Wednesday, May 7, 2014

Budget Blowout ... In Coalition Polling

Scroll to bottom for latest updates - updated through May and first week of June

Current aggregate: 54.2 to Labor (+0.3 since last week, +2.9 in five weeks)
Tracking graph (smoothed)




Advance Summary

After holding a slim and by no means election-winning polling lead for five months from November 2013 to April 2014, the Labor Opposition moved into a clear and widening polling lead in the weeks surrounding the 2014 Budget, and would easily win a hypothetical election "held now".  The polling position of the Coalition is roughly comparable to the two worst slumps of John Howard's first term, but many federal governments have recovered from similar, and in at least four cases worse, polling situations.

The federal budget is among the worst received in the last 27 years, second only to the 1993 "horror budget" in terms of its perceived impact on poll respondents and providing the worst ratings since the early 1990s in terms of perceptions of its impact on the economy.  Tony Abbott's personal ratings are among the worst recorded in one pollster's history (Nielsen), while in another's (Newspoll) they do not stand out so much, but are nonetheless the worst results he has recorded since late 2012.



The budget polling results also indicate that Labor is more competitive on economic questions than might have been expected after its bad loss in last year's election.

While it is true that governments tend to recover from bad polling if they take tough measures that prove to be justified, it does not follow from this that a government that polls badly while taking tough measures is doing the right thing.  In stark contrast to the 1996 Howard budget, the current budget is being seen less as necessarily tough and more as needlessly unfair.

=====================================================================

Original Text

With five different polls reporting in in this weekly cycle - the most in any week since the election - it's a good time to start another federal polling roundup.  This one is intended to run with updates added until the end of May, thus covering the polling picture over the next few weeks as the Budget is anticipated, delivered and sinks in.

Another reason that this is a good time for a national roundup is that, after five generally stable months with Labor maintaining an insignificant lead, something has actually happened.  And the news isn't good for the blue army:

Aggregate - Spiky
Aggregate - Smoothed
(The numbers along the bottom are just weeks since the aggregate started, which in turn was four weeks after the election. I can explain why one has one more week than the other if anyone really needs to know!)

This week's polls have seen my aggregate jump from 51.3% to 52.4% to Labor, meaning that the ALP would be very likely to win an election held "right now".   This is the best position Labor has been in since the election, by a whole point.

Some may be surprised it's not even higher.  But this rundown of this week's 2PP polling should explain why: We had Galaxy and Essential at 52 for Labor and Newspoll at 53.  We also had a Morgan reported as 55 and a ReachTEL reported as 54 (I get about 53.9 off the decimal primaries).  But the 55% for Morgan was, as usual, their respondent-allocated preferences, and the last-election result was 53.5 from a poll that's been shown to lean Labor on that measure compared to the usual suspects.  There is also just a little bit of ballast from the two previous polls left in the aggregate.  

As for ReachTEL, while their national polling has not previously shown any lean to Labor, and indeed at stages last year seemed to have a slight Coalition lean, in the last six months its 2PP results have very consistently displayed the same relationship to the aggregate as Morgan's current series.  I have no knowledge of any methods change that might be driving this, and it's especially plausible with robocalling that this sort of thing could happen even without a methods change.  But it's very unlikely to happen by chance.  One hypothesis that I've seen is that robopolls, no matter how well scaled, could be a teensy bit more prone to "motivated response" than other polls.  The idea is that there is an elevated resistance among some voters to taking automated calls compared to other survey methods, but whichever side of politics is angrier at a given moment is more likely to overcome it.  

For now at least I think there's enough evidence to justify applying an adjustment, though my past experience is that as soon as I apply an adjustment to deal with a poll behaving in this way, it often stops doing so.  I've also made a number of other minor changes that are listed on the aggregate methods page.  With all of these changes removed, the aggregate would be just 0.2 points higher for Labor, at 52.6%.   BludgerTrack later this week may deliver a higher result. [update: nope, it is lso 52.4.]

Follow-Up on the Green Vote

My previous roundup Nielsen: Greens 17 - What Is This? covered a record primary reading for the Greens in a single poll coming immediately after the WA Senate election.  Either that level of support wasn't there in the first place or else it's gone away very quickly (almost certainly the former) since nothing since has shown anything like it.  There is a quite high reading of 14 for the Greens in the current Newspoll, but the rest are all in the 10-12 range, and taking into account that pollsters tend to overestimate the Greens' support level, the real level of Greens support is probably about 11%.  I'd say it's a rather soft 11 at that, since there is lots of anecdotal evidence of temporary "parking" of votes by disgruntled supporters of both majors in the Green column.  

Much has been made of the Newspoll especially showing no change in Labor's primary, but given that the Newspoll has an atypically high Green vote, this seems to be a case of commentators looking at just one poll and ignoring the contrary data from the others.  While Newspoll has the ALP primary at just 34, all the others have it in the high 30s.  

Abbott Returns To The Mire

Only two of the five polls this week produced leadership ratings and both showed Tony Abbott with the worst ratings of his term.  ReachTEL showed just 26.5% rating Abbott's performance as PM "very good" or "good" with 56.8% saying "poor" or "very poor".  Using my standard conversion method for ReachTEL leader ratings (treating the "satisfactory" group as half weakly positive and half neutral) this is equivalent to a Newspoll reading of -22 (35% approve, 57% disapprove).  Hence I was not surprised when the actual Newspoll reading was -21 (35-56).

If that sort of Newspoll rating set for Abbott, looks familiar, it's because it's exactly the same one he polled three polls in a row as Opposition Leader in the first four weeks after Kevin Rudd returned.  Indeed, Abbott last had a higher dissatisfaction rating way back in January last year, but this is the sixth time since then that exactly 56% have given him the raspberry.  It's hardly a dreadful rating in the context of past incumbent slumps, but this week at least the PM is quite unpopular.

In the context of bad ratings for a newish PM, Paul Keating was again way ahead of Abbott, polling -32 four months into the job (incidentally, also in May!) But no other new PM has seen ratings this bad anywhere near so quickly.  Also, and this rather surprised me in the context of the awful ratings Abbott used to poll as Opposition Leader: the loss of 14 netsat points in the current Newspoll (admittedly over four weeks rather than the usual 2-3) is Abbott's largest Newspoll-to-Newspoll loss since just after the 2010 election.  Indeed both his higher Newspoll-to-Newspoll losses came under clearly atypical circumstances: the first Newspoll with Julia Gillard as PM (which was an outlier compared to other polls at the time), and the second poll after the 2010 election (when Coalition voters suddenly noticed their man had failed to form government.)

