Tuesday, May 16, 2017

Poll Roundup: Are Budget Bounces Ever Real?

Aggregate: 52.7 to Labor (+0.2 in two weeks since last completely pre-Budget reading)
Labor would easily win election "held now"
Voter budget ratings are historically about average

Commentary around the 2017 federal budget has been even more focused than normal on one of the Australian beltway polling obsessions, the idea of a "budget bounce".  The Coalition's moves on the National Disability Insurance Scheme and the announcement of a bank levy, in particular, were seen as bold moves to seize the middle ground from Labor in an attempt to make the Government popular again.  Some even saw this as a threat to Bill Shorten's leadership of the Labor Opposition.

That at least a temporary bounce was widely anticipated (and also widely seen as the aim of the game) is one thing, but another has been the common blight of house-poll myopia in which commentators from particular stables obsess over the single-poll-to-single-poll changes in the poll their network commissions, while ignoring both the aggregated cross-poll trend and even the longer history of their own poll.  In the case of Newspoll, it didn't help the Government that their previous result (48-52) was a shade on the friendly side of the recent run of polling, thus setting a pretty high benchmark for any budget bounce to be assessed by News Ltd scribes from.

However the analysis from the Fairfax stable has been ridiculous, with several pieces trumpeting the turnaround from the previous Ipsos while ignoring all the other evidence.  The previous Ipsos is in fact ancient (at seven weeks old - eleven other polls have walked the planet since that dinosaur), but what's more, it was an outlier.  For instance, Laura Tingle (who I single out because I reckon she knows better, whereas some of her colleagues are hopeless cases) declared:

 "The Fairfax/Ipsos poll will certainly be grounds for immense relief within the government: a decisive turnaround in the trend and a much-needed lift in the Coalition's primary vote, though obviously there is further room to travel."

It's much more likely that 80% of said government are too fixated on Newspoll results to even notice that the Fairfax/Ipsos poll exists, and that the 20% who do notice it remember that the last poll was that really odd one that had the Greens on 16% for the second time since the election, five points higher than any other poll has had them in that time.  Had Ipsos produced another shocker, even the 20% would have written it off for good.  

The Budget polls: voting intention

So far we've had five post-Budget polls; I'll add updates for any late-breakers.  Ipsos and Newspoll both produced 53-47 to Labor by last-election preferences, compared to 55 seven weeks ago for Ipsos and 52 three weeks ago for Newspoll. I aggregated both of these at 53.1 to Labor after including the primaries.  Essential stayed at 54-46 and was aggregated at exactly that.

The confusing ones to get a handle on have been the two ReachTELs (separate samples with quite similar primary votes) released by Sky News and Seven on Friday night, and both taken on Thursday.  Not only is ReachTEL's recent formatting of its polling results as if it were designed to create as much confusion and as many pointless differences with other polls as possible, but media sources then make the problem worse by reporting only basic details like the two-party preferred result, making it almost impossible to reconstruct what is actually going on.

The reported 2PPs of these polls were 53 (Sky) and 54 (Seven) to Labor, but these were respondent preferences, which have a historic tendency to usually skew to the ALP compared to what actually happens at elections.  After I was unable to reconstruct the 53 from the published data by normal means, the company were kind enough to confirm, as I suspected, that they are asking National Party voters for their preferences, and distributing those preferences between Labor and Liberal.

Now whatever might be said about the merits of respondent preferences, if you are going to do them at all, then this is wrong.  The reason it is wrong is that a quarter or so of Nationals voters will probably say they would preference Labor ahead of the Liberals, based on how Nationals preferences tend to split when given the chance.  But the vast majority of people who actually vote for the Nationals live in electorates where the Nationals finish first or second and hence those preferences are never distributed.  At the last federal election, the Nationals received 624,555 votes, but only 72,350 of those votes were distributed between Labor and Liberal on a 2PP basis.  Only 15,967 National votes wound up credited to Labor on the final 2PP, meaning that 97.5% of Nationals votes actually counted towards the Coalition's 2PP.  Labor would not really get more than 0.1 2PP points out of Nationals voters come election time, but ReachTEL's method could be giving them close to a point out of the Nationals polls.

As a result, the ReachTEL federal preferences in these polls are skewed to Labor and should be thrown away, though its possible this skew might cancel out any anti-Labor skew in their primary vote polling from time to time.  I was finally able to estimate last-election 2PPs from the Sky poll at 51.5 to Labor and the Seven poll, based on incomplete data, at 52.4.  

With all of that out of the way (phew!) my new aggregate reading is 52.7 to Labor, a change of 0.2 in Labor's favour since the last Budget-poll-free end-of-week reading from two weeks ago.

One thing worth keeping an eye on - One Nation has slumped from 10% to 6% in Essential's primary vote readings.  The party also declined from 18% to 15% within Queensland in a Queensland federal Galaxy at the end of April.  At this stage other polls have not yet picked this up to such a degree.

