2018 Victorian State Election Intro

It's very close to the 2018 Victorian state election for me not to have written a thing yet about it!  Largely this has been because my analysis model needs polling to work, and (in common with other recent state elections) there's hardly been any of it.  Anyway, this is an opening offering on some general issues in trying to forecast this election

Let's start with the important bit.  If voting in the Legislative Council (upper house) in Victoria, vote below the line for candidates, not above the line for parties.  You only have to choose five candidates for a valid vote, though you will make your vote a lot more powerful if you number a lot more.  If you vote above the line, your vote will be at the mercy of your party's decisions about where to send your preferences, and they may well choose to send it to a party who you'd be totally opposed to.  Unfortunately, Victoria is one of the two remaining states that has not got rid of the Group Ticket voting system.  Keep control of your own vote and say no to preference-harvesting which can lead to unknown parties electing unaccountable candidates off tiny percentages of the vote.



I hope to have time to analyse the potential impacts of Group Ticket Voting at this election, but I will be offline for a week a few weeks out from the election, so by the time I get onto it, people may already be voting.

The Weight Of History

As for the lower house, which will be the focus of my usual predictive efforts, the past history of state elections does not bode well for the Liberal opposition.  As I noted in my hopefully vaguely seminal post What Kills State Governments: Age Or Canberra the two biggest danger factors for state governments are long incumbency and being the same party as the federal government. The defeat of the Weatherill government in South Australia earlier this year marked just the seventh time in the last 35 such cases in which a government has lost while the same party was out of power federally.  So only 20% of governments not subject to what I call "federal drag" lose, while those governments that are the same party as the federal government are kicked out about half the time.

Moreover, of those opposite-party governments that have lost, most of the cases have involved leadership instability and/or loss of majority on the floor of the Parliament.  None involved a first-term government, let alone a majority first-term government.

If we want to find the last time a state government lost after its first term while also in opposition federally, we will be going back a long, long time!  The last twelve such state governments nationwide in a row have won.  The last one to lose was the Walsh-Dunstan SA Labor government in 1968, as a result of malapportionment.  Before that the previous such case was in 1955 (Cain Labor in Victoria as a result of the DLP split) and prior to that we have to go back to the Depression era, when there were a string of first-term losses (often also split-ridden) by federal Oppositions in NSW and SA.

So losing after one term in office while also in Opposition federally is an event from the days of the dinosaurs.  And while the Andrews government has a remarkable ability to get itself into scrapes (whether it's the Red Shirts Rorts, fights with firefighters or even fights between its own MPs) it's not clear that all that is enough to overturn the history of these things.  It seems more likely to just knock margin off what in all normal circumstances would have been a thumping win to a first-term government assisted by a Canberra Coalition regime that has lately been a constant source of farce.

A question is the potential of the rorts issue to cause Comey-style disruptions that hurt Labor badly during the campaign.  It is unclear though whether this issue has enough breakout beyond the readership of the Herald Sun to cause the party serious ballot box harm. 

Labor's Disadvantages

Having said that, Labor faces significant disadvantages in turning votes into seats.  Most new governments go into their first re-election attempt with a brace of seats firewalled by personal vote effects as a result of winning them at the previous election. While such seats will still often fall if there is a swing back, the two or three that don't can make all the difference.  Labor, however, made only four pickups (the four "sandbelt" seats) in 2014, and one of those was against Geoff Shaw, which reduces the personal vote bonus.  Labor goes into the election with ten seats on under 4% vs the Liberals compared to the Liberals' five (plus a 2-1 split of such seats vs the Greens), and this means that they are at more risk from random variation in swings than the Liberals are. Overall, all else being equal and if there is little net swing, the Liberals would be expected to gain a seat out of this.

The second issue is population growth.  Since the last redistribution the average population of seats that are Labor-held on a two-party basis has risen by 16% but that of Liberal-held seats has risen only 9.5%.  As Antony Green has noted, the Liberals are in for a lot of pain at the next redistribution, but that's another way of saying that population growth in the western suburbs has created a skew in the Liberals' favour - if all goes well for them they need fewer votes for the same number of seats.  It's quite possible pollsters won't be wise to the problem and that the current polling might underestimate Labor's 2PP slightly for this reason.