Bill Shorten, meanwhile, has benefited from the government's problems, improving his standing slightly to -6 in Newspoll, and I calculate the ReachTEL as equivalent to -2.  Given the 2PP result, a meaningless two-point lead to Abbott as "better Prime Minister" (Newspoll) - an indicator always tilted in the incumbent's favour - is about what should be expected.

Why Is This Happening?

Two things that usually affect polling much less than many people think they will are good and bad political news cycles, and Budgets.  But sometimes these things can have large impacts, and the Coalition is creating genuine concern about its various budget intentions.  Chief among these are: changes to the retirement age, the possibility of required co-payments for going to the doctor, and the "deficit levy".

The release of a harsh and nasty report on funding options by a new government is a trick as old as the hills.  A new government can release such a report, create a range of panic about the options contained within it, then point out that these are only options for discussion.  It can then impress everyone with its sensible nature by rejecting most of the most extreme proposals, and usually the sense of relief is so great that the government is able to smuggle through a few genuine nasties without anybody who would change their vote really caring (unemployment welfare being a common target).  By normal standards, those lefties who even reacted to some of the Commission of Audit's wilder suggestions would have been playing directly into the government's hands.  

For the time being, though, it's not clear that the government has control over the impacts of its political trollbait.  The aspect of this that's attracting a lot of debate as to whether it's a "Gillard moment" and election-loser or just a non-issue that will change nobody's vote is the proposed "deficit levy" on upper income brackets.  Details of the levy vary but it was initially widely reported (apparently inaccurately) as 1% of incomes above $80,000 and 2% of above $180K.  Subsequent reports have referred to 1% of that portion of income above $80,000 and 2% of that portion above $180K.

The exact form of it doesn't really matter at this moment, and nor by itself does the general perception (shown by Galaxy and ReachTEL) that the government is breaking an election promise. First-term governments have great public latitude to break election promises since they can generally blame them on having discovered that the budgetary mess left by their precursors is much worse than everyone thought.  

However, this particular levy is contrary to the Coalition's brand. The Coalition has a brand image that includes that it does not increase taxes for high-income earners, which is generally seen as a left-wing thing to do, with the economic right strongly believing that such taxes are not merely morally unjust but also economically counter-productive.  Now that the Coalition is prominently pushing a tax-the-rich plan as a solution to the deficit issue, some of its own supporters are deeply concerned, believing that this is not just a bad idea to begin with, but also electorally disastrous. Some of this sort of thing is mischief-making by anti-Abbott sectors of the party, and some of it comes from economic dries and libertarians who just aren't ideologically flexible, but this by no means explains it all.  Liberal voters of various kinds are warning their party not to take their votes for granted if they go ahead with this, and some are even saying they'll vote for minor parties if the idea makes it to the next election. 

Two polling questions on this that include party breakdowns illustrate the confusion the deficit levy idea has created in the electorate.  Firstly, ReachTEL:


Secondly, Essential: (click for larger versions):


The Essential question is a bit mean to even the strongest version of the deficit levy proposal, since no version of it targets all middle-income earners. Rather, it cuts in just above the median, so "high and upper middle income earners" would have been a fairer description.  The ReachTEL question effectively canvasses the 1%-of-all-income version, which I believe is not still being seriously discussed (if it existed in the first place).  But in both cases we see lukewarm support from the Coalition supporters and lukewarm opposition from Labor and Greens voters.  The confusion is that left-wing voters normally support taxing higher income earners, and right-wing voters normally oppose it.  But because the proposal comes from the Coalition and is addressed at one of the Coalition's totem issues (the deficit) many, but nowhere near all, Coalition supporters are dragged into supporting it, while many, but nowhere near all, Labor and Greens supporters think they have to oppose it.  Partisan loyalties are trumping ideology, but not for everyone.  

That said, if the question involved taxing high and upper middle income brackets to spend more on something more concrete and direct than the national budget, the results would be rather different.  And if Labor were in government and proposing the same thing, the results would be utterly different - Labor and Greens supporters would be strongly cheering and Coalition fans strongly opposed.  

The nothing-to-see-here case is argued by Peter Brent (who by the way has already called the next election). He points out that the general voter split is much more even on this issue than on doctor co-payments, which are quite strongly opposed (33.5-56.5 in ReachTEL).  I'm not completely sure about all this, since on the doctor co-payments issue the opposition is coming strongly from Labor and Green voters, with higher Coalition-voter support than for the deficit levy. I also think the high don't-know rate for the deficit levy could represent confusion as well as indifference.

It's especially perplexing for some Coalition voters that the deficit levy is being even considered while Abbott's luxurious paid parental leave scheme remains on the table.  Here Galaxy shows a remarkable level of cross-party consensus in opposition, with only 26% of Coalition supporters thinking now is the right time for this measure given the state of the budget (62% opposed!), joined by slightly lower levels of Labor (23%) and Green/PUP/Others (c. 18%) supporters. A question on this from Essential, not mentioning the state of the budget but following various other questions about the budget, also finds supporters of all parties preferring the current policy, though to a more muted degree partly because of Essential's high don't-know rates.

I think that in its strongest rumoured forms, the deficit levy idea would be a misjudgement of what the electorate wants.  The debt issue is a political totem: people will feel comfortable if the incumbent government is making progress towards repairing it and noises about doing so, but it is not so paramount that the government needs to take radical, brand-damaging steps in obsessive immediate pursuit of it. [Note added later same day: not surprisingly, this evening's rumoured versions of the deficit levy saw further scaling down of the proposal, to the degree that it is probably now harmless.]

The Myth Of The "Budget Bounce"

There's been little of it this year, with many commentators expecting a budget slump.  But a while back there used to be a view that the delivery of the federal budget would give struggling governments a second wind.  Quite why anyone believed this is unclear, given that budget telecasts are among the most boring television events in existence and given that voters are more likely to be worried about the nasties in the average budget than impressed with the goodies.  Anyway, back in 2007 Possum looked at this in Budgets Bounce Like Dead Cats - to that point, there had been no effect of budgets on government polling, or if anything perhaps a negative one.

