Talk:Statewide opinion polling for the 2020 United States presidential election

New polling

There's new polling for GA and NC: [1] Prcc27 (talk) 18:33, 22 July 2019 (UTC)

How about indicating which polls are qualifying per DNC?

Thoughts? Humanengr (talk) 03:12, 31 August 2019 (UTC)

If you're talking about qualifying for the debates, then I believe that is not for the general election polls (this page), but for the primary polls, where they are already marked. --Spiffy sperry (talk) 20:36, 3 September 2019 (UTC)

map?

How about compiling a map (or rather one for each potential dem candidate against the incumbent) that quickly gives an overview of the red and blue states, and projects the electoral college? Should be relatively easy to do from the tables, and would provide a good overview. --91.41.47.165 (talk) 00:44, 11 September 2019 (UTC)

I added a map. Heitordp (talk) 23:15, 17 September 2020 (UTC)

Illinois

What about Illinois? Greenparties1 (talk) 17:44, 3 January 2020 (UTC)

Former candidates

It will be good to hide all former candidates polls, it'll make page much clearer. — Preceding unsigned comment added by 78.8.185.127 (talk) 22:08, 13 February 2020 (UTC)

I agree, and have for some time. I will make adjustments on the pages that are replicated here. --Spiffy sperry (talk) 16:57, 14 February 2020 (UTC)

Date format

I suggest that the dates format be changed to one rationally table sortable. Deancarmeli (talk) 13:23, 16 April 2020 (UTC)

CivicScience/Microsoft Research/MSN (MRP) Polls

Hello there, in 2016, we had decided to report on Wikipedia 2016 statewide polling for 2016 presidential election only individual state polls. As a consequence of this decision, tracking surveys polling all States in same conditions had been banned from Wikipedia statewide polling page. It appears « CivicScience/Microsoft Research/MSN (MRP) » polls recently reported on this page is a tracking poll of all states. Previous rules would have banned these CivicScience/Microsoft Research/MSN (MRP) polls. I personaly also think we should ban these polls. What are your opinions about that? Sthubertliege (talk) 16:30, 27 April 2020 (UTC)

They don't appear to be representative. Trump is currently losing approval, but according to Civic Biden would currently win 435 electors is not realistic. --Petruz (talk) 22:51, 27 April 2020 (UTC)

I agree, they are not very serious. The margins of lead are so unrealistic ! I suggest we should ban these polls. Sthubertliege (talk) 06:41, 28 April 2020 (UTC)

I agree that what they project seems optimistic for Biden in most states (though they're pretty reflective of state-level polls in others), but they're not push polls and we've put MRP polls up before. These are different from the MRP ones we've put up elsewhere because these used a nationwide sample instead of just a statewide one (though there are still at least 3000 responses from each state even in these MRP polls). If you consider this sampling method to be disqualifying, then I'm happy to leave them removed, but alternatively we could put (nationwide MRP) in the 'sample size' column. PutItOnAMap (talk) 12:34, 28 April 2020 (UTC)

Polling aggregation in swing states section graph

Hi, I'm not aware of whether it's just me, but the chart in the "Polling aggregation in swing states" section appears to be broken, as it's not loading. I've tried three different browsers, and when I tried to edit it to find what was broken and why it wasn't showing up, it began to load and then transmogrified into a black bar that caused the page width to expand substantially. Phinbart (talk) 14:33, 25 August 2020 (UTC)

98.116.236.116 (talk) added a data point to the North Carolina line without adding a corresponding date on the x-axis (here). I believe this is causing the error. Viewing the page version before this edit may not help, because I find that these charts are always broken when viewing old versions of pages. I would add a date myself, but I have no insight on the source of the data. --Spiffy sperry (talk) 15:03, 25 August 2020 (UTC)

Thanks. Removing the datapoint seems to fix it (at least in the edit preview), but I'm going to leave it for now so someone more experienced with the graph and the data can ascertain as to whether the data point is correct or not. Phinbart (talk) 16:33, 25 August 2020 (UTC)

The figure is very crowded. I would remove New Hampshire and Kansas and other states, which are not swing states. 2800:200:E200:763B:A14B:69BF:E2F0:C04E (talk) 17:52, 30 October 2020 (UTC)

Polling Average map like 2016

Can someone create a Polling Average Map.Alhanuty (talk) 00:45, 12 September 2020 (UTC)

I added a map. Heitordp (talk) 23:15, 17 September 2020 (UTC)

1st Oregon poll of 2020 released

Please include in the list.