A History Of The Budget Bounce Yeti

In a general and excellent piece about the futility of most attempts to deliberately "bounce" the polls, Peter Brent recently wrote that "Budget bounces are about as common as the Yeti, but Turnbull’s belief in his own abilities to move the immovable remains undiminished."

As there have been no reliable sightings of a wild budget bounce in the field for many years, some have even questioned if the thing exists at all.  David Crowe weighed in with a piece (data tables here) which showed that he believes the truth is out there, and has plaster casts of footprints in the snow to prove it.  At least he is referring to the evidence, which is more than can be said for most playing this game.

Unfortunately if we look closely at these sorts of methods, it all shows just how shadowy Yeti-hunting can be!  Measuring from one Newspoll to the next to try to find a bounce is fraught with peril.  Firstly, random variation alone should have caused an average shift from one Newspoll to the next of about 1.6 points, even if the Budget caused no change in voting intention and nothing else did either.  Taking rounding into account as well, now and then there would have been three or four point shifts from "pre-Budget" to "post-Budget" even if the Budget had shifted no voting intentions and nothing else was happening either.

Secondly, at the same time as any Budget there are often other things going on that influence voting intention, and sometimes those things can overshadow it.  Thirdly, measuring a bounce by just considering the polls right before and the polls right after the Budget is risky anyway, because there is usually advance information or speculation about the Budget floating about in the few weeks before it.  The poll before may thus itself be influenced by the Budget contents, or it could reflect inaccurate fears or hopes about what the Budget might look like.

Here's a look at some of Crowe's examples to see how easy it is to see a Yeti when it isn't there.

1996: The Coalition's primary rose from 47 (last pre-Budget poll) to 50 and then stayed around that mark, so supposedly a budget bounce.

The context: The 1996 budget was considered tough but good for the economy by voters and hence the Howard government was not punished for it, but they didn't get a big bounce for it either.  The 47 was an atypically low reading, which was probably sample noise or else may have reflected mild apprehension about the new government's Budget.  Here's the run of Coalition primary vote readings either side: 51-49-47-(budget)-50-49-50.  

2000: The Coalition's primary rose by five points and was still up four three months later.

The context:  Here's the run of Coalition primary vote readings either side: 42-44-40-(budget)-45-40-39.  Any immediate bounce was a sugar hit that completely vanished the Newspoll after - but more likely what we see here is mostly volatility.  Whether the Coalition's good run of primary vote results two and three months later had anything to do with the Budget is pure speculation. 

2009: Labor's primary rose by four points.

The context: Again just poll-to-poll volatility, mostly.  Here's the run of Labor primary vote readings either side: 47-47-42-(budget)-46-43-41.  (The 41 was influenced by "Utegate".)

If anything, these examples support the idea of a pre-Budget dip rather than a Budget bounce!  

In Search Of Budget Bounce ...

I thought I'd have a look for this monster too, but I found that the more you look for it, the harder it gets.

Firstly to get serious about this, it's necessary to use multiple polls, and not just a poll-to-poll reading.  So using Newspoll, I tried comparing the rolling average at the end of the month before the Budget, with a straight average for the four polls immediately afterwards.  (I used the end of the month before the Budget to avoid the issue of the immediate pre-Budget poll being sometimes contaminated by pre-Budget speculation, depending on the time gap.)

Even so, often over such a time scale there are things that have nothing to do with the Budget.  So I thought that if some Budgets produce polling bounces, it would make sense that they would correlate with the ratings given to voters by the Budget.

However, that corellation is weak.  Here, for example, is a graph showing the estimated actual change in polling for the last 30 budgets, as it relates to a voter "budget rating" derived from the Newspoll budget polls:

It's a noisy relationship (to be expected) but the relationship between the two explains 27% of variation, which isn't bad ...

...except that nearly all the explanative power comes from the 1993 shocker in the bottom left hand corner, and once that one's removed, the r-squared drops to 0.06.

I deal with a lot of messy data and I'm willing to get excited about 6% of variation explained if I have 2900 or even 290 data points.  But not if I have just 29 - one stray data point and it's dead.  And those poll-to-poll changes cited by David Crowe?  They have no relationship to what voters thought of the budget at all!

A few points we can draw from this exercise:

* Budgets never create large immediate bounces that have any staying power.  I couldn't find any polling change much above two points, and the chance that those bounces were really that big (as opposed to a bit of bounce and a lot of noise or other factors) must be low.

* Budgets don't tend to help government polling in the short term on the whole. The average result was a loss of 0.4 2PP points in the wake of the exercise.  That's not to say Budgets themselves cause such a small average loss, but they at least don't tend to cause gain.

* Unless you are the Howard Government, perhaps.  The Howard Government on average gained 0.4 points in the wake of budgets, while other governments on average lost 0.9.  Most likely this just reflects that the economy was that government's perceived strong point and was going well during its reign, and that time spent talking about the economy was time not spent talking about everything else that from time to time would get the crew in trouble. 

