go to next posts Nicholson in The Australian

welcome to               Mumble 2004: Go to post redistribution pendulum

September 25 Nielsen soars to Howard

In a week commentators almost unanimously scored to Latham, first TNS in Makin, then Morgan(!) and now Nielsen show the government moving ahead. It goes to show that no-one really knows exactly what makes voters do what they do.

Here's the table, the Age and SMH: 54 to 46. Even Nielsen pollster John Stirton seems doubtful.

September 24 Morgan: 51.5 to 48.5   !

Malcolm Mackerras on the Senate, always required reading.

'Tiser in Makin: 48 to 52

Leaked polling allegedly has Labor in good position. Here and here. Appropriately vague.

September 22 The other Malcolm

Crikey newsletter the other day carried this brief item: 

Know your demographic
Some enterprising soul has been at work on a Malcolm Turnbull poster on Edgecliff Road. Rather than reading "Putting Wentworth First" it now says "Fisting Wentworth".

But ACNielsen in the Age and SMH suggests the roles reversed. The survey in Wentworth shows Malcolm Turnbull in danger of losing - though not to Peter King, but the Labor candidate, David Patch. (SMH incorrectly declares King "is directing his preferences to Labor".)

Nielsen results: Turnbull 34 Patch 27 King 25 Greens 12.

Eliminating everyone but the top three gives Patch 36, Turnbull 35 per cent and King 28, which goes to 50 50 two party preferred, a roughly 50 50 split in King's preferences.

The last election saw primary votes of Lib (King) 52.1, Labor 29.5 and Greens 9.8; two party preferred: 57.9 to 42.1. So Turnbull and King's combined Nielsen vote is about 5 more than King's in 2001, and Labor's primary vote is down marginally.

The amazing thing is King's preference flow. I'll believe it when I see it.

September 21 Morgan in the Senate

Morgan, alone again - naturally - sees the Coalition going backwards - "losing up to five seats" - in the Senate. (Everyone else sees the government doing well, possibly ending up with half the 76 positions as its poor 1998 result finally recedes.) Readers may notice I stay away from the Senate as it hurts my head, but a certain Mr Bludger gives the whole thing a jolly good going over today. 

Marky Mark gets a Newspoll lift

The first post-debate Newspoll has Labor ahead 52.5 to 47.5 (to, uncharacteristically for Newspoll, half a percentage point).

While we're on the topic, it's ...

Time for a Newspoll graph (about Crean and Latham)    

This is about how hapless Simon Crean (but not his successor) got the raw end of Newspoll. Long-term readers may recall regular scolding of the pollster for its lackadaisical attitude to preferences. The story is rather tedious but much of it can be read here.

The upshot is that until February 2003 Newspoll didn't bother with preferences (apart from during election campaigns). Only primary support was measured - and published - which, without taking into account the unprecedented Green support which would flow overwhelmingly to Labor, made poor old Simon Crean look even worse than he was. (See this.)

In Feb last year they began calculating 'notional' two party preferred based on preferences at the 2001 election, but only using the aggregate preference flow from non-major parties. (That is, 58% to Labor, 42% to Coalition.) Better than a poke in the eye for Crean, but still not doing him justice, because the make-up of those non-major votes had changed dramatically - One Nation had collapsed, while Greens were firing ahead.

Then, from January this year, for the first time ever (outside a campaign) Newspoll began asking respondents for preferences (albeit just second ones). This coincided with a new Labor leadership. 

The point about all this is the notional numbers dudded Simon Crean, because they nearly always under-stated Labor's two party preferred support.

Newspoll 2004 - two Newspoll methods

The graph above shows Labor two party preferred Newspoll numbers throughout 2004 - beginning in January and ending last weekend. (Only Labor's is shown; the Coalition's at any point is, of course, 100 minus Labor's.) The red line is the number published in the Oz - that is, by asking respondents for their preferences. The second, purple, is what would have been published under the old 'notional' method if the poll had been taken in 2003. (During 2002 no two party preferred were given by the Oz.)

Today's Newspoll, 52.50-47.5, would have been called 50-50 if it had been taken last year. And the two before it, both 50-50, would have been described as 51-49 and 52-48 respectively. Most of the earlier ones would have been understated too (where the purple line is below the red one).

Crean wasn't much of a leader, but Newspoll and the Oz made him look worse than he was. Now, I don't have to tell you that Newspoll is the survey most watched (especially mid-term, when apart from Morgan no-one else does it regularly) and what it says (and how Mr Shanahan and the Oz interpret it) ripples right across the commentariat. 

