Talk:Opinion polling for the 2017 United Kingdom general election
This is the talk page for discussing improvements to the Opinion polling for the 2017 United Kingdom general election article. This is not a forum for general discussion of the article's subject. |
Article policies
|
Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL |
Archives: 1, 2, 3, 4 |
![]() | Elections and Referendums Unassessed | ||||||
|
![]() | Politics of the United Kingdom Unassessed Low‑importance | |||||||||
|
Justification
I'd suggest that this article would be justified if any such polls actually existed. Martinevans123 (talk) 13:04, 12 May 2015 (UTC)
- There's one there. More will come, on a regular basis. DrArsenal (talk) 15:27, 12 May 2015 (UTC)
- And if it was less than 4 years and 360 days until the next one... Martinevans123 (talk) 15:33, 12 May 2015 (UTC)
- It does make you wonder with all the effort on the previous article to what appears to be regarded in the press as a flawed system we need to detail every poll for the next five years in such detail if at all? MilborneOne (talk) 15:47, 12 May 2015 (UTC)
- The only poll that turned out to be accurate was the exit poll published on the night of the election itself. Don't you think there's a lesson there? Martinevans123 (talk) 16:12, 12 May 2015 (UTC)
- There is indeed a lesson here, but perhaps not one relevant to Wikipedia editors. Wikipedia follows reliable sources. Reliable sources go on printing polling results. These go on attracting extensive interest from political parties, commentators, academics, journalists and the public at large. Ergo, it makes sense for Wikipedia to capture and record the results of such polls, and as long as there are editors happy to do that task, I don't see how it's bothering anyone else!
- What we should make of such polls is another matter, but one that falls foul of WP:OR and WP:NOTAFORUM. Bondegezou (talk) 16:37, 12 May 2015 (UTC)
- It's true that the polls for the last General Election were not as good at predicting the outcome as they had hoped. The polls had Conservatives about 3% too low and Labour about 3% too high compared with the result - in each case, at the outer edge of the margin of error. The BPC is holding an inquiry, and we can expect changes in methodology that will attempt to reduce the gap. But that doesn't make polling worthless. The polls did correctly predict that LibDem support had dropped dramatically from 5 years ago, that UKIP support had increased, that SNP support had increased to such an extent that they would win a large majority of seats in Scotland and that Green support would increase - the polling record of the last 5 years even shows us when that happened. ...and I came to the page to see whether there is any evidence _yet_ for my hunch that the abysmal showing of the LibDems will lead to a further fall in support for them. There is more to be looked at than a prediction of who will win the next election. DrArsenal (talk) 16:52, 12 May 2015 (UTC)
Northern Ireland
Won't there be any polls for Northern Ireland, or why isn't there a column for them? --Maxl (talk) 16:55, 17 May 2015 (UTC)
- Polling in Northern Ireland is infrequent. Something can be added when we actually get our first NI poll. Bondegezou (talk) 19:08, 17 May 2015 (UTC)
SNP Polling Figures
As some pollsters regard the SNP and Plaid Cymru as one bloc, would it make sense to have either a column for Plaid Cymru (so joint figures can have merged columns) or a joint "Nationalist" column (as opposed to an SNP column)? This would avoid having to have a note saying "<pollster> combines support for the Scottish National Party and Plaid Cymru in its headline voting intention figures." Clyde1998 (talk · contribs)
- Agreed. Bondegezou (talk) 17:13, 27 June 2015 (UTC)
- I too agree. (Except for Scottish or Welsh polls, of course.) But is it possible to aggregate these figures in all cases? --Wavehunter (talk) 20:34, 27 June 2015 (UTC)
- The problem with a Plaid column is that in GB wide polls their share is likely to remain static in the *% to 2% range due to the relatively small Welsh portion of most sample sizes, making the usefulness of including such a column questionable. It should be possible to obtain a pure Ipsos Mori SNP figure from examining the tables rather than the headline figures - I don't know if that would be a more preferable option? Returningofficer (talk) 10:44, 7 July 2015 (UTC)
- That would probably be better. Clyde1998 (talk · contribs)
- I've edited the table so that the Ipsos Mori figures now link to the data tables rather than the headline figures, giving a specific SNP figure. For future reference the relevant data table (at present) seems to be Table 4. Returningofficer (talk) 09:46, 5 August 2015 (UTC)
Ipsos Mori sample sizes
Re: the edit by 92.17.196.117 on 26 July, should we be only listing the headline figure sub-sample when reporting voting intention? I think for other polling companies we're reporting the full base sample sizes every time. Returningofficer (talk) 09:51, 3 August 2015 (UTC)
- Yes, I think we usually list the full base sample size. Bondegezou (talk) 11:19, 3 August 2015 (UTC)
Graphical summary and the SNP
Shouldn't the SNP be on the graph? They are polling of the same order as the Greens and sometimes ahead of them. Saxmund (talk) 19:19, 15 August 2015 (UTC)
- I just came here with the same question! The graph was uploaded by User:Pi - I'll prompt him/her!--Super Nintendo Chalmers (talk) 18:18, 16 August 2015 (UTC)
- The reason I didn't put the SNP in is because they are nationally on 5% of the vote or so, which isn't all that indicative of how well they do, when in reality they're polling 40-50% but in Scotland only. I'm open to change it though if most people think that's best. Pi (Talk to me! ) 06:17, 28 August 2015 (UTC)
