I am not going to revert the edits of 77.101.6.102 any further, as I have done this too often already (WP:3RR), but the IP has made no attempt to justify its edits (either in edit summary or any of the relevant talk pages). Jmorrison230582 (talk) 16:23, 10 March 2017 (UTC)
Scottish Social Attitudes Survey
Trying to mentally reconcile the continuing majority against independence in polling with the appearance of support for independence in the Social Attitudes survey data, I realised that 'No Parliament' is in fact (an extreme) Unionist position. The splitting of the opposition to independence in this way seems like sleight of hand to create the illusion of a preference for independence (green cells). I'm sure that the referenced report is quoted correctly, but I think the presentation is clearly biased, and should be replaced with a simple two column table pro/anti independence. — Preceding unsigned comment added by 88.106.194.185 (talk) 12:41, 25 August 2018 (UTC)
But that isn't the way the SSAS asks the question. They have always asked the question in the three-option way, as they started polling when devolution was enacted and have continued to ask the question in that way since then. It would be original research to combine the two "unionist" options. Jmorrison230582 (talk) 17:21, 25 August 2018 (UTC)
Sorry, I disagree. Your argument might make sense if the table displayed stats on all five options. Combining all the nationalist options and only two of the three unionist options seems to be a device used in the 'indyref1 to indyref2' report to create the illusion of support for independence, not the SSAS data. In fact the SSAS supports the commonplace evidence that a majority of Scots still favour the Union, as they did in the referendum. Painting the cells green creates the illusion of change when there has been none.
I'm sorry, but you're talking complete rubbish. Page 28 of the 2017 report clearly shows the three options, dating back to 1999. Up until 2014, independence support varied between 23% and 35%, with devolution the most popular opinion in every poll (varying between 44% and 62%). Since 2014, independence has become the most popular option (albeit a minority) in this three-option format, with 46% in 2016 and 45% in 2017. Devolution support fell in those years to 42% and 41% respectively. There has been a clear shift, presumably due a significant number of those who had supported devolution now supporting independence. You're basically accusing Sir John Curtice of being biased in favour of independence. Utterly laughable. Jmorrison230582 (talk) 09:16, 27 August 2018 (UTC)
Apologies: the article text had mis-led me into believing that the cited paper 'Indyref1...' had some relevance, which it doesn't seem to. The partition is indeed from the British Social Attitudes survey. However, the presentation is still misleading: as stated in the text, the options are very dated. While it is the case that support for Independence has risen, as clearly documented in the other polls cited in the article, there has been no transition to a majority for independence, as a casual reading of the table suggests. There has not even been a transition to a plurality for independence, given the other options are all unionist options. Whether or not John Curtice is in favour of independence is irrelevant. This presentation is misleading. One might also question why the BAS is included in a page on opinion polling at all. One could argue that such a dated view should be struck out entirely.
There's been very little Scotland-specific polling so far this year. There was one poll by Survation in early March, which included some independence questions, but that's about it. Jmorrison230582 (talk) 19:33, 24 April 2019 (UTC)
The poll link goes to a twitter account that is not Panelbase, there are no don't knows included this poll should not be included until there is a link to the actual DRG data.Stevenxlead (talk) 13:38, 14 October 2019 (UTC)
Lord Ashcroft poll
He is not a member of the BPC, but this is based on a technicality (he only does work for himself). He is still a reputable pollster, as he publishes full details, and often uses BPC members to conduct his fieldwork. Per the BPC: "This issue of who is regarded as responsible for a poll has arisen on a couple of occasions recently. One of the most active pollsters in recent years has, of course, been Lord Ashcroft, operating under the banner ‘Lord Ashcroft Polls’. Lord Ashcroft Polls does not have the ability to conduct its own fieldwork and thus sub-contracts this part of its polling to a number of companies, many of them BPC members. However, Lord Ashcroft Polls is responsible for the design, weighting and question wording of its polls, and thus it is the body that is ultimately responsible for its results. As it happens, Lord Ashcroft Polls is not a member of the BPC (and as an organisation that does not do work for multiple clients is not eligible to be a member), but as it happens it publishes full details of its polls in much the same way as a BPC member would be expected to do."
Right. Normally we would exclude non-BPC polls, for example there was a poll by "Scotpulse" that put Yes on 59% (!) immediately after the Brexit vote. This has correctly been excluded from Wikipedia because a) Scotpulse is not a BPC member and b) we could not find data tables for that poll. The consequence of this is that we have no idea how they obtained that result. Ashcroft is in a category of his own in that he acts like a BPC member in terms of his transparency, but for commercial reasons he can't be one. Jmorrison230582 (talk) 18:29, 10 August 2019 (UTC)
Yes. This seems to have no source, except from a pro independence blogger. Can’t find it anywhere.... can someone else find it? TTFTAKM (talk) 19:45, 10 February 2020(UTC)
Should there be an additional section for opinions as regards having a referendum?
In the latest poll from YouGov, it's striking that 56% of those polled believed that there should not be a referendum this year, though 44% believed there should be one in the next 5 years. On the post-Brexit referendum polls page, this particular question has its own section (although I'm not sure it's up to date). I think it would be of little surprise to anyone if the main reason people visit this article is to consider the legitimacy of the arguments for a second referendum, which the SNP has done everything in their power to keep in the political scene, in much the same way other political parties have maintained that there's a case to be made for the popularity of a second Brexit referendum. To this end, it seems like it would be adding relevant, topical information, for which there's precedent. Thoughts? — Preceding unsigned comment added by 90.252.156.191 (talk) 00:45, 12 February 2020 (UTC)
Very good point, such a section should be included. That recent yougov poll is not the only poll done that has asked about timing of a referendum so including such polls would make sense.There are few paragraphs mentioning about polling on if there is support for a second referendum on the main article about the proposed referendum, though that information is outdated, and a table of polls about if there should be a referendum and when, on this article, seems sensible. RWB2020 (talk) 01:02, 12 February 2020 (UTC)
Latest Poll data NOT available
Latest ScotGoesPop Panelbase poll - ffs the data link links to a tweet! Take it down until data is published or mark it provisional Stevenxlead (talk) 08:35, 6 June 2020 (UTC)
AS at 7/6/20 it now goes to a Herald Newspaper article quoting James Kelly - can we delete until actual data is available or show as provisional Stevenxlead (talk) 17:26, 7 June 2020 (UTC)
Updated chart, Excel File
Having figured out how to create the chart without using unreliable Macros, I'm now happy to share the source to allow others to update. Here is a link to Dropbox for the Excel file. Basically just paste in the table and save the chart. — Preceding unsigned comment added by RERTwiki (talk • contribs) 11:06, 29 May 2020 (UTC)
Might I suggest that the year be displayed at each tick-mark along the x-axis? In its current form, it isn't immediately obvious that the "peak" for No occurred in 2017 (in fact, it takes a bit of effort to figure out exactly which year corresponds to each tick-mark). Other than that, I think the chart looks brilliant. --Homgran (talk) 19:53, 21 August 2020 (UTC)
re: Post referendum polling
Hi. The lead-in paragraph to the section on post-referendum polling is becoming dated. There is a clear attempt being made to give annual data on the number of Yes and No polls seen, but it is now over 15 months since the end of the last such year and there has been no update. Could some text be inserted to bring the paragraph up-to-date?
On average the 13 polls of how people would vote in a second independence referendum conducted during 2018 found (after excluding those who said they did not know how they would vote) that 45% now back independence, while 55% would prefer to stay in the UK. These figures replicate exactly the outcome of the September 2014 ballot.
