Go to next posts

welcome to               Mumble 2004: Go to post redistribution pendulum

Jan 31    

In today's Financial Review [subscriber only; will post next week], on the Antony Green and John Cherry tete a tete at Crikey re all those exhausting preferences in Queensland in 2001.

A sub-editor confusingly changed this sentence: 

"Green writes that if Queensland used CPV, Beattie’s current majority would shrink by eight, and Springborg’s required swing would only be 6.8 percent" 

to 

"Green writes that if Queensland uses CPV, Beattie's majority will shrink by eight, and Springborg's required swing will only be 6.8per cent." [My italics]

It's the little things, but Queensland won't, of course, use CPV. Will fix it when posting here.

See the ALP National Conference                            

Jan 30

Morgan says Labor 55 to 45

January 26

Useful summary and review of Latham's books at Crikey.

January 20

Pop the corks. Latest Newspoll reveals Latham's support .... equals Crean's final result - 50 50. (No link.) See this table if you haven't already.   Plenty happening in Queensland

January 19

The federal pendulum gallery is up - almost completed. 

Has the federal pendulum - and much much more. All electorates now have graphs showing two party preferred votes over last two decades.

January 16

Morgan puts it, federally at  55.5 to 44.5; in Qld 61 to 39

Queensland has moved to a new page

January 12

Federal pendulum is updated. Data from the APH. Hopefully won't change when AEC calculates theirs.

January 5

Mackerras pendulum in the Oz. I'll use it to update mine. Main difference will be the retention of two NT seats, so leaving total HoR at 150.

In his accompanying blurb, Professor M writes: "I confidently predict that there will be a result, and that result will lie somewhere in a range between a landslide to Howard and a landslide to Latham".  He wrote the same before the last election, and it invites derision, but his points - that landslides can happen easily in single member electorates (especially if you don't actually define what a landslide is), and that not knowing who's going to win doesn't mean the result will be close - are worth making.

He could have added one more: a close vote doesn't necessarily translate into a close result. Fifty fifty can translate into a double digit seat majority for one side without breaking a sweat. (Single digit seat majorities are rare.)

M also postulates that Victoria and Queensland will be Labor's weak spots at this year's election, while SA and NSW will be the other (all relatively speaking, of course). None of that surprises except perhaps Queensland: if the ALP does even worse there than it did in 2001, then they've really got an uphill battle.

See also this table.

December 28

Morgan gives it to Labor 54 to 46

December 17

No bounce for Latham. Why ?

He has taken Crean's approval rating, preferred PM and primary vote way up, but two party preferred has gone nowhere. Two party preferred is how you win elections, remember. See table.

Maybe this really is his bounce, and next year he'll settle into a pattern of something like 47 to 53. Now that would be the sort of train wreck Crean only had nightmares about.

December 16

Another Newspoll gives 51 to 49 (although my calculations from their rounded primary votes make it 50 50 - Newspoll, remember, doesn't ask for preferences, they calculate a "notional" two party preferred.)

This survey says the same as its predecessor: that Latham's elevation is attracting buckets of approval but no more two party preferred votes - you know, those things that win elections - than his predecessor. And this is the honeymoon.

December 15

In Canberra Times today - you can guess the topic.

And here's Saturday's Fin Review "Lies and Stats" on Latham's ordinary poll performance. Table below compares Latham's first Newspoll performance with those of Howard and Downer. 

December 13

Take a good look at the man pictured right. He looks highly intelligent. His face exudes thoughtfulness. Yet he voted for Mark Latham over Kim Beazley to take the Labor party into the next election. He is NSW's best Labor Senator, but Sussex Street won't give him top ballot slip position. Sussex Street backed Beazley. Is that why this man backed the other fellow? (Or was it too many lunches with Alan Ramsey?) Every Labor MP, obviously, would have voted on one criterion only: who is more likely to lead the party to victory at the next election. Wouldn't they?

