Talk:Opinion polling for the 2015 United Kingdom general election/Archives/2014/May

Polling in Individual Constituencies

I have started the process of adding tables for the remaining individual polls and I am about halfway through and it's just getting ridiculous. It's making the article much bigger then it needs to be. I really think we should split these individual polls off into a separate article, whilst obviously ensuring that they are well linked to other relevant articles such as this one. Can we try and achieve consensus for splitting this article. I notice that CH7i5 is in favour and others have concerns about the sheer size of the article and the fact that somewhere between 20 and 25% of the data has yet to be entered as the election isn't for another 375days. Owl In The House (talk) 10:44, 1 May 2014 (UTC)

Also lets not forget that there will more than likely be more graphical summaries to include as well and heaven knows whatever else, the measures, data and forms of analysis for elections seem to be growing and growing. Owl In The House (talk) 10:46, 1 May 2014 (UTC)

Boldly done. New article created at Opinion polling in individual constituencies for the next United Kingdom general election. Bondegezou (talk) 13:20, 1 May 2014 (UTC)

Many Thanks, problem solved Owl In The House (talk) 14:02, 1 May 2014 (UTC)

ICM Wisdom Index

I don't think the ICM Wisdom Index should be included, as it is not a normal opinion poll. The question is along the lines of "who do you think will win?" rather than "who will you vote for?". It has never been included before, and I think this is the right policy. Saxmund (talk) 07:58, 4 May 2014 (UTC)

I agree that it certainly shouldn't be included in the normal tables: it asks a quite different question. I'd have no objection to a separate table just for the Wisdom Index if people thought that useful. Bondegezou (talk) 09:25, 5 May 2014 (UTC)

I have decided to delete them, partly because the editor posts anonymously and doesn't have a talk page, so I can't discuss it with him. Have also added the most recent two YouGovs. We need some different polls! Saxmund (talk) 21:21, 5 May 2014 (UTC)

Hello. I added the most recent Widom index. I don't see the objection to them. When they first came out, ICM said they'd been using them internally before the 2010 election, and they'd been more accurate than their standard poll: " In our final poll before the 2010 General Election among a random sample of just over 2,000 people across Britain, ICM added a few questions asking people what they thought the result of next day's actual election would be. We were surprised to find that the collective judgment of the crowd outperformed our own 'classic' opinion poll methodology, which itself still produced the most accurate prediction of the election, according to the British Polling Council's evaluation."86.2.12.245 (talk) 08:32, 6 May 2014 (UTC)

I have no objection to including the Wisdom index somewhere, but they are a distinct and different thing to regular opinion polling, so including them in the same table is misleading, I suggest. By all means, include them in a separate table. Bondegezou (talk) 09:29, 6 May 2014 (UTC)

The problem with the Wisdom index is not whether or not it is accurate (we will find out at the General Election) but that it asks and entirely different question. So by including it in the main table you would be comparing different types of things. This seems to have been the first one published where they say:

"The Wisdom of Crowds (Surowiecki, 2004) suggests that the collective estimation or decision-making of a random crowd is superior to that of the smartest people within it. Utilising this logic, the Wisdom Index is a brand new way of predicting elections or assessing the current state of the parties. In contrast to traditional political opinion polling, it moves away from a focus on individual human behavior (how will you vote?), to one where outcomes are the primary concern".

It will obviously be interesting to see how it performs, and as Bondegezou suggests, maybe a separate table would be best. Saxmund (talk) 17:21, 6 May 2014 (UTC)

Populus Data

The individual or individuals inputting data from Populus are using two different numbers for Sample Sizes. Some of the information uses the Total Number of Respondents. Some of the information uses the Likely to Vote Respondents. We should make this data point consistent instead of bouncing back and forth between the two choices. — Preceding unsigned comment added by 129.110.242.7 (talk) 15:31, 8 May 2014 (UTC)

I always aim to use the Total Number of Respondents for all polls, I try to follow previous practice and this seems to be what is usually used for all pollsters Saxmund (talk) 19:33, 10 May 2014 (UTC)

Actually it appears to be only one poll (30 Apr - 1 May) so I have amended it as clearly being out of step with all the others Saxmund (talk) 19:36, 10 May 2014 (UTC)