The largest alleged bounces in my sample (Prime Minister/Treasurer noted) are:

2001 Howard/Costello, 2.4 points
2012 Gillard/Swan, 2.3 points
2004 Howard/Costello, 1.9 points
1991 Hawke/Keating, 1.5 points 
1988 (May) Hawke/Keating, 1.2 points
1995 Keating/Willis, 0.9 points
1998 Howard/Costello, 0.8 points
1988 (August) Hawke/Keating, 0.7 points
1996 Howard/Costello, 0.6 points
1999 Howard/Costello, 0.7 points

But 2012, 1991 and 1995 are fakes.  These budgets rated badly with voters and there was other stuff going on in these cases - leadership spills in the first two and an Opposition coming off a bounce for a leadership change in the third.

If there's anything to see here it's that the Howard/Costello team were pretty good with the pre-election sweeteners, since 2001, 2004 and 1998 (all years they were re-elected in) all appear in the hall of fame and all could have been real.  But there were other things going on in 2001 that suggest that even in that year, the Budget wasn't the full story.  

The Yeti probably has existed, but most sightings of it are bogus, and proving it has been seen in any given year is near-impossible.  

This Year's Budget

Here's where this year's budget shapes up in terms of voter perceptions of its economic and personal impacts, compared to other recent and notable budgets (labelled):

The 2017 budget has eaten the dot for the 1989 budget, which had exactly the same combination of net personal and economic scores.  See also the full Newspoll tables here.

The shift to new Newspoll methods last year appears to have produced an increase in the uncommitted score both for the economic impact question and for the question about whether the Opposition would have done any better.  So while belief that the Opposition would do better is at an eight-year low and belief the Budget is bad for the economy is at a nine-year low, this seems to be because more people are choosing the don't know option as a result of the change away from live-interview methods.  

The views of third-party voters about this Budget seem to be similar to those of Labor voters.  I estimate the third-party voters as breaking 21-35 on economic impact (22-39 for Labor voters), and 13-50 on personal financial impact (11-59 for Labor).  Of course they are more sceptical than Labor voters (I estimate 25-43, vs 64-16) about Labor doing better.

There are a great many polls around with specific questions on budget measures - these can be found on the websites of individual pollsters as usual, save for the Newspoll ones mentioned above.  At this stage, because of the amount of work I have on, I don't have time to summarise and unpack them all, but there are few surprises.  Voters mostly approve on specifics of this Budget when asked, but are cool on the whole package, and it isn't changing their vote anytime soon.  


Lastly a quick run around the latest leadership figures.  Newspoll had Malcolm Turnbull up five points to a net rating of -20 (33-53) and Bill Shorten down two to -22 (32-54).  That's twelve in a row at -20 or worse for Turnbull (again, Abbott never had more than seven such) but would give some hope that he might get out of that range soon.  Turnbull leads 44-31 as "better Prime Minister", again a large lead given the 2PP deficit.

Ipsos' leadership ratings tend to be benign generally, and for whatever reason they are particularly kind to Malcolm Turnbull.  They now have him at a net +1 (45-44) and Bill Shorten at -5 (42-47).  Shorten is up 13 points since the March Ipsos, but only six on Newspoll in that time.  Here, Turnbull leads 47-35 as preferred leader.

ReachTEL - which doesn't have a natural skew to incumbents in its preferred leader polling - has Turnbull preferred to Shorten 52:48, the closest the two have so far been.  ReachTEL has Turnbull on a net -23.8 personal rating - easily his worst to date, but that's no surprise given this is their first reading since the election.  Shorten is at -21 and was lower than that for most of the nine months through to March 2016.  

ReachTEL also has Turnbull still the preferred leader of the Liberal Party with 38.1% support, to Julie Bishop 28.5%, with Tony Abbott now up to 17.3% and Peter Dutton recording his first double-figure score from anyone so far, at 10.5.  Bill Shorten, however, is in third place on the ALP list with 26.0%, behind Tanya Plibersek (30.7%) and Anthony Albanese (26.2%).  At least he has a modest lead among his own party's voters.  While some would wish polling was close enough to pressure Shorten's leadership, this doesn't look like becoming the case anytime soon.


  1. Thanks Kevin, Very informative, as usual. Your commentary is plainly far superior to any produced by the MSM.

  2. Granted, the idea of a big budget bounce, or any budget bounce, has been a bit of a tradition in Australian politics with little to no evidence behind it.

    That said, politicians and the public seem to believe in it. The fact that there was no bounce to a very 'kind' first year budget will worry those in Canberra.

    I fear for Libspill 2.0 and Prime Minister Dutton now... As he is seemingly the only one putting his hand up.

    Out of curiosity, that newspoll was also conducted prior to the budget, wasn't it?

    1. The current Newspoll (53-47) was entirely post-Budget (Thursday to Sunday of budget week).

      I wonder about the whole thing of removing Turnbull before the next election in terms of what he would actually do in response and how much damage he would be able to cause if he wanted to. There was a time when he could have silenced any thought of it simply by threatening to start his own party - that time is past I think, but still ...


The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.