See also this chap, left, who holds a safe western Sydney seat.  A couple of years ago, Latham described chappy as "one of the great embarrassments. The only reason he is in parliament is his father had the numbers in Granville. The only reason he is on the front bench is his brother had the numbers on the Left ... It is an embarrassment to the labour movement to think of him as a future minister" Said chap was Latham's most vocal backer last week. And, yes, he's still a shadow minister.

Humans are indeed complex. That's what makes life interesting.

Anyway, I've updated Lemmings to include senators as well.

December 12

Mr Morgan, he of the Roy Morgan Research Centre (his name's Gary; Roy was his father) believes it to be 52.5 to 47.5.

But Mr Morgan usually had Crean in front too. See his poll database (best to select two party preferred)

Michael Costello in the Oz makes one of my points from tomorrow's Fin Review "Lies and Stats": that Mark Latham's bounce in the opinion polls was absolutely zilch. Costello compares Latham with new opposition leader Alexander Downer in 1994 but the same, and obviously more telling (because he was successful), one can be made with John Howard in '95.

So here's another table

New opposition leaders' "bounce"

This has opposition leaders' first and last poll results. The main points are clear: Downer and Howard oversaw big jumps in voting intentions, and took their party from behind to comfortably in front.

Latham, by contrast, added three percent to Labor's primary vote, but sent another five scurrying to the Coalition, so taking his party backwards in two party preferred terms.

* my estimate of 2pp

+ excluding weekend of limbo, between Crean's resignation and Caucus leadership ballot.

It's still early days, of course. And one poll is just one poll.

December 9

Newspoll shows Labor's two party support ... pretty much unchanged under Latham (51 to 49). Doesn't stop the Oz popping corks.

ACNielsen has 52 to 48.

Danger for new leader in turning over new leaf and repudiating past misdeeds is that he becomes just another "white bread politician" - and a not particularly articulate one at that.

And "I love Australian larrikinism". Pass the bucket. Bob Hawke never had to use the word, he let his persona do the talking.

December 8

According to a little birdie, the final Mackerras Pendulum - to go into the next election - should make an appearance in the Oz this week. When it does I'll pilfer his numbers and modify mine. The only big change, I think, is the retention of two NT seats.

December 7

Sydney Still Centre of Universe

Sun Herald commissions Taverner to survey five Sydney seats and declares "the result would comfortably hand the western Sydney politician the keys to the Lodge".

(Sunday Age also has it, with this strange mathematical equation: "support for Labor had jumped by 5 percentage points over the November 2001 election result, from 39 per cent to 42 per cent.")

The numbers actually point to none of the five changing hands - the Libs keep Macarthur, Lindsay and Parramatta, and Labor retains Lowe and Greenway. Labor's vote is up, however.

That's assuming the poll is reliable. They don't tell us sample size (in online version) - presumably the individual seat ones are too small - and for some reason don't measure preferences or tell us minor party support.

Most useful info is that 77 percent thought Latham's elevation a good idea. That's a positive start.

But there are 150 seats in the lower house of federal parliament.

Update: I've seen the Sun-Herald, and the total sample size is 500, which makes individual seat results meaningless. Five hundred has a margin of error of about 4.5%, which is just usable - just.

Also breakouts in the hard copy version for each seat, but the data is a term out of date - based on 1998 results, so the information relevant to going into the 2001 election.

December 6

I went on about Caucus's decision, didn't I?

Let me recap:

The opinion polls, plus the Howard government's electoral record, plus the electoral cycle, plus common sense and gut feeling, have through-out this term pointed towards one thing: the natural way of things is for the Howard government to go down to the ALP, and if the opposition had a decent leader this would happen.

Even under Simon Crean they had a chance.

Beazley would have won. (Almost) no question. Had he gotten the opposition leader's job, the dynamics would have changed immediately and stayed that way until - and including - the next poll. 

