Voodoo polling corner
No proper national polls in the papers this weekend (though the Scottish Mail on Sunday does have a Scottish Opinion poll), but there is what proports to be a selection of constituency polls in the Sunday Express.
The Sunday Express claim to have constituency polls in Salford, Worcester and Basildon, which show the BNP ahead in Salford and the Conservatives ahead in Basildon and Worcester. Of course, in theory these could be properly conducted polls carried out with appropriate sampling techniques and weighted to be representative of the relevant constituencies. If so, the swings are off the scale.
I suspect however, they are nonsense - straw polls of the first 500 people the Sunday Express could grab off the street. We can't tell, but the warning signs are there. As well as the implasibility of the figures themselves,
there is no mention of the pollster who conducted the survey in the report, suggesting there wasn't one. Secondly, the results a quoted to a single decimal place, given most polls have a margin of error of around 3% false precision like this is normally a sign that the person who produced the figures doesn't know what they are doing. The cherry on the cake is that Salford and Basildon constituencies are both to be abolished at the next election anyway, to be replaced by Salford & Eccles and Basildon South & East Thurrock.
On the subject of voodoo polls, there was a vintage example on the London Daily News website, who took an open access poll on their own website which allowed multiple voting seriously. They updated the story saying "The poll referred to in this article has been removed due to a clear attempt by a political group to sway the results". I hate to break it to them, but the poll would have been meaningless even if there hadn't been orchestrated attempts to fix it. Polls that are open to all don't work, since people who are interested in the subject are more likely to take part, people who take part will not be demographically representative of the county, and because people are free to manipulate and fix the poll.
UPDATE: Even worse - Wales on Sunday had this shocker of a Voodoo poll, done just using people who happened to visit their website. They've even tried to extrapolate figures for each individual MP in Wales from the figures, which must be unrepresentative, unweighted samples of about 40. Naughty step, now.