Years since then have made no difference to this conclusion.  The Rudd government lost some support soon after Budget time in 2008, though that was from a stratospheric high to a still stratospheric high and therefore meaningless. 2009 was a similar story and 2010 happened to coincide with Rudd's popularity falling off a cliff for unrelated reasons.  The 2011 budget also appeared during a slide and 2012 and 2013 both had no obvious impact.  

Anyway, I thought I'd have a look at the average level of government 2PP polling for polls taken entirely or partly in each month of the year in the whole history of Newspoll.  Here's the result:



The average government 2PP by month in Newspoll history (since 1986) has varied by not much more than a point between the worst month (September) and the best (December). Budgets in this time were issued in August eight times, September once and all the rest in May.

Lest anyone think that this gives much credibility to the idea of a holiday "silly season" at the end of the year, it actually doesn't.  The December sample is quite small with only 29 readings included, and these happen to include three from 1992, when the Keating Government had just enjoyed a massive polling resurgence on the back of problems with the Liberals' Fightback! policies and alarm at the behaviour of the then newly elected Kennett Government in Victoria.

If anything, the weak and shallow (and probably not statistically significant, though I don't even see a need to check) pattern is for government ratings to be on the slide from April through to September (doubtless explaining why Julia Gillard twice considered August-September such an excellent time to go to the electorate).  Perhaps budget anticipation contributes to this; perhaps it just happens anyway and budgets make no difference to it, but in any case the variation in average polling through the year is slight. Anyway, any idea that budget time tends to produce big shifts in any consistent direction is just wrong.

Some Other Polling Snippets

Also of interest recently are:

* An Essential finding that 10% of Coalition supporters say that obviously positive ALP reforms would make them less likely to vote Labor.  (The 2% of Labor supporters who agree with them are probably the honest ones.)

* Essential findings on the economy, showing again that voters rate Labor better at handling most suggested aspects of the economy but still prefer the Coalition as overall handler by a large margin 

* Essential preferred Treasurer polling, marred by a high don't know rate but showing Joe Hockey leading Chris Bowen by a modest 33:27.

What Does Labor's Lead Mean Predictively?

In a historic sense, little or nothing.  Historically about 65% of governments fall this far or further behind at some stage in their polling history, but about 65% of those get re-elected.  Of course the chances are higher if the government's worst position in its term is something in the high 40s rather than much lower.   It's unprecedented for a first-term government to be this far behind this quickly, but if it doesn't get much worse then it won't matter; indeed, better now than closer to the election.  About 20 months into the Howard government's first term it slumped to a rolling average in the low 46s, and was still re-elected easily in seat terms despite a low 2PP vote.  

Governments have to be willing to fall behind in polling in the knowledge that they will probably recover. Governments unwilling to take this risk (Labor in so many ways in 2007-10 being a case in point) are likely to suffer for it later.  If this is all there is, then the current slide is far from a disaster, and some cool heads in the Coalition party room would probably think 47:53 in Newspoll in the circumstances isn't actually so bad.  This is only a problem if it is only the start: if we now see the Coalition's vote crash to something outside the competitive zone (ie well below 46) and stay there.  

This article will follow polling developments over coming weeks, probably with updates each week as new polls arrive.

9 May 2014: Twin Layers of Lightning:  A new ReachTEL has been released in the Fairfax press, which just like the other one comes in at 54:46 to Labor, though on the available primaries it is looking like this one is worse for the Coalition than the one from five days before, unless the rounding is very friendly to Labor.  Anyway the easiest way for my aggregate to cope with the double ReachTEL queue was for me to declare that it was Saturday a few hours early, so I've done that and the weekly reset is applied.  While I wait for further details of exact primary votes (we so far know only the whole numbers: ALP 40 L-NP 38 Green 11) I've added the poll to my aggregate based on an estimated 2PP of 54.4.  Labor gains 0.3 of a point from this at present, but depending on the rounding the gain might be as low as 0.1 (ie 52.5 to Labor).  The poll unsurprisingly finds that Coalition voters are fairly evenly split on the proposal to increase the pension age to 70, while non-Coalition voters are strongly opposed.

10 May: More on the ReachTEL:  The decimalised primaries are out and my 2PP estimate from them is 54.2% to ALP, adjusting the aggregate to 52.6.  The poll also includes findings that 53.7% of respondents support imposing a deficit levy of 1 to 2 percent on "high income earners", compared to 32.4% opposed; echoing last week's finding that a 2% levy on incomes above $180K had broad support. Bearing in mind that what is being proposed now is not even a levy on a person's full income, it is possible real support for the cut-down levy, if described precisely, would be higher.  Oddly, a rather high 42% say that the levy would make them less likely to vote for the Coalition; without the benefit of the party breakdown yet, I suspect the lion's share of this number are locked-in ALP or Greens supporters who are fibbing.

There is a clear difference in polling between versions of the deficit levy that could have targeted "upper middle" income earners and those that only target very high income earners.  Given that the deficit levy will be of the latter form, it now seems the blowout in Coalition polling is either not still largely related to it, or else is related to misperceptions of it.

13 May: Essential: After the deluge of last week some expected a Nielsen this week but not to be, so we have only Essential to add to this stage.  Essential came in with an unchanged 52:48 to Labor, leaving my aggregate also at 52.6 (which will be a new week-ending high if nothing else comes out by Friday).  Essential leadership ratings also showed Abbott on his worst result in that series since the election (-20, down 14 since last month); it is also his worst rating since January 2013.  With Shorten on a mundane -2, the most noticeable (if not necessarily notable) result is that Essential is the first poll to show Shorten as preferred as Prime Minister over Abbott (37-36).

Next week we should get some signs of how the budget contents are settling in the electorate.

15 May: The current BludgerTrack reading is also 52.6, with BludgerTrack also now implementing a response to the recent ReachTEL lean to Labor.  The difference is that BludgerTrack's weekly headline rate has been this high before; mine hasn't.

18 May: Galaxy: Hopeful expectations on the left of a savage post-budget plunge in the Coalition's numbers were not supported by the first post-budget poll out, Galaxy.  Galaxy found Labor ahead 53-47 (up 1), which takes my aggregate to a new but hardly massive high of 52.8 to Labor.  The poll finds a massive 75% believing they will be worse off, compared to 11% better off, but also finds that 41% think the budget will be good for the economy, compared to 46% bad.  That's a very high disconnect between the two net ratings there.