Forty five members of Labor caucus voted for the person they thought most likely to win the next election. The other 47 were motivated variously: spite; retribution; "not to reward treachery"; loathing of Beazley; it seemed like a good idea at the time. Probably a handful actually thought Latham had a better chance of success.

On the occasions I've found myself within earshot of Labor circles - well, staffers at least - I've been surprised at some of the things they think. All that stuff journos and pollies mouth in public, that one had assumed people in the industry knew to be nonsense but were incorporating in their "message of renewal" - it turns out they actually believe.

So they too reckon that if Beazley hadn't pursued "small target" he would have won. That policy is more important than presentation, and the ALP needs to "cut through". Because voters say they're interested in education and health (and they are), an opposition should talk about those things until they eventually do "cut through".

But the biggest howler: Howard is immensely popular and unbeatable.

There's an explanation of the caucus result: thinking you have no chance makes it easier to throw the election away.

Still, anything's possible, and Latham might win. I still put it at 5 to 1. More likely they've needlessly handed another term to the government.

Best result for Labor would be a win, obviously. Second best one would be to go down by a big margin. Worst outcome would be a close loss, enabling the leader to stick around.

Note Beazley said he was sticking around after the next election. If, a few months out, Latham looks like winning, don't be surprised if Beazley changes his mind and announces retirement.

Oh, and don't believe that rubbish about only Beazley being able to win his seat of Brand (and that's why he's staying on). Like most party leaders, he's never been much of a vote getter in his own seats. Plus it's very safe for Labor.

December 5

Mr Morgan, who is a very nice man, says 48 to 52 on the weekend, ie after Crean's announcement but before Lathams' election. So of not much use. Newspoll put it at 53 to 47 on the same weekend.

December 3

See the lemmings!

Want to see the faces of the Lower House Lemmings (don't have the Senators) who think "larrikinism" plus fiscal and social conservatism "cuts through" to win elections from opposition?

History may well record December 2 2003 as the day Federal Labor snatched a generation of defeat from the jaws of victory. 

When the Coalition doubles its majority next year, these are some of the people you'll have to thank. 

December 2

Put it this way. [Stop me if I'm raving.] Latham lives in Werriwa, which borders Hughes, Lindsay and Macarthur. These are the three outer suburbans that the ALP lost in '96.

Now, it seems that not all seats are equal, and these ones are really, really important to the ALP. They "must" win these three back.

They're typical outer suburban - that is, white-flighters, not short of a quid - and no doubt when Mark crosses the border and queries the blokes in 4 wheel drives what they want from a political party, they say Mate, let us keep the private health rebate, but throw in a top bracket tax cut, keep the refos out - oh and for f**k's sake, don't talk to me about Reconciliation.

These fellows are probably impressed with the lad's testosteroned fisciality, and as I said, many in the ALP would gladly trade ten other seats for these three. They call them "battlers", but the real battler seats, further in (like Fowler), are safe so safely ignored. (See this table of seats and top tax brackets.)

One problem with the outer suburban strategy is they're aren't enough of them for victory. Another is that they are held by the Libs by large margins. And yet another is that they're doing very nicely thankyou right now so why would they vote Labor?

Latham's message is tailor-made to blokes in the outer suburbs - to the exclusion of everyone else. That's why he'll lose in a landslide everywhere else - and won't get those seats either.

(I still blame the chirpy Rudd on Lateline - see below.)

Latham wins

Until 10am today the ALP was in with a good chance at the next election - even under Crean.

How long now? Its last spell in the wilderness was 1949 to 1972; the Great Split in 1955 sent them packing for another seventeen. It's now almost eight years since Howard's election; in another fifteen Latham will be 57.

Caucus committed huri-kuri today. Watch out for a landslide.

Latham's leadership will impress about ten percent of the male population, most of whom still (albeit reluctantly) vote Labor anyway. A couple of percent of Greens might return to the fold (until the penny drops that the new leader loves mandatory detention and thinks Reconciliation's a waste of time), but Labor had their preferences anyway.

