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

This page 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.

Is it just me, or did Lord Ashcroft's poll earlier in the week actually give a Tory lead of 1% (32 v. 31) as opposed to the published tied vote (32 each for Lab and Con)?

Tables here

weighted sample without adjusting for don't know/refusers: 501

Lab 159 (32%) Con 158 (32%) UKIP 87 (17%) LD 39 (8%)

assuming 50% of 2010 don't know/refusers are added "back" to their party

159+12 = 171 Lab 158+18 = 176 Con 87 = UKIP 39+12 = 51 LD

New weighted sample adjusting for DK/ref = 543

New percentages:

Con 32% Lab 31% UKIP 16% LD 9%

regards, Sunil060902 (talk) 23:52, 1 October 2014 (UTC)

I would think a discussion like this would be more appropriate on a forum like ukpollingreport. The likely explanation btw is rounding. The 171 people wasn't exactly 171, but weighted and not an integer. I would deduce it was weighted to somewhere 171.36 to 171.98, made up by adding two numbers that get rounded down to 159 and 12 respectively, but because they add to >171.35, the total gets rounded up in the calculation of the %.

Dr Arsenal 146.227.159.61 (talk) 13:00, 2 October 2014 (UTC)

If I might chip in. That may explain the Con and Lab ending up equal, but not the UKIP and LD figures. Remember that UKIP "people" is unchanged in the adjustment. 87/501 v. 87/543. 89.243.100.230 (talk) 01:32, 3 October 2014 (UTC)

I don't know, I leave you alone for a bit and this happens....what the hell!?! This two tables thing is just so much nonsense....I see the "debate" got really confused, can't really see Saxmunds or Owls point...but surely this isn't the answer? If/when the "detailed" table gets done, what on earth is the point of the "normal" table? How this was in any way the consensus is beyond me... On another note, you appear to be reporting 0% as <1%. The two are not the same. EDIT:Notice you've already noticed that. Also, how come no Respect column? I know they're dead, but not so much as BNP. Iliekinfo (talk) 16:55, 3 September 2014 (UTC)

As far as I know (and Impru's been doing most of the hard work of adding these), the detailed poll table includes any party listed by the pollster. If they've not listed Respect, we don't list Respect.

I would suggest we stick with two tables for now, see how it goes, and then return to the issue if people think it's the wrong approach. Bondegezou (talk) 17:24, 3 September 2014 (UTC)

Looking into it, it seems that some pollsters include the BNP & some don't, but only yougov include Respect. Allowing for this inconsistency is one (possibly the only) advantage of my old suggestion. I don't see the two tables working out, I certainly don't see how people are against the "Proposed Changes to Inclusive System and Aesthetic Design" but in favour of this....which is exactly the same but doing it half all over again. I'd support doing it the same as every other opinion polling page (Proposed Changes to Inclusive System and Aesthetic Design), or doing it my way, or just the "old" table, or just the "detailed" table. This doubling is madness....every new poll has to be added twice? Let alone the task of backdating everything, and fixing those <1%s. Iliekinfo (talk) 17:34, 3 September 2014 (UTC)

The "0" of YouGov polls are shown as <1% for the simple reason that YG tends to round its vote results to the nearest integer. As an example, it is perfectly possible for 1 person to have stated they would vote for BNP, yet in % terms they may only score, let's say, 0.1%. That's not 0%, but YG rounding would effectively show them as 0%. This applies not only to BNP but also to 'Others', which YG show as 0% in many polls, and which is obvious that would not get 0 votes. Impru20 (talk) 17:38, 3 September 2014 (UTC)

Actually, those polls I changed to 0 are exactly 0, pollsters usually report figures such as 0.1% with an * (or, they actually give 0.1%), these are polls where the BNP really did get 0 votes. Iliekinfo (talk) 17:43, 3 September 2014 (UTC)

YouGov does not do that. They have never shown parties with 0.1% with a *, nor do they show the number of respondents saying which party would they vote. YG just shows percentages rounded to the nearest integer. Otherwise, it's just impossible that in some polls, YG records 0 votes going to 'Other' parties (seriously, no one?). Impru20 (talk) 17:47, 3 September 2014 (UTC)

