1. The AWU "scandal" involving Prime Minister Julia Gillard's former links to a union-related "slush fund" recently dominated the news cycle and resulted in accusations of criminality by Opposition members including Tony Abbott against the PM.
2. Current polling shows that a movement to the Labor Party over the course of several months appears to have, at least temporarily, halted and probably slightly reversed.
3. Although the AWU debate may have contributed to the slowing of momentum towards Labor, there is insufficient evidence that it has caused damage to Labor's vote.
4. Questions surrounding the AWU affair by three pollsters show varying levels of design quality, and many are unsatisfactory.
5. Those polls that have produced the worst results for the Prime Minister are typically the worst designed, while the better designed polls show that the PM's handling of the issue is fairly well approved of.
6. At the height of Parliamentary debate about the issue, both the Opposition (for its handling of the matter) and its leader recorded very bad ratings, without pro-Coalition voting intention being affected.
7. The impact of the issue upon the standing of the Opposition Leader is not yet clear because of insufficient evidence.
8. Comparisons between this issue and the Ozcar/Utegate blunder by Malcolm Turnbull in 2009 are simplistic as there are many differences in the nature of the issue and the existing popularity of those affected.
In the last two sitting weeks of Federal Parliament, the last week especially, the news cycle was dominated by the so-called AWU "scandal". The rather complex "scandal" features a movable feast of accusations about the level and nature of Prime Minister Julia Gillard's involvement as a then-solicitor in the early-1990s establishment of the "AWU Workplace Reform Association", a re-election "slush fund" for union officials Bruce Wilson and Ralph Blewitt. Wilson and Gillard were in a romantic relationship from 1991 to 1995. The "scandal" also involves the subsequent purchase of a house in Blewitt's name. Wilson and Blewitt have been the subject of police suspicions of massive fraud involving the Association, but while Blewitt has confessed to involvement in fraud and is seeking immunity from prosecution, Wilson has denied guilt, and to this stage no-one has ever been charged. Far more detailed summaries of the issue can be found elsewhere for those attempting to absorb it all. Good luck; even Bob Carr is struggling.
This issue has intermittently affected Gillard's political career for the past seventeen years. For instance it appeared in a "heavily lawyered" article by Glenn Milne just before the 2007 election, and Milne was back for another go in August 2011, in a story that was very quickly replaced by an apology and retraction. Andrew Bolt had highlighted the matter at that time in an article entitled "A tip on something that may force Gillard to resign" (which was taken down, together with the Milne piece). It's a sign of how much the stakes have been raised since that Bolt's argument included these words:
Available Polling Evidence
- Galaxy (released 1 Dec, voting intention and AWU affair - also covered four-way preferred PM polling)
- Essential Report (released 26 Nov, voting intention and 3 Dec, voting intention and AWU affair)
- Roy Morgan by phone (conducted 27-29 Nov, voting intention, leader ratings, AWU affair and preferred leader of each party)
- Roy Morgan face-to-face (conducted 24-25 Nov and 1-2 Dec, voting intention)
As everyone should be well aware by now, single federal polls are in most cases useless when it comes to indicating whether voting intention has shifted since similar polls a few weeks earlier. Mostly, movements from poll to poll are within each poll's margin of error, meaning that a shift that might actually be quite large cannot be distinguished reliably from random noise in a single poll.
As the recent US Presidential election demonstrated in spades, subtle changes in voting intention are best monitored through aggregators which compile polling from a range of different sources, often weighting according to evidence of the reliability or party skew of different pollsters.
I am aware of at least four aggregators now operating in the Australian pseph scene:
Bludgertrack (displayes on the right-hand side)
Mark the Ballot
(Note: it appears the Insiders Poll of Polls is another aggregator, but one that is only updated monthly.)