Meanwhile several percent of voters will probably run a mile into the safe arms of John Howard. Labor's primary vote remains constant while its two party preferred one goes down.

There's your landslide.

Wonder how much part was played by:

1. Shots of Beazley in gym clothes

2. Kevin Rudd demonstrating the reasons behind Caucus's apparent detestation of him, with an annoying preening effort on Lateline in which he backed Beazley.

-------------------------------------

Newspoll says Carr most wanted [no link, surprisingly], but the question is a leetle bit pushy:

"If Bob Carr, the Premier of NSW, was available to lead the Labor Party in Canberra before the next federal election, who do you think would be the best choice to lead the federal Labor Party?"

Carr gets 32, Beazley 22, Rudd 11 and Latham 10

December 1

Morgan again on the Labor pretenders. Beazley 49, Latham 29 without Rudd in the equation. Open PDF.

Western Sydney is the centre of the universe, part VIX

The member for Reid, Laurie Ferguson, says Latham deserves the top job because he would "win back" western Sydney.

If you had a candidate for the Liberal leadership putting a premium on reclaiming the  waverers on Sydney's north shore, you'd know they had rocks in their head.

Ferguson says more people live in Western Sydney than South Australia - and he's right. Western Sydney has 16 seats, of which Labor holds 13. South Australia has eleven, of which Labor has just three*. Another three Coalition held South Australian seats would fall with swings under four percent. Just one Western Sydney seat would.

And then there's Queensland, where Labor holds seven out of 28 seats* and would win another six with swings under four percent.

I know where I'd be throwing electoral capital. Latham's backers have no idea what wins elections.

Look to the regions.

*notional post redistribution numbers

I go on about the regions alot, don't I? Here's why: the regions' pendulum - where the marginals are.

November 30

A Newspoll Time Capsule: nine years ago (give or take)

It's January 1995. Two years earlier Paul Keating had pulled off a win against the odds, securing a rare pro-government swing predominately on a single issue. The Liberals are in a shocking state, with no idea what they stand for, and Keating is looking unbeatable.

Alexander Downer's short spell as Liberal leader is spluttering to an end. Newspoll asks who should lead the Liberals and gets these numbers:

John Howard  44 

Peter Costello  14 

Downer  7

Which party would they vote for? Labor 47, Coalition 40, Democrats 7

Better PM? 47 percent say Paul Keating, 21% reckon Downer. 

All similar to today (with parties switched) but for two things:

1. Back then, Keating was widely loathed, with 35% satisfied and 54 percent dissatisfied with his performance. Howard's respective numbers today (at last month's Newspoll) are 53 and 37 - close to a perfect reversal.

2. However, Keating was way ahead of Downer in voting intentions, especially if you factor in Democrat preferences. Today the vote after preferences is close.

Here's the PDF

Several months earlier (October 1994), preferred leaders were:

John Howard 20

Alexander Downer 18

Nick Greiner 14

Bronwyn Bishop 13

Peter Costello 10

John Hewson 9

And in January 1994, preferred leaders were

Bronwyn Bishop 48% 

followed by John Hewson on 34

and then daylight.

Today's polls

Mr Morgan says Beazley preferred by 38% of Australians, Rudd by 21 percent and Latham 17.5%. This will open pdf in new window.

Sun Herald says something similar, as does Sunday Telegraph.

November 28. Crean is gone. Tomorrow's Alan Ramsey column in Sydney Morning Herald should make for interesting reading. He's Latham's strongest press gallery fan; it's a fair bet he'll flay Mr Beazley something rotten.

Speaking of Kim, I have a guest post on John Quiggin's blogspot on why I like him for top job. [Now here]

November 27 Mumble odds revisited!

In early September I posted odds on the next election result depending on who's leading the ALP. In light of current shenanigans, here's an update.

