Tuesday, December 25, 2018

New South Wales 2019: Battle Of The Unknowns

It's a personal tradition to release something on this site on Christmas Day if possible.  This year it was tempting to go with the annual Ehrlich Awards for Wrong Predictions, but one of the predictions on the shortlist still has a remote chance of being fulfilled in the few days left in this year, and also, it's a bit mean on the recipients to drop this year's collection at such a time. So that will come out early in the New Year, and in the meantime, nominations for any false predictions I may have missed regarding political events in the year 2019 are welcome.  See the original Ehrlich piece for the award rules.

Instead I've decided to go with a curtain-raiser for this year's biggest state election in New South Wales.  Currently the expected date of the NSW election is March 23, with the federal election generally expected to follow seven or eight weeks later.  An alternative scenario is the federal election being held on March 2, in which case it is possible to delay NSW until April 13, at the expense of a slowed post-count because of Easter.  There are some other scenarios, but it's highly likely the NSW election will be within about two months of the federal one, with a lot of rather anxious interest surrounding the question of whether it will be before or after.

A very major theme of state election analysis on this site is that the federal picture makes a massive difference.  For whatever reasons, voters are more likely to re-elect state governments when the opposite party is in power federally.  Opposite-party state governments are almost always re-elected while same-party governments are a not much better than even money propositions.  Age also matters - the Berejiklian government is eighr years old.  No Coalition state government (of any state) has won a third or later term since 1986, and it will be forty years since a state's third or later Premier led a state to a majority win at their first election. 

If the election is held before the federal poll, it will be interesting to see how the influence of federal factors plays out.  While voters kicking same-party state governments is certainly a thing, it's not clear whether it's because voters like to ensure power is dispersed, like to send federal governments a message, have their perceptions of the state party coloured by perceptions of the federal party, or use state elections to punish the federal government for its misdeeds.  Perhaps it is all of these things to a degree.  Some of them point to state voters potentially focusing on state factors if they are convinced the federal government will be gone soon anyway (cf Tasmania 1996).  There are not so many recent cases of state and federal elections held close together, especially not in big states, to use as a guide.

The Liberal-National Coalition under then very popular Premier Mike Baird won the 2015 election convincingly, with a 2PP vote of 54.32%, winning 54 of the 93 seats.  In by-elections since, it has dropped Orange (NAT 21.7% vs ALP) to the Shooters, Fishers and Farmers and Wagga Wagga (Lib 12.9% vs ALP) to independent Joe McGirr.  Orange was very close but the Wagga Wagga flogging was so heavy that Labor very narrowly won the 2PP vs the Liberals as well.  The record in NSW is much the same as the record elsewhere - government defeats at by-elections tend to predict 2PP swings against the government at the next election.

Uniform Swing Is A Risky Assumption

The NSW 2019 election, more than any other recently, looms as one where just reading uniform swing numbers off an electoral pendulum could give a misleading result.  The main reason for this is the strange distribution of margins in Coalition seats in the 2016 result.  Six Coalition seats were held on 3.2% or less, but then there are no more between 3.2% and 6.2%.  Yet then there are another five in the range 6.2% to 7.6% and another eight on 8.2% to 9.7%. On uniform swing assumptions there is no difference between a 51.1-48.9 result to the Coalition and a 51.9-48.1 result to Labor.  However, because swings are never uniform, it makes a lot of difference.  If the swing is around the former level, it's likely Labor won't take all the sub-3.2% marginals, especially as Labor's vote is likely to be disproportionately boosted by personal vote swings in the seats they took from the Coalition last time.  On the other hand a swing of around 6% would almost certainly take out some of the seats in the 6-8% range, and perhaps one or two higher up the tree.

Currently the Coalition holds 52 seats, Labor 34, the Greens 3, Shooters, Fishers and Farmers 1 and independents 3.  Losing a net six seats puts the Coalition into minority, while a gain of thirteen would give Labor a majority.  The first would happen on a uniform swing of 3.2% if the Coalition fail to recapture the two seats lost in by-elections.  If those two seats are recaptured, the uniform swing required rises to 6.6%.  But in reality for the reasons above, the swings needed for these two targets would be closer together than that.  The target for Labor for majority government, all else being equal, is 8.7% (there are three seats on exactly that).   Some of any swing to Labor will be wasted on personal vote effects in seats that will become safe if the election overall is at all competitive; on the other hand Labor would hold hopes of taking a seat or two from the disunited (to say the least) NSW Greens.  Very uneven swings would seem to be needed for Labor to win a majority without winning the 2PP by a commanding margin, while the Coalition might get away with a fairly narrow 2PP win, or even if very lucky a 50-50ish result.

