Tuesday, September 3, 2013

Poll Roundup and Seat Betting Watch: All Disastrous Things Edition (September 3)

Election Night Update: I will be providing live commentary at The Mercury on election night from 6:30-ish til late.  An exact link will be posted on the night.

2PP Aggregate (Tuesday 3 September): 53.1 TO COALITION (+0.8 since last week)
Individual Seat Betting: Labor favourites in 54 (-5.5: Brand, Lilley, Parramatta, Kingsford-Smith, Lyons and Lingiari all to Coalition)
Seat Total Market: Labor 57 seats (-4) (This figure is probably slightly skewed by longshot bias.)

This is week ten in a regular weekly series in the leadup to the federal election.  Week nine was here and through it you can click back to the previous weeks.  Or just click the "betting" label at the bottom.  As stated before, the aim of this exercise is not to claim that seat betting markets have predictive value, but to test whether they do, and to see which of the markets and the aggregated polls see the ultimate outcome of the election first.  

This Week's Polls

The weekend's national polls were as follows:

* Galaxy 53-47 to the Coalition
* Newspoll 54-46
* Essential 52-48, including the information that this week's sample was 53:47
* Morgan multi-mode 52.5:47.5
* AMR (online panel poll similar to Essential) 53:47

Also late last week ReachTEL released a 53:47 and Morgan released a 54:46 (special multi-mode via combined internet and phone sampling).

All up these polls have knocked my aggregate out to 53.1 to Coalition, which is actually one of the more Labor-friendly readings doing the rounds (aggregators that use the last election results to align readings have it slightly over 54).  My aggregate uses very simple mathematics compared to the others, and it assumes that the data of previous weeks are useful in smoothing out the impact of specific polls in a given week.  This does mean it will tend to lag when there are strong and persistent trends in one direction.  My Newspoll rolling average is at 53.5, which is about as good for the Coalition as at the same point in the 1996 campaign. 

The apparent slowing in the Coalition's rise last week has turned out to be a false flat and the more or less linear decline of the three weeks before that has continued. 

Newspoll leadership ratings showed Kevin Rudd with a net satisfaction of -26 (32 satisfied, 58 dissatisfied) and Tony Abbott considered better Prime Minister 43:41.  (As noted before, preferred/better Prime Minister polling is skewed in favour of the incumbent, so when the Opposition Leader is leading, the government is usually in a world of pain.)  This is the first time Abbott has led Rudd according to Newspoll and indeed the first time anyone has led Rudd since John Howard did so by similar margins in Rudd's first three Newspolls as Opposition Leader around the end of 2006.  The only bad news for the Coalition in the Newspoll was that Tony Abbott is still not exactly popular, with a netsat of -10 (41:51).   It is notable that this Newspoll had a rather low "upfront exclusion" percentage of 6% (4% uncommitted, 2% refused).  The very last campaign Newspoll always has a very low upfront exclusion rate, but this is not always true of the one taken with one week to go.  So for those who were hanging on to hope that was based on a high undecided rate, there's nothing much to see there at the moment.

Rudd's personal ratings are in fact still better than Paul Keating's at the same stage of the 1993 election (Keating had a netsat of -31 and trailed John Hewson by six points as PPM) but there is a big difference: in 1993 the 2PP polling was level.  And ultimately, that's the bit that counts.

Galaxy showed Rudd still preferred over Abbott as better PM, 45-39.  Labor supporters almost all preferred Rudd (91-2) while Coalition fans were a little more hesitant about Abbott (80-9).  Green, independent and other voters still don't rate Abbott at all (I derive a lead for Rudd of 47-8 with 45% undecided). 

Essential, which seems to provide very benign personal ratings compared to Newspoll, showed Rudd with a netsat of -5 with Abbott on -9 and Rudd as better PM 44-38.  Essential's party breakdowns are fairly similar to Galaxy's, but Essential semi-usefully provides a Greens breakdown on better PM (74-6) making it possible to estimate the breakdown for Others (I make it 23:17 to Rudd with 60% undecided). 

Personal ratings from the Morgan special multi-mode were discussed in the updates to last week's edition, as were a slew of marginal seat polls that generally painted the usual grim picture for the ALP.