(Read the fine print.)

Phillip Adams (1) ingratiates himself in case of a good election result, and (2) adumbrates the company line if there's a bad one - better to die on your feet than live on your knees. Strange, because Latham is a bigger serial capitulator to Howardism than Beazley ever was, and on mandatory detention, which Phillip feels strongly about, Latham's complaint about Beazley was always that he wasn't tough enough.

• Gerard Henderson - a commentator who actually understands preferential voting.

September 20 (evening) Malcolm leaps off the fence

It's 7pm, and Malcolm Mackerras was just on ABC Radio National. Til now uncharacteristically prediction-free, he now expects a Coalition majority of "about ten seats" in the House of Reps. There, he's said it. Malcolm never does anything without reference to his pendulum, which means he expects about 50.5 to 49.5.

Glenn Milne returns to work

The Australian's Glenn Milne often does a good impression of inscrutable non-partisan reporting .. and then there are columns like today's in which he thunderously warns that the Green-Labor preference deal makes 'the already strong likelihood of a hung parliament even likelier' - but never gets round to explaining why. (Let's reiterate, a hung parliament is very unlikely - and the preference deal makes it microscopically less unlikely. To make such a difference it would have to result in a Green win in a seat currently held by the Coalition.)

Morgan: 53 to 47

Galaxy poll in Courier Mail: and Adelaide Advertiser: 51 to 49

• Electiontrack might be worth a look.

September 19  Green preferences ado about nothing

Here we go again, a vital deal for the ALP in getting Green preferences in 30 seats.

From the Adelaide Advertiser, the marginals included are Hindmarsh, Makin, Solomon, Deakin, Richmond, McEwen, La Trobe, Dickson, Longman, Page, Bowman, Brisbane, Stirling, Hasluck, Banks, Lowe and Rankin

Most preference commentary implies that Bob Brown can actually direct them in the House of Representatives. He can't; all anyone can do is make suggestions on 'how to vote' cards. Then it's up to the voters, and very few Green ones follow them.

This Parliament House research note of a couple of years ago looked at parties' success in directing preferences, and found the Greens least successful of them all. "For the Greens, in divisions where preferences were directed to the ALP, its share of Green preferences was 3.5 percentage points higher than in divisions where no direction was given."

That's 3.5 percent of the Green vote, mind you, so in an electorate where the Green candidate, to be optimistic, gets 10 percent of the vote, we're talking about 0.35 percent of the total vote. Obviously, that can make a difference if the contest is very tight, but that's not many electorates.

Green voters, in other words, tend to think for themselves - and about three quarters will preference Labor ahead of the Coalition come what may. 

Once upon a time, replacing the first word in "Compulsory Preferential" with "Optional" was high on Labor's to-do list - they got it in NSW and (later, under different circumstances) in Queensland - to split the conservative vote. They are so lucky they never succeeded federally: under OPV Mark Latham would be heading for a dreadful hiding next month.

Senate preferences (as the Greens are getting from Labor in exchange) are a different story, because something like nine out of ten people vote 'above the line', so putting the flow in the hands of the parties. Good deal for Greens. 

Read also this.

Some people will believe anything Tony Abbott and Joe Hockey allegedly in trouble in their posher than thou North Shore Sydney seats. The 'Doctor's Wives' syndrome, apparently. Possibly the silliest report of the week.

September 18  Nielsen says 51 to 49 here and here.  Graph at right is from SMH site, as is this largish table. Nielsen came closest of any major pollster to the actual result in 2001. Unlike Newspoll - but like Morgan - Nielsen measures full preferences. Speaking of Newspoll, the Oz has them in the Coalition marginals. Despite the headline, it shows Labor making some ground there since last election.

• A 'Tiser poll shows His Lordship safe against Brian Deegan.

More on the pollsters

Following yesterday's post, a reader asked "How is the Morgan Poll so out of whack with the others? The method might have something to do with it normally but a phone poll should surely give results more like the other polls. Are the questions different, or does it have more to do with way preferences are dealt with?"

Answer: not sure. I wrote about the polls in the AFR last year. Morgan and Nielsen ask for full preferences - full marks. Newspoll just asks for second ones and calculates from there. Not too bad.

