Friday, June 28, 2013

Rudd Replaces Gillard

"Being prime minister isn't a job you have a shot at and then you come back again if you think you've learned a lesson" - Gary Gray, Feb 20 2012.

Indeed not, Mr Gray.  The last egomaniac to pull that little trick off lasted only seventeen years the second time around.

This post discusses various polling history and projection aspects of the replacement of Julia Gillard with Kevin Rudd.  Of course, this major event has rendered a lot of poll-based projections that were premised on the idea of a Gillard prime ministership at the election void, and all over Australia psephologists are scurrying like spiders to repair their webs.

Historical Aspects

Kevin Rudd is the fourth Australian Prime Minister to recapture the role after losing it, after Deakin (twice), Fisher  (twice) and Menzies.  Fisher and Deakin played musical chairs in an era before majority governments were the norm (though Fisher later lost an election outright before winning the next one outright).  Menzies lost the support of his party and resigned, then led his party back from opposition.  Rudd is the first departed PM to return to the post without an intervening change of government.

Even at state level, the recapture of the top job (as opposed to the Opposition top job) is rare in the modern world.  It was common in the chaotic nineteenth and early 20th century state parliaments, such that NSW has had three five-time Premiers.   Since the 1950s, however, only Eric Reece in Tasmania (1972, recovering government against a short-term Liberal minority that collapsed during its term) and Don Dunstan in South Australia (1970, ditto).  So Rudd is the first leader at state or federal level for 41 years to lose his job and recover it, and has done so in difficult and unusual circumstances (albeit with a huge assist from the hung-parliament situation and the generally hapless self-marketing and trust problems of his successor's regime.)

Julia Gillard is the fourth Prime Minister, following Gorton, Hawke and Rudd, to be removed as a result of a leadership spill by their own party.  Also, both Hughes and Menzies (first time) averted that fate by resigning.

In an old piece As Gillard Recovered, So Can Abbott? (the Abbott-related content since reviewed and some comments recanted in The Abbott Factor Revisited)  I looked at unpopular Prime Ministerships in terms of the Prime Minister's worst ratings and their subsequent recovery when removed or at the election.  Julia Gillard's worst approval netsat was -45, and her subsequent peak was -11, which was actually the third largest recovery by a sitting PM in a given term in the Newspoll era.  However, by the time of her removal her rating had slipped back to -34, and it's doubtful it would have improved that much from there had she been retained. We'll never know for sure.

Gillard is frequently described as the most unpopular Australian Prime Minister ever, but that actually isn't correct.  Again looking at Newspoll netsats, Paul Keating had a career median of -24 and a career average of -24.4 and never polled a single positive netsat.  Julia Gillard had a career median of -19.5 and a career average of -15.9 and polled positive netsats consistently over her first seven months in office.  In all Gillard recorded 13 positive, 52 negative and one zero netsat ratings.  There are various ways of comparing the popularity of Gillard and Keating (eg just their last terms in office, just their worst patches of a certain length) and in general Keating had a slightly worse popularity record than Gillard.  But despite that, Keating was able to recover from bad polling six months prior to win an election outright, and to then record competitive polling and see off two Opposition Leaders in his next term.  Gillard was unable to make her unpopularity an electoral asset by generating grudging respect of the sort Keating sometimes attracted, and outside of the post-election honeymoon period, led an uncompetitive party for all but five months of the last two years.  When the party was competitive, it was still trailing.

History that Wasn't

We will never know what exactly would have happened had Gillard led the party to an election, but there was an increasing sense among psephologists that in broad terms we'd already seen how the movie ended and the question was only whether the government would get soundly beaten, thrashed or smashed into little pieces that would take several years to reassemble into a competitive party.  In contrast, the Rudd return is a massive wild card - we haven't seen the likes of this before and no-one really knows how it will play out. 