- Yeah, I can see the point - though it's not like the vote of the other parties is spread evenly across the country either. While the SNP are consistently polling at >2-3% I think they're worth having there. --Super Nintendo Chalmers (talk) 08:48, 28 August 2015 (UTC)
- Agree with including the SNP on the graph. Understand that we're unlikely to see much movement in their polling figures, however I feel that as we're including them in the GB-wide tables it makes sense to display them as well. It's also what seems to happen on the Canadian opinion polling pages with the Bloc Québécois. Returningofficer (talk) 14:28, 29 August 2015 (UTC)
- While the SNP seem to be polling as high as, or even slightly higher than, the Greens, it seems appropriate to include them (permanently), even if there is unlikely to be much movement in GB-wide terms. The argument for adding UKIP and latterly the Greens in the previous page was that they were polling as well as, or better than, the LibDems. Thanks to Pi for updating the graph. Saxmund (talk) 11:45, 27 September 2015 (UTC)
- Saxmund has, I think, been over-simplifying the history of criteria for inclusion in the previous page - certainly polling as well as, or higher than, a party already listed was ONE argument, but I don't think it was ever taken to be conclusive: it was only when other arguments were relevant too, that there was inclusion.
- As it is, the only problem I see in including SNP in this graph is that some pollsters don't report SNP figures separately from a combined SNP/Plaid Cymru figure, but if we can get reliable data without O.R., I don't see any problem. DrArsenal (talk) 22:02, 13 October 2015 (UTC)
- While the SNP seem to be polling as high as, or even slightly higher than, the Greens, it seems appropriate to include them (permanently), even if there is unlikely to be much movement in GB-wide terms. The argument for adding UKIP and latterly the Greens in the previous page was that they were polling as well as, or better than, the LibDems. Thanks to Pi for updating the graph. Saxmund (talk) 11:45, 27 September 2015 (UTC)
- Agree with including the SNP on the graph. Understand that we're unlikely to see much movement in their polling figures, however I feel that as we're including them in the GB-wide tables it makes sense to display them as well. It's also what seems to happen on the Canadian opinion polling pages with the Bloc Québécois. Returningofficer (talk) 14:28, 29 August 2015 (UTC)
Events in opinion poll tables
I'm wondering whether we should have specific criteria for including important events in the opinion poll tables? It just strikes me that we now have Jeremy Corbyn's election listed separately in the GB, Scotland & Wales tables, but the summer budget and Tim Farron's election appear only in the GB table. Conversely, there's no mention of Kezia Dugdale's election as Labour's leader in Scotland. Should all important 'GB-wide' events be reproduced in the Scotland & Wales sections, or is it worth only listing specific Scotland & Wales related events? Or indeed, given the rarity of Scottish and Welsh polls at this stage in the cycle, is it worth including events at all in those sections? Returningofficer (talk) 11:03, 15 September 2015 (UTC)
- I am strongly against most "events" in these tables. They become far too subjective, in terms of what to include and what not to include. Opinion poll tables for other countries usually don't have any events at all: e.g. Next Australian federal election. Bondegezou (talk) 13:37, 15 September 2015 (UTC)
- I would remove Corbyn's election from the Scottish and Welsh tables, and add Kezia Dugdale's election to the Scottish table. You're quite right, though, that the inclusion criteria are fairly subjective; we could try to draw up a list of what to include and what to exclude, but it could easily be overtaken by events. --Wavehunter (talk) 18:15, 15 September 2015 (UTC)
- I looked at equivalent polling tables for Australia, Poland, Italy, India, Germany, Greece, Canada and Spain. None of them showed any events in their polling tables (other than election results). I say we should do the same. Anything else is editorialising/WP:OR. If there are marked shifts in polling, that reliable sources describe as being causative, then those can be described in accompanying text. But let's stop adding in "events" into polling tables at the whims of individual editors. Bondegezou (talk) 21:48, 15 September 2015 (UTC)
- Agree. Requiring events to have been identified as signalling a shift in support would keep the focus on polling and prevent the tables from simply degenerating into timelines of political events. Returningofficer (talk) 12:13, 17 September 2015 (UTC)
- I looked at equivalent polling tables for Australia, Poland, Italy, India, Germany, Greece, Canada and Spain. None of them showed any events in their polling tables (other than election results). I say we should do the same. Anything else is editorialising/WP:OR. If there are marked shifts in polling, that reliable sources describe as being causative, then those can be described in accompanying text. But let's stop adding in "events" into polling tables at the whims of individual editors. Bondegezou (talk) 21:48, 15 September 2015 (UTC)
- Would we then have to define what constitutes 'a marked shift in polling'? I think it will be fairly easy to find text in broadsheet newspapers saying this policy or that gaffe led to a fall or rise in support for party X. --Wavehunter (talk) 19:56, 17 September 2015 (UTC)
- I don't think it's as common as you think, but if we can find reliable source discussion of polling shifts, I'm all for including that in some accompanying text separate to the table. We should be trying to define as little as possible and let ourselves be guided by what reliable sources say. Bondegezou (talk) 08:29, 18 September 2015 (UTC)
- I'm convinced that tables should NEVER have more "events" than polls, and as such would say there is currently a problem with Scotland and Wales.
