Thursday, March 7, 2013

Yet Another Western Sydney Article

Advance Summary

1. The Western Sydney area has received great attention in commentary on the 2013 election although the widespread view that it will decide the election result is almost certainly wrong.

2. Recent ReachTEL polling of some Western Sydney confirms suspicions that very large seat swings are possible in the area.

3. The polling is also notable for returning much better results for Tony Abbott than national polls, and very bad results for Julia Gillard, including in comparison with Kevin Rudd.

4. The polling is not a post-event verdict on the Gillard Western Sydney visit as it was taken just before it.

5. Although voters voting for candidates/parties other than Labor, Liberal, Green in 2010 were Coalition-leaning, there is some reason to suspect (even from very small sample sizes) that self-styled Others voters in ReachTEL polling are less Coalition-leaning than Others voters from 2010.

6. In a quiet week of federal polling, an interesting if messy new development is a Morgan "multi-mode" poll with over 9000 respondents.  However, as this poll is impossible to benchmark for possible house effects because it is a new poll form from that pollster, it is difficult to make use of it yet.


The Usual Background

Western Sydney, an area of tediously intense focus during the 2010 election (among others) has been so much in the spotlight this week that you're forgiven if you switch off as soon as you see that this article mentions it.  It's of current interest in the leadup to the 2013 election partly because there is now quite a deal of polling (not all of it peddled in shady alleys by dubious sources saying "Pssst! Wanna used opinion poll?") suggesting that Labor is facing monster swings there, larger than the NSW state average, with even seats on low double-digit margins now in play.  ReachTEL polls conducted on Feb 28 showed Labor trailing slightly in Chifley and Blaxland (12.3% and 12.2% swing needed) and not even competitive in McMahon (7.8%) and Werriwa (6.8).  I have some differences with some of ReachTEL's 2PP preferencing assumptions, which I find often give a more Coalition-friendly figure than I believe would be the case off the same primaries, but the magnitude of the results is such that it makes very little difference.  Morgan has also compiled results from the past two years showing Labor facing an average c.8% swing in Greenway (0.9%).  If that was mostly Face-to-face, it might be even more.

Peter Brent has been pointing out for ages in many articles that the line that Western Sydney will decide the election is, in all probability, rubbish.  Seat swings are, historically, distributed normally around a national average, meaning that if a local brace of seats is swinging against the government by double figures, then that's very likely because there is a substantial national swing against the government anyway, meaning that it will lose office on the basis of seat losses in other areas alone.  If, on the other hand, there is no national 2PP swing, but there are moderately large swings in Western Sydney, this is likely to be because Labor is picking up seats with swings in its favour elsewhere, especially Queensland.

Another recent article by Gordon Graham makes the point that if Labor is going to make a serious attempt at winning, then picking up seats in Queensland and elsewhere is at least as important as saving seats at risk in Western Sydney. The suspicion, then, is that a Labor strategy that is intensely focused on one area is a strategy that's premised on defeat and damage-limitation.

 (Pedantic note: Labor only necessarily must win the 4-5 seats mentioned by Gordon if it wants to win majority government. If every seat in the current parliament was retained by its current owner except for Dobell and Fisher reverting to Labor and Coalition respectively, then we would have another hung parliament and another Labor government for much the same reasons as before.  But with Lyne and New England at great risk of falling to the Coalition, and given that there is usually seat transfer in both directions (especially if an election is at all close) we can very safely assume that if Labor is to retain office in any form, it will need to take at least a handful of Coalition-held seats somewhere.)

To me, one instructive aspect of the Western Sydney focus is the derogatory and uniform stereotyping of voters who live in the area by many political chatterers who don't.  The stereotype of a region which, like all large regions, is far from being the same all over, is that the Western Sydney problem concerns "aspirationals" - voters, sometimes derided as bogans, who mortgage themselves to the hilt to buy houses and raise families in the proverbial boondocks because they can't afford to buy anywhere else, then find they are tremendously vulnerable to infrastructure problems, fuel price spikes and general cost of living crises.  This in turn ties in with the economically-motivated xenophobia that both parties are trying to play up to at the moment.

But what I find instructive about all this is the extent to which posters on forums and social media and so on, blame those who live in Western Sydney for the government's campaign focus on them, instead of just blaming the government for its preoccupations with its own heartland mythos.

