same-sex marriage than live phone polling.

Update: Same-Sex Marriage And Issues Polling

The Examiner has now published the SSM and issue findings and the results were much as expected.
The question was "Do you support or oppose same-sex marriage?", a seemingly harmless but actually suboptimal question wording, as some people would not be prepared to say they personally "support" same-sex marriage but would nonetheless not think it should be illegal.  "Do you support or oppose legalising same-sex marriage?" would be more to the point and would probably have received a more positive response.  My favourite analogy for almost all gay rights issues is football. Just because someone won't say that they "support football", doesn't mean they think that football should be banned.

In Bass 47.9% "support" and 40.7% oppose with the remainder undecided.  In Lyons it's 45.3% to 40.4%.  Both electorates show a predictable partisan breakdown with about 60% of Liberal supporters opposed.  A breakdown for Ind/Others voters doesn't seem to be given and I'm assuming that those given as "Undecided" are the soft undecideds on voting intention who were leaning to a party - we know they were leaning more to Labor in the Bass sample and this is reflected in them being more pro-SSM in this one.

The Lyons sample has a whopping gender difference which probably just reflects the gender difference in basic voting intention (as noted above) and is probably mostly not a real reflection of the electorate.  All up given the unfavourable question design, the fact that it was a robopoll (in which voters tend to be more frank about divisive views) and the nature of the electorates in question, even a plurality in support is consistent with the national polling picture (strong support) on this issue.

The issues importance questions are summarised in another piece and none of these tell us anything we didn't already know.  Health, the economy and jobs in some order always come out on top of these things in this area.  Much is made of only a small proportion ranking same-sex marriage as the most important issue, but this is no surprise given that relatively few people are directly affected by it, and it is just about a universal finding in this sort of poll.  Any real relevance of same-sex marriage when it comes to voting intention is not that it swings votes, but that it probably discourages some voters from swinging their vote to the right when they would otherwise do so.

There will be reporting of reactions to the poll results tomorrow.

Category 5 Headline Fail

Thanks to David Mohr in comments for alerting me to a grossly inaccurate headlining of the poll in the Examiner hard copy.  The blunder (at least one hopes it wasn't deliberate) didn't make it to the online version, but yes, it is true, the headline really says "Most Northern Tasmanians oppose same-sex marriage" ... over a poll that shows 40.7% opposition in one electorate and 40.4% in the other, and support exceeding opposition in both.

Seriously, in what universe does 40.7% of the population equal most?  Is there a need for remedial maths education in some part of the Examiner office, or was taking the time to even read the article properly before headlining the problem?  With this kind of thing it's no wonder people make unkind jokes about Tasmanians supposedly having two heads.

The only proper response to this is a second outing already for:

Rainbow Fish Award For Fishy Poll Headlining Connected To Same-Sex Marriage


Tuesday Updates

The Examiner has corrected its blunder from yesterday with a small note on page six that yesterday's headline was "not correct", and an article opening with the fact that the poll showed "more Northern Tasmanians support same-sex marriage than oppose it".

Today there are some new findings on the long-proposed Bell Bay pulp mill, which amusingly is becoming more reliably polled as it becomes still more imaginary.  I've updated last year's article to deal with those.

ReachTEL have also released some new breakdowns for Bass and Lyons on their site.  In both electorates Liberal supporters rank all of health, jobs and the economy highly.  Labor supporters are concerned about health (especially in Lyons) and jobs, but less likely to rate the economy as their top priority.  We knew from surveys in 2013 that voters ranking same-sex marriage as their top priority tended to be young and female; we can now add that they tend not to be Liberal voters.

Some other news in Bass was the announcement of Paul Mallett as an independent candidate.  He's doing the right thing by hitting the ground running at an early stage because it will be a marathon if he wants to get near getting elected in this seat.  Mallett's platform invites at least passing comparison with Andrew Wilkie's as a "progressive-left" independent running on a greenish platform with a focus on political honesty, though I don't think Wilkie's website comes with its own social theory section.

I think Bass is an extremely difficult electorate for any independent to win in at the next election, and that to even get the standard "nuisance value" vote for prominent independents (10-15%) would be quite challenging there.  To use Wilkie's surprise win in Denison 2010 as a comparison, Wilkie was able to get ahead of the Greens and gain their preferences, hence passing the Liberal candidate, and then passing the Labor candidate on Liberal preferences.

He was able to do this partly because Denison had a naturally high (but as it turned out, soft) Green vote, because the Labor incumbent had retired with Labor botching the replacement process, and because one of the major parties (the Liberals) was so weak in the seat as to make it possible to get ahead of them on Greens preferences.  The challenges facing an independent in Bass are much greater because both majors will be running flat-out.  Also, Bass does not have anything like Denison's prevalence of moderate "Turnbull Libs" (who I believe switched more or less entirely to Wilkie in 2010).  I strongly expect we'll be talking about a conventional Liberal-ALP 2PP contest in Bass come election night.


  1. Hi Kevin
    Good analysis as per usual. With regard to the Same Sex Marriage poll and it clearly showing "support" was well in front, it was alarming that the front page of The Examiner declared today that "Most Northern Tasmanians oppose same-sex marriage"

  2. Long time to go before the next election, although it's interesting that the movement is so small. At least Labor is putting up a capable candidate in Bass this time around; should be interesting.

  3. Very encouraging that support for Marriage Equality for those couples in same-sex relationships is actually polling ahead in such conservative seats as Bass and Lyons. Urban area would have much higher levels of support, so it is very positive for the future of this issue, especially when the GLBTI community are under such constant negative attack from the churches and the conservative media around the country. The victory will be won in due course. Still chuckling about Barnaby stating that trade with Asian countries will be affected if Australia gets Marriage Equality. New Zealand has had Marriage Equality for over 2 years and has had 3 billion worth of trade with China in that period. The excuses and denials are getting more absurd by the minute and the public are not being fooled.