- I can see the merit of the suggestion that events should have been identified by RS as causing a shift in support. I don't think there is any problem in this only being possible retrospectively, because until that data is available, any event that is sufficiently important will be fresh enough in the memory/news sites for those interested to know of it well enough anyway.
- However I see a problem. Often apparently similar events don't cause a shift in the polls - for example most budgets don't, but the "omnishambles" budget was widely reported as having done so. Only including the events that are reported in RS as causing a shift in support itself gives a misleading impression as to how likely it is that events will cause a shift (indeed, I suspect that in fact "reliable sources" when written by journalists rather then political scientists tend themselves to over-report likelihood, reporting things to have been caused by a single "obvious" event at about the right time, when in fact the causation was more complex or not/mis-understood). DrArsenal (talk) 20:58, 18 September 2015 (UTC)
- The solution then, I suggest, is to remove events from the polling tables and to have a section of text discussing, based on good quality RS, shifts in the polling over time. That is, the focus should not be on the events: the focus should be on the polling, and if polling shifts, we can discuss any issues (be they specific events or longer term issues, like a recovering economy) that are thought to be related. Bondegezou (talk) 12:43, 19 September 2015 (UTC)
- Alternatively we can continue to include a small number of major political events that may cause cause a shift in polling (or, in the case of elections, signify that one is taking place). Then we are not editorialising. The effect on the GB table where historically there have been a lot of polls is to punctuate the table and help the reader to look for trends. It is a shame that there are more events than polls in the Scottish and Welsh tables but I don't see it as too much of a problem. Saxmund (talk) 11:39, 27 September 2015 (UTC)
- Fine, Saxmund, you don't see it as too much of a problem. But this page is here to compile a list of the headline results of opinion polls. The Scotland/Wales sections just don't meet that spec. At the moment they are a list of GB political events punctuated with the odd opinion poll. And most of those events are repeats of events mentioned at other places on the same page. I would advocate changing Scot/Wales to only list events not listed in the GB wide table, unless it is an event in Scot/Wales with wider impact, such as a parliamentary by-election. DrArsenal (talk) 23:02, 27 September 2015 (UTC)
- I'd be happy with that as well, there are so few non-GB polls at the moment that it is looking a little silly. Saxmund (talk) 10:50, 28 September 2015 (UTC)
- Fine, Saxmund, you don't see it as too much of a problem. But this page is here to compile a list of the headline results of opinion polls. The Scotland/Wales sections just don't meet that spec. At the moment they are a list of GB political events punctuated with the odd opinion poll. And most of those events are repeats of events mentioned at other places on the same page. I would advocate changing Scot/Wales to only list events not listed in the GB wide table, unless it is an event in Scot/Wales with wider impact, such as a parliamentary by-election. DrArsenal (talk) 23:02, 27 September 2015 (UTC)
- Alternatively we can continue to include a small number of major political events that may cause cause a shift in polling (or, in the case of elections, signify that one is taking place). Then we are not editorialising. The effect on the GB table where historically there have been a lot of polls is to punctuate the table and help the reader to look for trends. It is a shame that there are more events than polls in the Scottish and Welsh tables but I don't see it as too much of a problem. Saxmund (talk) 11:39, 27 September 2015 (UTC)
- The solution then, I suggest, is to remove events from the polling tables and to have a section of text discussing, based on good quality RS, shifts in the polling over time. That is, the focus should not be on the events: the focus should be on the polling, and if polling shifts, we can discuss any issues (be they specific events or longer term issues, like a recovering economy) that are thought to be related. Bondegezou (talk) 12:43, 19 September 2015 (UTC)
- I don't think it's as common as you think, but if we can find reliable source discussion of polling shifts, I'm all for including that in some accompanying text separate to the table. We should be trying to define as little as possible and let ourselves be guided by what reliable sources say. Bondegezou (talk) 08:29, 18 September 2015 (UTC)
- I would remove Corbyn's election from the Scottish and Welsh tables, and add Kezia Dugdale's election to the Scottish table. You're quite right, though, that the inclusion criteria are fairly subjective; we could try to draw up a list of what to include and what to exclude, but it could easily be overtaken by events. --Wavehunter (talk) 18:15, 15 September 2015 (UTC)
Moving average
