A little while ago, some of you may remember this saga – which was all good and well, but there was something in that spiel by Morgan which was a little too cheeky and needs to be addressed. In fact, they’re very serious allegations.Morgan said:

It is important to understand that there are some key differences between the ‘face-to-face’ Morgan Poll and all other polls in Australia.

The first major difference is that the Morgan Poll is based on ‘face-to-face’ interviews which obtain a response rate at least double the polls conducted by telephone and at least 10 times the polls conducted via the Internet.

In addition, the Morgan Poll is weighted by age and sex in the region interviews are conducted whereas the ‘Other’ telephone polls are weighted by spurious systems (such as ‘past vote’). At the present time, weighting by how respondents say they voted at the last Federal Election changes the voting intention by 3 or 4 points towards the L-NP. This is because when you ask people how they voted at the last election; the number of voters who say they voted L-NP is between 30% and 35% whereas actually the L-NP vote at the last election was 42.1%!

“Other” telephone polls are weighted with things such as “past vote”?

Oh really?

Weighting by past vote is a rather peculiar thing to do, let alone make an allegation about, since such weighting would effectively change a poll from being an estimate of the current state of public opinion, into being a projection of what a pollster believed public opinion ought to be instead.

It would be making the assumption that people don’t actually mean what they say in surveys, and as a result, those naughty and untrustworthy respondents require their answers to be “modified” by the wisdom of the pollster (who obviously understands people’s real thoughts better than they do), before their opinion becomes fit for public consumption.

It would effectively be fabricating survey results, giving not the best current estimate of public opinion, but the equivalent of market research navel lint.

Weighting polls by past vote would be, according to one seasoned Australian pollster that has worked just about everywhere with just about everyone over the last 30 years:

The stupidest f%**& thing I’ve ever f%*&$&* heard in my whole f^%^%* life, and I’ve heard plenty.

…..not to put too fine a point on it.

So I asked our two regular phone pollsters in Australia – Nielsen and Newspoll – a simple question:

“Do you weight your polls by past vote ” ?

The official Nielsen answer was a definitive “No, we do not. Never have“. Nielsen clearly state on their site regarding their polls that “The data is weighted to reflect the population distribution.”

The official Newspoll answer was likewise a definitive “No!”. In the fine print of every Newspoll release is the statement “The data has been weighted to reflect the population distribution”. Have a look for yourself the next time you read one.

Pollsters weight their data by various demographic variables like age, gender, income, geography etc, where different pollsters have their own chosen set of variables and weights – usually well kept commercial secrets – which they use to turn a given random or quota (or hybrid) sample of the population into being as close to representative of the total target population as is possible – which in Australia with political polling is the electoral roll. The battle of the pollsters is effectively a battle of who can recreate the most accurate population distribution from their samples.

Weighting data by measures like past vote would not only completely and utterly defeat the purpose, but would be bad science and a scandal in the Australian market research industry.

Which begs the question of just who Morgan is actually making allegations against here. The obvious candidates of Nielsen and Newspoll don’t weight their data by such nonsense, which means Morgan is either wrong and has been a very naughty boy, or he’s talking about someone else.

So Gary – over to you, just who is it that weights their data by past vote?

Who are these ‘Other’ telephone polls?

Inquiring minds want to know.