Saturday, May 18, 2019

Election Eve: Final Newspoll and Seat Betting Roundup

2PP Aggregate (Final): 51.6 to Labor (Last-election preferences)
(51.3 with One Nation adjustment)
Seat projection assuming polls are accurate 79 Labor 66 Coalition 6 Others

The final Newspoll marks the end of my poll aggregation for this term (barring any late offers) though in the last few weeks, I really needn't have bothered.  Aggregation is most useful when polls are bouncing around, when they have marked house effect differences relative to each other or when there is something happening that is causing voter intention to move.  In the last few weeks, almost nothing has happened and seventeen consecutive polls have landed their 2PPs somewhere in the range 51-52.  There's really nothing for an aggregate to do!

Anyway the final Newspoll has come in at 51.5 to Labor, following this week's 51 from Ipsos, 51.5 from Essential, 51 from Galaxy and 52 from Morgan (who currently stand to take the 2PP marbles if there turns out to be even a whiff of a Bob Hawke effect).  My aggregate has finished up at 51.6 to Labor by previous-election preferences.  If you want to make a Newspoll-style One Nation adjustment to that you can get 51.3.  It actually makes little difference to the seat projection.

The Newspoll gave Scott Morrison a final net satisfaction rating for the term of +1 (46-45, the 46 being his highest satisfaction score to date) and Bill Shorten one of -8 (41-49).  This saw Bill Shorten out of negative double figures for the first time in 85 Newspolls, and the 41% was his highest satisfaction rating since February 2015.  He continues a record run of 88 negative Newspoll netsats, but one way or another this is extremely likely to end at the election.  The Preferred Prime Minister score (47-38) was a nothing-to-see-here score given Labor's 2PP lead.

The Newspoll and my aggregate more generally suggest a small Labor majority. Attempts may be made to spin it as a hung parliament based on the idea that Labor might flop out in region X, fall short in region Y, etc, but I've experimented a lot with this in my seat model and generally what goes up comes down somewhere else.  Of course, it's well within the margin of error of such things that a 2PP in the 51s could be a hung parliament, but it's more likely it wouldn't be.

I'd like to be in a position to offer some magical objective insight that points to these herded polls being wrong in some specific direction at the time they were taken, but I've got nothing.  I'll be less surprised by at least a little blowout in Labor's favour than if the Coalition manages to make this really close or even win, but we will see tomorrow night.  If there is a blowout, it will be widely attributed to the "Hawke factor" but determining whether that is the real cause or the polls are just wrong will be tricky and we may never really know the answer.  

In its rawest form my seat model expects the Coalition to lose Dunkley, Corangamite, Capricornia. Forde, Gilmore, Robertson, Banks, Petrie and it has a bunch of seats on 50-70% retain chances that would make up another three expected losses to Labor.  It expects Labor to lose two to the Coalition, but doesn't have any specific Labor seat at less than 50% (Herbert and Lindsay are the worst).  The obvious adjustments are for the general trend of good polling for the Coalition in NSW, poor polling in Victoria, urban Qld and WA, and help from right-wing parties on preferences in regional Queensland. When these are added in it now predicts losses in Dunkley, Corangamite, Forde, Gilmore, Petrie, Hasluck, Chisholm and La Trobe.  I subjectively think that Swan would go as well (the Beazley surname!).  My model also, post these adjustments, has Robertson, Banks, Page, Flynn and Dickson (and on the Labor side, Herbert and Lindsay) as very precarious, but doesn't project any of these as quite falling on the aggregated poll result.  

Dickson is a very hard seat to predict.  On Peter Dutton's side are his increased prominence since the last election and the departure of Malcolm Turnbull, which should play better for him in an outer suburban seat.  On the other side there is his gaffe about his opponent Ali France using her disability as an excuse to not move into the seat, the scale of the campaign against him (even bigger than last time) and embarrassing news about him letting two suspected murderers into Australia that has broken in the last days of the campaign.  I've seen some Labor sources saying they're very confident of winning it and others say they wouldn't have the foggiest.  I don't really have the foggiest either.  