Of these Pollytrend is the longest-running while the last two are very new. These aggregators vary in the sophistication and the methods of their modelling, in how they weight polls and in which polls they include (eg Poliquant excludes Morgan face-to-face polling while Bludgertrack gives it a low weighting). But the overall picture is rather similar. A steady trend to Labor that had been running since at least July seems to have at least flattened out since around early November, and probably (all now say so) moved slightly in the Coalition's direction. We are now, most likely, somewhere in the range 47-53 to 48-52. While this would produce a Coalition win if an election was held right now, in view of the past history of government resilience it really means nothing by itself, and anyone who thinks they know who will most likely win the next election based on current polling is just kidding themselves.
All the same, the size of the recent swing back to the Coalition, if it is real at all, is still quite small (about a point maybe) and there is no clear evidence that the AWU saga is the cause of it. Indeed, the point at which the momentum in Labor's favour stopped seems to have been before the latest ramping up of AWU claims. It is possible that the AWU saga is contributing to either the loss of Labor momentum or even some small move back to the Coalition, but the evidence is unclear and if there is such an effect, it is not presently a big one.
It's true that sometimes events can take a while to resonate in public discussion to the point that they affect polled opinion. But in the case of an event failing to greatly affect voting intention, a more common explanation is that the event was mostly just Parliament House and press gallery hot air that voters were not all that interested in.
(As a side-note, it is so obvious now that good aggregators are more meaningful than individual polls that Gallup, ignominiously the worst pollster of the 2012 US Presidential election, has claimed that since it is cheaper and easier to aggregate polls than do your own, the rise of aggregators like Nate Silver's could lead to a "tragedy of the commons" style decline in actual polling, thus rendering aggregators less useful.
Greatly moved by this concern, or more likely just plain lazy, I have refrained from inventing my own aggregator, though I am almost tempted to take the Gallup sulk to its ultimate conclusion by inventing a super-parasitic meta-aggregator, which should be even more accurate than the aggregators and require almost no work at all!)
The Morgan Poll on AWU
Roy Morgan were the first to rush in (where angels dare to tread?) with an attempt to frame questions on the AWU saga. Their polling asked these questions:
“Recently there has been much discussion and publicity about the so-called AWU 'slush fund' scandal from the 1990s when Prime Minister Julia Gillard worked at Slater & Gordon and was the legal adviser to AWU members Bruce Wilson and Ralph Blewitt. Before today, were you aware of this?”
Those who were at least vaguely awake were then asked:
“From what you know or have heard, do you believe Julia Gillard was aware that the AWU 'slush-fund' was illegal when she resigned from Slater & Gordon in 1995 or do you think she only found out about it later?”
“Do you approve or disapprove of the way Prime Minister Julia Gillard has recently explained her involvement in the AWU 1990s 'slush fund' scandal?”
“If the AWU 1990s 'slush fund' scandal allegations against Julia Gillard are true, do you believe Prime Minister Julia Gillard should resign, or not?”
The Morgan Poll - as with many of said company's releases - is rather confusing (William Bowe called another of their recent efforts "headache inducing") because its primary results are for all Australians aged 14+. Given that 85% of the polled 14-17 year olds are not aware of the AWU issue, that the number of 14-17 year olds in the sample is less than 50, and that most of them will not vote next election I cannot see any use in including them in these figures at all (with apologies to my doubtless vast teenage audience!) It is necessary to look at the third table under each question to see the results for all electors.
The questions are far from perfect. The second question asserts that the "slush fund" itself was illegal, and this could be taken as a reference to either its establishment (in which Julia Gillard had a role) or its operation (in which she did not). The second question also asks about Gillard's knowledge of the fund at the time of her resignation from Slater and Gordon (and the end of her relationship with Wilson) but it might have been more useful to ask what Gillard thought about the fund at the time of its establishment and in the years before 1995.
