Talk:Opinion polling for the 2021 Dutch general election

Peil polls

I've noticed that many consecutive polls from Peil.nl have identical numbers for all 13 parties that are currently in Parliament. For example 13 and 20 May 2018, 29 Apr and 6 May and 15 Apr and 22 Apr 2018. The probability of getting identical numbers for all 13 parties in two polls would be practically zero. Even if there had been no changes in support over the week, sampling errors would lead to slightly different results in a new poll. I'm therefore suggesting that many of Peil's polls are thus "recycled" for a second week. They should be removed from the table and graph, unless there is evidence and sources that in fact Peil did on many occasions take two fresh polls and get exactly the same results. AV85647 (talk) 23:06, 29 May 2018 (UTC)

The methodology of the Peil poll is different is because it's an online non-probability panel in which respondents, upon returning, are shown the last response they submitted. The published numbers for new polls each week are based on the shifts versus previous weeks rather than the raw levels, which should explain the relative stability of the published seat estimates. (Source is here.) I usually try to avoid voicing particular methodological concerns, but I'll admit that I'm suspicious of the unusually high values for right-wing parties that de Hond's polls have historically shown (PVV in the past, FvD now) – but the overall consistency of levels each week makes sense given the stated methodology. Mélencron (talk) 00:44, 30 May 2018 (UTC)

Color

The page is a bit confusing as it is. PVV is of a lighter shade than VVD on the graphic, but it's the darkest one of the pair in the column. Quite misleading on a quick read. Which one is right?--Aréat (talk) 11:35, 29 July 2018 (UTC)

Recall line

Tijskeuk's edit was clever. The list is long enough to be adorned with a recall line at the top. Up until now, and perhaps 2022, the party seats will be those gained in March 2017. The one who wants to get a glimpse of the state of Dutch party politics today only needs these two lines: last week and last election. Rather than going by the 100, and soon 200-line-long list. Kahlores (talk) 01:10, 31 July 2018 (UTC)

I disagree – the list is meant to be strictly antichronological and I don't see the obvious necessity of listing it twice within the same table. Furthermore, it's not generally used on other polling articles; off the top of my head, the only current instance is for the 2018 Swedish parliamentary election. (I'd also note that the rate at which this list will expand is likely to be considerably slower with only two regular pollsters – of which one seems to have gone silent in recent weeks, despite not usually taking a July/August polling hiatus.) Mélencron (talk) 03:48, 31 July 2018 (UTC)

Bad logic, but a curious discovery indeed

So we should add one line for every week now, even if there hasn't been any release?

With all due respects, you misinterpreted the pdf. The polling report doesn't state that four polls have been taken. It says "there have been no shifts in political preferences during the past four weeks". Perhaps they've accumulated data throughout these weeks, but we cannot speculate, as they never release such details. And even if they said they did, it wouldn't be a reason to create four lines. This is a list of published opinion polls.

You suggest to look at the graph. First of all, the graphs are not the results. Merely a representation of them.

So I looked at the graph's source code. There's a javascript tag containing weekly data. Indeed, each week is filled, even those without a release. However, does that mean that they've got a secret average for every week?

Let's compare with last summer, where there was a long break. The last poll released in July was 8 July. The first poll in August was 19 August. So there were five Sundays without a poll release. Yet the graph does have data for these dates. So by the same token as yours, we should add these weeks. Here's how it looks in the code:

[' 8-7', 12, 10, 11, 30, 15, 17, 6, 13, 3, 7, 5, 15, 6, 0],
['15-7', 12, 10, 11, 30, 15, 17, 6, 13, 3, 7, 5, 15, 6, 0],
['22-7', 12, 10, 11, 30, 15, 17, 6, 13, 3, 7, 5, 15, 6, 0],
['29-7', 12, 11, 11, 29, 15, 17, 6, 13, 3, 7, 5, 15, 6, 0],
[' 5-8', 12, 11, 11, 29, 15, 17, 6, 13, 3, 7, 5, 15, 6, 0],
['12-8', 12, 11, 12, 29, 15, 16, 6, 13, 3, 7, 5, 15, 6, 0],
['19-8', 12, 11, 12, 29, 15, 16, 6, 13, 3, 7, 5, 15, 6, 0],

The only difference after the summer break were VVD (30 to 29), GL (17 to 16), SP (11 to 12) and PvdA (10 to 11). According to these lines, VVD and PvdA swap one seat on 29-7, and GL and SP one seat on 12-8. We cannot speculate on what basis these dates stand for. Maybe their computer took the results from 8-7 and 19-8 and decided to swap seats in the middle and at the end of the break (we'd need a first swap at the start, but there were just two). Maybe these are actual unreleased results from Peil.nl. But if it were so, then we cannot list them. These are not newsworthy pieces of information. These are crumbs for opinion polling maniacs. I wouldn't mind if you were using them for the .png graph. But keep in mind that the data may be biased, for instance by a lack of a significant number of people surveyed.