Closer to the election, when more sitting MP retirements are customarily known, I will fire up a seat-based model to see what different levels of 2PP swing should be expected to do.  (After the 2015 election I estimated that Labor would need about a 50.6% 2PP for a 50% chance of winning.) But it is worth already noting some of the seats where personal votes are a factor.

Firstly the seat of East Hills (Lib 0.4%) was a miracle hold for the Liberals in 2015, but the miracle was tainted by a nasty and multiply illegal smear campaign against the Labor candidate. Liberal incumbent Glenn Brookes' campaign manager was charged but found not guilty, and the culprit(s) have not been identified.  Brookes was also caught out in a campaign donations scandal and, after leaving then rejoining the party during the term, has decided to rejoin.  Given that the seat is a vacancy, is on a tiny margin and was won in circumstances unlikely to be repeated, it's generally considered a goner.  You can get $5 on the Coalition in it if you don't agree.

Secondly the seat of Lismore (Nat vs Green 2.9%, Nat vs ALP 0.2%). Lismore and its neighbour Ballina were involved in similar three-cornered contests with Labor and the Greens, but the vacant seat of Ballina fell, whereas Thomas George's personal vote helped him record enough primaries and hold off the preference flows from Labor to the Greens enough to survive.  This time George is retiring, which is problem enough for the Coalition, but there is more.  The seat is closer than its 2PP margin suggests.  George defeated the Greens by over 2000 votes, but he only beat Labor on the 2PP by 192.  Labor were only not George's opponent because the Greens got over them by 417.  So it takes only a 0.5% swing from Green to Labor (surely plausible for multiple reasons right now) before the seat reverts to being a 2PP seat, at which point George's retirement means that it should fall if there is no statewide 2PP swing.  So unless there is very good reason to expect that the factors driving the swings in Lismore and Ballina in 2015 have gone away, this seat could be very difficult to hold.

(Those who like the ultimate in electoral wonkery with their Christmas cheer will be delighted to know that a recent paper by Blom, Stuckey and Teague uses Lismore as an example of why the final margin of a seat is not a reliable indicator of easy it would be to change its result by hacking.  This also applies to changing the result by convincing voters to change their minds!)

Thirdly, in Goulburn (Lib 6.6%) prominent sitting member and eleven-year incumbent Pru Goward is retiring.  In theory this should reduce the effective Liberal margin in the seat making it one likely to fall if Labor is getting up into the 4-5% swing range.  However, Goulburn saw a remarkably large swing to Labor (just over 20%) in 2015, and hasn't been won by Labor since 1965.  Did the swing to Labor just about "max out" last time, or could the seat actually fall?  In Victoria there were examples of a large swing at one election being no barrier to another one at the next.

It's also worth noting that Nationals seats (Lismore, Upper Hunter, Monaro and Tweed) are over-represented among the low-hanging marginals.  This means Labor's task will be much easier if it can appeal to these sorts of seats and win them back.  Upper Hunter is another seat where Labor got an extremely large swing (in this case 20.8% off the retirement of a long-sitting incumbent) and it's a seat Labor hadn't historically been even competitive in for decades until then (and hasn't won in over 100 years.)  Feel free to mention other seats of interest in comments.

Recent polling and leaderships

Most of the polling since Gladys Berejiklian became Premier showed the Coalition either modestly ahead or around even on a 2PP basis.  Since Michael Daley replaced Luke Foley, however, two polls have shown Labor with small leads (a 51-49 ReachTEL and 52-48 Galaxy).  This sort of lead, if repeated at an election, wouldn't be likely to be majority-winning, but would give Labor a good chance of forming a minority government. 

