August 18 Poll Report

August 18, 2019
 

August 18 Poll Report

Good afternoon.   My poll staggers on, wounded but not yet down.   In the poll of Friday, August 16, Kirsten Gillibrand (a) significantly exceeded expectations and (b) won the poll, for the second straight time:

Scores

Gillibrand

234

Moulton

103

Delaney

116

de Blasio

92

 

 

 

 

 

 

 

 

 

Predicted

Gillibrand

43

Moulton

19

Delaney

21

de Blasio

17

Actual

Gillibrand

57

Moulton

10

Delaney

23

de Blasio

9

 

Second straight Gillibrand poll in which she significantly exceeded expectations and won; she also did this on August 9.   In the poll of Saturday, August 17, Julian Castro had a great result, beating expectations by 14 points:

Scores

Gabbard

375

Bullock

165

Castro

370

Steyer

58

 

 

 

 

 

 

 

 

 

Predicted

Gabbard

39

Bullock

17

Castro

38

Steyer

6

Actual

Gabbard

32

Bullock

9

Castro

52

Steyer

6

 

Since my last report two days ago, Elizabeth Warren is up 100 points and Donald Trump is down 52 points, mostly as a result of the removal of the July 4th poll from the relevant data.  In the poll on Independence Day, Warren got 58%, Trump 27%.   But Warren’s current score is more than four times Trump’s score, which suggests that Warren should beat Trump something like 81-19.  (It suggests this wrongly, I agree, but this is the mathematical implication.)   Anyway, when you take the Warren/Trump poll out of the system, the remaining data tries to push Warren and Trump into an 81-19 relationship or ratio.  Since their numbers are so large, Warren being now at 2,000 or 20%, the removal of that poll has large effects on the Support Scores. 

Since my last report two days ago, in addition to the changes for Warren and Trump:

Kirsten Gillibrand is up 29 points as a result of Friday’s poll,

Pete Buttigieg is up 12 points as a secondary adjustment,

Michael Bennet is down 10 points,

Bill de Blasio is down 12 points as a result of Friday’s poll,

Steve Bullock is down 29 points as a result of yesterday’s poll,

These are the current, and, it may be, the final Support Scores from my audience:

Rank

First

Last

Support

1

Elizabeth

Warren

2000

2

Pete

Buttigieg

1063

3

Kamala

Harris

1002

4

Joe

Biden

755

5

Andrew

Yang

507

6

Amy

Klobuchar

481

7

Bernie

Sanders

443

8

Donald

Trump

416

9

Cory

Booker

413

10

Julian

Castro

390

11

Tulsi

Gabbard

356

12

Beto

O'Rourke

275

13

Kirsten

Gillibrand

263

14

Jay

Inslee

251

15

Michael

Bennet

250

16

Bill

Weld

221

17

Howard

Schultz

162

18

Tim

Ryan

160

19

Steve

Bullock

136

20

Marianne

Williamson

111

21

John

Delaney

110

22

Seth

Moulton

94

23

Bill

de Blasio

80

24

Tom

Steyer

59

 

I’m not 100% on this yet, so I am going to sit on it for 24 hours before I make a final decision, but I have tentatively decided to end the series of polls at this point.   The reasons for this were explained in the August 16th poll report, which I will re-publish at the end of this article.   Basically, the twitter polling series has to end when the point is reached at which the campaigns or persons supporting the campaigns begin driving their supporters to the poll in numbers sufficient to seriously undermine our ability to draw conclusions from the poll.   This happened in spades on Thursday, with Andrew Yang supporters, but it happened again on Friday, with Gillibrand supporters, and again on Saturday, with Castro supporters. 

There are several signs that this is happening in a given poll:

1)     The poll results for one candidate deviating strongly from expectations.  

2)     Mid-day "surges" in the poll; that is, a period of time in which a candidate suddenly goes from 20% to 75% in the poll, and then retreats slowly but ends up at 45%,

3)     A larger number of responses to the poll than are expected,

4)     Twitter notifications about the poll being re-tweeted, which, when checked out, can be traced to supporters of a campaign. 

