Oz Politics Blog

My last post has produced the usual cries that Morgan must be wrong. I got the usual arguments saying the real Coalition two-party preferred result is something like Morgan plus five. Someone even suggested a fudge factor of plus ten.

I am not convinced. While I think Morgan may be to the left of the population parameter, it may well be less than five percentage points.

I agree that Simon Jackman’s work on the lead-up to the last election suggested that Morgan under-estimated Coalition support by 4.7 percentage points and Newspoll under-estimated it by 2.7 percentage points on average. Jackman found that ACNielsen was pretty well on the money.

But simply because that happened then, it does not mean it is happening now.

Anyway, it is illogical to apply Jackman’s bias-set [4.7, 2.7 and 0] to the current trend-lines. While ACNielsen was on the money in the lead-up to the 2004 election, it is now hovering off to the left with Morgan. It is simply not consistent. If anything, ACNielsen is currently validating Morgan, and Newspoll is out of sync.

Furthermore, if you go back and look at the polls in the lead up to the 2004 election, it is possible there was a late swing to the Coalition (starting in early August) that was accurately picked by ACNielsen, but largely missed by Morgan and Newspoll. If you accept this thesis, it also could be used to argue that currently Morgan is doing a better job than Newspoll at calling the mood of the electorate.

Now don’t get me wrong. I am not saying that ACNielsen is right and everyone else is wrong. I have two concerns with ACNielsen at the moment. First, I think it has the Green primary vote implausibly high. Second, I think it is sheeting too many non-Coalition-non-Labor preference votes away from the Coalition. Indeed, I suspect that along with Morgan it is tracking to the left of the population parameter.

I am just not convinced that Morgan is as far out as five points. Besides, I put little store in simplistic rules like Morgan plus five.