Talk:Opinion polling for the 2023 Dutch general election

Language

Hey, I do not really edit wiki pages so I don't want to interfere with your stuff, but I would suggest to replace the word "fuse" with "merge". — Preceding unsigned comment added by 86.92.89.111 (talk) 11:47, 19 April 2023 (UTC)

Thank you. done. Uwappa (talk) 16:27, 28 July 2023 (UTC)

Color blindness nightmare

Hi, the graph showing the ups and downs of each party is impossible to be read by a color blind person. Please lets make wikiPedia more acessible. Color blindness is not rare among men (15% of men have some degree of it). If the graph contained a small arrow wih name of each party it would be helpful 85.147.184.52 (talk) 17:52, 15 April 2023 (UTC)

Good call. But not easy to solve. The spaghetti caused by the multitude of parties, the dots to represent the different polls make these graph very hard to read even for people with very good colour vision. There are only so many colours (I once heard about 5-7) that are sufficiently distinguishable to use in a graph and the number of parties in this schem goes way beyond that. In addition colour schemes seem chosen to be as close as possible to the self-chosen party colours so that may be part of the problem. Ideally we should indeed apply some kind of additional distinguishing features - like dashed, dotted or otherwise distinguishing lines. Arnoutf (talk) 19:28, 15 April 2023 (UTC)

I'm certainly willing to implement some improvements, however it really is hard to come up with a good solution. Sure, dashed lines could be a part of the solution but it certainly doesn't solve the color blindness issue on its own since it's impossible to come up with 17 types of lines. Additionally, I don't really believe in a solution like this, since very often, multiple lines end up on more or less the same spot and then the assignment becomes ambiguous. Gbuvn (talk) 19:36, 15 April 2023 (UTC)

There are too many parties for a clear line graph. An alternative could be a stacked area graph, with parties sorted, current smallest on top, current largest at the bottom. Uwappa (talk) 20:58, 15 April 2023 (UTC)

Examples of similar charts:

• 
  stackedarea
• 
  stackedbar

Uwappa (talk) 08:27, 5 September 2023 (UTC)

Events in table

I think the way events are shown in the table violates the accessibility of the table. I would guess that screen readers are not able to properly identify those as seperate. Dajasj (talk) 16:04, 20 July 2023 (UTC)

Opinion poll reference consistency

Most of the polls' sources are hyperlinked on the polling firm's name, but the most recent ones are instead added to the footnotes with a link in a separate column of the table. We should do one or the other for consistency. (I don't really mind which one.)

The advantage of the hyperlinks is that the table is less cumbersome (one less column) and readers can more directly go to the source (with one click). The advantage of the footnote references is that we can have a proper citation of the source in case the link dies (which probably follows wikipedia guidelines better).

In any case, it's just confusing to have part of the article in each style. We should make it consistent. ElwinOwen (talk) 17:58, 21 July 2023 (UTC)

I started with the correct way, but have had no time to finish it yet. Footnotes are imo the way to go, to avoid stuff like linkrot. Dajasj (talk) 18:17, 21 July 2023 (UTC)

@Serolhak, why delete a start of proper references instead of external links? Dajasj (talk) 05:40, 25 July 2023 (UTC)

Removing column Lead

If we're removing columns, can we remove Lead. It isn't that important in a proportional system and the largest party is already highlighted Dajasj (talk) 05:44, 25 July 2023 (UTC)

Unreadable table with spammed texts

The table with these short "descriptions" of new party leaders, expelled members.. makes it unreadable and you can't compare older polls with new poll. Please remove them as it is unnecessary. Shadow4dark (talk) 04:28, 26 August 2023 (UTC)

I agree. The texts are taking over as it was a blog of political events. FantinoFalco (talk) 07:05, 26 August 2023 (UTC)

I disagree. Some of the breaks in the trends over the last few weeks (particularly the jump in support for GL/PvdA and the high listing from 0 to 31 of NSC) can only be understood by such happenings between polls. As is now only important (leader, new entry) changes of major parties are listed. Also given that all of it has calmed down now, there should be no (to hardly) any of such lines between now and election. I agree that it is a bit chaotic now, but so is the data. (PS in my view the graph is far worse as the splines do not respond well to breaks in trends and the multitude of lines make it more a plate of spaghetti than a readable graph). Arnoutf (talk) 22:08, 26 August 2023 (UTC)

But with that logic, I would leave out most events that have no effect. This includes the election of Bontenbal, Jetten, Yeşilgöz-Zegerius, Timmermans and all the unnecessary departures.