How many results are being averaged? How can you calculate a moving average from the first result? 217.38.103.51 (talk) 08:10, 25 September 2015 (UTC)
- I would be interested in knowing the algorithm being used to average polls, just to be transparent really. Saxmund (talk) 11:47, 27 September 2015 (UTC)
- Pi? --Dweller (talk) 12:19, 30 September 2015 (UTC)
- Hi, at the moment it is the last 10 polls, I was intending to make it last 30 days but at the beginning there just weren't enough polls being made. Pi (Talk to me! ) 16:17, 9 October 2015 (UTC)
- Thanks for following up, Pi. Agree with Saxmund that this should be spelled out in the article. 217.38.99.110 (talk) 18:26, 16 December 2015 (UTC)
- Yes definitely, there have been comments below from people saying it is incorrect so would be good to have the methodology available so it can be checked. Absolutelypuremilk (talk) 19:49, 16 December 2015 (UTC)
- @Pi if it is ok with you I will add into the graph caption: includes polls up to 20th November. Moving average calculated from last ten polls (I think thats the last poll!) Absolutelypuremilk (talk) 14:29, 22 December 2015 (UTC)
- Yes definitely, there have been comments below from people saying it is incorrect so would be good to have the methodology available so it can be checked. Absolutelypuremilk (talk) 19:49, 16 December 2015 (UTC)
- Thanks for following up, Pi. Agree with Saxmund that this should be spelled out in the article. 217.38.99.110 (talk) 18:26, 16 December 2015 (UTC)
- Hi, at the moment it is the last 10 polls, I was intending to make it last 30 days but at the beginning there just weren't enough polls being made. Pi (Talk to me! ) 16:17, 9 October 2015 (UTC)
- Pi? --Dweller (talk) 12:19, 30 September 2015 (UTC)
- Thanks Absolutelypuremilk. That would help. DrArsenal (talk) 20:28, 22 December 2015 (UTC)
Time to remove the greens from graph?
No longer beating the lib-dems. 78.144.221.190 (talk) 18:31, 12 October 2015 (UTC)
- The decision was never as simple as just whether Greens were ahead of LDs or not. What we should be doing is following Wikipedia policy and following reliable sources. So, 78.144.221.190, can you point to reliable sources that do give all of Lab/Con/UKIP/LD/SNP but NOT Greens current polling? DrArsenal (talk) 23:03, 12 October 2015 (UTC)
- Consensus, as far as I remember it, was to include every party individually listed by the pollster in their detailed table, i.e. an inclusive approach. If the Greens are still recorded by the pollsters, and I believe they are, we should report them. Bondegezou (talk) 17:17, 13 October 2015 (UTC)
- The decision was never as simple as just whether Greens were ahead of LDs or not. What we should be doing is following Wikipedia policy and following reliable sources. So, 78.144.221.190, can you point to reliable sources that do give all of Lab/Con/UKIP/LD/SNP but NOT Greens current polling? DrArsenal (talk) 23:03, 12 October 2015 (UTC)
- Consensus changes. When it was UKIP the standard was emphatically 'not until they pass the lib-dems', obviously this is out of date. So when're the BNP & respect being added to the graph? 78.144.26.65 (talk) 21:35, 16 October 2015 (UTC)
- Greens are still consistently polling above the margin of error, unlike BNP or Respect which, when included in tabs at all, only very rarely get anything other than an asterisk. Additionally, as much as Green support may have fallen, the figure itself remains significant in terms of highlighting a decline in support since the election. If at the start of the 2016 polling table the party's support has drifted into consistently insignificant territory, there may be a case for rethinking the party's inclusion in the new table, but for now I see no reason for removing them. Returningofficer (talk) 21:46, 16 October 2015 (UTC)
- I was refering to Bondegezous recollection of consensus, but seriously wiki should reflect reliable sources: pollsters give headline figures only for the big 4, all newspapers & the BBC follow, and ofcom list the same 4 as significant. I can't see any justification for the current state of the graph. The argument for including the Greens was made ad nauseum on the basis that they were level with the lib-dems, which (IMO self evidently) is an argument without merit. wiki should either follow RS, or have a precise unbiased qualifying standard. The current system is A: there used to be 3 clear parties, B:when a 4th clear party emerged in RS people argued against its inclusion (raising the bar as needed) until it overtook the smallest of the old 3, C:the smallest of the old 3 shrank to briefly the level of a small 5th party, which spawned an incessant argument for the 5th parties inclusion on an obviously fallacial 'fairness' point. 6th small local party grew to the level of 5th small party, and was added on the basis of 'doesn't make sense, but what's the harm?'. Changing the listed parties at year end makes sense, to stop them keep flashing in and out....but only if there's clarity on how a party qualifies. 78.144.26.65 (talk) 02:56, 17 October 2015 (UTC)
- Additionally, I just checked & IDRC, some polls lump GB nationalists in together, meaning the 'SNP' figure listed in the table is often just plain wrong, being both not given & contradicted by the source. I don't know if the current grapher is aware of which polls to ignore, but either way the table is wrong. 78.144.26.65 (talk) 03:04, 17 October 2015 (UTC)
- It's not the case at all that polling companies only give the 'big 4' headline figures: see for example YouGov Ipsos MORI and ComRes. The issue of SNP polling figures has already been discussed on this page - see discussions above. Where a headline figure gives a combined nationalist score, the practice has been to extract the SNP figure from the detailed polling tables. On the more general point of inclusion, I think we need to be careful not to drift too much into questions of fairness or whether party x/y deserves inclusion - it's ultimately more a question of whether the inclusion of a party adds or subtracts from the article. If I had to suggest criteria for inclusions, I would argue that a party should be included in any given year's table if the following criteria are met: 1) the party is regarded as significant by substantial elements of the media/political world (for 2015, inclusion in debates would meet this criteria) and 2) the party is polling at a meaningful level for at least a portion of the year in question. Returningofficer(talk) 07:50, 17 October 2015 (UTC)
78.144.26.65 is going back to arguments I thought long since settled. More recent consensus embraced the idea that we report whatever the pollsters report, reflecting practice on non-UK polling articles. Take a look at polling articles for forthcoming elections in Germany and Spain, or the recent election in Greece. Bondegezou (talk) 09:06, 17 October 2015 (UTC)
- 78.144.26.65 "pollsters give headline figures only for the big 4, all newspapers & the BBC follow". Which pollsters, when? Which articles in newspapers, please? Can you provide links? Which BBC stories - are they on the website? Returningofficer indicates in the opposite direction with real evidence, while so far all we have had from 78.144.26.65 is unverifiable generalities which cannot possibly amount to evidence for a change in what Reliable Sources do. DrArsenal (talk) 06:40, 20 October 2015 (UTC)
Survation - UK-wide poll
I've just added a new poll from Survation to the GB table. While it looks very much like a standard poll (and appears to have been treated as such by general media sources), upon close inspection of the polling tables it unusually appears to include a small number of respondents from Northern Ireland. As a stop-gap measure I've added a footnote to make it clear that it's not a standard GB-wide poll, but I wonder what the general consensus was on this? Should it be moved to a new UK-wide section? Returningofficer (talk) 12:34, 13 November 2015 (UTC)
- Presumably this was done because the main aim of the poll was the EU referendum. For now, I think your approach is sensible of including a footnote. If more like this crop up, then we can re-think. Bondegezou (talk) 13:02, 13 November 2015 (UTC)
problem with graph
the graph doesn't seem to include the latest poll figures and those it does include don't seem to be plotted properly. — Preceding unsigned comment added by 90.222.73.109 (talk) 20:23, 23 November 2015 (UTC)
Missed one or two?
ComRes Daily Mail Political Poll November 2015. 217.38.99.110 (talk) 16:40, 16 December 2015 (UTC)
- Also on the Opinion Bee list but missing here: 22-Sep-15 Survation / Huffington Post. 217.38.99.110 (talk) 16:59, 16 December 2015 (UTC)
- And just while I'm in OCD mode, the fieldwork for the September Ipsos MORI / Evening Standard poll was done from the 19th to the 22nd, not the 19th to the 23rd. 217.38.99.110 (talk) 18:21, 16 December 2015 (UTC)
- I've added the Survation and ComRes polls. Saxmund (talk) 21:27, 20 December 2015 (UTC)
- And just while I'm in OCD mode, the fieldwork for the September Ipsos MORI / Evening Standard poll was done from the 19th to the 22nd, not the 19th to the 23rd. 217.38.99.110 (talk) 18:21, 16 December 2015 (UTC)
- Not just missed. Several polls have previously been included which are now missing! I recall there being survation polls in which Labour trailed by double-figure percentages, but these have all now been removed. Difficult to identify the specific point at which this was done because a flurry of edits hides the change, but could do with someone taking the time to go over and fix this so everything's includedWikiditm (talk) 15:57, 17 December 2015 (UTC)
- Additionally, the results from some of the polls are now wrong. Table now needs significant review.Wikiditm (talk) 16:33, 17 December 2015 (UTC)
- I just came here to look something up and noticed this. The most recent poll (Ipsos-MORI 12-14 Dec), for example, should be Con 37 Lab 33, but has been edited to Con 34 Lab 36. Why the anon editor in question decided to make this huge slew of incorrect adjustments changes I don't know, but whatever the cause, they were (thankfully) so blatant about it that said changes jumped out at once. Loganberry (Talk) 17:05, 17 December 2015 (UTC)
- I've tried to correct the errors introduced today, but am not sure I've got them all. --Wavehunter (talk) 18:15, 17 December 2015 (UTC)
How often is the graph updated?
