Monday, March 2, 2015

EMRS: It Looks Like A Hung Parliament Poll, But ...

EMRS: Liberal 42 (-9 since election, stable since Nov), Labor 34 (+7/+3), Green 15 (+1/-4), Ind 7 (overstated), PUP 1, Other 1
EMRS historically skews to Greens and Independents and against Labor
Based on this poll Liberals would have a slightly better than even chance of retaining majority government in an election "held now"
Median seat distribution taking house effects into account 13 Liberal 10 Labor 2 Green
Without considering house effects, 12-9-4

A new EMRS poll of Tasmanian voting intention has been released.  The headline figures show the Liberals stable on their iffy result of 42% from November, Labor up three to 34 (their highest reading since August 2010) and the Greens down four on a rather hard-to-credit result of 19 in the November sample.

It may look with the Liberals so short of 50% that this poll could only deliver a hung parliament if an election were "held today".  However, EMRS polls have a known tendency to favour the Greens compared with other parties, and an increasing tendency to favour unstated "independents" (though independents don't poll much worth noting in actual elections).

Considering what we know about EMRS's historical house effects, an even swing by electorate based off this poll's results would leave the Liberals with good chances of retaining their three seats in Lyons, and thereby retaining majority government.



To explain why, here's a couple of interpretations of the poll translated to seats

1. Without house effects

Assuming that EMRS results do not lean to any particular party, and assuming an exactly uniform swing around the state, this is what the current poll would look like:


It's reasonably clear-cut.  In Bass Labor seems to be in a close race with the Greens and Liberals, but Labor would go into the election with only one incumbent, which makes life difficult on preferences, so a 3-1-1 split would be most likely.  In Lyons it isn't completely impossible that the Liberals with 2.56 quotas and three incumbents would beat the Greens with 0.75 quotas and none, especially as minor party preferences in Lyons were very hostile to the Greens in 2014.  But most likely the Greens would just get that one back, especially as it's unlikely the Liberals will have such an even spread of primary votes for their three incumbents as last time.

Thus it's very easily within the margins of possible errors involved that the Liberals would retain 13 seats if we just take the poll straight, but on the balance of probabilities, probably not.

2. With house effects

The history of EMRS polling at elections is that it overestimates the Green vote (as do Newspoll and as did ReachTEL in 2014) and greatly overestimates the vote for Independents.  An "Independent" result, except when it comes from someone like me who actually voted indie at the last election, is generally an easy way for the voter to say they are sick of Labor, Liberal and Green and would like someone else.  But if no someone else comes along who they have heard of and can stand, they will often vote for the major parties anyway.  Pollsters that include "Independent" as a stand-alone item in a read-out always seem to end up with a figure for it that is much too high.

I've found that typically what the Greens get at elections is about what EMRS has them at in the table including undecided voters (Table 2) and that distributing the undecided voters evenly between the major parties with none to the Greens works reasonably well.  Ignoring the issue of whether the 7 points for Independent and PUP is still too high, this is what happens if I do just that: +2 points to Labor, +1 for the Liberals, -2 to Greens, -1 to Others.


Now, likely I should be giving even more to Labor here on the back of that still dodgy looking eight points for Others.  But it doesn't really matter if I do.  On this simulation, the Labor resurgence in Bass means that they win a second seat at the Greens' expense there.  But more importantly, reducing the Green vote in Lyons means that the Greens no longer gain a seat there.  With only 0.63 quotas the Greens would suffer from leakage from their minor candidates, lack of preferences because their lead candidate would lack profile as a sitting member, and a high risk of losing because of the Ginninderra Effect (warning: that link is extremely wonky) if they managed to overcome these issues.

Based on the issue of pro-Green house effects, as experienced not just by EMRS but everyone who has tried to poll Tasmanian voting lately, this 15% is actually a rather bad poll for the Greens.  Indeed, Labor and the Greens combined would be in line to win more seats between them if Labor could give excess votes to the Greens, especially in Lyons.  (In practice, these votes do not flow well as preferences, especially because preferences exhaust).  Even if there is a very strong resurgence towards Labor at the next election, if the Green vote goes backwards again then the Liberals could still end up in majority.

A quick comment on the Independent vote: in times past EMRS have sometimes had scores of 5-6 for Indepedendents, but the Independent score has tended to peak around May when Legislative Council elections may confuse some voters.  In the November poll it was 6% and now 7%.  It is possible these are just Liberal voters parking their vote because they're underwhelmed with the Liberal government but wouldn't vote Labor or Green in a fit, but it's also possible we are starting to see a Jacqui Lambie factor contaminating state polling.  There have still been no Senate polls that try to measure how Lambie is actually tracking in terms of possible re-election at the next half(?) Senate election.

Leaderships

The poll continues to show Will Hodgman with a hefty lead as preferred Premier, but it has shrunk by six points over three months, to 22 points (48:26).  It is normal for a new Opposition Leader to improve on this score after they have had time to become more established in the job.  It is also normal, unless a Government is in trouble, for an incumbent leader to lead by a number well into double digits.  In my view, not much can or should be read into this leadership result.

New Aggregate

My aggregate was last updated in November (Liberals Fall To Four Year Low).  Since then, aside from this new EMRS poll, there were Morgan samples in January (Lib 38.5 Labor 39.5 Green 17 PUP 1 Other 4) and February (Lib 39 Labor 39.5 Green 14 PUP 2.5 Other 5).  These Morgan samples have extremely small sample sizes, and even in the absence of house effects they would be pretty useless.  It looks like they have a major skew against the Liberals, though how much is to Labor and how much to the Greens remains unclear.  For benchmarking purposes, I'm now adding 4 points to the Liberals in these polls, and taking three from Labor and one from the Greens.

As data as old as November is very old rope now, I'm weighting the new EMRS poll at 50% of the aggregate, and the two Morgan polls (as corrected) at 5% (Feb) and 3% (Jan), with the old reading at 42%.  I'm also deducting another two points from the Ind/Other score in EMRS and giving one point to each major party.  This is how the aggregate comes out:



The apparent strengthening of Labor and slight weakening of the Greens in the recent results means that the aggregate projection now has the Greens losing Bass and failing to gain Lyons, meaning that the current median result is 13-10-2.  All the same the Greens would be quite unlucky to just lose both of those (and might hold Bass off Kim Booth's increased profile as leader), and the Liberals' majority would be touch-and-go.  In this respect, not a lot has changed since late last year.

No comments:

Post a Comment

The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.