51% Biden 39% Trump

https://www.dhmresearch.com/joe-biden-leads-by-12-points-among-oregon-voters-worry-about-covid-19-spread-remains-high/

https://www.dhmresearch.com/wp-content/uploads/2020/09/DHM-Panel-Presidential-and-COVID-memo-September-2020-2.pdf

https://www.dhmresearch.com/wp-content/uploads/2020/09/DHM-Panel-Survey-Presidential-and-COVID_tables-September-2020.pdf

Thanks.

--The Pollster (talk) 14:04, 19 September 2020 (UTC)

Done WittyRecluse (talk) 01:40, 20 September 2020 (UTC)

Can someone update the image? I tried to but its named .svg even tho its a png and I can't figure out how to not get the image to fail the wikimedia upload. I have the new updated image if you need it. WittyRecluse (talk) 02:33, 20 September 2020 (UTC)

The file is SVG. To update it, first download it from the "original file" link, not a PNG version. Then open the file with a text editor like Notepad. In the beginning of the file, there are 5 lines with lists of states, one line for each color. For example, to change Oregon from dark blue to light blue, move ".or," from line 5 to line 4. I just made this change and also updated two other states. You can try yourself in future updates. Heitordp (talk) 06:21, 21 September 2020 (UTC)

Thank you, I really appreciate that. It is difficult to find how to do such things. WittyRecluse (talk) 05:06, 22 September 2020 (UTC)

1st Vermont poll of 2020 released

Here is the first VT poll of 2020 by Vermont Public Radio:

Biden+24% vs. Trump

http://projects.vpr.org/vpr-vermont-pbs-2020-polls-september

https://www.vpr.org/sites/vpr/files/202009/methodology_report_september_2020_vpr_vermont_pbs_poll.pdf

Please include in the state table. Thanks. --The Pollster (talk) 11:29, 22 September 2020 (UTC)

Map criteria

I recently made the map, using the most recent poll for each state. However, another user pointed out that the polls for some states were outdated or unrealistic, and I found that using only the most recent poll would require many frequent updates for some states. I also noticed that the map in the previous election used different colors for polls older than September. So I propose the following criteria for a more realistic and stable map:

• If a state doesn't have a poll since September, color it dark blue or dark red based on the result of the previous election. States with no recent polls are the most likely to maintain the same result between elections, so even after a poll appears it will probably just change the state color from dark to light but not between blue and red.
• If a state has a poll since September, and any of these polls has the top two candidates within the margin of error or different polls indicate different winners, mark it purple. These states are the least predictable, and this way they remain purple in the map without requiring further updates.
• If a state has a poll since September, and all of these polls indicate the same winning candidate outside the margin of error, mark it light blue or light red.

What do you think? Heitordp (talk) 05:20, 24 September 2020 (UTC)

More of a question than an opinion, but is this an adjustment of the way the current map is laid out, just with the cutoff being September first as opposed to not having a cut off? If so, why is September the cutoff? WittyRecluse (talk) 06:39, 24 September 2020 (UTC)

September is usually the month where pollsters switch from registered voter samples to likely voter samples, which are the more reliable versions to pin down a state's voting behaviour. I'm definitely OK with using September 1 as the cutoff date. Using polls before that, such as the outdated Arkansas poll, doesn't make any sense really. I'm also OK with the other proposed changes.--The Pollster (talk) 14:10, 24 September 2020 (UTC)

Proposed changes sound good, with the change of grey for purple for colorblind concerns below. WittyRecluse (talk) 15:47, 24 September 2020 (UTC)

I would prefer gray over purple. To my colorblind eyes, purple looks too close to blue, and we used gray during the last election. --Spiffy sperry (talk) 14:49, 24 September 2020 (UTC)

