Tuesday, March 12, 2019

Poll Roundup: Medevac Fails To Shift The Dial

2PP Aggregate: 53.4 to Labor (+0.4 since last week)
52.9% with One Nation adjustment
Labor would easily win election "held now" with about 90 seats.

In the wake of the Coalition's 50th consecutive Newspoll 2PP loss, here's another roundup of the state of federal polling, and I also include some comments about the state of seat betting, which I track in the approach to each federal election.

A few weeks back, just after the last roundup, there was a lot of hot air about a possible "Tampa moment" for the Coalition in the form of a close Ipsos poll immediately following the passing of "medevac" legislation by Labor and the crossbench.  Since that Ipsos 51-49 (with the Coalition lucky to get 49 on the published primaries anyway) we've seen Newspoll at 53-47 and 54-46, Essential at 52-48 and 53-47 and a commissioned ReachTEL at 53-47.  There was also a Queensland-only YouGov-Galaxy with a 6.1% swing to Labor in that state.

Overall, the other polls provide no evidence to support the movement seen in the Ipsos poll.  Someone might say that voters initially reacted to the "medevac" controversy and then reacted back when politics returned to its regular fare of Coalition scandals, stuffups and infighting.  But I've never seen any evidence that voters react so waywardly, with hundreds of thousands jumping back and forth in response to Canberra stuff that doesn't affect them at all or to behaviour patterns that have been commonplace through the current term.

It's far more likely that the Ipsos was simply a bit off the ball as a result of its bouncy nature.  This is the eighth time out of its 15 polls this term that Ipsos has differed from my aggregate by more than a point even after it was itself included.  Of those eight, this is the fifth time the movement in Ipsos has failed to foreshadow a movement by the other polls in general, compared to two cases where other polls did follow the path Ipsos blazed, and one (the 55-45 in the final days of Turnbull) for which Ipsos was probably overstating real movement, but we'll never really know. 

After considering the preferences and applying last-election preferences, I aggregated the Ipsos at 51.4 to Labor, the Newspolls at 53.9 and 54.6 and the Essentials at 52.2 and 53.  My aggregate is currently at 53.4 to Labor.  Here's the smoothed tracking graph:

The Coalition is faring just a little bit better than late last year, but it's not much, and there are still no signs of a sustained and serious recovery or even the ability for one to occur.  Also, it is again in the position (assuming an election in May) that no government that has been this far behind with so little time to go has won.  The two governments that were challenging it on that basis (Whitlam's in 1974 and Howard's in 1998) were both recovering to some degree by this stage (at least if assuming public opinion was changing gradually between one poll and the next).

Essential had some questions about tax policy last week (example) that raised a few (mostly conservative) eyebrows and I'm not really sure what they thought they were doing with those.  The questions seem designed to test how potential Labor policies would go when brought into mortal combat with a cardboard-cutout defence of Coalition policies.


There continues to be little to see when it comes to Scott Morrison's personal ratings.  His Newspoll satisfaction and dissatisfaction ratings currently remain very similar (he's at 43-45 for a net rating of -2).  As with Howard in 2007, voters don't think the PM is the problem, and so the normal relationship between PM ratings and the 2PP has broken down.

Opposition Leader Bill Shorten continues to poll somewhat poorly but less so than when he was opposing Malcolm Turnbull.  This week Shorten yet again recorded a net rating of -15 (36-51).  There is little to see on the misleading Better Prime Minister rating with Morrison continuing to "lead" 43-36, which as regular or even irregular readers of this site will know, means he's effectively behind, because Better PM has a massive house edge to incumbents.

A source of much well-deserved opprobrium this week was Peter van Onselen's claim that "It seems that we can now conclude that Shorten is the least popular opposition leader in Australian political history, more so even than the deeply unpopular Abbott." This claim was premised simply on Bill Shorten having racked up four years in a row with a negative Newspoll netsat.