I'm just asking, since the updates seem to be rather infrequent at the moment. Also, I am curious as to how the graph is created? What software is being used? FloppyCatfish96 (talk) 13:40, 31 December 2015 (UTC)FloppyCatfish96
- It is done by Pi (see the "Moving Average" section above. Perhaps post on their talk page? Absolutelypuremilk (talk) 18:04, 31 December 2015 (UTC)
- can the graph not be made to auto-update when data are added to the table?
- The graph is done on Excel/Numbers (for Mac). I have made a new version as I don't think Pi is active on Wikipedia any more. I am working in Numbers and I have not been able to extend the moving average line to the most recent polls (it works off the last five polls and appears to place the line at the start of the five rather than at the end) so if anyone knows how to to do it properly then it would be much appreciated. Absolutelypuremilk (talk) 23:35, 16 January 2016 (UTC)
- thanks Absolutelypuremilk: it is really good to have the graph updated. Can I ask, though, that it returns to a ten-poll basis? The five-poll basis gives too much noise. Thanks. DrArsenal (talk) 21:29, 17 January 2016 (UTC)
- Unfortunately as I wrote above, I can only get the line to extend to the first of the points it is taking the moving average from. If I make it ten points, then the line stops significantly before the points end and it looks very strange. When the next poll comes in then I will upload a ten-poll graph (and then revert) so you can see what I mean. Absolutelypuremilk (talk) 21:39, 17 January 2016 (UTC)
could the average of the last ten polls be included as an additional column in the table? — Preceding unsigned comment added by 2.126.247.113 (talk) 12:12, 21 March 2016 (UTC)
Panelbase 8-14 Jan
I have added the Panelbase poll which has been reported by ukpollingreport, Opinion Bee and Britain Elects. However I can't find figures for the SNP or Greens, or work out how to derive them from the tables. Can anyone help? Saxmund (talk) 22:07, 24 January 2016 (UTC)
- I am not sure why the Westminster voting intentions have been left off. Perhaps email them and ask? Absolutelypuremilk (talk) 23:12, 24 January 2016 (UTC)
- I did email them and had a swift reply from Ivor Knox (MD), the tables have now been updated so I have been able to put the figures in the usual format. Saxmund (talk) 21:14, 27 January 2016 (UTC)
- Thanks Saxmund. Now I'm glad I didn't get round to emailing them! DrArsenal (talk) 23:11, 27 January 2016 (UTC)
England Polls
I've noticed that the England polls listed in the sub-national polling section is including sub-samples (from GB-wide polls) from ComRes. There are about 1,500 respondents from England in these polls, but they're not weighted to England's demographic - they're weighted to the demographic of Great Britain as a whole. I'm wondering: should this sub-sample data should be included in that list? The Survation poll listed is an England-only poll, so it's weighted for England's demographic. If yes, a Survation poll (derived from a Scotland-only poll) has a Scotland sub-sample where over 1,000 people were asked (weighted down to count for 180 respondents, however) - should this be included in the list? Clyde1998 (talk · contribs)
- I've got no firm opinion either way, but it's worth noting that this approach was taken to derive England-only results for the 2015 page with no objections. The argument is presumably that England makes up such a large proportion of the electorate that any error is very small, and probably within the margin of error. Saxmund (talk) 21:10, 5 February 2016 (UTC)
New Polling Graph Proposal
Hello everyone,
I've created a new proposal for the polling graph, just to clean it up and make it more clear and informative. I can't upload the image directly as I haven't signed up here yet, so I've uploaded it online (link below).
I'll now outline how the graph is composed and why I have chosen the methods I have.
The graph is made up of a moving average of the aggregate percentages of each polling company (rather than each poll) that release a new poll at least once per month. Those companies are YouGov, Ipsos Mori, ComRes, ICM, Survation, Panelbase and Opinium. I have chosen to use polling companies rather than individual polls because each company posts new polls at different rates, meaning under the current method, a pollster that may generally overstate a party's support, but posts a new poll every day, could skew the moving average of individual polls. However, under the new system, a polling company which may only post a new poll every two weeks will be given equal weighting to one that posts a daily poll.
As you can also see, there is a semi-transparent line behind each of the aggregate lines. This is a long-term trendline, which allows one to see whether a party is performing above or below their long-term trendline at any point.
I have also used two sets of gridlines, one for each axis, with the x-axis gridlines highlighting a period of 3 months. Again, this allows one to quickly access more information without cluttering the graph.
This is only a suggestion, but feel free to use this current version and get back in touch for the new graph as I update it.
