Talk:Nationwide opinion polling for the 2020 Democratic Party presidential primaries/Archive 2

Aggregate polls
Should we also include poll aggregates as we did in 2016? SecretName101 (talk) 19:21, 6 July 2019 (UTC)

's justification for removing them was "just a personal view, but the RCP average is terrible, deliberately refuses to use the numbers preferred by pollsters themselves (the YouGov/Economist polls, for example, are already among likely primary/caucus voters so it's pointless to filter them down to "registered voters"... people can register before the primary), and also ridiculously selective in which polls they include/exclude"

I argue we should include aggregates.

SecretName101 (talk) 19:24, 6 July 2019 (UTC)
 * I argue that we shouldn't use an aggregate that 1) deliberately excludes many polls, 2) is entirely based on the whims of arbitrary decisions as to whether drop off single polls from the average after a certain point, 3) refuses to use the numbers actually used by pollsters (e.g. YouGov/Economist every week and more recently ABC/Post), 4) often contains errors for a long time (e.g. by listing the 5 weighted respondents for Gabbard in a Harvard-Harris poll as "5%" for weeks before it was corrected), and 5) also repeatedly has issues with correctly listing polls (e.g. repeatedly incorrectly listing candidates in Suffolk University polls as earning 0%, calling the weekly Morning Consult poll "Politico/Morning Consult" when it has literally nothing to do with Politico's entirely separate weekly tracking polls). I check RCP with some regularity, but they're also careless and something of a joke. Mélencron (talk) 19:35, 6 July 2019 (UTC)
 * Yes, let's have a separate list of polling aggregations. Raising issues that individual aggregations may have is purely original research. We do not include polling results in Wikipedia because they are correct or because there is any degree of accuracy in predicting current or future political support. Polling data is notable in its own right, and is only more notable when there are significant errors. Onetwothreeip (talk) 23:20, 6 July 2019 (UTC)
 * I'm not making a comment about the accuracy of polls – I'm saying that there are errors and omissions in the RCP aggregate, and it should neither be included nor treated seriously. It's essentially nothing but a right-wing blog site that happens to list some but not all polls. (I would also argue that it's not OR to raise and point out obvious errors or issues with sources: if a source contains known errors, then don't use it, period.) Mélencron (talk) 01:36, 7 July 2019 (UTC)
 * These are not serious errors at all, these are just choices that you disagree with, and errors of fact that we could expect to find from any reliable source. Their polling averages are well known and are relied upon, for better or worse, regardless of how they are created. We shouldn't rely on RealClearPolitics for the individual polls, it's just that their aggregations are content themselves. Onetwothreeip (talk) 01:55, 7 July 2019 (UTC)
 * Sorry, I must've missed the part that missing 33 polls since the start of April and misreporting the results of 14 of them isn't considered a "serious error". I actually don't have the time to engage with you right now, so I'm not pursuing this further for now. Mélencron (talk) 02:00, 7 July 2019 (UTC)
 * If they are as unreliable as you think they are, we should be able to ascertain that from other reliable sources claiming so. Omitting 33 polls is not a serious error. Onetwothreeip (talk) 02:37, 7 July 2019 (UTC)

Swalwell has dropped out
Eric Swalwell has ended his campaign as of July 8. This should be noted in the table. 2600:1014:B062:15C4:BC62:10C5:C6EA:21AC (talk) 20:52, 8 July 2019 (UTC)


 * Tom Steyer meanwhile has joined the race. That should be noted too. 97.116.77.170 (talk) 21:45, 10 July 2019 (UTC)

Versions of the same poll
Okay, what are we doing here? There seems to be a misguided idea that we should only report polls as they are used for debate qualification. I would think by precedent we should use polls from those who report as supporting or leaning towards the Democratic Party. We should also think quite seriously about reporting more than one version of the polls too. Onetwothreeip (talk) 05:23, 11 July 2019 (UTC)
 * Can you be more specific about which polls you think we should add? — JFG talk 19:49, 11 July 2019 (UTC)

Suggestion: split on April 1
The current tables are split in October 2018, January 2019, and March 2019. I would argue for moving the latest split to April 2019 instead of March. That would fall on a trimester boundary like the previous splits. Moreover, the data from March still shows mostly 3% support for Klobuchar, and includes two polls featuring Clinton at 8% and 11% even after she declined to run. Therefore, pre-March columns would still make sense in March. After April 1, Klobuchar support drops to 1-2%, and Clinton is no longer featured. — JFG talk 06:55, 10 July 2019 (UTC)
 * Yes, I would combine January 2019 to June 2019's results, and then make a new table for July onwards. We should also create a completely separate table for polls which have asked about Hillary Clinton and Michelle Obama, since they fundamentally change the context and can't be compared with other polls that exclude them. Onetwothreeip (talk) 07:25, 10 July 2019 (UTC)
 * Meh. January to June would be too long. The proposed split by trimester is more reasonable, and ties in well with the removal of Clinton's column. I agree that pollsters should not ever have asked about Michelle Obama, but they do as they please, and we must report her popularity, even though she never gave any hint of personally moving into politics. Giving a separate table to Clinton/Obama/Winfrey and other such non-candidates would be undue weight imho. — JFG talk 19:48, 11 July 2019 (UTC)
 * The main disadvantage of a trimester-based split, which would otherwise be neat, is that Buttigieg got a few 3% poll results in March, plus one 4%, so should probably be included in any table which has March in- whereas January+February polls mostly don't include Buttigieg at all. Chessrat  ( talk, contributions ) 20:45, 11 July 2019 (UTC)
 * I think people should have a column when they are at or above 3% during a significant part of the trimester, say averaging 3% over one month. Buttigied didn't reach that sustained level until April. — JFG talk 05:22, 12 July 2019 (UTC)
 * ✅ - Now split on quarters at April and July. Looks good afaic. — JFG talk 22:29, 15 July 2019 (UTC)

Gabbard and Yang in graph
Now that Gabbard and Yang have been added to the poll aggregate table, should they be included in the main graph that is featured in the article? Pinging editors who have been active in past discussions ---. { [ ( jjj 1238 ) ] } 22:11, 15 July 2019 (UTC)
 * Thanks; I got your ping. I personally don't think those candidates should go on the graph yet; let's wait a little first to see if their numbers increase. --Comment by Selfie City ( talk about my  contributions ) 22:16, 15 July 2019 (UTC)
 * Right, they're still too low for now. If the uptrend continues, they may enter the graph while Booker may leave it. — JFG talk 22:28, 15 July 2019 (UTC)
 * My opinion: not yet- they're above Booker in the polling aggregate table at the moment, but it's too early to tell if the trend will continue. Chessrat  ( talk, contributions ) 22:43, 15 July 2019 (UTC)
 * On the contrary, Booker should be removed from the graph. Onetwothreeip (talk) 23:31, 15 July 2019 (UTC)
 * I'm not opposed to removing Booker, but if we do, shouldn't we remove O'Rourke too? His downward trend line seems to be quite parallel to Booker's. --Comment by Selfie City ( talk about my  contributions ) 23:55, 15 July 2019 (UTC)
 * It doesn't matter what their current polling is, what matters is the highest they have polled. We really need to stop thinking in terms of the last week, and edit as if this is ten years from now and we are looking at the period in its entirety. Onetwothreeip (talk) 00:00, 16 July 2019 (UTC)
 * Even in that case, O'Rourke either should be removed or is close to being removed. He polled fairly high for a short while, but Booker has polled reasonably high in a few polls as well. --Comment by Selfie City ( talk about my  contributions ) 00:07, 16 July 2019 (UTC)
 * The graph isn't about current performance though, it's about the entire campaign. Regarding Booker, it's up to us to decide if we consider a 4% average to be high enough to show him on the graph. Onetwothreeip (talk) 02:45, 16 July 2019 (UTC)
 * We don't have to delete the whole line; we just need to make it stop around the end of this month. --Comment by Selfie City ( talk about my  contributions ) 03:46, 16 July 2019 (UTC)
 * That might not be necessary- after all, Buttigieg has a line on the graph in February despite having no support back then. Chessrat  ( talk, contributions ) 13:54, 16 July 2019 (UTC)
 * Well, that line is a good way of demonstrating how his support was low and then increased dramatically. --Comment by Selfie City ( talk about my  contributions ) 14:28, 16 July 2019 (UTC)

Time to revisit who's included in the table soon?
The table so far contains nine polls from July (which would be a good point to start a new table from, given the recent debate, etc). Of those nine polls, the candidates have average scores of:

Biden 24.89% Sanders 15.78% Harris 14.89% Warren 14.78% Buttigieg 5.78% O'Rourke 2.56% Booker 1.67% Castro 1.44% Yang 1.44% Gabbard 1.00% Klobuchar 1.00% All others: 0.56% or less

Similar results can be seen in the polling averages calculated by RealClearPolitics and 270toWin (seen in the table in the article). Now, I'm not sure what the best inclusion criteria are here, but due to the very small gap between Booker and Castro/Yang, I think we probably need to revisit which candidates to have columns for, and which to relegate to the "Others" column. There are a few options, as I see it:
 * A) Down to five candidates with columns; move Booker and O'Rourke to "Others"
 * B) Down to six candidates with columns; only move Booker to "Others"
 * C) Seven candidates with columns: current table; no change
 * D) Eleven candidates with columns: everyone over 1%; i.e. add Castro, Yang, Klobuchar, and Gabbard.

It might be best to wait a week or two before making any changes, to see what trends are sustained, but I thought it'd be worth raising this topic now to see if people generally lean towards more or fewer columns. If the number of columns is changed, it would make sense to start a new table from the start of July, I think. Chessrat ( talk, contributions ) 18:11, 9 July 2019 (UTC)


 * Arguably, this has become a four-candidate race: Biden, Sanders, Warren, and Harris. At the same time, however, we want our article and information to be informative, so the best thing is a balance between key political figures and containing the amount of information people would expect. --Comment by Selfie City ( talk about my  contributions ) 00:37, 10 July 2019 (UTC)


 * For comparison purposes, the Nationwide opinion polling for the 2016 Republican Party presidential primaries article contains all candidates in the main table... even when there were 17 of them... but that looks crowded and it's not that easy to glean the data from it. For this article, personally I'm leaning towards the option D that I mentioned above– that's basically showing as many candidates as possible in the columns, without making it too crowded. Chessrat  ( talk, contributions ) 02:08, 10 July 2019 (UTC)


 * That’s a lot, though. I’d probably go with option B for the table starting August 1st (assuming some other candidate doesn’t suddenly get high percentages) but keep Booker on the graph for now. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 02:22, 10 July 2019 (UTC)


