Tuesday, May 15, 2007

Brown Bounce, Blair Bounce or Dead-Cat Bounce ?

I am not one of those who argue that opinion polls are worthless, but I do think that newspaper headlines based on them have to be taken with not just a pinch, but a bucketful, of salt.

You can usually find something in the small print of almost any poll which either in isolation or by comparison with the previous poll in the sane series can be used to spin any message you want to try to justify.

This morning's Times headline is a classic example - it refers to a "bounce for Brown" and says that the prospect of a new man at Number 10 has given the chancellor a "poll boost."

Is Conservative support down in this poll ? No.

Is there a significant drop in the Conservative lead when people are asked how they would vote in a general election (with Cameron and Brown as party leaders) ? No, Cameron still has a lead of 10%, compared with 11% last month.

As it is based on a survey of 1,504 people. this change is not statistically significant. A 10% lead for David Cameron, is however, statistically significant on a sample of that size.

And the "justification" in the text for the deadline? When interviewees were asked to rate leaders on a scale of 0 to 10, Brown came out 0.05 points ahead of David Cameron. Again, this is not statistically significant.

So The Times ignored evidence from their survey which gave good news for Cameron and was statistically signifcant, and printed a headline favourable to Brown based on tiny margins for him which were not statistically significant.

Whoever wrote this rubbish did not deserve to pass a statistics GCSE.

And whenever you see a survey, political or otherwise, quoted in the papers, always look at the small print!

No comments: