Voodoo corner
As we get closer to the election I expect we'll also get a flurry of what Bob Worcester calls "Voodoo polls" - open access polls that do not make any attempt to gather a representative sample.
In yesterday's Sun and Times, and today's Indy, we have a "poll" from a website for mothers, that claims to show mothers rushing to support Cameron and intending to vote in record numbers.
As far as I can tell, it was just an open access poll on a website for mothers. It will, therefore, be skewed towards mothers interested in politics who would have been likely to take such a poll(hence the high figure of them saying they will vote), probably towards mothers with an interest in the sort of issues discussed on websites for mothers, and certainly by the demographics of the people who read the site. We know that last one for sure, since the poll gave some demographic breakdowns of respondents which show that stay-at-home mums and part time workers were over-represented, full time working mums were under-represented, and low income mums and single parents were grossly under-represented. With that income skew it should be no surprise they were very Conservative!
Philip Webster in the Times can be partially forgiven, since he did at least add the caveat that the survey "cannot be regarded as in any way conclusive and is not an opinion poll [...] is not representative because it is self-selecting and the organisers accept that people responding to such a survey are more likely to be politically engaged than those who do not."
Unfortunately, he then went on to report the figures as if they did mean something - far better to have done what one editor who mailed me about it on Tuesday, and one veteran political editor who I spoke to today did, and not touch it with a bargepole.
If you see something in the papers proporting to be an opinion poll and it is not from an established polling company, read it very carefully. How did they gather the sample, who are they claiming it represents the views of, and what measures did they take to make sure the it matched the demographics of that group. If the answer is they took whoever came along, and didn't take any measures at all to make it representative, I think you know how useful it is.