Whoa, you're right, my apologies. I was going through tabs of polls and still thinking in populus layout where they give full figures. Iliekinfo (talk) 17:52, 3 September 2014 (UTC)

We had an RfC, and the closure of that RfC was reviewed and approved. I see little point in re-opening the discussion right now given that and given most have accepted the resultant plan. Let's see how it goes and if lots of people are unhappy at some future date, we can of course re-examine the issue. Bondegezou (talk) 20:21, 3 September 2014 (UTC)

Well yeah, I wasn't trying to remove it all on my lonesome, just think it's a really really nonsensical conclusion is all. Also that closure seems a bit of a nonsequitor from the RfC :)Iliekinfo (talk) 21:05, 3 September 2014 (UTC)

I wasn't all together keen on the 2 tables idea but thinking about it, it is reflective of what our reliable sources report. Without doubt the headline polls (which only include the 4largest parties) are the main focus, the pollsters themselves highlight them, the Newspapers, online news sites and TV networks only focus on the headline polls. That said other parties are consistently represented in more detailed tables, so I do think this in many ways is the best of both worlds, provided this second more detailed table remains a minor piece of the article and not the main focus (that would not be reflective of reality at all).

As for the original proposals, well that discussion has been had, voted on, rfc'd, you name it we've done it. The reality is it did not reflect reliable sources or indeed political reality.

On the subject of adding the Respect Party being added to this detailed table, I have raised this before in a kind of indifferent way but I would be in favour of including them in this table if there is sufficient data for them (we don't really want a collum mainly filled with dashes or merged with the others column. I hold the same position for the SWP or whatever other party (I realise Respect is probably a borderline case - hence I'm on the fence). Owl In The House (talk) 21:53, 3 September 2014 (UTC)

A largely blank column poses some problems. If a party is only occasionally mentioned in polls, could we use footnotes or something? Bondegezou (talk) 21:58, 3 September 2014 (UTC)

The reason I say it's borderline in my view is because YouGov do include them as far back as at least late 2011 (polls before that don't open) and YouGov do account for most of the data and since this is a detailed table, part of me thinks that we'd be skimping on detail. I guess the question is would there be more numbers then dashes or dashes then numbers. Owl In The House (talk) 21:01, 4 September 2014 (UTC)

Opinium also list Respect. Thinking about it, I'm more inclined to say they should be included in this table, it is meant to be a detailed table afterall and it is detail we have, I can't see a good reason not to include them given that we now have this second table. Owl In The House (talk

) 15:38, 5 September 2014 (UTC)

I would tend to agree with Owl in the house's earlier view (4th Sept) that there nothing to be gained if there is no real data in the column. Can anybody find a national poll where Respect weren't <1%? Thinking about criteria so we could make a rational decision for what to do after the General Election, I would suggest that a column isn't justified for a party if most of its polling (perhaps over the last 10 polls) is <0.5% - so not even rounded up to 1%. I would make an exception if/whenever the party manages a 2% in a single reputable poll on the basis that we have some interesting data, so it does deserve a column to show that interesting data. 85.133.27.19 DrArsenal, normally editing from 87.112.138.6885.133.27.19 (talk) 07:22, 2 October 2014 (UTC)

undent The columns can only really change at the start of each year, unless we make each month or whatever separate tables (which seems a bit finicky), so the 'last 10 polls' thing doesn't really work. I'd suggest whatever criteria should be based on how many pollsters included them in the previous year. Maybe any party that gets 3 pollsters including in a year, gets a column in the next years table? Or something along those lines. Iliekinfo (talk) 11:04, 3 October 2014 (UTC)

Sorry, I didn't make myself clear. I didn't mean a constant running 10 polls, so much as a 'last 10 polls' when a decision is made - which I would be happy to be an annual decision. But equally, the number of different pollsters that included them in the previous year would be a reasonable criterion. DrArsenal 14:19, 3 October 2014 (UTC) — Preceding unsigned comment added by 87.112.138.68 (talk)

