Thursday, February 18, 2010

The super soaraway subsample swindle

Anyone who follows discussions on polling in the blogosphere will know that whenever anyone (such as yours truly) has drawn attention to Scottish subsamples from UK-wide polls, it's always been quickly pointed out that the results are highly unreliable - not merely because of the small sample size, but more particularly because the sample is not properly weighted and will not reflect the characteristics of the Scottish electorate as a whole. My response has always been that it's equally foolish to go to the other extreme and suggest that the figures are utterly meaningless and 'just a bit of fun', and I stand by that, but nevertheless it's been quite jaw-dropping over the last eighteen hours watching some (I emphasise only some) of the exact same people undergo a Damascene conversion on the matter. Apparently, a subsample is utterly authoritative just so long as the media (ideally the Sun) tells them it is - and naturally some very mildly favourable figures for the Scottish Tories only help to consolidate this thrilling new world view.

The figures are -

Labour 37%
SNP 21%
Conservatives 21%
Liberal Democrats 15%

There is some confusion over at Jeff's site as to whether these figures might come from a full-scale Scottish poll (understandable, given the Sun's deliberate, shall we say, lack of clarity on the matter). However, if they did, given YouGov's standard practice the detailed breakdown would normally be on its website on the first working day after the headline figures were released (ie. today). It's not there, and the Caledonian Mercury's reporting leaves little room for doubt that this is simply an aggregate of two subsamples from UK-wide polls.

No comments:

Post a Comment