Caution should be exercised in calling close-ish seats with large enrolment growth too early on the night, so here's a list of some of those: Bendigo East, Yan Yean, Albert Park, Cranbourne, South Barwon, Bass and Prahran.  But Prahran is bound to cause other problems anyway ...

Polling

Which brings me to polling, or more pertinently, the lack of it.  It's a common theme in state elections lately but with four weeks to go til the big day Victoria has had just seven statewide polls all year! These were two Newspolls, one YouGov Galaxy for the Herald Sun, two Age ReachTELs, and two YouGov Galaxys for the Victorian Bus Association (for the first of which only incomplete results have surfaced.)  All these polls have shown Labor with narrow leads (between 51% and 53% 2PP), and every poll conducted in the term has shown Labor ahead with the exception of one tied Galaxy and two private polls (one of unknown provenance) back in 2017. The last three polls have been 52s and 53s, one of which could easily have been a 54, and this is the best polling the government has seen since 2016.

The Herald Sun Galaxy found no difference from the 2014 election except a modest and statistically insignificant gain to Others at the expense of Greens.  The most recent Age ReachTEL found Labor on 37.1% after distributing soft voters (who ReachTEL calls "undecided" voters), down a point on 2014.  The Coalition were on 40, down two points.  The Greens were on 10.5, down a point, and Others were on 12.4, up four.  However the latter may have something to do with the poll's unusual design, which named four micro-parties (Reason, Shooters Fishers + Farmers, "Hinch Party" and Animal Justice Party) that will all be focusing mainly on the upper house and probably won't run a lot of lower house candidates.  Polls have struggled with underestimation of Others votes at recent by-elections but I am not convinced this is the answer.

The VBA Galaxy, released minutes after I first released this article, has Labor on 40 (+1.9 since election), Coalition 39 (-3), Greens 12 (+0.5) and Others by implication 9 (+0.5).  The published 2PP of 53-47, in the absence of a One Nation campaign, seems a little stingy; by last election preferences it would be at least 54-46.  This poll should be treated with more caution than the others given that it is commissioned by a lobby group.

While we have not seen much approval polling, it is also worth noting that Daniel Andrews' ratings have been generally harmless, including a -4 net satisfaction in the last Newspoll back in April, a -2 from YouGov Galaxy in September, and improvement to +9 in the VBA Galaxy now.  Very few Premiers survive the next election from net ratings well into negative double digits, so that's another box for predicted defeat that Labor has so far avoided ticking.  One would think if there was massive outrage about any of the issues Labor has had in this term, Andrews' polling would be worse than that.  Matthew Guy continues to poll poor ratings (most recently a net -18) but Opposition Leader ratings are not such a big deal (just ask Steven Marshall!)

Seats of interest

The following are some seats that are of special interest in this election for reasons other than just being close on the 2PP; feel free to suggest others in comments.  Noel Towell has a good rundown of seats to watch including the close 2PP seats and the possibility that Bass (Lib 4.6%) could be undone by demographics.  Matt Johnston and Tom Minear have another one with lists of issues for the electorates.

Prahran (Green vs Lib 0.37%, Lib vs ALP 0.03%)

Prahran was a micro-close three-cornered contest at the 2014 election.  The Liberals will be a 2PP combatant in the seat but Labor might entertain some hopes of jumping the Greens and making the final two.  This, plus the complete lack of evidence on how Greens go when defending close Green-vs-Lib seats (in Green-ALP seats their primary vote seems to rise at Labor's expense, based on a sample size of two) makes this seat difficult to model.  At this stage I am inclined to treat the seat as Green vs Lib, giving Sam Hibbins the benefit of a double sophomore effect in that contest (which means his on-paper buffer is more like 2%) but a general lift in the Labor vote or slump in the Green vote could make that not the best way to go.  Such is the closeness of the seat that a primary vote swing from Liberal to Labor could cost Hibbins the seat without Hibbins losing any votes himself.

As I write a commissioned ReachTEL by Bike Melbourne claims a large swing to Labor at the expense of the Liberals, which if true would put Labor into the final two at the Greens' expense and easily win Labor the seat.  However, warnings about the general dodginess of seat polls (reinforced in Wentworth where the polls were total garbage) need to be amplified in this case with a sample size of only 600 and because polls (ReachTELs especially) have been especially poor in these inner-city contests involving the Greens.  The poll also as reported has a two-candidate preferred between Labor and Green, which is silly - even on those numbers, the Liberals would make the final two in Prahran.

