As I suggested it would on Monday, the Galaxy Poll of four Queensland marginals has done more to mystify and shape the media narrative than to enlighten. Witness stories like this one in the SMH which asserts that Labor needs to win 6 seats in Queensland to form government (as QUT Political Science professor Clive Bean keeps emphasising, 2 would do and the Mexicans can take care of the rest) and that the swing shown is state wide (and thus Labor won’t win Blair) – which it isn’t.
It’s also unleashed a veritable flood of leaked party polling. It’s worth demystifying both Galaxy and the motives and significance of the leaking of “private� polls.
Galaxy is designed like the “tracking polls� the parties themselves run – usually on a sample (not necessarily representative) of marginals or safe seats. These polls are repeated on a regular basis and are used to discern things like the effectiveness of messages and to track leaders’ images as much as to measure the vote in the seats concerned. They’re like focus groups writ large. Similar polling was leaked in enormous detail last year by Sussex Street to help Rudd oust Beazley.
Because the sample is spread across several seats, the margin of error in each seat would be 7%. But as I’m emphasising, they’re not used by the parties primarily to measure voting intention. The other problem we have with interpreting Galaxy is that we’re only seeing one poll, whereas party strategists have a succession of them (hence the terminology “Track 33� in the infamous leaked Crosby/Textor research).
So Galaxy honcho David Briggs is quite right to say the good numbers for Rudd from their poll are good news for Labor.
The parties sometimes also poll individual seats with a proper sample (which is about the same size as a national sample for statistical reasons). This is very expensive, which is why the newspapers don’t commission them, and why leaks of such polls should be poured over for their motivation. There always is one.
Labor leaked polling in one state seat, Redlands, last year, showing a lineball contest to suggest Beattie wasn’t home and hosed. But, of course, that was one of very few seats where for local factors the vote was close.
Yesterday we could read in The Australian that Labor has “written off� Longman, and the Libs have been touting polls which show them in contention in Moreton for yonks. No one I’ve spoken to believes that, though maybe Gary Hardgrave’s anti-Sudanese ramblings (which began long before Andrews came to the party) have had their impact. Sometimes the parties give journos actual figures (as with the Senate polling I was leaked a while back) but most of the time it’s just “this is what our polls say�, with no numbers. Access to the polling data itself is rare.
Who gains? is the question that demands an answer. With Moreton, the Libs are probably trying to ensure that their troops are geed up to fight the good fight, and donations keep trickling in. With Longman, Labor is playing the opposite game – trying to depress expectations and drive away complacency about any Queensland Ruddslide.
It’s also very important to note that these polls and leaked whispers are being seen as contradicting the national trend. That may be so, but it might also be a way to play the misinterpretation of the national polls as static rather than dynamic. Just because they haven’t moved much doesn’t mean they’re frozen. Queenslanders are going to be bombarded with Rudd parochialism, and voters in Longman (for instance) are the direct targets of “working mum tax cuts�.
I await Shanahan’s claim that Labor needs to pick up thirty seats in Queensland in order to topple the successful and popular Peter Costello.
Thanks Mark,
qui bono, indeed. On this interpretation of the leakers’ motivations, can we assume that parties will only leak (or indicate roughly) private polls that show the party is doing poorly? I can see that if voting weren’t compulsory, a party wouldn’t want complacency to set in….
Did anyone else hear the GG pollster interviewed on radio yesterday (Wed.)? He was saying, “the public HAS to….” which was interesting because the prior phrase in “If the Government is to be returned, the public has to….” was never uttered. It sounded as if he knew the proper outcome of the election, and was figuring out what intellectual tasks lie ahead (for the voters), to ensure that outcome.
It made me wonder if the GG was perhaps a little biassed, forsooth. Heavens to betsy.
I find all this seat by seat and marginal analysis to be a bit silly. It seems to me that if the ALP win 52% or more TPP they’ll win. Between 51 and 52 will be close. Anything less and they lose.