Although now that I think of it, the two events you describe (NSC and PvdA/GL) are also visible in the table itself. Dajasj (talk) 09:29, 5 September 2023 (UTC)

I agree, the stepping down should be trimmed Shadow4dark (talk) 18:56, 5 September 2023 (UTC)

Suggestion; use footnotes like a-g in 2023_Dutch_general_election#Political_parties_and_lead_candidates. Events that may impact polls will be visible on mouseover or scroll after click. Uwappa (talk) 17:22, 11 September 2023 (UTC)

References

Bar chart

Why include a bar chart? Right now, it is confusing because it suggests there is already an outcome. There isn't. And when this is all over, such a graph will be included in the election page, so they will be removed here. So I'm in favour of removing it. Dajasj (talk) 18:33, 14 October 2023 (UTC)

@Uwappa, 1100px is way too wide for most screens. Dajasj (talk) 18:38, 14 October 2023 (UTC)

Is there a way to rotate the labels 90 degrees so the chart could be much smaller? Uwappa (talk) 18:41, 14 October 2023 (UTC)

Yes, but my other point is that the bar chart adds little value on this page. The line chart already starts right after the election, so it gives the same information. And it is also already in the table. Dajasj (talk) 18:44, 14 October 2023 (UTC)

How to rotate the labels? Please do so the graph can be way smaller.

The bar chart makes it very easy to draw conclusions. It tells a story.

10

20

30

40

VVD

NSC

GL PvdA

PVV

BBB

D66

PvdD

CDA

FvD

SP

Volt

CU

SGP

DENK

JA21

BVNL

BIJ1

50+

•   Maintained
•   Gained
•   Lost

1. NSC is something special, did not have any seats yet. It is competing with VVD and GL/PvdA in becoming the largest party.
2. D66 is looking at a big loss, even bigger than CDA. D66 used to be the second biggest party but is now heading to become one of the medium sized parties.
3. CDA is loosing a lot too. The gain of NSC is way more than the loss of CDA. So NSC must gain voters from other parties too.
4. coalition parties VVD, D66, CDA and CU are all looking at a loss
5. BBB gains a lot, yet is far from becoming the largest party
6. BIJ1 and 50+ might loose their seat
7. something special for BVNL, no seats in last election and possible no seat in next election. What is the story?
8. PVV and Volt are at a small gain
9. PvdD, SGP and Denk are stationary
10. Smaller parties FvD, SP and Ja21 are looking at a relative substantial loss
11. No party is anywhere near a majority of 76 seats, a coalition is likely.
12. VVD, NSC and GL/PvdA are currently in the lead, candidates to participate in a coalition. PVV and BBB could be candidates to join a coalition.

It is possible to draw all of these conclusions from the line graph and the table, but that is much harder to do:

• The line graph is quite hard to read, especially for the smaller parties. See #Color_blindness_nightmare above and suggestions for an alternative graph, a stacked area chart. Results of previous election and latest poll are far apart which makes it hard to see seats gained/lost.
• The table is way to large, requires scrolling. It is impossible to see previous election results and latest poll at the same time. The table does not list change. To get an idea of seats gained/lost the reader must start computing. It is impossible to compare seats gained/lost of several parties at the same time.

