Poll Roundup: Two Steps Forward, One Step Back

2PP Aggregate: 52.2 to ALP (+0.2 since last week, +1.2 in three weeks, highest in twelve weeks)
ALP would probably narrowly win election "held now" outright

Four weeks ago I declared the Abbott government's budget blowout "over", their aggregate position having come back to 49% 2PP, which was about where it had been back in April.  It had taken almost half a year to get back to that point, but assuming that the Coalition was now going to poll competitively for a while, it was easy to believe that they were on the track to re-election.

What's happened since may bring just a little smile to ALP supporters, since no sooner did the government finally get back to effective parity than its ratings turned around and started heading down again.  As the Abbott government approaches the first anniversary of losing the lead, this small shift back adds a little more interest to the question of when and under what circumstances they might actually get it back.



This week's polls

Four polls have been released in recent days.  The first federal Fairfax-Ipsos came out with a 51:49 result to Labor.  We don't know nearly enough about this poll's behaviour yet to have a clear idea what this means; its NZ counterpart was at times very high for National in the leadup to the NZ election, but that doesn't mean the Australian poll will be likewise.  This was followed by a 53.5-46.5 result (by last election preferences) from Morgan, equivalent to 52:48 after accounting for house effect, a 54:46 from Newspoll (the best for Labor since late July) and a 52:48 from Essential.

It's notable that the Ipsos poll had a bad state breakdown for the Coalition in Victoria (43:57); Morgan has had similar, and federal polls are fully consistent with the Abbott government being a dead weight for the struggling Napthine government less than a month out from its state election.

After adjusting all these polls taking into account the published primaries, my aggregate has Labor on 52.2, their highest level since the second week of August.  If there is no further polling by midnight Friday, they'll get another tenth of a point on weekly reset.  Here's the smoothed tracking graph:

The recent death of Gough Whitlam, plus the "bowser bandit" Herald-Sun headline about petrol tax rises, are two events that have been cited in shifting voting intention back a little bit Labor's way.  But most of this small shift back appears to have happened before these events.

Primary/2PP Differences Continue

As noted in the recent arthouse favourite Wonk Central: What Do We Do With The Poll Rounding Problem?  I've thrown some hybrid vigour into my aggregate model by using an average of the 2PP published by the pollster and the 2PP implied by the primaries.  This is designed to better deal with cases in which the published 2PP and the published primaries are at odds with each other, as will happen randomly from time to time.

But what's looking likely since I wrote that article is that these discrepancies are not just happening randomly.  Four of the last five Newspolls (that might just be a lucky run) have had a higher 2PP for Labor than the most likely reading implied by the published primaries, as have at least the last six Morgans in a row.  For Essential, however, this pro-Labor effect in the published 2PPs does not occur.

If there is something in the state mix of party support for the Greens and for Others, or in the nationwide breakdown of Others support, that is driving these discrepancies then it's possible the minor-party voter landscape really is better for Labor than at the 2013 election. (That's even before we come to the question of whether supporters for given parties might have become more pro-ALP).  In that case my methods shift might even be underestimating Labor's standing.  But I don't think there should normally be that much in it.

Something that could cause large differences however, is if the Others category contained a lot more self-professed independent voters than at the last election, bearing in mind that indie preferences went quite strongly to Labor in 2013.  However, intention to vote Independent tends to get over-reported by poll respondents (as shown by Nielsen's dabbling with the category), and those who say they'll vote Independent then don't are not likely to be as ALP-friendly as the indie denizens of Denison and Indi.

When 2PPs consistently or frequently differ from a pollster's published primaries in the same direction, does the published 2PP tell us everything (because the pollster knows the breakup better than we do) or nothing (because the pollster is making questionable assumptions)?  For now my methods sit between these options, but I'll be watching this issue for a while.

Leaderships

The only new leadership figures this week were from Newspoll and Ipsos.  Newspoll had Tony Abbott's netsat unchanged at -15, Bill Shorten's up three to -8 and Tony Abbott with a one-point better PM lead (39-38, unchanged).  The Ipsos net approval ratings seem to be continuing in the Nielsen tradition of being slightly milder than Newspoll: Abbott kicked off with -7 (42:49) and Shorten with +3 (43:40).  Ipsos had a 41:41 tie for preferred Prime Minister (a rather good result for Shorten, given the closeness of the 2PP).

Ipsos also had some within-party preferred leader polling, which as usual is heavily distorted by the tendency of Opposition supporters to want whoever the PM of the day is removed.  For the Liberals, ratings were Turnbull 35 Abbott 20 Bishop 20 Hockey 8; I believe this is the first time Bishop has tied Abbott in such a poll.  For the ALP, Shorten (30) leads Plibersek and Albanese (18 each) by more than in other recent such polls, but his lead is still hardly commanding.

There are, as usual, numerous issue questions out there but I'm too busy this week to deal with them in any detail.  Save that the Ipsos Perils of Perception survey has been hailed as an insight into the cluelessness of voters about issues such as the number of Muslims in Australia or the unemployment rate (both of which voters massively overestimate).  Perhaps what it really says is that voters are not at all good in thinking about percentages and converting their intuitions accurately into the right number.  Otherwise, it's no wonder that political debate is often so hopeless, and a wonder that it isn't even worse.