2001: The Final Frontier

Advance summary:

1.  Combined results of all current polling suggest that the Gillard Labor government is now in a worse position than any past winning government - in terms of the amount by which it trails for the time remaining to the election - except for Howard's in 2001.

2.  If Labor does not have a 2PP position better than 46% by late March, it will fall behind Howard's 2001 recovery curve.

3.  It is also possible that very bad polling in the next few weeks would cause Labor to reach a position worse than that Howard recovered from in 2001.

4.  Being in a position from which recovery is without precedent or has few precedents does not mean a party cannot win an election.  However, it suggests they are unlikely to do so.

5.  This article also includes subjective waffle about Labor's current leadership and direction problems and the difficulty finding a clear and competitive way out of the mess the government has got itself into.

Update (25 Feb): It is now debatable whether even Howard's 2001 recovery is a valid precedent for recovery from the Gillard Government's current situation.



---------------------------------------------------------------------------------------------------------------------
Another two and a half polls went into the mix last week, continuing an unusual few weeks of polling.  Nielsen released a 56:44, Essential curiously dropped back a point to 54:46 and Galaxy released a poll of 800 female voters with a 2PP of 53:47.  (Newspoll followed its usual habit of postponing to avoid clashes with Nielsen, an action which always provokes a wave of conspiracy theories.)

Last year, a female-only 2PP of 53:47 would have been matched by a male 2PP many points higher, but what little evidence exists on the gender divide in polling this month suggests it has come down a lot.  The last two Morgans have averaged a one point difference, compared to an average of 7.5 points for the previous six, and the Nielsen breakdowns point to about a three point gap. The gender gaps in Essential's approval ratings are also modest.  Even assuming a three-point gap, the female-only Galaxy implies that a full sample would have been somewhere in the 54s, and that from a slightly Coalition-leaning pollster.  So we've had a rather odd two weeks in which the normally balanced or Labor-leaning pollsters have given savage readings for the government while the normally Coalition-friendly ones have barely moved. I am including a loosely estimated gender-adjusted fudge of the Galaxy female-only poll in my calculations (but downweighting it), and that is some part of my current 2PP estimate of 54.4 to Coalition (45.6 for ALP) being a bit more Labor-friendly than Bludgertrack (55.3) and the new Pollytrend (looks a bit below 55).  But the exact estimate doesn't really matter, because the picture across all the pollsters (and apparently across the dreaded "internal polls" too) is the same no matter how it is sliced: for the first time since last August, Labor is, for now, not even competitive.

We have also seen two polls now (Essential last week and Nielsen this) that have shown Gillard with worse netsats than Abbott, the first such since last August.  In the case of Nielsen the netsats have gone from -4 for Gillard and -29 for Abbott two months ago to -16 and -13 respectively.

Leadership speculation has been rampant for days (though it died down at the end of the week in the absence of fresh polling oxygen) and at present Sportsbet has Rudd a marginal favourite at $2.15 to lead Labor into the next election with Gillard at $2.30 and a rapidly closing Shorten at $4.  Sportingbet currently has Rudd $1.80 Gillard $2.50 Shorten $4.00 to be PM at the time of the next election (very likely, but not necessarily, the same thing).  It's always hard to tell whether bets on these things, on which there is a dearth of empirical information, are being placed by well-informed insiders (in which case they may mean something) or by less-informed punters who are overly reliant on media hype.  However, if you are confident Gillard will stay in the top job, and you are right, then you can make some pretty good money.  It seems there's plenty of belief around that Gillard will be retained, but not a lot of confidence in that belief.  This is not the first time Gillard has not been favoured to make it to the election as leader (see for instance this graph of implied probabilities based on betting odds here) but her fall from a position where her leadership was considered settled has been swift and dramatic.

A very bad Newspoll this weekend will certainly lead to another round of leadership feeding frenzies this week, while even a less bad result (like, say, 47) isn't likely to put it to bed for long unless matched by improvements in other polls.

We have even just seen the first for a while of a rather dodgy genre of polling (hence the fish), the leadership replacement poll.  In this case a Galaxy - more interesting for its measure of the current Queensland 2PP than the hypothetical exercise - purported to show federal Labor would increase from 45:55 in Queensland to 53:47 if Rudd was reinstalled.  This is meant to happen via a halving in the votes for both the Greens and Katter and the peeling off of five points from the Coalition, for a net primary gain of 14 points.  This species of poll is generally bulldust because voters are not very good at predicting (or in some cases being honest about) how they would vote under different scenarios.   I usually divide any claimed change in polling intentions as a result of them by five and even then prod suspiciously at the output.  Probably this one should be taken a bit more seriously than most because Rudd is a scorned favorite son of his home state in which the poll was taken, but there's nothing there to suggest Rudd will lift Labor immediately to a national election-winning lead of any durability.  Presumably if a shift in Queensland is only good for some fraction of eight points, then the impact of the shift would be minor in other states and perhaps even negative in Victoria.  Polls of this sort were also common in the leadup to the 2010 leadership challenge, and all of them were completely discredited by the time of the election, but that probably won't stop people taking them seriously this time around too.