 * In prior discussions, candidates were selected based on getting 3% or higher in any of the 10 most recent polls. On this criterion, Booker should soon be out, because he only got 3% once in the last 10 polls, while O'Rourke still gets up to 4% in several recent polls. Therefore I'd support option B. I also agree that switching can wait a couple weeks, to confirm that Booker remains on the decline, and as we see whether Steyer gets any traction in the polls. — JFG talk 06:38, 10 July 2019 (UTC)
 * Even if we wait a few weeks to change columns, I'd argue that a natural cutting point would be July 1st (a trimester boundary, and a change in trend after the first round of debates). — JFG talk 06:40, 10 July 2019 (UTC)
 * I would argue that it should be based on getting 3% (or 4%) or higher in the last month. When you have a debate you have a ton of polls that come out at once, and if someone had a bad debate they could dip, but then come back up. - Nablais (talk) 13:50, 10 July 2019 (UTC)
 * We really need to avoid the criteria of "the last 10 polls" as it's an awful case of recentism. We need to assess whether the candidate has polled a sufficient level throughout the table's period, not just the end. I would also strongly caution against the Others column having results of over 20%, so I don't think it would be wise to remove both Booker and O'Rourke. Onetwothreeip (talk) 00:07, 11 July 2019 (UTC)
 * Agreed, "has reached 3% in at least one of the last ten polls" is too much recentism. If that criterion were applied now, Castro and Yang would have columns, but Booker wouldn't... if a criterion like that is used, maybe it should be something more like "has reached 3% in at least 10% of the polls in the previous two weeks, and reached 2% in at least 40% of those polls". Chessrat  ( talk, contributions ) 22:27, 11 July 2019 (UTC)
 * I agree too: I think people should have a column when they are at or above 3% during a significant part of the trimester, say averaging 3% over a month. — JFG talk 05:23, 12 July 2019 (UTC)

The "polling aggregation" section specifically
I think this could be changed more immediately, as there are only two polling summaries in that section, and both already all show Booker as not having meaningfully more support than some of the candidates in the "Others" column. Would suggest switching to one of these two options for that particular section: Chessrat  ( talk, contributions ) 22:26, 12 July 2019 (UTC)


 * Agree, and just trimmed the list.. However O'Rourke should be removed too, because he is well under 3% average support. — JFG talk 18:24, 16 July 2019 (UTC) Update: I have removed him. — JFG talk 18:28, 16 July 2019 (UTC)
 * Yes, probably O'Rourke should be removed from the list. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 18:30, 16 July 2019 (UTC)

Question for split
Should the polls be split based on how the candidates could qualify for the third debate(s)? That seems like a more sensible split than using the months, and wouldn't that include the polls conducted after the first debates that are currently listed in the April 2019 to June 2019 section? 170.135.176.108 (talk) 12:53, 17 July 2019 (UTC)
 * I'm not sure I quite understand your proposal, but to me the main reason for splitting the polls at all (rather than just having one long table) is so the columns can change as candidates' support does. For example, the July 2019 - present table doesn't have a column for Booker because he no longer has as much support as he used to. This wouldn't really be possible if the polls were split based on whether they could be used to qualify for debates. SCC California (talk) 17:06, 17 July 2019 (UTC)
 * I could be wrong, but my interpretation of what the IP editor said is that the polls considered "major" by the DNC (aka those that can qualify candidates for the debates) be split from the other polls considered more minor that do not count for qualification. Personally, I think it is a good idea to denote some sort of indication of the major polls versus the non-major polls. <b style="color: #AB2B2B;">{ [ ( jjj</b> <b style="color: #000000;">1238 ) ] }</b> 17:39, 17 July 2019 (UTC)
 * I've had this thought before as well. Maybe we could color code the specific polls that count towards debate qualification? Just a thought. - EditDude (talk) 17:51, 17 July 2019 (UTC)
 * This was the intent, yes. Sorry if that was unclear. 170.135.176.108 (talk) 18:07, 17 July 2019 (UTC)
 * I think highlighting the major polls with a different color is probably the easiest and most obvious way of handling it. <b style="color: #AB2B2B;">{ [ ( jjj</b> <b style="color: #000000;">1238 ) ] }</b> 19:57, 17 July 2019 (UTC)
 * It would be madness to split qualifying polls into a new table. I might agree with highlighting them a certain colour, but only until this really matters for debates. Onetwothreeip (talk) 23:42, 17 July 2019 (UTC)

Error in the chart
The chart is quite outdated now, particularly the polling results for Kamala Harris have changed significantly. It would also be helpful for readers to know up to what date the graph measures. Onetwothreeip (talk) 00:50, 2 July 2019 (UTC)
 * Yes, Kamala Harris is definitely wrong on the chart. --Comment by  Selfie City  ( talk about my  contributions ) 00:59, 2 July 2019 (UTC)
 * I was planning to update it on Wednesday given the number of post-debate polls that are going to come out over the next couple days (three national polls and an Iowa poll within the next 48 hours); right now most of the post-debate polls come from just two pollsters. Mélencron (talk) 01:16, 2 July 2019 (UTC)
 * It's clear we need a way for anybody to be able to update the chart. Onetwothreeip (talk) 01:17, 2 July 2019 (UTC)
 * I think Mélencron's explanation is perfectly reasonable. --Comment by  Selfie City  ( talk about my  contributions ) 02:59, 2 July 2019 (UTC)
 * I understand what they are staying but the chart is currently inaccurate and this should be fixed. I would fix it myself if I could, which is why we need a way for anybody to be able to update the chart, not just one person. At the least we should be hiding the chart until the problem is solved. Onetwothreeip (talk) 03:02, 2 July 2019 (UTC)
 * I have removed that chart/information accordingly; it can be re-added when the chart is updated. --Comment by  Selfie City  ( talk about my  contributions ) 03:26, 2 July 2019 (UTC)
 * Would it be possible to use a model like the one on this polling page? It seems to update itself on new polls being entered, rather than someone manually editing a chart. 20:34, 2 July 2019 (UTC) — Preceding unsigned comment added by 192.76.8.64 (talk)
 * That chart is not updated automatically either. It comes from File:Opinion polling for the next United Kingdom general election.svg, and as you can see, Mélencron is updating it accordingly as they are with the chart on this article. I do believe that updating the chart weekly when polls are coming in everyday is undesirable, but updating it everyday is also a lot of work. A chart that can be edited by everyone seems to be the best option, but how can that be implemented. <b style="color: #AB2B2B;">{ [ ( jjj</b> <b style="color: #000000;">1238 ) ] }</b> 20:06, 2 July 2019 (UTC)


 * I have created this graph in my sandbox based on a template, making the graph easily editable. It's not perfect by any means, but would enable anyone to edit the graph. How do we feel about using this choice instead? --Comment by  Selfie City  ( talk about my  contributions ) 21:13, 2 July 2019 (UTC)


 * Melencron's graph has been updated. --Comment by  Selfie City  ( talk about my  contributions ) 21:15, 2 July 2019 (UTC)
 * While Mélencron's graph has been updated, this does not prevent the concern we raised today from happening in the future. If Gabbard suddenly rose 10 points in the next four days, would we still have to wait for a weekly update to even see her on the chart? I think the graph that you created in your sandbox is a start; we need a way that anyone can update the graph at any given time. <b style="color: #AB2B2B;">{ [ ( jjj</b> <b style="color: #000000;">1238 ) ] }</b> 21:20, 2 July 2019 (UTC)
 * Just ping me, I edit every day. In general I tend to update these charts pretty frequently (UK with the release of every other poll or so, Germany every week, German states with every new poll, France/Netherlands every month given the volume of polling data, Dem primary frequently from Wednesday through Friday as well as additional updates as I see fit). In this case I chose not to update immediately because almost all the data came from just two pollsters, but now there are five in the mix with post-debate data with another three or so likely to be released within the next 18 hours. (Also, sorry, but the quality of editable charts on Wikipedia is abysmal, which is why they're basically never used in favor of user-maintained ones – Impru20, Avopeas, and myself each maintain such charts for multiple countries.) Mélencron (talk) 21:24, 2 July 2019 (UTC)
 * Would you be okay with maybe an every-other-day update (provided there is relevant polling)? That would probably suffice. <b style="color: #AB2B2B;">{ [ ( jjj</b> <b style="color: #000000;">1238 ) ] }</b> 21:25, 2 July 2019 (UTC)
 * Given the higher volume of polling, I'd be fine with doing daily updates (as I've also streamlined the process for updating these somewhat). Mélencron (talk) 21:28, 2 July 2019 (UTC)
 * Well if we can commit to that then it's fine with me. The weekly updates were really my only issue since if you're doing daily updates there's no need for everyone to be able to do the updates themselves. <b style="color: #AB2B2B;">{ [ ( jjj</b> <b style="color: #000000;">1238 ) ] }</b> 21:32, 2 July 2019 (UTC)
 * It all should be good if we do regular updates. --Comment by  Selfie City  ( talk about my  contributions ) 22:47, 2 July 2019 (UTC)
 * It's been 9 days since the last update, during which time we've had three polls. I don't think it's sustainable to have complete reliance on one person to keep this up-to-date 09:52, 12 July 2019 (UTC) — Preceding unsigned comment added by 109.94.138.129 (talk)


 * One mitigating solution would be to have the chart prominently state when it was last updated, so a reader can gather for themselves how up-to-date it is. These edit wars temporarily removing a great visualization of months of data everytime it's a few days out of date seem really counterproductive. When there's a lag, the chart is still an accurate representation of historical information and should remain accessible. 209.6.126.227 (talk) 15:46, 12 July 2019 (UTC)
 * The date should absolutely be provided. What is very urgently needed is a way for any editor to update the graph. We can't have outdated and misleading information. Onetwothreeip (talk) 00:18, 13 July 2019 (UTC)
 * This is getting absurd. The date, although better than nothing, isn't even that visible and could easily be missed. We are now 15 days since the last update. I vote either we have something that any editor can update or we remove it again.Nablais (talk) 13:58, 18 July 2019 (UTC)
 * We might as well just use the Wikipedia template then. It may be bad but is better than nothing. I’m sure it would be possible to add more regular updates than months. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 14:56, 18 July 2019 (UTC)
 * I have made a Wikipedia graph using the original graph’s data and added the new one. It only includes Warren, Biden, and Sanders, so please add the others as needed. Thanks. When that’s done we can remove the original. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 15:30, 18 July 2019 (UTC)

Qualifying Pollsters
Considering qualifying polls for the debates have been shaded, I think we should make special note of Reuters and The Las Vegas Review-Journal only counting towards the first and second debate, as they don't count towards qualification is the third debate. Political Vacuum (talk) 19:24, 18 July 2019 (UTC)
 * That is covered in the "Background" section. Now that the period of qualification for the July debates is over, those pollsters just will no longer be shaded. <b style="color: #AB2B2B;">{ [ ( jjj</b> <b style="color: #000000;">1238 ) ] }</b> 19:30, 18 July 2019 (UTC)

Out of Date Graph (07/18)
It looks like the user who was updating the graph showing nationwide polling results and trend lines has not been active on Wikipedia for 11 days.

