Thursday, April 28, 2016

Queensland: Compulsory Preferencing And Recent Polling

Because of other distractions (see articles below) I've been a bit slow dealing with Queensland's sudden return to compulsory preferencing, but fortunately in this time a new state poll has arrived that is highly relevant to the issue.

For those who spent a couple of hours asleep last Thursday, that was about all it took for Labor to amend an LNP motion to expand the parliament from 89 to 93 seats and make it conditional on a return to compulsory preferencing.  The LNP voted against the amendment but the KAP and ex-Labor crossbenchers supported it and in the end the LNP didn't even vote against the amended bill.  It was vintage roughshod Queensland politics in its execution - moving it as an amendment meant it could be passed without prolonged discussion or a committee process, and Queensland of course has no upper house to get in the way.  It all gave the appearance of a tactical disaster for the Lawrence Springborg-led LNP (allowing such a major concession while fishing for crumbs) and again raised concerns of friends and foes alike as to whether the Queensland Opposition has all that much upstairs strategically.

The Impact of The Change

Immediate reactions suggested Labor might have voted themselves an easy win at the next election. After all, Antony Green had calculated that had voters whose votes exhausted distributed their preferences the same way as those whose votes didn't, Labor would have won an extra eight seats in 2015 and would have had a comfortable majority.  The 2PP might have been about 53.4% to Labor, 2.2 points higher than they actually obtained.  (A more accurate estimate could be found from the distributions, but probably isn't necessary.)

However, the number of Labor misses caused by optional preferencing probably wasn't really that large.  Especially at the 2015 election, at which the rate of preferencing by non-LNP voters was very high because of a successful "put the LNP last" campaign, it's likely that minor/micro-party voters who exhausted their preferences were unrepresentative.  Scott Steel (@Pollytics) on Twitter suggests, based on internal campaign research, that the winnings would have been more like three or four extra ALP seats.

Steel offered some descriptions of the sort of non-major voters who still didn't preference in 2015, suggesting they were "a different political cohort [..] more angry, more hardcore" and suggested examples like anti-asylum-seeker Greens voters, haters of organised politics, "unusual idealogues" and "a whole bunch of fad voters that think "this is the only way", even if that "only" changes from week to week".  I suspect that low-information PUP supporters could be added to the mix.

I'll add that projections don't always take account of the votes Labor disproportionately loses to the informal pile when voting is made compulsory.  However, in my view the key point is that this change is not about what would have happened last time, it is about Labor's fear of what will happen next time.  Quite aside from the waning of the anti-Newman factor, there is bound to be disenchantment from Greens voters about the environmental direction of the Palaszczuk Government, for instance over the approval of the Adani coal mine.  Voters for other parties such as KAP may have other reasons for concern.  If Greens and Others voters under LNP were to revert to 2012 preferencing behaviour, that could have put the Palaszczuk Government into opposition under optional preferencing even without any swing on primary votes at all.

There is probably nothing that Labor can do about the reversion of non-Green preferences to type sooner or later, after an extraordinary result in which 61% of KAP and 67% of PUP preferences that were distributed went to the ALP (though both parties were carrying votes from other candidates).  However, it seems that the wild swings in Green preferencing behaviour could be connected to a certain type of Greens voter - the type who will preference Labor when they detect a strong difference between the parties, but exhaust their vote otherwise.  There is strong reason to believe that under compulsory preferencing this voter type will nearly always preference Labor, and hence I see Labor's real goal here as a license to take Green preferences for granted.

(I did wonder how the Queensland 2015 election would have panned out had preferences been compulsory but had voters preferenced as at the 2013 federal election.  Labor would have done worse than they did in the real state election, with 50.1% 2PP, which highlights how extraordinary the preferencing behaviour of KAP and PUP supporters was.)

Problems With This Form Of CPV

Readers may be aware that I am not a big fan of compulsory preferencing, nor even of "compulsory voting".  I believe that a voter should be allowed to exhaust their ballot for any of these reasons:

* as a strategy to discourage major parties from taking their preference for granted
* when they reach a point where they have no opinion between the remaining candidates
* when they reach a point where they find all the remaining candidates so objectionable that they do not wish to endorse any of them over each other.

The first probably has something to do with the original support for optional preferencing as an anti-corruption mechanism.  I think the third reason is stupid since there is nothing wrong with choosing to preference the lesser evil, but I think voters should be allowed to make their own decisions on that.

However, big-picture debates about the merits of compulsory preferencing vs optional preferencing aside, what disappoints me is that advocates of compulsory preferencing are often so unwilling to propose doing anything at all about informal votes arising from unintended numbering errors.

The new Queensland legislation is among the worst cases in this regard, and shows why even a vestige of consultation would have been a good idea.  It will allow voters to leave the final candidate blank after voting for all the others, but any other omission or repetition in numbering will see a vote discarded as informal.  This applies even if the vote has a primary vote for a party that will clearly not have its preferences distributed, and even if the error involves candidates who will not be in contention for the seat.  There will be a significant increase in unintentional informal voting, though most likely the party that proposed this amendment (Labor) will be the biggest victim.

Others such as Antony and William Bowe have made the fine point that this sudden move by Queensland Labor to smuggle this change through without consulting voters at all stands so sharply at odds with the federal party's complaint about supposedly insufficient consultation in the Senate voting system reform process.  I could go on about that too, but it would get pretty ugly and ranty.  Let's move on.