This polling might well represent a honeymoon effect for new Labor leader Michael Daley.  The previous Labor leader Luke Foley was always controversial and was widely regarded by progressive politically-engaged voters as a bit of a dud (both tactically and politically).  Daley faces a challenge making himself known to the electorate in time, but voters are usually more concerned with expressing a view about a government, rather than what they think of the opposition. 

If Daley wins, he will be one of the shortest-serving and therefore least-known state opposition leaders to win in the past several decades.  The only similar case at state level in even remotely recent times was Tasmanian premier Ray Groom, who won the state's election in 1992 having replaced former Premier Robin Gray just two months before the election.  At federal level there was Bob Hawke, but in those days ACTU leaders were major political figures, and Hawke's ratings had been being polled and reported by pollsters for years.

Going into the 2019 election we know very little about what the voters think of either Berejiklian or Daley.  Daley's approval rating as leader has never been polled.  Berejiklian's last personal rating was a net satisfaction of +10 from Newspoll back in March, but that's probably ancient history already.  The recent Galaxy had a preferred Premier lead of only 33-31 to the incumbent, and a forced-choice ReachTEL had Daley ahead 54-46.  These results seem unflattering to Berejiklian.  Alongside the understandable concern that the federal Coalition alone could cost the state Coalition the election, there have been plenty of internal concerns about whether the Coalition is selling its message effectively.  Somewhat contradicting that, there is also concern about whether the Coalition should be selling its message of achievement or needs to be offering the voters something new (which both Baird and, in Tasmania, Will Hodgman did successfully in the most recent Coalition victories at state level.)  There is also, somewhat unfortunately, concern about whether the Premier has enough of a public personal story, as if the electorate want their politicians to be celebrities and aren't going to assess them on how they do their job.

Much has been said already about the impact of One Nation.  At the party's current polling level of 8% (Galaxy) it seems unlikely it will win a seat, but if its voters are conservative-leaning but cheesed off with the government they could have a strong impact under optional preferencing.  A high exhaust rate off One Nation voters would spell trouble for the incumbents.

All up the Coalition's position in the Lower House looks pretty fragile at this stage, but it's possibly too early to have a justifiably confident view about which way it is going, especially with the impact of federal factors still to unfold. 

I will be looking at the NSW Upper House closer to the election.  Thankfully it doesn't have Group Ticket Voting and also it has a statewide electorate, so it's a lot less work for us all.  On the other hand, it has constitutionally embedded random preference sampling, which is ridiculous.  And it has eight year terms.  It's possible Mark Latham, David Leyonhjelm and Jeremy Buckingham will all be in there until 2027!  Hope I haven't ruined too many Christmas evenings by mentioning that ...

Saturday, December 22, 2018

EMRS: Small Swing To Labor In Tasmanian Federal Poll

EMRS Tasmania (federal) ALP 40 (+2.1 since election) Lib 33 (-2.4) Green 11 (+0.8) Others 15 (-1.4)
Overall poll suggests a no-change seat result in Tasmania would be likely if election held now
Individual seat results must be treated with caution because of small sample size.

The Liberal Party suffered a major blow early on federal election night 2016, losing the three northern Tasmanian seats that it had captured from Labor in 2013.  In the leadup to the 2019 election, the Tasmanian federal seats have so far not attracted much attention.  This is consistent with a national feeling that with a substantial swing to Labor likely, only a few Labor seats are likely to be in play, and typically where so for unusual local reasons (Lindsay, Macnamara for examples).  The Coalition's strategy is likely to be focused on trying to save as many of its own seats as possible, and going hunting for gains in Tasmania (where PM Morrison has never been high-profile) doesn't seem like a high-payoff-chance strategy.  The northern Tasmanian seats are also protected by personal vote effects for new sitting members, making them harder to shift (not that this stopped them from flipping last time.)

All the same the Labor Party's vote did appear a bit wonky at the Braddon by-election, caused by incumbent Justine Keay's Section 44 problems.  Unlike in Longman, Labor obtained no 2PP swing to speak of there.  There's a case the seat would have been very close, or even have actually fallen, had the Liberals not spent the campaign fighting with the charismatic independent fisherman Craig Garland.  Another seat where incumbents always need to be tokenly nervous is Bass.  While Ross Hart holds it by a comfortable 5.3%, the seat has changed parties seven times at the last nine elections.