 

It’s not really difficult to tell when someone is trying to swamp the poll.  The fact that it is not difficult to tell is a primary reason that I didn’t worry about it happening before, when some of the citizens of BJOL thought that I should be worrying about it.  But I don’t see any reason to believe that this will stop happening if I continue doing the polls.  It seems clear that it will just get worse and worse.  It was always clear that this would happen; I never had any doubt that it would happen.  I just didn’t know when. 

If I continue the polls now, I will just be measuring the competitive ability of the different campaigns to drive their supporters toward my poll, or how much interest they have in doing so.   While I can envision a scenario in which that might become very interesting, (a) this was not my intention in starting the series, and (b) the interim period before we reach the point of everyone making an equal effort to do that would yield ugly, essentially uninteresting results.  I’m just not going there.   So the polling series is probably done at this point.  I thank you all for participating, and I thank you for your comments about and insights into the process. 

 

 

 

August 16 Poll Report

 

Good afternoon, everybody.  I have not done a poll report for the past two days, have not done one since Tuesday, because I have had obligations I couldn’t dodge.  Tuesday’s poll went more or less as expected; Elizabeth Warren over-achieved by a couple of points, but the poll went 95% as expected:

Scores

Warren

1881

Buttigieg

966

Biden

773

Klobuchar

461

 

 

 

 

 

 

 

 

 

Predicted

Warren

46

Buttigieg

24

Biden

19

Klobuchar

11

Actual

Warren

48

Buttigieg

22

Biden

20

Klobuchar

11

 

That doesn’t stir the numbers much.   The poll of Wednesday also went pretty much as anticipated, although Howard Schultz got a shot of adrenaline somewhere:

Scores

Sanders

424

Schultz

149

Hickenlooper

275

Harris

976

 

 

 

 

 

 

 

 

 

Predicted

Sanders

23

Schultz

8

Hickenlooper

15

Harris

54

Actual

Sanders

24

Schultz

12

Hickenlooper

13

Harris

51

 

As you may remember, I have also begun deleting two old polls a day from the data considered relevant; until this week I was eliminating one a day, but now two a day, shortening the list of relevant polls from 50 to 40.  We’re now at 46. 

And now I have some bad news that I have to share.  This polling exercise appears to be near an end.  I don’t think I can sustain it much longer. 

In yesterday’s poll the four contestants were Beto O’Rourke, Donald Trump, Marianne Williamson and Andrew Yang.  The predictions for that poll, based on previous polling, would be that Andrew Yang would win the group with 38% of the vote.

Scores

O'Rourke

269

Trump

489

Williamson

110

Yang

500

Predicted

O'Rourke

20

Trump

36

Williamson

8

Yang

37

 

  Early in the voting, however—four or five hundred votes in—Beto O’Rourke was out-performing expectations by a huge margin.  I didn’t take notes, but he was at something like 58%. 

Well, you know what that means; it means that somebody has re-tweeted the poll to a group of Beto O’Rourke supporters.   That’s fine; the polls can survive that up to a point.  Dave Pinto and Pat Porter re-tweet the polls regularly, which is great, and there’s a guy who re-tweets the polls to a more conservative audience, which I appreciate his doing, because the Twitter audience leans so far to the left that it is in danger of falling off the mound and landing on first base in the middle of its delivery. 

Sometime about mid-day, however, an Andrew Yang supporter noticed this, and responded by re-tweeting the poll to an Andrew Yang support group of some kind.  This kind of got out of hand.   My polls usually get about 1,150 responses, but this one got 3,569 responses.  Yang, whose track record would suggest that he would get about 38% of the vote, a little bit better than that, honestly, since he was running against Donald Trump, but he got 62% of the vote. 

Nothing against Andrew Yang or his supporters or his enthusiastic supporters, but this is how the poll ends.   I gather that this is called "brigading". . .. don’t know what the derivation of that expression is. 

