Tuesday, August 6, 2019

EMRS: Labor Down, But Will The Others Voters Please Stand Up?

EMRS July raw figures: Liberal 38 Labor 30 Greens 16 Others 16
Also retro-released EMRS May: Liberal 38 Labor 34 Greens 13 Others 15
Also retro-released EMRS March: Liberal 38 Labor 34 Greens 14 Others 14

Possible "interpretation" figure for July poll: Liberal 41 Labor 32 Green 13 Others 14 (maybe)

Liberals could retain majority in an election "held now" (13-9-3 or 13-10-2), but this would probably depend on what happened with Sue Hickey.

Tasmanian pollster EMRS has released a poll of Tasmanian state voting intention, and has also released the two previous polls in the series (which were not released at the time they were taken; the last released poll was in December).  The polls show a general pattern of a slim lead to the Liberal Government, support for which in the series crashed not long after the March 2018 election, but this particular poll has that gap widening to eight points, with Labor dropping four to 30%.  Labor also polled 30% just after its election loss, and prior to that we have to go back to March 2017 to find it polling worse.  The main beneficiaries are the Greens, who EMRS has long tended to have too high compared to their actual support at elections, but there is also a trend of "Others" continuing to rise, although less than 7% voted for "Others" at the last election.  Who are all these people saying they would vote for someone else, and what are they thinking?

The Labor slump would raise some concerns - as at federal level the party is currently struggling to work out what it stands for, and much of its oxygen on issues is being taken by Sue Hickey.  However, at this stage it is just one reading and we need to see the next one to see if it's a blip or a lasting loss of enthusiasm.

EMRS And The Polling Failure

There is bound to be scepticism about any polling in the wake of the severe national polling failure at the 2019 federal election.  However EMRS did not do any federal polling this year and are relatively exempt from the direct fallout.  They did do a federal poll late last year, with primaries of Labor 40 Liberal 33 Greens 11 Others 15, compared to actual results of Labor 33.6 Coalition 34.6 (the Nationals polled significantly because of the Jessica Whelan situation) Green 10.1 Others 21.7.  It appears that poll's major failing was to not specifically include Andrew Wilkie in its Clark readout, and had it done so it may have been much more accurate (especially given the amount of time before the election).  EMRS's last test close to an election was its Hobart City Council mayoral poll, which was remarkably accurate for the difficulty of polling such an election.

Nonetheless, some of my leading suspects for the national poll failure, such as oversampling of politically engaged voters, are also suspects for EMRS's habit of having Greens and Others votes too high.  EMRS also has rather too many yellow flags from a transparency perspective.  Its minimalist current report format does not disclose whether it is still conducting its state polling purely by phone polling (and with what share of mobiles if any) and does not disclose whether phone numbers are chosen randomly, from a panel-style list, or by some mixture of the two.  The report discloses that responses are weighted but does not say what the weightings are based on.  Also, this is not the first time EMRS has retrospectively released results rather than releasing them soon after they were taken.  It is good science to always release results as soon as possible after taking the poll so as to avoid any suspicion that either the decision to release the results of the previous poll or any data handling in it may have been influenced by subsequent results or events.

What Does It Mean?

Converting EMRS polls to an estimate of voting intention is difficult mainly because of the Others problem.  Although less than 7% of voters voted for Other parties or independents at the election, supposedly 16% would do so in an election held now.  This has been a long-standing issue with EMRS, though at actual elections (when the Others vote tends to fall away) their estimates of the Others vote have ended up pretty good.  So who are these Others?

Here are some possibilities that could explain where 9% of extra Others vote has come from:

* The voter is confused with federal politics and is thinking they would vote for Andrew Wilkie or Jacqui Lambie. (EMRS thinks this is likely.)
* The voter believes Sue Hickey will run as an independent and wants to vote for her.
* The voter is unhappy with their party and "parking" their vote in the Others pile but will return to their usual party at the election
* The voter is genuinely seeking to vote for someone else, but may return to the big three if no suitable and high-profile independent runs.

In trying to project seats off these polls (a dubious exercise predictively anyway, because of the late bandwagon effect seen especially in 2006 and 2018) I have deducted points from the Greens and, when the Others vote is high, from Others as well.  I've given those votes to the major parties, especially the incumbent government, because EMRS has tended to have incumbent governments of either side too low in final polls.  However, my adjustement of the Others vote (2 points) may well be too low - there is just not enough information about why these voters are saying they would vote for someone else.  It would be very useful to ask respondents for Others who they are thinking of voting for and to note how many give Wilkie, Lambie or Hickey as responses, so as to have clear evidence of the extent to which the Others vote is likely to be bogus.

Anyway this is an attempt at a projection off these numbers, with adjustments as discussed above:

The Others vote in each case is assumed to be dispersed between minor parties and independents.  Even in Braddon, if one of them got, say, half of it, they would still lose to Labor because of the even split observed between Shane Broad and Anita Dow at the last election.

In Lyons the Liberals start marginally behind the Greens but would swamp them on preferences from Others.  In 2018 Others polled 10% (0.6 quotas) in Lyons and the Liberals gained 0.2 quotas from those preferences. There was also a strong flow from Others to major parties in this seat in 2014.

The Greens' projected regaining of a Bass seat should be treated with some caution because of adverse preferences and lack of an incumbent.

The big issue with the Government's apparent retention of majority here is the seat of Clark, where their edge over Labor is relatively modest and there is great uncertainty regarding the future of renegade Speaker Sue Hickey.  The projection assumes Hickey either runs again as a Liberal or retires quietly, both of which could be considered unlikely.  If Hickey runs as an independent in Clark, the Liberals do not win two seats on anything like these numbers; either she holds her seat or Labor wins three.

At this stage I still have not restarted an aggregation process for state polling as there is still only one pollster that has been in the field.

White Still Preferred Premier

Even despite Labor dropping to 30% in this poll, Rebecca White continues to lead Will Hodgman as Preferred Premier, this time by 45-40.  White has now led Hodgman in seven of the ten polls since she became leader, the exceptions being two surrounding the election and her first poll as leader.

My suspicion is that a very high proportion of Greens voters prefer White to Hodgman, so this is being driven by the Labor plus Green primary vote exceeding the Liberal primary except for the two polls around the election.  However, that's not all, because the poll is not showing the skew in favour of incumbent Premiers that is normally seen in preferred leader polling.  Even when White was behind around the election, she was not behind by as much as she should have been given the voting intentions and the normal skew to incumbent leaders.

Having observed this for a few years now I think it might be a property of Rebecca White specifically - that voters consider her very likeable and as a result mark her up as a potential leader, but without switching their vote to her party on that account.  However, approval rate polling would need to be released to test if this is still the case (White had extremely high net personal ratings in the state election leadup).   EMRS are known to have conducted approval rate polling for the leaders and also Sue Hickey in the last two polls, but no information regarding the results has been released.

No comments:

Post a Comment