Talk:Opinion polling in United Kingdom constituencies (2010–2015)

Latest comment: 7 years ago by Dunarc in topic Plymouth Sutton & Devonport

East Dunbartonshire poll

edit

A Survation poll commissioned by the LibDems (like the Hornsey & Wood Green we've included before). Sorry, don't have time to add myself right now. Bondegezou (talk) 14:50, 16 April 2015 (UTC)Reply

I think polls commissioned by the parties themselves would fail as a reliable source. The party can use any wording for the question to get the answer they want - the polling company simply asks the questions that they have been told to. If you do want to add this, it should be added as a LibDem poll rather than a Survation Poll as it wasn't done independently by the polling company. Survation have actually publically reproached the libdems for leaking privately-commissioned poll data, and emphasised that " These polls should therefore not properly be described as “Survation polls” http://survation.com/in-reference-to-recent-liberal-democrat-polling-shared-privately-with-the-media/ Little Professor (talk) 20:25, 28 April 2015 (UTC)Reply

The way Hornsey & Wood Green is done is problematic. Fair enough, it's been labelled as a LibDem poll with fieldwork by Survation, but how do you decide which set of figures from the poll to use? The report has raw normal weighted figures of 38.1% for Labour and 35.4% for LibDem once don't know/refused are excluded. But then the next table has the undecideds and unknowns excluded and replaced by a 0.3 factor of the 2010 vote. This leads to the 37.3% Lab/36.1% LD figures that are in the table, making the race look closer than it is. Other survation polls either don't have this methodology (taking a factor of the 2010 vote), or if they do, the table on wiki still uses the non-factored figures. Little Professor (talk) 20:39, 28 April 2015 (UTC)Reply

Sample size

edit

For the Ashcroft polls, some of the tables are using the total number of people surveyed (i.e. either 1000 or 500) as the sample size. Others are using only the "Certain to Vote" subset, which is generally any number between 500 and 700. I don't really have an opinion one way or the other as to which figure we should be using, but we should at least have consistency Little Professor (talk) 20:19, 28 April 2015 (UTC)Reply

New (but which I mean old) poll

edit

Should we add this? Bondegezou (talk) 15:13, 19 May 2015 (UTC)Reply

Plymouth Sutton & Devonport

edit

The poll here seems to have an error as the lead is give as 8% over Con, but it look like the Lab Con gap reported was 13% (39-26 - or are the figures wrong?) Dunarc (talk) 19:27, 2 June 2017 (UTC)Reply