Abbott Fastest Ever To Lose Poll Lead

Note: updates after Monday 16 Dec may or may not be slow for a while - other commitments.

 Advance Summary:

1. Following the recent Newspoll, the new Abbott Government has lost the two-party preferred polling lead.

2. This does not necessarily mean the government would lose an election if one was held now.

3. The Abbott Government has lost the 2PP polling lead much faster than any other new government elected from Opposition in federal polling history.

4. Tony Abbott has also recorded negative personal ratings much faster than any new PM elected from Opposition in federal polling history.

5. While polling taken at this stage has very little if any predictive value, governments that have lost the lead very early in their terms have a historically greater risk of defeat at the next election.

6. Bill Shorten's polling as Opposition Leader appears good, but is nothing unusual by the standards of other Opposition Leaders at the same stages of their careers.

7. Furthermore the strength or otherwise of an Opposition Leader's personal polling after only two months in the job has no relationship with their success at later elections.
======================================================================



This week's Newspoll was one of those polls.

Most opinion polls this early in a new government's tenure attract little interest.  Findings that look even vaguely intriguing are dismissed with the response that it is too long to the next election for the poll to mean a thing.  Some even wonder why the pollsters are already polling, and some of those especially sulky about their party's fate at the last election think they shouldn't be allowed to, as if the frequency of polls is somehow the cause of their party's defeat.  A bit like blaming a thermometer for the weather being cold, but some people are like that.

But suddenly a single Newspoll has hugely excited the online left, who desperately hope that Tony Abbott will be a one-term Prime Minister or, better still, be rolled by Malcolm Turnbull (or less preferably Joe Hockey) before his term ends.  I'll wager many of those who are metaphorically rubbing Liberal faces in the PM's current polling fate are the same ones who were calling Newspoll an instrument of "Mordor" (=Murdoch) when it released a 56-44 to Coalition just six weeks ago.

So is there really something going on, or is this just another poll and are people (including some of the 1200+ who Facebook-liked the AFR article on this poll) getting excited over nothing?  In fact, a lot has changed in the four weeks since the initial version of my Early Abbott Era Polling Roundup showed the Abbott government with a lukewarm if solid polling lead.

The Newspoll

This week's Newspoll, 52-48 to Labor, is noteworthy in a number of ways.

Firstly it is actually the worst single poll for the Abbott government since the election.  Yes there was a Nielsen that was also 52-48 in Labor's favour, but that was based on respondent-allocated preferences, and it was really more like 51.4.  And yes there was a Morgan 51.5-48.5 last week as well, but that too was based on respondent-allocated preferences, and from a new Morgan methods combination (SMS/face-to-face) that thus far seems to be skewing heavily to Labor.

Secondly,  while the Nielsen seemed like a bit of a blip at the time, a second such reading from a major pollster gives a lot more credit to the idea that the government really does have a bit of a problem on its hands.

The last time Labor led in a Newspoll at all was a possibly rogue poll in March 2011; the last time they led by this much was November 2010, and the last time they led by more was an unrepresentative poll taken just after Julia Gillard became PM. 

The government has now declined in three consecutive Newspolls.  This might not seem so unusual,  but it is actually quite a rare event, since a movement in one direction is most commonly followed by a movement in the other.   This pattern last appeared in late 2009-early 2010, immediately after Tony Abbott replaced Malcolm Turnbull as leader.  The seven previous instances were in 1992 (when Labor's derived 2PP declined in four Newspolls in a row), 1997 (twice), 1998 (twice), late 1999-early 2000, and 2001. 

The Newspoll was also the first poll by a major pollster to show Abbott with a negative net satisfaction rating.  A little while ago I started a Not-A-Poll for readers to predict which leader would be first to fall below zero and 82% of voters correctly picked Abbott:

At -5, Abbott's netsat is still one point above his pre-election rating of -6, but no-one will be surprised should it fall further. Past Newspoll history shows that an elected PM's netsat always improves after the election (this applies whether they were elected from Opposition or re-elected) but that four of the past nine winners had regressed to their pre-election netsat after two or three months, so it won't be unusual should Abbott do the same.  (John Howard was the best at making the netsat honeymoon last - he was above his pre-election level for over a year after the 1996 and 1998 elections - but off a very modest base in both examples.)

