Sunday, January 4, 2015

Queensland: Beautiful One Day, Lineball The Next

Estimated Queensland state 2PP based on late 2014 polling and first 2015 poll: 50.2-49.8 to LNP (revised from 50.1 to ALP)
LNP would probably retain office in an election "held now", but would be at substantial risk of loss of majority
(This is not a prediction of the election result yet, as voting intentions may change)

With the Victorian election and its unsurprising result out of the way, it is time to look north to a much murkier contest.  The Queensland state election could in theory be held as late as June, but the government's three-year anniversary comes up in late March, and there are persistent rumours that it will be called slightly earlier, for late February or early March.  I previously covered Queensland state polling in four articles last year:

Is Campbell Newman Actually In Trouble? (late April through June)
Careful With That Queensland Pendulum, Eugene (July)
Recent Queensland and Victorian Polling (August)
Recent Polling In Four States (October)

The story for much of 2014 was pretty similar.  The Newman government held a small to modest aggregated 2PP lead and was facing a swing that would cost it dozens of seats but not necessarily government.  Particular polls were sometimes interpreted as showing that the government would actually lose, but either those polls were outliers or the interpretation was dodgy.  The problem always was that one of the doomed seats, all else being equal, was Premier Campbell Newman's own seat of Ashgrove, and all the polling in Ashgrove suggested that all else was in fact equal and that if the election was even remotely close then the Premier's seat would be toast.  Furthermore, Newman at various stages through the year was mildly to distinctly unpopular both in his own seat and in Queensland as a whole.



Voting Intentions

When I left this story at the end of October to focus on Victoria, the LNP government had just had some encouraging statewide polling, picking up vote share on the back of the same bursting of the Palmer United bubble seen everywhere else in Australia.  Since then, however, we have had the following voting intention results:

* Morgan SMS (Oct 24-27) 50.5-49.5 to Labor off primaries of  38.5 LNP, 38 ALP, 10 Green, 6 PUP, 2 KAP, 5.5 Others.

* Galaxy 18-19 November50-50 off primaries of 37 LNP, 38 ALP, 9 Green, 3 KAP, 7 PUP, 6 Others.

* Morgan SMS (Nov 21-24) 50.5-49.5 to LNP off primaries of 39 LNP, 36.5 ALP, 9.5 Green, 4 PUP, 3.5 KAP, 7.5 Others

* ReachTEL 27 November: 51-49 to Labor off primaries of 39.2 LNP, 37.3 ALP, 7.9 Green, 6.5 PUP, 9.1 Others (including KAP).

* Newspoll quarterly (Oct-Dec) 50-50 off primaries of 37 LNP, 36 ALP, 10 Green, 17 Others (including PUP and KAP)

Of these the least trustworthy are the Morgan SMS polls, which returned volatile and inaccurate results in the Victorian campaign.   In this case though their results are not much different to the others.

Also of note: another Ashgrove ReachTEL (4 December) showing pretty much the usual: Campbell Newman losing 55:45 by 2012 preferences (55.7:44.3 respondent-allocated).

Leaderships and other results

The Newspoll showed a slight popularity recovery for Campbell Newman, with his net satisfaction improving from -19 to -13 (38-51). His "better Premier" lead was up from six points to nine, but that's nothing to write home about given the lead incumbents usually enjoy on this measure.  Opposition Leader Annastacia Palaszczuk has a solid if unspectacular +4 netsat and I don't quite agree with Michael McKenna's claim that voters are "unimpressed" by her performance.

Both Morgan SMS polls actually had Palaszczuk as preferred premier (52.5-47.5) and it's possible that Morgan SMS polls screen out the normal advantage to incumbents because they don't allow a middle option.  (It's also possible that these results are rubbish.)

The ReachTEL showed Campbell Newman's performance rated good or very good by 33.8% and poor or very poor by 46.7%, for a net rating (-12.9) basically the same as in October, but substantially better than in June through September.  Palaszczuk polled a net -6.3, but ReachTEL net ratings are a bit harsher on little-known Opposition Leaders because of the use of "satisfactory" as the middle option.  Palaszczuk's ratings have continued to improve in successive ReachTEL polls and a useful comparison here is that she is now rated "very good" by 11.5%, compared to only 8.1% for Bill Shorten federally.

The ReachTEL poll shows that voters don't believe the government deserves to be re-elected (35.2-51.7) and don't believe Labor deserves to win either (32.8-51.1).  The party status drilldown is interesting however: the bad result for the LNP is driven by overwhelming rejection from Green, PUP and Others voters as well as (predictably) from Labor, but the bad result for Labor is driven by 14.1% of Labor supporters being unconvinced their party is up to it. 

The Ashgrove ReachTEL shows Newman's ratings as local member to be more or less neutral, and his ratings as Premier to be not too bad in his own electorate.

