Tuesday, February 27, 2018

Tasmania 2018: EMRS Has It A Little Closer Than ReachTEL

EMRS: Liberal 46 Labor 34 Greens 12 JLN 4 Others 3
EMRS has historically skewed against Labor and to the Greens by a few points, but will not necessarily do so this time (given Green vote is so low and methods changes)
On this poll Liberals would probably win a majority (c. 13-10-2) but would depend on seat breakdown 
Aggregate of all polling 13-10-2
Current prediction 13-10-2

The second independent poll of the Tasmanian election campaign (now in its final week - see guide) is out and shows a 12-point lead to the Liberal Party over Labor, compared with the 15.8 point lead in the ReachTEL after redistributing "undecided" voters.  This is a 12-point gain to the Liberals compared to the December EMRS (which showed the parties tied) and I believe that this is the largest poll-to-poll gain by a party in EMRS history.  However, there are close similarities with the recovery of the Lennon Labor government from a seemingly majority-losing position in late 2005 to a winning one by the time of the March 2006 poll.

The handy trend tracker makes it look like the jump has come from the Greens, the Lambie Network and others, but more likely voters have moved from minor parties to Labor and from Labor to the Liberals, and the flatlining Labor vote at 34% four polls in a row is not all the same voters.  The trend tracker shows the Green vote of 12% at its lowest level in the last six years; I believe it is actually the lowest since about 2000 but I do not have full archives of the early EMRS results.

Especially if considering EMRS' track record of skewing to the Greens and against Labor this poll is very similar to the aggregate I posted after the ReachTEL.  While clipping the wings of any runaway victory idea for the Liberals, the poll otherwise doesn't say much that we don't already know.  It also says the election is still close enough that a hung parliament remains possible - with the most plausible scenario for that now being a pokies-driven revolution winning Labor three seats in Denison.  It also confirms that the Greens are struggling to match their 2014 vote of 13.8%.

The poll was taken 24-26 Feb with a sample size of 1000.  It may have caught the tail end of the Greens' announcement that they would move a no-confidence motion against the Liberals, but I am not convinced that will greatly affect vote shares anyway.

EMRS has a history of getting the Liberal vote almost spot on and having the Green vote a few points too high and Labor a few points low.  At the 2014 election it had Labor 4.3 points low and the Greens 3.2 points high; at the 2010 election the errors were larger, while at the 2006 election the error in the Green vote was only slight. It is also not clear whether EMRS tended to get the Labor vote a bit low because of an effect specific to Labor or because of an effect specific to incumbent governments. (Prior to this term there had been a Labor government for the whole history of this pollster.)  Also, EMRS has started including mobile phones in its sample for the first time during this term, so past house effect conclusions might not hold.

My aggregate following ReachTEL had figures of Liberal 46.6 Labor 36.8 Greens 10.5 JLN 4.3 Other 1.7, pointing to a result of 13-10-2.  If the division of support around the state (based 60% on that poll and 40% on previous breakdowns in much older polls) is correct then the small differences between this poll and the aggregate do not change that picture.  That applies whether a house effect adjustment is applied for the Green vote or not.  However if no house effect is applied then the Greens are reasonably close to retaining Bass at the expense of the third Labor candidate.  If any house effect is applied this prospect disappears.

If a house effect approaching the size of past elections (say, three points in the case of the Greens) is applied, putting the Greens down to 9%, then both the Greens seats get a little messy.  Denison becomes at some risk of loss to Labor and Franklin becomes at risk of loss to the Liberals.  My suspicion is that with the Green vote in this poll already so low, the house effect might be less significant this time, assuming it exists at all.  So I still have them on for two seats (perhaps three if things go well for them and they get lucky) but on the downside the possibility of the Greens being reduced to one or even zero seats can still not be discounted.


The EMRS poll shows Will Hodgman leading Rebecca White as Preferred Premier 48-41, a massive turnaround from White's 48-35 lead in December.  While the December poll was probably an outlier, I read from this that the honeymoon effect for White has disappeared and that some voters will have developed reservations about her leadership, given that Labor has taken a massive risk (and one I still don't understand) with the nature of its anti-pokies policy.

The Liberal government was in poor shape in late 2017.  It did not appear to be governing the state actively and had just come off a poor result following an inappropriate campaign in the Pembroke by-election.  Voter opinions of Hodgman may also have improved because his government has campaigned with more energy and motivation than it was displaying last year and has looked more like it is willing to act on the problems that have become apparent in this term.