The big news today is regional polling from Queensland, with a Nielsen poll of Queensland returning 53:47 to the Coalition - implying some prospects of Labor gains - while a Newspoll of all Labor-held Queensland seats except Griffith showed 51:49 to the Coalition, implying multiple Labor losses.  In the Nielsen the leaders are tied as preferred PM but Abbott has a net trustworthiness (!) of -4 compared to -18 for Rudd. In the Newspoll, Rudd "leads" as better PM 47-40 while Rudd's netsat is -7 and Abbott's is -5.

The Nielsen poll used respondent-allocated preferences to compensate for a burgeoning Palmer United Party vote in state polling, with slightly more than half of PUP and KAP voters indicating they would preference Labor (albeit with a massive margin of error in these estimates).  However PUP how to vote cards preference the Coalition so whether Labor will get better than the standard ind/others 45:55-ish flow from PUP is unclear.  Another doubt I have about the theory that last-election preferences massively under-predict Labor's performance in Queensland is that the PUP vote is mostly regional but the seats Labor needs to win from the Coalition are mostly urban or urban-fringe.  Furthermore because of sophomore effects it is not sound to project the 53:47 to substantial Queensland seat gains even assuming uniform swings. But the PUP/KAP preferencing situation might also take the sting out of some of the worse results for Labor in marginals polling.  (It is also possible that the polls that read out PUP as a party name are getting an inflated PUP vote, as has tended to happen in the past with the Greens).

I have a suspicion that there is some logic in this possible change in the preferencing behaviour of KAP supporters compared to previous elections.  That is, that Kevin Rudd has been campaigning in a mildly Katter-ish fashion (northern development, a bit old-fashioned about foreign land, move the navy to Brisbane etc).  Not only is Labor support apparently drifting to the right-populist parties, but those already disposed to right-populist parties may be seeing Rudd as preferable to Abbott (who refused to be drawn into all this and gave a sensibly mainstream response to the foreign land ownership question asked in the most recent People's Forum.)

There are very detailed discussions of the Queensland polling mess by William Bowe here and Peter Brent here.  In my view the most likely upshot of all this is that the truth is somewhere in between what is implied by Newspoll and Nielsen.  I also agree with William Bowe that the chances of Clive Palmer winning the seat of Fairfax should not be too lightly written off.  Some time ago Scott Steel (@Pollytics) noted that in internal polls (presumably union polls) PUP were polling double figures off the bat in their best areas even prior to any significant campaigning.  This is a seat where individual polling would be interesting.

One more state result that caught my eye was that the Morgan multi-mode sample for Tasmania, albeit of only 160 votes this week, had a 15% swing from the mid-55s for Labor that Morgan has been getting most of the time.  With that included, the last four weeks of Morgan polling now average 52:48 to Labor (MOE about 4%), but that's over four weeks in which Labor's national 2PP has slid, and on an indicator (respondent-allocated preferences) that favours Labor. Considering all this the last four weeks of Morgan data now point to a swing in Tasmania not much smaller than that implied by ReachTEL, further confirming that Bass and Braddon are well below the expected waterline, the loss of Lyons is entirely feasible and Franklin remains shaky.

Epitaph for ALP Chances

In the last week, Labor's chances of winning the election as measured by analysis of polling data and historic trends have gone from remote to token.  Polling one week out from an election is quite a deal more tightly predictive historically than at two weeks out, and Labor's raw 2PP position is much worse than it was last week.  Also the suggestive evidence that the Rudd bounce might have stopped deflating has been overturned, leaving us with a five-week downwards trend at over half a point a week. It would be brave to assert that there must be recovery this week rather than still further decline.

Even the naive version of the model I've been using for assessing Labor's chances now gives Labor only a 3% chance of victory and an expected 47.6% 2PP (when modelled by party, which is at this stage much more predictive).  Modelling by the fate of the government of the day gives a higher 10% chance with an expected 48.3% 2PP, but is much less predictive at the best of times.  Even the former estimate set is in my view too high, because it ignores the extent to which Labor's vote still, even now, appears to be declining as the Rudd-return bounce continues to self-destruct.

If there is still decline ongoing this week then not only should a further blowout rather than narrowing be expected, but also rolling-average type aggregators like mine may be too conservative, in which case it is likely we will be over 54 or perhaps even out to 55-ish (should the pollsters collectively be as Labor-friendly as in 2010) by polling day.  After five weeks of decline in a row it is hard to accurately weight the possibility of further decline this week against the possibility that this will finally start behaving like a normal election.  Even weighting the former at only a half (which is probably much too low) Labor should be expecting about 46.5% on average, and as for winning chances, that $14 on offer for Labor is not something I think anyone sane should feel remotely tempted by. 

