A tale of two cities
Forget different demographics, Melbourne's two daily newspapers seem to be operating in different universes.
Over at the People's Republic of News Limited:
Whilst over in the far more Bracks-friendly Commonwealth of Fairfax, it's a very different story:
As William over at the Poll Bludger explains, this is more than just different spin on the same data:
A difference of 4% is a substantial gap given the similarity of the questions asked, techniques used, and population sampled. And more worryingly for Bracks, it seems that Galaxy has some form at picking results, if Crikey's wrap-up of the Queensland election is any guide:
My hunch: the truth lies somewhere between the two. Given their low starting point, the Liberals would be incompetent not to receive some sort of swing toward them. The game they are playing at is a two term victory: win back some seats this time, and get themselves in a good position to win in 2008. I suspect that we might see a reverse of the 1999 election, with the conservatives winning a swathe of seats in the country, but struggling to make an impact in Melbourne. This quote from the analysis of The Age poll is telling:
Hmmm, just what does country Victoria have in store for us this time? Kennett Mark II?
UPDATE 24/10, 9:03pm: Do'h. 'Anonymous' in the comments has shown the value of proof-reading your work. Next election is indeed in 2010. 27 November, 2010, in fact.
Over at the People's Republic of News Limited:
Poll reveals Labor backlash
EXCLUSIVE
Ellen Whinnett, state politics reporter
October 24, 2006 12:00am
THE Bracks Government is facing a voter backlash and could lose up to 16 seats, an exclusive Herald Sun poll shows.
Almost 60 per cent of those polled believe the Government is out of touch while half say Labor does not deserve to win the November 25 state election.
The poll shows voters deserting Labor in droves, with a swing of almost 6 per cent against the party.
Whilst over in the far more Bracks-friendly Commonwealth of Fairfax, it's a very different story:
Liberals face crushing loss at poll
Paul Austin
October 24, 2006
THE Victorian Liberals are in danger of a crushing defeat at next month's election, with the latest ACNielsen/ Age Poll showing Premier Steve Bracks on track for a landslide victory.
But the poll, conducted exclusively for The Age one month before the November 25 election, finds overwhelming public support for Liberal leader Ted Baillieu's promise to slash the number of poker machines in Victoria.
As William over at the Poll Bludger explains, this is more than just different spin on the same data:
Galaxy Research has created a buzz by showing a lower-than-expected Labor lead of 52-48 in its poll in today’s Herald-Sun. However, it comes on the same day as an ACNielsen poll in The Age showing a Labor lead of 56-44, in line with general expectations.
A difference of 4% is a substantial gap given the similarity of the questions asked, techniques used, and population sampled. And more worryingly for Bracks, it seems that Galaxy has some form at picking results, if Crikey's wrap-up of the Queensland election is any guide:
The final polls from both Newspoll and Galaxy picked the result almost exactly.
My hunch: the truth lies somewhere between the two. Given their low starting point, the Liberals would be incompetent not to receive some sort of swing toward them. The game they are playing at is a two term victory: win back some seats this time, and get themselves in a good position to win in 2008. I suspect that we might see a reverse of the 1999 election, with the conservatives winning a swathe of seats in the country, but struggling to make an impact in Melbourne. This quote from the analysis of The Age poll is telling:
Labor's vote after preferences is stronger in Melbourne (61 per cent compared with the Liberal/Nationals 39 per cent) than in the rest of Victoria (43 per cent compared with 57 per cent for the conservative parties).
Hmmm, just what does country Victoria have in store for us this time? Kennett Mark II?
UPDATE 24/10, 9:03pm: Do'h. 'Anonymous' in the comments has shown the value of proof-reading your work. Next election is indeed in 2010. 27 November, 2010, in fact.
Comments
But what is a very important and an oft overlooked element of polls is the statistical error. This is error is completely independent of any poll biases and crops ups in any statistical figures from coin tosses to nuclear physics.
The formula for this error is the square root of the sample size expressed as a fraction of the sample size. For example both polls were of around 1000 people. The square root of 1000 is 31 and 31/1000 is 0.0316. That means that both results had an unavoidable 3% error.
This really needs to be declared more often in papers especially when claiming a swing of 1 or 2 percent.
It a case of, Bracks leaping 1% in the polls plus or minus 3%, not having quite the same impact.