Tuesday, April 9, 2013

Nelson Legislative Council polling (Updated)

 (Note: for more on the Legislative Council elections see my rampagingly popular if somewhat sprawling candidate preview article.  And there is so much more to come, perhaps including a look at the dreaded Section 159!)

UPDATE: Following the publication of partial results of another ReachTEL on Tasmanian Times this article now includes discussion of two separate polls.  The second is discussed at the bottom of this article.

There have been some mysterious media reports of a Nelson Legislative Council ReachTEL poll.  I now have a copy of the full results (though there is a suggestion that an official corrected version may exist too - see first update below).  This is one of perhaps as many as four polls and surveys of the electorate conducted by or circulated among candidates; one of the others is discussed below.

The publication of partial details of polling commissioned by unknown sources in the mainstream media is a source of continual frustration because it means that the public are not in a position to know what the polling really says, or to access fully informed discussion of what the polling really means.  Nor are they in a position to be aware if there are errors in the polling results, as is apparently the case this time around.

I am here publishing the full details for the purposes of non-profit critical review and study.  Anyone thinking of copying these details for profit-related or any other purposes should be aware that the poll is copyright ReachTEL and obtain appropriate permissions.  I should note that while I have the data, I don't have permission to disclose (or know with absolute certainty) the identity of the commissioning source, and results should be treated with caution for that reason.  That said there is absolutely nothing in this poll, beyond some errors in the presentation of the data discussed below, that I have any reason to doubt is true. 

Firstly, the state-level voting intention of 442 Nelson residents on the night of 21 Feb 2013 was ALP 17.9 Liberal 49.5 Green 22.3 Ind 6.4 Other 0.1 undecided 3.8.  This compares with 2010 election booth figures of ALP 21.7 Liberal 41.8 Green 28.5 Ind (Wilkie) 7.9.    That an already Liberal area has become more Liberal in the prevailing political climate, about equally at the expense of both governing parties, is absolutely no surprise.   As the poll was conducted relatively early in the cycle it is not surprising that it has not asked what we really want to know: who are people actually going to vote for? 

Secondly voters believed that being an MLC should be a full-time job (58.8-24.9).  Of the main parties, support for this view was slightly strongest among Liberal supporters, who were also more likely to be undecided than to have a clear view that it should not be. 

However the differences across parties are not of great meaning because of the small sample size of voters for each party, and I don't suggest too much be read into them. 

Thirdly the table for voter ratings for Wilkinson's performance is available.  However there are severe errors in the version of the table that I have seen.  The Total column is repeated in the Undecided column, the Others column has figures in it despite the number of Others voters being negligible, and the Independent column does not have figures.

There are also some obvious irregularities in the table such as Wilkinson being supposedly rated highly by Labor voters and poorly by Liberal voters.  I believe that what has happened here is that most of the columns have been mistakenly copied and pasted one column to the left of where they should be and that the results that have been reported for Wilkinson's approval in the media are in fact wrong.

The question is "Please rate the performance of Jim Wilkinson as your Legislative Council representative" and these are the circulated incorrect tables:

The Good, Average, Very Poor and Don't Know figures in the Total column were reported by the ABC as figures for all voters.  But it appears that they are actually the figures for Labor voters only.

This is (roughly) what I believe the table should look like:

These figures show Wilkinson with a very respectable net satisfaction rating of +14.5 (including a +38.9 netsat among Liberal voters) and a much lower non-awareness rating than was quoted by the ABC. (Amusingly Wilkinson tried to put a positive spin on the quoted figures on non-awareness, which were wrong.  The actual figures are much more satisfactory for him.)

And this is why the common practice of selective media reporting of unpublished polls is terrible practice.  If there are errors in the poll then there is no opportunity for analysts to scrutinise and correct them.  At least, journalists who do not have the expertise to check an unpublished poll they are reporting on for themselves should run it past someone who does, but ideally, this silly business of people commissioning polls and showing them to journalists (edit: or them being leaked then shown to journalists) without the poll ever being published should end.  Let's not forget this stupid internal-poll-selective-results media/players game has recently been part of the process of removing an elected Prime Minister during his first term of office - it is not as if this is a trivial issue.