I have aggregated this poll at 40% weight, giving it about the same weight as ReachTEL, because although the ReachTEL sample is much larger, previous elections do not provide much evidence that either poll is more accurate than the other.  I have only adjusted the EMRS Green and Labor votes by two points, which is lower than I'd normally apply.  However I have also applied some subjective changes to the aggregate to account for certain problems.  The first is that the Others vote was running too high in Denison and too low in Braddon.  Many voters would have told pollsters through the term that they wanted to vote independent (especially in the hope that Kristie Johnston might run) but no suitable candidate has emerged.  The second is that the aggregate was projecting the Greens Braddon vote down to as low as 3%, which doesn't seem credible, so I have compacted the Green vote across the state and adjusted other parties .  The following is the current aggregate:

Comments on each seat:

* Bass - while it currently looks like Labor will claim the Greens' seat, if polls are not repeating their past skew to the Greens then the Greens are still in with a chance.  Labor's support candidates are all fairly modest in profile and there could well be a leakage issue for them.  However JLN preferences would help.

* Braddon - the Lambie Network is still lurking as an in-theory threat to Labor's second seat, but given the likely rate of leakage within the Lambie ticket, I don't think it's a serious chance.

* Denison - most likely 2-2-1 but it's still conceivable Labor could get three.  On the current aggregate it seems the Greens are more at risk if this should happen, but the Liberals might also struggle especially as the pokies issue heats up in the final week.  Bear in mind that because of leakage Labor will probably need 3.1 quotas to win three.

* Franklin - if the polls have read the Greens accurately (for once) then Rosalie Woodruff will do better than this and be easily safe.  But on the current aggregate figure, 0.8 of a quota does not make her completely safe from the risk of being stuck behind competing Liberal candidates.  I would still expect her to hold on the aggregate figures, with the help of Labor preferences.

* Lyons - allowing for sampling error and the possibility that the polls might not be skewed to the Greens, it's still possible the Greens could challenge the third Liberal on raw quota figures here.   However aside from the same risk applying as in Franklin, it's also worth noting that preferences from minor candidates in Lyons were very unhelpful to the Greens last time.  Even in a clash with the Liberals I would expect the same again.

Only four days to go! I will be live blogging for The Mercury on election night and posting detailed comments on the counts on here overnight Saturday/Sunday.


  1. Kevin, it would be appreciated if you could explain what are the 'methods changes' in this poll that you mention in the summary.

    1. This will be done in the final version.

    2. I've now been advised that the shift to an online component doesn't apply to this poll and it was phone only (mixed landline/mobile).

  2. Kevin do any of the polls take into account the fact that many more young voters will be voting perhaps for the first time having registered for the SSM 'poll' and are now on the electoral roll - I am wondering if EMRS et al have adjusted sample target audience to try and pick this up ?

    1. Sorry about the slow reply - data on the age breakdowns of voters are available including updates from after that change. Pollsters should be taking it into account in their scaling but I don't know for sure which ones are. I've seen some stats re Batman that suggest that the enrolment differences produced by the survey have since retreated to at least some degree.

  3. The Greens in Braddon are an interesting one to pre-ponder. As a Braddon local I'd observe that Paul O'H was remarkable and had excellent ability to be 'votable' for those not rusted on.

    Scott J is a far more divisive figure, and IMO will struggle to gain swinging local voters. Every time he is in Hobart talking up local issues I suspect he is degrading his vote.

    Locally it feels like a campaign that expects to be an 'honorable failure' not a serious attempt to win the 5th seat.

  4. Between this and Reachtel, we are within the margin's of polling error and candidate preferences of the Libs getting either 1 in Dension or 2 in Lyons and ending up with minority government.

    That said all things being even, you'd expect Libs to get 13 seats, with 12 or 14 not improbable and 11 being a long shot. Can't see Labor getting to 13 seats unless the election results and the polls don't line up at all.

  5. Kevin what does the Liberal vote need to fall to in Denison for a "one seat for the Liberals" scenario become a chance?

    1. It becomes a chance around 32% but only if the Greens are at quota or very close and the Labor vote is over 3 quotas.

  6. Whilst the poll suggests otherwise, it will be interesting in Lyons if Michael Kent gets ahead of the Greens in the count for 5th. I can't see the Greens getting a lot of preferences from Kent/JLN voters, but Greens voters may preference Kent over the Liberals, particular those interested in the salmon farm issue.