The fourth question is the most problematic because it refers to whether Gillard should resign "if the [..] allegations are true", but it does not say which specific allegations the question refers to, nor whether the question means some of them or all of them. Given that 26% of electors were not even aware of the "scandal" at all, it's very likely that many electors were not aware of the full range of allegations - which even shifted during the polling period. So a "yes" on this question might mean that Gillard should only resign if she was fully and knowingly complicit with and financially benefiting from the establishment of a fund for deliberately corrupt purposes (of which there is no evidence whatsoever) or it might mean that she should resign if the worst she was guilty of was poor judgement about personal conflict of interest and failing to open a file. My suspicion is that most respondents would give Gillard the benefit of the doubt and only consider whether she should resign if at least some of the worst of it was true, and that most respondents would know that the worst available accusations were very serious, but the question is unhelpfully imprecise. Another problem is that some respondents may have already concluded that the accusations are definitely not true, so how do they answer a question based on a premise that they think is false?
The most useful question is the third. The third question shows a 37-28 approve-disapprove split on Gillard's handling of the matter. Compared to the Morgan leader approval ratings for the same period, this shows that Gillard's net approval rating over AWU (+9) is higher than her net approval rating generally, or netsat (-6). If all those not aware of the saga are excluded, Gillard's net approval rating over AWU among the remainder rises to +12. Gillard's net approval over AWU (with those unaware of the AWU issue removed) is higher than her overall netsat among supporters of all parties, but this is especially so for voters for Others (+24 vs -24), voters who could not say what party they support (+41 vs -6) and Coalition voters (-45 vs -69).
On the resignation question, 43% of electors gave a "yes" response, whatever that means, to 25% for "no". It's notable here that a significant minority of Coalition voters (19%, or 24% of those with a view about the issue) didn't consider that the matter (as they understood it at that time) was even potentially resignation territory.
The Galaxy Poll on AWU and the Rule of Divide By Five
The Galaxy Poll (link may expire; more detail here) asked this question:
"Julia Gillard has been embroiled in allegations surrounding an AWU slush fund. In your opinion, has Julia Gillard been completely open and honest in responding to the allegations, has she been economical with the truth or do you think she has lied?"
William Bowe has already taken this question to the cleaners , arguing that "Particularly where complex or half-understood issues are involved, choices like this are known to activate the strategy of “satisficing” (“choosing the easiest response because it requires less thinking” [..]" Another point against it, made by my partner before I potentially thought of it and therefore duly credited, is that there is a general cynicism about the idea that politicians would ever be "completely open and honest" about anything and that alone could cause some voters to dismiss the notion out of hand. A further issue is that the word "completely" might deter even those who believed Gillard to be almost completely open and honest from choosing that answer. So the overall figures (21%, 31%, 31% respectively) are not very useful.
There is also a question about whether Gillard should provide a "full account of her involvement in the AWU "slush" fund through a statement in parliament." (This is the Australian's quote; I have not been able to find the exact form of the question.) William Bowe calls this "an over-elaborate proposition that feels tailored towards eliciting a positive response" and my own observation is similar; the question carries a degree of loading in that it hints that Gillard's accounts thus far have been less than full (a leading assumption). I don't see the strong split in favour of this question (60-26) as therefore meaning a great deal. The result can also be taken along the lines that ideally the PM should do such-and-such, but people won't necessarily care if she does not.
The third question is the one of most interest to me because it is one of those questions where voters are asked if an issue makes them more or less likely to alter their vote. Such polls should always be treated with extreme caution. The Australian's Samantha Maiden writes:
"According to the Galaxy Poll, 26 per cent said they were less likely to vote Labour and 9 per cent were more likely to vote Labor, suggesting a net loss for Labor in voter intention."
But the way a net loss in voter intention is measured is by looking at polling of actual voter intention, not by inferring from polls of this sort. The voters who say their intention might change in response to a question like this are often locked-in voters for either side who delude themselves that the issue makes any difference to their vote (or if it does, a shift in probability of voting the other way from .1% to .08% is hardly interesting.) Even among voters who correctly claim an issue makes them significantly less likely to vote a certain way, that change of likelihood doesn't necessarily mean a swinging vote, and we don't know whether the change in likelihood is the same per voter on one side compared to the other. Bowe provides a neat list of issues also supposed to have elicited similar or stronger responses (the budget, power prices in WA, poker machines, the Labor leadership switch), and I can think of other examples like the Bell Bay pulp mill and same-sex marriage. Typically, the issues that feature in these sorts of polls do not have that much impact - and the issues on which a minority will swing very sharply don't feature in these sorts of polls, because the parties know this, and avoid policy shifts on such issues like the plague.
