Thursday, November 14, 2013

Early Abbott Era Polling Roundup

(Note: an update has been added at the bottom following the surprising first Nielsen result and other signs of narrowing in late November.  Another has been added for the release of official 2PP preferences.)

We are now just over two months into the lifespan of the first Abbott government and it is time for a brief summary of its polling history so far, and that of its leader. 

The following polls have been published since the 2013 election: nine Essentials (one of them including a week of pre-election data), four Morgans, two Newspolls and one and a half ReachTELs.  (The half ReachTEL was a poll of NSW and Victorian voters only, that canvassed voting intention along with Shorten-vs-Albanese ALP leader polling.)

Two-party preferred

Compiling a meaningful two-party preferred picture from this polling has not been easy, for the following reasons:

* The performance of Essential and Morgan in the period between the return of Kevin Rudd and the 2013 election.  Essential tracked in a contrary manner to other pollsters (below them for Labor after Rudd returned but then above them for Labor as the election drew closer).  Morgan (using preferences modelled by the last election) initially tracked favourably for Labor but once the election was called tracked neutrally.

* A method change by Morgan to drop online surveying from their most recent survey. 

* The lack of precise preference distributions from 2013 (these are expected very soon). Pollsters can model them, but the only pollster that has attempted to do so (Morgan) has been getting results that don't make sense from the given primaries (the figures given are far too pro-Coalition). Analysts observing the results can also model the preference distributions, but the problem is that in doing so we have only the rounded primary figures made available by the pollsters (except ReachTEL which publishes to one decimal).  Therefore we are losing accuracy both with the base primary and with the estimated preference flow.

* The strongest poll for the government came from the most reliable and established of the pollsters, Newspoll, but in this particular case it may have just been an unrepresentative sample.

* The best trending data comes from Essential (and shows a gradual trend to the government) but this is coming from a pollster whose trending was untrustworthy in the two months prior to the election.

Despite this, there are some conclusions that can be drawn:

* In general, the government is ahead.  No poll that appears to imply a significant lead for Labor using 2013 preferences has been published.  Two Morgan polls might translate to a slight lead for Labor, depending on rounding, but one of those is the most recent (which involves a reduced combination of methods and therefore cannot yet be benchmarked).

* Few polls have showed the government with big leads.  Only one Newspoll (56-44 as published, but really 55-45) has shown an advantage clearly exceeding the election.

In all the government seems to have a moderate lead of around the size of the federal election result or slightly less.  My own aggregator is currently at 52.8 (and has not been above 53.1) and Bludgertrack is currently at 52.7 (but reached 53.6 last week).  At this stage, the predictive value of the government's early ratings is negligible.

On my current assumptions, this is what the tracking of my aggregate over the last six weeks looks like:

Oooh, a tracking graph! Sorry it's such a primitive one.  (Dear Santa, all I want for Christmas is a microchip that will implant immediate mastery of R into my brain.) 

In any case, that's very rough, because I may well decide to modify my treatment of some pollsters when I review their 2013 federal election performance thoroughly (something I will get onto once I have the final election results.)  The same broad trajectory has been seen in Bludgertrack but a lot of the blowout is down to the re-entry of Newspoll to the mix, and BT has a less close starting point.

What is noticeable in this polling overall is the lack of a visible honeymoon bounce period for the new government.  At a similar stage, governments newly elected from opposition were running above their election results by, on average, about four points 2PP (1950, Menzies), 2.5 points (1973, Whitlam), six points (1983, Hawke), 5.5 points (1996, Howard), and six points (2008, Rudd).  Only the Fraser government in early 1976 did not get a bounce above its election result (running at it or slightly below it after a couple of months), and given that its election result was so enormous to begin with (55.7%) that's understandable.

Generally these bounces do not last all that long.  The Rudd government ran well above its incoming election result for just over two years but that was an exception, and we all saw what happened to that.  In other cases a new government that gets a bounce has returned to its election result (or, often,  worse) after between three and fourteen months.

Personal Ratings

The good news for Tony Abbott is that he has finally returned to positive rating territory.  A period of 65 consecutive negative Newspoll netsats going back to November 2010 finally ended when Abbott returned a rating of +13 (47-34) in the late October Newspoll, followed by +7 (45-38) last week.  Even with his party on course to a decisive victory Abbott had not managed to poke his head above zero in the election campaign, finishing his Opposition Leadership on -6.  (Incidentally, Abbott's 65 straight sub-zero netsats was a new record for an Opposition Leader, but Paul Keating had 109 of them in a row as PM.)