Looking at similar questions in Newspoll historically, as Peter Brent has done, we see that usually voters think the Budget will make them worse off (even if they like its economic impacts), but rarely to this degree.  However the only previous Budget provoking a more extreme net "worse off" result in Newspoll (1993) was also considered an economic disaster.  This one, at least according to Galaxy respondents, is not.  Nonetheless, no Howard government budget received a net negative Newspoll economy rating, so it will be interesting to see where this one stacks up in the Newspoll survey. Brent has also pointed out that the order of the questions may be different in the two surveys.

In 2013 there was relatively little difference between how Galaxy (14-48, net -34%) and Newspoll (11-44, net -33%) respondents rated the impact of the budget on themselves.  In 2012 Galaxy had a 23-46 (-23%) personal rating compared to 18-41 (-23%) for Newspoll.  So from a very limited set of data points it looks like Galaxy has similar net ratings on this issue but a lower undecided rate.

Some more on Newspoll budget data historically: on average the net (better off - worse off) rating trails the net (good for economy - bad for economy) by 33 points, and the highest previous disconnect in Newspoll between the two was 58 points, in 1996 (-21 personal, +37 good for economy).  So if Newspoll replicates Galaxy then the number of people thinking the budget is bad for them but good for the economy will be very similar to the first Howard budget in 1996 (at least c. 30% in each case), but the number thinking it is bad for both will be much higher now.

(I've moved the graph showing where the Galaxy result fits in to the Newspoll pattern to the bottom and replaced it with the Newspoll result as they were nearly identical.)

One poll down and probably three or four this week to come.  

18 May Nielsen:  Nielsen has come out with a relatively shocking poll for the government, though one that still didn't match some of the wilder expectations out there.  It's 56:44 to Labor with Abbott on a -28 netsat, with a 62% disapproval.  Given that Nielsen netsats are a little bit mild compared to Newspoll's that's a rather extreme figure.  William Bowe notes:

" The net approval rating of minus 28% was exceeded by Julia Gillard in Nielsen on only two occasions, in September and October of 2011, and equalled in July 2011. The figures resemble Gillard’s final rating from June 2013 of 36% approval and 61% disapproval."

I can add to that that this is Abbott's second-worst net rating from Nielsen ever; he managed -29 as Opposition Leader in December 2012, a time at which his leadership was looking rather shaky.  And that Howard's worst Nielsen net rating ever (in a sign of how benign Nielsen normally is) was just -21.

Noting that Nielsen has been erratic to say the least since the election, let's wait for more polls this week before anyone gets too carried away by this result.  

10:35 pm: The wait will be short; Newspoll is apparently also a shocker for the Coalition!

11pm Newspoll: Yes Newspoll has joined Nielsen in returning something pretty ugly: 55:45 with Abbott on a -30 netsat, Shorten with a 10-point PPM lead, and only 5% deluded enough to believe they will be better off under this Budget.  (The Coalition's previous worst there was 7% in 2002, while in 1993 the figure under Labor was 4%).

This has sent my aggregate to 53.8 to Labor but for the time being it has only the three new polls and a lot of ballast from the week before last.  So I would not be surprised to see it go up when Morgan reports tomorrow. 

Abbott's rating is obviously his worst as PM; his worst as Opposition Leader was -36 and he was -30 or worse five times in late 2012.  As for Shorten's 10-point lead as better PM (for the first time in Newspoll) this is matched or exceeded only by the following Opposition Leaders since 1986:

* Hewson over Hawke and Keating at the time Hawke was removed, four times with a high of 15 points.
* Hewson over Keating following the 1993 horror budget, four times with a high of 16 points.
* Howard over Keating twice in March 1995 with a high of 12 points.
* Rudd over Howard eight times scattered through 2007 with a high of 13 points.
* Abbott over Gillard in June 2013 just before Gillard was removed, with a lead of 12 points.

But the all-time champion for PPM lead over the incumbent is of course Alexander Downer (you can win bets on this one surely) who led Keating by 20 points at the height of his short-lived career in 1994.  (Newspoll stopped polling the question briefly around this point.  It is not recorded whether they decided it was broken.)

Some may be wondering how this all compares to the low points of the first Howard government's polling.  It's already about the same.  The first Howard government had a Nielsen low of 44% 2PP and a Newspoll low of the same.  On my Newspoll rolling averages the Beazley-led opposition was a touch below 54 twice: in Nov-Dec 1997 and July 1998.  The latter had the very bad Nielsen results so it's likely the 2PP for Labor was really a shade over 54 around then.  The Whitlam government in its first term also seems to have been down around the 46s at some times.

More on Newspoll and Nielsen later.

19 May: Nielsen gives Shorten an 11-point lead as PPM (51-40) and a quite respectable +8 net approval (47-39).  The Nielsen poll still has the Greens primary vote suspiciously high on 14%, a result other polls have not generally been getting, following on from the record 17% last time.  The budget is rated fair by 33% and unfair by 63%.

Newspoll's figures for Abbott are satisfied 30, dissatisfied 60.  The 60 is his biggest dissatisfaction rating since November 2012 (his worst ever was 63).  The better PM figures are 44-34; one comparison is 45-33 to Abbott over Gillard just before Gillard was rolled.

The budget net "good for the economy" score in Newspoll is very similar to Galaxy's at -9 (39-48) and the net personal impact score is exactly the same at -64 (5-69). (Newspoll offers a "neither" option, unlike Galaxy, resulting in lower positive and negative scores.)  Here is the graph (previously included in the Galaxy section) that shows the Newspoll budget ratings compared to those of previous governments:



Blue for Coalition, red for Labor, data since 1988, selected years labelled.  We can see that while this budget is not in quite the horror class of 1993, it is still the Coalition's worst-rated budget on both scores by quite a distance.  

When asked if the Opposition would have done any better, voters apparently don't really think so (39-46).  But that 7-point gap is actually the sixth lowest since 1987, and the lowest in that time for a Coalition government.  Also, in all five cases where the gap was lower or reversed, the incumbent government lost the next election (I am not saying this is causal!).  This seems to be one of those indicators, like Preferred Prime Minister, on which the incumbent government has quite a house advantage, with No on average beating Yes by 17 points.  Moreover, the Coalition under Howard had a better average margin (23 points) than Labor under Hawke, Keating, Rudd or Gillard (13 points).  So this result is actually also fairly bad, albeit highlighting that the Coalition's perception edge on economic management remains even when people think the budget is quite unfair.  There has only been one budget (1993) for which more voters have answered Yes than No to this question, with a tie in 2013. 