Is there another user with the technical know-how to update it? If not, perhaps we should remove the graph so as to not give a false impression. --2610:48:100:A9E:0:0:0:60 (talk) 13:43, 18 July 2019 (UTC)
 * ✅ There is now a new graph that can be updated by Wikipedia users. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 15:53, 19 July 2019 (UTC)

Editable graph
@SelfieCity - thanks for adding the up-to-date graph to this page. I did not know there was a such thing as a built-in WP graph before. Would you please share here how you got the numbers for the first and 15th of each month so that others (including myself) can help keep it updated and help add other candidates? Thank you. SCC California (talk) 16:30, 18 July 2019 (UTC)
 * I just used the numbers from the other, older graph that’s apparently gone now. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 16:42, 18 July 2019 (UTC)
 * Oh, I see. I removed the other graph because I thought yours was more updated with a July 15 number. We should figure out a regression to use so we can update it moving forward. In the mean time, I'll add Booker, Buttigieg, Harris, and O'Rourke based off the old graph. SCC California (talk) 16:52, 18 July 2019 (UTC)
 * Thanks! --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 16:53, 18 July 2019 (UTC)
 * Also, feel free to take a look at User:SelfieCity/sandbox where I made the graph. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 17:06, 18 July 2019 (UTC)

The graph currently ends at July 1. What's the point of replacing the other graph without adding new numbers? --Gbuvn (talk) 17:53, 18 July 2019 (UTC)


 * Be patient. The point is now we can add new numbers.  I could take a stab at it, using the polling aggregates that were in this edit (the last one here that didn't include data after July 15).   --Spiffy sperry (talk) 18:02, 18 July 2019 (UTC)
 * Thanks to those who added the next two weeks of information. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 18:14, 18 July 2019 (UTC)
 * Yes, thanks Spiffy sperry for adding new information. Would you please tell how you got the numbers so we can consistently use the same method? SCC California (talk) 18:40, 18 July 2019 (UTC)
 * See my link above, which is the last version of this article (I think) with no polling aggregate data after July 15. I averaged those polling aggregates, and rounded to the nearest 0.5% to match the precision of the other numbers in this graph.  I won't claim that this is the only way to do it, but it seemed reasonable for a first shot.  --Spiffy sperry (talk) 18:58, 18 July 2019 (UTC)
 * The term aggregates refers to 270toWin and RealClearPolitics, right? Thanks. That makes calculating easy in future. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 19:31, 18 July 2019 (UTC)

The editable graph would probably be best as a separate template, similar to how we deal with route maps for railroads. SecretName101 (talk) 00:11, 19 July 2019 (UTC)
 * It is a separate template, right? --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 01:08, 19 July 2019 (UTC)
 * I think separate template means, there are no data in the source code of this article, just something like and then add the data there.--Gbuvn (talk) 09:32, 19 July 2019 (UTC)
 * I guess we could do that, but does it really provide any advantage? You just move the same content to another page. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 15:53, 19 July 2019 (UTC)
 * I also think it would not be advantageous to create a new page just for this chart. SCC California (talk) 16:11, 19 July 2019 (UTC)
 * In my opinion, it is more convenient just to edit the graph right were it is displayed. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 16:19, 19 July 2019 (UTC)

Favorability ratings
There are so many candidates, all at roughly the same spot on the political compass, that vote-splitting is going to heavily skew polls that ask "Which candidate will you vote for?"

We should show the results of favorability polls, too, which are immune to vote-splitting. For example: https://i.imgur.com/bHXevGo.png https://poll.qu.edu/national/release-detail?ReleaseID=2622 — Omegatron (talk) 22:07, 20 July 2019 (UTC)


 * I added a few. It took a very long time.  My spreadsheet includes favorable and unfavorable for all 22 candidates, so I could add more. Hopefully others are better than me at collecting the data.    I think favorable / (favorable+unfavorable) is a better overall metric, but favorable-unfavorable is what everyone else reports, like 538, etc. — Omegatron (talk) 06:07, 26 July 2019 (UTC)

Unconfirmed debate qualifying polls
Why is the NBC News/SurveyMonkey poll from Jul 2–16 marked as unconfirmed debate qualifying poll? - Ich bin es einfach (talk) 16:04, 27 July 2019 (UTC)

Graphical summary colors
Several editors have been editing the colors in the graphical summary, resulting in it changing almost daily. To me, there seems little reason to have the colors anything but what's easily distinguishable, but let's try to come to some consensus before changing it more. SCC California (talk) 16:56, 29 July 2019 (UTC)
 * Let's stick with the current. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 19:54, 29 July 2019 (UTC)


 * The one change I would make to the current version is to make O'Rourke's color a little less pale. I think it would be easier on the eyes if it were brighter. Jacoby531 (talk) 20:02, 29 July 2019 (UTC)
 * Right. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 21:14, 29 July 2019 (UTC)
 * Buttigieg and Warren look too much alike in my opinion. Other than that it is fine. - Ich bin es einfach (talk) 16:55, 30 July 2019 (UTC)
 * Same remark. I'll make Warren a shade of red. — JFG talk 12:16, 1 August 2019 (UTC)
 * For Beto, I'd suggest grey instead of pink. See what you think. — JFG talk 12:17, 1 August 2019 (UTC)
 * Let's stick to those colors. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 12:32, 1 August 2019 (UTC)

Starting point of graph- January or March?
The starting point of the graph was truncated to only begin in March. I extended this back to January (where it originally started), but was reverted by asking to discuss on talk. I really can't see any reason not to begin the graph in January- all other polling articles contain as much data as possible and don't truncate it, so why should this situation be any different? Chessrat ( talk, contributions ) 22:14, 3 August 2019 (UTC)
 * My apologies. I was reverting your edits adding Yang into the graph (please see above discussion). I, for one, have no issue with you extending the graph back to January. SCC California (talk) 22:18, 3 August 2019 (UTC)
 * Great. So now we have a lovely white, blank space at the beginning of the graph. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 23:19, 3 August 2019 (UTC)
 * Have updated it to the start of December 2018, to avoid whitespace. Chessrat  ( talk, contributions ) 01:45, 4 August 2019 (UTC)
 * Thanks. That looks much better now. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 01:51, 4 August 2019 (UTC)
 * Yes, great work! — JFG talk 11:50, 4 August 2019 (UTC)

Correction/ Missing Poll
Gabbard most definitely did NOT get 100% in the most recent poll. Fix it please. 199.59.117.115 (talk) 01:50, 5 August 2019 (UTC)

Can someone please add the Change Research poll from August 2-4?: https://www.changeresearch.com/post-debate-poll-august-2-4-2019

Booker/Yang
Yang's numbers are now significantly higher than those of Cory Booker. Isn't it about time that we added Yang's numbers to the graph? --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 12:17, 31 July 2019 (UTC)


 * I have no objection to this, but how would you find/calculate the historical numbers? I think we should use the same method as the other numbers that are in the graph.  Since the numbers prior to July are based on another user's image of major candidates (this one), then Yang's pre-July numbers could be added based on that user's other image of minor candidates (that one). --Spiffy sperry (talk) 14:12, 31 July 2019 (UTC)
 * I think our metric for inclusion on the graph had been that the candidate had reached 3% at any time. Because Booker has been at and above 3% since the beginning of the graph (March), he was included. Yang has not ever been at this level of polling and isn't included. In short: neither of them has the numbers to be included based on July (which is why neither is included in the polling aggregation table), but Booker is still in because of his historically higher numbers. SCC California (talk) 15:19, 31 July 2019 (UTC)
 * I would also add that, in several months, once the chart only displays past March May 8, Booker would be removed (unless his poll numbers reach 3% again). SCC California (talk) 15:22, 31 July 2019 (UTC)
 * Thanks; your explanation really helps. Soon, though, the table will only include after March 8, as it starts in March now. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 20:05, 31 July 2019 (UTC)
 * Sorry, I meant to say May 8. That looks like the date when Booker dropped below 3%. SCC California (talk) 20:45, 31 July 2019 (UTC)
 * Yes, the removal of May 8 is a little while away. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 11:43, 1 August 2019 (UTC)
 * @2604:2DC0:200:2AF:0:0:0:0: we have a consensus-based policy here to include all candidates whose polling averages have reached over 3% in the relevant timeframe and no others. You are welcome to propose a change to this policy, but putting all of the candidates would be extremely cluttered and hard to read (like this (this is frolm before the user became inactive)). SCC California (talk) 18:33, 1 August 2019 (UTC)


 * Yang is now on the graph, and rightly so, in my opinion. While I see SCC California's point about the 3%, it did not feel right that we were not allowing Yang on the graph but allowing Booker. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 21:33, 3 August 2019 (UTC)
 * Hi - I undid that revision by Chessrat per WP:BRD. I am happy to make changes, but until we have some consensus, let's stick with the old policy. One possible policy is to include anyone above some percentage, maybe 2%, at any given time, but that would have the effect of Booker dropping off of the graph while he didn't drop out of the race. This is why we have Booker still included even though he's now polling far below what he used to. We could also include anything above the lowest line on the graph by the 3% rule, but that would include a lot of low-polling candidates because of Buttigieg in March. Maybe something like to include:

1) Anyone whose polling average has been at or above 3% at any time, and

2) Anyone whose polling average has been above the lowest line on the graph per rule 1 at any time, given their polling average was, at the time, above 1.5%.

Thoughts? SCC California (talk) 22:17, 3 August 2019 (UTC)


 * Those two rules would seem reasonable, but I would point out that they'd mean Amy Klobuchar would be included- on March 15, Klobuchar was on an average of about 2.2%, ahead of Buttigieg on 1.5%. (Klobuchar's peak polling average was 2.8% in February, so she wouldn't pass criterion 1, but it looks like she would pass criterion 2). Chessrat  ( talk, contributions ) 23:08, 3 August 2019 (UTC)


 * Rule 1 is enough. We can insert Yang and others whenever they reach 3% over a couple months. — JFG talk 11:49, 4 August 2019 (UTC)


 * Beto and Booker have also polled below 3%. Why are they included, but not Yang who has similar poll number? Isn't that an unfair double standard towards Yang because he's not a politician? Pennsylvania2
 * @Pennsylvania2: There's nothing in the rule about whether anyone is a politician or not. Booker and O'Rourke's polling averages have both been at or above 3% (Booker from March and earlier; O'Rourke from July and earlier), so they have been included in the graph. Their line doesn't end, even though their numbers have since dropped, so as not to convey that they have dropped out of the race. Yang's polling average has never reached 3%, but if and when it does, he will be added to the graph, according to the current policy. If you have another idea, you are welcome to share it. My proposal above would add Yang and Klobuchar into the graph (as well as possibly others, I'm not sure). SCC California (talk) 04:47, 5 August 2019 (UTC)
 * Just looked into the detailed graphs by Mélencron... and on about 13 January, Klobuchar was on 1.5%, Gillibrand was on 1.55%, and Castro was on 1.9%. So Gillibrand and Castro would be included too in your proposal. And then, on 15 July, Gabbard was on 1.65%, ahead of Booker on 1.5%... so Gabbard would be included too. So I think those criteria would be too lenient/lax... including Yang, Klobuchar, Gillibrand, Castro, and Gabbard is too much IMO. Chessrat  ( talk, contributions ) 14:31, 6 August 2019 (UTC)
 * Yes. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 14:49, 6 August 2019 (UTC)


 * With regard to the suggestions by SCC California, I support any proposal that provides the readers with additional information (including adding Klobuchar or others). I'd especially like to see the polling data of Andrew Yang and Tulsi Gabbard added, as both of these candidates have surpassed Booker in recent polls. Iff it's undoable to add additional candidates, only then would I prefer that we remove Cory Booker.     Vanilla           Wizard      💙 23:39, 7 August 2019 (UTC)
 * Graphs with many lines at very low levels are simply unreadable; this is why it was decided earlier to only include candidates who have polled at or above 3% for a while. If Yang or others get there, they will be included. If Booker and O'Rourke keep dropping lower, they will be removed. — JFG talk 00:52, 8 August 2019 (UTC)
 * It would be best to avoid the way RealClearPolitics has gone. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 16:30, 8 August 2019 (UTC)