My own aggregate of the 2PP at the time of Gillard's departure was 56.1:43.9, which on average would have resulted in loss of about 26 seats.  The serious modelling debate was between the "narrowing" model (that the 2PP would improve to something around 46.5 and a seat tally in the sixties) and the "random-walk" model (that the 2PP would not improve in a sustained direction and was about as likely to go down as significantly up, and that a seat tally somewhere in the 40s was most likely.)  I was finding the random-walk model increasingly attractive after three months of essentially flat polling and given Labor's historically poor recovery abilities.  A third version was the aggravated random-walk model (ie that Labor would not improve its 2PP and would lose a lot more seats than the pendulum implied, and possibly be reduced to 30-35), and for those who care, I was never too convinced of that one.   That Gillard would have lost is just about beyond serious dispute; by the end of her career all of us who tried our hands at crude Aussie versions of the Nate Silver game (most of them unpublished except for Pottinger's, which was the least crude one) were getting victory chances, by various means, ranging from less than one to a few percent.

Maybe in future, snippets of what would have been the full Gillard campaign will come to light as eerie crossovers from an alternative reality in which she was retained, like this ghost of what would have been the 2010 Kevin Rudd campaign.

Where To From Here?

In trying to model Labor's voting intention now, both Mark the Ballot and Pottinger have decided to declare a discontinuity - the data of the last months of the Gillard era are basically no longer predictively useful.  I'd already independently done the same thing in my own simple aggregation exercise.  Probably there will be a rush of quick polls and within a couple of weeks a pretty decent aggregate will be possible.  Meanwhile it's useful that the first poll to come out, the Morgan snap SMS (51:49 to Coalition 2PP based on last-election preferences) is in the ballpark of hypothetical polls about the impact of a change to Rudd.

(On the downside, the Tasmanian sample of 110 voters supposedly 59:41 for Labor before the spill and 63:37 after is impossible to take seriously even allowing for a 9-point MOE.)

What Value Change?

The large bounce for a change to Kevin Rudd apparent in the first Morgan poll (about five points compared to the aggregated national 2PP) is not surprising.  In the past I've found that the usual value of a leadership change is about three points with the benefit peaking from three to eight months after the change.  In this case Labor probably doesn't have three months and the closeness of the election is likely to make the bounce wash out of the system faster, leaving Labor only with whatever lasting benefit they derive from changing leaders. 

But the impact of leadership changes varies a lot from bounce to bounce.  The bounce from Downer to Howard was worth a seven-point 2PP gain in two months, about four points of which proved permanent.  The bounce from Beazley to Rudd was worth six points within three months (and eight points after six), although in the end Rudd's result was only 1.5 points better than the 2PP under Beazley when he was removed.

An important result of past polling is that attempts to "bounce" an election with a late leadership change have had underwhelming results.  Labor tried that last time and by the time the poll was reached their polling position was worse than before the leadership change.  More significantly, the classic case of this sort of tactic came when Bob Hawke replaced Bill Hayden on the eve of the 1983 election.  In the last three polls under Bill Hayden, Labor's primary lead averaged five points and in the campaign polls under Hawke the party averaged 9.3.  But the final margin was just 5.9.  Hawke's popularity provided added confidence that the election would not slip away as 1980 had done, but the bounce essentially didn't last and it's not clear Hayden would have actually done any worse.

There's some expectation that the initial ratings will be as good as it gets for the recycled PM and that it will be all downhill from there.  That actually isn't how it has usually worked in the past, but because Rudd is a known quantity it seems a bit more likely to be that way this time.  Especially, people know what they liked about Rudd but can be surprised by the re-emergence of forgotten ways in which they found him irksome.