There is naturally a strong link between what voters think of a budget's impact on the economy and whether they think the Opposition could have done better.  This budget falls almost exactly on the line of best fit for those things:




The current budget is the only blue one left of the vertical axis.  The budgets well above the line are those where voters were unusually prone to doubt the Opposition could have done better, given their impression of the budget's economic calibre.  These include 1988 and 1989 (because the Liberal Opposition at that time was a rabble) and 1996 and 1997 (because the Labor Opposition at the time had been economically discredited.)

Well below the line there's that little cluster of red dots near the zero.  These are 1995 and the last three Swan budgets - all cases where the government was seen as struggling for economic credibility.  That blue one way below the line on the right is Howard's last budget, which was very well received but by that stage voters were positively enthused about the Rudd opposition.   

I think the closeness of the question on whether the Opposition would have done better is actually a positive for Labor.  The gap is lower (by two points) than it has ever been in a case where the Government delivering the budget was re-elected.  That said, the Menzies government handed down a notoriously nasty budget in 1951 before we had these sort of polls and went on to win.

19 May - More Poll Data Snippets:

Some data out now allow for my familiar hobby of finding the unstated breakdown on an issues question for minor party voters.

In Newspoll, the breakdown for the 39-48 result on the budget being good for the economy is 81-10 good-bad for Coalition supporters and 15-77 for Labor supporters.  For minor party supporters (Green/PUP/others/ind) I get 16-58. The disenchanted Coalition supporters who strongly disliked the budget have already shifted their voting intention (temporarily or permanently) in protest; the 81-10 is the reception from the diehards.

Newspoll: better or worse off personally: Coalition 11-57, Labor 0-83, others 3-65

Newspoll: would Opposition have done better?: Coalition 8-87, Labor 70-13, others 37-38.

Galaxy: good or bad for economy: Coalition 81-11, Labor 19-68, others 13-67 (!)

Galaxy: better or worse off personally: Coalition 22-60, Labor 4-87, others 5-80.

There is a small ray of light for the Coalition from a Nielsen question that found a 49-48 split on whether the budget was economically responsible, but given that the 49% would include those who thought it was economically responsible but bad in other ways, that's not really too much to shout about.

Some more charts and analysis drawing generally similar conclusions to the above at Poll Bludger.

19 May: Morgan: This week's Morgan (and it is just this week's; post-budget only) comes in with a whopping 57.5:42.5 headline result for Labor, but once this is converted to last-election preferences (I get c. 55.5) and then adjusted for Morgan lean, it's been defanged to something in the area of Newspoll and hence has relatively little impact on my aggregate - another 0.1 of a point goes to Labor.  It is quite possible that PUP fans (a record 6.5% of them) are presently slightly Labor-leaning on preferences and hence that last-election preferences underestimate the damage of the Budget on the Coalition's numbers; on the other hand I have the average lean of Morgan's last-election preferences now running at 1.3 points (compared to the 1 point I correct them by).

We have Essential, probably tomorrow, to probably complete the full set.  If we considered just this week's polls my aggregate would be running at 54.6 so it is possible the Coalition will fall further behind in coming weeks. That said, there won't be much polling next week.

Morgan Update: The published last-election 2PP from Morgan is a surprisingly high 56.5 to Labor.  Assuming that's correct, the explanation would be the state distribution of the minor party votes (and not just rounding).  Pollsters can look at the preference split from the last election by party in each state rather than just nationally.  With that included my aggregate has hit the magic 54 for the first time, with potential to move into the high 54s should results like this week's persist.

May 20: Essential:  Essential has done what Essential does; it has showed no 2PP change and only very slight primary vote changes while three of the other four polls went gaga.  It remains at 52:48 to Labor for the fourth week in a row.  My aggregate stays at 54% and will stay there at the end-of-week reset if there isn't anything new by then.  However Essential's use of fortnightly rolling most likely means that about half of its sample was taken post-Budget.    Essential respondents disapprove of the Budget (30-52) and think it is good for the well-off (51-15) but bad by similar or greater margins for the average worker, low income earners, the elderly and the young.  Voters apparently agree there is a "budget emergency" (56-32) but are not that convinced this Budget will fix it.  Increasing the pension age is the most disapproved of various budget measures tested (17-61).  There were ticks for some measures: making graduates pay HELP debts faster, cutting the ABC's budget and cutting foreign aid.

May 22: BludgerTrack:  The new BludgerTrack estimate is 54.2 to Labor.  Note that the BludgerTrack sidebar now carries retrospective aggregation graphs for 2004-2010.  This is as a result of me managing to extract a full set of Nielsen polling from 1996 to 2011, that used to be on the Nielsen website, from the Wayback Machine.  If anyone would like the Nielsen PDFs I can email them.

May 23: Where Does The Abbott Slide Rate?

I thought I'd have a look at how Abbott's loss of 23 Newspoll netsat points in the last two Newspolls compares with other PMs of the Newspoll era who have suffered such sharp losses in their personal rating.  A loss of this magnitude or greater over a period of at most seven weeks is a fairly rare event, happening about once every two years.  The table gives the size of previous netsat losses, the number of weeks the loss covered, and the eventual fate of that PM in that term.


Every Australian Prime Minister in the Newspoll era has had a similar rapid loss of net popularity to Abbott, although in most cases it came from a much higher base.  I can also say based on incompletely scribbled data from old Morgans that similarly quick and severe losses probably happened at least for the following: McMahon in 1971, Whitlam in 1974 and 1975 and Fraser in 1978.  However as the table shows, a hefty proportion of such losses represent the fading of a honeymoon effect after a PM won an election or returned to the job.  Also, the fates of PMs who suffer such losses are all over the place.

It's also not unusual for PMs to experience ratings as bad as -30 at some stage in their career; indeed, since leadership ratings were first measured regularly in 1969, only Gorton has not polled comparable ratings at some time.  In terms of Abbott's ability to retain office, the question is not how unpopular he may become (unless he starts polling -50s and worse, maybe), but what is his ability to rebuild his personal standing as PM. That's something that polling doesn't yet tell us, although Abbott did achieve the second-best personal ratings rebuild for an Opposition Leader.

