Democratic Primaries

highplainsdem

(49,004 posts) Mon Dec 23, 2019, 03:09 PM Dec 2019

Reuters/Ipsos did 2 different GE polls Dec. 18-19, two different sets of voters.

They got two different results.

One of the polls has a higher percentage of Republicans than the other poll has. That's the poll Sanders is doing better in, the poll that a Newsweek article is hyping, and that a couple of DUers who can't stand Biden posted, 45 minutes apart.

This is the poll where Sanders beats Biden by 2 points, while Biden beats Trump by 1 point, and Warren loses to Trump by 2 points:

https://www.ipsos.com/sites/default/files/ct/news/documents/2019-12/topline_reuters_impeachment_vote_12_19_2019.pdf

These are some of the findings of an Ipsos poll conducted between December 18-19, 2019 on behalf of Thomson Reuters. For this survey, a sample of 1,108 adults age 18+ from the continental U.S., Alaska and Hawaii were interviewed online in English. The sample includes 444 Democrats, 439 Republicans, and 119 independents.

In the other Reuters/Ipsos poll also conducted Dec 18-19, the numbers of respondents and their political affiliation were different:

https://www.ipsos.com/sites/default/files/ct/news/documents/2019-12/topline_reuters_2020_democratic_primary_tracker_12_19_2019_writeup.pdf

These are some of the findings of an Ipsos poll conducted between December 18-19, 2019 on behalf of Thomson Reuters. For this survey, a sample of 1,117 adults age 18+ from the continental U.S., Alaska and Hawaiiwere interviewed online in English. The sample includes 481 Democrats, 408 Republicans, and 139 independents.

In that poll of Democratic candidates versus Trump, Biden and Sanders both beat Trump by 4 points, Warren beat Trump by 2 points, Bloomberg beat Trump by 1 points, and Buttigieg lost to Trump by 1 point.

Don't ask me why they did two separate polls this way.

They messed up the math they're reporting, too. In the first poll there, the one with nearly as many Republican as Democrats (which is wrong for a GE poll), their internal table at the top of the press release says the respondents were 43% Democrat and 38% Republican. Nope. Takes seconds with a calculator to check that and see both parties are 40% (well, the Republicans were 39.6%).

Anyway, their polling is messed up.

And the poll that's most messed up is the one showing Sanders doing slightly better against Trump than Biden does.

If I were to vote in a presidential
primary today, I would vote for: Joe Biden

13 replies

= new reply since forum marked as read

Highlight:

Reuters/Ipsos did 2 different GE polls Dec. 18-19, two different sets of voters. (Original Post) highplainsdem Dec 2019 OP

Unskewing the Polls? bluewater Dec 2019 #1

Just reporting the data. Check the links. Two polls, same days, one overpolling Republicans. highplainsdem Dec 2019 #5

Sure seems like "unskewing" the polls though. How do you know it's over polling anyone? bluewater Dec 2019 #9

Thx for breaking that down... tempting to ignore the whole thing Thekaspervote Dec 2019 #2

No, this isn't a good look for Ipsos. highplainsdem Dec 2019 #6

Polls rso Dec 2019 #3

The polls were probably reweighted to reflect the final reported distribution of R vs D Dream Girl Dec 2019 #4

If they're going to do two different GE polls on the same dates, getting different results, they highplainsdem Dec 2019 #7

+1 TexasTowelie Dec 2019 #8

Polls stlll are kind of meaningless at this point. LiberalLovinLug Dec 2019 #10

" a couple of DUers who can't stand Biden" TryLogic Dec 2019 #11

Polls, schmoles. littlemissmartypants Dec 2019 #12

KR! Cha Dec 2019 #13

bluewater

(5,376 posts)

1. Unskewing the Polls?

Reply to highplainsdem (Original post)

Mon Dec 23, 2019, 03:13 PM

Dec 2019

The polls are what the polls are.

We can all relax, real voting is almost upon us.

If I were to vote in a presidential
primary today, I would vote for: Joe Biden

highplainsdem

(49,004 posts)

5. Just reporting the data. Check the links. Two polls, same days, one overpolling Republicans.

Reply to bluewater (Reply #1)

Mon Dec 23, 2019, 03:26 PM

Dec 2019

If I were to vote in a presidential
primary today, I would vote for: Joe Biden

bluewater

(5,376 posts)

9. Sure seems like "unskewing" the polls though. How do you know it's over polling anyone?

Reply to highplainsdem (Reply #5)

Mon Dec 23, 2019, 03:42 PM

Dec 2019

Likely voter models are a black art, we all know that about polls.

"unskewing" any one particular poll just seems like cherry picking results.

As I always say...

The polls are what the polls are.

But, we can all relax, real voting starts within a handful of weeks.

Then we will have real data to discuss.

If I were to vote in a presidential
primary today, I would vote for: Joe Biden

Thekaspervote

(32,778 posts)

2. Thx for breaking that down... tempting to ignore the whole thing

Reply to highplainsdem (Original post)

Mon Dec 23, 2019, 03:18 PM

Dec 2019

Not a good look for ipsos

If I were to vote in a presidential
primary today, I would vote for: Joe Biden

highplainsdem

(49,004 posts)

6. No, this isn't a good look for Ipsos.

Reply to Thekaspervote (Reply #2)

Mon Dec 23, 2019, 03:27 PM

Dec 2019

If I were to vote in a presidential
primary today, I would vote for: Joe Biden

rso

(2,272 posts)

3. Polls

Reply to highplainsdem (Original post)

Mon Dec 23, 2019, 03:19 PM

Dec 2019

National polls are next to useless. The only polls that matter are those in the 6 or 7 key battleground States.

If I were to vote in a presidential
primary today, I would vote for: Joe Biden

Dream Girl

(5,111 posts)

4. The polls were probably reweighted to reflect the final reported distribution of R vs D

Reply to highplainsdem (Original post)

Mon Dec 23, 2019, 03:26 PM

Dec 2019

Typically you’d want the same sample size for both subgroups so that you can break it out by demographic breakouts, subgroups, etc. however when they report the totals they apply a weighted average to bring it back into line. Hope that makes sense. BTW I do this for a living.

If I were to vote in a presidential
primary today, I would vote for: Joe Biden

highplainsdem

(49,004 posts)

7. If they're going to do two different GE polls on the same dates, getting different results, they

Reply to Dream Girl (Reply #4)

Mon Dec 23, 2019, 03:34 PM

Dec 2019

should make a note of that and link to the other poll.

Right now, if someone just refers to the Reuters/Ipsos GE poll done Dec. 18-19. there's no way to know which one they're referring to.

That's particularly important to anyone concerned about Warren's polling versus Trump, since in one poll she loses to Trump by 2 points, and in the other she beats Trump by 2 points.

And since the poll where she loses to Trump by 2 points is the one where they apparently overpolled Repubcans, I don't think they fixed the problem with reweighting. That poll with too many Republicans also had Biden and Sanders doing a couple points worse against Trump.

If I were to vote in a presidential
primary today, I would vote for: Joe Biden