Graph "as of" date has gone missing again
For a brief period of time, the visual graphs were contextualized with what polling date they were up to date with. The graph is being updated more frequently and by more editors now, which is great! But it's still not always up to date with the most recent polls in the table. Could those who are updating the graph please make a consistent practice of including the date when they update? 209.6.126.227 (talk) 21:38, 7 August 2019 (UTC)
 * I believe the standard is to chart the graphs based on the aggregates (RealClearPolitics and 270toWin) rather than the polls. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 20:17, 8 August 2019 (UTC)

Average of polling aggregations
The average row in the table of polling aggregates is extremely unscientific - the two sites likely use some of the same polls and aren't necessarily over the same time period, so the error is unknown which makes all the numbers mostly meaningless (and I'm unsure but would us taking the average to create a new statistic violate No Original Research?). This is a general trend throughout the article that I'm going to work on fixing - without including error (and whether it's likely, registered, etc. voters too really) the candidates' percentages give the reader no meaningful information. Does anyone have a problem with me removing that row from the table? Jmg5vk (talk)
 * Hi - I added that row mainly because it is the source of the data in the graph above it. While I agree that it is imperfect, it does still show the trends of the candidates as well as their approximate polling numbers. Each source is a reliable polling average, so in my opinion averaging the two of them like people have been doing for the chart should only minimize any variations in averaging methods. That said, I would be happy if you have ideas to make the sample information of the polls more helpful. SCC California (talk) 04:24, 9 August 2019 (UTC)
 * I feel this row provides useful information even though it is imperfect. 199.212.18.130 (talk) 15:18, 9 August 2019 (UTC)


 * I do not think that the No Original Research rule is a problem, due to the routine calculation exception. --Spiffy sperry (talk) 15:29, 9 August 2019 (UTC)

Graphical summary update frequency
I just wanted to put into writing a system for updating the graphical summary that has been working. Most of the datapoints should be one week apart on Wednesdays (simply based on when it has been updated in the past). The data for this should come from the average row in the polling aggregation table below the graph. After the most recent Wednesday datapoint, there can be one more recent than that, which should be maintained to the current day from the average row in the table. SCC California (talk) 16:51, 12 August 2019 (UTC)
 * I have updated the hidden note in graphical summary section in the article as well. SCC California (talk) 16:55, 12 August 2019 (UTC)
 * It's good to see you thinking about this; however, if you expect others to contribute to the graph, I'd recommend that you do not make the rules too hard, in case it discourages those people from updating it. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 21:53, 12 August 2019 (UTC)

Frustration with Polling aggregate sources
It's really annoying that RealClear Politics and 270 to Win are incorrectly reporting poll results (and from what I hear, this is not a first). The latest YouGov/Economist Poll result from Aug 10-13 is misrepresented by both groups and their errors are identical: Biden: Actual 21 (Reported 23), Buttigieg: Actual 5 (Reported 6), Yang: Actual 1 (Reported 2). Why is it so hard to correctly report the headline figures? --Mrodowicz (talk) 02:57, 15 August 2019 (UTC)
 * They put the RV results of this poll in their tables. I think they do so to match other polls which only provide RV numbers. Anyway, I am not happy with the way they do polling aggregation. - Ich bin es einfach (talk) 10:11, 15 August 2019 (UTC)

Graphical summary
I want the graphical summary as displayed in this revision of the article https://en.wikipedia.org/w/index.php?title=Nationwide_opinion_polling_for_the_2020_Democratic_Party_presidential_primaries&oldid=910615520 to be in the article. It adds to the information provided by the polling aggregators average graph. Neither is perfect. - Ich bin es einfach (talk) 10:18, 15 August 2019 (UTC)
 * That graph seems very inaccurate. For example, Warren is far below Sanders while they have been polling very similarly recently, and there was a dramatic drop in Warren's line at the far right which isn't reflected in polls. SCC California (talk) 14:49, 15 August 2019 (UTC)
 * The drop mostly stems from the HarrisX polls that aren't included in the poll aggregators. It's also why the graph is different than the other one, it takes into account every poll, which isn't the case for the poll aggregators. It's why I wanted to add the graph, it's more comprehensive. I'm the one who originally made that change (adding the graph). I didn't know that you had to talk about in the talk page beforehand. Sorry about that, I'm relatively new to Wikipedia. P.S. The graph also doesn't include any polls after Aug 11 yet, which is also why there's a discrepancy.MikkelJSmith (talk) 17:20, 15 August 2019 (UTC)
 * So, can it be added or not? MikkelJSmith (talk) 21:14, 16 August 2019 (UTC)
 * I support inclusion of that graph. Chessrat  ( talk, contributions ) 03:13, 17 August 2019 (UTC)
 * Leaning against this graph because it's mostly WP:SYNTHESIS. — JFG talk 11:58, 17 August 2019 (UTC)
 * I oppose the new graph. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 12:40, 17 August 2019 (UTC)

Graphic summary and "aggregation" are a total joke
Despite Sanders leading Warren in 4 of every 5 polls with increasing margins, the graph shows Sanders falling as Warren surges past. The graph has no relationship with reality. 2607:FEA8:A6A0:494:5D77:EB4B:828E:6B7D (talk) 15:20, 17 August 2019 (UTC)
 * If you look more closely at the polls, a lot of them are done by HarrisX. We don't count one pollster over and over again. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 15:40, 17 August 2019 (UTC)
 * In addition, the recent dip in Sander's numbers is due to one recent Fox News poll. This will likely smoothen out over the week. In general, the polling aggregation and the graph based on it will never be perfect, but as an average of two reputable polling averages, it will definitely show candidates' general standing over time and currently. SCC California (talk) 16:54, 17 August 2019 (UTC)

HarrisX
Our list now includes a "new" HarrisX poll every day, each of them being conducted over a rolling 4 days: Shouldn't we keep perhaps just one per week? — JFG talk 17:01, 17 August 2019 (UTC)
 * Yes. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 20:40, 17 August 2019 (UTC)
 * Absolutely. Onetwothreeip (talk) 23:17, 17 August 2019 (UTC)
 * Absolutely not. Our "job" here is to track all available data. - Ich bin es einfach (talk) 23:37, 19 August 2019 (UTC)
 * I agree with Ich bin es einfach. We shouldn't be removing polls for any reason. MikkelJSmith (talk) 01:42, 20 August 2019 (UTC)
 * Those are not separate polls. Per the ScottRasmussen/HarrisX methodology statement: Each day's results are based upon a rolling 3-day average of approximately 1,300 to 1,400 Registered Democrats and Democratic Leaning Independents. We should only list polls conducted over a well-defined sample population and a well-defined polling period. — JFG talk 10:01, 20 August 2019 (UTC)
 * Very well, then. So, if we include the weekly poll we should have their results. So, I agree, it's fine to remove the 3 day average & keep the weekly poll. MikkelJSmith (talk) 15:05, 20 August 2019 (UTC)

270toWin misrepresenting polls
Why is 270toWin's most recent poll the one from Pew Research, listed as Aug 18 but actually conducted Jul 22-Aug 4 (see FiveThirtyEight)? This is shameless cherrypicking.

In my opinion, we should make our own aggregation based on FiveThirtyEight's, weighted by the pollster's rating. If there aren't any major objections, I can do this.
 * Presumably, this is a data entry error. I have reached out to the 270ToWin team and hopefully they will correct it (such things have happened before; in 2016 I noticed a swap between candidates in Huffpost Polling's Republican Nomination aggregation, and after reaching out they had it fixed within the hour). We actually already have our own aggregation, shown on the aggregation tab as "Average". This is the calculation used on the graph. Cookieo131 (talk) 05:00, 20 August 2019 (UTC)
 * OK. I do still think that FiveThirtyEight's pollster ratings should be utilized in some way so that "good" polls get a larger say in what the candidates' averages are. Schnittkease (talk) 18:05, 20 August 2019 (UTC)
 * Here is their response: "Thanks for the note. We list polls as of the date they are released. This is usually within a couple days of the end of when the poll is conducted, although the gap was quite a bit longer here. Separately, Pew seems to have changed the URL where the poll was published; we've updated it with the link you sent.  Thanks for taking the time to let us know. There are a couple new national polls out this morning which have been added to https://www.270towin.com/2020-democratic-nomination/."


 * Indeed, it is true that the data was not released until August 16, so this is internally consistent with their policies. Cookieo131 (talk) 19:17, 20 August 2019 (UTC)

Suggestion: Add an "Others" line to the graphical summary
At the moment, it looks like all 5 of the major candidates (Biden, Sanders, Warren, Harris, and Buttigieg) are simultaneously dropping in the polls, with none of the 7 displayed candidates gaining traction as a result of their decline.

As a reader first and an editor second, this leaves me confused. Their support must have gone somewhere, but there is no graphical summary showing the change over time for the support of the candidates not currently displayed.

This proposed "Others" line can be modified if and when we add more candidates (such as Yang), but it would be greatly appreciated if we could have this information visible on the graphical summary. I'm asking here rather than performing the change myself as I know that the graphical summary has been subject to much discussion already & any modifications should be discussed first.

Cheers,     Vanilla           Wizard      💙 21:20, 20 August 2019 (UTC)
 * I think that's a pretty reasonable idea. SCC California (talk) 22:05, 20 August 2019 (UTC)
 * Here is how it would look with an "others" line since we started having polling aggregators :

SCC California (talk) 01:40, 21 August 2019 (UTC)
 * That looks great, and it especially helps to see how the major candidates' poll numbers declined recently. Thank you for taking the time to make that!     Vanilla           Wizard      💙 03:00, 21 August 2019 (UTC)
 * I have added this line to the article. If anyone has a way to get this data from before July 3, that would be much appreciated. SCC California (talk) 05:49, 21 August 2019 (UTC)
 * Very informative, thanks! — JFG talk 13:37, 21 August 2019 (UTC)

Adding events to the timeline
Looking at the graph, I was thinking that it would be a nice addition to include the various events that has taken place, similar to the graphical overview of the length of all the candidates, as seen on the 2020 Democratic Primary timeline. My suggestion is to add a black linear line that signifies the two debates, and later on lines of other colors signifying the same dates as is found on the timeline. — Preceding unsigned comment added by Fjantelov (talk • contribs) 08:24, 21 August 2019 (UTC)
 * Sounds like a good idea, but I don't see how to implement this easily with this particular graphing module. Also, we would need extra elements in the legend, and that might be confusing. Feel free to try a draft version here, though. — JFG talk 13:40, 21 August 2019 (UTC)
 * I also think this is a good idea, but I don't know how it would be implemented either. SCC California (talk) 18:34, 21 August 2019 (UTC)

Add site to the poll aggregators
I wanted to know if we could at 10at10 (https://goodgawdanotherblog.wordpress.com/10at10-strict-poll-average-2020-democratic-primary-national/) for the poll aggregators? I know the site is a Wordpress site, but the author has been retweeted by polling firms and he has been doing this for a long time.

