Tuesday, December 13, 2016

Poll Roundup: 2016 Year In Review

Aggregate 52.0 to ALP (+0.1)
Labor would win an election "held now"

It's about the time of year when polling shuts down for a few weeks, so as usual I will post an annual recap.  If there are any late polls then I will edit the text to add them in.

In the three weeks since the last roundup, we've seen some evidence that the anti-Coalition trend in polling has softened, but not all that much.  Newspoll has come in from 53:47 for Labor to 52:48, while Essential's last three readings for Labor have been 51, 52, 53.  I aggregated the Newspoll at 52.1 and the Essentials at 51.3, 52 and 53.1.  There was also the Ipsos discussed in the update to the previous piece.  All up I have things at 52.0 to ALP, down from 52.4 three weeks ago:

Whatever is going on in February and March is likely to shape polling much more than whatever (if anything) is causing the very small comeback that we see here.


The most recent Newspoll saw Malcolm Turnbull's Better PM lead over Bill Shorten drop to an all-time low of 9 points (41-32).  This is the same margin Kevin Rudd had against Tony Abbott in 2010 when Labor dispensed with Rudd's services.

Both leaders polled poor personal ratings with Turnbull backsliding a few points to a net -23 (32-55) and Shorten doing the same to -17 (34-51).  Essential had Turnbull on a new personal worst of -12 (down 4, 34-46) with Shorten on -3 (up six, 35-38) and Turnbull's lead as Better PM down one to 11 points (39-28) - the 39% is Turnbull's lowest rating as Better PM from Essential so far and 11 points is his equal smallest lead over Shorten.  However these Essential findings should be considered as in the context of a sample that may have been a little bit Labor-friendly.

Essential also released ministerial approval ratings, which were pretty nondescript with the exception of Foreign Minister Julie Bishop, who recorded a whopping net +29.  Foreign Minister is an easy portfolio to be popular in because the incumbent is not responsible for harsh policies and often gets opportunities to look good on the world stage.


I was unimpressed with an issues-importance Newspoll that had "jobs and growth" (36%) as the "issue" nominated as the most important priorities by voters ahead of budget repair (16), same-sex marriage (14), asylum seekers (14), national security and, from memory, something to do with energy.  Employment and economic growth are two distinct priorities, even though they have a degree of causal interrelation.  So it would be more useful to offer them separately.  Moreover, offering an election slogan as a poll alternative could lead to some form of abstract identification with the slogan that might skew the results as an indicator of what voters really want.

I am also not that impressed with Essential's questions about whether particular parties are becoming "more progressive" or "more conservative" - over what timescale?  A more interesting question was "If a new conservative party was formed and included people like Tony Abbott, how likely would you be to vote for them?" but an 11% "likely" response from self-proclaimed Greens voters smacks of one or more of satisficing to get through the survey (see previous issue), vote-parking or having no real idea what the word "likely" means.

Another Essential finding that looked a bit suss was 15% of Liberal and 12% of Labor, but only 52% of Greens and 36% of Others, saying that they usually voted for minor parties or independents.  Sure there is always some mobility, but this much from a usual voting intention to current seems unlikely, and would 13% of voters really not know who they usually voted for?  And a question on Gonski funding had a preamble that was primed to the gills.


That - unless there are any last offers - brings us to the end of another year of Australian federal polling.  This year saw 108 public national polls, down fifteen on last year (see 2015 review) in spite of 2016 being an election year.  The mid-year cessation of Morgan multi-mode and the scarcity of post-election polling were the major factors here.   Before the election we had 25 Essentials, 12 each from Morgan multi-mode, Newspoll and ReachTEL, seven Ipsoses, three Galaxys and one each from Morgan (phone) and Lonergan.  (There was also a commissioned Community Engagement poll without a published 2PP or adequate methods details - I've ignored it in this section).  Since the election we've had 25 Essentials (two of which included some pre-election data), eight Newspolls and one each from Ipsos and Morgan (phone).

As with 2015 the year falls into two polling halves, but this time the election rather than the removal of a sitting Prime Minister marks the boundary.  Before the election the Coalition won the published 2PP (last-election preferences used where available) in 30 polls, tied in 22 and lost 21.  I had their average aggregated 2PP at 51.2 in this time.  The highest published 2PP for them was 55.5 from Morgan multi-mode, followed by a 55 from ReachTEL (both at the start of a very long year) and the worst was 48 (Essential twice in April-May and Morgan multi-mode in May).

Since the election the Coalition has tied three 2PPs (all of these dubiously so as they were all based on 2013 preferences), won none and lost 32.  Its best was obviously 50-50 (two Newspolls and one Essential) and the worst was a 45 from a small-sample Morgan phone (followed by a raft of 47s).  I have the average aggregated 2PP in this time for the government at 48.4.  Overall as an average for the year I have the parties tied at 50.0-50.0. Given that the government under-performed my aggregate by 0.44 points at the election (0.17 points of this caused by preference-shifting) the Coalition's share of the lead for the year is as debatable as it is irrelevant going forward.  This average also excludes the period immediately after the election, when my aggregate was switched off.  Overall the government has been behind in more polls than it has been ahead in, but it fattened its average with some high readings very early in the year.

Malcolm Turnbull started the year with a Newspoll netsat of +22 but stayed in positive territory for just three Newspolls, then recorded seventeen minus netsats in a row, with a low (twice) of -28. His average for the year was -12.5 (-5.5 pre-election, -23 since).  As already noted, he was unique among PMs in failing to get a post-election net satisfaction bounce.

