Comments on: Poll data, poll dancing

By: CJE

CJE — Sat, 05 Jun 2010 19:43:02 +0000

In the UK, yellow is a political color – the Liberal Democrats, who are now part of the governing coalition. Hopefully this graph was contextualized in a larger paper about UK polling/elections. otherwise the only hints are the choice of these three colors (red for labor, blue for conservative, and yellow for LD) and the “bang on” for the point of perfect accuracy.

By: selika

selika — Thu, 03 Jun 2010 21:29:22 +0000

“Red and blue are always political colors”

How exactly are red and blue political colours anywhere but the US?
In Australia there is no correlation between the two major parties and any kind of colour as far as I have noticed.

So I agree the graphs need a key, but assuming everywhere is like the US, is as always, not an accurate assumption.

By: cim

cim — Thu, 03 Jun 2010 13:14:41 +0000

I’d disagreed with that. There does seem to be a pattern of general improvement in the difference between the 1-day figures and the actual result, with the later elections doing better than the earlier ones (ICM, who were within the margin of error for the sample size on all three 1-day polls, are a slight exception). The polling companies have generally improved their performance over that time (if 1992’s election was included this would be even more obvious)

Meanwhile, there’s no reason you’d expect – even with a hypothetical perfectly accurate poll – the 1 year (or even 1 month) results to necessarily be close to the real results, because events can and do change people’s minds, and the question asked by the pollsters is not “how will you vote at the next election?” but “how would you vote if the election was tomorrow?” (a question that only strictly makes sense for the 1-day polls) so the lack of pattern there doesn’t necessarily mean a lot.

UK Polling Report – http://ukpollingreport.co.uk/blog/ has done a lot of good articles on the subject of polling accuracy.