That said, Abbott's new ratings are modest compared to other PMs elected from Opposition.  Whitlam was polling +44 (Morgan Gallup) a few months into the job, Fraser's first Morgans after the election (not counting the caretaker period before it) were +33 and +19, Hawke started with +43 and +46, Howard kicked off with +33 and +38 (Newspoll) and Rudd's opening ratings were +48 and +55.  Some of these are not quite fair comparisons because of the unusual delay in Newspoll's resumption of polling this year (caused by the Labor leadership ballot) but it makes little difference.  It took Whitlam and Fraser most of a year, Howard just over a year, Rudd over two years and Hawke around three to poll netsats as modest as Abbott is polling now.

On the other hand, Prime Ministers who take office mid-term are much more divisive.  McMahon started out at +47 but was below net zero in four months.  Keating never polled a positive netsat as PM and always had a disapproval higher than Abbott's is now (and yet today is widely lauded!) and Gillard was in barely positive territory within weeks of her appointment.  Abbott's current rating also exceeds any polled by Kevin Rudd on his return.

It's known that PM net satisfaction ratings for all PMs in the last 30 years have been very strongly correlated with party standing and that they act as leading indicators for changes of party ratings (rather than the other way around).   It will be surprising if Abbott does not follow this trend.  But it doesn't necessarily mean that lukewarm ratings for Abbott are a bad sign, since each past Prime Minister has had a different break-even point.  For Rudd's first term, his projected netsat if the 2PP was exactly level was -20.  For Keating it was -15, Hawke -6, Gillard +3, Howard +7, and Rudd's brief second stint +11.  If the break-even point for Abbott is towards the lower end of the scale then he will be able to poll quite bad personal ratings before his party has cause for concern.  Notably in state politics Coalition Premiers O'Farrell and Newman continue to poll mediocre personal ratings while their parties enjoy massive leads over Labor.

Relatively poor results have also been seen for Abbott in the relatively useless indicator that is preferred/better Prime Minister polling.  For instance Newspoll has shown leads for Abbott of 19 points (47-28) and 16 points (46-30) over Bill Shorten, despite Shorten not having even been voters' preferred choice as Labor leader.  As Canberra Times journalist Ben Westcott noted on Twitter, it took Kim Beazley a year to get a 30% PPM Newspoll score, while Turnbull and Nelson (up against a rampantly popular Rudd) never did.  At the equivalent point, Nelson was struggling to register single figures while Beazley was hovering around 20%.  Abbott's current lead (of 16 points) is the sort of lead that normally indicates a very close contest, and early in a new PM's term is when a PM is most likely to register large leads on PPM.  I think here that we are seeing a continuation into government of the high level of polarised dislike of Tony Abbott by Labor supporters that we saw when he was Opposition Leader.  While winning the election has boosted his stocks by several percent, it seems that around three-quarters of the voters who were firmly in the anyone-but-Abbott camp are still there.

Results from other pollsters who have canvassed leadership issues have been broadly similar.  Essential had Abbott at +5 in late September, +11 in late October and +5 now. Essential also had  a 19-point lead for Abbott as preferred PM in October (41-22) and 15 points this week (42-27), in both cases with higher don't-know ratings than Newspoll.   An online-only Morgan surprised with a very close 40-36 PPM lead for Abbott.

ReachTEL on 10 Oct (with no don't know option) had Abbott with a net rating of only +0.3, and a very party-polarised breakdown with almost no Coalition supporters rating his performance poor and almost no Labor or Greens supporters rating it good.  In my view the use of "satisfactory" as an intermediate option probably drags the net rating down a little, since the term "satisfactory" would probably capture mildly positive sentiment as well as views that the leader had been average.

Bill Shorten polled an opening +4 (31-27) net approval from Essential this week, having polled +8 (32-24) and +13 (37-24) from Newspoll.  In fact, Shorten's average netsat of his first two Newspolls as leader (+10.5) doesn't compare all that well with his predecessors: Rudd +37, Turnbull +24.5, Beazley I +22.5, Howard II +21, Hewson and Downer +20.5, Latham +19.5, Beazley II +15, Nelson +13, all started off better than Shorten. Only Crean (+5), Abbott (+3.5) and Peacock II (a big outlier at -23.5) had weaker starts.  Given that so many of those ahead of Shorten ended up being flops while two of the last three went on to two near-election-wins and a decisive win between them, the sort of ratings he is polling right now say very little about Shorten's future.  Labor would want him to stay above zero for a while though.