MikkelJSmith (talk) 18:23, 15 August 2019 (UTC)
 * So, can it be added? Unlike the graph that I wanted to add, it doesn't count HarrisX every time (it only counts it the weekly data like Aug 2-9 & Aug 9-16).MikkelJSmith (talk) 20:09, 17 August 2019 (UTC)
 * I am in favor of it. - Ich bin es einfach (talk) 23:39, 19 August 2019 (UTC)
 * I disagree. It is not, in my opinion, a very reputable source, unlike RealClearPolitics and 270 to Win, which are frequently cited by other WP:RSs. This 10at10 doesn't even come up when I search 10at10 in Google. SCC California (talk) 03:51, 20 August 2019 (UTC)
 * He's changed the name of his average so I guess that's why it doesn't show up on Google. If you search for Doug Johnson on Twitter, you can see his model on the pinned tweet. It's updated every week on Twitter (with delegate estimates) and every day on the site.MikkelJSmith (talk) 03:58, 20 August 2019 (UTC)
 * This new aggregator excludes certain low-polling candidates. It is apparent that you are interpreting this a zero for purposes of calculating the average.  This is not correct. --Spiffy sperry (talk) 17:03, 20 August 2019 (UTC)
 * I will revert this addition until this issue is sorted out. We cannot assume data. SCC California (talk) 17:14, 20 August 2019 (UTC)
 * I think the current version of the page solves this issue. It only calculates the average of the Others & Undecided for the 3 aggregators where we have all the data (RCP, 270toWin & the Economist) & ignores the 2 aggregators where we don't have all the data for those two categories(NYT & 10at10).JonathanScotty (talk) 00:56, 22 August 2019 (UTC)
 * It actually does solve the problem. This new version is perfect. What I did previously (assuming the data for certain candidates) was stupid. Thank you to the person that gave us this solution. MikkelJSmith (talk) 01:00, 22 August 2019 (UTC)
 * So, that means from now on, when doing the average for all the candidates polling below 3% on average (those that don't have their own column in the poll aggregators section) we don't use 10at10 or the NYT (unless they give us all the data). MikkelJSmith (talk) 01:58, 22 August 2019 (UTC)
 * Can we add everyone who shows up on the graph (ie. O'Rourke and Booker as of now) as proper columns on the table? Also, there has to be some formula or function we can agree on to calculate aggregation. Shahanshah26 (talk) —Preceding undated comment added 14:14, 20 August 2019 (UTC)

RCP still having Inslee
So, when removing Inslee's numbers earlier I didn't realize that RCP kept him, the others (270towin & theEconomist) have removed him, since he dropped out. Should I put Inslee back for the RCP section or is that for nothing since they will probably remove him soon (as in a few days)? MikkelJSmith (talk) 01:06, 23 August 2019 (UTC)

Yet more misrepresentation from RCP & 270
If it's not enough that RCP & 270 to Win got the figures wrong for last weeks Economist/YouGov poll (see my comments above at Frustration with Polling aggregate sources), they've both done it again for the Economist/YouGov poll of this week. This time Warren's numbers are incorrectly reported as 18 (actually 17) and Buttigieg as 8 (actually 7). I think we should keep track of these errors and use only the most reliable Polling Average data sites. Not sure how good 10at10 are overall, but they actually reported the Economist/YouGov results correctly on both occasions. --Mrodowicz (talk) 08:22, 23 August 2019 (UTC)

Need to Remove Economist from Aggregate
This article is misrepresenting the Aggregate data because RCP already included the Economist in their aggregate. therefore, the economist results are being weighted twice. Isaacquelly (talk) 23:19, 26 August 2019 (UTC)
 * The Economist has both its own polls, and an aggregation of various polls (including its own). The Economist's aggregation is not used by RCP. --Spiffy sperry (talk) 14:18, 27 August 2019 (UTC)

Anonymous - I think we should remove The Economist anyway, since they do not report dates or details on which polls are included. NYT should also be removed since they are incredibly slow to update it. — Preceding unsigned comment added by 207.107.70.194 (talk) 14:33, 27 August 2019 (UTC)

I agree. NYT has Warren 4 points ahead of Sanders when she is trailing him in the last 11 polls. Schnittkease (talk) 23:17, 27 August 2019 (UTC)
 * Please look at the pollster &mdash; HarrisX does many polls but they count as one. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 23:33, 27 August 2019 (UTC)

What does HarrisX have to do with the NYT aggregation? It isn't even included. Anyway, the larger point is that NYT hasn't been updated since 23 August. How can this site consider it any acceptable addition when it isn't a live tracker that updates daily? That is useless for those who want up-to-date information on the race. — Preceding unsigned comment added by 174.115.0.59 (talk) 01:54, 28 August 2019 (UTC)

OP here, i have no clue how to implement the change. someone else should do it. i only just created my account. lol Isaacquelly (talk) 01:27, 29 August 2019 (UTC)

Suggestion: addition of an "Undecided" line in the polling aggregation graph
Should an Undeciced line be included in the polling aggregation graph?

I think that would be relevant information to include, as people who don't vote constitute around 10% of the voter base on average, which is more than some of the candidates listed in the graph currently.

WittyRecluse (talk) 06:32, 29 August 2019 (UTC)

Suggestion: Add a line for Yang to the graph
At this point, Yang consistently matches or outperforms Booker, who has a line in the graph; thus not including Yang is no longer logical. Also, Yang is listed separately in the polling results table starting with the August section, so this would only be a consistent step. Silver Scroll (talk) 15:06, 29 August 2019 (UTC)
 * Earlier discussions said that if a candidate consistently polls around 3% or higher they'll be included. Yang hasn't consistently reached that mark but looking at the upward trend it seems likely he will be added. Catiline52 (talk) 01:27, 30 August 2019 (UTC)

Inclusion of "10 at 10"
Why is there an un-notable WordPress blog in the poll aggregations? Searching it comes up with no news reports or anything. Surely we'd be using aggregators that come from reputable major sources? Catiline52 (talk) 01:17, 30 August 2019 (UTC)
 * Agreed. There seems to reason to include this blog as a source. SCC California (talk) 17:16, 31 August 2019 (UTC)

How about a column for DNC-qualifying?
Thoughts? Humanengr (talk) 03:10, 31 August 2019 (UTC)

I got that started. Anyone care to help? Humanengr (talk) 02:51, 1 September 2019 (UTC)


 * Oops — hadn’t noticed the legend. Humanengr (talk) 14:52, 1 September 2019 (UTC)

Fav ratings- Inslee
I saw that Inslee (who dropped out) is still listed in the favorability section. I'm wondering if someone could please remove him, as well as any candidates who may drop out in the future. I would do it myself, but I don't want to mess up any formatting by mistake. Benwitt (talk) 20:08, 23 August 2019 (UTC)
 * Why would dropped candidates be removed? From an encyclopedic perspective, once candidates become notable to the election, shouldn't they remain notable even if they end their candidacy? Otherwise by Summer of next year this section will have only one candidate in it. This is an encyclopedia article about the primary as a whole, not a current events summarizer. 209.6.126.227 (talk) 05:44, 7 September 2019 (UTC)

--

Significant figures in polling average
The New York Times' polling aggregation only uses two significant figures for higher-polling candidates and one for the rest, while the rest of our sources use three for higher-polling candidates and two for the rest. Our average has been "making up" an extra figure of .0 on the New York Times numbers. I guess our two options are to either not include that source in the average or to reduce all of the significant figures in the average. I have gone ahead and done the first one for now (so we don't lose the precise data from the rest), but what do others think? SCC California (talk) 18:40, 6 September 2019 (UTC)
 * The same issue appears to be there with the Economist. SCC California (talk) 17:36, 7 September 2019 (UTC)
 * Not true of the Economist - all you have to do is click on the individual candidates to get their average to 3 sig figs. 172.251.70.104 (talk) 23:44, 7 September 2019 (UTC)
 * True, sorry. The NYT issue still stands. SCC California (talk) 02:58, 9 September 2019 (UTC)

abc/post- rvs
Hello, I just wanted to note that I added the info for RVs instead of adults for the ABC News/Washington Post poll. Both outlets today (Sep 8) based their stories on reg voter preferences as opposed to adults in general. Most aggregators (RCP, etc) show only voter prefs. --Benwitt (talk) 14:01, 8 September 2019 (UTC)Benwitt (talk) 14:00 8 September 2019--Benwitt
 * The DNC is using the "all adults" criterion; see . "Shortly after publication, the DNC told POLITICO that the sample with all adults will count toward qualification." David O. Johnson (talk) 04:16, 9 September 2019 (UTC)

Adding Steyer and Gabbard
Should Steyer and Gabbard be added? Steyer just qualified for the 4th debate and Gabbard has qualified on three polls now, i.e. both are getting 2% or higher on some polls. --24.101.156.239 (talk) 19:30, 8 September 2019 (UTC) Sockpuppet of permabanned troll A Nobody. <b style="color: Maroon;">Reyk</b> <b style="color: Blue;">YO!</b> 07:07, 9 September 2019 (UTC)
 * Gabbard didn't get a third poll that actually counts; among all adults, she only got 1% (and that's the one that counts); see ; "Shortly after publication, the DNC told POLITICO that the sample with all adults will count toward qualification." Steyer did cross the threshold, though. David O. Johnson (talk) 04:18, 9 September 2019 (UTC)

Democracy Corps / GQR Research
538.com made a mistake in their listing of the Democracy Corps / GQR Research Sep 7-11, 2019 poll. They list this poll as "likely voters" (LV) result but display the numbers for the "registered voters" (RV) result. The result for "likely voters" excludes from the polling population those voters who said they will probably not vote. E.g. Registered voters = 800. Will probably not vote = 25. Likely voters = 800 - 25 = 775. Using likely voters as polling population makes more sense than using registered voters because in an election the population used for calculating percentages is also excluding those people who did not vote. Therefore we should list the percentages based on population = likely voters, see page 11 of the poll data. Xenagoras (talk) 21:43, 16 September 2019 (UTC)

Yes, that makes sense. Thanks for clarifying. Benwitt (talk) 23:16, 16 September 2019 (UTC)

We still don't know the sample size of the Dem LV group. Benwitt (talk) 23:46, 16 September 2019 (UTC)

Yang & the graph
Should Yang be added to the graph? He's getting 3% or close to 3% and he's been higher than Booker and Beto for a while now. JonathanScotty (talk) 19:41, 8 September 2019 (UTC)
 * As per previous discussions, he will be added as soon as his average hits 3%. SCC California (talk) 03:01, 9 September 2019 (UTC)
 * He's now in. — JFG talk 03:00, 19 September 2019 (UTC)

HarrisX
Didn't we agree to only keep weekly HarrisX polls (I'm not talking about variants like The Hill/HarrisX or Harvard/Harris)? JonathanScotty (talk) 02:12, 11 September 2019 (UTC)
 * This was discussed twice already here and there and it seems the consensus was to not include HarrisX polls with overlapping time frames, but only one poll per week. Currently there are overlapping and small-sized polls from HarrisX in tables of several months, but some weekly HarrisX polls are missing. Xenagoras (talk) 00:03, 17 September 2019 (UTC)
 * Agree with prior consensus: keep just their biggest weekly poll. No time to edit myself though. — JFG talk 03:01, 19 September 2019 (UTC)

Colors on the graph
Have they been changed again? --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 12:30, 21 September 2019 (UTC)
 * I think Booker's color was adjusted to accommodate/contrast with Yang's. Chessrat  ( talk, contributions ) 21:53, 21 September 2019 (UTC)
 * Contrast looks good to me now. — JFG talk 22:14, 21 September 2019 (UTC)
 * Okay, thanks. Hopefully they can stay like this for a while! --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 16:39, 22 September 2019 (UTC)