The article isn't really helpful, being focused on the interactions with Brexit rather than polling trends. The fact quoted tends to obscure rather than illuminate the data,
since comparable figures would have shown much higher support for independence in the immediate aftermath of the referendum. The scattergraph in the 2014 polling section helpfully showed trends, and could be either extended to include post-referendum polling, or a post referendum polling scattergraph included here. What do you think? — Preceding unsigned comment added by 89.243.190.32 (talk) 14:45, 27 January 2019 (UTC)
Hi. Having check NOR and V, they do not apply. The suggestion is to display the data which you tabulate in graphical form. This is done in several other articles in Wikipedia on political polling, not least the article on the 2014 Scottish Independence referendum. For sure the form in which the data is displayed cannot affect whether it is Original Research, or Verifiable. For the same reason, this isn't an issue of whether sources are used or not. Just to clarify, I'm not volunteering you for work, I would be content to add this myself if I felt it would not provoke an editing war. — Preceding unsigned comment added by 89.243.190.32 (talk) 11:44, 30 January 2019 (UTC)
I thought I would put this in here before attempting to edit it into the article. Feel free to comment or ask questions. The process of getting to this from the table is fairly automated and the format can be tweaked if needed. I've kept the exact colours from the table. I think the chart is a reasonable way to clarify the trends in the data.--RERTwiki (talk) 12:02, 12 March 2019 (UTC)
A chart displaying the margin of all polls since the referendum
(Nice chart) latest YouGov data still not up, but What Scotland Thinks has a breakdown 44 yes 45 no 10 dk & 1 refused (I add dk's and refused together.
Shouldn't we display this as a two curve Pro vs Con graph to keep it neutral though? "Lead for No" sounds very biased as if a lead for No is what we're espousing. Akerbeltz (talk) 10:57, 20 August 2020 (UTC)
That looks a whole lot better, though I'd probably add a trendline of sorts, like the one on the Parliamentary polling? Akerbeltz (talk) 21:19, 20 August 2020 (UTC)
Thanks. I'm off way for the next few days so I'll not add it in to the article just now, although if someone else wants to then go ahead. When I get back I'll provide a link to a Google docs version of the spreadsheet so that anybody can add data in future so it can be kept up to date without being reliant on me. Grinner (talk) 07:32, 22 August 2020 (UTC)
I really like that you've chosen a more neutral visualisation. Now for some comments/suggestions. (1) I think the plot would look neater if we used the same point-type (ideally, a circle) for every data point, rather than making the point-type (triangle, square, circle) conditional on the response. The different responses are already uniquely identifiable by their colour. (2) I think the ten-poll moving average should be centred on the average date of the ten polls that were used to calculate the average -- this isn't the case for the current version of the plot. On a related note, the grey line representing the "Undecided" response doesn't appear to be a moving average at all (it's just a horizontal line); we should be consistent here and use a moving average for all three options. (3) We should consider using semi-transparent colours for the data points and fully-opaque colours for the moving averages -- this would certainly make the plot a bit clearer (and it would solve the problem of two data points lying on top of each other, in the case of a tie). (4) We might want to consider moving the legend to outside the plot-area. (5) The latest Savanta ComRes poll (6-13 Aug 2020) doesn't appear to be represented in the current version of the plot, so that should be included. --Homgran (talk) 12:09, 23 August 2020 (UTC)
UPDATE: Here's an idea of what I'm suggesting. In the attached plot, I've given three smoothing options: (a) a five-poll moving average, (b) a ten-poll moving average, and (c) local regression (loess) smoothing. Happy to hear opinions on which option people think is best, and if it should go into the main article. The idea would be to keep one of the three options in the attached plot (without the title). I could also provide the R code for generating the plot so that anyone can easily update it in future. --Homgran (talk) 15:00, 23 August 2020 (UTC)
I'm ok with anything that's balanced to be honest. Are you proposing putting all three as a merged graphic or just to pick one of them? Akerbeltz (talk) 17:01, 23 August 2020 (UTC)
I'm suggesting that we just pick one of them, because having multiple trend-lines for each response would probably be confusing. Let me know if you have a preference and I'll pick that one. Personally, I'm leaning towards the loess regression curve. Since all of the raw data points are included in the plot, then the real purpose of each curve is to capture the overall trend -- poll-to-poll variability is easily seen from the data points. That said, I'm happy to hear other opinions/suggestions. I agree that the most important thing is for the plot to faithfully represent the data. --Homgran (talk) 17:50, 23 August 2020 (UTC)
I've changed the loess fit to have a narrower span, because I think this does a better job of capturing local trends. What do you think? I see that the other page has omitted the "Undecided" category... what are everyone's thoughts on this? I think having an "Undecided" category is helpful, but I'm happy to go along with the consensus view. I agree that we should be consistent across the two pages. --Homgran (talk) 08:38, 24 August 2020 (UTC)
I agree that they should be the same on both pages and that keeping undecideds is a good idea. I like the latest option (though a bit pixelly) as the 10 poll average has a bit too much lag. Akerbeltz (talk) 09:28, 25 August 2020 (UTC)
Excellent! I've updated the plot on this page (and on Proposed_second_Scottish_independence_referendum). I think the previous versions appeared pixellated because Wikipedia was converting them from pdf. The most recent plot has been generated/uploaded as a png file, so it should be much sharper. Happy to make further modifications to the plot as needed. I'll keep this up-to-date as the main table is updated. --Homgran (talk) 13:56, 25 August 2020 (UTC)
Proposal of removal of "Non-standard question" entries from the "Post-referendum polling" table
Wanted to discuss this before performing an edit. For the purpose of neutral data, the proposal is to remove entries in the condensed table featuring biased or in other ways non-standard questions. These questions can be seen to drastically impact the result of that specific survey, and therefore are not comparable to surveys using the official question. The reader does not gain any useful information from comparing surveys with these included, but can rather be argued to specifically favour the Remain side. Although note this is not a political side-picking, I would be equally against any survey featuring a question that would bias towards Leave side as well.
Considering the graph already excludes these kinds of surveys, it would help to increase the neutrality and accuracy of this table. People generally only see "how much red and how much green", and the amount of people actually following up on - or understands the meaning of - the "Non-standard question" note is arguably small. Foorack (talk) 15:32, 25 September 2020 (UTC)
The problem with this argument is that although there is a "standard" question (i.e. just using the same one as was used in 2014), we don't know what the question for a second referendum would be (or, indeed, if there will be one). I think the current position is reasonable - it shows the poll results, but makes it clear for certain polls why their results may be different. Jmorrison230582 (talk) 16:57, 25 September 2020 (UTC)
Disagree. I think it would be politically biased to exclude polls with different wording when there is no set question. Just because wording was used in 2014 does not mean it would be used in any future referendum. Polls with other wording should be included, though yes it should be noted that its used a different styled question. RWB2020 (talk) 20:03, 25 September 2020 (UTC)
I disagree with this proposal too, we should include all relevant surveys, with the appropriate notes to highlight any discrepancies. That said, I would have some sympathy with moving surveys that use "remain/leave" into a separate section from "yes/no" as this clearly has an impact on the result. Grinner (talk) 07:00, 28 September 2020 (UTC)
I think Grinners suggestion of moving surveys using a "remain/leave" question (instead of a "yes/no" question) to its own table is a good idea. My problem was not with those surveys being in any way "incorrect", just that they are mixed in with surveys using a very different question, as can be seen to impact the result. If this is implemented then nothing would have to be deleted, and it would be clearer to the reader. Foorack (talk) 15:28, 28 September 2020 (UTC)
There's currently a difference between the value for the September survey by JL Partners here and at Proposed second Scottish independence referendum. Looking at Table 4 in the reference, I see the figures to be Yes 51%, No 40%, DK 7%, and would not vote (WNV) 2%. They haven't filtered out a version with WNV removed, so how should this be reported? Currently this page says 51%/42%/7%, whilst Proposed second Scottish independence referendum says 50%/42%/8%. Based on the raw figures (Y/N/DK) of 521/408/68 I think I get 52%/41%/7% with WNV excluded, but this may fall foul of the original research rule. Grinner (talk) 08:57, 3 November 2020 (UTC)
Add the refusers (WNV) to DK, or leave them out completely. The Yes and No figures shouldn't be grossed up unless the pollster has done that themselves. WP:NOR. Jmorrison230582 (talk) 11:39, 3 November 2020 (UTC)
I think I'll leave the WNV out, otherwise that's kind of original research (although a simple addition would probably not actually be a breach). I'll alter to Yes 51%, No 40%, DK 7%. Grinner (talk) 08:01, 4 November 2020 (UTC)
Restore Graph:Chart of historic polling data
User:Akerbeltz: "reverting to the old chart for now, that new curve was just totally confusing. Please take it to talk"
The WikiGraph chart is not "totally confusing", it represents the data just as well as the attached image.