Another issue with these kinds of projections is that one of the previous data items also contains a bounce-washout, but that was one in Labor's favour (2001, one of only two cases where Labor has gained from polling a week out to election day). The straws that those hoping Labor would still win were clutching at have been one by one disappearing; the remaining hope is that the defeat might still be close, but I do not think it will be.  At the moment I'll be a little surprised if Labor gets above 48% on election day or below 44.5%; the midpoint of this range being something in the 46s.  1996 again, or perhaps a little worse, in other words, and with a seat tally that might not be much better.  I expect my seat analysis model (under heavy development!) to point to something lower than the 61 seats of the historic-trend prediction I have mentioned.  

Labor did the right thing in replacing Julia Gillard with Kevin Rudd, whatever the outcome, because under Gillard a loss was certain and it would have likely been a very heavy one.  They had a duty to attempt to win, but it hasn't worked, and in retrospect they needed to try it much earlier.  The late timing of the change of leadership has meant that the whole business of how election policies would be generated under the new regime has not been adequately sorted through.  The gamble was that a presidential-style campaign around Rudd would work, but it has not, leaving a tattered campaign strategy in which a leader with no clear authority announces policy via thought-bubble without consultation.

In contrast to the 2010 campaign, in which the two sides attempted to outcompete each other in their ability to dumb it down and say absolutely nothing, in this campaign I think that Labor has actually been trying to run substantial messages - but has been utterly confused about what those messages are.  Rudd Labor v2 is a poorly defined experimental product that doesn't seem to communicate too much clearly beyond that it just isn't Tony Abbott.

The Labor campaign has also had a lot of familiar death rattles about it.  Not only have we seen Labor fail to knock a hole in the Coalition's policy costings but we have also seen Rudd retreat into pseudo-psephology , picking fights with Murdoch and even praising Media Watch.  Yes the blatant partisanship of the Telegraph especially would be annoying for Labor but the media are a part of the game in this country, and to be attacking an outlet for bias (even when true) is to be conceding failure to have played that game effectively on its historically established terms. 

On the policy costings bit: some people think Labor deserved sympathy because the documents it released contained clear disclaimers about their own limitations.  However, Labor did make exaggerated claims about the ability of those documents to prove a hole in Coalition funding.  (And in general, I've not been completely impressed by the standards of fact-checking sites in this campaign, but the one linked to above gets the correct reading on the latest of Rudd's many spurious claims about evidence for Labor's chances.)

For a while, it was difficult to rule out the possibility that Labor's competitive polling from July would last until the election, although that is not normally what polling bounces do.  But in the end, what we are seeing affirms the past pattern that you cannot "bounce" an election.  Whether Labor ends up losing by more under Rudd than under Gillard or not, there are no miracles here.  Dead men rise up never, or if they do, they fall back down again.  A government that will probably get a better rap from history than now for its policy performance, but that has been a recurrent flop at selling its own message, and farcically inept at resolving its own leadership problems, is about to be put out of its misery.  Whether or not Tony Abbott is right that Labor could not fix its problems while in government, fixing them and preparing an effective campaign strategy in just a couple of months has been too hard.  Electoral oblivion may come as a relief of sorts; with both Gillard and (presumably) Rudd out of the way, the party may be able to unite under new leadership and gradually work over time on repairing its internal culture.

The best hope for an exceptional Labor recovery between now and polling day would be that the Coalition's belated costings announcements do contain significant flaws and that Labor manages to make headway on this issue in the last 48 hours.  But the record of message-selling by this government is so poor that even if there are substantial errors, the government will probably make a fist of exploiting them, and in any case, quite a lot of voters will have already voted via pre-poll or postal voting.  (I should clarify: I am not saying it is objectively impossible for Labor to win, just that it seems extremely unlikely.  Something would have to happen that defied all that we previously knew about the way polls and elections behave.)

Betting Roundup

This week the Centrebet/Sportingbet exact seats market has seat totals from 90 to 92 seats as joint favourites (this would probably leave Labor with about 57).  The correct election result market, which is probably slightly skewed by longshot bias (but which has pretty much dropped the ranges from Labor 80 seats up, which are all over 100-1 now), also currently implies 57 Labor seats. The handicap market has a Coalition margin of 29.5 (equivalent to about 59 Labor seats; for some reason this one is consistently a bit more generous to Labor than the others.)