The Unusual Poll

Anyway, this week the Prime Minister has staged a much-publicised voyage to Rooty Hill, the need to be accommodated there overnight seeming to be tacit admission of the transport problems in getting into and out of the region quickly.  And a very handy Channel 7 - ReachTEL poll says some useful things about attitudes to leadership in the region, some of which may also shed light on the identity of ReachTEL's sometimes high "others" vote.

The poll was taken on March 1.  There's a trap for beginners right there, because the poll actually pre-dates the PM's artist-in-residence tour, and so it is a snapshot of attitudes to the tour and the leadership before it happened.  Despite this, we get bizarre interpretation such as:

"Prime Minister Julia Gillard’s controversial five-day visit to key Labor electorates in western Sydney could have had a negative effect on her election campaign."

So a visit could have had a negative effect before it's happened and this can be known before it's finished?  Hmmm.  We'll have to wait to see what the net post-visit impact (if any) will be, not only in the area but elsewhere.  Not that I'm expecting roses for Labor there.

The poll includes the Labor-held electorates of Blaxland, Chifley, Fowler, Lindsay, Greenway, McMahon,  Parramatta and Werriwa, the Coalition marginals Macquarie and Macarthur and the utterly safe Coalition seat of Mitchell. Collectively this lot returned 53:47 in 2010 and there's surprisingly no figure given as to what they're returning now, but with 61.3% of the surveyed voters preferring either Abbott or Turnbull as PM to Gillard and Rudd, it's obviously not in Labor's favour.  However, seat betting markets either don't believe the earlier ReachTEL, or think Labor will turn  it all around, with only three seats currently favoured to go (two of them heavily) and Werriwa thus far lineball.  Indeed, the seat punters haven't believed either the commentariat or the available seat polls and believe that Labor will suffer as many or more losses in the rest of NSW than in Western Sydney. That said I suspect that seat markets take a lot of convincing that seats on big margins may fall.

The poll returns a massive 74.5:25.5% preferred-PM lead for Kevin Rudd over Julia Gillard.  Regular readers will be well aware I've rubbished the constant pundit/media focus on PPM scores that compare Prime Ministers or Premiers with Opposition Leaders in terms of voter preference for a job one holds and the other does not.  PPMs involving alternative leaders of the same party, or PPMs involving multiple alternative leaders, are a different story, but they too have their hazards.  So when looking at a preferred-PM poll between leaders from the same party, we shouldn't pay too much attention to the preferences of the Opposition voters, since most of them won't vote for the government whoever its leader is.  There's also a tendency to demonise the current leader of the opposing party and to say that leader is the worst leader that party has ever had.  So the finding that Liberal voters prefer Rudd by 83.1:16.9 is hardly news.  The most interesting findings there are (i) that Labor voters in Western Sydney have a marginal preference for Rudd (ii) that voters who give their preference as "Other" are the strongest Rudd-preferrers of all (84.5%).

On the opening question of whether voters would be more or less likely to vote Labor under Rudd, Labor, Green and Others voters in the area give Rudd a big thumbs up (the Others most of all), while Liberal voters claim to be more or less split.  It's possible that some Liberal voters are being cagey here - they don't want Rudd back as they think he is more dangerous than Gillard.  I certainly suspect the honesty of some of the Liberal respondents to Question 3, in which a staggering 60% say that Gillard's visit makes them less likely to vote Labor.  Many of these, in my view, would not even be considering voting Labor and are simply expressing negativity about the Prime Minister and her visit.  But again here we see the "Others" voters, who we already know want Rudd back, regarding Gillard's impending visit almost as negatively as the Liberals.

Then, there is a 4-way preferred Prime Minister question.  This is the second poll I know of that has asked voters to pick a leader out of Abbott, Turnbull, Gillard and Rudd, the first being a national Galaxy early last December. The following table compares that national Galaxy (Gal) to this Western Sydney ReachTEL (RT) after excluding undecideds from the Galaxy:

*The Rest figure for Galaxy is an approximation derived by me.  Thanks to the Hobart Mercury for publishing the Labor and Liberal breakdowns of the Galaxy.