Thus far Abbott seems to be at least as much a slave as almost any other PM to the trend of leadership ratings being highly and apparently causally correlated with the party's poll performance, and this makes sense because what's going on is all about the Government.  Bill Shorten's netsat is up 14 points in six weeks, but it's not because he's doing anything differently or right or because people like Labor suddenly; he's just getting free hits from the Government.

Some might suggest that if Keating won an election with a net personal rating of -25, Abbott could just as easily do so too.  Not too likely.  Keating's first term as PM from 1992-3 is the one period in Australian polling history in which the relationship between the PM's rating and the party's 2PP score didn't exist (indeed the relationship was between voting intention and the Opposition Leader's ratings), whereas with Abbott so far it exists and explains 87% of variation.   Figures, because politics at the time was about whether the Opposition were fit to take over from a government that was ready to be turfed.

May 27 Essential:  Essential remained at 52-48 for the fifth successive week.  This had a very small impact on my aggregate, but if it turns out that Essential is wrong then other pollsters will pick up on this next week and the aggregate will move into the 54s.

Essential included Abbott-vs-Shorten attribute polling and it was a bit of a no-contest with Shorten winning every head to head comparison except "visionary" ("won" by Abbott 31-30).  On average Shorten's rating on the fifteen attributes was 14.4 points better than Abbott's.  Since April Abbott's scores have worsened on every attribute (by an average 7.5 points)  while Shorten records an average rise of 4.3 points with improvement on all attributes except "aggressive" (+3). 

A very stark comparison appears in the article by Essential's Peter Lewis and Jackie Woods - Abbott is also outperformed by Gillard on these measures, based on results immediately before Gillard was removed.  Gillard's ratings were better on 13 of the 15 attributes, with one tie and Abbott shading her by three points on "more honest than most politicians".  On average Abbott scores 7.6 points worse than terminal-era Gillard.

May 29: Bludgertrack: After a few weeks of being more or less identical there is suddenly a 1.1 point divergence between my aggregate and BludgerTrack, which has gone to 55% to Labor.

The full size of this divergence won't last long, because in the absence of further polling mine will go to 54.3 this weekend as old polls drop out of the system.  Beyond that about half the remaining difference is caused by differing treatments of Essential.

June 2: With only one poll last week we really need this week's data to round off where the Coalition's budget blues are settling.  First cab off the rank is Morgan with 55:45 by last-election preferences, down 1.5 points on last week's reading.  As Morgan leans to Labor this result actually knocked 0.1 of a point off this week's starting value.  Newspoll and Essential to come and another ReachTEL is probably not far away.

Newspoll: Another poll suggesting things at least aren't getting worse for the government, with Newspoll returning 54:46 to Labor.  Abbott's netsat has recovered four points to -26, while Shorten's has dropped eight to -5 for no evident reason.  Shorten maintains his lead as better PM from last time, 45:35, and the 45% is in fact the equal highest score for an Opposition Leader since Kevin Rudd faced off against John Howard in 2007.  Tony Abbott as Opposition Leader held the same PPM rating exactly twice: in his last poll against Gillard and his last poll against Rudd.

I think now that we have a fair fix on the net damage this Budget has caused to the government's polling position: about three points.  For all the more excitable commentaries in the last few weeks I don't think we are seeing anything catastrophic in this polling yet.  Some might even think that the Coalition has got off light. Tomorrow Essential will bring us Abbott-Turnbull matchup polling, and while the result is likely to be very bad for the PM, I think the results will be for entertainment purposes only.

Ipsos I-view: As noted by Pollbludger today (I actually hadn't noticed it yet) there may be a new bird on the polling lagoon with the arrival of Ipsos I-view.  From their website I've determined they run a fortnightly online panel survey with each survey taken over five days and a sample size of about 1000, and have recently asked voting intention questions.  I've found three recent 2PP results from them, 50-50 released April 14, 51.5 to Labor April 28, 53 to Labor May 13, and with the third they qualify for inclusion in my aggregate, albeit with a low weighting until they have completed eight polls (assuming this happens).  The three results available so far are similar enough to what other pollsters have been recording.  However, there are no voting intention results available in the May 28 release.  (There are, however, leadership and budget results, including a 66% dissatisfaction for Tony Abbott in the latter, and a range of other figures similar to what other pollsters have been getting.)

As one pollster possibly enters, another will depart with the end of Nielsen polling in Australia.  The decision was apparently not related to current Fairfax budget cutbacks.  Fairfax, which has been publishing Nielsen for a long time now, will be running a different poll, the form and source of which remains to be seen.

3 June Essential:  Essential has finally gone to 53:47.  Most voters think Labor should vote against parts of the budget (53%) rather than pass it all (18) or block it (18).  Even 41% of Coalition supporters think Labor should oppose some parts.  Findings on voting for or against specific questions are posted and are similar to what we've seen all along, with co-payments, university cuts and fees, and increasing the pension age the most opposed.  "Party trust to handle issues" scores show Labor improving relative to the Coalition on most issues since four months ago, the main exceptions being national security and asylum seekers.  And finally, the Turnbull-vs-Abbott showdown is a bit of a fizzer this time around with Turnbull leading only 31-18, and a high proportion of "don't know"s and mostly token "someone else"s (they should have been then asked "who?").  Among Coalition supporters, Abbott leads 43-27.  The gap has actually closed since the question was last asked during last year's campaign.

This article has now tracked the government's polling through the pre-budget phase and the weeks after the budget.  I intend winding this one up at the end of this week and starting a new roundup when something else interesting happens.

28 comments:

  1. A really informative article as usual.

    congratulations

    ReplyDelete
  2. Kevin I have highlighted this article today in my modest blog. I also said your blog is a must read if one is a political junkie.
    i do hope you get a few people reading this article from this.

    Keep it up

    ReplyDelete
  3. In some ways I'm surprised the polls have been so slow to respond. It's been fascinating to see Labor wedged into making the economic rationalist case, while the Liberals have to (try to) defend economic profligacy.

    Just hoping the state government makes a better fist of it than the federal government has; their task is arguably more difficult. ( I think it's reasonable to say that you would have to trim around three times the percentage from the federal budget for it to be as challenging a task, and the cuts required are about equal in percentage terms).