The individual seat betting picture paints an appalling picture for the government.  Here's the graphic:

 The colour-coding again:

Medium blue: A seat in which the Coalition is favoured in all betting markets.
Pale blue:  Coalition favoured in some markets, level in others.
Grey: (none this week) All markets tied or both parties ahead in some markets and behind in others.
Orange: Labor favoured in some markets, level in others.
White: Labor favoured in all markets.

Bold shows a seat that has changed in colour in the last week. Petrie and Hindmarsh still have Labor favourite but with some ties on specific markets.  Lingiari and Parramatta have gone direct from white to blue, Lilley from white to light blue, Brand from grey to light blue, Kingsford-Smith from orange to light blue and Lyons from orange to dark blue.  All up, that's a loss of five and a half seats in a week to leave Labor with 54 expected seats and a loss of nineteen (94-54-2).  The suggested result is as bad as expected by the pendulum for a national 55:45 outcome, mainly because of the expected Tasmanian trainwreck; markets now believe (albeit weakly) that the voters really will retire Dick Adams from his 12.3% margin after 20 years.

This week's trend tracker (of seats that have seen changes in status) sees the addition of Petrie and Lilley:

Since the return of Kevin Rudd as Prime Minister, five seats (Brand, Lilley, Parramatta, Kingsford-Smith and Lyons) have moved from Labor to the Coalition according to betting markets and just one (Page) has moved the other way.

There was also a comparison with the final day of the Gillard era, in which Labor was favourite in 50 seats, by Peter van Onsolen in the Australian today.  The conclusion that might be drawn is that since Labor is only favourite in a few more seats now than then (and may fall further by election day) that the change from Gillard to Rudd is seen by the markets as having had no net benefit and that Labor may as well have kept Gillard.

Any such argument - at this stage - is fallacious.  The market expectations for individual seats on June 25 did not give the expected result if Gillard remained leader.  Rather, they had an estimated chance of Rudd returning, and its possible impacts, priced into them. Seats the markets would have been writing off had the markets known Gillard would remain leader were at shorter odds because of the possibility that a big Rudd bounce would save them.  I think this was especially true of the betting for Lyons at that time.  The market expectation of the seat result if Gillard remained was probably somewhere in the low to mid 40s.

On to the close seats analysis.  It would have been expected that the loss of five seats from the Labor pile and the addition of six to the Coalition's would have eased last week's 17-4 difference in close seat vulnerability.   It has, with Labor now favourites in twelve close seats compared to the Coalition's eight.  One of those eight seats is the Coalition's own seat of Brisbane.  The second-most vulnerable Coalition seat in the country according to seat betting is Indi, where Cathy McGowan is now at $3-$3.25, just outside my close-seat cutoff.  For Labor, Griffith is off the deathwatch but Bendigo is now on it.

I expect to post a detailed attempt at a seat projection on Friday, and then update it during Saturday should late polling demand.  I may also post a brief guide to election viewing sometime this week in an attempt to reduce the confusion habitually caused by issues with the ABC's coverage. If, as looks almost certain, I am not posting here for most of election night, there will be a link here to where you can find me.

(A few words in this article were ripped off Swinburne here; I could have ripped off so many more.)

Update 4 September: ReachTEL  A ReachTEL poll showing 52:48 to the Coalition this morning caused some surprise especially given that most ReachTEL national polls have shown a point or so in the Coalition's favour compared to other polls at the time.  This apparent "house effect" - an at one stage consistent impression based on relatively few national ReachTELs - seemed to be subsiding in the last couple of polls and Mark the Ballot now shows ReachTEL's recent form as inside Newspoll's.  As a result of this, ReachTEL national readings are no longer modified in my aggregate.  The preferred PM result (noting that ReachTEL's PPMs are very different from other pollsters' because of forcing) is 52.9-47.1 in Abbott's favour.