Uwappa (talk) 19:30, 14 October 2023 (UTC)

If the bar chart is to stay, might I suggest we use a polling average rather than individual polls? That way we don't give one particular polling firm undue weight and don't have to update the graph with every new poll. Luxorr (talk) 21:17, 14 October 2023 (UTC)

The top row of the table could show the polling average so the table provides the same numbers as the bar chart. However, that polling average is probably different from the latest values shown in the line graph which will be confusing. Also such polling average combines fresh and older polls. The bar chart shows the latest poll, maintaining a relation between the last dots in the line graph, bar chart and current top row of the table. No undue weight to a specific polling firm as the chart reflects the latest poll, irrespective of polling firm. Keeping the bar chart up to date is not a problem. Uwappa (talk) 07:32, 15 October 2023 (UTC)

With every update the polling average combines the latest polls of both reputable polling firms. Right now at any given time the bar graph visualises one individual poll from one individual polling firm, which I think is a bit arbitrary. There are lots of polls throughout the campaign period and any poll could simply be an outlier. Luxorr (talk) 07:44, 15 October 2023 (UTC)

So, should the top row of the table show the polling average? Is it acceptable that the end of lines in the line graph differ from the polling average? Uwappa (talk) 07:47, 15 October 2023 (UTC)

I think having the polling average either at the top of the table or in a separate section (like here) would both be good. Personally I don't think these averages not matching up perfectly with the ends of the line graph doesn't matter much. Luxorr (talk) 08:41, 16 October 2023 (UTC)

We could also add a poll aggregate table for percentages (as opposed to seats) and add Politico Europe's poll of polls. It might look something like this:

Poll aggregator	Update date	VVD	D66	GL/PvdA	PVV	CDA	SP	FvD	PvdD	CU	Volt	JA21	SGP	DENK	50PLUS	BBB	BIJ1	NSC
Peilingwijzer	11 Oct 2023	17.3%	5.1%	15.0%	11.4%	3.5%	3.2%	3.0%	5.0%	3.0%	2.7%	1.9%	2.4%	2.1%	0.5%	8.0%	0.5%	16.1%
Politico Europe	9 Oct 2023	17%	4%	17%	12%	4%	3%	3%	4%	2%	3%	1%	2%	2%	1%	8%	1%	18%

Luxorr (talk) 09:20, 16 October 2023 (UTC)

I prefer this above the graph Dajasj (talk) 14:24, 23 October 2023 (UTC)

Hypothetical scenarios

What to do with Hypothetical scenarios?

1. keep, as a record of history
2. merge into the main table?
3. purge, now outdated?

Uwappa (talk) 06:56, 16 October 2023 (UTC)

1 it has historical value. Shadow4dark (talk) 08:17, 16 October 2023 (UTC)

Election Poll Leaders & Colors.

Hello. When adding polls, I have been trying to figure out how to shade and color election poll leaders for the 2023 Dutch election. For example, if the NSC is leading a poll, their number of seats predicted (ie 28 seats) is shaded in yellow.

I am wondering how exactly you shaded in the poll leaders' color (in this case, NSC is commonly leading, and their lead is shaded in yellow).

I would like to know how I can be able to do this, for when I add future polls to election wikipedia pages. Can you explain how candidates' leading in polls have their poll lead colors shaded in? Thanks Giantsfann48 (talk) 21:02, 14 November 2023 (UTC)

It is a bit of CSS, e.g. style="background:#F8E280;" Uwappa (talk) 22:57, 14 November 2023 (UTC)

peil.nl

Given that peil.nl is either an outlier or a first sign of a trend, it is irresponsible for Wikipedia to present that one poll in a bar chart, without any margins or anything. So Ill remove it. Well have the results a few days from now, and we wont need that bar chart here after the elections anyway Dajasj (talk) 17:52, 18 November 2023 (UTC)

It is the most recent poll. It is big news now in the NL, see https://nos.nl/collectie/13960/liveblog/2498269-wilders-ziet-gamechanger-in-peiling-timmermans-hoopt-op-strategische-stem#UPDATE-container-73500057 Uwappa (talk) 18:08, 18 November 2023 (UTC)

Well, three things. First, the bar chart isnt usefull on the page after the election anyway. Because the results will be communicated on the election page.

But more importantly, we are not a news site. We shouldnt highlight the latest poll. We shouldnt highlight any single poll. It suggests that that is the current state, which it isnt.