The 2PP Problem

(Warning: this section is about Wonk Factor 2/5). The calculation of 2PP figures for Queensland is difficult.  The main issues are that Queensland has optional preferencing, that the ECQ does not conduct ALP vs LNP preference throws in non-classic seats, and that PUP did not contest the last election.  An added complication is that pollsters tend to overestimate the Green vote (in Victoria all seventeen polls taken during the two months before the election did this.)  Finally, the 2012 election was a record wipe-out for the ALP and it is to be expected that the preferencing pattern in 2015 will be friendlier.

So it does help in this case to look especially at the known known: primary votes.  In the 2012 landslide the LNP polled 49.65% to Labor's 26.66%, a 23-point gap.  The polls above have that gap at 0.5, -1, 2.5, 1.9 and 1 points, for an average LNP lead of a point.  All else being equal, that's an 11-point swing, which would seemingly make the 2PP about 52% to the LNP, enough for a still fairly comfortable win.  But see below.

The pollsters have the government's 2PP not at 52% but at about 49.8% on average.  In part this is because the mix of other parties from which preferences are available has changed.  At the 2012 election, 49% of third-party preferences came from Katter's Australian Party, 32% from the Greens and 19% from independents and micros.  In current polling the Greens are getting an average of 39% of the polled third-party votes, PUP about 24%, KAP about 12% and independents and micros 25%. There's only a very slight increase in the total polled third-party vote compared to the 2012 election, and this is entirely due to Newspoll which has been prone to high non-major readings lately.

It might seem that 9% of the total vote moving from KAP mostly to PUP shouldn't make as much as 2% difference to the 2PP, and in fact it doesn't (it is probably worth about a point).  An important trick in Queensland elections is that because some votes exhaust, a primary vote swing between the major parties produces a slightly greater swing in the 2PP vote.

The seat of Albert is an excellent example of how this works.  In Albert, the LNP polled 13888 primaries (49.6% of primaries) to 7943 for Labor (28.4%). The final 2PP after preferences was 15198 to 9358 (61.9% to LNP).  The percentage gap between the parties increased from 21.2 points on primaries to 23.8 points after preferences, but preferences actually favoured Labor slightly (1415 to 1310 with 3451 exhausting.)

If we just take 11% of the original primary vote (3081 votes) from the LNP and add it to Labor, Labor becomes the winner, 12439 votes to 12117, a 2PP swing of not 11% but 12.5%.

My estimate is that on average the 2PPs provided by the pollsters are pretty close to the mark for the 2012 election assuming preference flows from individual parties stay the same, and that the flow from PUP is similar to in the federal election and with similar exhaust rates to other minor parties.  I get 50.0:50.0 as an averaged estimate from the primaries in the polls above, but when the unreliability of Morgan is taken into account I get 50.1% to Labor as an aggregated estimate.  I haven't applied any recency weightings as none of the polls are entirely recent.


Likely Preferencing Changes

At both the 2013 federal election and the 2014 Victorian election, the preference flows to Labor both from Greens and collective Others (includes PUP and KAP where present) were somewhat stronger than at the previous elections of each type.  This is likely to happen again in Queensland for various reasons.  I would expect Green voters to be especially keen to throw the Newman government out and for that reason to exhaust their vote at lower rates and preference Labor slightly more strongly.

The experience of both the above elections was that pollster-released 2PPs were still accurate, because changes in preferencing behaviour cancelled out stronger than expected primary vote performances by the Coalition and weaker performances by the Greens.  So I am not inclined to mess with the last-election preference assumptions in estimating the 2PP.

What Does a 50-50 2PP Look Like?

In the second of the above four Queensland articles I looked at the output of a seat-based projection model similar to the ones I've used federally and Victoria.  The model makes adjustments for expected personal vote effects and projects a winning probability for each seat.  At the time of that article it was looking likely there would be a bigger crossbench in 2015 than 2012.  This is no longer the case. 

If the 2PP for the government is 49.9%, then my model expects it lose 30 or 31 of its 73 seats to Labor.  That would leave it with about 42-43 to about 39-40 for Labor, before considering the seven current crossbench seats. 

The current crossbench seats are Gladstone and Nicklin (held by long-term independents), Mount Isa and Dalrymple (won by KAP at the last election), Condamine (where the winner of the safe LNP seat defected to KAP) and Gaven and Yeerongpilly (where the LNP winners both became independents, then joined PUP, then quit PUP and became independents again. As you do.).

Of these I greatly doubt either Gladstone or Nicklin will be won by major parties unless their incumbents retire.  The LNP would be most likely to pick up Gaven, Condamine and Dalrymple if any of those fell, while if either major party won Mount Isa or Yeerongpilly it should be Labor.  My expectation is that the major parties will recover some of these seats, but probably not all.  The model does therefore provide a few paths to a scraped outright LNP win with a roughly 50:50 2PP, but it suggests that on that sort of vote share the government could well just lose its majority.  Any gains from the crossbench might also be counterbalanced by new unexpected crossbenchers raiding "safe" LNP seats.  The model also suggests that a Labor win of any kind with a 50:50ish 2PP is rather unlikely, but not impossible.