Vertical stacked bar chart proposal
I like this type of chart as it allows to estimate which candidates share the same voter pool and how these pools grow and shrink over time. The data comes from Morning Consult which polls 17,000 registered voters weekly and therefore achieves a much smaller margin of error than all other polls. I would like to insert this chart below the existing line chart with the leading text: ''The following graph displays the results from national polls conducted weekly by Morning Consult. These polls have a margin of error of ± 1.0%. '' Xenagoras (talk) 23:55, 17 September 2019 (UTC)
 * I like this, this is very interesting. The only question I have is why specifically Morning Consult? Couldn't we do a chart like this, but with the poll aggregation averages? WittyRecluse (talk) 05:27, 18 September 2019 (UTC)
 * Thank you WittyRecluse. We could do this with the poll aggregation average. Morning Consult polls approximately 17000 registered voters weekly which is 38 times as many as the 450 voters the other organizations poll on average. This means Morning Consult polls more voters than all other pollsters combined. (Only the weekly HarrisX polls come near with 3000 voters polled on average.) Morning Consult's very high number of voters polled results in a much higher precision (measured by a much smaller margin of error) than all other polls. This precision results in a smaller fluctuation of results than even the polling averages have. Morning Consult publishes the full data set since January 13 on their website which makes creation and adaption of the chart easy for everybody at any time. I would also like to have a graph displaying the results of early primary states (which is contained in the Morning Consult data set). Xenagoras (talk) 07:05, 18 September 2019 (UTC)
 * Early primary states sound like a good idea, and I agree that Morning Consult is very precise, but you can see why I would be concerned about bias if we only use one source. WittyRecluse (talk) 18:27, 18 September 2019 (UTC)
 * Conversation continues at — JFG talk 22:07, 21 September 2019 (UTC)


 * I'm against inclusion. Does not look as informative as the current, simple line graph. — JFG talk 02:59, 19 September 2019 (UTC)
 * I don't think this would be useful for the page, though I appreciate the work that's gone into assembling it. Stacked bar charts are sometimes useful for conveying meaning, but I can't see what it would add in this case. The rationale of showing which candidates share the same voter pool I think more clearly applies to the existing line graph. On the existing line graph, I can comfortably interpret that a lot of Biden and Sanders voters switched to Harris in early July. That's not as easy to do in the stacked bar chart. Ralbegen (talk) 09:53, 19 September 2019 (UTC)
 * Ralbegen, in the Morning Consult national data, there is no voter move from Sanders to Harris in early July. Sanders has been constant since begin of May. The voter migrations I can recognize in the data are: mid February until mid March: "others" -> bigger candidates. Warren + O'Rourke = pool. April: Biden -> Buttigieg. Late June: Biden -> Harris. Since late July: Harris -> Warren. Conservative pool = Biden + Buttigieg + Harris + Yang + Booker. This pool grew during January - April and shrinks since August.  Xenagoras (talk) 00:18, 21 September 2019 (UTC)
 * Actually, I agree with Xenagoras here (on the information provided by the secondary format). I didn't realize that Warren was getting her July-September upswing votes from Harris, until I saw the second graph. WittyRecluse (talk) 07:14, 21 September 2019 (UTC)
 * Neither the stacked graph nor the line graph can accurately convey which voters have moved from a candidate to another (and polls do not usually ask this question). The best we can see on the line graph is that it reflected the Biden/Harris confrontation during the first debate. For the rest of jockeying, each reader can interpret the graphs as they see fit. — JFG talk 22:03, 21 September 2019 (UTC)

Graphs for early primary states?
How about this chart for the early primary states? Data source: Morning Consult.

Same data as line chart:

As can be seen in the direct comparison of vertical stacked bar chart and line chart, the bar chart has the capability to display much more information without getting overcrowded. Xenagoras (talk) 23:53, 20 September 2019 (UTC)
 * Too crowded with candidates polling below 2%. Perhaps worth adding with fewer people. Also, perhaps better suited for the state polling article. — JFG talk 22:13, 21 September 2019 (UTC)
 * The bar and line chart for early primary states only contain candidates that currently poll at 3% or more, with the exception of Klobuchar which I included because she had a high of 4% in the past. I will add a bar chart without Klobuchar to the the state polling article. Xenagoras (talk) 22:33, 21 September 2019 (UTC)
 * As much as I love bright colors, that stacked bar graph makes my eyes bleed. Would anyone mind if I make them a little softer?


 * How about these colors?     Vanilla           Wizard      💙 22:55, 23 September 2019 (UTC)
 * Thank's for your proposal. I reduced the brightness contrast between neighbouring candidates' colors. Xenagoras (talk) 16:52, 27 September 2019 (UTC)

10at10
I think we should remove 10at10 from the polling aggregation section, not only for it not being notable, but also because it has recently been using every day of the 3-day HarrisX averages separately, essentially triple-counting the same information (this is the same reason we removed many HarrisX polls from the Wikipedia table). SCC California (talk) 01:07, 24 September 2019 (UTC)
 * Agreed. <b style="color: #AB2B2B;">{ [ ( jjj</b> <b style="color: #000000;">1238 ) ] }</b> 01:11, 24 September 2019 (UTC)
 * They haven't been using every HarrisX polls separately, only the most recent poll (only the polls in the colored lines). But agreed that 10at10 is not notable. I think it's still a good source for finding lesser known polls. Seismologist76 (talk) 20:29, 24 September 2019 (UTC)
 * Yeah, they haven't been using HarrisX separately, just the most recent one, so I think it's still a good resource. It also helps us find lesser known polls, since it's the only average not to exclude any. As for it not being notable, I've seen a pollster retweet 10at10 on Twitter and pollsters talking to the person behind 10at10, just beause it's a wordpress site doesn't mean it's a bad one. I'm guessing it should be added back now that there isn't consensus? — Preceding unsigned comment added by JonathanScotty (talk • contribs) 19:22, 28 September 2019 (UTC)
 * I'm kind of meh. I could see us keeping 10at10, since they aren't representing HarrisX separately, so there is no issue with that. Although, I don't really mind that it's been removed. So, I guess the page can stay how it is. MikkelJSmith (talk) 19:33, 28 September 2019 (UTC)
 * "just beause it's a wordpress site doesn't mean it's a bad one" WP:RSSELF sources are generally not accepted as a reliable source. Anyone can make a blog about a topic. Catiline52 (talk) 23:29, 28 September 2019 (UTC)

Polling aggregation: RealClearPolitics
From where do you get the RCP polling numbers for all candidates? I only see the top 13 candidates listed on their website but not the lower polling candidates. Xenagoras (talk) 14:50, 30 September 2019 (UTC)
 * There is a graph underneath the table. The graph legend includes the numbers for the lower polling candidates. --Spiffy sperry (talk) 15:47, 30 September 2019 (UTC)
 * Thank's. Xenagoras (talk) 17:19, 30 September 2019 (UTC)

"Raycroft Research"
This is so obviously fake that I can't believe that you dolts are falling for this. Literally just a blank website and single twitter account, both of which were registered in August and untraceable to any individuals, and the only actual links to anything are Typeforms and Facebook/Twitter surveys. Absolute joke. Unweighted. Absolute dolts. 73.145.173.55 (talk) 00:07, 5 October 2019 (UTC)
 * I understand that you don't think this poll is reliable, but "fake" is not a justifiable reason to remove content. Please be WP:CIVIL when interacting with other editors who are just trying to follow the rules. If you have information suggesting that this particular poll is unreliable, please give that reason in your edit summary or on the talk page. That said, it was not labeled as a qualifying poll (because it's not) and there is no reason at this time to infer that the poll was conducted in defiance of usual polling standards. WMSR (talk) 01:11, 5 October 2019 (UTC)
 * I do not care either way whether the poll is included. I came across it and so I added it. However, your analysis is misleading. The website (which also is not a "blank website") says the firm was founded in 2017 and simply began undertaking published polling in Q3 of 2019. As WMSR said, please remain civil when attempting to discuss changes you wish to make, or you risk ruining your entire argument. <b style="color: #AB2B2B;">{ [ ( jjj</b> <b style="color: #000000;">1238 ) ] }</b> 01:24, 5 October 2019 (UTC)

Others line in graphical summary
As someone who originally was in favor of adding the others line, I have found that it has not improved my understanding of the race and instead seems only to add clutter. Thus, I propose removing it. Are others (no pun intended, I promise) on board? SCC California (talk) 00:29, 17 September 2019 (UTC)
 * I am; it is unfortunate we do not have anything between January and July. WittyRecluse (talk) 05:09, 17 September 2019 (UTC)
 * I agree. --Spiffy sperry (talk) 15:18, 17 September 2019 (UTC)
 * I think we should remove the line ending in January but keep the one from July onward and keep it updated. The line is useful information for reference. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 20:35, 17 September 2019 (UTC)
 * I'd support restoring the "other" line since July. It's an informative explanation of "where did the votes go" when several top candidates see their support drop in unison. — JFG talk 03:03, 19 September 2019 (UTC)
 * I've added a combined line for "others" and "undecided" from the beginning, in a light grey and in the background to de-emphasize it. I realize that combining these two stats isn't necessarily the most accurate thing to do, but IMO it provides a good analogue for how much the race is/isn't converging, and looks nicer than the spontaneous "others" line in July. If anyone feels strongly opposed or would rather discuss it more first, feel free to revert, of course. —Tga (talk) 00:28, 3 October 2019 (UTC)
 * I appreciate that it is light gray and in the background, but I must say that I don't really see the point in the line at all. The real undecided numbers are far higher than reported in these polls as people usually answer their favorite candidate even when not certain of who they will vote for. SCC California (talk) 22:12, 3 October 2019 (UTC)
 * Like I said, I think it provides a decent analogue for how much the race is converging. Even if people aren't actually "deciding" on who they're voting for (which should be obvious, given how every line has up and down swings), the general gist is that as the grey line goes down, the race is closer to converging on the set of candidates the primaries/caucuses will actually be between. When it swings up, it means that there is a greater degree of uncertainty in this set. —Tga (talk) 22:54, 3 October 2019 (UTC)
 * I agree. User:PutItOnAMap (PutItOnAMap) 23:33, 14 October 2019 (UTC)

Graphical Summary - continued from archive
While I am happy to stick to the consensus of using aggregates' averages for our graph, I would point out that creating our own on the basis of all polls would not necessarily constitute WP:SYNTHESIS if we averaged polls according to a reliable method cited from 538. If we used their means to analyse polls instead of averaging aggregators' results, which have shared (but not equal) subsets of the same polls, I feel our results, overall, could reflect the reality of the primary polling a lot more accurately.