Opposition Leader Bill Shorten, meanwhile, continues polling positively.  He's on a net +17, and the percentage satisfied with his performance (44) is higher than that for Abbott (40) for the first time.  He's also very competitive with Tony Abbott as Better Prime Minister (the 41-34 "lead" to Abbott actually means that Shorten is in front, since preferred/better PM ratings that include a don't-know option favour the incumbent.) 

Peter van Onselen here reckons "it is how Abbott compares with Bill Shorten that will worry Liberals the most."  Actually if they have much familiarity with polling history, the Liberals will realise the comparison with Shorten is the least of their polling problems.  As one of my earliest pieces on this site The Abbott Factor showed, Opposition Leaders tend to be popular in their honeymoon phases, and their netsats tend to get worse as the uncommitted vote goes down.  Here's a graph from that piece (not yet updated for ratings since, but the pattern won't have changed that much):



This graph shows the average opposition leader net satisfaction score for a given percentage of voters who are uncommitted.  The better known an Opposition Leader becomes, the less the voters usually like them. 

Shorten's uncommitted rating is currently 29, so he's expected to have a netsat of about +10 while people give him the benefit of the doubt.  He's above that, but not so much that people should get carried away about it yet.  Especially not given that the fate of Opposition Leaders is usually grim.

We're still at this phase of the shelf-life of a typical Opposition Leader as chronicled in that article:

Stage 2 (8% of Newspolls).  Includes a lot of bounce phase ratings.  The leader is probably still fairly new but the change in leadership has kicked in and the Opposition's standing has improved (on average to almost equality, though with enormous variation of course).

There's also the problem that, as The Abbott Factor showed, Opposition Leader ratings just don't have that much influence on 2PP voting intention.  If Shorten can maintain a +17 netsat when his uncommitted rating drops below 20, then that might be worth a point in Labor's favour, compared to if his rating was zero.  But most likely it will be swamped by other factors.

And while Shorten's position might seem handy, two months into the job Opposition Leaders Rudd (won), Latham (lost), Downer (booted), Turnbull (booted), Beazley I (lost twice) and Hewson (lost the unloseable election) were all polling somewhat to much better netsats than Shorten is now.  Howard II (won) and Beazley II (lost) were polling very slightly better.  Only Howard I (lost then booted), Peacock II (lost), Nelson and Crean (booted) and Abbott (won at second attempt) were polling worse.  So not only is there nothing special about Shorten's ratings at this stage of his career, but there is no relationship between an Opposition Leader's ratings two months in and their eventual electoral fate.

An indicator Shorten is doing historically well on is the increase in his approval rating since his first poll.  An increase of 12 points from his first to fourth Newspoll sits him fourth on the list behind Rudd (+27), Latham (+24) and Downer (+23) - one winner and two duds.  However, since Shorten's first Newspoll had a very bad 2PP for Labor, that is not a very meaningful statistic.  If we start to see Shorten polling approval ratings of, say, 55, that will be much more interesting.

 The Overall Picture

We also saw weak confirmation of the move to Labor from the perennially sluggish Essential, which dropped a point to 51-49 to Coalition (probably in this case because this week's sample was pretty much 50-50).  Essential also gave Abbott a net negative rating (albeit only -1), with Shorten on +8 and Abbott's Better PM lead shrinking to 10 points (43-33).

I've recently recalibrated the aggregate on the sidebar and relaunched it. Those wanting to see how it's done can check out the methods discussion here.  It now includes pollster accuracy weightings and hopefully the model will be pretty stable from now til the next election.  Here's my tracking graph for the last ten weeks:



At this moment the aggregate, showing some caution about the recent change, has Labor with a slim lead of 50.2:49.8 (which probably wouldn't win them an election held now, but it would be very close).  On the weekend reset this will change to 50.4 assuming there are no more polls released til then, and then we will see what impact next week's polls have on these figures.  Next week's could be the last batch of serious polling before pollsters pack it in for Christmas.  Note that Bludgertrack shows Labor with a more serious lead at 51.2 - the difference probably being created by a combination of its faster disposal of old data, different treatment of Morgan's methods changes and heavier accuracy weightings.

