Deconstructing Newspoll

Newspoll, moreover, rings no mobile phones, just those people who happen to be home on Friday nights and not, say, at the pub. Just those people who possess home phones (many young people don't any more) and don't mind giving up half an hour of their evening when they might be cooking dinner. Those who do mind are listed as "refused". Only those who speak English, and speak English well, are persisted with, I am told; otherwise, I am told, Newspoll hangs up, and lists them as "refused".

If this is not so I ask any former employees of Newspoll to deny it.

A greater proportion of non-English-speaking migrants vote Labor, I am told. Of course they do. And leaving them out of the sample, if this occurs, puts the Liberal vote up. Of course it does.

And it's by these methods Newspoll came to the conclusion that Bligh Labor was losing in Queensland, and Bligh herself lost round 96,000 voters between Friday 6th and Sunday 8th of March. Have you met any of these people? Anyone who said "I've really gone off Anna Bligh" on Saturday 7th March? You should, if it's true.

We live in a world where numbers are manipulated (by Enron, by Lehman Brothers, by Bernard Madoff, by Robert Mugabe) so commonly, so routinely, that economic meltdowns result from their manipulation.

Yet it's thought that pollsters do not manipulate numbers, even when their clients would like them to. Clients who, for instance, need headlines.

It's usually thought they don't manipulate numbers because it's not wise of them to do so, that a wrong prediction of an election result would do them eventual harm. But this rule applies to only one poll, the one that comes out on election day. Before that they can print whatever they like and so confect headlines - Labor In Crisis, Shock Swing To Coalition - that influence early voting patterns (and early voters) and cause panic in the Caucus, destabilise Kim Beazley, undermine Malcolm Turnbull, influence Labor to support the imprisonment of boat people, and wavering Liberals to support the Iraq war, and so on.

They can put out 75 false counts of the figures in three years, and only one true count, and still keep their reputation. They have to be right, or close, only once.

All this is only my opinion, a hypothesis raised in the public interest that no doubt those more informed can shoot down by releasing raw figures or the rules of engagement explaining why they won't release them, although they used to; why they won't release them any more. Or reminiscing about their happy years at Newspoll, and what, in those days, the ground rules were. And why they changed, if they did.

And explaining why Newspoll has a CEO.

What is there for him to decide?

Just asking.