The news is not entirely good for Labor, in that the 2PP swing from the previous ReachTEL comes in the form of a loss of primary votes from the Coalition to the Palmer United Party - whose Queensland supporters told Nielsen they will lean towards preferencing Labor, but whose how-to-vote cards will preference the Coalition.  Views about whether this combination of circumstances will make PUP preferencing more or less Labor-friendly than the slightly pro-Coalition split of "others" preferences (I make it 56:44) at the last election vary widely.  I've taken the middle course and assumed no difference and am continuing to distribute Ind/Others preferences as per the last election. 

The other thing is that just because ReachTEL trims its headline 2PP doesn't mean we have to, when it has kindly supplied primaries to a decimal point.  And it turns out that by last-election preferences, this is only just a 52 rather than a 53; I make it 52.4 to Coalition.  With that added to my aggregate, but also with the house effect removed from the previous ReachTEL, the Coalition takes a minimal hit down to 53.0.

Update 4 Sep: Galaxy of Melbourne:  A new Greens-commissioned Galaxy of Melbourne has been reported as showing Adam Bandt (Green) on 40%, Cath Bowtell (ALP) on 30% and the Liberal Party on 26%.  This compares to a previous Galaxy showing Bandt on 48% that I did not take very seriously (and still don't.)  Reporting claims "Taken between Thursday and Monday, the poll found that Labor's Cath Bowtell might struggle to get home even with the help of Liberal preferences because of a collapsing Labor vote, a rising Liberal vote, and what appears a very high leakage rate of Liberal second-preference votes to the Greens." A respondent-allocated 2PP of 54:46 in Bandt's favour is provided, suggesting that Liberal voters would split only mildly in Bandt's favour ("just half" intended preferencing Labor and "forty per cent" preferenced Bandt; presumably the rest were undecided).

The problem here is that endless experience of past elections - most prominently 2004 in which Newspoll's use of respondent-allocated preferences caused one of its worst ever election-eve poll errors - shows that you cannot rely on what voters say they will do with their preferences when voters will be handed how-to-vote cards recommending otherwise.  This especially applies to major-party voters. In the Victorian state election (2010) in which Liberal voters were recommended via how-to-vote card to preference Labor, 66.4% of Liberal preferences in the district of Melbourne and 64.3% in the district of Richmond (the two seats that make up most of the federal electorate) did so.  (This includes votes received from other candidates, but these are not a major factor.)  On these sorts of distributions, Bandt would probably win, but only by a whisker (I make it just under 51:49).  Perhaps Bandt will be better at attracting Liberal preferences because of his profile, but Cath Bowtell is not exactly low-profile either.

Are the claimed primaries in this latest commissioned Galaxy poll credible?  In my view, they are.  Tracking for individual party primary votes (see the Poll Bludger estimates here) shows a surge for the Greens through the campaign in Victoria and South Australia.  The Green vote in Victoria currently appears to be tracking above the 2010 result of 12.66%.  Add in the sophomore effect for Bandt as a first-term member and the tendency that swings for the Greens concentrate in areas where they are doing well, and there is every reason to believe he should be polling in at least the high 30s.  The loss of points for Labor since the ReachTEL of Melbourne  is also consistent with state patterns.  I note that the BludgerTrack figures are weighted against past election performance so they have the tendency of pollsters to overcall the Green vote built into them.

Commissioned polls should always be treated with caution because of the possibility of selective release, and the remarkable tendency of those that are released to nearly always align with what their sponsor wants.  Nonetheless, even without this poll there is reason to extrapolate that Melbourne is much closer than the previous ReachTEL suggested.  Note also that the ReachTEL did not redistribute the undecided - doing so produced primaries of 35.2 for Bandt, 35.5 for Bowtell, and 23.7 for the Liberals, for a 2PP of around 54:46 to Labor.  The swing back in state patterns since then is probably worth at least a couple of points for Bandt.  There has been a tendency for commentators and punters alike to write this seat off, but the public polling evidence at least projects the seat as more or less line-ball.  Perhaps the punters know things from internal polling that analysts like me do not, or perhaps the state patterns and their possible impact on this seat have been overlooked.  On the evidence available to me, even completely ignoring this latest commissioned Galaxy (as I generally do with internal polls except where there is no alternative), Melbourne is more or less a tossup and there is no sound basis for writing off Bandt's chances of retaining the seat.

Melbourne has moved back into the "close seat" category on betting markets today, though the Greens are still at not much under $3. 

