The day has finally arrived when my headline must specify which party is in front. The latest Newspoll has the Coalition taking a 51-49 lead on the two-party preferred vote, which I do believe is the first time any pollster has had them in front since 2006. Certainly it is their first lead in Newspoll since August of that year. We are also told Kevin Rudd’s rating on preferred prime minister has fallen six points to 50 per cent, while Tony Abbott is up three points 32 per cent. More to follow.
UPDATE: Possum reports primary vote figures: ALP 35 (down eight), Liberal 38 (up one), Nationals 5 (up two), Greens 10 (steady), others 12 (up five). I’m more than ready to believe Labor would have suffered a solid hit in the period in question, but eight points seems a bit much, as does the huge hike in the others vote and the failure of the Greens to reap any ETS dividend.
UPDATE 2: Full tables here. Newspoll pointedly asked whether voters perceived the leaders as decisive and strong, but Kevin Rudd’s decline from the heights of early 2008 is probably not much different from most other attitudinal measures. However, it’s slightly bigger than on the other such question gauged, understands the major issues. The gender breakdowns here are interesting: Kevin Rudd performs much better among women on both measures, particularly decisive and strong. The figures also provide hard evidence, which has been elusive in the past, of Tony Abbott’s much touted problems with women voters. We are also told that Labor’s once huge lead as best party to handle climate change has vanished, whereas its lead on health and Medicare is mostly intact. Here the gender gap is that women are a lot more likely to be uncommitted.
UPDATE 3: And while I have you, here is today’s rather less exciting 53-47 Essential Research result. Note however that Essential Research is a rolling two-week average, so half the result was obtained from the same sample that made up half of last week’s 54-46 result. Furthermore, we get an inkling of how things might have changed from one week to the next from the fact that both samples were asked about leadership approval, and there was a sharp narrowing in Kevin Rudd’s ratings from one week to the next from a net positive rating of 17 per cent in the first week to 4 per cent in the second.
Below is some number crunching I related in comments. It should first be noted that a contributor whose understanding of statistics I trust more of my own is not as impressed with the 0.47694 R squared as I was; that I wrote this before I saw Newspoll’s primary votes, which have made me a bit more skeptical about the poll; and perhaps also that Essential Research has been a bit less favourable for Labor lately.
As of the start of 2008, when Rudd government polling began in earnest, we have three frequent and regular series to play with: Newspoll, Essential and Morgan face-to-face. Their respective average 2PP for Labor has been 56.0, 58.4 and 59.1. I think there’s pretty solid evidence that Newspoll is broadly accurate as these things go, so the best way to interpret this is that Essential has a 2.4 per cent and Morgan a 3.1 per cent house bias to Labor. We can’t compare Essential against election results because they only started in 2008, but we know the final Morgan face-to-face polls at the past three elections overrated Labor by 5.5 per cent in 2001, 3.7 per cent in 2004 and 3.8 per cent in 2007 (significantly, that accurate late Morgan poll you recall from 2007 was a phone poll).
Plugging Newspoll data into a linear regression suggests every 1 per cent shift in Kevin Rudd’s net satisifaction rating should send the Labor vote either 0.118 per cent up or down, with what to my untrained eye looks like a pretty healthy R squared value of 0.47694. The 53-47 from Essential Research combines two weeks of polling conducted from Wednesdays to Sundays, so the more recent half of the survey coincided with the Friday to Sunday period of Newspoll. We know that the sample from the first week gave Rudd a net approval rating of +17 but from the second week’s sample it was only +4. If I’m doing this properly, that points to a 1.5 per cent difference in the two-party score from the first week to the second: so perhaps 53.8 per cent in the first week and 52.2 per cent in the second, assuming the 53-47 figure didn’t involve much rounding.
The former figure is a nice fit for the 54-46 we got from Essential last week, which also included this sample as one half of its survey. It’s also very similar to the 54.5-45.5 from the most recent Morgan survey, which was conducted during the weekend before last – also corresponding with the first half of the Essential survey. Matter of fact, the 0.7 per cent difference between the two figures is almost exactly the difference you’d expect to see between Morgan and Essential, if you look back to the last sentence of my first paragraph.
The 52.2 per cent figure I’m surmising from the more recent week’s Essential sample is 3.2 per cent different from Newspoll – somewhat higher than the 2.4 per cent we’d normally expect, but not off the chart. It leads me to suspect that maybe Newspoll is off by a point or so, but with rounding and sampling errors and whatever you really can’t say this with much confidence.
As I’ve noted, this has come after a two-week period of strategic deck-clearing that in a sense was almost designed to give it short-term hit in the polls. Once the government is again giving the impression of being back on the front foot, and it may have done that already with its tax reforms, I expect its poll ratings to return to more familiar territory. I don’t think there’s anything dubious about this poll, but it does precisely nothing to shake my conviction that the government will hold its ground at the election.