Talk:Opinion polling for the 2015 United Kingdom general election/Archives/2016/January
Latest comment: 8 years ago by Bondegezou in topic Why opinion polls were wrong
![]() | This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page. |
Why opinion polls were wrong
A study showed the polls were biased against the Tories as they had too many Labour voters at the expense of the Conservatives http://www.theguardian.com/politics/2016/jan/19/general-election-opinion-poll-failure-down-to-not-reaching-tory-voters
(Coachtripfan (talk) 08:41, 19 January 2016 (UTC))
- I'm sure that there needs to be a comment on the page about the outcome of the BPC enquiry, but I don't think Coachtripfan is fairly summarizing. There is no evidence of intentional bias (and most commentators I've heard/read agree it worked to the Cons advantage, so if it was intentional, it backfired).
- I think a fair summary would be along the lines that "the BPC enquiry found that sampling techniques had not been good enough, and not enough conservative supporters were sampled, while young supporters of the Labour Party proved to be rather less likely to turn out to vote than had been expected by the pollsters." Thoughts/comments/suggestions/references welcome. DrArsenal (talk) 14:23, 19 January 2016 (UTC)
- Other thoughts: how about "pollsters kept finding too many politically engaged young people who said they would vote Labour, but didn't actually vote, and not enough over 70's who aren't politically engaged/interested, but actually turned out to vote Tory (it seems people who are over 70 and not politically engaged/interested aren't terribly likely to answer online opinion polls, or telephone polls where someone phones up and wants to ask them about their political views)". DrArsenal (talk) 20:14, 19 January 2016 (UTC)
- It's not that they didn't try to sample Conservative supporters, it's about what we'd call preferential non-response. I'd use wording like "sampling failed to reach as many... because of differences in response rates". Bondegezou (talk) 18:31, 20 January 2016 (UTC)