Poll Roundup: Recovery, Or Just Turbulence?

2PP Aggregate: 52.4 to ALP (-1.1 since last week)
Closest reading of 2017 so far
Labor would win election "held now" with a moderate seat margin

Five weeks since the previous edition, it's time for another roundup of the state of federal polling.  After some really bad readings from Newspoll in February and Essential in March, things seem to have settled down a little for the Turnbull government.  This week the government gained a 2PP point on both Newspoll (47 to 48) and Essential (46 to 47).  I aggregated the Newspoll at 47.8% and the Essential at 47.1%.  With a bit of help from the March Ipsos and (temporarily) last week's Essential falling out of the sample, these polls have improved the government's position on my aggregate by 1.1 points in a week, to 47.6% 2PP.

I normally show just the smoothed tracking graph of rolling averages, but here's the "spiky" graph of one-week end-of-week figures, because it has a story to tell.

The one-week aggregate readings in 2017 so far have been much more turbulent than in the second half of 2016. Perhaps this week's shift is just another case of this turbulence.  (Incidentally, BludgerTrack has been running somewhat lower for Labor than my aggregate lately after disregarding Ipsos, and is bound to hence show a smaller correction based on this week's polls.)  Here's the smoothed tracking graph:

Labor started the year in very strong shape, but haven't built on it since.  However, we shouldn't take that much notice of moves back towards the Coalition until we start to see sustained close or even polling (if that happens.)

Leaderships

The commentariat had decided that this week was all about Prime Minister Turnbull's announcement of changes to citizenship tests and working visa laws, and with articles probably half-written before Newspoll even came out, they weren't going to let such minor changes as a single Newspoll point on 2PP and four points on the Prime Minister's net satisfaction rating put them off.  These changes could have easily been about voters not wanting Tony Abbott back, or the military posturings of North Korea, or a lack of obviously bad news cycle items, or good old statistical random noise.

Anyway, Malcolm Turnbull is up four points on net satisfaction to -25 (32-57), but that change is actually below the average net poll-to-poll change in PM netsats in Newspoll history, which stands at 5.5 points.  So nothing special is needed to explain the shift, since such shifts in PM ratings happen mundanely all the time.  Indeed the PM's netsat changes have now reversed direction six times in a row (up-down-up-down-up-down-up) but he still needs another two reversals to tie the all time record for such things (set by John Howard in early 2000).

The bad news for the PM is that this his his eleventh Newspoll in a row with a netsat of -20 or worse (Howard, Hawke and even Tony Abbott maxed out at a streak of only seven such, though Paul Keating had 17 in a row and Julia Gillard at one stage 29).  Not coincidentally, it's also the eleventh Newspoll in a row in which the Coalition has lost the 2PP.  As Malcolm Turnbull used the loss of thirty consecutive Newspolls as an argument for rolling Tony Abbott, this ticking timebomb is watched with interest by political tragics.  If the clock strikes thirty, the expectation is that Turnbull will be ridiculed out of office, so one can well imagine that for all his denials the PM hopes desperately for a rogue 50:50 to reset the clock.

Or maybe people just generally focus too much on Newspoll!  George Megalogenis has a long-standing theory that the shift to fortnightly Newspolls kickstarted our revolving door of leaders, and apparently this theory popped up again recently.  But I don't think it's actually true.  If we take the ten years either side of the start of fortnightly Newspolls in 1992, the ten years prior had seen the PM change twice (Fraser-Hawke-Keating) and the Opposition Leader change five times (Hayden-Hawke-Peacock-Howard-Peacock-Hewson).  The next ten years saw the PM change once (Keating-Howard) and the Opposition Leader four times (Hewson-Downer-Howard-Beazley-Crean).  Even if we stretch the comparison to the longest possible, 25 years, we get six PM changes and 11 Opposition Leader changes apiece before and after - excluding caretakers.  (The "before" did cheat by one of their number being lost at sea, but Holt may well have been rolled soon anyway.)

The Hewson-Downer-Howard rollercoaster had nothing to do with fortnightly Newspolls and everything to do with Hewson being cooked as leader by defeat in the "unloseable" 1993 election, and Downer being just a dud.  At the time it was also seen as having something to do with Andrew Peacock's effective veto on Howard resuming the leadership, which ended when Peacock retired.