Wondering if this http://lordashcroftpolls.com/wp-content/uploads/2014/10/Post-Conference-Poll-Full-tables-Oct-2014.pdf should be listed as a VI poll. It was published in the Sun on Sunday without headline VI being published, presumably because of the SoS's arrangement with YouGov. However the tables weighted politically and for turnout give Con 31, Lab 34, LD 7, UKIP 18. As this is a major poll of over 5,000 respondents it would be a shame to miss it, however does it count as OR to lift figures off tables in this way? Saxmund (talk) 10:06, 12 October 2014 (UTC)

Yes, include it. Bondegezou (talk) 10:25, 12 October 2014 (UTC)

Could we add polls with seats?

http://www.itv.com/news/update/2014-10-11/study-ukip-could-win-128-seats-in-next-election/

Labour 253 Conservatives 187 Ukip 128 Lib Dems 11 Other parties 71

Last updated Sat 11 Oct 2014 81.58.144.30 (talk) 11:24, 12 October 2014 (UTC)

Not like that. That's nonsense speculation to make a good story, there's no easy way to convert national poll results into seats, and if there was an agreed method, it definitely wouldn't be that one. You might find this (incomplete) article

https://en.wikipedia.org/wiki/Opinion_polling_for_the_next_United_Kingdom_general_election_in_individual_constituencies interesting though: the actual number of 'UKIP in 1st place' seats according to polls is 7. Iliekinfo (talk) 22:00, 12 October 2014 (UTC)

While adding detailed poll results, I am sticking to the previous practice of indicating <1% for poll results that are >0 and not enough to be reported as 1% by the pollster. However, given practices of rounding, anything 0.5% and above would be rounded to 1%, so I keep feeling that <0.5% would be a more accurate description of that poll result for the party in question. Should it be changed to the more factually accurate <0.5%?

In a similar vein, Ashcroft these days appears to report every party on 1% or more in his 'headline' results, including all <0.5% in the 'Others'. Thus we know that any blanks in a row of Ashcroft are <0.5%, but the reader of the page could get the impression that they are >1% since they are then bundled into an 'others' that is >1%. Should we do something to disambiguate this? DrArsenal 87.112.138.68 (talk) 18:26, 3 October 2014 (UTC)

In the absence of objection, I've just changed <1%s to <0.5% (over 200 of them!). DrArsenal 46.208.137.165 (talk) 22:18, 15 October 2014 (UTC)

There seems to be some inaccuracy with the graph still in relation to both Labour and UKIP. While the other parties are accurately portrayed Labours last 20 results clearly shows only 3 results of 36 up to 38 while other polls show them as low as 30 and yet they are ahead on 36%. The current figures would put them down 2 on 33.8%. Even with the last 10 results they should have dropped by 2%. UKIP on the other hand despite quite a number of polls being at 16 or more percent including the recent 25% is still around the 15% mark when it should be up 2 on 16.9%. — Preceding unsigned comment added by Robhandford (talk • contribs) 23:46, 15 October 2014 (UTC)

Note the note below the graph. It was last updated Sept 26th. You'd be welcome to update it yourself, but it'd be a LOT of work. IIRC its from an excel spreadsheet of every poll since 2010. For what it's worth, UKPRs weighted average up to 10th Oct has Lab on 34, UKIP on 16, so you are ~right that they've changed a little in Oct. Iliekinfo (talk) 00:21, 16 October 2014 (UTC)

Why is there two ComRes entries for 16th October 2014? best, 188.222.179.74 (talk) 22:40, 18 October 2014 (UTC)

Because there are two ComRes polls, done with different methods over a splitted sample of 2,000 people. See it here. Impru20 (talk) 22:50, 18 October 2014 (UTC)

For 9th Oct it currently verbosely says - "The UK Independence Party wins its first elected Member of Parliament after the sitting MP switches allegiance from the Conservatives and resigns his seat to contest the resulting Clacton by-election. Labour narrowly retains its seat at the Heywood and Middleton by-election, despite opinion polls predicting a wide lead." I really can't see how such a lengthy description is appropriate in a table that is meant to be about opinion polls. Can we reduce it please? I would suggest that "UKIP wins first MP, after sitting MP defects and resigns Clacton seat to contest by-election. Labour retains seat in Heywood & Middleton by-election." is factual and sufficient to give the context to the changes in Tory and UKIP shares before and after. Dr Arsenal 46.208.137.165 (talk) 22:15, 20 October 2014 (UTC)