Per Wiki-guidelines SVG should be used whenever possible, especially in this kind of scenario.
The image adds a SPOF on User:Homgran, in the case they go inactive there would be no way to update the image. The WikiGraph version allows the chart to be easily updated by anyone.
For those reasons I Disagree with the revision 994165830. This change should not be implemented unless there is such a special reason which outweighs all the problems above. — Preceding unsigned comment added by Foorack (talk • contribs) 09:06, 15 December 2020 (UTC)
The smoothing of the curve is rather differnt, it looks like a yoyo whereas the other looked more like a trendline. I'm not averse to moving away from the png but is it possible to tweak the smoothing? Right now, we might as well do away with the line and just keep the dots. Akerbeltz (talk) 16:27, 15 December 2020 (UTC)
The WikiGraph chart has terrible smoothing. There's no quick fix. I think on some pages, users insert rolling average numbers rather than individual polling numbers as a means of manual approximation but that really is confusing, so it should only be done if some users maintain the graph. --Gbuvn (talk) 15:07, 16 December 2020 (UTC)
I've reverted it for now. That yoyo graph is more of a sandbox project that something we should play around with in live. Akerbeltz (talk) 22:47, 16 December 2020 (UTC)
Hi folks. ?time for a chart update? I could bung in a chart in the previous format pretty quickly, but better leave as is and update the data. Can someone do that?RERTwiki (talk) 13:41, 28 February 2021 (UTC)
Hi. You'll see I've updated the chart to the latest polls. I've no axe to grind about the format, if people want a new update in the more recent format, that's cool. However, the polls have been volatile and I thought it badly needed an update. RERTwiki (talk) 18:09, 11 March 2021 (UTC)
RERTwiki Thanks for the update. However, your one could be seen to violate NPOV because it is focused around to the "no" vote. I'm sure Homgran wouldn't mind updating their chart when they have a chance, but (as pointed out in the above section) it would be preferable to have a chart that anyone could update. The easiest option would be {{Graph:Chart}}, but the trendlines are inaccurate and it doesn't look as neat. I made a graph recently for Opinion polling for the 2021 Scottish Parliament election which is stored on Commons, but is constructed using the Vega online editor here so that anyone can update it. Would something like that be a good alternative if we need to update graphs quickly? PinkPanda272 (talk/contribs) 18:26, 11 March 2021 (UTC)
I agree with above comment, in that we should plot the three responses (yes, no, DK) rather than editorialise into a lead for one side or the other. But we also need to avoid being dependant on a single editor. Grinner (talk) 08:12, 12 March 2021 (UTC)
No problem with whoever put in the LOESS plot updating to current data. I'm sorry but I don't have (or possibly understand) the tech to update that chart. It is better than mine to the extent that it shows more context but in my opinion it obscures the bottom line, which is the margin. I agree with the thrust of your comments on graphs. I'm happy with a scheme which is updatable by anyone, but it has to be bulletproof and the instructions to update have to be crystal clear and posted somewhere here. If only one person can update it they have to be very assiduous in keeping the thing up to date. Either works, I think, though the former is preferable.
I'm open to suggestions about what the vertical axis title should be, if that is what is bugging people. I'm also happy to put it the other way up, though that obviously has no bearing on NPOV.
However, I have to disagree if there is a suggestion that the display of margin violates NPOV. Physics says this is a signed real number, and it cannot be against NPOV to have the thing one way up or another. Further, the margin is in fact the bottom line, so displaying it gets to the heart of the matter. Anyway enough said. I will only update again if the chart gets substantially out of date. I'll use my format adapted to subsequent comments unless there is consensus here on an easily updated alternative.RERTwiki (talk) 10:29, 12 March 2021 (UTC)
I just think that showing the lead is a bit of odd way of showing things. Generally graphs of voting intention in newspapers (and in article on other subjects) show the two lines. Also, doing it this way misses the don't knows off, and it is worth seeing where there is a shift to/away from DK. I'd probably be happiest going to an imbedded table so that anyone can update it, the one at Welsh_independence#Graphical_summary looks fine to me. Grinner (talk) 11:19, 12 March 2021 (UTC)
I'm not sold on the Welsh Independence graph. The trend lines are incomprehensible, to be kind. Also, that would require the table to be maintained in two places, one for the data table and one for the chart, which is a likely source of (hard to spot) errors. The ideal would be to create a single table which could be used to populate the chart and the table display, but I have no idea how to do that with Wiki tools. RERTwiki (talk) 17:08, 13 March 2021 (UTC)
I am about to update again, changing the title of the vertical axis to remove suggestion of POV. Please can we discuss this rather than simply revert edits. If people prefer the other format that's fine, but it is not appropriate to retain stale data in the page, particularly when elections are upcoming in Scotland. RERTwiki (talk) 17:31, 14 March 2021 (UTC)
I've reverted that edit because it looks like an update to the chart is in the works. I can see a 9-Mar update in Wikimedia, but no sign of that in the actual article. I'll keep an eye out and come back here if it doesn't come through in a few days. RERTwiki (talk) 17:45, 14 March 2021 (UTC)
Charting Non-Standard Question polls from 'Should Scotland be an Independent Country' table
Historically I've left out polls with non-standard questions, and the latest chart update does the same. I've added small print to that effect. I'm quite happy exclude or include the polls, and include or exclude the fine print, in any combination, if people have strong opinions. RERTwiki (talk) 10:58, 22 March 2021 (UTC)
Thanks for updating the table. I agree that it makes complete sense to only plot the polls using the standard question.Grinner (talk) 11:53, 22 March 2021 (UTC)
Given that Savanta ComRes have accepted their latest poll, 4–5 Mar 2021, is not up to standard and has been widely criticised as inaccurate, should it not be excluded?