I am, however, a bit cautious about the model's weighting for personal votes (it currently thinks personal vote effects are worth about three seats for the government).  While this held up reasonably well in Victoria, the landslide nature of the 2012 Queensland result means that the LNP won a lot of seats with candidates who normally would have had little chance, and these might not on average be such great first-term local members.  There will also probably be a lot of seats where Labor runs the former member for the seat again, which will mute personal vote effects.

In Ashgrove, the model expects that if the state 2PP is 49.9%, Newman would get only 44.6%, which is more or less exactly what he is getting in current polling.  I don't rule out Newman retaining his seat (especially as seat polling generally has a shaky record in recent elections), but unless the state 2PP lifts sharply from here it will be very difficult.  

Where to from here?

Just because the polling late in 2014 showed a 50:50-ish average doesn't mean that should be the expected result.  Looking quickly at Newspoll 2PPs for state elections since 2000, I found that in most cases the last quarterly or bimonthly aggregate that was not taken during the election campaign favoured Labor compared to the election result (on average by about 1.7 points, and with quite a lot of variation).  A possible confounding factor is that Labor were the government in most of these cases, but the finding is consistent with the historic federal pattern that the Coalition performs better relative to leadup polling than Labor does. 

However, I'd expect a strengthening flow of preferences to Labor to partly mute the Coalition's normally better resilience against bad polling.  Another reason for concern about the LNP's ability to gain vote share during the campaign is the Ashgrove issue.  There was also doubt about the fate of Ashgrove in the leadup to 2012, but at that election voters were voting to throw out Labor and mostly wouldn't have cared less who the new Premier was. The idea of voting to keep a government without any idea who might actually lead it seems like a tactical trainwreck in the making.

So while there is a general view that the LNP will be returned with a narrow but clear majority - which implies polling improvement from the current position - I don't think the case that that improvement will occur is clearcut.  A very close result is still a serious chance.

I'll be covering Queensland as much as I can in the leadup to the election, noting that I will be offline for several days in late January and early February.  

Update: Early Election Expected

There are now widespread reports that later today (Tuesday) the election will be called for Jan 31.  In that case I strongly expect to be offline for remote fieldwork Jan 27-Feb 6 so don't think I'll be able to provide any live or much post-count coverage.  Most likely I'll be watching the TV coverage without internet assistance (and fuming about the limitations of so doing).  This might change, but even if it does my coverage around election night will be on a modest scale.

Another possibility is Feb 7 (if there is not enough time to get writs issued for Jan 31).  In that case I think there is a good chance I will have full on the night and postcount coverage, but little else in the two weeks prior apart from a hastily banged-together forecast on the night of Friday 6th.

The early date (if confirmed) is being portrayed as an attempt to catch Labor on the hop, but a more obvious possible advantage is avoiding unwanted distractions from federal parliament.

First Campaign Poll: ReachTEL

The first poll of the campaign is already out and it's a ReachTEL with a sample size of 1583.  The LNP has a 2.2 point primary lead (40.3 to 38.1) with 7.6 for the Greens, 6.3 for PUP and 7.7 for others (including KAP), all the last three having shed votes to the major parties.  The released 2PP is 50:50 but I'd say this is a slightly better poll for the LNP than any of those above cited, and that on such primaries by last-election preferences I'd expect them to be around 50.5. 

This poll deserves to have a high weight put on it given that it is fresh while the others are all entirely or on average over a year old - not that putting a high weight on it makes any great difference since it is so close to the existing average.  The one cautionary note about polling for this election generally applies: January is a difficult time to be polling with so many voters on holidays - scaling and/or selective targeting can address this but may introduce more scope for errors.

Including this poll in my modelling raises the aggregated 2PP to 50.2 to LNP which gives them an expected average of 44 retained seats in my model, which would be slightly more likely than not to produce a slim majority with pickups from the crossbench.  That said, I will be looking carefully at whether there are reasons to predict a swing that is heavier in the mid-range seats (LNP seats on margins around 8-15%) than at the ends.

The ReachTEL shows Campbell Newman's ratings improving slightly, including a 4.2% increase in "Very Good".  Palaszczuk's ratings are slightly worse than the previous poll, and ratings for both leaders are more polarised.

In other news, Liz Cunningham is retiring and her seat will be an ALP target; however she has endorsed another independent candidate as her preferred replacement.


When there are a few more polls I'll post another roundup. 

3 comments:

  1. Thanks for the analysis Kevin,

    Liz Cunningham (indy member for Gladstone) is retiring - wonder how this will affect the crossbench

    ReplyDelete
  2. The seat was ALP-held before her and the previous Gladstone-based seat had a long ALP history so Labor would be looking to collect it.

    ReplyDelete
  3. Commentators' assumption that Labor no chance, and their failure to consider hung parliament possibility is noteworthy. I do think that negative evaluations of Labor's last term are a fundamental that may come into play as polling day approaches but currently too much reliance on rule of thumb assumptions about too many seats required? Geoff Robinson

    ReplyDelete