Just for the fun of it I've started a Not-A-Poll on the question of who will be the first to reach a Newspoll net negative rating!  Not-A-Poll is extremely unscientific and stackable and mainly for infotainment purposes; that said, the voting close to the election in its previous outing was accurate.

Polling Roadmap for the Coalition's Term

It's my view that it makes absolutely no difference whether the Coalition enjoys large polling leads or modest ones at this stage , and whether its leader has good ratings or more or less neutral.   The government's prospects of a long shelf life are more likely to hinge on perennial big-picture factors that are yet to emerge, but also on the answers to these two question sets:

1. Are voters' reasons for rejecting Labor at the last election dominated by the leadership mess and by policy and credibility failings that are ultimately explained by bad leadership and/or the hung parliament situation? Or are there more deep-seated reasons that cannot be fixed just by taking out the trash from the Rudd/Gillard/Rudd era, and that all other things being equal will not be forgiven quickly?

2. Assuming Labor can convincingly fix everything they got wrong in their six years in office, is it the case that voters would want Labor back and only put the Coalition in reluctantly and by default?  Or are voters not really fussed which major party is in power so long as it has its act more or less together?

A "yes" to the first parts of both 1 and 2 is important to Labor's chances of a highly competitive first term in Opposition.  My own tentative view is that it is "yes" to the first question for 1, and "yes" to the second question for 2, and that the expectation bar for the Coalition in its first term is not going to be all that high so long as voters are satisfied that their government is no longer turbulent and at least modestly competent.  But it could be I am wrong about this and that the cycle of voter dissatisfaction flagged by Laura Tingle will continue, and that Australia has become a kind of Poland of the South Seas where incumbency generally becomes toxic sooner rather than later.

Some on the left have held out hope that competitive polling early in the term gives Labor good chances of taking the fight to the Coalition and making this a one-term federal government (extremely rare as those have been in the past).  Modest as the government's early form has been, there is no real reason yet to believe that this is so.  If the government falls well behind after, say, six months or so, and stays well behind for a long time, then it will be time to talk about such things.

During the election campaign I wrote a polling roundup article weekly.  My current intention is to write something like this at irregular intervals of about two or three months until we get close enough to the next election to justify doing so more often, but to write stand-alone articles whenever a notable polling theme emerges.

Note added 17 Nov:  In the Not-A-Poll on the sidebar (in which early voters strongly expect Abbott to be the first into negative territory) I included the "neither" option for completeness, but it's already racked up a vote.

According to my records (which are slightly incomplete for the old Morgans), this is how long it took for a negative netsat to appear on one side or the other, using Morgan for pre-1986 elections and Newspoll thereafter, and which side's leader has been the first to do so.

A few explanative notes, mostly for the points marked (*) :

* The Duration figures for ratings before the 1990s aren't really comparable to the earlier ones.  In the earlier cases, polling was much less frequent.  Had it been more frequent, it's likely negative net readings for leaders would have been recorded faster.  So it's probably not really the case that 1969-1971 was a period of unparalleled leadership approval.

* In the case of Gorton as PM post-1969 and Peacock as LO post-1984, the leader in question was removed by their party without recording any negative netsats.

* In 1990, Peacock polled a single negative Newspoll netsat after the election.  His replacement, Hewson, then polled positive netsats for a year and a half.

* After the 1998 election, Howard's initial negative rating was an isolated -1 after seven months.  He didn't move into consistently negative territory until 16 months after the election.

* After both the 1996 and 1998 elections, Beazley first moved into consistently negative territory just a couple of polls after Howard did.

So, while there are a lot more LOs than PMs in the "culprit" column, it's also noticeable that those PMs who have appeared in that column were all Coalition PMs.  In 11 cases the Coalition leader was first to become unpopular, compared with only 3 (or 4, discounting Peacock in 1990) for Labor. 


Update: Nov 24 - Nielsen Bombshell!  A big surprise with the first Nielsen of the new era delivering a 52:48 lead to Labor based on respondent-allocated preferences.  This is especially amusing given that Nielsen, along with Galaxy, never even once showed Labor ahead during the entire 2010-13 term.  (Essential and Newspoll stopped doing so in early 2011.)  The poll also shows a close PPM situation (49-41) and Shorten with by far the better net approval rating (+21 to +1 for Abbott - and Nielsen ratings are frequently mild, so a small positive from Nielsen would probably be a negative from Newspoll.)

Since Nielsen were using respondent-allocated preferences I've calculated estimates based on 2013 primaries and approximate preferences.  By one method I get 51.2, by another 51.4, so I've aggregated it as 51.3 (for now).

