Wednesday, July 23, 2014

Poll Roundup: A Brief History Of Disaster Bounces

2PP Aggregate: 53.1 to ALP (-0.6 since last week)

"I actually think Tony Abbott is doing us proud.  I didn't think he would, I didn't think he had it in him, but he's put his foot down and now other countries are supporting him".

The above was one of a number of vox-pop style voices in a Tasmanian ABC radio segment about reactions to the MH-17 air disaster.  Interviewees, all or nearly all of them female, spoke about the way the disaster is being seen in the community and the issues involved in talking about it to children.

MH-17 coverage is inescapable in media of just about any kind, and some of the more lurid excesses of disaster-porn (Daily Telegraph and Herald Sun, that would be you) have just about had me reaching for the JG Ballard collection in disgust.  But what effect does this sort of thing have on polling and perceptions of the government?

When I put to various left-leaning online political circles the idea that this might see a rather large improvement in the Abbott government's fortunes, the consensus reaction was pretty sceptical: no bounce, or if any bounce lost in the noise.  After three polls out this week, we're seeing what looks like some movement back to the Coalition, but it's modest so far and far from conclusively real.  Untangling what is going on is made more difficult by the government having succeeded in repealing the carbon tax in the same polling cycle.

This week's results

What is expected to be the last ever federal Age Nielsen came in at 54:46, and included one day of data from before the MH-17 crash, and two days immediately after.  The poll showed a substantial lift in Prime Minister Abbott's stocks (from -25 a month ago to -18 now, a much greater improvement than in Newspoll), but that improvement was already present in samples from the first day of polling.  Obviously, there was nothing to see on the 2PP front.

The first poll to show a shift was Essential, which came down from 53:47 to Labor to 52:48.  Furthermore, the 2PP for the most recent week of sampling (which was mostly post-disaster) has been reported as being 51:49 to Labor, off primaries of 40 for the Coalition to 36.

While this was reported by Crikey  as evidence of "substantial recovery",at this stage that's still tentative.  One reason for this is that while a two-point move between one full Essential poll and the next usually means something has actually happened (sometime in the last two months anyway), a two-point move between two halves of an Essential poll (as in this case) is nowhere near statistically significant.  It makes lots of sense that the change would be caused by the MH-17 bounce, but we are not yet able to test this.

The other reason is that it's Essential, a poll that often seems to march to a different beat to the rest.  That said, Essential results since the last election do correlate with those of other pollsters.  While there is scarcely any correlation between poll-to-poll moves in Essential and those of other pollsters, this is also true for other pollsters as well.

A new ReachTEL has also shown Labor's lead down to 52:48.  Based on the published primaries, this looks to have been precariously close to rounding to 51 rather than 52.  Indeed the average reading for the published primaries by my reckoning would be 51.4 to Labor; I've aggregated it as 51.5.  The first poll to canvass approval of Abbott's handling of the MH-17 issue finds strong approval (51% positive, 23% negative and 26% "satisfactory" - a term which can denote weak positive responses), equivalent to about a +41 net response as measured by other pollsters.  That's being reported as "the Prime Minister's popularity soaring" but the full poll results show Abbott still in the doghouse overall, a 2.5 point swing lifting him only to ratings that I convert as equivalent to a -16.7 netsat from other pollsters. (By the way, something that should concern Labor marketers just a tad is that while the "meh" factor for Bill Shorten means that his ratings overall are better than Abbott's, it also means that the number of voters considering him to be a very good leader is steadfastly in single figures.)

All these results have my 2PP aggregate down from 53.7 last week to 53.1 this week, but it started this week at 53.9 after the weekend reset.  We will probably have a clearer view of where this is all going next week; in the meantime here's the "spiky" aggregate graph, the smoothed one having not yet woken up:

(The axes are 2PP vs weeks since aggregate started, the aggregate having started in October 2013.)

Past Disaster Bounces

John Howard received three significant bounces in his personal ratings in vaguely similar situations in the past.  These were:

* Following the 2001 World Trade Centre attacks, Howard's netsat rose by 18 points from one Newspoll to the next.  The Coalition's 2PP rose by 6 points from poll to poll, and around 5.5 points based on rolling averages.  The bounce occurred in the lead-up to an election, which probably shortened its duration; it was gone in seven or eight weeks.

* Following the 2002 Bali bombings, Howard's netsat rose by 18 points again.  The Coalition's 2PP rose by 4 points from poll to poll, and around 2.7 points based on rolling averages.  Howard's rating stayed at elevated levels for four months while Labor did not get back to quite where it had been for a very long time, though how much of that was caused by Bali is debatable.

* Following the 1996 Port Arthur massacre, Howard's netsat rose by 24 points, the largest poll to poll increase for a PM in Newspoll history.  The Coalition's 2PP was not affected, but since it was at 59 already, one would not really have expected it to be.  Howard's bounce lasted only a month, but it was probably shortened by the fact that he was also coming off honeymoon-level low disapproval ratings at the same time.

The second-highest poll-to-poll increase for a PM, +20, was also for Howard, in the wake of the MV Tampa incident in 2001.