I have updated the map, using the start dates as opposed to the end dates for determining what counts as September. This excluded California and Idaho, wondering if this should be changed. I do not particularly like that we have Kansas included with one poll with 800 voters of an undisclosed voter base btw. Also, what do we do for places like me2 that do not have margins of error listed? I just put them as outside margin of error because I counted the MoE as 0. Lastly, what do we do if it is the same as the margin of error, like with Mn? I just had it as within MoE. I also noticed the current map by current standards was pretty outdated. Przemysl15 (talk) 07:36, 25 September 2020 (UTC) Also should I have gone with a darker grey? Przemysl15 (talk) 07:39, 25 September 2020 (UTC)

Hey, thanks for the changes. A darker grey would be better for swing states, so the electoral votes are more visible. Also, you need to change Wisconsin and Maine's 2nd district to tossup. They are within the margin of error too. Yougov poll in Wisconsin is close and the Suffolk poll in Maine-02 too. Thanks.--The Pollster (talk) 09:06, 25 September 2020 (UTC)

Updated the grey, but none of the Me2 polls have any MoE and the YouGov Wisconsin polls have a MoV of 6 and 4 with a MoE of 4 and N/A respectively, so I didn't update those. Przemysl15 (talk) 15:39, 25 September 2020 (UTC)

Hi, the grey could even be a little darker. I don't see any difference. Also, Maine-02 must definitely be coloured grey because the difference in the Suffolk poll is only 2% and it only has a sample size of 220 people, where the margin of error is usally around +/-6%, even if it is not listed. It is definitely bigger than 2%. Also, the WI poll has a sample of 600, where the MoE is typically 5% and the Biden lead is exactly 4%. So, WI should also be a tossup (grey). Thanks.--The Pollster (talk) 09:22, 26 September 2020 (UTC)

There are also new September polls for CA and MD. --The Pollster (talk) 11:43, 26 September 2020 (UTC)

Thanks for updating the map. To answer your questions, I suggest including polls where most of the dates are in September. It's arbitrary, but eventually it won't matter as more polls get added. For polls that don't provide a margin of error, use the margin from another poll with a similar sample size, or use this calculator (for population size, put 250,000, about the smallest number of votes in any state or district in the previous election; increasing this number further practically doesn't affect the result). Also note that the margin of error is plus or minus, so the difference between the two top candidates needs to be more than twice the margin of error to be outside of it. If it's exactly twice, I also think that we should consider it as within the margin.

There were some extra commas in the code that prevented the gray color from displaying correctly, so I fixed it. I also updated the map based on some new polls and considering the margin of error plus or minus. But now 20 states and 2 districts are gray, making the map not look very useful. So I suggest expanding the criteria:

• If a state doesn't have a poll since September, color it dark blue or dark red based on the result of the previous election.
• If a state has a poll since September, and all of these polls indicate the same winner outside the margin of error, mark it medium blue or medium red.
• If a state has a poll since September, and all of these polls indicate the same winner but at least one poll is within the margin of error, mark it light blue or light red.
• If a state has a poll since September, and not all of these polls indicate the same winner (excluding ties), mark it gray.

If this is too complicated, an easier alternative would be to use the aggregate polls and color shades based on the leads, for example gray <5%, light 5-10%, medium 10-15%, dark >15%. What do you think? Heitordp (talk) 16:16, 27 September 2020 (UTC)

I would prefer the first option for clarity, but the second is probably more in line with WP:SYNTH Przemysl15 (talk) 08:40, 28 September 2020 (UTC)

There are now also September polls for Alaska and North Dakota. Please update the map. Also: isn't the margin of error for the whole poll ? Not for each candidate ? For example if the poll result is 48-42 Biden and the MoE is +/- 4%, shouldn't the map be light blue for Biden instead of grey ? I don't think that the MoE can be multiplied by 2 for each candidate, so that the MoE is 8% instead of 4%. --The Pollster (talk) 08:47, 2 October 2020 (UTC)

The margin of error is plus or minus for each candidate. For example, if a candidate is found to have 48% and the margin of error is 4%, it means that the candidate can be between 44 and 52%. Same for the other candidate. So for the lead to be "outside the margin of error", with no overlap, the difference between candidates needs to be twice the margin of error. This is the standard used in statistical analyses, see this article. Heitordp (talk) 10:01, 2 October 2020 (UTC)