Actually, they're the same as before, with the addition of Kevin Rudd. Plus I've lengthened Latham's (to hammer the point).

The five leadership scenarios now:

1. Simon Crean is still leader: Labor 2 to 1 to win

2. Bob Carr is leader: even money (1 to 1)

3. Kim Beazley is leader: Labor 1 to 2 to win

4. Mark Latham is leader: Labor 10 to 1 to win

5.  Kevin Rudd is leader: 2 to 3 to win.

(That is, Latham is least likely to win and Beazley most likely. Both Beazley and, to a lesser degree, Rudd, are more likely than not to win.)

Apart from the obvious - a good communicator, foreign policy credentials - Rudd is from Queensland, and I reckon Labor should take more leaders from the third most populous yet most parochial state.

Downside: he's untested and could go down in a screaming heap, like the man he currently shadows did in his brief tenure.

And don't forget this about former leader.

November 26. Another table (or three), apropos of Mark Latham's proposed top income tax cuts; Lindsay Tanner suggests they wouldn't deliver much electoral bang. Find out who's right. (It's Tanner.)

November 21: Mr Morgan reckons 50.5 to 49.5, from primary support of 42.5 to 35.5.

Greens on a record (or close to it) 9.5, their preferences splitting 88.5 percent to Labor, 11.5 to Coalition.

November 19

Last year Alan Ramsey strangely stated that "the Greens didn't come within cooee of winning a seat" in the Victorian election, despite one of them getting 48 to 52 after preferences.

More weird pronouncements on matters electoral  today.

He offers as " bleeding obvious" that " 1) For Labor to have any chance at the next election, it is absolutely crucial it significantly increases its primary (ie, popular) vote; and 2) The preferential voting system increasingly disenfranchises more than 2 million voters to the advantage of the big battalions. There can be no other conclusion when one in every five voters last election wanted to elect somebody other than the Coalition or the ALP."

Ramsey's wrong on both points. The first is self-evident: if they get enough preferences they don't need a high primary vote to scrape in (although it always helps). In the second case, it's not preferential voting that disenfranchises "minor party/other" voters, but single member electorates per se. Preferential voting actually gives them another bite of the cherry. Under "First Past the Post" in, say, the UK,  such votes would be wasted; they're more "franchised" here.

As far as I'm aware, the only minor party or independent to ever be disadvantaged by preferential voting was One Nation - specifically, Pauline Hanson in the seat of Blair in 1998, where she won the primary vote but not the seat. But having both parties gang up on you with preferences is unusual; and anyway, no-one forced Labor voters to put the Liberal ahead of Hanson - they were just advised to do it on their 'how to vote' card.

Ramsey also notes, as have others, that in 2001 the ALP got its lowest primary vote since 1931.  But so what? Its second  lowest was in 1990 - which it won, despite trailing by four points on primaries. And its highest was in 1954 - which it lost.

That's because two party preferred votes decide seats, not primary ones.

If Ramsey's saying that Simon Crean is a dud and his party should, in the current circumstances, be attracting more support, he's right. But that doesn't mean it "doesn't have any chance at the next election".

November 18. Newspoll is back to 50 50 ( no link). So last fortnight's is looking distinctly roguish.

November 15

Two new tables - seats in order of aggregate movement since 1993. Have a looksie.

And by two party preferred swing at 2001 election. Look at that one.

November 7 Mr Morgan says it's 53 to 47

More Newspoll musings

Two weeks ago the chat was about Labor's two party preferred 51 to 49 Newspoll lead. Now it's the Coalition's 54 to 46 one. (Here is a PDF of this week's Newspoll in a new window. Toggle or close it to return here.)

1. Newspoll really should measure full preferences.

Newspoll just asks respondents who will get their primary vote. In days gone by, this was kind of good enough. But not anymore. Minor parties and "others" are attracting record support, and while it's probably ok to notionally distribute Greens about 75% - 25% in Labor's favour, we have no idea who, for instance, the over 10 percent (on average in the current parliamentary term) "others" are. (In the last term "others" averaged about 4 percent).