Length of unpopularity is a poor measure of being the "least popular opposition leader"; what it actually measures is the most enduringly unpopular.  And that's a dubious measure, because several previous opposition leaders were given the flick before they could last long enough to see if they would remain in negative territory for as long as Shorten, and were often flicked because they were unpopular.  On the other hand, Tony Abbott's unbroken streak of unpopularity was broken by winning an election, the apparent difference between him and Shorten being simply that he came to the leadership late in his party's first term in Opposition, while Shorten did so more or less at the start.  Indeed of the 20 previous Opposition Leaderships since the start of leader rating polling, only three (Hayden, Hewson and Beazley) lasted as long in a single go as Shorten has been polling sub-zero netsats for.

While any measure has its issues, three more useful measures of opposition leader unpopularity could be career mean netsat, career median netsat and career worst netsat.  On carer average (mean), Shorten (-16.1) is the second worst to date, behind Alexander Downer (-17) and ahead of Tony Abbott (-14.42) and Simon Crean (-14.38).  (No-one else is worse than -10).  On career worst, using a mix of Newspoll and pre-Newspoll Morgans, here's the list of those who have at some stage been worse than -30:

-49 Downer
-47 Peacock
-42 Hewson
-41 Snedden
-39 Crean
-38 Shorten
-36 Abbott
-35 Howard, Beazley
-33 Turnbull

On career median (which takes out both initial bursts of popularity and brief bad spells) Downer is by far the worst with -34.5, compared to Shorten's -18, Abbott -16 and Crean -12.  So on all the measures I've looked at, Alexander Downer is the most unpopular leader.  Of course, we can debate whether Downer was there for long enough for that conclusion to be reliable, but at the least van Onselen's metric is flawed and other metrics give a different result, so we cannot now conclude what he asks us to.  Do I really need to write another sentence starting with "A professor of politics really should know better"?

Local/Seat Polls

A Unions ACT uComms Senate poll has been reported, but unfortunately full results have not been released. It appears - but I'm not certain - to be of the whole ACT whereas a previous Senate poll for the Greens rather sneakily polled only the Greens-friendly seat of Canberra (one of the ACT's three electorates!) Senate polling is notoriously unreliable, not least because it fails to capture how the voter will respond to a ballot paper with many choices.  This particular exercise has dubiously included "Independent" as a standalone option, which for some reason often results in inflated readings even compared to "Independent or other".  It's not even clear on the numbers in the poll whether Zed Seselja would actually lose, since in 2016 the Greens would have gained just two points on the Liberals had all preferences been thrown, and this poll doesn't give Labor much of a surplus, so it would all come down to those who preferred "independent".  I just about wrote off the Greens winning Seselja's seat when I first looked at it, and I still think it's a tall order, but it's not completely inconceivable.

More GetUp! uComms polls of Warringah (supposedly 57-43 to Zali Steggall) and Dickson (52-48 to Labor) have had their existence sketchily alluded to.  I have declined to use the word "reported" or link to the coverage at all given that the Guardian has failed to publish any detail of the primaries and they haven't surfaced anywhere else that I'm aware of either.  It's doubtless true that both the sponsor of the poll and the reporter were more interested in the poll's findings re "medevac", but it is pointless reporting poll results on issues questions unless they are reported with enough voting intention detail (preferably the full tables, not just the primaries!) to decide whether there was anything obviously wrong with the sample.  Something funny that has been showing up surprisingly often when tables for ReachTEL platform polls (uComms or not) have been printed lately, is weirdly conservative breakdowns in the 18-34 year age demographic.  My general response when I've seen commissioned seat poll polling tables lately has been that I'm often surprised to see things wrong with them that I didn't even expect I might see.  No wonder seat polling is such a mess.

Seat Betting Watch

From time to time (and roughly weekly once the formal campaign gets underway) I will be having a look at what seat betting markets are doing.  Seat betting markets are not reliably predictive (contrary to the unsourced and unsubstantiated claims now and then seen in commentary) but they are worth keeping an eye on for two reasons. These are firstly to see to what extent (if any) they do turn out to have special insight not available through other methods, and secondly to hose down the nonsense that sometimes appears about "what the market thinks".  The most common errors in reading seat betting are (i) treating the probabilities in all seats as completely independent of each other (they are not) and (ii) only looking at who is favourite in the most seats while ignoring whether there is any imbalance in how many seats each side is narrowly favourite in.