The methodology of the latest Savanta ComRes poll has changed. This has radically changed the figures. How can this be acceptable? surfingus (talk) 07:04, 12 March 2021 (UTC)
There's been a flood of polling over the last week, which broadly show the same picture. I don't think it's worth getting particularly excited about one poll that deviated somewhat from its usual methodology, presumably because a newspaper wanted to be first with the "polls now say x" story. This poll was weighted by the usual demographics, but was not weighted by the (self-reported) likelihood to vote. Are supporters of one side that much more likely to vote than another? It might have shifted it by a point or two, I suppose, but it doesn't seem plausible that it would make a huge difference. I think it's comparable to polls that exclude 16 and 17 year olds, which are included in the table with an explanatory note. I don't think the difference is so great that it warrants removing the poll entirely or creating a different section for it. Jmorrison230582 (talk) 07:49, 12 March 2021 (UTC)
I do not think it should be excluded. Instead what we have been doing, and should continue to do, is to add a note of explanation about limitations or differences of methodology of how a particular poll was conducted or the question asked, etc.--Literaturegeek | T@1k?16:58, 13 March 2021 (UTC)
Agree with Literaturegeek. These polls are being widely cited and Wikipedia is a prime source. These polls have a clear purpose in the short term. As years progress, they become an invaluable historical research too. Wikipedia should make every effort to avoid the appearance of bias. surfingus (talk) 13:28, 14 March 2021 (UTC)
Apologys if I may have appeared overly earnest. I was concerned with some of the apparent bias reported in some polls. But I respect Wikipedia and the contributors. surfingus (talk) 18:22, 27 March 2021 (UTC)
No Plot of one of Hanbury polls
Hi. Latest chart update excludes the Hanbury poll with 'don't know' as N/A, as it does not include all of the time series required for the graph. If someone can include the relevant data I'll include it on the next update.RERTwiki (talk) 14:12, 29 March 2021 (UTC)
Why a note in October PanelBase poll is needed
I added a note to the October PanelBase poll to say, "The 46–47% difference was purely due to rounding, as the difference was only one respondent. When undecideds were excluded, it showed as 50%–50%." It was reverted by Jmorrison230582 with an edit summary of "So?"
So:
The table ought to reflect the reality. If the survey reports 412 out of 887 as 46% (actually 46.45%) and 413 out of 887 as 47% (actually 46.56%) then 46% and 47% is what should go in those columns of the table, but there should also be a note that tells it exactly as it was, i.e. that the difference was only one person, and that support for independence was 50%, as reported in the news. Also, the figure in the "Lead" column does not come from the source, and it is grossly misleading. It should not be a big red 1%, but the true figure of 0.1%. 95.44.50.222 (talk) 15:07, 16 October 2019 (UTC)
I don't think this is significant. Every polling result is rounded to the nearest 1%. Sometimes this favours one side of an argument, sometimes it favours the other. For example, the actual result of the 2014 referendum was 44.7% yes (to one decimal point), but this is commonly rounded to 45%. Jmorrison230582 (talk) 15:33, 16 October 2019 (UTC)
In the 2014 Scottish independence referendum article it is not rounded; it's given as 44.7% in the lead and 44.70% in the infobox. And at the bottom of the table in this article – the very table that we are discussing – it is given as 44.7%. Anyway, this is a straw man. In the 2014 referendum the difference was 10.60%; if rounding errors cause it to be given as 10%, there's no significant information lost. But here the difference was 413 to 412. Rounding off the difference gives 0%, and 0% is a highly significant figure – it means that half of all Scots polled supported independence, as witness the news headlines above. A short note to state that significant fact, which is both verifiable and relevant, does not take from the article, it adds to it. Is there any policy-based reason that such a note should not be added, or is it just because it was added to your article by an IP? 95.44.50.222 (talk) 16:47, 16 October 2019 (UTC)
"Highly significant"? It's an opinion poll, for goodness' sake. Sometimes rounding to the nearest whole number will favour one side, sometimes it will favour the other side. It's not important. You made a bold edit, I reverted it and now we are discussing it. If there is a consensus to add this information, I wouldn't object. Maybe we should have a column which shows the results excluding don't knows? That's how most polls are reported in the media. Jmorrison230582 (talk) 17:57, 16 October 2019 (UTC)
It's an opinion poll that shows 0% difference between yes and no. In terms of opinion polls, that's as significant as it gets! Sometimes rounding to the nearest whole number makes no significant difference; sometimes it makes a crucial difference. It's not a question of whether it (slightly) favours one side, but whether it gives a true picture of what the poll results actually were. Here, it makes a crucial difference. All I'm asking is (a) that the difference shouldn't be shown as 1% when it is 0%, and (b) to add a short note explaining that the figures aren't what they seem. I get that you personally don't see the need, but I don't get how you think it harms the article. You still haven't answered any of my questions. Why should this article go out of its way to contradict the news stories, just for the sake of an aesthetically pleasing table? And what policy-based reason do you have for opposing the addition of relevant, sourced content? I have no objection to having columns that exclude don't knows, but adding that would still leave us with the question of how a 1% difference becomes a 0% difference when don't knows are excluded. 95.44.50.222 (talk) 22:06, 16 October 2019 (UTC)
@Jmorrison230582: You used the addition of a new poll to sneakily remove the note again. That is disruptive. I have asked you twice: what policy-based reason is there for not including that note, which is both informative and relevant? You have not answered, preferring to just edit-war. Will you answer the question now, please? 95.44.50.222 (talk) 09:20, 24 October 2019 (UTC)
Thank you for at least citing a policy. Of course, I completely disagree. It can't be called bias to add a note explaining the true situation. Are there other polls in this article where the table gives a misleading impression of the actual result? If so, they should be fixed as well. If not, then this one is unique and therefore it is not biased to say that it is unique. On reflection, I think that the note, "The 46–47% difference was purely due to rounding, as the difference was only one respondent. When undecideds were excluded, it showed as 50%–50%", should go into the Notes column as text, not as a footnote. And I repeat, the "lead" should be changed from 1% to 0%, which is the true value. 95.44.50.222 (talk) 21:46, 24 October 2019 (UTC)
External editor here, I think it's pertinent to add the figures without the undecided votes. It's common in elections to report the number of blank votes or absentees separately, and for surveys it's always good practice to show it. Furthermore, as the IP provided, there seems to be several reliable sources using the figure without the undecided votes, so this seems notable enough. But in this case, I wouldn't mention "difference was purely due to rounding, as the difference was only one respondent", which may seem WP:OR, unless the sources explicitly write such an info, otherwise I would provide the exact numbers in parenthesis, eg, "46-47% (412 vs 413 respondents), with 7% (xxx respondents) undecided". Another alternative would be to just keep the 46-47% figure, but add the number of undecided (7% I guess?), and add the margin of error if it's reported by the source, usually it will be of several points, and this is even better to show the reader a quantification of the accuracy. Hope this help, have a nice day --Signimu (talk) 23:25, 24 October 2019 (UTC)
I know this is a dead topic now, but returning to this idea of showing "Don't Knows excluded" - I like the idea of adding two new columns for "Yes" and "No" (plus lead?) with D/K excluded. I would be happy to add it but obviously I wont if there is no concensus and I can see it hasnt ever been done like that. I also note that there is a danger of "overload" if the table is too detailed/too many columns. Or you could just have one which is either "Yes Lead/Deficit" or "No Lead/Deficit" which gives the % of Y or N. Just ideas anyway - Lockdown eh — Preceding unsigned comment added by Dundee1991 (talk • contribs) 16:26, 3 April 2021 (UTC)
Chart Timeline
Is there any appetite for an additional chart to cover a more recent period with greater clarity? All the action is in the last few pixels, and it is actually getting hard to distinguish the updates. RERTwiki (talk) 14:17, 29 March 2021 (UTC)
The problem would be deciding where to start the series: there's no obvious natural date other than the 2014 referendum. I think on balance I'd prefer to keep a single chart. Grinner (talk) 19:13, 29 March 2021 (UTC)
I agree a separate recent chart would be much better. Why not from Jan 2020, that is the period most care about now. I agree that it is really hard to see what is going on now because all the recent bursts of activity are so close together (especially the spate of inquiry polling). Happy to help in any way. I also like the idea of a "last 4/6 months" graph, and that might be better to avoid the "where to start" problem, but also might be more work Dundee1991 (talk) 16:28, 3 April 2021 (UTC)
Herald Poll and Non-Standard Polls
I see that the Herald poll for SIU has been removed. That is not in keeping with previous practice, where all polls are held in the table. Non-Standard polls have been noted and excluded from the charts. By far the best course seems to me to be to hold to previous practice, and put the SIU poll back suitably caveated. While I understand this is an emotional topic for many of us, please can we make this a tiny corner of the interweb which actually does adhere to NPOV?