Question four refers to perceptions of whether party-endorsed candidates should run in the Legislative Council:

 It's a shame the commissioning source didn't include an option for preferring party candidates only, though I don't think many would have chosen it.  My own perception in the past has been that Liberals were more likely to prefer independent candidates but, perhaps as a result of the success of Vanessa Goodwin, that seems to be no longer the case.  The somewhat paradoxical finding that Greens supporters are as anti-party-candidacy as any others despite their party frequently running candidates is something I'd be inclined to not take seriously because of the small sample size.

Question 5 asks about the importance of different issues:

 There is nothing really surprising here - perhaps the low Education %age among Labor voters, but that is not too odd given that Nelson is a relatively well-heeled area where state education quality is probably a lesser concern than in some other parts of the state.  The concerns of each party's supporters are as would be expected, the sample sizes for Independent and Undecided are too small to be useful, and overall these figures just suggest there is nothing odd about the voter bases identifying as belonging to different parties.

Question 6 asks if voters "support the implementation" of something called the "Tasmanian Forrest Agreement", but sorry Ruth, I don't think this one's about you.

It's interesting that the tiny sample of Undecided voters are basically Greens on this question when on question 5 they were more like Green-tinged Liberals.  Overall the break is basically on party lines but with substantial portions of both Labor and Liberal supporters "crossing the floor".  Because there are more Liberal supporters than Labor supporters, and because support from non-major-party supporters is strong, the overall result is supportive.  It appears, however, that there is another error in the tables here because the Undecided breakdown by party and the Never Heard breakdown by party are the wrong way around compared to the totals.  My suspicion is that the Total figures are correct and the breakdowns are the wrong way around, since 2.0% would be a ludicrously low Undecided percentage but is believable as a percentage of voters who have never heard of this major controversy.

I note that these figures are dated since there has been a lot of TFA water under the bridge in the last six weeks, including the completion of the select committee's report.  It shouldn't be assumed that views now are still the same.

Finally question 7 refers to same-sex marriage.  As I expected in my initial comments, this is simply a question relating to general approval of the concept: "Do you agree or disagree that same sex couples should be allowed to marry":

Here it all adds up, but it does seem a little bit strange that the Undecided voters (all seventeen of them or so) are strident gaymarriagephobes (even more so than the Liberals) yet support the TFA as much as Greens do.   It might be real (very small samples will do odd things) or might not.  Anyway it makes little difference to the overall result.

Given the Liberal (but not necessarily ultra-conservative) lean of Nelson and the usual 60-30ish results in recent national SSM polling this is about what I would expect.  The key point is that the surveyed question is not about support for state-based same-sex marriage and therefore it should not be assumed that it demonstrates Jim Wilkinson's stand on the matter to be an unpopular one in his electorate. 

In summary, the poll shows that Nelson is now a heavily Liberal-leaning electorate and that Jim Wilkinson is passably popular among Liberal voters and has issue positions that are endorsed by most of them, although they would ideally prefer him to be a full-time member.  Not all Liberal voters will vote for him but this does suggest that Wilkinson's opponents will need to harvest a very high proportion indeed of the non-Liberal vote, with a tight preference flow, to have any chance of defeating him.

UPDATE (10 April): Was told tonight (not by the pollster) that a revised version of the results I've reported here also exists with errors corrected (most of them being the errors I spotted here, that were the foundation for incorrect reporting of Wilkinson's ratings by the ABC).  However I have not yet seen this claimed revised version or been able to verify this claim.

The Second ReachTEL (4th April)

The second ReachTEL of which some results are available - two questions out of six now published on Tasmanian Times -  was commissioned by "Nelson Voters for Change" (a small group, but yes there are more than one of them, headed by former Greens candidate Dr Juliet Lavers).