I have a rule I call the divide by five rule for interpreting polls like this one. The rule is that in virtually all cases where a poll question like this is asked, divide the percentage who say they are more/less likely to vote a given way by five to find the maximum percentage who will actually switch their voting intention. (Its most recent predictive success was the size of the debate bounce received by Mitt Romney in defeating Barack Obama in the first Presidential debate!) In most cases the percentage who actually switch, especially over a long period of time, is much lower. The maximum three-point swing suggested by the rule in this case would still be a very big deal if it happened, but there is no evidence based on voting intention polling that anything like that is happening.
The Essential Report Poll on AWU
The current Essential Report poll firstly shows that Essential's respondents are more aware of the issue than Morgan's, with only 15% either unaware of it or unaware of how aware they are! Essential's use of an online panel as opposed to phone polling may contribute to this difference, as may the timing of the two polls. Of those aware, Essential gets it right by asking simple approve/disapprove questions about the role of Julia Gillard, the Opposition and the media. The poll gives Gillard a net +4 approval rating on the issue (39-35), which is not quite as high as Morgan's, but the Opposition a net disapproval rating of -29 (20-49). It is especially notable that while there is hardly any level of "ratting" on this from Labor supporters (7% thinking Gillard has handled the issue poorly and 4% that the Coalition has handled it well), the level of "ratting" from Coalition supporters is much higher (17% praise Gillard's handling and 21% condemn their own side's).
A further question on whether the issue has affected voter views of Gillard shows exactly and neatly why the Galaxy question on whether the issue would move votes (and other questions of this kind) are so useless. Although 38% say they have a more negative view of Gillard, compared to 9% saying they have a more positive view, most of those claiming to have a more negative view are Coalition supporters (64%). Only 9% of Labor supporters have a more negative view, including a negligible 1% with a much more negative view - and given the incentive for Rudd-supporting Labor voters to talk down the Prime Minister, these figures are really exceptionally low. Furthermore, the disparity between those claiming a more negative view and those claiming a more positive view is to be expected from the nature of the issue - if the PM succeeds in clearing her name in the public mind, then that just puts her back where she was when the issue was last off the radar.
The Essential poll, then, as the best designed of the three polls, fails to show any evidence that the Prime Minister is being damaged by the matter at this point. And I take its findings to be strong evidence that the slight trend back to the Coalition in polling generally is probably not caused by the AWU scandal.
It is notable that voters who claim a large degree of knowledge of the issue tend to be more negative about the handling of the matter by all of the PM, the Opposition and the media, with the largest difference for the PM. These voters don't split strongly by party and there is not enough information to tell whether it is the well-informed Labor or Coalition voters, or both, who are creating this pattern.
Finally, there is no evidence in the Morgan poll of leader approval to suggest Gillard's Prime Ministership has been damaged by the issue. Her Morgan net satisfaction rating in the last days of parliament was -6 compared to -11 on September 17-20. In the same period her Newspoll net satisfaction rating has been static, but the most recent Newspoll was released before the last week of Parliament.
What About Abbott and The Coalition?
Given that the Essential result shows strong disapproval of the Coalition's form of the issue, it also makes sense to ask whether the Coalition's pursuit of the issue is blowing up in their faces. There is reason to suggest it might, including Abbott's poor attempt to back up his accusations of criminality and Bishop's bad day on the attack trail and unnecessary meeting with Blewitt. (My impression of Julie Bishop's form has been that she thinks this is just like being a lawyer again, and has lost sight of the way that in politics making weak accusations and meeting with dubious figures can count more against you.)
The voting intention evidence does not suggest that the Coalition's support levels are being affected. After all, a trend that was running against them for months has either stopped or reversed, at least temporarily. However, the situation for Tony Abbott and Julie Bishop might not be so rosy.