Around the time of the Queensland election an email from Antony Green contained the following:

"By the way, I think you are a bit tough on Newspoll. I am not exactly sure that asking preference questions in every poll necessarily gives you better data. Past research has shown that people make up their mind on the Senate and on preferences later than they do their main lower house vote.

Remember that Newspoll and Morgan have always asked primary vote questions differently. Morgan offers a list of minor parties, Newspoll ask Labor, Coalition and other and then ask a second question as required. Traditionally Newspoll has had a lower minor party vote as a result."

My spy at Nielsen sent me these questions earlier this year. I don't have the others'.

But Weighting is probably the main explanation. All the pollsters adjust their results (or, at least, their samples) to make them more 'representative'. They all have their own methodology which they keep close to their chests and Morgan probably does something different to the others. That's my guess. As indicated earlier, there's a small chance (1 in 10?) that Morgan was the most correct in 2001 and things changed drastically in the last few days.

We'll probably find out on October 9.

September 17  The always delightful Mr Morgan, post debate, says 54.5 to 45.5

A telephone survey; Morgan's specialty is the face-to-face which he keeps insisting is superior.

Roy Morgan Research has been in the doghouse since the last election, which they got wrong. The other pollsters Shadenfreuded all over the place, but in truth none had much to be proud of. Yes, the rest correctly predicted a Coalition win but they all overstated the margin.

Morgan has always maintained (as he would) that his polling was right and that images of boats and chaos in the final days turned it. It's not out of the question, and if Morgan really is the best Labor is heading for a landslide win this time.

Litmus seats: Eden-Monaro and Macarthur

Many eyes are on NSW's Eden-Monaro, whose biggest town is Queanbeyan next to Canberra, as the bellwether seat. It's gone to the election winner ever since (and including) 1972. But there's actually an electorate that's had that characteristic at every election since its creation in 1949 - Macarthur, around the southern highlands of NSW.

It's expected to remain Liberal no matter what this time, but funnily enough, the big redistribution before the 2001 election was on paper in Labor's favour, so much so that new Lib candidate Pat Farmer needed a small swing to hold it - which he got, and more. 

Macarthur's complexion has moved away from regional towards outer suburban mortgage belt, and like nearby Lindsay and Hughes is looking difficult to shift.

• Have added Australian Financial Review to papers' election sites at top right. Has some informative if slightly gimmicky graphics.

• Advertiser poll in Wakefield has 53 to 47, a swing of a couple of percent to Labor.

• Mr Shanahan in the Oz does the seat sums and comes up, wouldn't you know it, with a hung parliament, although he doesn't describe it as such (see last par). His summary is useful in broad, relative if not specific terms. South Australia does seem to be Labor's best chance, but unfortunately, it's a small state with few seats. Queensland has a smorgasbord on offer, but the signs there are mixed. WA probably go backwards for Labor a little.

Apparently Labor expects to lose Hasluck in WA. Not fair, as the member, Sharryn Jackson, was no lemming.

More on Dennis: in his column today, he writes "the Coalition got 51 per cent of the vote after preferences and lost the election in 1990." That should read "50.1 per cent". There's a big difference; 51 percent, which Beazley got in 1998, was the highest losing vote in federal history. 

See rankings 18 and 22 in this table.

September 16 Newspoll in the SA margins

Perth yesterday was where Marky Mark bellowed (and Senator Peter Cook behind him jumped out of his skin) that he was "Jack of it!". (At least it's an Aussie colloquialism, unlike his favoured 'turn it up', which is as Pommy as Sun Hill coppers. Isn't all this self-conscious 'ridgy didge' stuff wearing thin?)

Anyway, over in South Australia the Oz today reports a Newspoll in the marginal trio Adelaide, Hindmarsh and Makin giving 54 to 46, a six percent swing to Labor since 2001. As the report says, "the polling contrasts with recent data compiled by The Advertiser" (see September 14 below).

The Oz doesn't say when the poll was taken, but my guess is that this is the first since Sunday's debate; if so, that's a big boost for Latham.

Tasmanian trees

No compunction prevented Mark Latham getting all populist over Governor Butler's payout, and Tasmanian Premier Paul Lennon returns the favour by rubbishing Labor's policy on trees.

Twenty-one years ago the Messiah Bob Hawke took Labor to victory with a platform of which the Franklin Dam had  a high profile. He got a three and a half percent national swing comprising large positive contributions in every state except one - Tasmania, which went to the Fraser government by three and a half. 

Back then all Tassie seats were in Liberal hands anyway; today all five are Labor. Poor Ms Byrne.