I updated the map based on the expanded criteria above and newer polls. I think that the map reflects the situation better now. Heitordp (talk) 01:37, 5 October 2020 (UTC)

Hi, could you please update the map again ? There have been numerous new polls and many states are now within the margin of error, such as NV, MN, MO, WI, PA etc. - Question: how do you apply the criteria for colouring a state ? You said you use 2x the margin of error, so for example 2x 4.0 = 8%. If only one poll since September 1 is for example below that, is the state automatically grey ? Or do you use an average of all polls since September 1 ? I think using an average of ALL polls since September 1 would be better because it smooths out outliers. What do you think ? Also: I liked the darker shades of the colours that were used before much better.--The Pollster (talk) 08:31, 12 October 2020 (UTC)

Please see the revised criteria above, which I wrote on September 27, and the legend on the map. I use 3 shades of blue and red, plus gray. The dark shade is for states/districts with no polls since September 1 (now only used for 2 Nebraska districts because all states already have polls). The medium shade is where all polls since September 1 indicate the same top candidate outside the margin of error (such as CA, IN). The light shade is where all polls since September 1 indicate the same top candidate, but at least one poll is within the margin of error (such as the states that you mentioned). The gray color is where polls since September 1 indicate different top candidates (such as FL, OH). The reason why I revised the criteria is because there were too many states within the margin of error (as you noticed), so if I used the previous criteria the map would be mostly gray, not very useful.

I think that calculating averages on our own would be too complicated and would violate WP:OR. If you prefer averages, it would be better to use the aggregate polls that other sources already calculate. But the problem is that we'd have to specify certain percentage ranges to determine the color shades, and some states would change shades back and forth between ranges, requiring frequent updates. With the criteria that I wrote, the colors can only change in one direction, lighter then gray, so the map becomes more stable. Heitordp (talk) 09:22, 12 October 2020 (UTC)

Would it be OK to set a new deadline of October 1 for inclusion of polls ? There are going to be lots of new polls out and therefore Oct. 1 as deadline would be better. Also: Michigan as tossup/grey is very misleading. Almost all polls have Biden clearly ahead, only one Trafalgar poll from September Trump slightly. And even a later Trafalgar poll had Biden ahead. All averages also have Biden ahead. If you are not changing the deadline to October, could you at least use the newer Trafalgar poll and toss the old ? --The Pollster (talk) 13:49, 15 October 2020 (UTC)

I agree that October polls are better, but not all states have them. I also agree that the Trafalgar poll is an outlier, but we need to specify a consistent set of criteria for all states. I also noticed that the article already lists averages of the aggregate polls, so I'm reconsidering your previous suggestion about averages. I propose the following:

• Use the average of the aggregate polls for each state/district. If there is no aggregate poll, use the most recent individual poll. If there is no poll at all, use the result of the previous election.
• Use the following colors, based on the difference between the top candidates: <5% gray, 5-10% light red/blue, 10-15% medium red/blue, >15% dark red/blue.

This set of criteria might require more frequent updates for states near thresholds, but I suppose that it will make the map more realistic. What do you think? Do you prefer different or fewer thresholds? Heitordp (talk) 02:20, 18 October 2020 (UTC)

That sounds good. What would the map look like right now with your proposals ? --The Pollster (talk) 13:09, 18 October 2020 (UTC)

Some states would be darker, MI light blue, and AK NE1 gray. I'll make the change. Heitordp (talk) 19:02, 18 October 2020 (UTC)

What do you guys have against Survey Monkey?

I noticed that SM does polling for all 50.5 states, and they released them last week. They were put up by 538.com, and then by you guys. Then you took them down. Why? There are several states where these are the only data available. What do you have against these people? could someone put them back up?Arglebargle79 (talk) 12:08, 28 September 2020 (UTC)

You are right, they conduct rolling interviews in all 50 states + DC. But they only publish weighted poll data for 4 states: WI, MI, OH and PA. The other 46 states are using unweighted data, which cannot be used for this page. It's not scientific. Only weighted polls are. --The Pollster (talk) 08:33, 29 September 2020 (UTC)

SurveyMonkey polls have been added back to the page for SD, DC, and RI. Should these be removed? Przemysl15 (talk) 12:51, 4 October 2020 (UTC)