Removing the SIU poll means, if nothing else, going back through history and throwing away other non-standard question polls for consistency. That seems like an act of vandalism. Having separate tables for non-standard questions would be ugly in my opinion, though it might possibly be interesting to know the polling for 'Should Scotland leave the UK?'. An extra column to include a non-standard flag, so that users can see only standard polls by sorting, is a thought. It might also help in constructing the chart. What do people think?
I note that we have the "Remain/Leave" Polls in a separate section which I think is a good thing because it shows you the data on that all together. Is it biased to put in a little bit of commentary about those polls - e.g. Prof Curtice has talked about that question a lot and some context might be helpful (e.g. (a) what the criticisms of it are, and (b) why the methodology is used in the first place/ what the justifications are for asking the question according to that rubric. Does that violate NPOV? Dundee1991 (talk) 16:38, 3 April 2021 (UTC)
I think that's a good idea, I was meaning to add something based on what the prof wrote recently on the subject. He is pretty much the goto expert on Scottish polling so the best source to use. Grinner (talk) 10:32, 4 April 2021 (UTC)
PinkPanda272 Kindly, forgive my not knowing but rounding error?.. I would think Savanta ComRes should be professional enough to get their maths right. Also I cannot find the poll online, only a reference to it in the article. Does anyone have a link to the survey itself? Dava4444 (talk) 15:29, 25 April 2021 (UTC)
Error is probably the wrong word, if the numbers are (say) 44.8, 47.8 and 7.5, then they add up to 100%, but rounding them to the nearest whole number makes the total 101%. The tables usually turn up in a couple of days, normally figures are based off of newspaper reports until then (so it could be a typo on the part of the Scotsman). PinkPanda272 (talk/contribs) 15:55, 25 April 2021 (UTC)
The latest entry for times/yougov 39:45:10 has a large 'other' vote - yes+no+undecided is 94%. This is because won't say and won't vote are excluded. Last time this poll was entered they were included. I think the table entry should be adjusted to be consistent somehow. My preference would be to keep the totals adding to 100% (to rounding) but any consistent approach would do. I will change this next time I update the chart unless there are objections, to 39:45:16 RERTwiki (talk) 18:03, 28 April 2021 (UTC)
Missed that yesterday, but it's done now. As the last poll, the undecideds now include don't know, won't say & won't vote. RERTwiki (talk) 12:11, 30 April 2021 (UTC)
Savanta/The Scotsman not to be trusted
Please don’t trust polls from the above-mentioned organisations. I imagine, like the Daily Mail, that these sources are using nefarious protocols in their polls. I feel that the numbers of their most recent poll do not reflect the general trends over time. Please consider getting rid of them. Thank you. 95.148.142.9 (talk) 00:41, 30 April 2021 (UTC)
No, the onus is on you to provide some sort of evidence as to why a poll should not be included. Your feelings are irrelevant. Any attempt to delete content without justification will be treated as vandalism. Jmorrison230582 (talk) 03:45, 30 April 2021 (UTC)
I am confused: Savanta/The Scotsman have shown Yes in the lead by 4% and 10% previously and now show No in the lead by between 2 to 7 points, depending on the poll. Are you saying they are biased towards Yes or No? These polling agencies select a diverse group of about 1,000 people and then reinterview them every few months or so, so they will be detecting sways in opinions. So you would need to argue against both the highs for the Yes side as well as the No side being unreliable.--Literaturegeek | T@1k?12:42, 30 April 2021 (UTC)
I think the main issue here is that we are dealing with a sway of over 10%. People don’t change their opinions that significantly. My opinions have no changed like this at all. I think the idea here is to consider the fact that some of the polls are a bit not right. I don’t see much bias here, but I think that their polls should be a lot less changeable, therefore their methodology is probably flawed. 2A01:4C8:1405:6380:2D96:9823:4226:47E5 (talk) 21:51, 30 April 2021 (UTC)
There have been very big changes recently. There was a very successful vaccine roll out compared to the EU so harden views on EU and the Tories will have softened, also lockdown is coming to an end — stress levels are dropping, etc. and then the fallout from the Sturgeon vs. Salmon battle and sleaze, etc. Not sure about the methodology, but you can look into it, like looking at the question being asked was flawed, etc. and if you find anything in their report you can raise it here rather than making suggestions without evidence. The reverse happened with Brexit and then the Tories winning by a landslide followed by a pandemic, these three things undoubtedly pushed, at least temporarily, the Yes side consistently into the lead. Politics and opinion polls are never stable. --Literaturegeek | T@1k?22:11, 30 April 2021 (UTC)
Savanta are definitely broken. Their latest poll is entirely wrong and is considerably different polls that were taken around 3-4 days before. As I had said before, these things don’t swing that hard. Also, the British Polling Council should not be trusted on Scottish matters. I wish for the expulsion of these propaganda merchants from this fine wiki. These polls are not helping anyone, and therefore should not be hear spreading their lies and their mistruths. Please take this seriously, as I am losing the will to live. Thank you. 95.148.142.9 (talk) 21:47, 5 May 2021 (UTC)
1. We will find out in a few days whether it is "entirely wrong" or not. 2. Even if it does prove to be "entirely wrong", shouldn't it be put on record here so that people can look back and say "ah, they were wrong"? Jmorrison230582 (talk) 04:44, 6 May 2021 (UTC)
Good point, I would like for their wrongness to be more obviously. Hopefully some polls will come out that make that clear. 95.148.142.53 (talk) 13:30, 6 May 2021 (UTC)
We need a new separate graph for last 6 months - current one is useless for current trends etc
It has occurred to me that we are going to have weekly (possibly more) polls potentially all over the place on Independence from now on, and I note that the Graph is totally illegible for 2021 and up to the election in May because the graph is over such a long time-span. The graph will continue to be a mess of red and green circles all clumped together in the far-right. I really would support any effort to do a "last 6 months" graph because that one is totally meaningless insofar as recent history is concerned- although I accept its purpose is to represent the historic position which it of course does well (mainly because there used to only be one poll a month!). I am actually happy to do these things by the way but I am aware that there are others who also do them and who have always done them. Dundee1991 (talk) 11:23, 8 April 2021 (UTC)
OK, I agree as above. I'd propose to add a chart of recent history as well as the story since 2014. I might get to that at the weekend, but it shouldn't be hard at all.RERTwiki (talk) 16:47, 8 April 2021 (UTC)
Cool! Sorry my thing wasn't meant to be rude and also I didnt want to start doing anything because I know there are already people who deal with all these things. Thanks! look forward to seeing itDundee1991 (talk) 13:10, 10 April 2021 (UTC)
The graphs look good, but the side by side doesn't really work on my computer screen due to the width, they go off centre. On my phone they are on top of each other. I get that this is obviously specific to my screen size but I wonder if it might be better to either shrink them a bit or leave them one on top of each other rather than side by side? Grinner (talk) 14:43, 12 April 2021 (UTC)
Just had a thought, perhaps we could use a Switcher like there is a election polling:
Agreed, that's better, I'll try that soon. FWIW as is it works OK now on my laptop and phone, so I did check! RERTwiki (talk) 12:53, 13 April 2021 (UTC)
I've tweaked the chart a little, breaking the axis to reduce the wasted space between undecided and yes/no, reducing 'recent' to last 20 weeks, and centering the titles. Hopefully not controversial. RERTwiki (talk) 13:42, 18 May 2021 (UTC)
Data on Polling Organisation results
I don't think this is suitable for inclusion in the article, but it does have bearing on some of the conversations here.