The Morgan Poll covering the most intense period of the scandal showed Abbott's Morgan net satisfaction rating at -35, adding Morgan to Newspoll, Essential and Nielsen in giving Abbott his worst ever rating in November. The change since September for Abbott was greater than his rolling average Newspoll change in the same time, although this could be down to sample variation. What would be especially concerning to Abbott fans here is that the 7-point drop in Abbott's overall standing since the September Morgan is coming not from locked-in Labor supporters, but mainly from those voting for the Coalition (a 16 point netsat drop from +22 to a terrible (for voters of his own party) +6, with approval down by 8 and disapproval up by 8) and those voting for Others (approval down 15, most of which shifted to neutral.) Gillard's netsat among Labor voters, by comparison, is now +59.
The Morgan Poll on preferred leaders is also nasty news for Abbott. While only 15% of voters overall think he should be leader (cf a spectacular 50% for Turnbull and 18% for Hockey), the most important figures are those dealing with the attitudes of supporters of the Liberal Party. In this case Turnbull has opened up a large lead (40% to 28% for Abbott and 22% for Hockey). This appears quite a big shift - although it is only borderline in statistical significance in view of the small sample sizes of each poll - since September, when Liberal supporters still marginally preferred Abbott (33%, Turnbull 32%, Hockey 21%).
It's also notable that at a time when Julie Bishop was front and centre of the Opposition's case against Gillard, and hence in a fine position to push her leadership credentials if any exist, Bishop's own rating as preferred Opposition Leader continued to slumber in single digits (and actually went meaninglessly backwards).
Not too much should be read into this Morgan Poll, a single result based on a sample of only 523 electors at the peak of the "scandal", albeit by the more reliable and neutral of Morgan's two polling methods. All the same, given that it is consistent with the findings of Essential, it does seem that Abbott's pursuit of Gillard over AWU has, in the short term, done her no damage and himself, at best, no favours. If he has done himself further damage, we may find out in next week's Newspoll.
(A widespread theory has been that Tony Abbott arranged for the AWU matter to be reheated to take pressure off his own struggling leadership and reduce potential for an end-of-year spill. Given that there was no sign that anyone else was in an organised position to mount a challenge, I think this one can be dismissed out of hand.)
Why This Isn't The Second Coming Of Utegate
Although there is ample evidence that Ralph Blewitt is an even more dubious associate for the Coalition than Godwin Grech - and that they would or should have known this all along - Gillard's attempt to compare this saga with Ozcar/Utegate and suggest it would backfire to the same degree seems more hopeful than realistic. It would be a source of frustration to the PM that while evidence that the public strongly dislikes the Opposition Leader continues to grow and grow and grow, his party nonetheless remains, for now, in front.
The Ozcar/Utegate saga in 2009 had the following polling impacts:
* Upon discovery that the critical "smoking email" was clearly forged, Malcolm Turnbull's Newspoll netsat dropped by 40 points from +7 to -33, the largest fall in a single poll by any leader in the history of Newspoll.
* A trend to the Coalition in voting intention, which had seen them gain about three points in two months, reversed, with about half the gain disappearing in the next two months and no recovery until Malcolm Turnbull was replaced.
* Turnbull's ratings did rebuild slowly (back to -14 by five months later when he was removed) but never fully recovered.
* Kevin Rudd's ratings, which had declined by 24 points in two months, reversed direction and improved steadily over the next few months.
I do not believe the polling impacts of AWU on Abbott and the Coalition will be as large, for the following reasons:
1. The chief allegation in the Ozcar/Utegate affair was proven to be false. In the AWU case there are multiple allegations, some of which are difficult to prove true or false.
2. The AWU case involves multiple allegations. A voter may well believe that the worst of them are false while still believing that they show enough evidence of suspect judgement on the younger Gillard's part to at least give the Opposition some degree of leeway to raise them.
3. The Ozcar/Utegate affair represented a major failing by a to-that-point reasonably popular Opposition Leader. It was massively damaging to his leadership "brand", although he has since recovered. In contrast, Abbott is already very unpopular, and cannot possibly fall as far.
