26/04/2007

Lies, Damned Lies and Welsh polls #3


I have already explained on this blog* why I don't like Welsh opinion polls; basically all that have been published in the past have proved inaccurate.

The most interesting thing about the poll published by the Western Mail today was the percentage turnout that it predicts. The WM suggests that an unprecedented 64% of people will vote. The detail of how each party will do according to the poll is being saved until next week. A delay in publishing a poll, by definition, makes the poll unreliable and out of date. However if 64% of the electorate vote there can only be one result: a Labour majority. Labour will have at least 33 seats. Without having any inkling of what the Western Mails seat distribution claims will prove to be, I can (almost) guarantee that it won't show a Labour majority of three seats.

The Western Mail's poll is conducted in a different way to ITV's.

ITV takes a random selection of people from all of Wales and distributes the result equally thorough the country (weighted acording to UK trends). Because there is almost no relationship between the way the wind is blowing in the Rhondda and the way it blows in Aberconwy trying to predict the result in either place from an all Wales result isn't going to work.

The Western Mail poll has a fairly equal number of respondents from all 22 local authority areas, which might in theory appear to give a more balanced view and take local trends into consideration. In practice, what the Western Mail has is 22 mini polls rather than a single poll - but each mini-poll will have fewer than 50 respondents. The margin of error in a poll of just 50 people is enormous, so again this poll has no credibility.

The Western Mail used the same polling method to predict the results of the National Assembly referendum and predicted a huge majority in favour of devolution. (The yes vote in the referendum only scraped a majority by the skin of its teeth.) Plaid used the same company for its internal polls in 2005, but failed to realise that the party was to lose Ceredigion and fail to re-capture Ynys Môn.

Once again, the only conclusion that anybody interested in Welsh politics can come to, is that Welsh polls are bollocks; and that Wales deserves better from the polling companies and the news media that employ them.

See also: Wales Elects 2007 for more about the poll and links to other blogs discussing the subject

*Lies, Damned Lies and Welsh Polls
*Lies, Damned Lies and Welsh polls #2

3 comments:

  1. Cheers
    Valleys Mam

    ReplyDelete
  2. Another spot on post from the miserable one. Wales is just too diverse for the pollsters.

    ReplyDelete
  3. I've never trusted opinion polls since a student told me he'd worked on one for the Brecon and Radnor by-election back in the 80s. He filled in the forms in the pub and the poll was spectacularly wrong...

    ReplyDelete