This is a table which shows the average difference between a polling orgs polls and the 10-poll average I compute.
So, for example, Mori is on average overstating 'Yes' by 4% and understating 'No' by 0.7%, relative to the average of 10 polls, including Mori, at the time of its research. Thus overstating the margin for 'Yes' by 4.7%.
It is hard for the predominant polls to be far from the average (including many of their own). However, interesting data.RERTwiki (talk) 16:27, 28 May 2021 (UTC)
Polling organisation
Number of Polls
Yes Difference
No Difference
Undecided Difference
Panelbase
41
1.0%
1.5%
–2.0%
Survation
40
–0.3%
–0.6%
1.2%
YouGov
32
–1.9%
–0.4%
1.2%
Savanta
12
–1.0%
0.9%
0.3%
Mori
11
4.0%
–0.7%
–3.2%
BMG
6
1.1%
–2.1%
1.4%
Opinium
3
0.6%
–1.4%
–1.4%
Other
9
0.2%
–0.8%
1.0%
Ipsos MORI Poll April 2021
I added the recent Ipsos MORI poll to the leave/remain section, with note about the question. I think this a valid poll and not just a subsample given how it has been reported, see [1], however I thought I ought to mention this as obviously pulling single figures out from the tables can potentially be dodgy, however I think MORI are using correctly weighted groups for each nation here. Grinner (talk) 10:27, 19 April 2021 (UTC)
Couple of comments: firstly, the question used here is a hybrid of 'leave/remain in the UK' and 'Become an independent country'. Second, the weighted sample is 718, not 1889. Which is right? The figures in the table certainly add up to 718. Third, you are very right, picking sections out of Polls is getting dangerous. Usually it is far from obvious that each individual geographic region is individually demographically balanced. It also sets a difficult precedent - are we obliged to include all 'Scotland Subset' results from now on? Is anyone going to go back and look for similar data in historic polls? They might not even have made this page. Obviously, picking individual eye-catching results risk violating NPOV.
I wouldn't myself include this datapoint. If it is to be included, I think it fits better in the main table, where we document polls with an eclectic mix of questions. The results of the poll, which are bang in-line with the main table, and out of line in both position and trend with the leave/remain table, support that view. My 2 Cents, RERTwiki (talk) 13:21, 29 April 2021 (UTC)
I think you're right, I'm going to remove it. It is perilously close to original research, and it's not like we are short of polling on independence at the moment. Grinner (talk) 11:16, 30 April 2021 (UTC)
I was amused to look at the Scottish section of the recent YouGov UK poll and see that apparently 2% of Scots would vote for Plaid Cymru... RERTwiki (talk) 14:57, 5 June 2021 (UTC)
Proposing update to intro paragraph
I'd like to propose that we drastically cut back the paragraphs immediately starting the section 'Post Referendum polling'.
The first paragraph is a wordy attempt to summarize three years of polling data. The next two paras are slightly breathless descriptions of the twists and turns of the last couple of years. 2018 and most of 2019 seem to have been memory-holed. The last paragraph is getting out of date, as No has established a small lead in the course of the election campaign.
The section as a whole is not really written from a NPOV, but perhaps more importantly, it's just in the way. I think people come here for the latest polling news, and flick past this text as quickly as they can.
Here is a suggested alternative, better words/suggestions gratefully received -
Support for Yes and No to Independence has varied widely since the referendum. Based on averages of polls, No has been the most popular response over most of the period, but with margins generally lower than the referendum result. Recently, between March 2020 and April 2021, Yes was generally most popular.
The charts below display result for polls which ask the same question as the referendum. The table lists these polls and those which ask very similar questions.
Aye that seems fair. I agree that the current wording is a bit "this happened and then this happened and then this happened". Grinner (talk) 09:29, 27 May 2021 (UTC)
That's reasonable. I still think we can improve the intro as follows. I'll give a proposed amended version, and explain later the motivation for the changes.
Since the referendum in September 2014, opinion polls have asked how people would vote in a hypothetical second referendum. These polls have been carried out since six weeks after the referendum.[1]
Sentiment in favour of independence was high immediately following the referendum, with the majority of polls published in the next six months showing a plurality in favour of 'Yes'. Over the next two years support for 'No' rose and support for 'Yes' fell. A Survation poll carried out in the two days prior to the UK general election on the 8th June 2017 showed the largest margin in favour of 'No' of 56% to 36%. By September 2017 five consecutive polls had shown a margin for 'No' greater than the 2017 referendum result. To date this has proven to be a peak for 'No'. Support for 'No' declined slowly until the end of 2018, and more quickly from the spring of 2019, at the height of parliamentary gridlock over Brexit. In October 2020 as the second wave of the COVID-19 pandemic began, 'Yes' reached its peak to date when an Ipsos MORI poll for STV News showed a margin in favour of 'Yes' of 52% to 39%.[2]. Since that point sentiment swung steadily back in favour of 'No', coinciding with the period when Brexit trade deal was finally completed and the COVID-19 vaccination program was rolled out.
Throughout the period, events have caused poll results to swing sharply. In the few days after the Brexit referendum in 2016, polls registered support for 'Yes' which was out of line with polling immediately before that referendum, and which disappeared a few weeks later. In July 2019, immediately following Boris Johnson's election as leader of the Conservative Party, a poll by Lord Ashcroft showed a the first majority for 'Yes' in over 2 years. Again this poll was an isolated event, contradicted by polls immediately before and after, though it occurred at a time when there was a clear trend towards 'Yes'. Polls conducted in early March 2021, following testimony by Alex Salmond and Nicola Sturgeon at a Holyrood parliamentary inquiry, showed narrow leads for 'No' which were similarly transient.[3]
Professor John Curtice said in mid-2019 that there had recently been a swing towards 'Yes', and that this was concentrated among people who had voted to "Remain" in the 2016 Brexit referendum.[4]. Since then support for 'No' has increased.
OK, what did I do. I've replaced the first paragraph: it was boring and incomplete, not covering 2018 and 2019. I've taken on board the suggestion that we should narrate the background to polling, including references to major political and external events which seem relevant. I've tried to balance the text, which was previously written from a 'Yes' perspective. Thus I included reference to the rise in 'No' after the referendum. Again for balance, I've included references to the peak support for both 'Yes' and 'No'. I've changed the data referred to in those polls to include undecideds, to avoid apparent inconsistency with the table. In the second paragraph, I've re-cast the previous references in the context of special events: of themselves they don't justify inclusion in a discussion of hundreds of polls over many years, but the reaction to events is an interesting feature of the series. I've balanced the comment from Curtice referenced at the end with the simple fact of the trends since: I think that is more appropriate than regarding it as out of date, since views on Brexit do seem to have shaped politics.