We would expect Abbott's position in the ReachTEL to be somewhat better than in the Galaxy and Gillard's to be somewhat worse, even if the ReachTEL was a national sample.  We have no real basis for knowing what the story would be with Rudd and Turnbull on that basis. Here are some things we can determine from this comparison:

1. Western Sydney voters are much more pro-Abbott than voters nationally (the national change since December would not account for such major differences.) This is especially because Western Sydney Liberal voters are very pro-Abbott.  This is no suprise; the rich, moderate, small-l liberal Turnbull just wouldn't be expected to have much appeal to Liberal voters in the outer city.  It is also consistent with the four ReachTEL seat polls, which all found Abbott with slightly positive netsats (+3 to +14) with Gillard's netsats in the -30s to -40s. Over 40% of voters in all four rated Gillard's performance as Very Poor.

2. The levels of support for Rudd and Gillard in Western Sydney overall, and among Labor, Liberal and Green supporters in that area, are about what would be expected following the Galaxy poll given Gillard's current difficulties and given that the area is pro-Abbott.

3.  Western Sydney Greens voters don't rate Abbott any more than Greens voters in the rest of the country.  Probably even less.  But probably even some of those who would preference Labor prefer Turnbull as PM to Rudd or Gillard.

4. The Others category is the most interesting.  The leadership preferences of this group of voters is similar to that of the Green voters, except that their split of the Coalition vote is somewhat less pro-Turnbull with a modest share for Abbott.

ReachTEL's "Others": Who Are They?

In my recent and critically acclaimed* update to Silly Lilleys: Is Wayne Swan Losing His Seat? , in referring to those who pick "Others" in ReachTEL's polls, I argued that:

 "the Others column probably includes voters who are basically Labor-friendly but just don't like Julia Gillard and are therefore parking their vote in the Others column (although they would probably vote Labor or perhaps informally at an actual election.)"

In this, I think that picking the option "others" is for some poll respondents just a glorified "don't know" or "don't want to say". 

The following is a table of the tiddler parties that actually contested these 11 Western Sydney seats last time around.  The Others vote ranged from zero in Werriwa (because there were none) to 10.3% in Greenway (where there were many). 

Here I've given the total share of the Western Sydney vote attained by each of these famous electoral behomeths (the Christian Democrats far and away the most successful with 37% of the micro-party total) and their national Coalition 2PP shares.  I've used the national figures rather than local partly out of laziness but also because for such small totals it's possible the luck of the donkey vote could mess about with it a bit.  The exception is for Independents, for which I used the 2PPs for the actual contesting Western Sydney candidates.

The point to be made here is that actual Others voters in Western Sydney were tending to vote for Coalition-leaning parties more often than not in 2010, and this in electorates that, on average, Labor won 53:47.  Now, with the same electorates preferring one or the other Liberal as Prime Minister 61:39 and picking Tony Abbott almost twice as often as Turnbull, the self-styled Others voters prefer the two Labor PMs by insignificantly more than the Greens voters do, and were much more likely to prefer Turnbull than Abbott. 

Now, I'm not quite sure what a "Christian Democrat" [double sic] does when asked to make a choice between two Catholics (one of whom supports same-sex marriage), a PM (who does not believe in God) and Kevin Rudd (who thinks he is God!)   Pray for divine assistance, maybe? And I doubt the Others sample size here was large - perhaps only something like 60-70 voters.  But again, the ReachTEL Others don't have anything remotely like the leadership support profile that would be expected from the way that Others voted in the last election.  And I believe this is because they are substantially not the same set of voters. A possible contributing factor is KAP picking off some of the right wing ones, but I think there is more to it than that 

The National Polling Scene: The Calm Before ...

A quietish week for national polling so far as we await the resumption of the House of Reps and the most critical, decisive and eagerly awaited Newspoll since, oh, the one last week.

This week has so far seen only two national polls, from Essential as usual and from Morgan.  Essential sat on 56% after its two-point jump last week; though it may not have been far from hitting 57.  Anyway my own aggregation attempt called that a 55 and didn't budge.

The strange one was the Morgan - a monstrous (in more ways than one) "multi-mode" poll, combining Face-to-face with internet polling, and having a sample size of 9101 (that's not a misprint) and a 2PP (by last-election preferences) of 54:46.

Morgan continue to find new ways to confuse.  Presumably this one is more internet than face-to-face, since face-to-face is more expensive, and if the internet sample was several thousand then it would be better to just publish the two results separately so that benchmarking would be possible for similar internet-only polls in future.  As one of the sampling modes used in this "multi-mode" is known to have a large house effect, and as we don't know the story yet with the other one, it's impossible to know yet what to make of the new data.  If this sample method continues then through the year it will become possible to benchmark it and use it prominently in aggregates but for the time being we will have to wait.

