Saturday, February 28, 2015

New South Wales Roundup: Is Even This A Bridge Too Far?

Time to kick off some NSW state election analysis, much of it in broad and general terms.  I haven't done the usual seat-modelling yet; that will follow in a week or two.

Feds Will Destroy Everything

A common theme on this website is the massive influence of which party holds power federally on state elections.  While this impact is not obvious in every election, and isn't even present in a few, there is a long-established pattern that is the very first thing anybody looking at state elections needs to know.  Once a party has been in government federally for any length of time at all, it starts tending to shed seats in state elections.  This process continues until it has lost most if not all of the states, and not long after that it will probably lose federally as well.  A new federal government comes in with massive majorities in many states, and then starts losing those.  Rinse and repeat.

Since the Liberal-National Coalition came to power federally it has faced four state elections, winning only Tasmania where an ancient Labor government could not escape retribution for a deeply disliked coalition with the Greens.  This happened while the federal government was still relatively new.  However, the Coalition has also lost office in Victoria and Queensland and failed to win in South Australia.  The Victorian result was no surprise, SA was a vaguely forgiveable case of electoral geography 1, 2PP vote 0, but the Queensland result was an earthquake.  The government that had won office with a record majority only three years earlier, almost wiping its predecessor out of Parliament, was sent packing (by one seat) with a 14% swing against it.  The LNP's aggressive and divisive born-to-rule approach in office had done it no favours but things would have been very different had Julia Gillard still been in the Lodge.

With declining levels of rusted-on adherence to both major parties, there's just no such thing any more as a win so big that you cannot possibly waste it in a single term.  That applies especially when the prevailing wind from Canberra is not in a party's favour.  In fact, the idea that a seat margin at one election determines the seat margin at the next is overrated anyway - what winning big really does is just gives a party a lot of seats in which it has new personal votes.  It turns out that in NSW even that doesn't apply much this time around.  So is there such a thing as an election that is too hard to lose, or is nothing actually safe in state politics anymore?

Polling Trajectory

A poll tracker posted by William Bowe last September shows that the Liberal government in NSW has led comfortably through its term.  There may have actually been a slight dip followed by a rebound around the time (April 2014) that Barry O'Farrell resigned after unintentionally misleading ICAC about the receipt of a bottle of expensive wine.  However, there weren't enough polls at the time to be sure.  In any case, the switch to Mike Baird saw the Coalition carrying on with 2PP results in the mid-50s.

This picture was maintained late last year with leads of , for instance, 55-45 in Newspoll (September-October) with 56-44 in Nov-Dec, 54-46 in Essential (Oct), 56-44 in Galaxy (Nov), 54-46 in Ipsos (Nov), 56-44 in Morgan SMS (Oct) and 55-45 from the same poll in November.  There were some notes of caution: Essential had a 52:48 in December and the November Ipsos was only 51:49 by respondent-allocated preferences (a bit sobering given Victorian and federal Ipsos polls have seemed to have a slight Coalition lead).  Overall though it seemed that the new Premier, Mike Baird, was going very solidly and that his overall position was a lead of at least 54:46, perhaps 55:45, by last-election preferences.  You could argue about how much preference distributions might change compared to the 2011 walloping where Labor polled very poorly, but it was clearly an election-winning position at the time.

A few things have changed since.  The first is that the tenure of Labor's initial choice for leader, John Robertson, came to a grinding halt when it was discovered that in 2011 he had written a letter supporting his constituent Man Haron Monis, a serial political nutter and all-round nasty piece of work who later became the gunman in the Lindt Cafe siege.  Robertson was replaced by Luke Foley, who proved to have his own skeletons in the form of two old drink-driving convictions.

Another is that the federal government has again become deeply unpopular, after a recovery of sorts last spring, and has now become embroiled in leadership tensions that have dominated the last few weeks.

These are the polls known to me since Foley took office as Opposition Leader:

A few notes about this table.  The 2PP is the last-election preferences published by the pollster.  In the case of the Ipsos poll, a respondent-allocated result was released of 53-47 to LNP.  In the case of the small Morgan phone poll, the method by which the 2PP was obtained is not stated.  There are a wide variety of methods used for estimating 2PPs under optional preferential voting and that probably accounts for some of the contradictory 2PPs above.

Taking account of sample size (broadly), recency, pollster reliability and the usual (but not in Queensland) problem of the Green vote being overstated by pollsters, I aggregate these polls at 53.7 to Coalition by 2011 preferences, off primaries of 43.7 LNP, 35.7 Labor, 10.1 Green, 10.4 Other.