Gillard has created a lot of her own problem through her many colourful denials of leadership chances in 2010: flying to Mars, playing full-forward for the Western Bulldogs, starring opposite Brad Pitt (etc).  With such denials so fresh in public memory, everything said by prospective leadership aspirants is scrutinised for what it didn't say as much as for what it did, and there is very little Rudd, Shorten etc could possibly say that is immune to being twisted by the commentariat as a sign of leadership interest.  If Rudd were to say nothing for a week, he would be suspiciously silent and "clearly" plotting the numbers in hiding; while if any possible candidate were even to take the Sherman pledge then it would be a suspiciously emphatic denial.

Some rather odd things have been said, and those who once hung on every decimal point of a dodgy Lindsay-voters opt-in are suddenly hardened poll-sceptics come to save the body politic from the demeaning blizzard of numbers - because the current numbers do not suit them, and the talk about those numbers suits them even less.  My favourite piece of curious logic was Simon Crean's claim that leadership change in result to bad polling is a "broken and discredited" model.  So, does this mean that the correct response is to keep the leadership team that by the same logic should not have been installed in the beginning - when an available alternative is to return a leader who by the same argument should not have been removed?  (After all, the model doesn't really apply to Rudd's original winning of the leadership.  Labor was polling quite healthily, on a 2PP basis, at the time that he replaced Beazley.)

Evaporating Precedents

Stephen Smith has been one of many pointing out that the government's polling position is by no means hopeless, in the light of past recoveries:

"At various times the Hawke, Keating and Howard governments looked done for but bounced back to win, he said."

But in terms of the amount by which the government now trails, and the relatively small amount of time remaining, the relevance of most of these claimed precedents is gone.

To start with, in the leadups to the three re-elections that Hawke won, he was never in serious trouble.  Apart from a few bad months in mid-1985 way out from an election (at worst, around 46.5% 2PP), a rather roguish looking Newspoll in September 1986 (that comes out to a 46:54 2PP) and the odd poll with a slender 2PP lead to the Coalition in mid-1989, Hawke was typically no worse than equal.  The publication of just primary results in those days may have made it look like Labor was up against it, but it wasn't real.  In the term where the party under Hawke ran into major trouble, he was removed - not during his first batch of crisis polling in early 1991, but after recovering to 47s and 48s later that year and then falling back into terrible numbers again.  (Sounds strangely familiar.)

Within three months of the shift to Keating, Labor's polling improved markedly, and while the party still trailed for most of the election leadup (as mentioned before), at no time in the twelve months leading up to the 1993 election was Labor as far behind, on average, as Labor is now. No doubt the subsequent election win was just another example of the "broken and discredited" Labor model of changing leadership in response to dire polling.

That leaves only Howard, and really only 2001.  Recently Peter Brent posted this graph which showed the ups and downs of the Howard and Gillard governments approaching election time.  The graph shows that prior to the introduction of major and contentious legislation (the GST in one case, the carbon tax in the other) both governments suffered.  Both lifted after the legislation was enacted without the sky falling in, but both nosedived early in the election year.  The Howard government started to get its act together from about May and took the lead in late August - early September following the Tampa, having trailed narrowly before it.  The S11 attack then deprived Labor of any oxygen in the last two months of the campaign, although the massive bounce it gave Howard washed out of the system by election week.

One of the few things you don't get from the Brent graph is an impression of the comparative urgency of Labor's plight.  After all Howard mucked around in the mid-40s for months; does this mean Gillard might do the same and yet still win?  The difference in the comparison is that the 2001 election was held two months later.  At the equivalent time Howard had only just hit the front.

The following is a very rough graph of my comparative Newspoll rolling 2PP averages (designed to blunt poll to poll bounciness, at the cost of a fair degree of lag), graphed by months to go to the election.  Where there was no poll for a given "half-month", or two polls just to either side, I've just averaged the surrounding values. 

[This graph is now updated to include the Newspoll released 25 Feb. See update comments at end of article.]

The comparison above is generous to Labor given the results of other polls (which suggest the red line has gone much closer to the blue one in the past few weeks) and given that the most recent Newspoll is a few weeks old.  It doesn't really matter, since to keep up with the Howard line, Labor will need to be above 46, more or less consistently, by about late March.  That is, assuming Labor doesn't get some more really terrible polls and fall below the Howard line before then (a 43 or worse this week would do the job immediately.) 