All these results could be explained by an unrepresentative sample, and any poll that is four points off the previous aggregate sets off the rogue siren while we wait to see if anyone else replicates this result.  Perhaps within a week or two this poll will just be being laughed at by all but certain sections of the ALP faithful and the grassy-knoll anti-Newspoll crew.  Nonetheless, it is very unusual for governments newly elected from Opposition to trail in even a single poll conducted by conventional telephone polling, since normally the honeymoon effect pushes them to such a lead that even a rogue isn't likely to do it (and also, there is less polling detail following most prior changes of government).  I'm not aware of prior examples of a federal government newly elected from Opposition polling a single result this bad anything like this soon after its election.  (Re-elected governments, on the other hand, have done this several times, including after the 1961, 1974, 1990, 1998, 2001 and 2010 elections and probably others.)

For the moment, this poll has had a seismic 1.8-point impact on my 2PP aggregate, which moves to its closest position since the election at 50.6 to Coalition.  However this will be quickly watered down if other polls out in the next week do not confirm its findings, and the expected back-inclusion of Morgan at an adjusted value from next week might be worth half a point or so in the Coalition's favour. 

The state breakdowns of the Nielsen suggest the Queensland sample is dodgy.

Nov 25-6: Polls released today (basic results for Essential and last week's ReachTEL) have not exactly replicated the Nielsen result, but they've not been that crash-hot for the Coalition either (bearing in mind that Essential are still using 2010 preferences).  So it seems we now have a genuine narrowing (to 51.1 on my aggregate).  Newspoll confirmed this further with a 52-48 by 2010 preferences, which is probably a 51-49 by 2013 preferences (though rounding might mean it is, say, 51.7). 

All up the Coalition has shed about 1.4 points on my aggregate in a week!  To put this in context, it comfortably exceeds the loss in the ALP's worst weeks in the 2013 campaign.

The ReachTEL figures are finally up, and I make the 2PP for that poll 52:48 to Coalition on 2013 preferences rather than 51:49 as published.  And ReachTEL has Abbott at a net rating (good/very good minus poor/very poor) of -9.8 and Shorten at a net -8.3, but as noted above I believe the use of "Satisfactory" as the middle option means that these net ratings are very harsh compared with those of other pollsters.  This is especially so in the case of Shorten whose ratings are much less polarised than Abbott's.  Indeed it is tempting to count a "Satisfactory" rating as a positive, giving Abbott +7.1 and Shorten +33.2, except that that's probably too generous since the Satisfactory rating would certainly capture voters who had no view at all and were forced to make a selection.  Counting "satisfactory" in a forced poll as half positive and half neutral gives Abbott -1 (compared to +10 in October) and Shorten +12, and I think this might be the most accurate way to compare these ratings to those of other pollsters.

Here's my revised 2PP tracking graph for the last eight weeks - noting that this is only approximate and there will be some revisions made to (i) the treatment of Morgan (ii) the treatment of all polls based on exact preference distributions (whenever we finally get them!):

It might look like Labor will take the lead very soon but I don't expect that to happen.   

28 Nov Poll Recalibration

The release of final preference data for the 2013 election permits me to recalibrate my assessment of the polls so far based on the preferences recorded at the last election.  There will be some errors because the pollsters (with the noble exception of ReachTEL) shave decimals off their primary figures, and hence some detail is lost in the rounding.  However, pollsters who have been using 2010 election preferences have been getting some clearly incorrect results.  The following are my estimated 2PPs for every national (or in the case of one ReachTEL, NSW and Victorian federal) poll published since the last election that I'm aware of:

By way of explanations, the first Essential partly included data from before the election.  The last two Morgans are flagged as "MM-" because online surveying has been dropped.  Some polls segregate PUP from the remaining Others, in which case these numbers appear in the relevant columns; others combine them simply as Others, in which case the column "OT+PUP" is used.  Nielsen's strange decision to include "Independent" as an option is ignored; these are simply lumped with OT+PUP.  In the Pref Method column the preference method used by the pollster is given: LE=last (2010) election, TE=this election, RA=respondent allocated.  The main pattern here is that the 2PPs of those using 2010 election preferences (Newspoll and Essential) have been consistently overestimated, and this is not surprising since the 2010 and 2013 preferences were quite different.