• I've changed it back to what is normal for by-election results throughout the rest of the article. No need for analysis or bias in a section about data. 123.2.85.195 (talk) 20:24, 23 October 2014 (UTC)

Thanks 123.2.85.195 fo the comment here. I note 86.163.137.11 has been adding 'narrowly' back in without edit summary - I have undone (for now at least) DrArsenal 46.208.137.165 (talk) 20:53, 23 October 2014 (UTC)

I believe the 'narrowly' adjective should be kept in the table for the case of the Heywood & Middleton by-election. Opinion polls shortly before the elections were predicting wide Labour leads which were finally not materialized; thus, the election result came as a surprise to many. The fact that Labour only narrowly won the seat against UKIP could have had (and indeed it seems to have had) some influence in the recent UKIP's boost in opinion polls, so I believe it is a noteworthy event in this particular case. Impru20 (talk) 21:36, 23 October 2014 (UTC)

If there had been a similar poll boost without the Clacton by-election, I think it would be justified to include 'narrowly', but as it is, the text starts with reference to UKIP winning in Clacton, which could be adequate explanation by itself. Meanwhile, Impru20, should there be a Notable Event row to explain the SNP's part of the increase in 'others'? And what do you put the increase in the Greens' portion of 'others' down to? DrArsenal 46.208.137.165 (talk) 21:53, 23 October 2014 (UTC)

The table of opinion poll data simply isn't the right place for political analysis. The best place for this is their respective by-election articles. The results of elections and other notable events should be summarised as short as possible to convey the actual results. 123.2.85.195 (talk) 05:17, 24 October 2014 (UTC)

Agreed. We should avoid editorialising and keep such notes plain and simple. Political analysis, properly referenced, can go in the by-election article. Bondegezou (talk) 12:17, 24 October 2014 (UTC)

Trouble is non-logged in editors keep changing it without even using the 'edit summary' to justify, let alone coming here to discuss. DrArsenal 46.208.137.165 (talk) 14:50, 24 October 2014 (UTC)

I know support for the Green Party in the polls has been all over the place recently: some polls have its support as low as 2% but the Lord Ashcroft poll I just added in there has their support at 8%, above that of the Liberal Democrats on 7%. However, I have had to lump the number for them in with the 'Others' category. It is with this in mind that I would like to ask: just so we're clear, in what circumstances would we add the Green Party to this table, in a separate column from 'Others'? Would it only happen if a larger number of pollsters included them as a 'main option'? Would they have to be consistently above the Liberal Democrats after a certain number of polls? Any assistance you can give in this matter would be much appreciated --Rayboy8 (talk) 09:52, 21 October 2014 (UTC)

I don't think we can make up a rule ahead of time (see

WP:LOCALCONSENSUS

for the dangers in doing that). It's something that will have to be dealt with when we get there. There have been many long debates about this: first around including UKIP and now around including the Greens. At the moment, we have two sets of tables: the first table with Con, Lab, UKIP, LibDem and Others, which reflects how most polls are reported in the UK (although some press reports do include the Greens nowadays), and a detailed table that includes SNP, Plaid, Greens and BNP, which reflects the detailed results seen from polling organisations. This compromise of two tables is not to everyone's tastes. One suggestion is that, once the detailed table is more complete, we just use that and drop the first table. However, there is far from a consensus behind that.

