Pollapalooza – The Turnbull Effect Part 1

With a Nielsen and Newspoll out this morning and an Essential Report out this afternoon, we’ll have to break the polling goodness down into two parts. Later on today, we’ll plug the numbers into Pollytrack and look at the qual metrics, but first up we’ll answer the question of whether Turnbull had an impact on the polls, and the answer is it’s almost certainly the case.

Newspoll today comes in at 55/45 to the ALP (Labor down 1) off the back of primaries running 42/38 the same way – with the ALP dropping 2 on the primary vote and the Coalition picking up 1. With Nielsen we have the headline TPP down 3 for the ALP to come in at a close 52/48, yet the big news is the primary vote with Nielsen having the Coalition taking the lead with 42 (up 2) compared to Labor’s 41 (down 2).

Yesterday on Nerdy Sunday, we looked at the value of LOESS Regressions for measuring turning points in polling data and today that little primer comes in handy (so give it a quick squiz if you start going WTF? over the next few charts)

If we run two LOESS regressions through the all polls series of primary and TPP votes, one with low sensitivity (Degree=1, Bandwidth Span =0.3) and one with higher sensitivity (Degree=2, Bandwidth Span =0.3) we end up with the following charts (just click to expand).

While the low sensitivity regression shows a slight decline for the ALP and a slight rise for the Coalition, the regression with the higher sensitivity shows quite a sharp upturn for the Coalition at the expense of Labor.

Similarly, if we do the same for the voteshare margins (ALP vote minus Coalition Vote) and run three Loess regressions of varying sensitivity through those charts we get:

This again shows that Truffles seems to have had an effect on the Coalition vote, where the more sensitive the regression the larger the effect. We’ll be able to see more when the Essential Report comes out later today and we can plug some Pollytrack numbers in to get an estimate of the magnitude – but Turnbull certainly got a bounce.

What’s amusing about these polls is if only the pundits at The Oz could have owned Nielsen this morning rather than Newspoll – instead of talking through gritted teeth about the mild dangers facing Labor because of obscure metrics on which leader is preferred by the public on what issue, they could have unleashed a tirade on “the honeymoon being over” of the type they’ve been desperately wanting to do since December 2006!

Tough break eh guys

While we’re stirring the commentariat (it is a Monday after all, and what better way to start the week), Tony Wright of the Age today went:

To understand the boost today’s poll figures will give the Coalition’s long-flagging spirits, simply imagine the guffaws if any commentator had dared suggest even a week ago that the Coalition was on its way to its strongest position since November 2006, when Labor was in despair over Kim Beazley’s failure to chip bark off John Howard.

Ah, Tony – that’s the problem with reading commentators, some of them might own Newspoll, but they haven’t really proven themselves the sharpest tools in the shed when it comes to these things.

Over here at Pseph Central on the Crikey Blogs, Pollbludger and Pollytics asked deliberately “What Size a Turnbull Poll Bounce” because we knew that it was almost inevitable. That’s what the data suggested – and really, who are we to argue with the data.

Similarly, you folks reading thought much the same if we look at the results of our online poll. This is what the poll data was last night just before Nielsen and Newspoll hit the press:

82.1% of responses (and we’ll say responses here rather than respondents because some of you cheeky buggers voted a little early and a little too often) predicted a Turnbull poll bounce using Newspoll as the yardstick. So well done on that.

The “pick the bounce” contest on the Newspoll figures came down to a two way contest between Gusface and Gabhran, with both getting the TPP and the Coalition primary vote right, and both making the same primary vote prediction for Labor of 45% – being equally wrong on that metric. Going to the next metric in order, Truffles Preferred Prime Minister rating, Gus chose 26 vs Gab’s 27 making Gus the winner, being the closest to the actual result of 24.

So Gusface, drop me an email at my yahoo address via the contact page and I’ll set you up with a free Crikey sub. Also well done to Gabhran, missing out by a single point on the PPM.

Later today we’ll look at the qual metrics, Pollytrack and a few other things. Until then, there’s plenty happening elsewhere: Pollbludger, Larvatus Prodeo, Ozpolitics, Blogocracy