In my view, the decision by Julia Gillard to declare she would quit politics at the election if she lost, and her encouragement to Kevin Rudd to accept the same, has greatly increased the chance that Labor will be able to go to the election without constant severe infighting.  Also, it looks like the danger of the parliament falling prematurely on the floor of the House has been averted.  These are two key ingredients in increasing the chance that some of the gain from the change will persist til the election.  It is hard to see that Labor will crash back to 44:56 in two (or even three or four) months from here; the terrible polling we have been seeing in the last few months was probably artificially deflated rather than natural.  So once the size of the Rudd-return bounce is clearer, the question will be how much of it can be conserved to the election?  Things will get particularly interesting if the Rudd bounce takes Labor into a substantial lead, although there is not yet any evidence that this will happen.

There are a lot of comments around to the effect that Rudd should go to an election quickly to maximise the size of his bounce.   Rod Cameron, projecting a rather generous 10-point bounce  (on primary vote maybe; it would be unprecedented on 2PP) reckons this is the way to go.  I agree (again) with Peter Brent's view that poll bounces are not "real" and that when massive shifts happen very close to an election there is much faster adjustment than at other times.

That's not to say Rudd should necessarily avoid an election before September 14, just that he shouldn't be too concerned with the poll bounce issue as an argument for doing so in my view.  Arguments for going (relatively) early may include shutting up businesses wailing about possible postponement, and creating public relief that the endless campaign will be over a bit sooner than expected.   Arguments against include that a longer period of rule could give Rudd more time to make serious reforms and demonstrate that he really can govern effectively again. 

Hard For Labor

Although Kevin Rudd's popularity compared with Tony Abbott's, and the country's generally good economic situation, gives Labor some chance of winning the election now, there are many reasons to believe it will be difficult.  If polling settles at somewhere near 50:50, the following are reasons why it will not be easy for Labor to win:

1. Opposing sophomores: At the 2010 election the Coalition won eleven seats from Labor, most of them with members new to Parliament.  These members build profile as sitting members during their first term, making them more resistant to swings against them at this election (called the "sophomore effect".)  Labor in contrast won only two such seats.  In a close election each side would be winning seats from the other, but it will be harder for Labor to do this with so many opposing members having this boost.  The good news for Labor here is that seven seats are in Queensland, so they would be hoping that with Rudd restored there is a big enough swing back in that state to swamp the sophomore problem.

2. Retiring members: At this stage eleven Labor, nine Coalition and two Independent members have announced they are throwing in the towel.  But while the Coalition members are mostly in very safe seats, the Labor departures include the seats of Rankin, Capricornia, Kingsford-Smith, Barton and Perth, all of which have been on the radar as loseable seats on margins less than 7%.  The Queensland effect may well protect the first two.

3. The credibility problem: Tony Abbott is already flogging the line that Labor voters voted for Rudd in 2007 and got Gillard and voted for Gillard in 2010 and got Rudd, so how are they to know who they will get if they vote for Rudd this time?  It's a threadbare line factually since in each case the chosen PM ruled over the great bulk of the parliamentary term and was replaced only in the leadup to the election at the end of it, but the history of rolling leaders does raise some questions.  There's no evidence that the restoration of Rudd really shows that those who deposed Rudd in 2010 have realised they did the wrong thing.  They have reinstated him not in contrition but in self-preservation in the interests of the party's electoral chances, in realisation that Gillard specifically was not a competitive leader, and it is easy for the Coalition to market the line that should Rudd win it is only a matter of time before Bill Shorten seizes on a batch of dodgy polling and makes his move.

4. Resourcing: Labor has been campaigning on the assumption of a primary vote of around 31%, writing off ultra-marginals and sandbagging second-tier seats that actually won't be near being lost if the election is close.   It now faces the challenge of furiously trying to save seats it had been neglecting as lost.  It's been interesting to watch a number of the formerly doomed members for Labor's ultra-marginals bouncing up in the final Question Time to get their faces on TV again.