Thanks. — Preceding unsigned comment added by 2A01:388:385:151:0:0:1:1BB (talk) 15:59, 4 February 2016 (UTC)
- Thanks for your suggestion, I agree it would be helpful to see a long-term trend line but I think it is also helpful to see the individual polling points and I think it looks a bit cluttered with both of these. This is probably going to be down to personal preference as to whether we stick with the current one (possibly with gridlines on the x-axis and/or with a white background), move to your suggestion or to have both of them, but this may cause confusion, especially if they are not updated at the same time. Comments from other editors would be much appreciated. Absolutelypuremilk (talk) 16:48, 4 February 2016 (UTC)
- Thanks User:2A01:388:385:151:0:0:1:1BB. I can see multiple advantages to your version. I like the fact that the line is thinner, and I think there is merit in the idea of compensating for differing frequency of polls by different pollsters. I also think the trend lines may be valuable, but we must have clarity about how they are calculated, and we need some way for them to take account of trends changing direction. If you look at
you will see that the Conservatives clearly had declining support 2010-May 2013, but that decline stopped in May 2013, while Labour had increasing support May-Dec 2010 and declining support after mid 2012. In either case, no meaningful trend line could be drawn for the entire period from May 2010 to May 2015.
- Between User:2A01:388:385:151:0:0:1:1BB's, the current graph and the graph at File:UK_opinion_polling_2010-2015.png, it is the 2010-15 graph which I prefer - it has the individual polling points, but they are feint/small enough to not obscure the graph in the way that the current graph allows them to. If a way can be found to include a longer-term trend line, in the way proposed, that works, that would be good. DrArsenal (talk) 16:12, 5 February 2016 (UTC)
- Thanks User:2A01:388:385:151:0:0:1:1BB. I can see multiple advantages to your version. I like the fact that the line is thinner, and I think there is merit in the idea of compensating for differing frequency of polls by different pollsters. I also think the trend lines may be valuable, but we must have clarity about how they are calculated, and we need some way for them to take account of trends changing direction. If you look at
- Thanks for your feedback Absolutelypuremilk and DrArsenal. I've made an amendment to the graph, based on your suggestions. I've removed the 'long-term' trendline and replaced it with the individual polling results. I think this works a lot better than my previous one. Link is below:
- UK 2020 Poll Graph (New) — Preceding unsigned comment added by 5.151.170.197 (talk) 18:52, 5 February 2016 (UTC)
- My feeling is this is an improvement on the graph currently presented, but is too jagged. It needs to be taking rolling means over a longer period for each pollster, I think. DrArsenal (talk) 00:01, 7 February 2016 (UTC)
- I've taken the feedback on board, and I have to say I agree it is a little too jagged. I have modified it so it is now a 2-polls-per-pollster rolling average (for the pollsters who conduct a new poll at least once a month). For the earlier parts, where a pollster hadn't yet conducted 2, I inputed the GE results as a placeholder where relevant, preventing the graph from getting really jagged at the beginning.
- UK 2020 Polls (2 Poll Rolling) — Preceding unsigned comment added by 2A01:388:385:151:0:0:1:125 (talk) 13:44, 7 February 2016 (UTC)
Undent - I reckon that looks a lot better, muchly thanks to the thinner line. Is it just me, or does the current graph have some proper weirdnesses? eg, around september it looks as though the UKIP line is temporarily travelling back in time. 80.42.16.12 (talk) 01:04, 1 March 2016 (UTC)
- On my graph (the one that is currently being used) I have increased the number of polls that the moving average uses (to remove some of the weirdness) and decreased the thickness of the line. I will upload the new version when the next poll comes out. Absolutelypuremilk (talk) 09:11, 1 March 2016 (UTC)
- I've updated my graph so it's now up to date; the consensus seems to be that this should be used, so feel free to use it, and I will update it frequently.
- 2020 Poll Graph 5.151.170.197 (talk) 10:35, 22 March 2016 (UTC)
Wales only chart
Is there a similar Wales only chart being created? I see we have the data, but not the chart? A similar one for Scotland only would also be appreciated. Thanks. Llywelyn2000 (talk) 07:37, 8 March 2016 (UTC)
- There are only four polls for Wales so far and two for Scotland. When there are a few more then I will make a chart. Absolutelypuremilk (talk) 10:12, 8 March 2016 (UTC)
- There would be far more point in having one at National_Assembly_for_Wales_election,_2016#Regional_Vote_.28AMS.29, based on the polling there (to parallel the one that already exists for polling for the Scottish Parliament election). DrArsenal (talk) 23:13, 9 March 2016 (UTC)
- @Absolutelypuremilk: Many thanks for the offer!
- @DrArsenal: Both need doing, but as this is the UK GE, I stuck to that. Llywelyn2000 (talk) 10:32, 11 March 2016 (UTC)
ICM Poll
Today ICM released a poll with Tory 36.4, Labour 35.6, which rounds to Con 36, Lab 36 as shown. However the Tory lead was 0.8%, which rounds to 1%. Should we have for the lead either: Tory lead of 1%, Tory lead of 0% or Tied (clearly we can have explanatory notes for all of these scenarios). I put it in as tied originally to avoid confusion but now think that Tory lead of 1% is best way to go, as a note can be put in to explain why this is the case even though the headline figure is that the two parties are tied. Absolutelypuremilk (talk) 20:39, 14 March 2016 (UTC)
Budget and EU referendum dates
The Budget and date that the PM announced the EU referendum have been included in here as notable dates, I don't think that they are notable in relation to Westminster polling and don't think they should be included in the table, what do others think? Absolutelypuremilk (talk) 16:15, 16 March 2016 (UTC)
- Take out both: I thought we'd already agreed to remove things like that! I'll do it now. Bondegezou (talk) 11:09, 17 March 2016 (UTC)
- I removed both. An IP put both back, arguing we've had budgets/autumn statements before. I've re-removed the referendum announcement (as it's not a budget/autumn statement).
