Poll Roundup: Onion Edition

2PP Aggregate: 52.5 to ALP (-0.4 in one week, -0.6 in two weeks)
ALP would win an election held "right now" with small to moderate majority

==================================
'"Of course," Cantling said. "The onion had a dual function.  On one level, you did it just to prove how tough you were.  It was something none of the others hanging out at Ricci's could manage.  It gave you a certain status.  But on a deeper level, when you bit into an onion you were making a symbolic statement about your appetite for life, your hunger for it all, the bitter and the sharp parts as well as the sweet."'

For the source of this week's mystery quote, read on ...

Six weeks ago, the Coalition was in all sorts of trouble.  The bizarre decision to knight Prince Philip, the unthinkable loss of the Queensland state election, a leadership spill motion attracting almost 40% support, a regular barrage of leaks and appalling 2PP polling all combined to make it look like Tony Abbott would not be long for the top job.

As last fortnight's edition noted, the polling blowout caused by these events was surprisingly brief, and within a few weeks polling had reverted to modest, 53:47-ish ALP leads. There was some thought that the blowout, rather than the reversion, was the new normal, based on the idea voters were factoring in Malcolm Turnbull becoming PM, but that idea has clearly been refuted.



This week's polls

This week has seen three polls so far, with wildly varying results.

Morgan recorded a 2PP of 56:44 to Labor (a 2.5 point increase) by respondent-allocated preferences, its usual headline method.  However, on last-election preferences, the move was only half a point in Labor's favour (53.5 to 54.)  The odd thing was that after Morgan's 2PPs for Labor had seemed strong for the released primaries 11 times in the last 14 polls, in this case the last-election 2PP for Labor was lower than normal for that set of primary votes, which would normally have come out at about 54.7. On that basis, it's likely the real Morgan poll-to-poll move in Labor's favour was actually about a point.

Newspoll came out with a rather startling result of 51:49 to Labor, a four-point move in the Coalition's favour.  (Yes, we had also seen this score from Ipsos, but Ipsos is a newish poll that so far leans slightly to the Coalition.)  Rounding probably exaggerated the movement, as last fortnight's 55:45 outlier would likely have had the 2PP for Labor rounded up, while in this case it was likely rounded down.

Essential then produced a rather surprising two point shift from 52:48 to Labor to 54:46 to Labor. Essential usually shows very little poll to poll movement, partly because of its use of fortnightly averages, and this was its first two-point weekly shift since last June and just its fourth for this parliamentary term.  Again, rounding has probably exaggerated the size of the shift, but this put a damper on an otherwise large gain for the Coalition in my aggregate.  The result with all three polls included was a 0.4 point move to the Government which climbs to 47.5, its least worst reading since early November.  Here's the smoothed tracking graph:

Newspoll Volatility

The difference between Morgan and Newspoll, and the recent volatility of Newspoll's results, have both created plenty of reaction.  But the first is a non-issue - of the five point gap between Morgan's 56 and Newspoll's 51, two points are explained by Morgan's use of respondent preferences and 1.5 points by Morgan's skew to Labor, leaving only 1.5 points to explain.  My aggregate currently puts the remaining 1.5 points down to Newspoll being bouncy, but we'll see how that pans out with further evidence.

Morgan has responded by gloating about their performance at recent elections, but this is misleading since, for starters, none of their final pre-election polls included the face-to-face component that is the cause of the skew to ALP in their current multi-mode series.

The government has now had two four-point Newspoll rises in the last three fortnights, punctuated by a two-point fall.  It is actually very rare for a government to get two such surges in such a short time, with these being the only precedents I could find:

* In September 2012 during a brief recovery from otherwise terrible polling Labor under Julia Gillard recorded a run of 45-50-46-50.

* In April 1995, as John Howard was coming off a honeymoon bounce as new Opposition Leader, Labor under Paul Keating recorded a run estimated off the primaries 42-46-50.  (This was followed by reversion to 47s and 48s suggesting that the 50 was a case of bouncing.)

* In August-September 1994 as voters realised Alexander Downer was a dud and amid Coalition infighting there was an extremely volatile run of Labor government 2PPs, estimated at 46-51-48-53-57-49-51.  This is the only time a government has obtained three four-point lifts in four successive Newspolls, but the 57 was clearly over the top.