NOS btw mostly reports on the news around the poll. They only report the averages of the Peilingwijzer. That is the only responsible way to report on polls. Dajasj (talk) 18:17, 18 November 2023 (UTC)

Maintain a neutral point of view. Try to get consensus to remove the bar chart. I do not agree to remove it. It show the latest status, does not favour any polling station. It is the latest poll until another poll. Uwappa (talk) 18:30, 18 November 2023 (UTC)

Or get consensus to maintain the bar chart perhaps? Looks like 2v1 here.

And it isn't about favouring any polling station (or having a POV). I don't favour any, I have voiced my objections in the past. But this specific poll is significantly different than the others. And it is simply irresponsible to highlight one specific poll. That's how reliable media in the Netherlands also approach individual polls. I don't see why we should change that just to accommodate your desire to have this specific type of bar chart, to be honest. Dajasj (talk) 18:41, 18 November 2023 (UTC)

It is very special to

1. 17:52 post your own opinion on a talk page, claiming a majestic plural "we won't need" opinion
2. 17:53 remove the bar chart referring to your own opinion on the talk page without describing that you removed the chart
3. 18:41 claim that consensus should be reached to maintain the bar chart.

The latest poll is a major subject in the news, the Dutch NOS 8 o'clock TV news just started with this news. Uwappa (talk) 19:06, 18 November 2023 (UTC)

Regarding point 1: Do you believe it makes sense to have the bar chart after the final results on this page? Given that it will already be present on another page and this page should be focused on the polls, not the results.

Regarding point 2: "So Ill remove it."

Regarding point 3: I don't see your point here really. Dajasj (talk) 19:11, 18 November 2023 (UTC)

And to be clear, I don't propose removing the poll from the table. I simply don't want it to be highlighted in a bar chart. Because that was irresponsible to begin with and even more so with such an outlier. Dajasj (talk) 19:16, 18 November 2023 (UTC)

Please read WP:3RR and restore the chart. Uwappa (talk) 19:32, 18 November 2023 (UTC)

Thank you. Uwappa (talk) 19:49, 18 November 2023 (UTC)

Can you now return to answering my questions? Dajasj (talk) 19:52, 18 November 2023 (UTC)

Please

• Take back these words: "Stupid rule protecting nonsense".
• Respect the rules of Wikipedia.
• Respect the work of other editors.

Recommended reading: Graham's_Hierarchy_of_Disagreement based on the 2008 How to disagree. Uwappa (talk) 20:30, 18 November 2023 (UTC)

Thanks for this lecture. Regarding your first point, I do have some objections to the rule so I am not able to take these words back. They were however not aimed at you, although I disagree with the discussed content.

Can we now get back to the real questions, such as: do you want the bar chart present on this page after the election? Dajasj (talk) 20:37, 18 November 2023 (UTC)

O and given that we are citing rules now, please read WP:NOCON. Given that I have objected to the bar chart shortly after it was added, I think it follows that with no consensus we return to the version before the bar chart was added. Dajasj (talk) 20:43, 18 November 2023 (UTC)

I will repeat my position that the bar chart should be removed. As I have explained before, the most recent poll may not be representative of the state of the polls in general, because any particular poll may simply be an outlier, and visualising it so prominently gives it undue weight. Luxorr (talk) 22:58, 18 November 2023 (UTC)

It seems that peil.nl was the first to signal a trend, as it happened to be the first poll after the SBS6 debate, see https://maurice-nl.translate.goog/2023/11/18/sbs-debat-werd-de-gamechanger/?_x_tr_sl=nl&_x_tr_tl=en&_x_tr_hl=nl&_x_tr_pto=wapp

I suggest we reinsert the graph and keep it up to date, based on the latest poll (I&O Research, 20–21 Nov 2023):

10

20

30

40

PVV

VVD

GL–PvdA

NSC

D66

SP

PvdD

BBB

CDA

FvD

CU

Volt

DENK

SGP

JA21

BVNL

BIJ1

50+

•   Maintained
•   Gained
•   Lost

Uwappa (talk) 15:07, 21 November 2023 (UTC)

My issue was not with the particular poll being visualised, but with the principle of visualising any single poll at all, for the reason stated above. If there is to be a graph at all, it should visualise a polling average, preferably with margins of error. Luxorr (talk) 15:32, 21 November 2023 (UTC)

The problem with a polling average is time. Voters preferences change over time and polls measure the current preference, which may differ from yesterday's preferences, e.g. after a game changing debate. A polling average would mix old and new polls, giving undue weight to outdated polls. In a fast changing world I think the bar chart should show the latest poll and be updated frequently, a fair go for all polls. Usually the differences between poll results are within the margin of error, see 2 examples above and below.