Anyway, that's a pleasant historical diversion from the extremely boring task of writing about Bill Shorten's net rating of -20 (33-53.  Actually, feeble as it is, it's his best of the year), or the preferred prime minister score (42-33 to Turnbull, the same lead as before.)

Essential a few weeks ago had Turnbull up from -17 to -12 in a month, Shorten up from -19 to -13 and the better PM lead to the incumbent at 39-28.

Other polling

I'm pleased to say that I've seen less obviously dodgy polling in the last few weeks than I was seeing in the average day at the time of the previous episode.

Essential polled approval ratings of crossbench Senators, finding that Senators Xenophon, Hinch and Lambie are all reasonably well regarded, Senator Hanson polarises opinion but is more disliked than liked, and Senators Leyonhjelm and Bernardi don't have massive fan clubs.

Essential also polled a voting mobility exercise, which showed that voters who report voting for the major parties (and to a lesser degree the Greens) tend to have voted more stably than those who vote for the smaller parties.  The table takes quite a lot of unpacking - each vertical column gives a party a respondent has said they have voted for in the last ten years, and then lists the chance that they had also voted for each of the other listed parties.  The average voter who has voted Liberal-National reports voting for 1.69 of the listed parties in the past ten years at state and federal elections combined, compared to 1.83 parties for Labor, 2.34 for the Greens, 2.53 for One Nation, 2.95 for NXT, 3.15 for Family First, 2.82 for independents.  The ultimate party-hoppers were those who had voted for PUP (3.57) but that makes sense since no-one could have voted exclusively for that party over a decade.  I suspect voters are over-reporting the extent to which they have voted for One Nation, given that the party's share of the vote in state and federal elections was very poor from 2007 until the 2016 election at which they managed over 4% in the Senate.

Essential voters overall narrowly approved of the recent US bombing of Syria (41-36), but don't think we should get involved (31-50).  There are plenty more issues polls over at Essential for anyone who finds the opinions of a panel of 100,000 grumps interesting, but I thought this one on breaking the law was especially odd. 47% of Greens voters think it is never justified to break the law?  Do these people not even know who Bob Brown is?

A JWS survey taken in March found that 50% of voters wanted longer parliamentary terms (of some sort) while only 13% wanted more elections.

2016 Warringah entrails

The most interesting poll-shaped object of the last few weeks has been the supposedly "leaked" poll purporting to show that Tony Abbott had been in danger of losing Warringah to Labor or an independent until Malcolm Turnbull saved his embattled mate with a robocall.

As appealingly undignified in a classically Abbott way as this story is, the most likely explanation for all of this is simply that the poll results were wrong.  It had a sample size of only 400, and the poll was reportedly conducted by Sexton, rather than Textor, at Abbott's request.  The preferencing seems to have been screwy too, since the reported 57-43 to Labor "aided vote" sample (with Abbott named) off primaries of 34-19 would have required an 81% flow of all preferences to Labor (in reality Labor would have got about a 2PP of about 52-53%, rather than 57%, off the stated primaries.)

The robocalls were real as attested by some (but not very many) sources at the time.  But if a robocall from the boss and the other named interventions were really worth seventeen (or even seven) points in Warringah, it would be likely that the same sort of thing would have been worth quite a lot in other electorates where the Prime Minister was then still regarded well.   And while naming or not naming a candidate sometimes does make a difference, I'd be surprised to see it matter as much as ten points and for that effect to then be repeated at a ballot box.

If the story about Abbott requesting a different pollster is true then that shows a lack of judgement that Turnbull was only too happy to take advantage of.  In reality Abbott's opponents were a rabble and Abbott's vote was never likely to be anything other than, as Antony Green put it, "a whale surrounded by minnows".  If there really was such a savage primary vote slide against the former PM then a ReachTEL conducted by the Australia Institute six months earlier, while suggesting voters wanted Abbott to retire, had shown no signs of it.

Anyway, on we go to the next prism through which every slightest poll move can be misconstrued, the Budget!  Budget polls always need to be put in their historical place, so another roundup is probably not too far away.