Rogue Pollster

Of interest to us poll anoraks amateur psephologists, Anthony has an interesting post on the different methodologies used by the different polling companies to produce their results.

One point that I think is worth making on the different results is that while the methodologies can produce a variability in the reported results which is especially notable now when different companies place different parties in the lead, when viewed together they can show what the current trends in voting intention are. For instance, polls over recent months have shown Labour losing support but the voters that are deserting Labour are moving to the Liberal Democrats or Others rather than to the Conservatives.

To my mind, watching the trends over time is a useful way of firstly, noting who's winning the current arguments and what way voters are moving in response, and secondly, they provide a good way of spotting what may be rogue or 'blip' polls - the latest Guardian/ICM poll which shows Labour gaining votes from the Conservatives against the current trends of the other polls may well be a rogue poll (though it may also herald the start of a new trend). The important thing is that if a company has produced a rogue poll in one month, then it's results for the next month may then overexaggerate any trend as it corrects itself.

What would be useful, of course, is a Poll of Polls averaging out the data from the four companies to give a better perspective of the trends over time. I don't know if Anthony's patented Gnome Industries election swingometer can manage that though...