* In Feb-Mar 1992 as Labor recovered from the disposal of Bob Hawke there was a run estimated off primaries at 43-48-47-51.

(Estimates c/- Peter Brent's old conversion tables.)

Over the past six weeks the Coalition has risen six points in Newspoll.  Government rallies of this magnitude and speed (or better) have now happened 16 times in nearly 30 years of Newspoll, but they are much less common recently than in the Keating and early Howard days.  Indeed there were only two since the 2007 election: a rogue poll in early 2008 and the return of Kevin Rudd in 2013.  The important thing is that usually such quick bounces are exaggerated by statistical error or short-term factors, and in 13 of 15 such cases the Opposition gained on the next poll.  The average gain for an Opposition in the poll immediately following such a large government rise is a massive 2.6 points.

In terms of measuring short-term Newspoll volatility, a possible measure is a sum of the poll-to-poll 2PP changes over a number of polls.  So for instance over the past five Newspolls, and hence four poll-to-poll differences the total of movements has been 13 points.  The long-term average is 7.6 points (representing a typical move of just under 2 points per poll, mostly caused by purely random changes).  For all the constant talk about Newspoll being a very bouncy poll, over the last nine months it had actually been more stable than normal.

The last run this volatile was in early 2013 leading up to the failed Labor spill attempt initiated by Simon Crean.  The recent run is obviously more volatile than normal, but not outlandishly so - I have found 20 past Newspoll runs of such volatility, several of them extending much longer than five weeks and some up to 50% more volatile than this one.  It's to be expected that when there are actual sharp changes in voting intention then random sample noise on top of that may lead to a poll looking volatile.  Evidence from other polls is overwhelming that the late Jan - early Feb blowout was real so in my view there is nothing unusual to see here.

I don't think it's clearcut that the move in Newspoll is just random bouncing.  Given that Newspoll has shown a strong result for the Coalition while the fortnightly polls have not, it is possible Newspoll is picking up on something that the others have missed.  The obvious candidate is contamination from the New South Wales state election - the generally strong polling performance of Premier Baird may be rubbing off on federal voting intentions.  That would be quite consistent with the Coalition rising while Prime Minister Abbott continues to poll poorly.

The other thing to bear in mind with the current polling, and the Coalition's improvement in recent weeks (even if the Newspoll has exaggerated it) is that just maybe the government is starting to get it.  It, in this case, being that voters do not want to be continually frightened with harsh savings measures about which they have not been consulted, and are actually nowhere near as fussed about the pace of a return to Budget surplus as the Coalition thought.  Hence the advance noises about a harmless and boring Budget that will be seen by experts as irresponsible but that may reassure voters afraid of more harsh medicine and nasty surprises.

Leaderships

As noted above the PM remains in Newspoll netsat hell, though he's slowly climbing out of it with another three points gained this week.  That puts him "up" to -32.  I commented two polls ago on the disconnect between Abbott's ratings and the Coalition's and that's again a theme this week:

The graph (click for larger version) shows the relationship between recorded PM netsats and Opposition 2PPs.  Abbott's last four ratings are shown in different colours to the rest - yellow three polls ago, red two polls ago, light green the last poll and orange for the current poll.  As with two polls ago Abbott appears in a sparsely populated region of the graph, well off the usual line of relationship between PM netsats and 2PPs.  The few cases surrounding him, with a relatively close 2PP but very bad PM netsats, were all recorded by Paul Keating during 1992 and 1993.

Two stray data points is nothing to panic over, but if we keep seeing this sort of thing that will be a major development and very bad news for Labor indeed.  The idea that Australian politics is presidential and that if the PM is disliked, so is the party, has been locked in for all PMs since Keating.  Keating was an exception partly because of his personality but also because he had the good fortune to be up against very radical Opposition policies that became more of the story than he was.  Oppositions know not to do that sort of thing anymore.  All that, plus the history of past spills, was my basis for thinking Abbott would probably be gone within months after the Queensland disaster, but this sort of thing makes me less confident.

