Don’t tell him that you read the Guardian or the Indy

Being the one alternative voice in the UK polling industry Peter Kellner, boss of the internet firm, YouGov, is going to come under intense scrutiny in the coming weeks and none more so than over the way those his firm surveys are “self-selecting”.

For instead of a pollster going out to find “random” voters to interview all of YouGov’s polls are confined to a panel of more than 50,000 people who have pre-registered for the opportunity of being paid to be polled. About 3-4000 of them are invited by email to take part in each survey.

A concern expressed repeatedly on this site is that that those who bother to register have a keen interest in politics and in no way can be described as typicial. On the face of it this seems a good argument and one that the firm has taken into account. For one of the questions when you register is which daily newspaper you read.

If the Guardian or Indpendent is your main paper then you are probably going to receive less invitations to take part in surveys and when you do your views will not register anything like the “value” in terms of the top-line results as, say, a Sun or a Mirror reader.

For the last YouGov poll for the Mail on Sunday two weeks ago this is how readers of the different papers were “rated”:-

Express/Mail 1.0
Sun/Star 1.24
Mirror/Record 1.11
Guardian/Independent 0.64
Financial Times/Times /Telegraph 0.87

So the party preference of every Sun reader was multiplied by 1.24 which is nearly double the value given to the choices of Indy/Guardian readers – a move designed by YouGov to ensure that their sample was representative.

The precise numbers change with every poll because it all depends on who responds. But invariably the views of the readers of “heavy-weight” papers, who are probably more likely to take part, are discounted while those of the “red-tops” are put at a premium.

Is it a good system? We will know better the day after May 5th – if that is, indeed, polling day. In the meantime YouGov will point to their record to show how much better they have been than the conventional pollsters in getting elections right.

© Mike Smithson 2005