We don't have so much data for major-scale natural disasters. Following the 2009 bushfires, Kevin Rudd's netsat rose a modest five points for one poll only, but that rise wasn't even necessarily connected to the event.  During the 2011 Queensland floods, Newspoll was on holidays, so any bounce to Julia Gillard was gone when they returned to the field a few weeks later.  The event did however deliver a large but temporary bounce to the state government, which did not stop it being swept out to sea a year later.

The "that's different" line for those sceptical that MH17 will deliver much of a bounce for the Coalition comes from this being apparently not a deliberate attack on Australia or the West generally (as were 9/11 and Bali) but rather it being apparently a combination of disregard for possible consequences by both those who shot the plane down and those who allowed it to be there.  But while this doesn't seem to have been intended as an act of terrorism, there's a sense in which it feels like one.

To suggest that Abbott is handling MH-17 especially well is to suggest that any of his predecessors would have been any worse, and that I think is rather unlikely.  No-one gets to be Prime Minister without having the skills to comfort the nation in these sorts of times.  All the same, events like this often do improve the public image of incumbents under siege on the domestic front.  In this light, ALP strategists must have been cursing when they heard Russia kick a goal for Abbott by bothering to denounce his public statements on the matter, which said little more than what probably close to every Australian believes.

We now await Newspoll and Morgan next week.

In Other News

The Nielsen poll also included a head-to-head Abbott-vs-Shorten attribute survey, and these were actually rather good news for the Prime Minister.  Although voters saw him as closed-minded, untrustworthy and lacking a "firm grasp of social policy" (translation: much too socially conservative) compared to Shorten, they also scored Abbott as the stronger leader, with the clearer vision, slightly better on foreign policy and more able to make things happen.  While an advantage to an incumbent on attribute polling is to be expected, this is much better than some attribute polling that has shown Abbott losing out on pretty much everything.

A state ReachTEL showed the LNP continuing to cling to a narrow lead in Queensland (with Campbell Newman trailing in Ashgrove as per normal), while in Victoria the last state Nielsen showed a sudden resurgence by the previously well-behind Napthine government, to a very competitive 49:51 deficit.  Frankly this one needs confirmation from other pollsters before we get too excited about it, but even assuming it is completely real, the line that this is about attacks on Daniel Andrews' CFMEU links  doesn't really work given that Andrews' own ratings have barely changed.  I'd take the result as most likely a combination of the temporary absence of Geoff Shaw, perhaps some impact of MH-17 even at state level (Victoria had a high toll), and sample noise.  If the result isn't rogue then Denis Napthine's popularity (at +20) is a worry for Labor, and a portent of what is likely to be a very leadership-focused campaign by the incumbents.

ReachTEL Update

Full ReachTEL findings have been released and the most interesting addition is an economic management question that sees the Coalition with a very narrow 43-42 lead over Labor to 5.3% for the Greens and 9.8% for Palmer United.  Compared with the primary votes, this mainly shows that minor party voters (especially Greens, and this is nothing new) are less keen on their own parties in this area than those who vote for major parties, but that Labor voters unimpressed with their party's economic policies are more likely to opt for PUP instead of the Liberals.

It was not surprising that the Nielsen poll delivered an unflattering one-point lead for Joe Hockey as preferred Treasurer, given that it was a poll with a bad 2PP of only 46 to the Coalition.  Since the ReachTEL has the Coalition in a much stronger 2PP position than the Nielsen, it might be expected it would show a bigger advantage for the Coalition in a traditional area of strength.  This could be down to ReachTEL's tendency to reduce the advantage for incumbents in such questions by disallowing the "don't know" option, and could also be an indicator that voting intention in the ReachTEL poll is affected by the MH-17 bounce.  I hope to find time to look much more closely at past economic management questions in ReachTEL polls and see exactly what is going on here in the next few days.


  1. G'day Kevin,
    What is the margin of error on your aggregate and what sort of change over what time period would you consider significant?

  2. I'd estimate the MOE of the aggregate at a typical point of the polling cycle - considered just in isolation and ignoring the surrounding readings - to be about 1.3 points. It is higher during quiet periods and lower when there's lots of polling. It's actually a question I can't answer precisely even at a given time, because while pollsters supply a sample size, they don't detail the amount of scaling in their samples, and the more scaling of a sample, the more variable the results are for a given sample size. Also there is the question of whether to take into account the observed "underdispersed" nature of certain polls in calculating the aggregate's MOE - if this is done then it becomes smaller.

    As a general rule, I'd be confident that a gap of a point or more between any two readings (whatever the time gap between them) is real *unless* one of those points represents a one-week spike, or was taken during a quiet period. On the spiky aggregate graph in this article there are such one-week irregularities in weeks 19 and 35 - these were caused by individual polls that were both bordering on rogue status. While the difference between two independent readings could be approaching 2 points without being "statistically significant", if such a difference occurred it would most likely be because the two readings were at opposite extremes of the non-rogue error spectrum, meaning that they would be highly likely to be corrected in the next reading.