Northcote (ALP vs Green 6.0% but disrupted by by-election (Green vs ALP 5.6%))

In classic 2PP seats where one party loses the seat at a by-election there is a usual trend to use the margin from the previous election, but the evidence (which I discussed in the leadup to the 2015 Queensland election) is that these "disrupted seats" behave differently from normal seats.  They tend to produce a result that, after adjusting for any swing between the elections, lands about between the last general election and the by-election result.  We have no idea how this works in ALP-vs-Green contests.  To further complicate matters, this area has been moving to the Greens for demographic reasons (which would suggest a Green retain in the absence of further swing), but at the Batman federal by-election it swung back to Labor.

While the Greens should hold Melbourne (2.46%) without much trouble with the assistance of personal vote effects, two other ALP-Green seats are interesting, Brunswick (ALP 2.22% but with Jane Garrett retiring) and Richmond (ALP 1.86%).  Third-time Richmond Greens candidate Kathleen Maltzahn faces opposition within the left over her long-term but recently recanted support for "Nordic model" sex work policies, but that was also factored into her vote last time.

Adding interest to these contests, the Liberals have not announced candidates for any of the four inner-city Greens targets and have long been expected to pike on them completely (which will mess up the 2PP vote).

Morwell (Nat, 1.8%)

Morwell is on a slim margin and the sitting MP has quit the Nationals for reasons deserving of sympathy but unlikely to be conducive to re-election; indeed it is not clear he will even be a candidate. Normally Russell Northe's departure taking a 12-year personal vote with him would be enough to put this seat on the wire but a further complication is that in 2014 Morwell saw an extremely large 2PP swing of 11.5% to Labor, likely to be attributable to the long-burning Hazelwood coal mine fire.  Normally, extreme swings at one election increase the chance of a swing back at the next.  Whether the Nationals are in any position to bounce back from that swing remains to be seen.

Ovens Valley (Nat, 16.6%)

One wouldn't think there was much to see here from the margin and from odds like National $1.01 Labor $12 (alas no independent option!) but this rural seat has hallmarks of indie bait.  For starters, the incumbent (Nationals' Tim McCurdy) is facing fraud charges during the campaign and has been criticised within the Coalition for even contesting the election.  Secondly the seat, a successor to Murray Valley in which the Nats hadn't faced a serious challenge since 1952, overlaps the federal Division of Indi and is the subject of an organised attempt to clone Cathy McGowan's federal win there at state level.  The primary challenger, Tammy Atkins, is an escaped National, and Labor seem to be doing the swan-dive on this one, finally announcing their candidate this weekend after months of speculation about whether they would even bother.

Hung Parliament?

Counting Morewell as Coalition and Melton as Labor, Labor goes into the election holding 46 out of 88 seats, the Coalition 38, the Greens 3 and one independent (Suzanna Sheed, who I'm assuming will be re-elected barring strong evidence otherwise.)  A loss of just two seats would take Labor out of majority, and if the Greens have a good run in the inner city then that could be enough to do it all by itself.  However, the Greens have had their issues in Victoria (including fallout from the Batman by-election).  So far the statewide polling doesn't show anything to suggest they're on track for a 5/5 result, but state polling wouldn't necessarily pick it up if they were, since Green voting can intensify in inner cities without any accompanying broader change.

At present there is not a strong expectation that any particular seat anywhere will change hands (whether between the majors or from the majors to third parties).  That presents an environment in which unlucky results could easily cause a hung parliament (and based on Labor's risk profile they seem to be more at risk of this.)

I think the big question here is whether Labor can pick up a swing.  If there is no significant change from the last election in voting intention, it seems really touch and go whether Labor will hold on in what currently looks like another low-seat-turnover election.  A swing to the Liberals of even a small size would make it very difficult to maintain a majority.

When I get enough fresh juice in terms of polling (another state poll, preferably "public", would be nice) I'll unveil some estimates from my polling model of classic and ALP-vs-Green seats.

Comments on seats to watch welcome.