Indi Update: Border Mail "Poll":  Today Indi became not only a "close seat" on betting markets but in fact the Coalition seat considered most likely to fall.  Although there is no public polling since a ReachTEL that showed Cathy McGowan behind but competitive, there are increasing signs of Liberal concern that the campaign may unseat Sophie Mirabella, who is a vulnerable link in the Coalition chain for good reasons.  But whether or not it will, the Border Mail "poll" "showing" 65% support for McGowan provides no evidence at all, because it is quite obviously an opt-in of newspaper readers.  Such "polls" are prone to the usual cocktail of motivated response, skewed response by readership bias, multiple voting and co-ordinated stacking and should always be ignored.  Disappointing to see Crikey reporting on this one so credulously in an otherwise interesting report here, and again showing it is sometimes an echo rather than an alternative to the political-reporting foibles of the mainstream press.

Porcupine Fish Award for Ultra-Fishy Polling (image credit)
I have seen the argument advanced that a win for an independent should be ruled out because of the current political mood. I do not buy it.  The mood is not rampantly pro-Coalition nor generically anti-independent; it is anti-Labor.  There is no in-theory reason why an unpopular Coalition MP on a less than massive margin is immune to being picked off by an indie.  Whether this happens or not depends on the strength of campaigns in the seat in question.

Sex Party Into Kinky Polls?  I was today robocalled on apparent behalf of the Sex Party.  The poll asked if I was aware that Weetbix was owned by Seventh-Day Adventists, that church-owned companies were tax exempt, and that taxing church-owned companies would raise <blah>.  Then I was asked if I supported taxing church-owned companies (a question probably wasted on me since I need very little encouragement on such a score).  It is almost identical to the techniques demonstrated in the Yes Minister excerpt that I link to on the sidebar.  The preliminary questions drive up support for the final question which can then be used in media splashes claiming a huge percentage of agreement with the party's policy (a classic "skew-poll").  Whether the intent is to publish the results or just to directly drive up support for the party is unclear at this stage, but any result appearing from this survey should be sent back to the bedroom.

Petrie Flips (5 Sep): Petrie (Qld) is now a Coalition favourite on all markets.

ReachTEL DuJour (5 Sep):  I stayed up slightly after my normal bedtime to check the results of today's ReachTEL (yep, two in two days).  It's 53-47 to the Coalition with the rounded ALP primary down 2 to 33, the rounded PUP primary up 2 to 6 (yikes!) and the Green and Liberal primaries unmoved from yesterday (10 and 44 respectively).  With those primary moves I've taken a punt that it's actually 53-point-something, which moves my aggregate back to 53.1; if the ReachTEL turns out to be 53.0 or lower then mine should also be 53.0. There is, at least, in these two polls no sign yet that things have got worse for Labor since last weekend.  Abbott is PPM by the entirely unremarkable (by ReachTEL standards) margin of 54:46.  More details later.  [Update: the primaries are up here and I make the 2PP only marginally south of 53.5:46.5, with the usual room for argument about KAP/PUP preferences.]

Voters are more or less split on calling a double dissolution to remove the carbon tax, and strongly opposed to breaking election promises for the sake of the mythical budget surplus monster.

ReachTEL Bass (5 Sep): There was a time when fresh polling of the seat of Bass alone was worth a whole article here, but these days more of what we already know is but a footnote.  The latest poll of what seat betting markets consider to be one of the furthest gone seats in the land shows Andrew Nikolic with a thumping 59:41 lead, a negligible increase on the 58.4% of the Mercury's poll in August. (Incidentally the game of ReachTEL-commissioning ping-pong between the two main Tasmanian newspapers is resulting in each ReachTEL release showing only the past polls commissioned for that paper and not the other one, which is probably causing much confusion.)

Despite being apparently headed for a big loss, Geoff Lyons is still rating not too badly; it's apparently about party and not personal for the incumbent with a netsat of -3.1.  Andrew Nikolic, however, enjoys a massive +32.2.  The Green vote is somewhat better than in the August poll, which doesn't hurt their Senate prospects (which I think are pretty healthy now anyway).  The collective Others vote is up since August by a margin that is almost statistically significant; it makes sense that PUP and Family First are the main beneficiaries.  In a very sad day for religious freedom, the Secular Party polled an awesome result of zero.  (No, I can't give you the margin of error on that one.)