I would consequently support inclusion of a graph based on the aggregation of all polls according to a tried and tested method, which might look something like this - could we do better to replace our current graph with this, or add both? PutItOnAMap (talk), 16:56, 14 October 2019 (UTC)
 * Performing the average seems SYNTHy because it involves picking the polls according to some threshhold. Polling is not non-partisan. PunxtawneyPickle (talk) 23:58, 14 October 2019 (UTC)
 * It would be if we arbitrated according only to internally defined characteristics, which is why I advocated using 538 ratings and the elements (in the public domain) of the model used for previous primaries (primarily, LOESS regression). While there is no undisputed gold standard of aggregation, there is a case to be made for adhering to one of several firm heuristics that can be cited as in use elsewhere. We could alternatively use 270towin's method of simply picking the last six polls from different sponsors within the last 30 days if we wanted to stick to previously cited sources, or use one of the other models in the public domain. I appreciate there is room for synthesising, but we do that to an extent by aggregating aggregators and thus weighting some polls over others (and, indeed, ignoring a third set completely) according to no reliable heuristic. While I'm happy to accept continuing with that, I do lean slightly towards trusting this alternative solution more. If pollster rating weights were unhelpful but the loess regression-based model still brought benefit, we could cut them out from the final model.PutItOnAMap (talk), 1:25, 15 October 2019 (UTC)

Amy Klobuchar's line in polling aggregation
I cannot find the average polling so I can include Amy Klobuchar in the polling aggregation graph. The criteria is to include a candidate is "candidates whose polling averages have been at or above 3% at any time." And from what I can see from RealClear Politics, The Economist and 270 to Win, Klobuchar reached over 3% in February and March 2019. Correct me if I am wrong and help me find the statistics if I am right. LoveCervin (talk) 21:58, 11 October 2019 (UTC)
 * In the past we used the graphs of User:Mélencron, and those graphs only showed Klobuchar reaching a maximum polling average of 2.7% or so, so we didn't include her. But now we're using those polling aggregator sites... and you're right, they do seem to show Klobuchar at over 3% in Feb-Mar. I'll sort out adding her line if nobody else does. Chessrat  ( talk, contributions ) 00:47, 12 October 2019 (UTC)
 * Yes, and thanks to the person who added them &mdash; I'm guessing you did? --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 17:00, 14 October 2019 (UTC)
 * I did, but only thanks to this page bringing it to my attention. PutItOnAMap (talk) 19:39, 14 October 2019 (UTC)

I appreciate the effort to add her line, but the graph begins to get too crowded at the bottom. I'd suggest tightening the criteria to keep only people who are currently above 2%, so that Klobuchar and Booker would go. — JFG talk 00:55, 17 October 2019 (UTC)
 * Wouldn't that risk removing some candidates who had more meaningful levels of support previously from a graph meant to show progress over time? E.g. Beto O'Rourke assuming his downward trend continues, or candidates who drop out. It's certainly a case for removing them from the aggregators, but while I don't personally believe the graph is too crowded, I'd have thought a better way to go about clearing it would be to just move the overall threshold required to have a spot on it from 3% to e.g. 5% (which would knock off Booker, Klobuchar and Yang). - PutItOnAMap (talk) 12:07, 18 October 2019 (UTC)

October polls
Cory Booker has not reached the threshold for us to include a column for him in the October polls, right? Shouldn't we remove that column and relegate his October poll results to footnotes? SecretName101 (talk) 17:40, 8 October 2019 (UTC)
 * I'd say we wait a week or so to see if the trend continues, and remove Booker then if it does. Chessrat  ( talk, contributions ) 13:47, 10 October 2019 (UTC)
 * What is the threshold for inclusion in the aggregates, as opposed to the graph? >PutItOnAMap (talk), 16:27, 14 October 2019 (UTC)
 * Not sure if there was general agreement on inclusion in the aggregates. I'd suggest being at 3% or close to that. That's the same as the threshold to have an individual column in the poll tables. — JFG talk 00:50, 17 October 2019 (UTC)
 * I would definitely remove Booker and Klobuchar who are both presently way under 2% in all tracked aggregators. — JFG talk 00:51, 17 October 2019 (UTC)
 * I believe that any candidates on the graph (who are still running (not currently a relevant distinction)) should be included in the aggregates table below it. SCC California (talk) 02:46, 17 October 2019 (UTC)


 * I have removed Booker's column for October, as he has not reached 3% in any of the last 15 polls. — JFG talk 01:04, 17 October 2019 (UTC)
 * Just to address SCC California's point - I wonder if we'd do better, in general, to attempt different things with the graph and aggregator - specifically, to include candidates in the aggregator on the basis of some threshold defined by recent poll results, and to include candidates in the graph based on historically reaching some level of support (possibly higher than that needed for a place in the aggregator). PutItOnAMap (talk), 12:11, 18 October 2019 (UTC)

Did I miss something?
I haven't been on this page in a while, since it's pretty much a machine that (I feel) barely needs my input. I still look at the edits when I can though. But, I was wondering, why was the NYT removed from aggregation? I'm not against it, just curious. Is it due to the lack of sig figs? JonathanScotty (talk) —Preceding undated comment added 14:40, 19 October 2019 (UTC)
 * See Archive 2 - it was removed because it only displayed averages to two significant figures, while our other aggregators displayed them to three. The debate there was over whether to reduce the overall average to 2 sig. figs or remove the NYT aggregator, but I wouldn't be against averaging the NYT's 2 sig. fig. result with the other aggregators' 3 sig. fig. results. PutItOnAMap (talk), 20:56, 19 October 2019 (UTC)
 * Another factor brought up in the archives (and which I agree with) is that NYT updated the numbers less often than the others, so their data was usually out of date. --Spiffy sperry (talk) 15:11, 21 October 2019 (UTC)


 * On another issue, I don't see why we present an average of poll aggregates. It's OR because it implies that the three aggregates are of equal value, which is a judgment. TFD (talk) 05:36, 24 October 2019 (UTC)
 * I agree. WMSR (talk) 15:15, 24 October 2019 (UTC)
 * It's hardly OR if it's a simple average - just a clean presentation of the pulse aggregators claim to have on the state of the polling. If it blurs any rules, it's SYNTH -- see the above discussion about the Graphical Summary for my opinion on that. If there is still significant dissent here, perhaps the best solution is to use both? PutItOnAMap (talk), 11:30, 28 October 2019 (UTC)

Inconsistencies in Ipsos/Reuters Nov 1-4 poll
The Nov 1-4 Ipsos/Reuters poll has some weird numbers in it that I cannot understand. Could anyone please explain them to me? 1115 were interviewed. 450 Democrat + 429 Republican + 139 Independent = 1018 voters. Why is the sum of Dem+Reps+Indies not equal 1115? Where do the other 1115-1018=97=8.7% voters come from? There is no "third" party with 8.7% voter share. Why are Independents only 139/1115=12.5% of voters although in all other polls Independents are the largest block with about 40%? Why is the "All voters" sum in the candidate list 686 although there are 450 Dems + 139 Indies = 589? Where do the 686-589=97 voters come from? Did Ipsos add the 8.7% "third" party voters to the Dems+Indies and why? I thought Libertarians and Greens etc are not eligible to vote in Dem primaries? Xenagoras (talk) 00:33, 6 November 2019 (UTC)
 * Some polls define "independents" differently: most independents lean toward one party or another. A lot of polls treat "independent leaners" as partisans because they usually end up voting like partisans. About 40% of voters are self-described independents, but only about 10% are "true" independents without a partisan preference, so this is pretty typical. It looks like about 10% of the sample gave "other/don't know" responses on the partisanship question, which probably accounts for the fourth group. Some of the number differences may be related to weighting, but I don't see anything unusual about any of that.
 * I would add that it is usually a waste of time to worry about the partisan breakdown of a survey sample. Party ID is not a demographic variable, it is an attitudinal one. It shouldn't be the same in every survey. Nblund talk 00:04, 7 November 2019 (UTC)

Where is the source for the polls chart?
How can one verify it? Guarapiranga (talk) 04:10, 15 November 2019 (UTC)
 * there is a separate polling page. Each and every poll has a reference there. They do a brilliant job, BTW.Arglebargle79 (talk) 14:13, 17 November 2019 (UTC)
 * So… where is the page? Where can one verify that brilliant job, ? Guarapiranga (talk) 19:57, 17 November 2019 (UTC)
 * The references on the agrigation of polls that make up the graph are referenced right there on the polls page. Arglebargle79 (talk) 21:16, 17 November 2019 (UTC)
 * Yes, I looked at that,, but 2 of the 3 cited references—270 to Win and The Economist—only show the last average, not past ones. The only one that shows past averages is RealClear Politics, and only by reading directly off the chart there. Guarapiranga (talk) 21:54, 17 November 2019 (UTC)
 * Look at the history of the aggregator websites at the Wayback Machine. Xenagoras (talk) 13:13, 18 November 2019 (UTC)
 * I'm sorry, but that's all I got. Arglebargle79 (talk) 22:26, 17 November 2019 (UTC)

Article is no longer being updated
For four days now. What gives? — Preceding unsigned comment added by 174.92.111.57 (talk) 16:11, 17 November 2019 (UTC)
 * There have been some small updates, but no polls have been added in a few days because none have been released. SCC California (talk) 17:38, 17 November 2019 (UTC)
 * And why, then, have none been released for most of a week now? — Preceding unsigned comment added by 174.92.111.57 (talk) 16:26, 18 November 2019 (UTC)
 * It's not uncommon for there to be patches of several days where no polls come out. There has been plenty of statewide polling recently, for what it's worth. Jacoby531 (talk) 16:35, 18 November 2019 (UTC)

Starting graph in January 2019
The summary graph currently starts in December 2018. Given that most candidates entered the race in January and February 2019, I would suggest trimming the lone 2018 month, and starting the graph on January 1st, 2019. What do y'all think? — JFG talk 19:15, 26 November 2019 (UTC)
 * Personally I think the December 2018 month is useful for showing O'Rourke's past popularity, so I support retaining it. Chessrat  ( talk, contributions ) 20:57, 26 November 2019 (UTC)
 * , I also support retaining it. MikkelJSmith (talk) 21:54, 26 November 2019 (UTC)
 * O'Rourke's had name recognition in late 2018 on the heels of his Senate bid; I don't think it's very relevant to the presidential race. Now that he has withdrawn from this race, his 2018 popularity polls are not reason enough to keep December 2018 on the chart. — JFG talk 05:13, 27 November 2019 (UTC)
 * I think it's somewhat relevant in that it goes some way to explaining his decision to run, amongst other decisions to run or not run made during this primary (there have been several waverers throughout). Where I think it adds the most value, though, is in showing the state of the field before the 'announcement bumps' caused by the deluge of people announcing their entry into the race was. Granted, this part of the graph isn't the most important, but I'd be in favour of keeping it. User:PutItOnAMap (talk) 16:54, 27 November 2019 (UTC)

Mistake in polling average
Pete Buttigieg's aggregates all place him betweeen 11 and 12, but the average is shown as 12. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 12:16, 6 December 2019 (UTC)
 * --Spiffy sperry (talk) 14:31, 6 December 2019 (UTC)