Another contribution on the one of my pet subjects, historical recoveries from lousy polling, came from the Insiders Poll of Polls aggregator segment.  This very nicely presented review started with an aggregate of 55:45 to Coalition and then went on to point out polls in which PMs Keating, Howard, Howard and Howard (1998, 2001, 2004) had been trailing by similar amounts and yet won.  However, this attempt to say that views on what will happen are just pure opinion on this basis had a flaw, which was rapidly sprung:

The point: a month of bad polling is worse than a single bad poll.  (Additionally, Brent's more accurate 2PP calculations give that October 1992 2PP as only a 53:47, though it was followed by a 54:46 before the Coalition fell off an 8-point cliff.)  To compare present positions with past recoveries it's necessary to compare like with like.  I've done a crude version of this using just Newspoll and a rather sluggish rolling average method, eg 2001: The Final Frontier, and on that basis most of the claimed precedents disappeared.  An aggregator using old data from all polls might get a slightly different result - or not. But the item is worth watching anyway.

* Bitchy PS (which most readers can safely ignore): One kindly reader - not some random idiot either - accused me re the Lilley article of ""my politics lead my analysis" horseshit" [..] ""ifs, buts and maybes" from people that should know better", [..] and "this same thing" "the tories did" in 2007.

What's funny about this - apart from the suspect's cluelessness re relevant aspects of my political preferences between different Labor figures - is (i) his attack is hopelessly conflicted, since he chose to pick on an article where I'd dissed a poll commissioned by a close ally of his employer (ii) his own recent federal Twitter form has been infested with one-eyed Ruddery (if that's a case of analysis leading politics instead then the analysis is suspect) and he needs to get the beam out of his own eye first!

But seriously though, if an analyst thinks they have evidence that another analyst is basically a partisan fraud in their considered published work, then the accuser should do that case justice by posting a considered argument to that effect and granting the target unlimited right of reply.  Or even post it here in reply to what you're criticising.  Aggressive sniping with semi-joking, poorly-substantiated, belittling accusations on social media is shabby form and should the suspect behave this way again, he can spend a week in my killfile along with his new friends @istand4marriage, @marriagedefence, @marriagedefense and @stop_gay_sex_ed! 


  1. All serious analysis aside, they're derided as bogans because they are bogans. Western Sydney consists of two basic demographics: Rudd's "aspirationals" (Howard's "battlers") and the underclass. There is little else to be found west of Parramatta and that's speaking as someone who spends a fair bit of his time in those parts.

    The mortgage-belt, aspirational (bogan) voters tend to swing and that's where the contempt of inner-city luvvies like me comes to the fore. Given that the major parties really haven't changed their basic platforms awfully much in decades, swinging voters act out of narrow self interest and parties tinker around the edges in order to appease them.

    Perhaps they don't deserve my contempt for that but they certainly deserve my contempt for coming into my neighbourhood on a Saturday night and starting fights where none previously occurred, but that's another story altogether. At least the underclass hogans are inclined to leave everyone else alone ...

    On a more serious note, Western Sydney may not be truly crucial in the coming election but I'm prepared to bet that the party that prevails in that neck of the woods will almost always win (as mentioned by you, almost certainly because they're doing well elsewhere and particularly in other marginal electorates). Because of the high concentration of swinging voters, I'm also guessing that it's a pretty good laboratory to try out electioneering strategies, although it would definitely seem that all Labor has left is "give Rudd a go - it's not as if he's likely to do any worse", even if that will be regarded with considerable suspicion by an increasingly cynical electorate.

  2. At the moment the question I'm pondering is whether Labor is going to make a serious attempt at winning, or whether they will try for a stop loss. Under stop loss conditions, Western Sydney is significant. Under winning conditions, not so much. Focussing on Western Sydney pushes the case that this is a stop loss election for them; if the focus remains or returns there after this 'expedition'...

    The other thought I've has is that this 'expedition' is more to do with the rest of the country than it is with Western Sydney. For the obvious reasons, it is difficult for the PM to campaign in Qld and WA. But there are probably a goodly number of people in those places who perceive of themselves of living somewhere 'like' Western Sydney, whatever their impression of that is.

    Oh, and on a less serious note - considering Western Sydney as an amorphous mass is probably disingenuous to start with, not that the bog warmers east of Strathfield would notice. It's an incredibly diverse area.