In addition, a polling average does not yield better results. Please look at the 2021 polls and how they all had a margin of error, even one day before the election. An average would not decrease that margin of error, especially when polls are far apart in time.

The line chart already shows the history of all polls with an average. Unfortunately the line chart is not updated frequently and being an image, it can only be updated by the one Wikipedian having the right software and chart source data.

The bar chart can be updated frequently, being wikitext which all Wikipedians can update.

I have worked on an alternative design that shows the margin of error. The margin could be shown with a fuzzy colour shading between

• the border of maintained-lost
• the top of gained
• the top of maintained (for stable parties)

Showing the margin of error would not add much, a higher bar would still be higher. At the moment the line chart and table do not show such margin of error. So a bar chart showing an error margin would be out of tune. In polls the margin of uncertainty is high. Swing votes are usually as high as 40%. At the moment around a whopping 61% of voters are not sure yet. So my recommendation: describe the margin of error in a general text that applies for graphs as well as the table. Uwappa (talk) 16:23, 21 November 2023 (UTC)

The chart for the latest ipsos poll, see https://flo.uri.sh/visualisation/14486373/embed

10

20

30

40

VVD

PVV

GL–PvdA

NSC

D66

BBB

SP

PvdD

CDA

CU

Volt

DENK

FvD

SGP

JA21

BIJ1

BVNL

50+

•   Maintained
•   Gained
•   Lost

Uwappa (talk) 15:38, 21 November 2023 (UTC)

An average such as Peilingwijzer takes into account time. Really simple averages should do that as well, such as our graph. And again, we are not a news site, if things changes too fast we should perhaps not report on them. We focus on the long term.

You also point to the previous election, where Peil.nl overestimated PVV significantly. While no polling organization was perfect, in particular when it comes to D66, but a single polling org has a larger chance to have a biased result. Thats why experts try to avoid reporting on them.

I agree with your point about the line graph not being updated enough. But that could be solved by sharing the code and data. Dajasj (talk) 17:50, 21 November 2023 (UTC)

I am aware a polling average does not have a smaller margin of error than individual polls, but the margin of error tells us something about the precision of polls, while the issue I raised concerned the accuracy of polls. Every polling firm has its own "house effect", structurally underestimating or overestimating certain parties. A polling average takes those house effects out and reveals trends shared by multiple polling firms, generally improving its accuracy over individual polls.

As Dajasj mentions, polling averages take account of time, such that they generally show the current strength of parties across multiple polling firms. Individual polls are listed in the table, so if a reader is interested in checking the latest poll, he or she can do so. A graph, however, should not highlight a single poll. Luxorr (talk) 18:17, 21 November 2023 (UTC)

Please look again at the 2021 polls. Add up all deviations and see they all had a similar margin of error, sorted from small to large:

• 22: peil.nl
• 23: Kantar Pulic
• 24: I&O Research
• 26: Ipsos

Any source available for the 'house effects' that would cancel each other out in an average? Would that really work for peilingwijzer which is based on just I&O Research and Ipsos?

The main cause of the error margin is the 40% of swing voters. You can't solve that. The only solution is to accept the margin of error and describe it. An alternative, showing ranges should impact the table, the line chart and the bar chart. Please show an example of those 3 with ranges and how that would be an improvement.

I maintain: An average does not decrease the margin of error, it increases error by being partly based on yesterday's news. The easy solution: Treat all polls equally, simply show the latest poll. Yesterday's poll results will be replaced by today's poll. It is OK for an online encyclopedia to show recent results for an ongoing event.

Yes, the line graph already shows an average and dots for all poll results. The table shows all poll results, the latest, the most relevant on top. If you really want an average, show only the latest average, at the top of the table, close to the individual polls it is based on.