The following allocations are used for 2013 preferencing: Greens to Coalition 17%, PUP to Coalition 53.7%, Other (excluding PUP) 53%, Other (including PUP) 53.3%.  But I do something a lot of people (including pollsters!) probably won't be doing, and that is to deduct 0.14% for the impact of three-cornered contests on the Coalition's 2PP at the 2013 election.

The stuck-in-the-mud behaviour of Essential, which has returned six consecutive 53s in its official results, is again starting to attract some attention.  However, if the real 2013 preferences had been used, the run of Essential readings would have been 51-50-51-51-51-52-52-51-52-52-52.  (Still, of course, unnaturally steady for a poll with a sample of 2000, even with its use of overlapping weeks.)  There's not much to see yet in terms of clear evidence of any poll being way out of whack with the aggregate overall, and the revised tracking looks like this:

It looks less smooth than the previous one, but that's an accident.

This covers midweek readings (taken on Wednesday) for the last eight weeks only and the revised current reading (51.0 to Coalition) is the equal closest in that time; the least close was 52.8 in the week of the release of the first Newspoll.  We can now say that every poll but one has shown the Coalition below its election result, but only one poll has shown it clearly behind.

2 Dec:  Morgan has published new 2PP figures said to be derived from 2013 election results here.  However, they're still wrong: the Sep 21-2 multi-mode is published as 50:50 but there is no way to get the ALP above 49 from those primaries given Morgan's rounding to half a point.  We will have to watch to see if Morgan is using the right methods now.


  1. On your questions 1a, 1b, 2a, 2b, I'd have to vote Yes to all four. I don't see 1a and 1b as either or questions - the reason Labor got tossed out was evidently "dominated by the leadership mess and by policy and credibility failings", but I would suggest that these circumstances occurred because of deeper seated issues. These deeper issues are not being addressed. Neither, for that matter, are the policy failings. At least the leadership issue is settled for now.

    We may well find out the answer as to whether leadership was the difference at the next election, as both parties are proceeding without significant change.

    I hold a fairly simple view that Australia tends to prefer a Labor Government when there is excess, and a Coalition government when in deficit. We like to do all the 'nice' things that Labor offers, but also realise that at some point it has to be paid for. I contend that if Labor could balance their budget, they would rarely lose office. However they inherently cannot, and consequently we end up with Coalition governments until the damage is sufficiently repaired. Thus question 2a is a yes, and question 2b is also a yes. Incompetence matters.

    A real danger for Labor is that they will get stuck with the incompetence label.

    1. I have a similar sort of view on the economic cycle: that voters will often put the ALP in when they can sense trouble coming, because Labor is less harsh on those who are down on their luck, and more likely to spend to stimulate. Once the economy looks fairly robust they become more likely to put the Coalition back in again to be thrifty and stash more hay. It makes sense especially in the context of Keating getting re-elected in 1993 then losing in 1996 when the economy was nowhere near as bad. If "it's the economy, stupid" applies in an Australian context, it's more about the future outlook than the current condition. John Howard misunderstood this when he claimed the economic preconditions for a transfer of office did not exist in the leadup to 2007, and some Labor supporters who argued a similar thing probably made the same mistake this time around. That or the whole leadership mess overshadowed any balanced analysis of whether Labor had done well economically or not.

      On the "deep seated issues" I have a view that both parties have deep weaknesses that they never really address because it's easier to just wait until the voters are more focused on throwing the incumbents out. The prevalence of union/faction hackery has been one of Labor's for a very long time.

    2. "it's more about the future outlook than the current condition"

      This is quite true. In that context, regarding the election just past it could be said that Labor contributed substantially to the negative future outlook that unseated them.

  2. Kevin, I am of the opinion that if polls are conducted on the electioneering promises made by the then in opposition party, (in this instance it was the Federal Liberals) then we count up the number of disparaging and overblown commentaries against the Labor ministers and Labor party in general that appears is forever issuing from within the Abbott team, (and often highlighted by the major media outlets) so much so that it suggests that there is nil fabric or substance contained in the Liberals quest that can give credibility to the various polls so conducted.
    EG: the serendipitous fancy that is actually stated as what 'we the Federal Liberal Party will do for you or alter for the peoples benefit' soon becomes known as a string of lies.

    Having said that in the above,'yes this would account for the rapid decline of Abbotts standing in the polls' therefore I conclude that often the poll veracities are mostly based upon the surreptitious rumblings 'only' within the minds of those person called upon to express their pre-election voter viewpoint.
    Now I am not arguing against yourself and your regularly accurate predictions, my question is more to the point of; "how is it that such a quantum of false promises and lies can influence pre-election polls."


    William Boeder.