No, let them stay. Those new polls are different. They are for all of September, they are for „likely voters“, weighted and have been added to 538.com as well. They are OK. --The Pollster (talk) 19:28, 4 October 2020 (UTC)

They're messing around wiki

My gad, they're plenty of Trump supporters accounts are messing around the wiki edits..... This is so bad & disgraceful.... vandalizing anything on the internet. Can't they just report these guys? Syaz351 (talk) 13:15, 3 October 2020 (UTC)

Re: the polling aggregation swing states diagram

Since when are Alaska and Montana swing states? They have been "Safe R" since like forever, and have only 7 electoral votes between them.

They only clutter the diagram, and should be removed. Missouri and South Carolina could be removed on the same grounds, "Safe [X]". There's a reason California is not there.

HandsomeFella (talk) 21:40, 18 October 2020 (UTC)

You need to look at the recent polling. Several new polls show that these states are much closer than in previous elections. --The Pollster (talk) 05:29, 19 October 2020 (UTC)

Additionally, of prediction sources used on the main page, namely Cook, Inside Elections, Sabato, Politico, RealClearPolitics, Niskanen, CNN, The Economist, 270toWin, ABC News, NBC News, and FiveThirtyEight (excluding NPR and CBS who only use Lean, Likely, and Tossup), only CNN lists Alaska and Montana as "Solid R" (their phrase for "Safe R"), the rest using Likely or Lean, with the exception of Niskanen, which lists Alaska as a Tossup. Similarly, Missouri is a "Safe/Solid R" for only two sources and between Lean and Likely R for the rest. South Carolina is the most republican heavy and still only has 3 "Safe/Solid R" listings as the rest are all Likely R save 1 Lean R. By contrast, California received the most Democratic rating possible from every source. Przemysl15 (talk) 07:20, 19 October 2020 (UTC)

What happened to Wisconsin?

Anyone else see all the polling data disappear? I've got it all saved in another spreadsheet in any case, but I came looking for updates! — Preceding unsigned comment added by 69.123.249.40 (talk) 22:38, 3 November 2020 (UTC)

I see it too. Also, the following section, Wyoming, has an error message in the table. The data in the underlying pages are intact, so appears to be a problem with the massive number of footnotes. I may attempt to work on this later, by combining identical notes. --Spiffy sperry (talk) 23:12, 3 November 2020 (UTC)

I do not see the issue with Wyoming. Also, I have no idea what happened to Wisconsin, the transclusion appears to have been done properly and is linking to a section that exists. I have commented out the transclusion for the time being. Przemysl15 (talk) 01:26, 4 November 2020 (UTC)

Actually, when I went to make the edit I went into preview and it was displaying properly there so I just added a line, but it is still broken. May be due to the page having to load too much? I'm not really sure. Przemysl15 (talk) 01:29, 4 November 2020 (UTC)

The problem with Wyoming was fixed after I combined many of the identical notes. I am at a loss for the problem with Wisconsin. --Spiffy sperry (talk) 01:44, 4 November 2020 (UTC)

Should we try to manually transclude? Przemysl15 (talk) 04:29, 4 November 2020 (UTC)

I guess so. --Spiffy sperry (talk) 15:03, 4 November 2020 (UTC)

I have manually transcluded. In the mean time, where do we go/who do we ask to help with issues like this? Przemysl15 (talk) 22:37, 4 November 2020 (UTC)

I think the MediaWiki Support Desk is the place to start, though I don't have much experience there. --Spiffy sperry (talk) 22:58, 4 November 2020 (UTC)

I have opened a help request here: https://www.mediawiki.org/wiki/Topic:Vx54h8h6bi0ulfhi. Przemysl15 (talk) 23:59, 4 November 2020 (UTC)

Przemysl15, the support desk figured it out. In short, there is a size limit, WP:PEIS, which I am honoring by excluding some of the less relevant collapsed tables in each section (former candidates and hypothetical polling). I am about halfway through and have verified that it works. --Spiffy sperry (talk) 16:25, 5 November 2020 (UTC)

Oh fantastic, that's great! Thanks for your help. Przemysl15 (talk) 20:02, 5 November 2020 (UTC)