Polling and CPV

The switch to CPV means that the published 2PPs from Queensland polling since the last election are now junk, since all were based on the assumption of optional preferencing.  Some polls were assuming voters would preference as at the 2015 election, some that they would preference as at some average of the last two or three elections, and some were using respondent preferencing or perhaps just plain making stuff up, but in any case it's all irrelevant now.  (A possible partial exception could be Morgan - in their very earliest state polls I suspected they might be assuming compulsory preferencing by mistake.)

How we model the 2PP under compulsory preferencing for Queensland is a very difficult question!

The newly released ReachTEL is a useful example.  After redistributing those initially undecided voters who can be prodded (in other words giving a headline figure the way other pollsters do), ReachTEL has primaries of ALP 36.9 (-0.6 compared to election), LNP 42.8 (+1.5), Green 10.8 (+2.4), KAP 3.6 (+1.7), Other/Ind 5.9 (-5 mainly down to the demise of PUP).  This could be explained largely as the PUP vote coming back to KAP and the majors, plus some of the Labor vote going to the Greens.

Based on these primaries, Labor would have a roughly 50.4% 2PP by last-election optional preferences, but the published headline by respondent preferences is 52:48 to LNP.  I'm a bit puzzled by that figure, because multiplying the 60.5% flow to Labor in question 1b by the number of minor party supporters in questions 1 and 1a actually leads only to a 51:49 (50.8 to one decimal) result to LNP, but the core point stays the same.  It's not clear Labor will do as well at converting preferences to 2PP next election even with this change as they did last time without it.  (By 2013 Queensland federal election preferences, this poll comes out to about 50:50.)

Before we get too excited about the finding of a rather weak flow of preferences to Labor, we need to bear in mind that this is based on a sample of only 260 or so minor party voters, so the 60.5% flow has at least a 6% margin of error, and could easily be up to 1.2 points out on the overall 2PP even if we assume the respondent preferences are broadly accurate.  But even at best, it's sobering stuff for Labor: preferencing intentions might already have changed back enough that all this change in voting systems does is cancel that out, if that.  The temptation to now add three or four points to all Labor's 2PPs since the election and declare that Labor have been winning by heaps all along (given compulsory preferencing) should be resisted. It's also interesting that this ReachTEL respondent-allocated 2PP is no better than one in mid-2015 under the old system with an LNP primary vote three points higher.

The ReachTEL was very similar to a Queensland Galaxy in February (which had ALP 37 LNP 42 Green 9 KAP 4 others 7).  Three Essential samples earlier this year had a slightly rosier picture for Labor when pooled with both majors on 40, Greens around 9, KAP 2 and others 10 (the latter sounds a bit high).  Morgan's volatile and unreliable SMS poll had LNP 45.5 Labor 36.5 Green 9 KAP 3 rest 6 in February, but LNP 40 Labor 37.5 Green 12.5 KAP 3.5 rest 6.5 in April.  It's likely that a cross-pollster aggregate would look very similar to the current ReachTEL, and that any plausible model of the 2PP would show things to be pretty close.  On this basis any view that the recent electoral system change has affected primary voting intention is unfounded, and it wouldn't be expected to.

The next Queensland election will be a mess to model with the creation of four new seats forcing widespread redistributions, on top of the change in the preferencing system.  It's likely that the redistributions will dampen the sitting-member-effect advantages Labor will enjoy in many of the marginal seats it is defending, but that these advantages will still exist.  I am not going to even try to work out what sort of 2PP the Opposition might need to take government until the redistribution is done.  Soon we should also have the benefit of one by-election at which we can see preference flows under the new system.

Other results

The ReachTEL asked Queenslanders for their view of the resumption of compulsory preferencing and respondents gave it the raspberry (27.3% approve, 61.6% disapprove).  This question might well be a bit wording-sensitive; for instance "This change requires you [..]" might focus voters' minds on their own burden of numbering extra squares, and not on the benefit to their preferred party (if any) of other voters being forced to.  All the same support from Labor voters is lukewarm and other voters are strongly against the change (parties like KAP stand to benefit from it, but Others aren't specifically included in the breakdown.)

The poll shows Premier Annastacia Palaszczuk with a positive net rating of +6.8 and Lawrence Springborg with a net rating of -10.4.  These are very similar results to last May's ReachTEL despite minor changes in the options available.  7.7% don't know who Springborg is; how many elections does a man have to lose?  Galaxy earlier this year found +19 net satisfaction for Palaszczuk and -3 net satisfaction for Springborg, but with a middle option of "uncommitted" as opposed to ReachTEL's "average".   Seven News has been having fun lumping the "very poor", "poor" and "average" responses together to make both leaders look bad, but comparison with Galaxy suggests "average" might be a compliment when voters are describing their leaders.

Edited in later on same day: The full results of the ReachTEL just released also include a question on approval of the increase in seat numbers.  It's a pretty neutral response given the natural "what? more politicians?" kneejerk reaction with 33.2% support 42.1% oppose, but what's amusing is that Labor voters tend weakly to support the measure and LNP voters to oppose it.  The votes of their parties in parliament were the other way round!  It's possible that some voters for both parties may have wrongly assumed the increase was a government-supported measure; in either case the LNP hardly convinced their support base of a need to add more seats (a move aimed at preventing the loss of northern seats in the redistribution).

That should do for this wrap of where things are at in Queensland polling.  When I have any idea of who's actually winning anymore, I'll be sure to let you all know.


  1. What is the very latest possible date for the next Queensland state election?