Voters on this site are less sure too; the percentage thinking he would win the 2016 election in my sidebar Not-A-Poll was zero after 111 votes, 5% after 178 votes and 11% after 261 votes, meaning that 23% of voters in the last three weeks think he will win.  (That said, I'd bet this site attracts more left-wing readers when the Coalition is polling awfully and more right-wing ones when its stocks are rising.)

Meanwhile Bill Shorten polled a pretty ordinary -11 (the fourth time he has polled this along with one -14) and had his Better PM lead cut from 11 points to five (41:36).  However, that's in fact one of the biggest ever leads for an Opposition Leader while the party had a 2PP of 51 or lower.

Also in the doldrums was Treasurer Joe Hockey who recorded a -24 netsat in Essential (27-51), down from -9 (35-44) in August.  Hockey's lead over Chris Bowen as most trusted to handle the economy was down from 11 points (34-23) to a mere point (26-25) but the massive "meh" factor on this question makes it hard to read much into it.

The Onion! (rant alert ...)

Ah yes.  A curiously left-field episode in the past fortnight involved Tony Abbott plucking an onion off a conveyer belt and taking a bite out of it, raw, peel and all.   Abbott has received widespread condemnation on social media for this action, mostly from people who disliked him already, and has even been declared a global embarrassment because of it.  Some seriously thought or hoped the onion incident might be the end of him.

Some readers may know Abbott is far from my favourite PM of all time. I hold his long-term services to political religiosity, homophobia, sexism and punitive welfare policy against him.  It was recently revealed he was even to the right of Kevin Andrews on the last of these, a position I was sure did not exist.   But hey, when a PM engages in an unscripted moment of harmless eccentricity, can we please give him a break, instead of drifting into schoolyard-level conformity just because we think it might help bury him, and wondering why our pollies are so over-scripted?

I'd wager that most of those bagging Abbott for eating a raw onion have actually never even tried it themselves.  Further, that most of them would lack the guts and that many of them would think it is a much bigger deal to do such a thing than it actually is.

The quote at the top of this article comes from George RR Martin's Portraits of His Children (1985).  In said story, the protagonist, an author, is visited by characters from his novels.  One of these is a street kid who continually eats raw onions, despite not even liking them, to show off how tough he is.

Now, I don't usually feel I have to replicate the experiences of characters in books I read to understand them, which with practically all of the stuff I read is just as well.  But in this case, some years after first reading that story, I became curious about what eating a raw onion was actually like, and therefore tried it.

It was quite an anticlimax, and when I saw all those social media comments suggesting that Abbott's onion-eating must have required great reserves of courage, creepiness or insanity, I saw a lot of people who really just had no idea.  I also saw a lot of hypocrites, who disliked our PM for his history of politically picking on people harmlessly different to himself, but were quite happy to return the same to him.  Sorry online onion objectors, but you're no better than the people who complained re Gillard's earlobes.

Anyway, that ends my quirky little rant for this instalment.  Expect a more normal and possibly shorter release sometime after Easter.  And there will be live coverage of the NSW election this Saturday night.

More Reading: 

Some readers may want to check out the JWS True Issues survey (large PDF download) for the last four months which contains a vast wealth of issues polling data.  There's a summary at AFR.

Essential Best PM Of Last 40 Years:

A new Essential result has John Howard again with a large plurality as best PM of the last 40 years.  The results were Howard 34 Hawke 16 Whitlam 13 Rudd and Keating 6 Gillard 4 Fraser 3 Abbott 2.  The poll was taken while Malcolm Fraser was still alive; there is little doubt he would poll higher if it were taken right now.  (Indeed Whitlam gained five points since the last such poll last March).

A hazard of these polls is silly reporting that ignores the effects of one party having more PMs to choose from than the other (the Coalition vote is heavily concentrated in Howard, but there are actually more votes for Labor PMs, so Howard might be in danger from Hawke on preferences if they used them).  With a don't-know rate of about 16%, it's possible there would be more Labor supporters unable to choose between two Labor PMs than Coalition supporters having the same issue.

Another problem is that these polls can be deemed to have shown someone (in this case Abbott) to be the worst PM because the least voters vote him the best.  While I have no doubt Abbott would "win" a worst PM poll taken now, that is not the question being asked.

I am thinking about ways to conduct a reader poll of this question but if so it won't happen until much later this year to ensure the poll is fair.