- I can't remember what the previous consensus decision was on budgets...? Opinion poll tables for other countries never include any sort of equivalent. I see no reason to include them: it's editorialising/OR. I'd take everything except actual election results out of the tables, but what do others think? Bondegezou (talk) 14:06, 17 March 2016 (UTC)
- I'd personally include the budget, but keep the referendum date announcement out of the general election polling. I'd include: Elections/Important By-Elections, Changes in Party Leadership, Budgets/Other important statements, TV Debates and Other events that have a substantial impact on party support. The Events should purely say what the event was - such as "United Kingdom Budget 2016" rather than including any additional information. Clyde1998 (talk · contribs)
- I would agree with including the budget, considering the very noticeable effect it has had on the opinion polls. Tidus mi2 (talk) 12:50, 21 March 2016 (UTC)
- How do you know that it has had a "very noticeable effect ... on the opinion polls." User:Tidus mi2? Certainly recent opinion polls have a dramatically lower Conservative lead than those prior to early March. But is that really caused by the budget? The 11–13 Mar ICM/The Guardian poll was the first to show something different from what had been happening before, and that was before the budget. So, perhaps the cause was something else? It might be it really was the budget (in large part), but we don't know this, and to include the budget row on the grounds that the budget has had a "very noticeable effect ... on the opinion polls" appears to me to be commentary and editorialising, not encyclopedic reporting. By the very nature of decisions about what to include, there is an implicit "this is more important than other events" just in the inclusion of the event. DrArsenal (talk) 22:10, 28 March 2016 (UTC)
- I would agree with including the budget, considering the very noticeable effect it has had on the opinion polls. Tidus mi2 (talk) 12:50, 21 March 2016 (UTC)
- I'd personally include the budget, but keep the referendum date announcement out of the general election polling. I'd include: Elections/Important By-Elections, Changes in Party Leadership, Budgets/Other important statements, TV Debates and Other events that have a substantial impact on party support. The Events should purely say what the event was - such as "United Kingdom Budget 2016" rather than including any additional information. Clyde1998 (talk · contribs)
19–22 Mar Ipsos MORI
Not sure what the rationale is for adding Plaid to SNP here when the pollster reports the figures separately. SNP 5% and Others 2% would be more useful. 217.38.116.87 (talk) 15:19, 24 March 2016 (UTC)
- Someone tidied up the table without fixing this, so I did it myself. The Plaid vote is counted under Other rather than under SNP in the earlier Ipsos MORI results. 217.38.116.87 (talk) 22:10, 24 March 2016 (UTC)
Navigation template
I think it would be useful to have a navigation template for this and the other General Election opinion poll articles. --Dweller (talk) Become old fashioned! 13:05, 30 March 2016 (UTC)
New graph
Mirashhh added the following graph to replace the current one, but it was reverted. As far as I can tell, the main differences are that it uses vertical gridlines, and that it only runs to shortly after the final poll. What do other editors think about these two differences? I made the current graph and I think it works better, but am open to change if others disagree.
Absolutelypuremilk (talk) 14:31, 2 April 2016 (UTC)
- I like your graph and prefer it over the current one. What's your method? Would it be possible to have a bit more smoothing? Bondegezou (talk) 14:47, 2 April 2016 (UTC)
- To clarify, I made the graph which is currently shown in the article, using Numbers (Mac's version of Excel). Mirashhh made the one I have added to the talk page above (Presumably using Excel or Numbers). For either graph it is possible to smooth it, but for my graph it is slightly more complicated. If you want the running average of ten polls, then Mirashhh's graph appears to start plotting the running average before there have been ten polls, i.e. when there have been four polls it takes the average of those four polls, but for my graph it waits until there have been ten polls before it starts to calculate the running average, and then plots the running average starting at the first of those ten polls, meaning there is a gap at the end. You can't really tell in the graph currently shown, but if you reduce the dates shown as I have done below, you can really see it.
Absolutelypuremilk (talk) 14:59, 2 April 2016 (UTC)
I like both graphs about equally, so I think Absolutelypuremilk's should remain (should he want to continue to keep updating it!) as a recognition of the work he has put in so far. As a side note - thank you! I find this to be one of the most useful UK politics resources available, and I regularly check it. Atshal (talk) 08:44, 5 April 2016 (UTC)