Kahlores (talk) 20:12, 13 January 2019 (UTC)

P.S.: While checking to see what happened in Summer 2017, I noticed a mistake in the code for 30 July 2017:

['30-07', 16, 11, 11, 33, 21, 13, 5, 20, 3, 5, 3, 6, 3, 0],

50PLUS has one seat less (3 instead of 4), VVD has one seat more (33 instead of 32) than shown in the pdf. If there are such discrepancies, how credible is the graph?

poll numbers add up 10 150% of votes!

does anyone know why?

Seats, not vote shares, it says it right there... Mélencron (talk) 17:06, 17 April 2019 (UTC)

Peilingwijzer

Is it maybe a good idea to add the Peilingwijzer (https://peilingwijzer.tomlouwerse.nl/) in a separate table every time it is published? Thomasfowl (talk) 14:42, 23 May 2019 (UTC)

I think the boycott of them by Maurice de Hond makes it less meaningful... and I'm not sure about including the raw data for every single date covered by it. Mélencron (talk) 17:18, 23 May 2019 (UTC)

Graph

The user who made the graphic, Mélencron, is no longer active on Wikipedia. Is there others users here who know how to make a new one? It's badly in need of an update.--Aréat (talk) 17:03, 13 December 2019 (UTC)

Here's a guide if you want to give it a try: https://docs.google.com/document/d/1khuyz_552Gt26uIMGhAIV8Q8prwlNubzfnmIQ05mLOI/edit#heading=h.15uu2btlcbhf --Gbuvn (talk) 09:30, 14 December 2019 (UTC)

Thanks. That seem a bit too complicated for me, though.--Aréat (talk) 19:29, 14 December 2019 (UTC)

I'm no expert either, but I've actually tried to do it a few weeks ago, unfortunately Inkscape gets very slow for me with that amount of polling data. I might try again with a simpler graph if we can't find anyone else for this. --Gbuvn (talk) 23:53, 14 December 2019 (UTC)

Nevermind, I just updated the graph. --Gbuvn (talk) 21:54, 15 December 2019 (UTC)

Thank you Gbuvn! --Aréat (talk) 07:28, 16 December 2019 (UTC)

PvdT should renoved from list

They have no any seats now. Shadow4dark (talk) 00:57, 24 October 2020 (UTC)

Agreed, they haven't been more significant than other parties which have been denoted by footnotes. --Gbuvn (talk) 11:04, 1 November 2020 (UTC)

Maybe a separate section for the smaller parties? Some have participated in several elections and there is very little documentation on their results over time. Pallieter (talk) 21:56, 25 January 2021 (UTC)

Petition to remove vote share section

This section is badly outdated. It would be a lot of work to update it and, quite frankly, also pointless since we have seat projections which are more relevant. The page for the opinion polls of the 2017 election also doesn't include vote share (see here: https://en.wikipedia.org/wiki/Opinion_polling_for_the_2017_Dutch_general_election) --Spaastm (talk) 01:10, 28 October 2020 (UTC)

Ipsos publishes projections to the decimal number. We should keep these, as they are unique and more precise, leaving aside the margin of error. The source page for the Ipsos numbers gives access to all the results on a timeframe starting in 2017, so it's not going to be such a pain to add them. But indeed, Peil.nl's percentages do not add anything to the seat count, and should not be added.

The other advantage of the vote share is that it provides comparison with the midterm elections (provincial, european and municipal). Kahlores (talk) 23:22, 28 October 2020 (UTC)

Exit polls

They should be moved to main page not here. Shadow4dark (talk) 01:43, 18 March 2021 (UTC)

As I indicated in an edit, I was thinking as well about deleting the Exit Poll. However, I find them useful for people looking at this page, while the count is ongoing. Once the final results are in, we should override the Exit Poll numbers with actual results. Maybe not 100% by the book, but I think it is a good compromise --FantinoFalco (talk) 04:51, 18 March 2021 (UTC)