That sounds like a strong position with four weeks to go and given that it's not that common for results to differ from an aggregate taken one month out by more than a few points.  However ...

Dangers To The Coalition's Lead

The first problem is that we now know from Queensland that most of this lead might not be real.  In Queensland, the gain rate to Labor off all minor party preferences changed from .05 votes per vote in 2012 to .32 votes per vote in 2015, causing the LNP to under-achieve compared to what would be expected from the primaries and 2012 preferences by a massive 2.7 points.  Contrary to my expectation that changes in preference behaviour should matter less in OPV because of the percentage of preferences that exhaust, they can actually matter more.  One reason is that minor party voters who dislike one of the two major parties strongly are locked into preferencing the other if preferencing is compulsory, but may shift between preferencing the opposing party and exhausting their vote (when disillusioned) under OPV.  While just voting 1 seems senseless in the context of a single election, in the long term sometimes exhausting your vote can be a tactic to force your preferred major party to lift its game.  Both in Queensland and NSW in 2011-2, many Green voters may well have exhausted their preference in disgust with the sorty state of Labor in both states.

Potentially, a similar shift in preferencing behaviour in NSW (from Labor's feeble gain rate of .03 votes/vote in 2011) would have about the same 2PP consequences.  The minor party vote is apparently about the same, and if the pollsters are to be believed the Green share of the minor party vote will increase by a similar amount as it did in Queensland.  (I'm a little sceptical on the latter point.)  If Queensland is repeated, then the Coalition's 2PP vote of 54ish and apparently slightly falling is potentially more like 51, at which point things get slightly interesting.

That applies especially if we consider the 2PP might fall further should federal leadership tensions damage the Coalition campaign.

I do, however, have doubts that the preferencing shift in NSW will be as strong.  Mike Baird isn't Campbell Newman; his government may be disliked by left-wingers, but doesn't bring out the same level of loathing reserved for the Newmans and Abbotts of this world.  It is unclear whether a "put the Coalition last" campaign will have the same appeal as "put the LNP last" did in Queensland.  And whereas Queensland Labor was seen as a small Opposition bravely recovering against almost insurmountable odds (the haplessness of the later Bligh days now forgotten), the NSW ALP may be more scarred by the damage done by Obeid, Tripodi, Macdonald et al.

A recent article at The Tally Room shows that shifts in preferencing through 2003, 2007 and 2011 in NSW were relatively minor.  For instance, a return to 2003 preference flows (.1 votes gained per minor party vote) would make about 0.8 points 2PP difference.  However, a graph posted here in comments by reader Joel shortly after the Queensland result (see also associated comments on Reddit) shows that we cannot even rely on a return to the days of successful Labor governments as a preferencing model, and that Queensland had a similar history of preferencing stability until a month ago.  In the best prior Queensland preferencing result for Labor, 2009, Labor had still gained "only" .14 votes per minor party primary.  The 2015 Queensland preferencing pattern was not just a return to normal following Labor's recovery from the depths; it was way off the previous scale.

The other lurking issue is Mike Baird's current popularity.  A good thing for the Coalition?  Yes, in isolation, but we should bare in mind what might be called the Brent Paradox: if a leader is popular, then it is more likely they will become less popular, which may cause their party's ratings to fall. Which brings us to:

It's A Baird, It's A Plane ...

Mike Baird is seriously popular.  His latest Newspoll netsat is +33 (59-26), Morgan's phone poll had him at +22.5 (48.5-26), and Ipsos gave him +42 (60-18).  His Preferred/Better Premier leads are large too (Newspoll 55-25, Morgan phone 49.5-22, Ipsos 54-24, Galaxy 46-22).  (Having a fresh Opposition Leader who is not too well recognised helps there).

But this was also true (to just not quite the same degree) of the last Liberal Premier to go to an election in NSW, John Fahey.  In January and February of 1995 Fahey had a +25 netsat, an 48% party Newspoll primary and a 25-point lead over Bob Carr as Better Premier.  By two weeks out from the election Fahey's netsat had dropped to +9, though the primary vote was still holding up well at that stage.  It didn't hold til election day though: Fahey's government suffered a 3% swing in the final weeks and was defeated.  Fahey had been governing in minority and did share with Baird that he had taken over from a Premier forced to stand down, but Fahey did not even have to contend with having the same party in power federally.

What is odd about Baird's popularity is that he is a Liberal.  Looking at Newspoll ratings since 1985, it is generally Labor premiers (eg Bracks, Goss, Beattie, Bannon; there are several others) who can still poll netsats over +30 after over six months in the job.  It will be interesting to see how Baird's ratings stand up to the test of a campaign.