The first sample of Tasmanian federal voting intention of any size is a sample of 923 decided voters, presumably piggy-backed off the recent EMRS state poll (which has a standard sample size of 1000, including undecided).  The poll report describes it as conducted for Font PR though I was also told it was commissioned by the Mercury (I will revise if I have more clarity on the commissioning arrangements.)

EMRS federal polling is a rare beast these days though it was more common prior to the rise of ReachTEL.  EMRS polled in the leadup to the 2007 federal election and in that case did not overestimate the Green vote, which has been a hallmark of their state polling (in fact, they underestimated it, though their final poll was taken some time out.)  The current poll includes seat-by-seat samples, but with an average of 185 electors per seat, not much reliance can be placed on these.  At the best of times a 185-vote sample has a maximum margin of error of 7 points, but the problems that have beset seat polling in Australia are such that one should really divide the sample size by six, which means a margin of error more like 17 points!

Based on 2016 preferences, Labor would have a 59.8-40.2 two-party preferred statewide, a 2PP swing to Labor of 2.4%, and that's the main takeaway here.   It's worth doing a run-through of the electorate results (unreliable as they are) just for the benefit of those interested.  The primary votes at the top are from the poll, and the 2PP, 2CPs and swings are my estimates from last-election preferences but using the new boundaries.




The most eye-catching figure in the electorate samples is a supposedly massive swing from independent Andrew Wilkie to Labor in the seat of Clark.  After very narrowly winning Clark (then Denison) from third on primaries in 2010, Wilkie retained commandingly in 2013 and picked up a further 6% primary vote swing in 2016.  While Labor has been rebuilding strongly in the area (for instance picking up a large swing at the state election) I don't trust this very small sample at this stage, and I suspect there's more going on with the poll's 16% swing than random error.  As far as I can tell, Wilkie wasn't named in the readout, and failing to name prominent independents in the readout often leads to a depressed reading of their vote.

On the other hand, naming "Independent" as a separate category (which this poll did - I amalgamated Independent 13% and Others 2% above) tends to get an inflated reading in electorates where none of any note are running.  Thus the poll has 9% for Ind/Others combined in Franklin and 15% in Lyons, whereas in 2016 the actual votes were 4.4 and 10.2.

In general, also, there is more variation in the 2PP swings than would be expected for an actual election, and this is again because of the small sample size.  We haven't seen these EMRS electorate breakdowns with small sample sizes published for a while and I want to again reinforce the message that these are small samples and are not reliable on a seat by seat basis.

The poll also included age and gender breakdowns, notable only in that Indepedent/Others voters were 58% male (not statistically significant, but probably suggests some Lambie voters in the mix as they tend to be male).  Young and female voters tending to vote Green and older voters voting Liberal; none of this is news.

At present I am not aware of any other questions being asked in this poll.

Thursday, December 20, 2018

EMRS: Some Respite, But Labor At Nine-Year High

NOTE: Coverage of the EMRS federal poll will be online here at 9 am Saturday.

EMRS: Liberal 39 Labor 35 Greens 14 Ind/Other 12
Interpretation: Liberal 42 Labor 37 Greens 11 Ind/Other 10
Estimated seat breakdown if election "held now" - majority status touch and go with 12-13 Lib, 10-11 ALP, 2 Green

Three months ago Tasmania's second-term Hodgman Liberal government received a rather nasty wakeup call in the form of the largest poll-to-poll crash in local pollster EMRS's history.  As commented at the time, the timing of the poll was especially unlucky for the government, which was afflicted by fallout from the federal leadership change, unpleasant headlines in the Angela Williamson saga and more criticism of health services.

But in a way that poll belonged to a simpler age for the government, as it was rocked in November by its renegade Speaker Sue Hickey crossing the floor many times over transgender rights amendments.  The government, which had brought in legislation to rectify gender-change divorce anomalies following the passing of same-sex marriage, ended up voting against its own Bill as Labor, the Greens and Hickey added amendments expanding anti-discrimination law in the already-included area of gender identity, making the recording of gender on birth certificates optional and removing surgery requirements for the recording of a gender change on a birth certificate.  Questions were asked about the government's stability but Hickey reasserted her support on supply and confidence, and at this stage remains a member of the party.