    ReplyDelete
  4. Replies
    1. Just a colour change because I felt like it; the orange had been there since McGowan won Indi. Actually it occurred to me later that purple is my mother's favourite colour and I had changed it on Mothers' Day, though if that's the reason it would mark the first time I had ever taken any notice of such an occasion. Anyway it is nicely major-party-neutral though I may fiddle with the shade a bit sometime.

      Delete
  5. I have a question:

    What is the worst 2pp poll result that an incumbent has fought back from to win an election?

    ReplyDelete
    Replies
    1. In the days before 2PP was measured, in February 1953 the Menzies government polled a primary vote in a Morgan Gallup poll of 39% compared to 60% for Labor and 1% for others, equivalent to a 2PP of 39.5%. Polls were very rare in those days but this was no once-off as four months earlier they had been only one point better.

      The Fraser (1980 election), Hawke/Keating (1993 election) and Howard (2001 election) governments all dipped to average 2PP positions of about 44% at some stage in their term but all won; the worst individual 2PP poll result for these I'm aware of is the 40% for Howard in Nielsen, April 2001.

      Delete
    2. there have been some massive come backs by some governments!
      I wonder whether there is a predictor of such a comback? like preferred prime minister? or some other measure or is it just luck or voter amnesia?

      I wonder if this weekend we will see a record low 2pp for an incumbant

      Delete
    3. The best predictor of a government comeback I'm aware of is that historically the Coalition tends to recover much better than Labor does. Keating in 1993 is the big exception to this but that comeback was partly because Hawke caused bad polling by hanging around too long (and partly of course because of Fightback! scaring voters away from switching.)

      Preferred/better prime minister scores don't seem to predict recovery, but they sometimes look like they do because they are so skewed in favour of whoever is PM at the time. So they seem to be saying things are close when the 2PP does not say this. If the 2PP is 50-50 the incumbent PM usually has a lead of around 16 points. If the PPM is level then the government is usually way behind on 2PP. But actually incumbent governments that are behind tend to improve no matter what the PPM scores are. And there have been some poor results by PMs who were way ahead as preferred leader.

      The worst ever poll by an incumbent federal government, by the way, was a 33% primary for the Whitlam government compared to 60% for its opposition, so about 37% 2PP.

      The benchmark for horror budgets is the Dawkins budget in 1993 which caused one Newspoll to be nine points worse for Labor than the one before, though the real damage in the medium term was probably more like five points.

      Delete
    4. I would highly doubt a historic newspoll of any description this weekend.

      Essentially, we got a "back to status quo (trying hard not to make a down down joke) Budget after a change of government.

      Yes, they're tougher than 'normal' budgets; but the government can just sigh and blame it on the previous government.

      Plus there is the fact that this budget wasn't as tough as Hawke/Keating first or Howard/Costello first.

      Plus they debt and deficit story, while for the most part a clever fabrication on behalf of the Coalition has been bought hook, line and sinker by the wider community.

      And, probably most importantly, it was no where near as bad as the Commission of Audit suggested and as such Australians were expecting to become a reality; because of our generally useless political media coverage.

      Essentially, you'll get a probable dip of a point or two in 2pp, but when newspoll asks about the budget specifically, expect a lot more positive responses than you would generally be lead to believe.

      Delete
    5. Difficult as predicting polls often is, I'm also not expecting anything historically awful in this week's results. As you point out some of the damage has already been absorbed through the softening-up process. Those who perceive this to be a thoroughly and unnecessarily nasty Budget will generally be those who would not vote for an Abbott-led Coalition under any circumstances.

      As with the Howard government I also suspect that the Abbott government will be able to appease noisy elements on the right with some of the nastier measures in this Budget, but without having to live with the blowback of actually implementing them. The six month dole wait for under-30s, for instance, may well be watered down if not killed off by even the incoming Senate. (That said if I was someone affected by it personally, I'd be a bit nervous about putting my trust in the PUP/AMEP bloc to scrap it.)

      Delete
  6. Funnily enough, as I've mentioned before, I'm a University student living on benefits and savings in an area where newstart payments are above the national average. The problem is where to draw the line between genuine need and dole bludging. From what I see on a daily basis, a six month wait would be disastrous and would inevitably lead to minor crime (drug selling, theft etc) increasing. This is unlikely, very unlikely in my opinion to pass the senate though. Making newstart/youth allowance receivers either work for the dole, study or do volunteer work could potentially lower the amount of people in latter life (say, mid 30s) sitting on their ass all day doing absolutely nothing, selling dope to supplement their income and living in government supported rental properties. Which is something I, unfortunately, see a lot of around here. That said, making these people work earlier in life will not necessarily mean they will work later in life. What needs to be accomplished is a change in the view of what exactly centrelink is; at the moment it is tended to be viewed as a right: free money so you're not on the street. It needs to be viewed more as a safety net. How this is done without disadvantaging people is unknown to me.

    As for deregulation of the University Sector, it needed to be done and extreme left wing students really need to stop giving us all a bad name. Australia's best university ANU, isn't even in the top ten in the world and unless the government can come up with billions of dollars of extra funding per uni in the Group of Eight it will have to come from the private sector.

    Lastly, the fee for seeing a doctor is something most Australians have seen coming for a long time. Sure Howard's Liberals gave up abolishing Medicare but most Australians knew future Coalition Government's may not be so kind. I will say this though, attacking Medicare as they have was the most ideological part of the entire budget and the opposition would be smart to focus on that methinks.

    ReplyDelete
  7. 2 questions
    when does it get to a stage where recovery is impossible?
    is there some polling measure of the "soft " liberal vote?

    ReplyDelete
  8. If there is a level at which recovery is impossible when a government has even a few months of time remaining then we haven't found it yet. I mentioned that c. 37% 2PP result for the Whitlam government in July 1975. Morgan polling in early November 1975 briefly had that government leading, a few days before it was sacked.

    Apparent softness of voter support is most easily measured close to an election rather than this far out. Essential tend to start measuring it about 6-8 months out. Probably few voters at this stage would say they are sure who they will vote for next election. There's a general pattern that the Labor vote tends to appear softer than the Coalition vote when such questions are asked, but I'm not sure that appearance is entirely real. I'm not aware of any current attempts to directly measure the firmness of each party's support.

    ReplyDelete
  9. On Budget polling: Is there a connection between pre-budget polling vs post-budget polling? (As an example, the worse the expectation of the budget the better it is actually received. Or vice versa.)