My seat poll doesn't deal with independents but I think the government could recover Wentworth and Indi while losing Farrer and Warringah and a fair chance of losing Cowper as well.  (If I modelled Cowper as if it was a 2PP seat in my model, the Nationals would be projected to just retain it, but precarious.)  There is also a messy situation in Mallee where multiple independents are trying to dislodge the Nationals.  

Seat Betting

I continue expending effort on measuring whether seat betting might be more predictive than anything else, though in this case it looks like it's rather closely converged with other sources of predictions.

Not much change in the last few days (see here for methods)

Expected ALP classic-seat gains (not close):  Dunkley

Expected ALP classic-seat gains (close): Dickson, Petrie, Gilmore, Robertson, Chisholm, Hasluck, Swan, Corangamite, La Trobe, Forde

Split market in Coalition classic seats: Stirling, Reid, Capricornia

Expected Coalition classic-seat holds (close): Bonner, Dawson, Leichhardt, Banks, Page, Casey, Deakin, Flinders, Higgins, Pearce, Boothby, Flynn

Expected Coalition hold vs Greens: Higgins

Expected Labor loss to Coalition (not close): Herbert
Expected Labor loss to Coalition (close): Lindsay
Split market Labor held vs Coalition: Braddon
Expected Labor hold vs Coalition (close): Bass, Solomon

Expected Coalition loss to IND (close): Cowper, Warringah
Expected Coalition gain from IND (not close): Wentworth
Expected Coalition gain from IND (close): Indi
Split market Coalition vs IND: Farrer

Expected Labor hold vs Green (close): Macnamara

The colour-coded favourites tracker:

Based on a current expected swing of just under 2%, the seat markets are bucking the pendulum in Lindsay, Herbert, Flynn, Banks, Dickson (just), Hasluck, Chisholm, La Trobe and Swan.  They have 11 Coalition seats more likely than not falling, all of which have been falling all along.

For prediction-testing purposes I should define a favourite in the various split markets by averaging the odds, so these are the Coalition in Reid, Farrer and Stirling and Labor in Braddon and Capricornia.  However the most used market, whose name I decline to mention because it paid out on the election early, has the Coalition favourites in the latter two and Farrer as a tie.  So across the four markets being checked I have Labor as favourites in 80 seats, the Coalition 65 and others 6.

Seat total markets now have Labor 76-80 seats as favourite, very closely followed by 81-85.

Once all seat winners are known I will do another article on the success (or not) of the seat betting markets.

Other predictions:

I have not seen many predictions of seat totals this election.  William Bowe's Bludger Track has 51.7 2PP and 80-65-6.

I will have a look round for some more, by psephologists or high profile commentators, to add in the morning.


  1. Hi Kevin,

    Thanks very much for all your work on this election. In 12 hours time I hope there is a clear result. I am looking forward to your post election analysis.


  2. Great blog, wish i found it sooner.

    Just wondering how much the popularity of the leader correlates with vote for the party candidate in marginal seats? I think polls consistently have Shorten trailing Morrison as preferred PM (32 vs 34 or thereabouts?) while Labor consistently stays ahead as the 2PP. Assume the PM factor must be significant or they wouldn't bother asking.

    1. "Assume the PM factor must be significant or they wouldn't bother asking." Unfortunately media continue to make a big deal of Better PM despite overwhelming evidence that it skews to the incumbent (by around 16 points).

      The PM's net personal rating has a stronger relationship with voting intention, the Opposition Leader's does not. So if Morrison has a strong net personal rating (not better PM rating but approve vs disapprove) in a seat, his party is more likely to do well in that seat. But that's all else being equal.

  3. @Jim

    Kevin has written posts before about the preferred PM polls essentially being useless.

    Read this one, although it's from a while ago:

    He has said elsewhere that there's an incumbency benefit, I think.

    But remember for example that John Howard won many elections, despite trailing in PPM. Gillard was leading Abbott in the PPM when she got thrashed in 2013 too.