It's rather striking even on my version though: the new Abbott Government has lost 2.8 points of public support in three weeks.  That's about as much as the Rudd Government, clearly coming off a leadership transition bounce at a rapid rate, shed in the whole election campaign.  At this stage it doesn't mean anything predictively, but the Coalition's batteries seem absolutely flat after a long exhausting year, and it's time for them to get some rest and reconsider their approach.  A completely unnecessary diplomatic row with Indonesia and a hamfistedly explained policy backflip on education funding have been just two of many lowlights for the Liberals in the last few weeks.

How Fast Do New Governments Lose Leads?

And how fast do new PMs become unpopular?  The answer to both of these questions is very similar, and on both of them the Abbott government has been by far the fastest to fall behind after being elected from Opposition.  For pre-1986 polling I am using old Morgan Gallups.

We have polling back to Menzies (1949) for new governments, and while polls were taken very sparsely in those days, Menzies' government enjoyed strong leads for at least its first eight months into 1950.  It certainly lost the lead within six months after winning the 1951 election.  No approval figures were taken in those days.

Labor under Whitlam (1972) had a relatively short honeymoon, but was certainly still leading after five months.  Polling was close after seven and the party had lost the lead after 10 months.  Whitlam himself took a year and a half to start polling negative netsats.

1975 is always anomalous because Fraser was artificially installed as caretaker PM before the election. Ignoring that, the Coalition under Fraser (1975) had one close poll five months in but generally led until falling behind after one year.  That was also the point at which Fraser polled his first negative rating as an elected (as opposed to appointed) Prime Minister.

Labor under Hawke (1983) soared in the polls for two years.  Negative ratings for Hawke and poll leads for the Opposition did not start appearing until mid-1985.

The Coalition under Howard (1996) took fourteen months to record negative leadership ratings and nineteen months to lose the lead.

Finally Labor under Rudd (2007) took two years and four months to record a negative leader netsat.  It's debatable whether Labor were ever really behind on the 2PP front in the 2007-10 term (in my view not, despite a few polls showing Coalition leads), but the lead was certainly lost soon after the 2010 election.

And now we have a case where a new government has lost the lead and recorded negative leadership ratings after just three months in office.  The comparisons above - albeit based on sparser data for the early ones -  show that nothing like this has ever happened before.

But Surely Polling Means Absolutely Nothing So Far Out?

I wouldn't be so fast with that assumption just yet.  Looking at the full history of governments (elected from opposition or re-elected) that have lost the polling lead in the first six months or so of their terms, it's notable that most of the recent examples (including the final governments of Whitlam, Fraser and Gillard) have lost, while the final Hawke government survived its term only at the cost of its initial Prime Minister.  (Note: for the Newspoll era I am not counting falling behind in a single poll as losing the lead.) To find examples of governments losing the lead early and then winning under the same PM we have to go way back to the Menzies governments of 1951-4 and 1961-3.  (NB added Oct 2014: I have recently found that the second Hawke government fell behind a little over six months into its term, and at one stage was down by about 54-46.)

Is that pattern of poor performance from a small number of examples signal or noise?  Can't say; there's just nowhere near enough evidence. But what I will say is this: the defeat at the 2013 election was supposed in some circles to have dealt Labor a mortal brand-trashing blow from which it would take a decade or more to recover (if it ever did), committing the party to a long time back in the wilderness.  In reality it was not such a heavy defeat as all that, and indeed the loss spared most of the talent Labor went to the election with, doing the party the great long-term favour of eliminating both Kevin Rudd and Julia Gillard, and the possibly even greater one of forcing it to place some mild restraints on the various products of an internal culture of backstabbing and faction-hackery. (I don't think they have gone nearly far enough yet, and I think these problems will bite them again in the future, repeatedly, but that is another story.) Within three months of this supposedly shattering defeat for the ages, Labor is already competitive.  This does not mean that they are favourites to pull off the very rare feat of dislodging a first-term government at an election, but it does mean that the argument for completely writing them off for a long time to come has already lost some of its force. 