Hickey has since crossed the floor again to much less fanfare, and has continued to criticise her own party for underestimating her and not giving her a ministry.  The government is doing its best to route around the problem by blaming Labor and the Greens each time it happens, but it's an ongoing nightmare for them all the same.

So in the circumstances, had someone offered the government a three-point lift as a mark to end the year on, I think they would have taken it, meagre as it appears.  It could be taken as some suggestion of a little bit of random error in the previous result as well.  All the same, the government's lead over Labor is an anaemic four points, lower than all but two polls in the previous term.

Labor Scales The Heights! (Well, slightly ...)

The bad news for the government is that its gain comes at the expense of Greens and Others, and not at the expense of the Labor Opposition.  The ALP recorded a one-point lift that, while deeply insignificant statistically, still takes them to a nine-year high of 35%.  Labor last polled 35% in August 2009 and was last above that level in May 2009 when it polled 43% during then-Premier David Bartlett's honeymoon phase.

The other good news for Labor is that Rebecca White continues polling well personally although the party didn't get near winning this year's election.  White continues to lead Will Hodgman 46-40 as preferred Premier.  Hodgman himself wasn't found to be unpopular when his personal ratings were measured during the election campaign, and I doubt that much has changed.  It seems that White's persistent leads here are something unusual - a case where an Opposition Leader is much liked by voters without the incumbent or his party being badly on the nose.

The Clark Conundrum

EMRS's stocks as a pollster have recently been bolstered by a remarkably good reading of the Hobart City Council election, an extremely difficult race to poll correctly (see here and scroll down to "Well Done EMRS!")  So I am not inclined to cast aspersions at where it has the major parties.  But it does have a long history of having the Greens vote too high, sometimes much too high, in pre-election polling.  Also although its record with independent/other voters is more accurate, its readings for them seem to blow out between elections, possibly as voters engage in wishful thinking about who might be on their ballot paper.  Finally, at recent elections it has tended to underestimate the primary of the incumbent government, more so than any specific major party.

After adjusting for all of these things I take it that an election "held now", but unaccompanied by months of cash-splashing from third-party forces, might yield something along the lines of this:


The exact level of house effect corrections required might be contested, but on an even swing from the state election result, only one seat would be close to flipping, and that is the Liberals' second seat in Clark (formerly Denison), which is now the most marginal of the thirteen seats they hold.  As in the previous term, it is possible the Liberals could stay in office even with a primary vote lead as low as 5-6%, because of the size of their margins in all the seats they won in 2018.

On an even swing the second Liberal Clark seat would be very close to falling (depending on the vote breakdown within the party) and might fall to a third Labor candidate if Labor had someone good enough.  However the 2022 race for Clark is going to be greatly complicated by the question of what to do about Sue Hickey.

At the moment there is a perception that Sue Hickey won't be re-selected for the Government given that she not only seized the Speakership but has also crossed the floor and criticised the Government's tactics and elements within it (also unhelpfully implying it is "right-wing", which is more true of some of its members than of others).  Indeed, Hickey probably would have been kicked out of the party by now except that doing so would make the government a minority government (placing pressure on the Premier to quit) and might even result in a mid-term change of government.  The best case for the Liberals might be that Hickey retires at the next election, but even then they will lose her personal vote and her appeal to left-Liberal voters.  If Hickey runs again as an independent, then it will be much harder to hold two seats in Clark, whether she wins or not.

A lot will change between now and the very far-off 2022 election, and current polling cannot mean a lot predictively.  A change of federal government is likely next year, and a Labor federal government might be gearing up for its first defence or even into its second term by the time Tasmania goes to the polls again.   State governments seldom lose while the opposite party is in power federally.  However, possible 2022 state election scenarios might be in the background as the parliament again considers restoring the House of Assembly to 35 members.  A parliamentary committee set up to consider the matter will be deliberating in 2019 and reporting by August.  (On current polling, it wouldn't help the government - it would win 16 or 17 seats only in a 35-seat House).

Note that I am still not running a Tasmanian polling aggregate, but intend to resume one once there are data from other pollsters.