The ideal place for the bar chart would be in the table, near it's poll so readers can easily relate the numbers in the chart to the latest poll. An alternative is to describe the bar chart in text, like Dajasj version https://en.wikipedia.org/w/index.php?title=Opinion_polling_for_the_2023_Dutch_general_election&oldid=1185750311#Bar_chart_of_latest_poll Uwappa (talk) 19:15, 21 November 2023 (UTC)

When I said margins of error, what I meant was the range within which a party is estimated to be, based on the sample size of the poll. So for example, in the last I&O poll, the PVV is estimated to win between 25 and 31 seats.

If you were talking about the accuracy of polls, it may be true that all polling firms' predictions had a similar deviation from the actual result, but part of their deviations may be in different directions; one polling firm may overestimate party X while another underestimates the same party. Taking an average of the two may therefore increase accuracy. (See, for example, Jackman, 2006.) Moreover, it is important to keep in mind polls are not predictions of the election result, but snapshots of how voters would vote at the moment of fieldwork. Luxorr (talk) 19:32, 21 November 2023 (UTC)

Please show me how results of Dutch polling firms have different directions in 2021.

The chart for the latest peil.nl. It will be interesting to see how today's 3 polls relate to results of tomorrow elections.

10

20

30

40

PVV

GL–PvdA

VVD

NSC

D66

CDA

BBB

SP

FvD

PvdD

DENK

CU

Volt

SGP

JA21

BIJ1

BVNL

50+

•   Maintained
•   Gained
•   Lost

And yes, I concur, polls are snapshots, taken at a moment in time. That is exactly why today's poll is more relevant than yesterday's. Uwappa (talk) 19:47, 21 November 2023 (UTC)

You can see here that Peil.nl overestimates PVV and the other two overestimated GL. But another challenge is that polling organisations also change overtime and the house effect may be different each year.

And again, a more recent poll is more relevant. And polling averages account for that. But an old poll is not irrelevant, because a recent poll has its downsides. And polling averages can put this recent poll into context. Dajasj (talk) 19:53, 21 November 2023 (UTC)

Polling firms having house effects, at any time and in any country, is a given. It is also true that a polling average that takes account of these house effects tends to be more accurate as a result. Prominently visualising one single poll, which may be subject to house effects or incidental inaccuracy, gives it undue weight. Luxorr (talk) 19:58, 21 November 2023 (UTC)

The 40% of swing voters is not the margin of error. It is a challenge when trying to predict elections, which as Luxorr explains is not really the goal. The margin of error has to do with the sample size.

Regarding the house effects, see the explanation of peilingwijzer.nl: https://peilingwijzer.tomlouwerse.nl/vragen.html. In general, I believe this provides the answer to most of your questions, because you seem to misunderstand polling and the challenges that come with them. Dajasj (talk) 19:48, 21 November 2023 (UTC)

Dajasj, please read WP:NPA and take back your words. Uwappa (talk) 20:01, 21 November 2023 (UTC)

Hi Uwappa, it was not meant as a personal attack and WP:NPA does not seem to define my comment as such. But you really don't seem to understand how polling, or at least the margin of error, works. So I hoped that by reading that piece we could continue this discussion with a shared understanding.

One note to clarify, polling average is maybe a bit misleading. Peilingwijzer does not simply average, but uses Bayesian statistics to aggregate. Dajasj (talk) 20:07, 21 November 2023 (UTC)

Second request: please take back your words. Uwappa (talk) 20:10, 21 November 2023 (UTC)

No, sorry. I don't believe there is something wrong with noting that another user might not completely understand a topic. It was however not my intention to offend you. It was merely to come to a shared understanding that would have benefitted the valid discussion we were having.

I hope we can continue our discussion. Dajasj (talk) 20:17, 21 November 2023 (UTC)

Feel to request a block, but please don't remove something which doesnt qualify as a personal attack. Or better, lets continue discussing content. Dajasj (talk) 15:32, 22 November 2023 (UTC)

This article perfectly explains why media and Wikipedia should be careful to highlight individual polls. Dajasj (talk) 20:23, 21 November 2023 (UTC)

References