It seems to me that the decision of which parties to include in the first table (if it is kept) has to be based on basic Wikipedia principles, i.e. Wikipedia reflects what reliable sources do. So I don't think we should attempt to construct a rule based on how many times the Greens are above the LibDems. Whether pollsters include them as a 'main option' seems, to me, more pertinent. Whether newspapers, TV and academic sources include them in their polling reporting or summarising, that is the most important criterion. Bondegezou (talk) 11:26, 21 October 2014 (UTC)

Given that the detailed table is relatively recent, and I have only managed to backdate it to mid-April so far I also don't think now is the time. HOWEVER, I would suggest that it is appropriate, in the meantime, for there to be footnote on relevant row(s) of the table pointing out when LDs are behind, or level with Greens (or someone else, like the SNP), and the share of the party ahead/level with LDs. DrArsenal 46.208.137.165 (talk) 13:57, 21 October 2014 (UTC)

I strongly oppose a note saying when the LDs are behind or level with the Greens. We had this issue before with the long debate about whether to note who is in third place (UKIP or LD) and rejected that. It strikes me as too much editorialising: it fails

Next United Kingdom general election article or the articles about the relevant parties noting the decline of the LDs and/or rise of the Greens over this period, with an example of that being the Greens overtaking the LDs in some polls. Bondegezou (talk

) 13:59, 21 October 2014 (UTC)

If a note saying when LDs are behind or level with Greens fails

WP:BIAS, I suspect that wasnt the page you were trying to refer to. Dr Arsenal 46.208.137.165 (talk

) 22:10, 21 October 2014 (UTC)

You're right, I was thinking of

Next United Kingdom general election summarising party standings over the Parliament. Bondegezou (talk

) 12:25, 22 October 2014 (UTC)

I've looked at

WP:SYNTH (despite people ignoring other violations on the page), then OK the footnote wouldn't have to mention the mathematical relationship with the LD share. To illustrate what I mean, I will edit the page, pro tem. DrArsenal 46.208.137.165 (talk

) 14:25, 22 October 2014 (UTC)

There are three issues here. (1) Is it noteworthy that the Greens were higher than the LibDems in that one poll? My answer: maybe it is, maybe it isn't, but that's not a decision that is up to us as individual editors. We must follow reliable sources (and reflect what they say in a

boldness. I think, however, that it doesn't work. We end up back at the whole, tortured discussion of whether to have one table, or two tables, or which parties to include. You've given us another option there, but the same underlying questions remain. And why give a footnote for that one poll? That, it seems to me, is giving undue greater prominence to the Others result in that one poll. Bondegezou (talk

) 16:50, 22 October 2014 (UTC)

Ukpollingreport would, according to earlier discussion on this page, count as RS. That source doesnt just note Greens higher than LDs, but also that "it’s a symptom of the genuine rise we’ve seen in Green support over recent months." - so while it might be a 'blip' in being the only poll higher than the LDs for a while, the genuine rise in Green support, as noted in further reliable sources, has hitherto not really been noted on Wikipedia (eg in the "graphical summary"). By contrast, the UKIP rise since the turn of the year is noted with 9 Oct 'notable event' row, and the two for the May election results, and the 26th March debate, as well as the 'graphical summary'. Green gains of council seats or MEP are not noted there - on their own they might be reasonable decisions, but I think it would be hard to justify that the UKIP rise is more than 5 times as notable as the Green rise this year, and the footnote is a lot less noticable than a 'notable event' row, let alone the 'graphical summary'.

Yes, my edit was intentionally in the spirit of

BOLDness. I would have thought 'others' being twice (or more) the share of one of the parties listed justified giving a footnote for a particular poll - unusually high polling often gets mentioned in RS, indeed newspaper reports tend to accompany unusual numbers from a pollsters far more often and far more prominently than routine polling numbers. DrArsenal 46.208.137.165 (talk

) 17:47, 22 October 2014 (UTC)

undent DrArsenals edit is very similar to a suggestion I made last time there was discussion on this point. (see 'Experimental table changes' in this archive https://en.wikipedia.org/wiki/Talk:Opinion_polling_for_the_next_United_Kingdom_general_election/Archives/2014/June ) However, I reckon it's been rendered moot by the addition of the detailed table. I think we have consensus on this issue...though depending on whether the detailed table ever gets completed or not, it could change....so will be removing DrArsenals note. Iliekinfo (talk) 21:57, 22 October 2014 (UTC)

I would say that to a real extent the situation is different from June: The Sun (@Sun_Politics) are now routinely tweeting Green shares, UK Polling Report is routinely giving Green shares, newspapers are reporting on how Labour in Scotland is suffering loss of support to SNP and how Green Party polling is having an effect on Labour and LibDems, etc. So reporting by reliable sources is in a different place from in June.