Luke Foley is polling reasonable personal ratings, but he is so new to the job of Opposition Leader that his ratings are largely irrelevant.  Given the recent pattern that elections are there to be lost by governments (and that if a government deserves to lose the voters may give almost anyone a go) it doesn't seem to matter that his Newspoll netsat is +6, Morgan zero, Ipsos +9 (etc).

Some might argue that having been through a leadership change so recently makes Labor uncompetitive. It doesn't.  Quite aside from some apparent relief that John Robertson has been removed, the track record is that a state Opposition can be a shambles and yet win by default.  This happened in WA in 2008 when Alan Carpenter's government seemed to be cruising and fell over (in part because of a headstrong calling of an early election) although the Liberals in that state had been through four leaders in one term and had finally recycled a "failed" leader from the election before.

Seat Treatment

As noted above I haven't had time to roll out the snazzy seat model (dented by the preference issue in Queensland after a near-perfect performance in Victoria) just yet.  The election starts with 69 notionally Coalition seats, 20 notionally Labor and 4 crossbench (one of them the new seat of Newtown which is notionally Green).  The Labor seats include two gained in by-elections, both of which will be easy retains if the Coalition's 2PP is below 55 or so.

Antony Green has treated the bizarre case of Miranda (where the ALP incumbent retired at the last election, recovered the seat at a by-election, and is now retiring again) as a Liberal seat based on its huge 2011 margin.  My analysis of disrupted seats at state by-elections agrees with this.

There are 14 seats on Coalition margins of 6.8% or less, and with polling pointing to a swing of at least 10% or so, unless a lot changes it's safe to treat that tranche as gone and think about a revised start line of 55 vs 34, with Labor needing to knock off at least another nine to put the Coalition out of majority government.

Next up are Coogee (8.3%) and Kiama (8.6%), which are both double sophomore seats, so add about 2 points to the swing required for those.  The interesting thing is that while in Queensland the target seats for Labor were often bolstered by double sophomore effect, in NSW this isn't the case.  There aren't that many target seats, and of the next eight:

* Seven Hills (8.8%) is actually mostly Nathan Rees' electorate, so the Coalition gets a smaller personal vote bonus for his retirement.

* Holsworthy (10.7%), Gosford (11.9%), Mulgoa (12.4%) and Parramatta (12.5%) were all seats where a Labor incumbent retired in 2011, so there is only a single sophomore effect advantage.

* The Entrance (11.8%) is good for not even that - the Labor incumbent retired and the Coalition incumbent has been ICACed.

* Port Stephens (14.7%) and Penrith (16.1%) are Coaliton-held for more than one electorate.

The next seat above these is on 17.1%.

Port Stephens is the seat at which the Coalition lose their majority all else being equal, and I don't think that making allowances for personal votes changes this a lot.  We can say "by the pendulum" that the swing required is 14.7%, but because of the sparseness of seats around that level in the pendulum, it's possible that it might really end up being a point more or less than that (at least).  Thus, the 2PP deck is probably a little in the Coalition's favour.  That is, assuming they don't drop any safer seats to indies who then back the ALP.  (I'll have a look for those in coming weeks.)

All Up

It's quite possible the Liberal Party will replace Tony Abbott by the time the week is out, greatly altering the impact of Canberra factors on NSW and allowing the Baird government to go to the polls with a new PM who should be enjoying an electoral honeymoon.  If Abbott is indeed rolled, we may see quite different state polling in coming weeks.

Overall it is very hard for Labor to win this unless Canberra factors trash the Coalition's primary even more.  The current 2PP is enough to very likely withstand a Queensland-style change in preferences, the change in preferences probably won't be as large anyway, and the Canberra factor might be close to maxing out even if Abbott survives.  The Coalition won't be going to this one asking minor party voters to preference a party that was about to lose its leader's seat.

I do think it's possible for the Coalition to lose this one as well, however.  As well as the Canberra factor, any government that goes to an election proposing even the most modest of privatisations seems to leave itself open to a backlash.  I don't think Labor will win at this stage, but it's a more realistic chance than it might seem.  NSW Liberals who are nervous that the continuing presence of Abbott as PM might cause them to throw this one away too, have good reason to be.

I'll be back with an attempt at a seat model and projection, and also some Upper House modelling if time permits, around mid-March.

PS I was too busy writing this article to notice that in yesterday's debate, Mike Baird gave Labor significant oxygen by stating there is "no plan B" if his electricity network leasing plan to generate money for infrastructure is blocked by the Upper House.  It's the sort of line that launches a thousand attack ads, and should make for a bumpy campaign.