Essential FINAL (5 Sep): Essential's final result is 52:48 to Coalition off a sample of 1035 from Sep 1-4.  This actually qualifies it for a weighting of eight in my aggregate - for one day only!  We see the usual hierachy of firmness to softness of voting intention (Coalition - Labor - Greens - Other).  Voters claim to be most motivated by the economy, the interests of "all Australians" (aaargh!) and self-interest in that order, but with a big partisan divide:

* Coalition voters overwhelmingly nominate the economy (69%) as one of their big issues, followed distantly by needing a change of Government and ability to govern effectively.
* Labor voters nominate the three aspects listed for all voters but are about as likely to nominate "they have a better leader" (whether this is code for liking Rudd or disliking Abbott is not clear)
* Green voters nominate the environment, trust and their idea of the  national interest
* Others voters are perplexingly likely to nominate "We need a change of Government", but off a very small sample size.

16% of voters say their opinion of Prime Minister Rudd has gone up since the election was called compared to 40% down, with the predictable partisan skew.  For Abbott the result is 25-33.  The key difference here is that Labor fans are less likely to say their view of Rudd has improved (33-12 with 53% unchanged), than Coalition fans are to say the same of Abbott (51-5 with 42% unchanged).

Green voters are none too happy with either (9-42 for Abbott, 8-63 for Rudd.)  For Ind/Others voters I derive figures of 8-30 for Rudd, while for Abbott 49% say their view has gone down while the portion who say it has gone up is lost in the rounding.  The sample size for Ind/Others voters is very small, and this is in the context of a poll that has provided a lenient 2PP for Labor, but still, another little morsel of evidence that Others voters just might not break for Coalition this time.  That is not to suggest they will break in Labor's favour either, but an even break is possible.

ReachTEL Du Jour (Sep 6): Today's ReachTEL finds the Government with about 47.2% 2PP (compared to 46.5% yesterday and 47.6% the day before); the daily movements of 2PP so far are statistically insignificant and are moving my aggregate up or down by no more than 0.1 of a point.  The Palmer United surge continues, now to 7% nationwide, adding much uncertainty to preferencing estimates.

Even accounting for ReachTEL's tendency to not favour incumbents in preferred-leader questions, Joe Hockey enjoys a thumping lead over relative newbie Chris Bowen as preferred Treasurer (58.3:41.7), including being preferred by a quarter of Green voters.  (I add that at every election there are some people who believe some policy issue of the day will cause virtually all Green voters to preference Labor, and at every election they are wrong - as also are those who expect the proportion of Green voters preferencing Labor to crash).  Tony Abbott's PPM lead tracks the overall 2PP closely.

Voters are also asked who ran the most negative campaign, and Labor gets the gong 56.6:43.4.

Galaxy (Sep 6): Galaxy too has released a 53:47 taken Sep 2-4 and hence already relatively old data.  Palmer United are included for the first time, polling 5; the last two days of ReachTEL suggest it could have already exceeded this.  The economic management question of course favours the Coalition by the usual amount (47:34) with slightly more Labor voters uncommitted to their own party on the issue; non-major-party voters break about 24:20 to the Coalition.  Tony Abbott is seen as having out-campaigned Kevin Rudd 43:36 with Labor supporters again a bit lukewarm about their guy; non-major-party voters break to Rudd 29:20.

And then we have the usual Galaxy murky loaded question section. This starts with the especially brazen: "If the Coalition win the election on Saturday do you believe this will be because they deserve to be elected or because Labor deserve to lose?"  The question leaves the respondent no room to decide that a Coalition victory might be undeserved, and a significant proportion of Labor supporters clearly do believe that that is the case.  Anyway those willing to volunteer either of the offered responses (which is virtually all the Coalition supporters and a surprisingly generous 52% of Labor supporters) lean towards the latter option in both cases (44-50 and 18-34), while third-party supporters strongly prefer that Labor deserved to lose (10-48 with 42 either "uncommitted" or contemptuous of the question).

There is also a repeat of the question that offers a description of Rudd's former style ("chaotic and dysfunctional") and then asks respondents whether he has changed, without giving them the opportunity to indicate whether they agree with that description.  (So basically, to agree he has changed you must agree the description was correct in the first place, and if you don't agree he has changed then you also agree.  It is basically a "when did you stop beating your wife?" type question.)  This is the third time this one has been submitted to voters, and the results show a declining trajectory for the PM, but the questions defects make it not worth analysing further. 

No comments:

Post a Comment