September 15 Hung parliament? Me in Canberra Times

Schools and home internet use

Two Census 2001 additions to pendulum gallery, tables showing number of people attending private schools and numbers who use internet at home. The ranks of the two tables are, as you might imagine, similar in many respects. Canberra and Bradfield head both lists. On the other hand, Sydney is near top of internet list, near bottom of schools one, due to few kiddies. (APH 'Couple families with dependent Children' - not shown here - has Sydney at bottom with 20.2 percent, the top seat, Calwell, has 49.6)

Tassie seats bring up the rear of both.

Go to pendulum gallery, scroll to bottom of right hand frame.

September 14 Newspoll says 50 to 50 again, from primary support of 40 to 46 - a somewhat Creanesque situation, but taken before the debate.

• From South Australia, one time Labor Sturt candidate Phil Robins sends this about a Tiser poll in Boothby (no direct link). A safe Liberal seat.

Most important election since 1972/WWII/Federation/insert

It's "most important election" time again. Bob Hawke says it is - as he did at every poll when he was PM, and all since. I think Beazley has too. It's usually Labor politicians who use this line. Is it a good one to use from opposition, when safety is at the forefront of the swinger's mind? Or is the problem for Latham that everyone reckons he and Howard and peas in a pod on everything but Iraq, so if that's the important thing ....

September 13 Latham, nerves of steel, charms and wins debate  Any actor will tell you that getting on top of your nerves is 95 percent of the job. If you do that, you can "be in the moment" and reach in and bring out all sorts of stuff. Latham showcased his inner charmer last night.  If Howard has an inner charmer it remained unrevealed for the nerves, and he was shrivelled once more. In the 2001 campaign it was over two weeks before a post-debate Nielsen or Newspoll. They eventually showed a fillip for the winner Beazley. Hopefully won't have to wait that long again. (Tomorrow should see one or both of them, but they'll have been taken over the weekend.) (Counted one mention of "the vernacular" from Howard and a restrained two "fair dinkums" and one "dead seat" from Mark.)

While we're in an optimistic mood for Labor, I want to put on the record my nomination for the surprise Ohmygod! seat for Labor if they win the election. It's Cowper, around Coffs Harbour in NSW, only held once by the ALP (1961). The reasoning? It exemplifies the sort of seat Latham has a best chance of winning: outside the capital cities, low income, (see median income table link in right hand frame of pendulum gallery) and it swung to Labor a little in 2001 while the state went the other way. Also, there're hippies, ferals and a growing number of ex-Sydneyites in them there hills.

September 12 4:30pm Death in Kalgoorlie

A reader informs me that Kevin Richards, the ALP candidate for world's biggest electorate Kalgoorlie (certainly winnable for Labor), has died of a heart attack. 

Debate tonight

A horrible feeling John Howard will utter, at least once, "to use the Australian vernacular, [such and such] is not 'fair dinkum'", and "larrikinism", "Kilkenny Cats" and "ridgy didge" will come from Marky Mark's mouth.

ABC reports a poll in Tasmania that has Bass and Braddon being lost to the Liberals. But with 1,000 respondents across the state, such individual seat calls are not meaningful.

Update: Fairfax also has it. Might be better to use the aggregate Tasmanian two party preferred, which they don't give us but which from what they do is probably 54 to 46. Last election was 57.7 to 42.3. Either way you look at it, Bass looks shaky. But sitting member O'Byrne is a card-carrying lemming so there might be some justice here. I touched on this six months ago.

September 11 Another day off for the Drama Queen of the South Pacific

As the sookiest nation on Earth gorges on another whopping serve of self-importance and self-pity, it turns out that today was always going to be a rest day out of respect to the Twin Towers episode three years ago. Pass the bucket. Talk about selective mourning: do we honour any other world atrocities like this?

Thankfully October 12 (Bali bombings) falls after polling day. 

Anyway ... Adelaide Advertiser asks a wee sample of 419 voters in the seat of Adelaide who they reckon they'll vote for and gets two party preferred at 51 to 49. Thankfully they're not still doing their 2pp calculations to two decimal places, but their "math" (as we Americans say) looks poor. In primary support they've got Libs on 42, Labor 35, Greens 5 and Dems 3. If they're still calculating using the preference flow at the last election then their sums are wrong; it should be at least 52 to 48.

Also, a sample of 419 means, according to this formula, an error margin of 5% (with a 95% confidence interval). Let's say Adelaide could go either way.