Tuesday, February 24, 2015

Poll Roundup: Coalition Polling Improves, Abbott's Doesn't

2PP Aggregate: 54.6 to Labor (-1.1 since last week)
Labor would win election "held now" with a massive majority

Quite a lot has happened in federal politics in the last fortnight.  Shortly after the last round of polls (most of them disastrous for the Coalition) were released, there was a leadership spill attempt which Tony Abbott survived by the unconvincing margin of 61-39, even without any declared challenger.  This was followed by various declarations from the PM that he would be changing his spots, and then a steady flow of medium-level gaffes and leaks that mostly suggested otherwise.  None of this shows any sign of stopping, with new revelations about internal problems in the party (often arising from the Loughnane/Credlin conflict of interest) now arriving more or less daily.

Abbott was even the subject of a major article in The Australian, alleging that he had informally proposed sending 3500 Australian troops into ground combat against the so-self-labelled "Islamic State" in Iraq - an allegation that at this stage hasn't been verified by any named source.  Abbott's denials of the claim seemed somewhat coded, but maybe Abbott just wanted to be sure he didn't deny saying something he might have said in jest or as a brief thought-bubble. The polling fortnight ended with Abbott again focused on terrorism issues ahead of a major statement yesterday.

This week's polling

This week has seen a substantial improvement in the Coalition's 2PP polling.  When the Newspoll result (53-47 to Labor compared with 57-43) came out many observers, including me, expected that the change was almost entirely random poll-to-poll noise, but while the size of the shift back to the government wasn't replicated by other polls, its existence was.  Essential moved one point from 54:46 to 53:47 and Morgan moved two from 57:43 to 55:45 (by last-election preferences).  Considering that Morgan leans to Labor by about 1.5 points (my own estimate of the average gap has now come down to 1.3), that makes its result quite similar to the other two.

So we have a bunch of 53-ish results a fortnight after a bunch of results in the 54 to 57 range, which in turn came after a period of sparse polling.  This is a rather volatile polling picture and my aggregate still gives a lot of weight to the nasty stuff from 2-3 weeks ago compared to the three milder results this week.  But the combined weight of a 4 point 2PP swing in Newspoll, a 2 point swing in Morgan and a 1 point swing in Essential is enough to knock 1.1 points off Labor's lead.   There's very likely a large random-noise element in the Newspoll shift (with last Newspoll being a bit high for Labor and this one likely to be lower than reality) but to assume it is all random noise is not supported by the behaviour of other polls.  Here's the smoothed tracking graph:

Before I move on to some issues specific to Newspoll, let's assume this is strong enough evidence of at least a probable shift in voting intention back to the Coalition from two weeks ago and consider why it is occurring.

Sunday, February 15, 2015

Queensland: Final Results And Poll Accuracy

The final results of the Queensland election are out.  Three years after being reduced to something that could fit in a Tarago, the ALP has won office in minority by what may just be a single seat.  (Katter's Australian Party spent so long milking press cover over their agonising decision on who to support that in the end government was formed without waiting for their decision.)  In one amazing echo, Peter Wellington gives Labor the numbers exactly as he did for a few months seventeen years ago.  In another, Antony Green estimates the 2PP at 50.9 to Labor, so virtually the same as 2009.  It is much as if 2012 just never happened.

There are two competing baseline 2PPs for the 2012 election, one of 62.8% to LNP and one of 63.1% to LNP, mainly depending on how you treat Gladstone.  On the latter the swing comes out at about 14%, on the former slightly less.  There may be slight revisions to the 2PP estimate for this election too, but in any case the swing was not much short of 14%, if at all.

In fact, the LNP were a trifle lucky to get as close to hanging on in seat terms as they did.  If the numbers are plugged into the ABC calculator (and crossbench defector/retiree seats assigned as they fell), it suggests 48-38-3 to Labor.  My own seat model, taking personal vote effects and probabilities into account, suggests 45-41-3 to Labor for that 2PP.  The LNP managed one more mainly because they had more very narrow seat wins.

Sunday, February 8, 2015

Liberal Spill And Poll Roundup: The Trouble With Political Jokes

2PP Aggregate: 55.8% to Labor (+3.2 points in three weeks, highest reading of term)
(updated following Newspoll and Morgan)
Labor would win election held now with a massive majority.