For the time being it can still be said that recovery to win from this sort of polling position this far out is not unprecedented, and even if it becomes unprecedented that does not prove that it will not occur, nor even that it would be a miraculous fluke if it did.  But with only one precedent, and with most of the six governments that have lost since polling started having been better placed than this one at the same time, the government's polling position is now, by the standards of polling history, undeniably grim.  And it's just the last few weeks that have made all the difference to the comparisons.  If Labor recovers, again a fairly sizeable "if", it may come inside the recovery curves of other past winning governments once more. 

Indeed, until quite recently, the current government was polling worse than Howard in the 2001 leadup at an equivalent stage.  But at those stages, it was polling on a par with other governments that were successful.  Since the recent drop, this is no longer the case. 

Leadership Vicious Spiral

The current leadership speculation has only come about because of the latest raft of terrible polling for the Government.  When the party polls badly, the number of backbenchers feeling jittery about possible loss of their seats, and willing to say intemperate things to journalists anonymously, rises.  But there is a potential vicious spiral here, in that bad polling leads to leadership speculation, which leads to instability and a difficulty focusing on selling the Government's message, which (potentially) leads to more bad polling, rinse and repeat.  It can't have been what the PM was hoping for when she declared the election date many months in advance hoping to obtain a focus on policy rather than an endless distracting focusing on when the election will be called.  So there is no endless distracting focus on the election date, but instead there is an even more distracting focus on bad polling, leadership speculation, and other things that also make it hard for the Government to get "its message" heard.

I really don't know what the ideal solution for Labor is, because none of the solutions to the party's situation look too promising at present.  The present leadership team seems more or less discredited, since (for instance) you can only tell the people there will be a budget surplus so many hundred times, then apparently not deliver one, and expect them to believe a word you say.  Likewise, while the Mineral Resources Rent Tax earnings flop is bad enough if it was simply a case of dud forecasting, the more concerning possibility is that Gillard/Swan deliberately accepted a soft version of the tax with the intention of making it look like they could pay for all commitments without deficit, but while having a fairly good idea that they would not.

Not much I have seen of the key ideological planks of Labor's re-election strategy (to the extent that it exists) looks promising yet, and the recycling of "working families" as "modern families" (while the Coalition has moved on so dramatically from "Real Action" to "Real Solutions") seems to play directly into the Opposition Leader's hands.

Abbott has been running a big family-policy area through his paid maternity-leave scheme for some time, and he just loves an opportunity to trot out his family-man credentials, especially since they remind voters that he is surrounded by women.  That the family lives of leading politicians are generally very well-resourced by the standards of most voters, and hence atypical, will be a detail lost on the electorate.

Perhaps the most bizarre aspect of the "modern families" slogan is that there are even suggestions that the supposedly "inclusive" (except of those who are not parents!) slogan would help with the gay vote.  The "gay vote" leans heavily to Labor anyway (on a 2PP basis, that is, since the Greens would get a lot of it in primary terms), but in terms of mobilising, energising and enthusing it (so that, with luck, it convinces its straight friends as well) there is something extremely obvious that could have been done, that Barack Obama had the courage to do, and Julia Gillard did not. (And nor for that matter did Kevin Rudd.) If you are going to pitch convincingly to modern families, you need modern attitudes.

A return to Rudd is a fraught exercise.  It could be successful, but the risks are enormous.  It would probably lead to a large leadership transition bounce, but as Peter Brent observed recently, there is a sense in which such polling bounces are not real.  Between elections, leadership transition bounces may appear to last several months, though much of it just represents recovery from the damage caused by the previous leader to polling at the end of the previous leader's reign.  Labor tried to surf a leadership transition bounce to a big win in 2010 and it didn't work, and in the end was almost disastrous.

Quite aside from the possibility that a return to Rudd might simply result in the government collapsing (more likely from the inside than the crossbench), I don't see how Rudd as leader would be able to just shrug off all the vicious attacks against him by fellow party members that are already on the public record.  That is, unless all those involved could be marginalised and discredited, and the culture of the party substantially transformed, something that is unlikely to be accepted without a massive fight.  It's the work of a decade, and even with the option of postponing the election date slightly, Rudd would only have several months to get it done while at the same time campaigning against the real Opposition.  If Rudd wants the leadership now, the question is whether he would want it as a platform for a serious attempt at victory, or simply as an exercise in personal closure and vindication in which a narrow loss is an accepted endpoint. Without radical and rapid change, how would the party convince the electorate that if Rudd was re-installed to win the election, he would not just be removed again some time after it? 