Anyway, I’ve always known that this would happen at some point, and this is how the poll ends.  We’ve had a good run; we had 120-some polls before we got here.  But polls which have an open doorway—that is, polls which allow anyone who chooses to participate to participate—have an inherent problem which will at some point sink them.  Maybe we’re not at that point yet, but it has to happen sooner or later.  Once people learn that they CAN end the poll by doing this, somebody will.

I did not enter the Andrew Yang/Beto O’Rourke poll into the data.   It’s like this:  suppose that your exercise plan is that you swim 30 minutes in your pool every morning, but you wake up one morning and discover that your neighbor’s sewer system has backed up and the overflow ran into your pool. Do you do your swimming that morning, or do you skip it?   You skip it, right?  I am just skipping that pool, not entering the results into the system.   Again, nothing against the people who did this; it was inevitable. 

For now, I’ll continue to post the daily polls, see how it goes.   It’s really easy to see when this brigading becomes a problem; there are many different signals that let you know that’s what’s happening.  I posted a poll today; I’ll post one tomorrow, see what I get.   At some point, the political partisans will flood the system, and I’ll have to stop.   It’s OK; everything comes to an end.  If it isn’t a life or a marriage, it isn’t a big deal.  I’ve gotten a lot out of it, and actually, I think I can continue the process in a different venue.  We’ll see. 

 

These are the current Support Scores for the 24 candidates that I am tracking:

Rank

First

Last

Support

1

Elizabeth

Warren

1900

2

Pete

Buttigieg

1057

3

Kamala

Harris

998

4

Joe

Biden

757

5

Andrew

Yang

511

6

Donald

Trump

488

7

Amy

Klobuchar

475

8

Bernie

Sanders

439

9

Cory

Booker

415

10

Julian

Castro

389

11

Tulsi

Gabbard

360

12

Michael

Bennet

271

13

Beto

O'Rourke

270

14

Jay

Inslee

246

15

Kirsten

Gillibrand

234

16

Bill

Weld

227

17

Howard

Schultz

165

18

Tim

Ryan

165

19

Steve

Bullock

149

20

John

Delaney

116

21

Marianne

Williamson

115

22

Seth

Moulton

103

23

Bill

de Blasio

92

24

Tom

Steyer

58

 

John Hickenlooper-Scooper has dropped out of the race since my last post.  Hickenlooper had 275 points, or 2.75% of the support, so when he drops out, those 275 points have to go somewhere.  Because of that, there are 13 candidates who are up ten points or more since the last post.  Basically, they’re splitting up Hickenlooper’s 275 points, and everybody winds up a little bit ahead.  Since my last post:

Pete Buttigieg is up 91 points, almost entirely as result of the removal or an old poll, which was taken just after the South Bend police shooting.   What we learn from this is that that incident has NOT seriously wounded Buttigieg, at least with this audience. 

Also, Kamala Harris is up 22 points, Jay Inslee is up 20, Elizabeth Warren up 19, Howard Schultz up 16, Bernie Sanders up 15, Julian Castro up 14, Amy Klobuchar up 14, Michael Bennet up 13, Tim Ryan up 13, Andrew Yang up 11, Bill Weld up 10, and Tulsi Gabbard up 10.   

Joe Biden is down 16 points because. . .well, he’s Joe Biden; his campaign is slowly dying, he just hasn’t figured it out yet.  He’s down 168 points in the last month, and 436 points in the last two months. 

Thank you all for reading.  We’ll hope that we can continue for a little while longer, anyway.  

 

 
 

COMMENTS (9 Comments, most recent shown first)

OldBackstop
If the twitter method was viable until only recently, I'm sure we will see Bill publish a summation of his 100 breathless daily analyses.

If not....well, maybe some dipshtt will get an apology?

Ya think?
9:31 PM Aug 20th
 
MarisFan61
Chris: What do you think it adds to what we know from the fogey polls as they are, especially if perhaps supplemented with what we think we can tell from trends in betting odds (which I think is a lot), but even if you forget about that latter thing?