The Coalition is now in its worst polling position of the current term
(NB Scroll to bottom of article for updates for polls and other events added from Sunday 8 Feb on)

I've just returned from a week and a half on remote field work, and was going to put out my first poll roundup for the year on Tuesday.  However, all sorts of things might happen by then, so the data might very well be redundant.  I've decided to catch up on the mass of recent polling data first, and then follow up on the early week polling when it comes out.  But first, time for a look back at how (in my view at least) we got here.  My own contribution to the narrative-writing about what is going on now is rather long, so some may wish to skip the next two sections.

Sunday, February 1, 2015

Queensland: post-election wrap and postcount comments

Expected seat outcome: 44 ALP 42 LNP 2 KAP 1 IND

This is my thread for Queensland post-counting comments.  For the first week it probably won't be updated more than daily, if that, and probably won't be detailed.  After that if there is any life left I'll have a more serious look at remaining seats.

Thursday 12th 5:00: One of the few remaining close seats, Lockyer, has been declared, with Pauline Hanson missing out again, this time by a feeble 114 votes.  Non-exhausting Labor preferences split over 58% in Hanson's favour against the LNP.

Thursday 12th 3:30: What little chance existed of the Ferny Grove result being wiped has pretty much gone out the window with news that the number of votes exhausting at the exclusion of the allegedly ineligible PUP candidate is lower than Labor's margin in the seat.  This means that the argument that "had all those who voted 1 PUP and stopped instead voted 1 LNP the result would have been different" does not in fact apply and any remaining substance to an appeal will likely be disposed of based on precedent.

Thursday 12th: The LNP have survived by 126 votes in Mt Ommaney; had Labor won this seat it would have a majority for Labor.  Seats are being progressively declared, with most done now, and soon the election will be over.

Tuesday: I've added nothing more because there's nothing to add; all the leaders are keeping their lead and the expected outcome is done and dusted unless someone finds 500 votes under the sofa.  As noted by Antony Green, the LNP cannot try the usual gimmick of the defeated Premier going back to the House to test his support, because the defeated Premier is not a member of parliament, so it should be that Annastacia Palaszczuk is about to be commissioned as Premier of Queensland.  See also now Pollbludger.

Sunday: Chris Foley is too far behind in Maryborough and will be eliminated based on an indicative throw; Labor will win the seat.

Saturday: There are confused reports about whether Chris Foley has conceded Maryborough (he's apparently conceded he's unlikely to win, which is not the same as conceding he's definitely lost, not that concession makes any difference anyhow).  It appears that during an indicative preference throw he is about 900 behind Labor with about 4000 preferences (presumably from PUP) to add, which is a hopeless position.

Saturday, January 31, 2015

Queensland Election Night Comments

Seats Apparently Won: LNP 39 ALP 42 OTHER 3 
2 likely ALP:  Mt Ommaney, Ferny Grove
Unclear LNP vs ALP: Mansfield, Whitsunday
Unclear ALP vs IND: Maryborough

Outcome: ALP government highly likely, unclear whether in majority.  LNP minority government possible but currently looks unlikely.

10:55 Labor's real-count leads in Mt Ommaney and Ferny Grove are both at 1.1% now.  Those will be hard to close down even with the size of post-count and the LNP needs at least one of them if not both.

10:35 On last-election preferences the current primaries would pan out to about a 50.5% 2PP for the LNP which would probably just have been enough to retain office (especially with likely pickup on postals.) If Labor wins outright then it will likely be because the LNP has been a tad unlucky with the split of close seats as well.  A lot will be asked about why all the late polls were all 52:48, though that is not a particularly bad miss and well within each's margin of error.  It seems there's been a late swing to Labor as undecided voters made up their minds, though this factor is normally no friend of the ALP or oppositions generally.

Queensland: Election Day Comments

Greetings from Corinna!  I don't have much time or much internet but it looks like I will be online tonight, but without access to any TV coverage.

All three polls overnight (ReachTEL, Newspoll and Galaxy) came out with a 52:48 2PP to the LNP so any polling aggregator is likely to have the LNP at something around that.  Mine has finished on 52.1% off primaries of 41.4% LNP, 37.1% ALP, 6.9% Green, 4.5% PUP, 9.7% Other. The important thing is not that 2PP figure but how you model it, and my projected seat tally has barely moved since the first polls after the campaign started showed slight LNP leads.  Out of the 82 major-party-held seats it expects the government to win the 2PP contest in about 48 and Labor in about 34.  Considering the crossbench seats Labor have good chances of two gains, while the LNP have probably as much risk as potential for gain.  (They might pick up Condamine and perhaps Gaven but are likely to lose Maryborough and at risk in a few others).  So I'm expecting something about 48 LNP, 36 Labor and five others.