The spirit of the times here was very nicely captured in a letter from Dave Watts, Clareville, to the SMH:

"The thought of Tony Abbott as Prime Minister is unbearable. Desperate times require that Kevin Rudd is reinstated as Labor leader to win the election, then removed again in favour of Julia Gillard for the same good reasons as previously."

Third-candidate suggestions involve either those who would supposedly save the furniture or putting a prospective future leader in ahead of their time.  For some reason Simon Crean is considered with reverence in some quarters under some unusual Labor standard in which you become a wise old legend by hanging around long enough after failing.  I think this sort of thing is silly and I'm not even convinced it would actually limit the size of any loss.  When parties beg for mercy, voters usually just kick them harder.

The growing idea of putting a future prospect in before the election is a very left-field one which may not be quite as silly as it sounds, especially as it at least may end the Rudd/Gillard controversy.  The usual view is that Labor would lose, and that the loss would damage the future prospect's chances, and that on that basis that prospect would be stupid to even want it.  However, in the case of Bill Shorten, the conventional wisdom is that he's going to be the first Opposition Leader if Labor loses, which has historically been a miserable experience in any case. 

I'm a mild Shorten-sceptic at this stage.  I know he's articulate, and seen as pretty bright, and has been a strong ministerial performer.  (These things also apply, probably more so, to Greg Combet, who would probably already be the leader on merit if Australia had a more European-style political culture.)  My first reservation about Shorten is that I don't know if the Australian public sees him as anything more than some union dude who was at Beaconsfield and had an incident in a pie shop.  Supporters seem to think he is Bob Hawke the Second, but in terms of personality, where's the evidence?

My second is that Shorten seems to be the embodiment of the union-based faction-hack culture that causes the ALP so many of its ongoing problems - from self-advancement networks that are prone to outbreaks of corruption, to leadership decisions influenced by unionists who are not even sitting members, to a tendency to retreat to class warfare at times when the party should be broadening its base and reaching out to moderate voters who do not come from a labor-movement ideological background.  Shorten has certainly talked the talk in the past in terms of broadening the base, a useful asset at a time when Gillard and Swan seem too willing to risk contracting it, but I really wonder if anyone who is so closely associated with union culture can do what should be done to modernise the Labor Party and reduce its electoral weaknesses.  

It is no permanent obstacle to electoral success should Labor fail to modernise itself and retain a disproportionate reliance on the union movement, which should really be one of the planks of a successful modern broad-left party, not the primary one.  Both major parties have entrenched weaknesses that seem to be just all too hard to fix.  At times when each party is on the nose, those weakenesses seem like a recipe for spending decades in Opposition, but soon enough the other party in government finds ways to be worse, and the Opposition gets voted back in anyway.

Finally, this article by Paul Williams is probably the very worst attempt to mathematically analyse a proposed leadership transition that I have ever seen!  I can only hope that it was written tongue in cheek.  There are plenty of opinions but there is really no valid scientific basis for projecting what a switch back to Rudd might actually do. 

This piece will be updated following the coming Newspoll (results of which I expect on Monday night) and perhaps other polls in the coming week.

Update (Monday midnightish):  Newspoll came out 55-45.  While some observers might have expected this Newspoll to be even worse than the previous, it's no surprise it is not, since the previous Newspoll was much harsher to the government than other polls that week.  Indeed in 30 previous cases where a government had lost 5+ points in a single Newspoll, as occurred last time, that government had gained points in the subsequent Newspoll 24 times, stayed level 4 times, and gone down again only twice.  (The latter two cases were in March 1995 during the bounce for Howard becoming Opposition Leader, and in June 1996 early in Howard's term as PM).

On the graph the 2001 and the 2013 lines are now extremely close, and it doesn't help the 2013 side of the comparison that the Newspoll rolling average method I use is rather conservative, and that my current all-poll aggregate estimate (now 55.1:44.9) is even harsher for the government.  There's a strong argument that the red line is actually now below the blue one, but it's also worth noting that in trying to fit data to half-month comparison periods I smooth out the 4-6 May 2001 poll, in which the 2PP was 56-44 to Labor and the Coalition's rolling average via the formula I'm using dropped to 45.2.

The Newspoll also saw Gillard record her worst netsat (-28) since August and Abbott his best (-22) since the same time.  It is all, suddenly, as if the government's recovery in the last four months never happened.

Essential went back to its normal recent Coalition side of Newspoll (see here) with a 56-44, representing only the sixth time since the start of 2001 that it had moved by more than one point.

Update (26 March): This piece is updated in: The Gillard Bounce Barely Existed - What About the Spill Cliff? At this point, the line was finally crossed.