Bill Shorten stayed in negative territory for all 20 Newspolls but his average of -18.8 (-21 pre-election, -15.5 since) was actually a slight improvement on last year's.  His worst netsat, -35, came at the start of the year, and his best, -12, came in late May.  The only previous Opposition Leaders to survive a whole year with no netsat better than -12 were John Howard in 1988 (best -15) and Tony Abbott in 2012 (best -16).  Abbott led his party to victory the next year while Howard had to wait a fair bit longer.  Shorten has now been sub-zero for 22 months in a row, but that only puts him fourth in that particular hall of shame among Opposition Leaders, behind Howard (41 months), Abbott (34) and Hewson (25), and also behind Prime Ministers Keating (49), Hawke (30) and Gillard (27).

Turnbull was the "better Prime Minister" in Newspoll for the whole year, but this indicator skews to incumbents by about 16 points, so an average lead of 18.3 for the year is nothing special.  Still he's the first PM to keep the lead for a full year since Kevin Rudd in 2009 (Labor kept it all year in 2010 but with two different PMs), and these days for a PM even to keep their job for a calendar year is not to be taken for granted.


Again I note some current betting figures.  Betting at the end of last year had the Coalition heavily favoured to win (which they did) but expected hardly any net seat change at all (bzzt, wrong).  Checking three sites for current odds on the next election, I found Labor currently favourite on all, but with substantial variation (1.60 vs 2.25, 1.62 vs 2.25 and 1.85 vs 1.90).  If you assume there's no chance of anybody else winning the thing that actually comes out to a very small arbitrage, but you'd be better off, even these days, leaving any sizeable amount of money in the bank.  Anyway, that's an implied chance of between 42% and 49% for the government to win the next election.  One site's odds have Turnbull probably not even making the next election (2.00 vs 1.75), with Bishop (4.00), Dutton (6.00) and Abbott (7.50) next in line as his replacements.  I can't find anybody fielding on whether or not Bill Shorten will serve out his second term as Labor leader.

The overall picture

The year started with the Coalition still on Malcolm Turnbull's honeymoon high, but it was bound not to all last.  Misadventures on tax policy and some general disappointment with Turnbull's willingness to toe the conservative line saw polling slump towards parity from late February through to late April.  After a deadlock that lasted most of the very long election campaign, the Coalition picked up just a little in the last few weeks - which turned out to be just as well for it, since an unfavourable regional distribution of seat swings went close to costing it its majority .  Since the election it's been mostly if gently downhill, with the Coalition recording some of the worst early-term polling by a returned government ever, without yet plunging into what I think of as the serious trouble zone (46% and below.)

The road ahead

It may seem more than a little compulsive to be thinking about whether a government will be re-elected only a few months after it last was, but I think it's worth looking at what can be read predictively from a government's early polling.

The general perception is very little - that polling more than a certain distance out is just not predictive at all - but in fact governments that poll poorly in their first six months do have a track record of losing the next election almost as often as not.  The escape route found by Labor in the 1990-3 term and the Coalition in the one just passed, of replacing a Prime Minister who was either past their use-by date (Hawke) or never much chop to begin with (Abbott) seems unlikely to work again, as voters would be unlikely to tolerate it if the Coalition again couldn't find a PM and stick with them for a whole term in office.

Still there is nothing in the polling alone to say that Labor already deserves favouritism.  Those punters putting money on the ALP to salute in 2018 or 2019 are probably going on something other than polls - a subjective perception that there could be severe instability within the government, or a suspicion that there might be a recession.  Either that or they are not reading the track record of polling correctly.

Update 20 Dec: Today's Essential (47, aggregated at 46.8) did a bit more damage causing the Coalition to end the year on 47.8.  The annual figures above have been revised, the rest hasn't.


  1. Kevin, I notice your scepticism about the "issues" questions. As I suggested (under that other name I use) on pollbludger, "perhaps about 10% of people just give a random answer to any question they’re asked! [I guess you could call that satisficing, but I'd just call it laziness or wilful sabotage.] Perhaps polls should include a question or two designed to weed out those people, like “do you agree with the following – one of my ancestors came from Mars” or “do you agree with the following – deedle eedle oodle eedle?” My prediction – 2-4% will stronly agree and another 8-10% will somewhat agree."

    I suppose when it comes to 2PP and similar questions these random answers may cancel each other out, but perhaps the 10% who spew out random answers should be born in mind when calculating MOEs?

  2. When it comes to basic voting intention there is strong evidence people don't answer randomly - parties that have virtually zero support at elections actually do drop to virtually zero in polls. Example PUP at last federal election.

    When it comes to issues questions though I suspect that there are respondents who do answer pretty randomly. The best way to test this would be with a bunch of questions where answer, say, C was always ridiculous. That would weed out the random-answerers (as opposed to those who deliberately chose stupid options where available).

  3. Have you passed that suggestion on to the pollsters? (I imagine they'd take more notice of you than if it came from me. I'd also love to see them test the deedle-eedle-oodle one - I'll bet it would get some "agree"s from the flippant buggers!)

  4. No. Pollsters do their thing and I comment on it publicly; a number of them read here. Certain pollsters only contact me when I have hurt their feelings!

  5. Kevin, I agree that the polls are not usually indicative this far out. However, when was the last time we had a PM who cravenly threw out his core "beliefs" to appease the right-wing of his party. I've never seen the like. That is why the result of the next election is already baked in.

    1. Turnbull has few if any core beliefs, it is all just debating school all over again for him. Trump was unelectable even as a nominee because nobody had ever seen his like in so many ways.

    2. I presoom you mean "unelectable" in ironic quotes, Kevin? As to Malcolom, I'm prepared to believe he still has some small-l-ish beliefs somewhere deep inside, but he's just not prepared to stand up for them. I suppose the kindest thing we could assume is that he's hanging on to the (nominal) leadership to keep somebody worse from becoming leader.

    3. "I presoom you mean "unelectable" in ironic quotes, Kevin?" Correct.