In addition, what I was suggesting wasn’t by any means the same as that proposal from June. What I was suggesting was an occasional footnote, not something that would add to every row, let alone something that made extra content routinely visible.

Even though I wasn't in any way involved at the time, I am not trying to overthrow the outcome of the RfC process: indeed, I think I am putting many times more effort than anybody else into implementing the ‘detailed’ table. However, since circumstances have moved on (both in reality and in reporting in RS), I would still argue that an occasional footnote would enable the outcome of the RfC to adapt to the changed circumstances, while being a sufficiently small tweak to be essentially within the spirit of that outcome. DrArsenal 46.208.137.165 (talk) 21:10, 23 October 2014 (UTC)

Agreed, the circumstances have changed since June. I propose that we include the Greens in the table from, perhaps, September onwards (whenever it is determined they have been getting sufficient attention to be fairly compared to the other four), but don't backdate them. The RfC determined the Greens weren't popular enough to be included a few months ago, so it shouldn't be backdated past then, but now the circumstances are different, IMO warranting inclusion. Chessrat (talk) 12:49, 24 October 2014 (UTC)

Just seen at ukpollingreport comments a comment from 'Richard' at 10.22pm "I see the Greens are now important enough to be regularly included in the headline figures/ tweets." I am not 'Richard' don't know who 'Richard' is, but from following UKPR, haven't seen much sign of him being a Greenie. What 'Richard' says is true: Greens are routinely being reported in reliable sources. The circumstances at the time of the RfC no longer hold. Something needs to change on this page. DrArsenal 46.208.137.165 (talk) 22:57, 26 October 2014 (UTC)

Who are the key RS reporters of UK politics? There's the BBC: their poll tracker has Con, Lab, UKIP, LibDem and Others (no Greens). The Guardian's Datablog has an archive of their polls that just covers Con, Lab, LibDem and Others (no UKIP, no Greens). Looking at more individual articles, I tried searching for "opinion poll" at The Independent's website and one of the first hits was this, a poll for Rochester & Strood, including Con, Lab, LibDem, UKIP, Greens and Others. Ditto at the Evening Standard brought up this article, which is about polling, rather than including any particular polling results, and mentions Con, Lab, UKIP and LibDem, but not the Greens. Ditto at the Daily Mail produced this article focusing on UKIP that also mentions Lab, Con, LibDem and Green, and this one which does the same. It seems to me that the Greens are sometimes being routinely reported along side the other 4 parties, but this is not a universal phenomenon.

My favoured solution would be to dump (or at least de-emphasise) the table with only the top 4 parties and focus instead on the table showing all parties listed in the polling results. I would also favour adding some explanatory text saying how the pollsters highlight the main parties and not others. Bondegezou (talk) 14:20, 27 October 2014 (UTC)

The problem with only including the whole table is that it's rather unwieldy (too much formatting/footnotes that admittedly can't be avoided) and probably violates

WP:UNDUE. While the attitudes of various sources differ, it's clear that the only parties regularly mentioned are Cons, Lab, LibDems, UKIP, and Greens. If it's decided that the current layout of main party table+full table is the way to go, then it's clear that including Greens is more representative of the aforementioned reliable sources than excluding them. To keep the main table consistent within each year, I propose that the Greens are included in the main 2014 table, but not in the previous years when they didn't have the same levels of support, and better belong only in the Full Party Table.Chessrat (talk

) 23:43, 27 October 2014 (UTC)

Bondegezou - I'm not sure it makes sense to pay attention to BBC or Gruaniad poll trackers. Both have a strong element of wanting to maintain the sequence, and are likely to just keep doing what they have been until it becomes completely untenable, because they don't want to spend the resource of backdating. As it is, BBC don't provide any rationale for including some YouGov and not others, and their explanation of how they calculate 'others' doesn't match what they actually do (it must have been written when they included UKIP in 'others'). DrArsenal 46.208.137.165 (talk) 23:33, 28 October 2014 (UTC)