It's been over a week since I posted the above. Taking the lack of response to mean there are no objections, I've put the above into the article.RERTwiki (talk) 15:12, 5 June 2021 (UTC)
Hi - I trawled through the Stack Data tables, couldn't see anything corresponding to the 48/48/4 result noted in the table. Can anybody tell me where to find that data?
Am I correct to assume this is a Non-Standard question? The only data I can see is people being asked to rate support for independence between 0-9 (though nothing that looks like 48/48/4).RERTwiki (talk) 15:18, 5 June 2021 (UTC)
I agree, this survey is not designed to evaluate how people will vote in an indyref2/independence For example, 0-1 rating is judged strongly against independence and 9-10 is strongly for independence, with a rating of 2-8 considered “middle”. So basically this survey is only weighted to determine how the hardest of hardcore Yes supporters feel on various political questions and how the hardest of hardcore No supporters feel on the same political questions. The political questions are not directly related to indyref2. If the 0-10 rating split people between 5 and 6 it certainly would have value to this page as a non-standard question survey, but it isn’t. So really I would advocate quite strongly removing it entirely from this page, despite me being very inclusionist when it comes to indyref2/independence surveys.--Literaturegeek | T@1k?20:46, 5 June 2021 (UTC)
The tables at Stack Data's own website show the question as 'how would you vote if there was a referendum tomorrow with the question "Should Scotland be an independent country?', but I can't see the "don't knows" in there. What Scotland Thinks is a reliable source so I suggest we just link to their figures (which includes DK) as a reference. Grinner (talk) 06:59, 6 June 2021 (UTC)
Okay, thanks for that. I can see that on page 4 of the document. I have struck my above comment as it was clearly ill-informed. I cannot see the don’t knows either. I am happy with your suggestion re. linking to What Scotland Thinks.--Literaturegeek | T@1k?15:03, 6 June 2021 (UTC)
Thanks folks, should have thought to look for their website. The tables originally linked were a subset. Odd to be no explicit statement, but the weighted base for the yes/no question excluding don't know etc is 95% of the overall weighted base, so I guess that's close enough. Wish they didn't make it such hard work... RERTwiki (talk) 14:31, 7 June 2021 (UTC)
Margin Chart in Switcher
I've added a margin chart to switcher, showing the margin excluding undecideds. I can't think of another fair way to compare current polling with the referendum result, and since people have to click-through to it, I don't think it should be controversial. famous last words...
RERTwiki (talk) 12:00, 10 September 2021 (UTC)
Scotland-in-Union Sep 21
Hi - this latest poll is attributed to Panelbase/Scotland-in-Union by The Scotsman here
Also I couldn't follow the link to the source for the poll data, message included 'Access Denied'.
If it is Survation can the link get fixed, or else can we cite Panelbase and the Scotsman?
The SIU poll was done by Survation [2], published yesterday (9th September). I think the reference to "Panelbase" is a mistake by The Scotsman. All of SIU's previous polls have been done by Survation, and the fieldwork numbers match up with the Survation poll. Jmorrison230582 (talk) 11:29, 10 September 2021 (UTC)
OK, cool. What about the link? It doesn't work for me. Here is the message I get when I click on Survation/ScotlandInUnion in the table:
This XML file does not appear to have any style information associated with it. The document tree is shown below.
<Error>
AccessDenied
<BucketName>wordpresssurvationcdn</BucketName>
<RequestId>tx000000000000027af2ad7-00613b49a6-167fa22c-fra1a</RequestId>
<HostId>167fa22c-fra1a-fra1</HostId>
</Error>
Folks - the latest Yougov poll seems to have been treated differently from the previous Yougov poll.
In that case it seems like refused/won't vote were included with undecideds, in this case they have been left out.
Most other polls get close to Yes+No+Undecided ~= 100%. This one is quite a bit out.
I can change the numbers at the next chart update is no-one objects. Undecideds will obviously change, and it will have a slight effect on the average margin.RERTwiki (talk) 17:56, 29 November 2021 (UTC)
I hid this poll because we don't have full figures for it, only the headline figures that exclude don't knows (I think that was 49Y-51N). ComRes haven't added it to their poll archive yet. Jmorrison230582 (talk) 14:14, 10 March 2022 (UTC)
Ipsos MORI poll February 2022
Hi, the question used in the latest Ipsos Mori poll appears nonstandard, asking those questioned “If a referendum were held in Scotland on its constitutional future, would you personally prefer Scotland to vote for or against leaving the UK and becoming an independent country?” instead of the usual “should Scotland become an independent country?”. I was wondering whether this poll should be moved to a separate table to signal the difference in questioning, or whether it is appropriate that it is left in the normal table? GamerPerson12345678 (talk) 22:29, 22 March 2022 (UTC)
Polls with non-standard questions are all excluded from the charts. I will take no action to update the charts at present.RERTwiki (talk) 16:12, 27 March 2022 (UTC)
Proposed text update
I think the third paragraph of the text prior to the main table is becoming dated: I don't believe the Holyrood Enquiry is significant or memorable enough compared to the other
events. I'd offer this as an alternative, shorter concluding paragraph:
"Throughout the period, significant events have caused polls to swing sharply. For example the result of the Brexit referendum and the election of Boris Johnson as the UK Prime minister caused sentiment to temporarily move in favour of 'Yes', only to soon move back towards the status quo ante."
Every poll recorded on this page has a sample size of over 1000, this has been established as the minimum size of poll to give accurate data with a low level of Margin of error. I notice a poll by Techne Uk with a sample size of 500 has been included, this significantly increase the margin of error and therefore can not be directly compared with all the other polls.
Does anyone else share my view that only full polls ie 1000+ should be included? Soosider3 (talk) 14:32, 4 July 2022 (UTC)
No. There's no requirement to have a sample size of 1,000 to make a representative poll. 500-sample polls are unusual at a national level in the UK (or Scotland), but they are fairly common for constituency polling (e.g. both polls for the 2022 Wakefield by-election had samples of just over 500). The table as it stands takes the correct approach - displaying the figures, but noting that this poll was a little unusual in the way it was compiled. Jmorrison230582 (talk) 15:57, 4 July 2022 (UTC)
This is a subset and not a full poll.
All polling on a National level requires 1000+ to be considered reliable that is why every Polling Company uses this as its bench mark. Polling across Wikipedia at this level always shows only 1000+, which is why the Wakefield by election poll would not be included in any page showing national figures.
Given the wide acceptance of this benchmark across the Polling Industry and across wikipedia I think it is really for you to put forward an argument as to why this practice should now cease and include subsets and sample sizes significantly less than 1000
1,000 (or more) is the typical sample used for a national opinion poll, but it isn't a necessary condition. There are a few polls cited in the 2014 indyref polling article that had samples of well under 1,000 (two by Angus Reid, and one by ICM in the final week). A smaller sample just means that the margin for error is a bit higher; it doesn't necessarily make it unrepresentative. Jmorrison230582 (talk) 13:58, 6 July 2022 (UTC)
Thank you for the response, however you do not seem to be accepting the core issue here.
I recognise that you now have accepted that the Techne poll is not a typical poll, I would remind you again that what you wish to add to this page is not actually a poll but a subset. As stated to you previously BPC has developed the practice that 1000+ is now the 'norm' across the industry, that fact by itself should be enough for you to accept that this is not the correct place for this subset and would ask that you withdraw it from this page, or at least this section of the page.