This introduces statistical errors, though. The Number of Respondents and the Vote Choice percentages should be taken from the same part of the survey. Using the Total Number of Respondents and then the percentages from the Likely to Vote portion of the statistical analysis is misleading as Populus's data for the Total Number of Respondents includes a "Don't Know" category. — Preceding unsigned comment added by 67.198.48.19 (talk) 17:55, 12 May 2014 (UTC)

Lord Ashcroft poll 9-11 May

Was wondering why 86.128.203.224 undid my addition of the new Lord Ashcroft poll without posting on here or on my talk page. I have since undone the revision. Saxmund (talk) 16:15, 12 May 2014 (UTC)

Ashcroft is (for some reason) keeping the company who conducted it secret, which may raise suspicions (in fact technically I think that'd fail BPC), especially as he's Tory in Chief. A little look shows it to be fine (and Populus' work), but it's understandable why some might think it voodoo at first glance. 92.1.34.203 (talk) 18:39, 13 May 2014 (UTC)

My guess is that the reason for keeping the company that did the fieldwork "secret" is probably no more than contractual. Anthony Wells confirms it is Populus, and both he and Mike Smithson appear to judge it to be a proper poll by the way they discuss it on their blogs. Ashcroft's polls have been listed before, and we have had other non-BPC polls such as BPIX. Similarly, I don't think it is relevant to include Ashcroft's non-BPC status in the table, but maybe he should have an entry in the section on polling organisations. If YouGov left BPC tomorrow it would make no difference to the accuracy of its polls.Saxmund (talk) 21:26, 14 May 2014 (UTC)

Agreed entirely, was just suggesting why someone might have mistaken it for not-belonging. Gonna change that note, as commissioners are never BPC registered it makes little sense. 92.1.34.203 (talk) 14:02, 15 May 2014 (UTC)

I intend to add an item under Methodology under Lord Ashcroft and then remove the note in the table. I think we should treat him as a "polling organisation" who commissions the fieldwork from another polling company but is responsible for his own analysis.Saxmund (talk) 18:38, 18 May 2014 (UTC)

ComRes 40 marginals poll

Clearly this new poll surveying 40 marginal seats can't be included in the main opinipon polling table on this page as it is not a national poll. So I have taken the bold move of finding a way to accommodate it in the article individual constituency polling page. I have created a new table in order to do this. I dare say this won't be the last of this sort of multiple marginal poll. I've just used a standard polling table and added the column "seats surveyed" representing the number of seats surveyed. If anyone thinks of a more appropriate name for the table or anything else, do chip in. Even if there are only a few of this sort of poll, this table does seem to be the best way of reflecting this data. Including this specific poll and this sort of poll in that article further justifies it's existence. I somehow had a feeling another sort of electoral measure would crop up. Owl In The House (talk) 10:54, 21 May 2014 (UTC)

That's fine by me. In fact, if you look on Lord Ashcroft's website I think you will find a couple of earlier ones. Saxmund (talk) 19:04, 21 May 2014 (UTC)

Total Percentages

Are you guys checking the Total Percentages?

Most values fall between 99% and 101%, but there are a few that are too low or too high. I'm double checking the entire list going back all the way to 2010. I'm making corrections as I find these errors. — Preceding unsigned comment added by 67.198.48.19 (talk) 21:44, 26 May 2014 (UTC)

How are you calculating them? My practice is

Where Others figure is given in the polling tables, to use that

Where it isn't, calculate it arithmetically by subtracting the Big 4 figures from 100. I could do it by adding up all the minor parties but it seems to me this will introduce even more of a rounding error.

I do copy and paste but try to remember to replace all the figures. Surely the range 98% to 102% is possible (four major parties rounded up by up to 0.5% each) if the pollster gives a figure for Others?Saxmund (talk) 19:37, 27 May 2014 (UTC)

It looks like there are several instances of human error. For example, all the % for Conservatives, Labour, Lib Dems, and UKIP are correct. However, the person who entered the data also used the aggregate "Other" % without subtracting UKIP.

There is one instance, 10-12 Dec 2010 IPSOS MORI, where the numbers add up to 102% and can't be reduced down to the 99-101 range. — Preceding unsigned comment added by 67.198.48.19 (talk) 22:08, 27 May 2014 (UTC)