    On Newspoll post budget polling history: What is the average difference between the reception of a Government's first Budget and later budgets?

    Lastly, a partisan question: On average, whose Budgets are better received by the community in both the "better for the economy" sense and the "better for me" sense?

    ReplyDelete
    Replies
    1. Howard government budgets on average had a better reception than Labor budgets (either Hawke/Keating or Rudd/Gillard), the difference being about 19 points on the net "good for economy" and 16 points on net "better for me" since 1988. Including the 1986-7 budgets for which only the "good for economy" score exists the gap on the economy score drops to 15 points. I find it hard to say though if all that is really a reflection on the party in power or the times they are in power in - as the graph shows the boom years 2004-7 had a lot to do with it.

      The first Howard government budget was among their best received in terms of the economy and one of their worst received in terms of personal impact - it has the greatest disconnect between the two scores in Newspoll history by a margin of 8 points (the 1988 May mini-budget and August budget were the closest).

      The first Rudd government budget was the best received on both scores of Labor's most recent tenure, the best received Labor budget on personal scores since that question was first asked in 1988, and the best received Labor budget for the economy since the two 1988 budgets.

      Probably with only two data points the average differences for a Government's first budgets on being elected from Opposition don't mean a lot. What is noticeable is that Howard/Costello were especially good with the election-year sweeteners since in every case their election-year budget had the best personal impact score of the three-year cycle. All up from 1988 onwards I get the averages as:

      first year of cycle personal -25.3 economic +8.6;
      second year of cycle personal -17.9 economic +11.7
      third year of cycle personal -13.1 economic +14.3

      The first question is one I think I'd need a full data set of aggregated polling going back many years to answer. I do intend to construct such a data set eventually but it will be a long time before I have it.

      Delete
    2. Thank you for that, I appreciate the effort you put in to answering those questions and I can't wait for when the time comes and you have an answer for my initial question.

      I find it somewhat strange given the Coalition's reputation for making the tough decisions regarding the budget (whether this is actually true or not I'm unsure. That said Rudd/Gillard governments did nothing to either prove their economic prowess nor deny the narrative the Coalition were creating) that overall their "good for me" rate is higher than Labor's is.

      Very informative and very interesting.

      Delete
  10. Is the Newspoll due out tonight (Sunday) Kevin?

    ReplyDelete
    Replies
    1. Yep first figures are already out on Twitter and I'll post them in the article soon. 55-45 to Labor. Newspoll usually comes out Monday night for Tuesday morning but sometimes Sunday.

      Delete
  11. Wow! What an eventful evening that was. Cries from the left over on PB would have you believe this is Australia's worst ever budget and that 55-45 30 months out from an election is 'dire'. Hmm.

    As I've mentioned this is by no means a horrible budget considering Hawke/Keating and Hoqard/Costello and their respective first budgets.

    I think, if this shows anything, it shows that Labor is starting to reconnect with Australians, it shows the power the media has (when "horror budget" is being screamed on every channel) over public opinion and lastly it shows that the Coalition don't know how to sell a budget.

    That said, I'm pleased my prediction of a 1-2 point dip in 2pp. That said, I did not see ratings about the economy so low given the debt and deficit story Abbott and co wrote.

    Oh, a stray thought: maybe this shows just how dire Abbott's popularity really is. There is no wiggle room for the Coalition with Abbott in charge. Turnbull for PM? lol

    ReplyDelete
  12. "The better PM figures are 44-34; one comparison is 45-33 to Abbott over Gillard just before Abbott was rolled."

    I think you meant to type "Just before Gillard was rolled".....

    ReplyDelete
    Replies
    1. Ta, fixed. Some might say that was retrospective wishful thinking, but I couldn't possibly comment ...

      Delete
  13. I mean, Essential could be right, I guess. I doubt it, but it could be.

    So, if Essential is wrong, what could be causing it? I understand that Essential is a rolling average but even still, by now, something should have shown. So, what do you believe is happening in Essential land?

    ReplyDelete
  14. It is the same thing as happened about four distinct times in the lead-up to last year's election: Essential tended not to fully pick up most of the dramatic shifts shown by other pollsters - even allowing for the two-week rolling average - and sometimes even moved in the opposite direction. While Labor were obviously falling back off the Rudd bounce well into the campaign, Essential still had them rising and even when it picked up some fall it was not as much as other pollsters. Since the election we've seen two sharp shifts and Essential took two months to pick up the first and has yet to register the second.

    There are problems with online panel polling in terms of trying to get an always-representative sample - these are worse than with robo-polling let alone landline polling - but I don't know why this curious stagnating behaviour should be one of them. It's possible there is some kind of long-term bias towards the median result that is caused by an increased risk of re-polling the same person or by some kind of scaling assumption, but I really don't know what is causing it.

    It also seems to affect voting intention a lot more than their other questions.

    ReplyDelete
    Replies
    1. It's strange though, at least when it comes to state polling Essential seems right on the money. When I say that I mean it's corroborated by other pollsters case in point QLD and Galaxy.

      Seems strange to me that state polling is accurate but Federal is not. Any reason for this? Do they use different methods in State polling than they do Federally? Or, is there another reason?

      Delete
  15. I find that Essential are only sometimes wrong federally. Of their 29 readings since the election, 17 have been within a point of my aggregate at the time. However of the 12 that deviated from it by more than that, 10 favoured the Coalition, and eight of these happened over runs of four consecutive weeks (meaning at least two completely distinct samples of c. 2000 voters each).

    Not enough data to say for sure whether Essential's state polling is less prone to going off track. However, one thing I notice is that the Essential state polling seems bouncier, even considering the smaller sample size.

    ReplyDelete
  16. How will Nielsen's departure effect your weighting? Will you simply remove it or weight the others slightly higher to compensate for the departure?

    ReplyDelete
    Replies
    1. It won't initially have any effect at all. My model is already designed to cater for the possibility that any given poll might have recent data, or not have recent data, at any given time. If a poll has no sufficiently recent data then its weighting in the current aggregate is zero.

      I am a bit concerned that with Galaxy switching to a hybrid polling method we'll have a dearth of pollsters operating by stable methods that were successfully validated at multiple previous elections. Effectively Newspoll becomes the sole anchor for assumptions about the behaviour of the others. Could get messy.

      Delete