5. A House Divided: As discredited as the Gillard/Swan regime had become in the view of the general public, losing several ministers doesn't make it easy to campaign for re-election, especially not when the same ministers show up in attack ads quoted against you.  The Liberals' attack ads on Kevin Rudd have already started, and its easy because all they have to do is quote Rudd opponents who can't stand him.  Some have questioned the inclusion of former leader Mark Latham in these ads but while many people consider Latham to be one of Australia's top half-dozen political lunatics, it only takes a small percentage of viewers who still respect Latham as a "straight talker" to justify adding his footage.

Especially with items 3 and 5, Labor has to realise these are points they can't win by normal methods, and beat the Coalition at their own game of appealling to the irrational.  They have to aim to make voters lose interest in attack-ads and scare campaigns and respond to the situation in a different way to the way they have viewed politics over the last few years.  

Seat Betting: Back To The Drawing Board

It will be necessary to see new seat betting in many marginal seats to now get any real idea of where those seats settle following the change in Prime Minister.  Probably seat polls taken after Rudd has been back for a few weeks will be most instructive here, especially in Queensland where the initial surge is likely to be especially strong.  It will be especially useful to see new seat betting in seats that were the subject of "if Rudd was Prime Minister how would you vote?" polling, to see whether that polling is similar now that Rudd is PM again.

And Lastly - A Personal Note

I would just like to add a personal note of deep relief that - whatever his flaws, however it happened -  the Australian Labor Party now has a leader who supports same-sex marriage, the first Prime Minister to do so.  Although personally unaffected in any direct sense by the SSM issue, I feel a deep sense that it is a no-brainer that the vast majority of caucus members of a "progressive" party (including the leader) should support this reform, one that delivers equality and freedom from discrimination and social stigma, and that harms nobody.  It is especially important that they do so as a symbolic stand against the imposing or maintaining of senseless restrictions in the name of religious morality more generally, and such restrictions must in my view be mercilessly driven from public life wherever they occur (one hardly expects the conservative side to do it.)  The issue is also important as a symbolic stand against discriminating against others just because they are irrelevantly "different".  Opposition to this reform should be considered a disqualifier to ALP leadership or very high office in the future, and I believe that it will be.

In mid-May I wrote a scathing piece entitled Julia Gillard: Same-Sex Marriage Enemy #1, attacking the former Prime Minister for her history of opposition to this reform.  Initially this piece received not that many hits, but in the last few weeks it has been receiving many more, and in the day of the Prime Minister's removal it was the most visited article on the site by some margin, and with an increase in a range of related search terms to match.  I do not know the cause of it, and it may just be that a lot of LGBT(I) press writers found the article while working on pieces about the SSM implications of the leadership handover.  But for whatever reasons, it seems there was a groundswell of increased online interest in the positions of Gillard and Rudd on same-sex marriage, not just as Gillard lost office, but in the fortnight leading up to this occurring. 

Either that, or one person doing a very long homework assignment!


Updates (There's a lot of them ...)

28 June: ReachTEL update: A large ReachTEL just released shows a national 52-48 lead to the Coalition.  I have not seen the full state breakdown but someone evidently has, because Scott Steel (Possum Comitatus) sent out a link to an aggregated breakdown of the Morgan and ReachTEL polls on the ABC Elections Calculator that looks like this:

 Now, I don't entirely believe this projection.  For starters it incorporates the Morgan Tasmanian sample that as discussed above is plainly bulldust (and was so both before and after the spill, suggesting that Morgan's Tasmanian sampling is quite unrepresentative as opposed to just the most recent one being rogue) and only the incorporation of that Morgan Tasmanian sample stops Bass and Braddon from being allocated to the Coalition (in Bass's case by the sorts of precarious margins the seat has often been decided by in the past).  So I'm inclined to knock at least those two seats off, and even without knowing anything else about this projection, treat the median projected outcome as at best a hung parliament, not a Labor victory.  Some more cold water to be poured on that projection is that it gives another seven seats to Labor by margins of less than a point compared to only one for the Coalition, with five of those seven by margins of 0.3 points or less.  In practice, some of these seats would go to the Coalition. 

