The end of the trend is nigh. Plugging the poll figures from this cycle into our Pollytrend algorithm – the locally weighted polynomial regression we run through an All Pollster pooled aggregation weighted by sample size - shows how the medium term trend running against the ALP since April has now bottomed out, even popping back up towards them a notch. If we compare the current Pollytrend line with the trend line that we had during the last polling cycle two weeks ago, the turnaround has been quite sharp – what we’ve witnessed this week in the polling has been quite an unusual event:
If the Morgan Phone poll taken last weekend also shows an increase in the ALP’s vote when it’s released later this week, that upswing will be slightly sharper still.
Over on the sidebar on the top right, the longer Pollytrend series has been updated to place it in historical perspective and the poll averages underneath it are all current, including the latest Essential Report.
9 Comments
You’ve been having a great day, haven’t you, Poss.
As always, thanks for the succinct analysis!
It would be really interesting to see this updated: http://blogs.crikey.com.au/pollytics/the-long-view/
4 blogs in one day! You da marsupial!
Any chance that Turnbull is JWH MkII?
http://blogs.crikey.com.au/pollytics/files/2009/02/howard2.png
He seems to be the only one who survived some net negative territory.
Poss, has there been events like this before – such dramatic turnarounds in polling – not just 2pp, but the satisfaction ratings?
The smears against didn’t work for Burke, the dishonest manipulation of Scores, ANZAC etc up until TurnMail-gate and what are the murdoch media up to now, what are the Coalition up to now (and is there really a difference between the two?), the same old smearing. It must be like heroine for these people.
Rudd is nasty, he is the one attacking our honest, innocent Turnbull, who was just sitting there minding his own business. One day the world will find out how nasty Rudd is because, we will invent it, smear it and make the people believe he is evil.
It is instructive of the personal character of many Coalition members and instructive of the character of many a MSM journalist. It is hard to know who is more worthless, some of these murdoch journos or the Liberal Party.
With TurnMail-gate they got their pay off, their just payment. And if course it won’t change, the incessant smearing will continue in unison.
Rudd better make sure that media owndership laws forbid foreign ownership of Australian TV stations, and forbid ownership of both tv and newspapers. Don’t want to see Fake-News in Australia just yet.
Possum, what’s the theoretical justification for using a polynomial regression in this case? Is there a reason we’d expect the underlying voting intention to be polynomial in time?
I ask because the comparison of the blue and black lines makes it clear that slightly higher values in the future have the effect of pushing the trend estimate down a little in the past, which seems hard to reconcile with reality.
Wouldn’t a centre-weighted moving average be more defensible? (or some other low-pass filter).
Rationalist -I find that flash chart clunky. I’m planning on overhauling all the charts soon to be a bit more sophisticated – but it takes time, and I’m still learning how to do it properly!
Peter, there’s been turnarounds this big before – pretty regularly actually, but never in such a short period of time. Usually they take a few months (Downer, Latham, Crean, Hewson etc) – but happening in a week sets a new benchmark.
Caf – the reason I use a LOESS regression rather than weighted MAs or Henderson type affairs comes down to the treatment of end points and the flexibility of LOESS (where I can ramp up the aggressiveness if needed). I went over most of the reasons for it in detail here:
http://blogs.crikey.com.au/pollytics/2008/10/05/nerdy-sunday-youre-so-smooth-edition/
The reason the blue and black lines separated mostly comes down to fitting a curve to the reality. In essence, the polling was a V shape, but just about any curve fitted to that sharp point will bow out at the sides a little. You might notice that the difference between the blue and black trend lines at the bow is less than half a point – which, if you look at the pollytrend chart in the sidebar (where the trend line itself is just under half a point thick in it’s width) means it wont actually make much, if any visual difference.
The next round of polls will be very interesting (not in the least because more interesting twists might still be to come in this whole email saga). With such a rapid plunge for Malcolm it remains to be seen whether that was a knee-jerk reaction, and with time (even just a few weeks) some of those whose minds were changed will come back to support him once the dust has settled, or whether this marks a long term change.
The actually party voting results didn’t shift nearly as much as the satisfaction ratings for Turnbull, so it will be interesting to see if the ALP vote continues to drift upwards or whether the recent events simply arrest the slow decline of the last few months.
So when do the next round of polls come out?
Matt, there’s a Morgan Phone poll out Friday (but that actually polled over last weekend), then there’s the usual Monday Essential Report which is the only poll next week.
Tuesday week we have Newspoll and probably won’t see another Nielsen for a month.