I would respectfully suggest that what happened several years ago, on other pages is hardly the strongest of defences especially as they are so few, the page you referred to has well over 100 polls recorded and only 3 or 4 that are significantly below the 1000 response standard, not one of them was a subset. The industry has moved on from 2011, learned from its mistakes and operates to a higher standard than in the past and so should we and so should Wikipedia Soosider3 (talk) 14:21, 6 July 2022 (UTC)
Yougov/These Islands Poll 29-31 March 2022
Have just noticed that this poll uses a mix of the standard and a non standard question, 519 Standard and 510 non standard. I would intend to make a small change to the Notes to reflect this. Soosider3 (talk) 13:42, 8 July 2022 (UTC)
This poll appears in 2 places on this page, we should be looking at polls only being recorded in one place, as it asks the non standard question Leave/Remain perhaps that is the best section for it. Soosider3 (talk) 14:17, 8 July 2022 (UTC)
Actually the more I look the more it appears that this submission is very odd. Its the first poll ( i know of) that splits the question on Independence with 2 different Responses, Leave/Remain and Standard Question
It is recorded on Leave/Remain section showing 35% Leave and 50% Remain, yet shows a sample size of 1029 - when only 510 respondents were asked that question. In the main section it shows figures of 39% to 44% again claiming 1029 respondents while figures relate to only to the 519 respondents that were given the standard question, both are inaccurate and misleading
To my mind it throws up some issues that we should try to address
1 What do we do with any other poll that splits questions in this manner do we
a) count them as 2 small polls, in which case they would both fall to the sample size not being large enough, industry standard requires 1000+ this is well established practice and can be evidenced across many polling pages on Wikipedia
b) count it as one poll with note about mixed questions, but this demonstrates the problem that with small polls and variable questions do we actually introduce too many variables and lower the standard and accuracy of the data overall. We effectively have three results from this poll Leave/Remain, standard Question and an amalgamation of both.
c) Is it reasonable to amalgamate what is effectively 2 different sub polls in this manner, I tend to think it is not acceptable practice
d) Should we introduce a subsection of the report specifically for such hybrid polls
Personally I think we should introduce a new section for Hybrid Polling on Independence, it categorises the poll correctly and prevents it from polluting the other polls with less reliable and not directly comparable data.
I think both entries should be removed as both are inaccurate and misleading and would propose doing so in the next few days Soosider3 (talk) 15:51, 8 July 2022 (UTC)
Given that both entries have now been corrected to the actual sample size 519 and 510, I would intend to remove both entries as sample size are too small and introduce a significant increase in margin of error. Industry standard across the UK is that 1000+ are required to produce reliable and comparable data Soosider3 (talk) 16:42, 8 July 2022 (UTC)
I would agree with removing this poll from both tables, but not because of sample size. I can't see how the polling firm could have constructed two samples that were both representative of the population within the same set of fieldwork. I think they only did that as a test, to find out whether question wording does make a difference in the results obtained. Jmorrison230582 (talk) 19:37, 8 July 2022 (UTC)
Thanks for the response, surely this has to be the case for every poll that is substantially below the 1000 samples? That if a poll has less than this it is difficult to make it representative of the whole population. Suspect that splitting the question would more likely have been done at the behest of the client especially as polling companies are very well aware that different questions will get different answers, as does the ordering of the questions and the format of questionnaire. Soosider3 (talk) 10:47, 9 July 2022 (UTC)
Hyperlinks to data rather than Newspaper articles
I do appreciate that often headline data is available before the data tables have been published, however I notice that many polls hyperlinks are to newspapers articles (which often contain other material and opinions) rather than to available data tables.
The proposal is to change this and amend the hyperlink to data on polling companies website ( where available) I believe this is better practice and allows users to delve deeper into the data if they so wish
If acceptable I would propose changing these hyperlinks starting with most recent and working backwards
I would welcome views from others Soosider3 (talk) 11:10, 9 July 2022 (UTC)
Yes, that makes sense - the tables are more reliable and in-depth than news articles that just summarise some of the findings. The news articles are often posted in the first instance because they are available first. Jmorrison230582 (talk) 14:59, 9 July 2022 (UTC)
Yougov Times poll 16 to 20 April 2021
While updating hyperlinks to data source rather than newspaper article I came across this poll, it is odd in a few ways.
Firstly I am not sure if it is actually a Yougov poll as data seems to have been lifted from The Times webpage, at least the bit before it blanks us non subscribers out, the reference to Yougov in the article is ambiguous and may actually be referring to previous poll. It maybe coincidence but Savanta ComRes have a poll with exactly teh same fieldworl dates. I rather suspect that wires have got crossed.
Secondly There is no trace of this poll on the Yougov website
Thirdly there is no trace of this poll on whatscotlandthinks.org or on ballotboxscotland.com, both reliable sources of Scottish political polling
Fourthly the Undecided 16% have been made up by adding Undecided, will not vote and refused, this is statistically illiterate, but more importantly it hints that these are raw numbers and have not been weighted for likelihood to vote.
I suspect the author of entry has quickly added numbers taken from newspaper and just not got back to it.
This rather highlights the dangers of adding data straight from newspaper article but also that link to article is no substitute for data source.
This entry should be removed as clearly it is not reliable and no corroboration for data is available. Soosider3 (talk) 12:39, 10 July 2022 (UTC)
"Secondly There is no trace of this poll on the Yougov website". You couldn't have looked very hard. I found this link after searching the YouGov site for two minutes. I'm concerned that your reaction was to doubt the existence of the poll and suggest its deletion, rather than assume good faith in the other contributors. Jmorrison230582 (talk) 13:00, 10 July 2022 (UTC)
Thanks for that, I have been looking at so many data sources today I must have gotten a little addled, I will use your link to up date the hyperlink to the data rather than to the newspaper article.
One other small alteration is I will reduce the 16% Undecided to the actual 10% in the data set and put a small note to explain that 4% would not vote and 2% refused. Soosider3 (talk) 21:52, 10 July 2022 (UTC)
Re: New para on Demographics
Folks - I don't understand the placement and reasoning behind the new section on Demographics.
For one thing, the inserted data is from 2014 and before. It is far from clear why we would want to insert it now.
Second, it is just out of place where it is. This is a page on Opinion Polling, specifically eschewing polling before the 2014 referendum. Every one of the polls we tabulate is likely to have data attached on demographics. Why are we picking just this (historic) data?
Third, this kind of data is inviting the conclusion that 'demography is destiny', and that because younger people favour 'Yes', then yes will have a majority when the old people die. But that's just wrong: forever and always, older people have been conservative biassed, and younger people become so as they age. The same could easily be true of yes/no voting. It is a spurious talking point to try to elide to a coming 'Yes' majority because of demographics.
At a minimum, a space could be found for this data elsewhere on this page, or even in the page on the 2014 referendum, for the historic data. It's positioning right now is too prominent and unwarranted. My personal vote would be to leave this out unless we are willing to do this properly, and open a section to describe demographics in all the polls tabulated, which frankly I think is a bridge too far.
Agreed, it has more to do with the 2014 referendum than what is in this article and the different attitudes amongst various demographic groups is described in the results section of that article. Jmorrison230582 (talk) 16:05, 27 June 2022 (UTC)
The section was also taken from the subsample of polling and therefore grossly invalid.
--Signimu (talk) 23:25, 24 October 2019 (UTC)