But it doesn't matter because the margin of error in any projection based on just two polls is going to be substantial, and what is important about this is that it highlights how Labor will save this if they can somehow save it at all.  Assuming a reasonably close national 2PP, it's all about winning back a brace of seats in Queensland on the back of the return of that state's prodigal son, and containing the damage everywhere else.  And it is quite possible to lose the national 2PP narrowly and retain office under such a scenario - the reverse of the case under Gillard in which even a 50:50 2PP if it could have been miraculously achieved was still going to be a loss.  Outside of Queensland it is going to be very difficult for Labor to win seats from the Coalition that it has, to this point, not even been thinking about targeting.

A graph posted by Possum (here) has also made the point about Queensland strongly - for a given gain on Labor's 2010 2PP, a medium-sized swing in Queensland delivers a lot of seats. (What the graph doesn't show is losses in some states where the 2PP is likely to go backwards.)

Of course, we don't know yet whether Labor can maintain a competitive national 2PP position until polling day.

ReachTEL in aggregate: I am now including the national ReachTEL in my aggregate but I think the jury is still out on the question of whether or not national ReachTELs have a small pro-Coalition house effect.  This will be reviewed after we see a full set of polls from other pollsters and if necessary the global house effect correction will be altered.

Update (29 June):  We have now seen the first ReachTEL seat polling that canvasses seats where the Rudd question was previously asked.  In the seat of Maribyrnong (ALP 17.5%), held by Bill Shorten, the surprisingly thin 52.5:47.5 lead for Shorten on June 8 has jumped to 58.6:41.4.  This compares to a projected jump from the June 8 poll to 61.1:38.9.

In the seat of Blaxland (ALP 12.2%), held by Jason Clare, the 47.8:52.2 loss polled on June 8 has turned into a 58.9:41.1 win.  This compares to a projected jump from the June 8 poll to 56.2:43.8.

In the seat of McMahon (ALP 7.8%), held by new Treasurer Chris Bowen, the 38:62 loss polled on 28th Feb has apparently turned into a Labor lead (I have not yet seen the margin of win published).  This compares to a projected Rudd jump from the Feb 28 poll to a 47:53 deficit.

In the seat of Chisholm (ALP 5.8%), held by Speaker Anna Burke,  the poll shows a 55.2% result for the incumbent.  I have not been able to find a previous published ReachTEL result for this seat (though I have seen claims one exists) but I do note that JWS Research had Burke winning with 55.6% 2PP before the change.

Seat polls tend to have in-theory margins of error of around the 4% mark or higher, and because of scaling the real margins can actually be higher.  So individual differences between seat polls of several percent often don't prove anything.  But in summary of the above, there's no evidence that the initial Rudd jumps in these seats are any smaller than expected based on hypothetical polling.  Whether those jumps last, decrease or even amplify remains to be seen but it would not be a surprise now to see Galaxy and Nielsen come in somewhere around 50:50 on the basis of their own hypothetical polling.

Galaxy (30 June):  Galaxy was quite obedient to the above, coming in at 51:49 to the Coalition.  This was described as a "winning position" on Twitter by the Sunday Telegraph editor Mick Carroll, but as the above comments on state-based projections indicate, it's anyone's guess what 51:49 would actually add up to if there was a large swing to Labor in Queensland.

Kevin Rudd is shown as leading Tony Abbott as preferred Prime Minister 51:34, which isn't surprising and means little given incumbent-advantage and bounce factors, but will receive a lot of press attention anyway.   More usefully the party breakdown (full table here) shows 12% of Coalition supporters preferring Rudd over Abbott while only 1% of Labor supporters prefer Abbott over Rudd - and this is even after the Rudd return would have taken any "Rudd refugees" who had fled to the Coalition back.  (By comparison, in March 5% of Labor supporters preferred Abbott over Gillard and 4% of Coalition supporters preferred Gillard over Abbott.)  Among voters for the Greens (10%) and Ind/Other (8%) combined, the PPM breakdown is about Rudd 53 Abbott 6 Uncommitted 41.  By comparison in March (when the overall uncommitted rate on PPM was 30) it was Gillard 32 Abbott 5 Uncommitted 62.  In all Rudd is up 18 points on where Gillard was after the March failed spill, while Abbott is down three.  Voters who preferred Gillard over Abbott over Rudd are probably about as common as Night Parrots and the change is most likely coming mainly from voters who were sick of both Gillard and Abbott but find Rudd at least bearable, with a small percentage who prefer Rudd over Abbott over Gillard - at least for now.