So at the moment Labor has 4% to lose and my interest is in the liklihood that could happen, and not strange arguments about individual seats.
With a decent showing in NSW (which is the keystone for the ALP anyway), coupled with picking up Bass, Braddon and Solomon, they could get away with pretty ordinary results everywhere else including just 1 Queensland seat. See here.
Coincidence, read your article just after WIN South Coast NSW had a news item on a government “leaked” poll that showed Garry Nairn (LNP sitting member) was closing in significantly on Mike Kelly (ALP) for the seat of Eden-Monaro (a supposed bellwether seat).
There has been a massive amount of pork thrown at Eden-Monaro by Howard, including propping up an nonviable timber mill and the building of a dam.
Ambigulous:
While you make a good point about the speculation on marginals etc, your point about a critical level of the 2pp (52%) delivering victory isn’t necessarily true. For this to be true you must make an assumption about the underlying distribution of voter’s preferences seat by seat. For instance, a 2pp of 50.1 would delivery victory to Labor if there was a uniform distribution of votes within all seats, which would then deliver them 50.1% of all seats etc etc.
Thus finding out which seats are swinging is an important part of the story as it can provide us with an indication of what the 2pp would have to be to deliver victory. For instance, if only safe Labor seats where swinging the 2pp could be 60% and Labor could still lose! Although it must be noted that historically when a swing is on…a swing is on…and thus while statistically possible, the above situation is never this pronounced!
Adrain, bellwether isn’t the same as magic. If a swing back to Nairn is happening because a bunch of pork was thrown at the seat it won’t magically mean the rest of the country follows.
Adam in comments at Pollbludger said this:
176
Adam Says:
October 19th, 2007 at 11:08 am
Good morning
Queensland Galaxy Poll Unmasked
We have now got to the bottom of the supposedly bad-for-Labor Galaxy poll in the Queensland marginals. Peter Brent has posted the primary figures for that poll, which apparently weren’t made available at the time the (alleged) 2PV figure was released. The primaries were ALP 45, Lib 44, Greens 7, others 4. This could not possibly produce an ALP 2PV of 51%, as was reported. If we allocate the Greens 80% to Labor and Others 50% to Labor we get 52.6%. In those four seats, that’s a swing of 6.7%. If that swing was uniform across Queensland, Labor would win Bonner, Moreton, Blair, Herbert and Longman. If it was repeated nationally, Labor would win 28 seats. In other words it was an excellent poll for Labor. This would have been obvious if the primary vote figures had been released. I don’t like conspiracy theories, but I’m afraid this makes me rather suspicious of Galaxy polls.
Good morning mark.
You must be busy since you have not posted anything on the LN/P’s poll bounce, surely the biggest political story of the week. So I will direct readers attention to it. The Government are back in striking distance: LN/P 47 - ALP 53.
It might be a rogue poll, unlikely since both Nielsen and Galaxy got the same results. Or it maybe a temporary tactical advantage, gained by Howard’s unorthodox and clever campaign strategy, soon to be nullified by another bout of Rudd me-tooism.
This election, in the blogosphere, is turning into a test of professional competence in intellectual matters. The results of the test hinge on the predictive capacity of bloggers with scientific pretentions or ambitions. As mumbles says:
I reckon that people who aren’t prepared to have a go - who cleverly hedge their bets all year - lose all bragging rights. I earned mine three years ago but, I believe, exercised them with great decorum and moderation.
(NB: I beat Mumbles handsdown in the 2004 election, correctly predicting an LN/P victory and and increased majority more than 12 months out. Can I say with all the modesty that I have at my disposal, that I was the only commenter to make this call that far out?)
I am betting that this partisan alignment is most probably enduring. In fact I bet this some five months ago, when I predicted a 47/8-52/3 spread.
Mark has been behind the curve at all times. Although, to his credit, he has been honest about his failures.
Over a year ago, when the ALP and LN/P were neck and neck, I predicted that the ALP would win the election. And five months ago I predicted that the LN/P would bounce back to a result similiar to the 1983 election.