Tuesday, December 18, 2018

Poll Roundup: 2018 Year In Review

2PP Aggregate: 54.2 to Labor (last election preferences) (+0.2 since two weeks ago)
With One Nation adjustment: 53.6 to Labor
Labor would easily win election "held now"
Labor won all 66 public and three commissioned national polls released this year

With the release of this week's Ipsos and Essential polls, the polling year has probably come to an end.  If there are any late polls I will edit this piece and update it accordingly.

For a government that currently looks as stuffed as a Christmas turkey, the end of the year cannot come soon enough.  As the final poll of the year, Essential offered some respite having the government only six points behind (47-53) but this should be treated with some caution as there is an ongoing difference of opinion between Newspoll and Essential as to just how bad the Morrison government's situation is.  Since Scott Morrison became Prime Minister, Newspoll has had the Coalition primary on an average of just 35% and the Labor primary on 40%.  Essential, however, has had the Coalition primary only narrowly behind (on average 36.9-37.2).  On a 2PP basis Newspoll has had an average reading of just 45.25% for the Government, while Essential has had 46.6% - and this is even though Newspoll's preferencing method is more favourable to the Coalition's than Essential's.  Currently, with Newspoll and Essential coming out in different fortnights, my aggregate bobs around a bit depending on which one is out, rather than based on the Coalition making substantial gains or losses.  If this continues into the New Year I may apply corrections to both.

Saturday, December 15, 2018

Victoria 2018: Final Lower House Results, Poll Performance and 2PP Pendulum

LABOR 55 SEATS COALITION 27 GREEN 3 IND 3
There is no single method of calculating 2PP for this election.  The following are examples of possible figures:

57.57% to Labor (Uniform swing applied to Richmond - probably fairest method)
57.84% to Labor (Richmond treated as 100% to Labor)
57.30% to Labor (Richmond excluded)

With two-party results for all Lower House seats now available it's time to wrap up my Victorian election coverage for 2018, on a high because at least that's the one house where I can talk about the results without constantly losing my temper at the system.  The article again includes a 2PP pendulum.  While this will be of less use for the future than the 2014 one was, given that there is a major redistribution coming, I think it is still useful for looking at the results, and especially at whether the Coalition was lucky not to lose even more seats than it did.

Wednesday, December 12, 2018

Group Ticket Voting Wrecks 2018 Victorian Upper House Election

The buttons have been pressed on the Victorian upper house election.  In the end, none of the results were all that close, and all regions have been declared.  If anyone can find a legal basis for challenging the results, they will now have to do so in court.  On that, it would be nice if professional preference harvesting could be deemed to be a bribery offence under Section 151 (3) (d) but I  suspect that it doesn't work like that, and that that section is aimed at bribery connected with how-to-vote cards.  I can only assume what has happened is all legal, but history should record it as another upper house election that was trashed by Group Ticket voting. 

I should add that this post is not intended as an attack on the calibre of those elected to represent parties with small vote shares.  They may turn out to be excellent MPs.  Rather the point is that they were not elected by a proper electoral system and those elected on very small vote shares do not have a proper mandate.

Tuesday, December 11, 2018

Victorian Upper House 2018: Button Press Day

Introduction

I've decided to bring the Victorian upper house count to the top with a new thread for the day on which all the buttons are pressed, and also for any possible recount news.  My coverage of the count was here.  The schedule is for buttons to be pressed for each division from 2:10 pm at ten minute intervals, in alphabetical order by region name.  However, buttons are being pressed faster than scheduled.  Declarations are scheduled for 6:00 but it is possible that some division will be close enough (either at the end or at a key exclusion point) for a recount to be requested, or that some other issue requiring a recount might be identified.  In 2006 there were two recounts, one because of a 6,000 vote transcription error.

Provisional results will be posted as soon as they are available. I am now reviewing the preference distributions.  There may be some delays in posting analysis (if the distributions are up by then!) as I will be out between 3:45-4:45 but will be online for some of that time.

For all the analysis on the above thread and elsewhere there are some seats that are going to the button in significant doubt.  There are others where what is going to happen appears to be clear but it is possible that all the modelling thrown at these counts might still be wrong and something unexpected will happen.  I should note that at this stage we only have party totals, and do not know if there might be an unusually significant below-the-line vote for any otherwise irrelevant candidate (though there is no reason to think that there is.)