The split on the question of whether Labor made the right decision is 57-31 in favour.  Some might point to the inconsistency of this finding with pre-split polls that were reported as saying voters were against a change. However, if anyone thinks that this is solely a post-hoc decision that the change went better than expected, the view that Labor made the wrong decision by keeping Gillard back in March was nearly as strong (53-32).  Rather, those pre-split poll questions that supposedly showed opposition to change were defective because of overly restrictive wording; one asked people only if they thought Gillard should resign, another asked people only if they thought Rudd should challenge.  There would be voters who thought Gillard should have taken it to a ballot rather than resigning (as she did) but preferred Rudd in a challenge and there would be voters who thought Rudd should not challenge, but given that he did challenge, would still have preferred that he win.   It is actually Green/Ind/Others voters who are least supportive of Labor's decision (split about 36:36, 28% uncommitted).  This compares to a 57:37 endorsement by such voters of Rudd over Gillard in March, and I suspect that we're here seeing flow in both directions between Labor and the Green/Ind/Others pile - both "Rudd refugees" returning to Labor and smaller numbers of Gillard supporters switching to non-major party options.  Anecdotally, a few left-wing female voters are viewing the removal of Gillard through the prism of sexism (as if the ALP is expected to slaughter itself at the ballot box just for the sake of not having a male leader) and are extremely upset about it.

More On That Projection:  I've checked my hunch regarding the lopsided nature of the seat distribution in the Possum projection showing Labor supposedly winning via the following method.  I've allowed each seat to vary by +/- 2.534 points, which is the standard deviation for seat swings cf. the state average swing from 1993-2010 according to this MTB analysis.  I've found the percentage expectancy for each result assuming a normal distribution, which the past distribution is reasonably close to.  So, for instance, a seat showing as ALP +0.1 actually translates to a 51.6% chance that Labor would win that seat, and a seat showing as L/NP +2.5 has an 83.8% chance of being won by the Coalition.  (It might be possible to use individual state past values for this exercise but there is not much variation between those anyway, and I don't know if the states are even significantly different.)

The seats O'Connor, Denison, Kennedy and Melbourne are excluded from the check as no swing value was given for them in the Possum projection.  I've copied the original assignment of these (Coalition, Ind, KAP, ALP respectively).

The total Labor expected seat tally from the 76 seats projected to Labor is 69.05, so the Coalition would on average be expected to win just short of seven of those 76 seats.  On the other hand, the total Coalition expected seat tally from the 70 seats projected to the Coalition is 67.54, meaning that Labor would on average be expected to pick up two or three of those.  Adding the totals together the Coalition projects to 74.5 and Labor to 71.5, and the Coalition also has O'Connor.  This means the correct median outcome of the merged ReachTEL/Morgan seat samples is actually either 76-72-2 to Coalition or 75-73-2 to Coalition, and not the stated 77-71-2 to Labor.

That doesn't mean I think those are correct readings of the current polling.  I mentioned above that I dismiss Morgan's Tas samples out of hand at this point (except perhaps for swing estimation purposes), though all other things being equal that would mean their samples were a little too pro-Coalition somewhere else.   A bigger concern in the opposite direction would be if ReachTEL national polls had a pro-Coalition house effect.  All what I have done here means is that a given projection purporting to show a Labor win isn't actually valid with an assumption of no house effects.