Mark was gloomy about the ALP’s election chances right up until the time that Rudd won the leadership. And he has hinted that the result would be similiar to 1996 landslide.
So far my model has been tracking reality accurately. Although there is still time for something to knock reality “off course” predicted in the Strocchi-verse. (Notorious hideout of solipsists and grandiose thinkers.)
I am noisily confident.
Shorter Strocchi: Prepare for me to exercise the mother of all bragging rights.
Off to CT’s predictions thread for your confirmation bias, Jack.
Liam on 20 October 2007 at 4:21 pm
Thanks for the link. I will put up my predictions on that site. I predict Howard will win his seat. My reasoning is that the personal factor will overwhelm the general political swing. Also I think East Asians and Howard are a good fit.
Good to see you are at least having a go. Why for heavens sake dont some of the Larva Prodders at least put their electoral model out there and test it? I mean its not as if they are going to get the sack from the blogosphere for getting it wrong
There is no “confirmation bias” when there is a scoreboard to check. That is why I always put signs on my predictions and, where I have enough data, plug some hard numbers into the equation. Although I am not a skilled quantitative modeller by any means.
So far I have correctly predicted every federal election result back to 1993 (I passed on 1998 as it was too hard to call owing to the GST factor, which was correct.) I can only substantiate that boast for the 2001 and 2004, which were published in digital format. Since these results apparently dont speak for themselves I have had to add some amplification. Sorry.
Also, when I am wrong I try to point this out. For instance I predicted that Iraq-attack would instigate a wave of democratic change turning the Middle East into the next Eastern Europe. Wrong. Very Wrong!!! I predicted Latham was a good choice as Opposition leader. Wrong. I predicted that the housing boom would splutter and die around 2003. Wrong.
Confirmation bias is much more common for so-called “qualitative polling”. That is Mark’s speciality. Also Hugh McKay. I think it is a waste of time but some people seem willing to pay for it. Then again they pay for reiki and reflexology.
Mark is fooling himself if he thinks he will make any intellectual progress by publishing Leftist comfort commentary and abjuring prediction. You cannot get the reward of scientific progress without taking a risk. That means sticking ones neck out and making a prediction.
At least by doing so you are forced to make your assumptions clear. (Which sometimes exposes bias!) Then if you are wrong at least you know that one or more of your assumptions are wrong. SO you learn by a process of elimination. And a little competition makes it interesting!
Of course you could be right in your prediction for reasons extraneous to the model. But luck only works in an insignficant fraction of occasions. Most of the time reality behaves normally. Its just that normality is not always simple.
So you learn when you are right. But you also learn when you are wrong. You do not learn if you sit on the fence.
Science is what we have learned about how not to fool ourselves about the way the world is.
Richard Feynman
Hello Happy Chap,
I think you were replying to Jenny, not me. But you made interesting points.
Not wishing to take sides in the titanic battle b/w Mark and Signor Strocchi, I do agree with the Signor that specific predictions are more ‘testable’ than qualitative observations. e.g. how many seats to change hands in each State? Which seats held? Which seats “to go against the State trend” because of strong local factors?
Some time ago, a poster on LP was castigated for calling millions of his fellow citizens, political “morons”. I agreed with the castigation. Each of us has our own reasons for voting as we do. I don’t care to be dubbed “selfish” or “moronic” or “ignorant”… neither, I fancy, do my fellow voters. It’s as much a matter of the courtesy of demcracy, as anything else.
Likewise, a tone of superiority tends to enter the homilies we read, about how foolish are the MSM pundits. OK: let the acid test be your election predictions. Can you do better than the Venerable McKerras? Or good old Milney? or Professor Albrechtsen, Professor Emeritus of Jurisprudence (Commentariat Medallist)? Or Newspoll… can you do better than Newspoll, oh you mockers??
Jack, why not just make a webpage summarising everything you have to say into five bullet points.
That way, you can just post a link and instead of writing 500 words of the same thing every time.