4. The Ozcar/Utegate affair represented an unsound attack on a popular and well-regarded Prime Minister. Julia Gillard is not so popular and well-regarded and is hence more likely to be seen as, to some degree, fair game.
5. Gillard has recently made strong attacks on Tony Abbott, for instance during her now-famous sexism speech. There may be some perception that while the Coalition has made a fist of its pursuit of her over this issue, and gone too far, that she is not immune to going a bit over the top herself.
That is not to say that the issue cannot impact on Abbott's leadership prospects over the next few months.
Many issues are greatly hyped by media commentary that turn out to be little more than "parliamentary theatre". The impact of the AWU issue could very well yet change, for instance if major new evidence comes to light (which seems unlikely given the extent to which it has now been rehashed). However, it is not yet clear if it has contributed anything to the next election other than some entertainment, minor hysteria and embarrassment for both current leaders along the way.
Update (Monday 10 Dec): two polls out today by Essential Report and Newspoll both show the Coalition with 54:46 leads. Lately Newspoll has tended to skew slightly to Labor and ER slightly to the Coalition, and I suspect the Newspoll 54:46 is a bit favourable to the Coalition, by Newspoll standards, as a result of random error.
The two polls disagree about the Gillard leadership, with Essential showing her approval down by 4 and disapproval up by 4, for an overall netsat change of -8, since four weeks ago. However, in the same time, Newspoll has Gillard only down a point, even though they are normally the more variable of the two. Notably, Essential actually shows a polarisation in Gillard's ratings, with an increase in the proportion of voters strongly supporting her performance, but the 4 point increase in voters disapproving comes entirely at the "strongly disapprove" end.
Both polls show Abbott at a popularity level still close to, but not quite at, his worst. This can be taken as evidence that the saga has not damaged opinions of Abbott despite the high level of disapproval of the Coalition's performance over the issue.
These polls will strengthen the move back to the Coalition that is showing in polling aggregations, some of which will now show the Coalition at above 53:47. Nonetheless given the inconclusive evidence concerning changes in leader popularity, and the fact that the move back to the Coalition started before the AWU saga boiled over, it is difficult to be confident of the level of AWU's contribution to the picture. To say that it has caused a 3-point change in Newspoll is incorrect, since the change is probably exaggerated by random movement (the last poll was below trend and this one is above.) It is odd that the pollster that showed Gillard's response in the most favourable light generally is now also the one that shows her taking a personal hit that may be related to it a week later.
Update 2 (Monday 17 Dec): a poll out today by Nielsen contains some similar questions to the Essential poll. Gillard's handling of AWU receives a slight plus from those aware of the issue (47% approve 40% disapprove) while Abbott's receives a massive minus (24% approve 64% disapprove). Note that in the same sample Gillard's netsat was -4 and Abbott's was -29 (a new worst for him from Nielsen) so Gillard's handling of AWU exceeds her netsat by 11 points while Abbott's trails behind his, woeful as it is, by the same amount. The difference in the questions is that Nielsen asked about Abbott's handling, while Essential asked about the Coalition's handling. The result for Abbott's handling was even worse (net -40 compared to Essential's -29 for Coalition), but this may be further coloured by voter dislike of Abbott.
It is no longer a good idea to attempt to untangle AWU impacts from changes in leader approval ratings or party standing - it was difficult enough in the first place, and now another issue (the Ashby finding) may be contaminating the picture further. But to argue that the AWU issue has damaged Labor's standing in recent polling is to argue that against a general backdrop of swinging voters most likely approving of Gillard's handling, with even Coalition voters deploring Abbott's, one or two points of swinging Labor support has swum against the tide and switched to the Coalition. That is possible in theory, but a claim that requires evidence beyond the 2PP shift itself, and there is none. Now that we've seen that the voter view of leader handling of the matter, as first accurately measured by Essential, has lasted, it's far more likely that the late move back to the Coalition this year was not caused by AWU at all.
The Nielsen poll's 52-48 result will probably be the end of a lot of nonsense narratives started by the three 54-46s. A new set of nonsense narratives will take their place.