I had a go at finding out what size of ReachTEL national house effect would be enough to make 77-71-2 to Labor a valid projection, ignoring the Tasmanian swing issue.  The answer is about two points.

And Now, Newspoll (1 July)

Newspoll also came in at 51-49 so there is consensus between the different polls so far - an election  held now would result in a tight contest in which Labor might just hold.  This is, however, a result at the start of a bounce and I stress again that no-one knows if that bounce will now amplify or fade.

This was a 6-point gain on the last Newspoll, in circumstances in which nearly all of the gain appears from the other polls to be real.   It is actually extremely rare for a sitting government to gain six points in a single Newspoll, and the only prior examples (excluding those where there was an election between two polls) are:

* November 1992 (+8): The sudden polar shift that I've talked about before (often attributed to the Hewson GST becoming unpopular and the Kennett factor).
* September 2001 (+6): I think you all know what happened then.
* May 2004 (+6): Rogue poll (This appears as +7 on Newspoll's website, but that was using respondent-allocated preferences)
* March 2008 (+6): Probable rogue poll during honeymoon period for very popular new government

So it's only happened four times since 1986, and two of those were probably not real.

Leader ratings show Rudd with a net satisfaction of, amusingly, zero (+36-36 with 28 undecided).  Usually new Prime Ministers start with high net satisfactions because there is a low disapproval rate while people give them a "fair go".  Also, they often take over after winning an election, which doesn't hurt.  With the exception of Fraser's initial poor ratings as caretaker PM after the Dismissal (his post-election ratings were good), the only other leader to start a spell as PM with a non-positive netsat since the late 1960s was Keating, who commenced with -21 (21% approve 42% disapprove). 

Rudd has had very little time to reestablish himself prior to this question being asked, but the impression I get from this and from the available PPM polling (both Newspoll and Galaxy) is that the returned Rudd isn't going to be wildly popular.  He's not that guy who recorded netsats of +55 as Opposition Leader and +57 as PM.  Indeed, his first approval rating is no higher than when he was removed (it's still higher than Gillard's has been since January), though allowance should be made for the far higher undecided rate.   We'll have to see which way the uncommitted voters go in current polls (the undecided rate of 28 should soon fall to the mid to high teens) but Rudd-2 won't be setting any records in the leadup to the election.  Rudd's Preferred Prime Minister lead of 14 points is just about what would be expected for a poll in which Labor is trailing 49-51, and indeed Julia Gillard had the same lead over Tony Abbott twice late last year.  (At that time, however, Abbott's personal ratings were very much worse.)

Seat Betting: I've had a brief look at the seat betting market released by Sportsbet today.  At this time the Coalition leads Labor in 14 Labor-held seats, and is level in four.  I expect these are basically fresh odds released by the bookie with relatively little market input so, for what seat betting is worth, I'll be keeping an eye on shifts over the next week, and might put up a piece following these movements when the market has had more time to be informed or misled by the views of punters.  I follow seat betting more to test how predictive it is than because I believe it is necessarily useful.  Also, looking at seat total betting the market expectation at the moment seems to be that Labor will get around 57 seats.

Essential and Morgan: Somehow it's no surprise that Morgan (which has a history of producing what would be headline-grabbing contrary results if the mainstream media actually paid it attention) has come out with the first Labor lead, a 51-49 by last election preferences in this week's multi-mode.  Essential meanwhile had the Coalition in front 52-48 but off a fairly small one-week sample (970).  They also issued a rolled average for the last two weeks, which I'm ignoring.

Last year and early this year Essential was delivering very Coalition-leaning results compared to Newspoll, but this pattern shifted when Labor became uncompetitive.  It will be interesting to see how it goes if Labor remains competitive for any length of time.


  1. I get a moderate pro-Coalition lean for ReachTEL around 1.8 percentage points.

  2. I find a pro-Coalition lean for ReachTEL of around 0.8% +/- 1.6%


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.