Too many mentions of minor candidates in the tables
The tables showing the history of polls now contain yellow-highlighted lines for every candidate's entry and exit from the race. This looks too heavy, and is obscuring the most relevant events to polling. Earlier in this page history, we only highlighted those candidates who polled high enough to get their own column in any table. I would suggest reverting to this minimalist approach. — JFG talk 12:31, 23 November 2019 (UTC)
 * I support this change and agree with the point on how it obscures relevant events to polling. WittyRecluse (talk) 13:03, 24 November 2019 (UTC)
 * I have trimmed the announcements. — JFG talk 18:50, 24 November 2019 (UTC)
 * I don't support this change and would be in favour of reversing it. The polls are to some extent a horse race, and thus information regarding when candidates decided to retire from the field is useful and worth conveying in this format. — PutItOnAMap talk 1:30, 26 November 2019 (UTC)
 * I'm of the same opinion - MikkelJSmith (talk) 03:01, 26 November 2019 (UTC)
 * I think that the only announcements that should be in the polls are for candidates who are in the aggregate graph or the aggregate table, or would have been in the aggregate graph or aggregate table. In other words, only those who have ever polled at 3% are significant enough by our own standards to affect polling, and therefor it is self contradictory to include announcements from those who are under 3%. WittyRecluse (talk) 14:41, 26 November 2019 (UTC)
 * Right. The best practice is to mention entries and withdrawals of candidates who have a column in the relevant historical table. People who came and went without ever reaching 3% polling do not need to be highlighted (e.g. Gillibrand). Also, people who entered long ago (e.g. Yang and Gabbard) but only reached 3%+ recently, do not need to be highlighted in the early tables because they did not have a significant impact at the time they first announced their candidacy. — JFG talk 17:36, 26 November 2019 (UTC)
 * , oh, I see that argument is persuasive. Yeah, nevermind, I'm fine with the change. - MikkelJSmith (talk) 17:39, 26 November 2019 (UTC)
 * Having seen the update here since last time, we are further minimising the pre-existing rule. I'll go with the user consensus, but personally I still think this is unwise. Considering that unsure voters are not always distributed evenly (e.g. a subset of voters unsure whether to pick one of two similar candidates, but certain not to pick the third), candidates polling less than 3% can still distort other candidates' figures by making more voters who'd otherwise be sure to vote for a specific candidate undecided. Therefore, I think it worthwhile to at least include candidates who have featured in at lesat one month's table, and would prefer to include all candidates otherwise considered major.PutItOnAMap (talk) 17:42, 26 November 2019 (UTC) 18:12, 26 November 2019 (UTC)
 * Further to this, I think we should add the entry of candidates like Yang (i.e. those who made it to the aggregate table or graph but did not do so at the time of their entry into the race) back into the table and feel a fair bit more certain that this would be the correct move. I'd argue that charting the rise of candidates who essentially started nowhere is valuable, and demonstrating the relative frequency of that is also valuable because the primary has been filled with long-shot candidates insisting that they can win from a very low base - showing how often they did or didn't get halfway to achieving that is informative. PutItOnAMap (talk) 16:40, 6 December 2019 (UTC)

Adding Bloomberg numbers from February/March
I am having trouble inserting Feb/March numbers in the aggregate poll for Bloomberg. The numbers are taken from RealClearPolitics: 2/8: 4.2 2/15: 4.5 2/22: 3.3 3/1: 2.5 3/8: 2.5 3/16: 2.0 The graph does not render properly when I insert these numbers. I'm not sure why the graph is not rendering.—<b style="color:red;background-color:#000">SP E SH</b> 531 Other  15:51, 25 November 2019 (UTC)
 * I've added the early polling data, but I'm not sure it's relevant, as he was not a candidate then. What do other editors think? — JFG talk 19:17, 25 November 2019 (UTC)
 * I lean in favor of including Bloomberg's polls from February and March. After all, we also include polling for candidates like Biden and Sanders from well before they announced. Jacoby531 (talk) 19:20, 25 November 2019 (UTC)
 * , I agree. I have the same opinion on that. MikkelJSmith (talk) 00:45, 26 November 2019 (UTC)
 * This seems like a sensible idea. I will root around (if I have time) for earlier polling data re: Bloomberg. PutItOnAMap (talk) 01:31, 26 November 2019 (UTC)
 * For what it's worth, Bloomberg was formerly included in the "Other" line by default. So, if his earlier (prior to Nov.) numbers are added, the Other line should be adjusted accordingly. --Spiffy sperry (talk) 16:13, 26 November 2019 (UTC)
 * Just added his old numbers and saw this. Will do. PutItOnAMap (talk) 17:42, 26 November 2019 (UTC)
 * ✅ — JFG talk 18:08, 26 November 2019 (UTC)

I don't think that this is a good idea because it implied that Bloomberg had 0% support from March to November, when in reality he simply wasn't included in polls (because he said that he wasn't running). If he had been included in polls, he might have polled higher, but it doesn't matter because he wasn't in the race. Similarly, Buttigieg wasn't included in polls before January and thus isn't on the chart (instead of assuming 0% from December). Accordingly, I have reverted the change for now. SCC California (talk) 22:38, 26 November 2019 (UTC)
 * , we could do two lines for Bloomberg then? One for before and one for now and leave the space in the middle empty. That's what was done for undecideds a few months back when we didn't have the data in the middle. MikkelJSmith (talk) 22:43, 26 November 2019 (UTC)
 * I agree. Two disconnected lines would show Bloomberg's support from Feb/Mar without implying 0 support throughout the period from then to November. Jacoby531 (talk) 23:07, 26 November 2019 (UTC)
 * I believe this is not possible without creating two separate lines (with two legend entries). --Gbuvn (talk) 00:12, 27 November 2019 (UTC)
 * Correct; the graph module does not behave well when there is missing data within a series. Anyway, perhaps we should exclude Bloomberg from the graph until he consistently reaches more than 3% in upcoming polls. His relative popularity nine months ago, when he was not yet a candidate, is not very relevant to this chart. — JFG talk 05:20, 27 November 2019 (UTC)
 * He has recently polled at 3 percent, though it is, admittedly, only 1 data point. WittyRecluse (talk) 06:01, 27 November 2019 (UTC)
 * I think him hitting the threshold just the once should be enough. I do have my doubts about that, but they are somewhat reassured by the fact that there are plenty of people polling below him who've made it onto there to boot. PutItOnAMap (talk) 16:56, 27 November 2019 (UTC)
 * , we used to have two lines for undecideds with the same color, we could do that. We would just need to style the first one as Bloomberg (until ...) and the second one as Bloomberg (from .. . onward) MikkelJSmith (talk) 19:15, 27 November 2019 (UTC)
 * Done. PutItOnAMap (talk) 17:20, 6 December 2019 (UTC)

The Economist
So, much like some of the aggregators (and polls) that were removed the Economist has started duplicating polls : https://gyazo.com/f501218c32333b33f1a013203c2d7c5c. Due to prior consensus, when we removed polls and aggregators due to this, should we remove the Economist for this reason as well? What do you think ? - MikkelJSmith (talk) 21:36, 27 November 2019 (UTC)
 * Yep, and how can we WP:verify the series in the chart? Look at the history of the aggregator websites at the Wayback Machine, says 🙄 Guarapiranga (talk) 03:44, 28 November 2019 (UTC)
 * I'm interested in why this page uses a collection of aggregators for the polling, rather than displaying the polls from the page like every other election. It was also used for the 2016 primaries. Catiline52 (talk) 04:19, 28 November 2019 (UTC)
 * , apparently no one knows how to make them here. MikkelJSmith (talk) 04:41, 28 November 2019 (UTC)
 * I remember going through the history of this page and seeing something similar to the old 2016 graph, though? WittyRecluse (talk) 23:11, 28 November 2019 (UTC)
 * , the person who made that graph retired from Wikipedia. MikkelJSmith (talk) 01:20, 29 November 2019 (UTC)
 * Using aggregators as sources for the chart avoids WP:SYNTHESIS. Besides, the simple average of three reliable aggregators makes for a more legible graph than showing hundreds of scattered dots for all polls. — JFG talk 15:22, 30 November 2019 (UTC)].
 * I have argued before that while raising this guideline as a concern makes some sense, it does not avoid a synthesis as we are still synthesising aggregators and weighting the polls within them as opposed to representing all of the polls, rather than representing a subset with an uneven combination of outside sources' weightings. If a graph more similar to 2016's, or one in line with that used in the UK 2019 General Election article, is preferred, I would be more than happy to see to its creation. PutItOnAMap (talk) 16:34, 6 November 2019 (UTC)
 * , I would prefer it, especially since the Economist is now doubling polls, has removed Ipsos and most polls are now there own, so we could argue it's no longer even a real aggregator. I know that in previous discussions some of the editors were fine with it, they just didn't want a poll to be reused in the graph. MikkelJSmith (talk) 00:25, 8 December 2019 (UTC)
 * Tagging other users that were involved in this discussion to mention that has offered to make a grah, which I think would be better, would like to know your thoughts : ,,
 * Oh wow I'm cool enough to be tagged? If is willing to make it, I would much rather have that graph, seeing as that seems to be the wikipedia standard. However, I wouldn't have a problem keeping the polling aggregation as well, but I certainly do prefer the 2016 style of graph over an aggregator. WittyRecluse (talk) 08:07, 8 December 2019 (UTC)
 * , That would be fine, I think. It would look like the 2016 page, with the graph and the aggregators. MikkelJSmith (talk) 13:35, 8 December 2019 (UTC)
 * Especially, since I'm no longer sure the Economist even counts as an aggregator, they've removed Ipsos/Reuters and now their own polls make up the majority of the polls they use and they count some polls like Survey USA  more than once, which is actually one of the reasons we removed some of the older aggregators. MikkelJSmith (talk) 00:33, 8 December 2019 (UTC)
 * Nevermind, they've readded Ipsos/Reuters, but they still count some polls more than once though. MikkelJSmith (talk) 16:48, 9 December 2019 (UTC)
 * Nevermind, they've readded Ipsos/Reuters, but they still count some polls more than once though. MikkelJSmith (talk) 16:48, 9 December 2019 (UTC)

Unrelated question about The Economist
How do people find the detailed data for the Economist? The only data I can find on their website only shows polling figures to the nearest 1%. Chessrat ( talk, contributions ) 18:17, 28 November 2019 (UTC)
 * , click on the candidates' names. WMSR (talk) 18:34, 28 November 2019 (UTC)
 * Thank you. Chessrat  ( talk, contributions ) 02:53, 30 November 2019 (UTC)

Length of Harris' line
Hey User:JFG - I see that you reverted my edit ending Kamala Harris' line at the November 26 update. While she does indeed have polling data for today (December 3), I think that her line should still end at the last weekly update before she dropped out, which would be November 26. Otherwise, it will fallaciously show polling data for her until the next update (which is coincidentally tomorrow). Similarly, Beto O'Rourke's line ended on October 30 while he didn't drop out until November 1. SCC California (talk) 06:37, 4 December 2019 (UTC)


 * I'd end the lines at the closest possible point. October 30 is closest to November 1, but December 4 is closest to December 3. Ending Harris' line early implies she quit six days before she actually did. Chessrat  ( talk, contributions ) 09:49, 4 December 2019 (UTC)
 * Addendum: One possible alternative is to change the date of the line update from Wednesday to Tuesday this week as a one-off (i.e. updates on November 26 - December 2 - December 10, rather than December 3). Chessrat  ( talk, contributions ) 09:52, 4 December 2019 (UTC)
 * Sounds like a good idea, although I don't think most readers will notice either way. --Comment by <b style="color:#14866d">Selfie City</b> ( talk about my  contributions ) 12:37, 4 December 2019 (UTC)
 * We'll knowFjantelov (talk) 02:29, 6 December 2019 (UTC)
 * I would support Chessrat's suggestion to stick to the closest weekly score to the date when a candidate withdrew from the race. Not a fan of making a one-off change of the usual day of the week for everybody just because one candidate dropped out a day before that. — JFG talk 13:38, 4 December 2019 (UTC)
 * I'd prefer Chessrat's suggestion to the Nov 26 cut-off point, but I don't see why we couldn't just add an extra data point between weekly data points when/if a candidate dropped out. User:PutItOnAMap (talk) 16:38, 6 December 2019 (UTC)
 * , I would also go for Nov 26 for Harris MikkelJSmith (talk) 00:30, 10 December 2019 (UTC)