I likewise have found the methodology and the results interesting -- but not at all enlightening. I don't see that they've added to what we could gather from existing stuff, and I don't at all see that what it provides is nearly worth the work.
7:47 PM Aug 20th
 
chrisbodig
If this is indeed the end, I'll be sad. I found the methodology enlightening and hope it gets copied by one of the old fogies.
11:36 AM Aug 20th
 
DaveFleming
If this is the end, I want to thank you for running the project, Bill. I've found both the effort and the conversation around it compelling, and I appreciate the work that you've put into it.

Acknowledging that this project likely slants to my worldview, it's been nice to hear, at least for a while, a narrative of the race that better echoes my own interpretations of what's happening than the one that the traditional pollsters and media have continued to push.

I look forward to the next iteration of this project, wherever it shows up.
8:22 AM Aug 20th
 
packbringley
Bill is aware — we’re all aware — of the glaring flaws of conducting polls via twitter. It’s beyond obvious. The interest (if you’re interested) is in the head-to-head(-to-head-to-head) methodology and Bill’s contention that traditional polls present a false picture. I’d agree though that he could stand to do more explaining on those points, for instance on his confidence that Biden is weak. I gather that he believes that his method, were it reaching and polling a representative sample, would reveal Biden as much weaker than the old fogey polls do. But I still don’t get why that would be true.
8:12 AM Aug 19th
 
OldBackstop
I agree I would welcome a return to regular HeyBills.

The "brigading" if that is a term, is not one-off incident, it is a chronic condition. A twitter poll may get 5 percent return, a number cited by Bill. A bump of 2000 votes, then, might indicate a retweet to a group of supporters of 20,000.

Retweets may be visible in a poll....attention drawn by the various methods used in campaigns would not be...emails, Instagram's, etc. And the bump might be 150, not 2000....enough to put a big thumb on the scale.

Just my dipshtt opinion.


9:16 PM Aug 18th
 
meandean
Bill, clearly the Yang Gang was brigading, but at least based on what I know right now, I'm not at all sure that that's what's going on in the other scenarios. I think most of your followers are "doctrinaire" Democrats, and thus there is an advantage in each poll for the biggest-name mainstream Democrat. So, you have:

- Pre-brigading, totally mainstream Dem Beto was overperforming against Trump (Repub), Williamson (a very unconventional Dem), and Yang (a very unconventional Dem). [Additionally, Beto got attention and praise from Democrats based on his reaction to the shooting in his hometown.]

- Totally mainstream Dem Castro overperformed against Gabbard (who is far more pro-Russia and pro-Trump than other Democrats), Bullock (who has ittle name recognition, and you also named a Deadwood character instead of the actual guy ;), and Steyer (who has little name recognition and no government experience.)

- Totally mainstream Dem Gillibrand overperformed against Moulton, Delaney, and de Blasio. (I suppose this is the most surprising, but honestly, is it that surprising? Gillibrand's increased support came from Moulton and de Blasio. We've now debated four times without Moulton. Obviously that's an enormous disadvantage for him, especially since no one knows who he is anyway. de Blasio overachieved low expectations in the first debate, but came off poorly in the second debate, with attacks on Biden that were clumsily telegraphed. His campaign just in general is perceived as a joke.)

Obviously, you're tracking the polls more closely than I am, and if you saw obvious examples of tampering -- i.e., there were 1,000 votes for Gillibrand in an hour, or something -- then that's a different story. But at least based on the end results, I don't see that evidence. Again, I theorize that whenever you put a mainstream, well-known, non-polarizing (i.e., not Bernie) Democrat up against three schnooks, that candidate will "overperform" in your polls.
6:00 PM Aug 18th
 
bjames
Patrick--Thanks.
2:35 PM Aug 18th
 
packbringley
Hey Bill, speaking as someone who found this experiment interesting, mostly, I’d just like to express my hope that you revive “Hey Bill.” When it’s rolling it’s a fantastic part of the site, and I feel that twitter doesn’t even come close to replacing it.
2:12 PM Aug 18th
 
 
©2019 Be Jolly, Inc. All Rights Reserved.|Web site design and development by Americaneagle.com|Terms & Conditions|Privacy Policy