Presumably, Newpoll is notionally distributing "others" about 52 to 48% in the Coalition's favour, as they flowed at the last election (when there were just 4.4% of them). But who are these "others"? Who is voting for them? They might, for instance, be predominantly Coalition voters parking themselves elsewhere for a little while, in which case their preferences would flow overwhelmingly to the government - and published results greatly understate government two party preferred support. (For example, if that ten percent would really flow 75 to 25 in favour of the Coalition, its two party preferred lead is being under-estimated by 5 percent.)

2. We need competition in the polling industry.

Every second Tuesday the political class hangs off Newspoll. Rightly or wrongly, Morgan, the only other regular pollster, remains in their doghouse. ACNielsen (like Morgan) measures full preferences and was closest to the actual result during the last federal campaign. But Nielsen only runs once in a blue moon.

Regular Nielsen polls in the Fairfax papers would be a wondrous thing for public knowledge. It's unhealthy that one opinion poll (reported in one paper - a proudly opinionated one) is given such elevation.

3. An individual opinion poll, such as this week's, isn't meant to be taken literally.

Some deeply silly things have been said about this week's Newspoll. Eg the Coalition got a seven point boost from the two visiting Presidents; Bob Brown's shenanigans cost him two points support.

Opinion polls aren't that accurate. They should be looked at collectively; the trends matter. Each new one gives us a bit more information to add to our bundle of knowledge, and because it's the latest, it deserves more weight than any previous individual one, but as it's only one survey it's less deserving of emphasis than the aggregate of all the ones before it. As well, as a poll recedes into the past, so does its relevance today.

So, before this week, a series of Newspolls told us it's pretty close but the Coalition would be favoured to win an election. After reading Tuesday's Australian, we should favour the Coalition a little more. If the next survey gives a similar result to this week's, then we could say the Coalition really is comfortably in front.

Here's a statistical perspective

(Attention the numerically timid: have tried to make what follows as painless as possible.)

A rogue survey? Margin of error and confidence limit

I got this explanation of margin of error and confidence limit from a UK site:

"The margin of error of an opinion poll is often quoted as plus or minus 3%. This is a good approximation, but there is an exact formula that can also be used.

If an opinion poll sampled N people, and found a proportion p of support for a party, then the margin of error on that proportion is 1.96*Sqrt[p*(1-p)/N]. For instance if a poll of 5000 people estimates a party at 30% support, then the margin of error is 1.96*Sqrt[0.3*0.7/5000]=1.3%. Another example is a poll of 1000 people, and a party at 50% support, where the error is 3.1%, which is the famous "3%" margin.

This margin is not an absolute guarantee but a confidence limit. It should be expected that for 19 polls out of 20 the true levels will lie within the margin of error (95% confidence level)."

OK. In the case of Newspolls, their sample size is indeed about one thousand, and so the error margin is about 3%. But the confidence level means that, on average, every twentieth survey will not even be within the 3% margin; it will be a "rogue survey".

It's possible this week's was a "rogue survey". Next fortnight's will tell us if it was.

Now, we could take the last two surveys, average out the results, and our error margin would be about 2% (by putting 2,000 into the above equation). And we could take the last four surveys and our margin would be about 1.5%. So the more surveys you look at, the more accurate your data .... except for one important (and obvious) aspect: much of the data becomes increasingly out of date (four Newspolls would take us back 2 months.)

But that's another way of saying it's the trend that matters.

Finally ....

Once more: Newspoll really should measure preferences. They do it during election campaigns; they should do it all the time.

And over at Fairfax, instead of Alan Ramsey, Michelle Grattan et al quoting the Australian's regular pollster, the company might dig deep and jolly well run their own.

November 4. That's a huge jump and big lead for the government in today's Newspoll.