Things will get really interesting if by, say, mid next year we see the new government crashing to 46% 2PPs and -25 netsats for Tony Abbott.  But we're nowhere near there yet.  The concern could become that while Abbott got away with being an unpopular Opposition Leader (the most unpopular ever to succeed in the role), getting away with being an unpopular Prime Minister is harder, because the PM's popularity is a major driver of the 2PP vote share.  That said, it's OK for a PM to be unpopular mid-term (as were Hawke, Fraser, Howard and Keating at various stages in the leadup to elections that they won), provided they can repair enough of the damage by election day.

But Surely This Is Just The Silly Season And People's Heads Are Full Of Tinsel, They Are All Switched Off And Not Thinking About Politics And Polling Taken At This Time Is Not Representative?

Nope.  The idea that polling taken in December is in any way different to polling taken in November (or I suspect any other time for that matter) is a furphy that was dealt with last year.

Updates will be added to this piece to cover all polling up til the end of the year.  There may not be much more of it to come.

Addendum: State Precedents?

Had a look through old Newspolls trying to find any example at state level of a new government falling behind so quickly.  What I found was similar to the federal pattern: new governments typically get a honeymoon bounce above their election result, or else their result was so massive they don't need one to keep the lead.

However I did find that in its first quarterly Newspoll on its election, the Richard Court government in WA (1993) was most likely slightly behind on a 2PP basis, before opening a massive lead in later polling and going on to re-election. It is hard to avoid suspecting an impact of federal factors on those results - the Keating government had just been re-elected when Court polled badly, and it initially polled well, but soon became very unpopular.  The Gallop Labor government in 2001 polled badly out of the blocks (and continually for some time after) but it had already been in office five months when the polling period started. 

The Kennett government in Victoria (later re-elected) also slumped in its first year, with a very bad poll spanning the period from 5-7 months after re-election.  Before that, it had led.

And, in the past 27 years of state polling, those cases were all I could find!

Perhaps it is significant here that of all Opposition Leaders to have won government at state level, Kennett and Court were probably the two who went through the worst personal ratings in Opposition.

Update Monday 16 Dec: ReachTEL and Morgan:  Two new polls just through confirm and in one case strengthen the trend to Labor.  The Morgan, based on results from the past two weekends (hence weighted as 4 rather than 5 in my aggregate) gives a reading of 52.5 to ALP by both respondent-allocated or last-election methods.  Given the current aggregate figures this does nothing to change my opinion that this Morgan method (SMS/F2F) has an ALP lean and so it translates to a very thin lead for Labor (especially as it was surely on the 52 side of 52.5).  On the other hand, ReachTEL, which I consider to have no such lead, delivered a 52:48 to Labor as well, based off primaries to one decimal place that come out on my modelling as 52.3.  This has a bigger impact on my aggregate pushing it out to 51.1.  ReachTEL has Abbott on a rather nasty net rating of -20.4, but as I pointed out in the previous Abbott article, their netsats are harsh on the leader if you treat "satisfactory" as neutral.  Treating it has half-neutral half-positive raises him to -12.3 compared to +13.3 for Shorten.  Full results here show that nearly 10% of Coalition supporters rate both their leader and their party's performance poorly.

The original version of this article has been among the most visited pieces on this site this year, among the most reblogged and certainly the most tweeted.  One aspect that has attracted a bit of critical commentary (though only a bit!) is the apparent tension between the playing down of predictiveness of polling at this point and the playing up of the historical record of governments that got in trouble early.

I am certainly not saying that because 3/6 past governments (none of them first-term) that have trailed early have lost, that therefore and for that reason Abbott has a 50-50 chance of going the same way.  But I do think that the historical record does have some kind of underlying logic to it and isn't something we should be sure about ignoring entirely.  We're used to the idea that polling becomes much more predictive as an election approaches and that polls more than six months out don't mean a terrible lot.  But I wouldn't be so sure that extrapolating this backwards to very early term polling and concluding that very early polling predicts absolutely nothing has to work.

Dec 17: Essential has as usual declined to follow the birdie (again showing why a lot of people don't take it all that seriously).  My aggregate, which downweights Essential because of this sort of thing (and could arguably do so more) isn't much impressed by it registering a 51 to Coalition while every other pollster has Labor ahead, and has moved a whole 0.1 of a point in response.