At the moment I am aware of two active markets (Sportsbet and Ladbrokes).  I class a seat as a "close seat" if there are two or more candidates at $3.00 or less on any market, or for three-cornered contests where there is no candidate at $1.40 or less.  Close means the odds are reasonably close; it doesn't always mean the market expects the margins to be.

Counting the Coalition as from a base of 75 seats (including notionally Labor Dunkley, defecting independent held Chisholm and sort-of crossbenchy Page), Labor from a base of 70 and crossbenchers from a base of 6 including Wentworth, I get the following:

Expected ALP classic-seat gains (not close): Dunkley, Corangamite, Capricornia, Forde, Gilmore, Flynn, Robertson, Banks, Chisholm, Reid

Expected ALP classic-seat gains (close): Petrie, Dickson, Hasluck, Page, Boothby, Dawson, Bonner, La Trobe, Pearce, Swan, Leichhardt (tied in one market), Brisbane, Deakin, Flinders

Expected Coalition classic-seat holds (close): Casey, Sturt, Stirling, Canning, Bowman, Aston, Higgins, Hinkler, Bennelong, Kooyong

Expected Coalition gain from IND (close): Wentworth
Tossup IND held vs Coalition: Indi
Expected Coalition holds against IND (close): Mallee, Farrer, Warringah
Expected IND gain from Coalition (close): Cowper

Expected Labor hold vs Green (close): Macnamara
Expected Green hold vs Labor (close): Melbourne

So at the moment the seat markets expect the Coalition to lose everything on 4% or less, plus Cowper to Rob Oakeshott, and also Reid, Brisbane, Deakin and Flinders.  They also expect the Coalition to recover Wentworth with a tossup in Indi.  In this case, there actually isn't a serious imbalance in close seats.  What is interesting here is that despite the usual pattern of voting intention narrowing as a federal election approaches, the seat markets aren't factoring this in, and if anything expect it to get slightly worse.

The following is the start of the chart of favourite-tracking I will do for seats in which the incumbent isn't favourite across all markets at any stage.

Key to colours (more may be added):

Red - Labor favourite in all markets
Orange - Labor favourite in some markets, tied in others
Dark blue - Coalition favourite in all markets
Light blue (none currently) - Coalition favourite in some markets, tied in others
Grey - all markets tied or different favourites in different markets
Purple - IND favourite in all markets

Seat total betting in bands of five seats for Labor is a bit less bullish for the Opposition so far - one market expects them to land in the low 80s and another was expecting them to land in the high 80s but disappeared as I was writing this article.  The individual seat betting markets, after adjusting for close seats, can be read as giving Labor about 94 seats.

At this stage markets would be thin and likely to be shaped largely by bookie judgement rather than being dominated by punter money.

I expect to be back in a few weeks with another roundup for Budget polling, and after that it will be one roundup per week leading into the election.

Oh and voting is still open in the sidebar on a couple of Best Premier Not-A-Polls, leading into what I expect to soon be the start of the runoffs between the state winners.


  1. excellent as usual Kevin.

    Well done

  2. Libs favourite to win back Wentworth? That means that the majority of people with so much spare money that they're prepared to risk it by betting on elections think/hope that the Libs will win. Weighted, of course, by _how_much_ spare money each bettor is prepared to risk. And who are the people with so much spare money that they're prepared to risk it by betting on elections? Probably died-in-the-wool Liberal voters who can't imagine anyone not voting Liberal. Meaningless!

    1. Typically bookies only accept relatively small bets on seat betting to avoid being stung by bettors with superior local or inside knowledge. In the case of Wentworth I doubt all that many people have been betting so far at all.

  3. Yeah sorry, I only noticed your remark that so far markets have probably been shaped by bookie judgment after I posted the above. But even then, what are they judging? The likely flow of money, not the actual probabilities as to who will win or lose. And of course they offer some teasing tempatations to the mugs - like "Australian Conservatives $101, One Nation $251, Greens $251" in Ladbrokes' odds for the next federal election (see the PB site). Longs-odds horses dooccasionally win, but none of that bunch are going to win government! I wonder how many optimists will spend a dollar or two just in case the utterly impossible happens?