To the initial question "Do you know who your current Legislative Council member for the seat of Nelson is?", 44% said Jim Wilkinson, 11.7% said Andrew Wilkie (Independent MHR for the federal seat of Denison which overlaps much of Nelson), 5.7% said Vanessa Goodwin (Liberal MLC for Pembroke over the river), 3.9% said David O'Byrne (Labor MHA for the state seat of Franklin which overlaps part of Nelson) and 34.8% were unsure.  (Presumably, considering ReachTEL's methods, only these alternatives were offered.)

Some might think there is a massive discrepancy between this ReachTEL, showing only 44% know that Jim Wilkinson is their MLC, and the first ReachTEL showing only 12.3% say they have not heard of him.  But what this really probably means is that many voters have heard of Wilkinson and are familiar enough with him to express a view of his performance but are confused or unsure about which politician occupies exactly which role.  This sort of confusion would be very widespread and  similar questions involving a wide range of politicians would lead to similar errors.  To what extent the incumbent is to blame for this, for failing to spell out his exact role to voters directly and loudly and repetitively enough, or by not having a prominent (or any) shopfront, is debatable.  However, if voters have a generally good impression of him and a sense that they know who he is, then not knowing what his exact role is won't matter when they are forcibly dragged to the ballot box.

(I'd be interested to see polling in which each Senator for Tasmania was randomly inserted into a list of non-Senator politicians, and voters asked to pick which one was the Tasmanian Senator.  My suspicion is that some Senators, even some who've been there for many years, might be outpolled by the dummy options.)

So I think the massive difference between the two polls is just because they test completely different things.  One tests the level of confident awareness of Jim Wilkinson's role while the other tests voter impressions of him, irrespective of whether the voter is confidently aware of his specific role or not.

The second published question is "2. With an annual salary and allowances of over $160,000, do you think the role of Legislative Council member should be full-time?"  The response is 81% Yes, 8.2% No, 10.8% Unsure.  This question - which I would assume to be set by the commissioning source - is an example of a deficient question design practice I referred to in my recent article on Mt Wellington Cable Car surveying.  It is what I call skew-polling. The form of the question presents some facts, but presents facts that would be more likely to be used by one side of the debate.  Therefore, the response becomes biased.

I am inclined to reject the finding of question 2 as unreliable on the grounds of faulty question design, and accept the one in the previous ReachTEL (showing a smaller majority for change) as more likely to be accurate, although it is possible that opinions have shifted somewhat in the meantime as there has been more debate.

One might imagine if the question asked had been "2. Given that Jim Wilkinson has never missed a vote in the Legislative Council despite doing some law work on the side, do you think he should be prevented from doing other work?"  Such a question would also, apparently (at least no-one has contradicted Wilkinson's claim to have never missed a vote) have been factual, but the presentation of information favoring one side of the debate over the other would have skewed the results to high heaven.

(And as was neatly pointed out to me at the debate tonight - the author of this comment is very welcome to claim credit for it if he wishes - Wilkinson's pet defence here is missing a fair slice of the point.  The analogy that was suggested to me is this: if a schoolboy says he's never missed a class, that doesn't prove he's actually done all his homework!)

What are the other four questions and why have NVFC, who claim to support "transparency", not published them too?

Here A ReachTEL There a ReachTEL Everywhere A ReachTEL  (April 30)

The Mercury today reports a GetUp! commissioned ReachTEL that claims to show that 75% of 521 respondents believe being a Legislative Councillor should be a full-time role.

It is not stated whether this is 75% of all respondents or only 75% of non-uncertain respondents.  If the latter then the finding is only slightly larger than the 70% in the initial ReachTEL discussed above (commissioning source still unknown) and hence no great surprise.

Beyond saying that, as usual the basic information needed to assess the reliability this commissioned poll is lacking - the wording of the question is not released, the full results are not released, the full list of questions is not released (prior questions can sometimes skew answers).   Until there is more detail this claimed finding is best ignored, though GetUp! may not care having already got their media hook.

No comments:

Post a Comment