July 19 Poll Report

July 19, 2019
 

July 19 Poll Report

            Good morning everybody.  Amy Klobuchar, the biggest name among the four candidates being polled yesterday, had a pretty decent day, outperformed predictions a little bit.   Well, she did and she didn’t.  In order to get "expectations" for the poll, I had to pencil in a number for Tom Steyer, the California billionaire who has recently entered the race.  Since we don’t have a polling history for him, one previous poll, it was just a guess as to how he would do in this poll.   It turned out that my guess wasn’t a particularly good guess, so Steyer got less than half of the 11% vote that I had anticipated for him.  The extra 6% had to go somewhere else, so Klobuchar naturally soaked up most of them, thus "overperformed".   But she didn’t really overperform; she didn’t get any real benefit from it in terms of increasing her Support Score, and the very modest gain that she did make will almost certainly be eroded over time by secondary adjustments.   As Steyer is polled again and again, he is (we can now guess) very likely to settle in at a lower figure, which will make Klobuchar’s performance in yesterday’s poll slightly less impressive, which will lower her score.  She gained 8 points yesterday (.08 of one percent); she’ll probably lose those 1 point at a time over the following weeks as Steyer is re-evaluated, and thus, yesterday’s poll is re-evaluated by what we will now know then, later on.   I hope that makes sense.  Anyway, this is yesterday’s poll result:

Scores

Williamson

90

Steyer

80

Klobuchar

449

Bullock

98

 

 

 

 

 

 

 

 

 

Predicted

Williamson

13

Steyer

11

Klobuchar

63

Bullock

14

Actual

Williamson

16

Steyer

5

Klobuchar

67

Bullock

12

 

            The predictions for the poll were 84% accurate, with most of the error caused by the fact that my estimate of Steyer’s starting point in the race was too high.  These are the changes in the standings which have occurred since yesterday: 

            Joe Biden is up 26 points as the result of the removal of the May 29th poll data from the data now considered relevant.

            Beto O’Rourke is up 17 points as a result of the removal of the May 29th poll data.   The May 29th poll was a poll of heavyweights—Biden, Sanders, Buttigieg and O’Rourke.   O’Rourke at that time, the morning of May 29, had a Support Score of 558.   Based on that, he was projected to get 18% in that poll.  He actually got only 10%, with the remaining 8% being distributed pretty evenly among the other three candidates.   That really was the start of his serious decline in the polls—this poll and others. 

            But, while Beto has not rallied in the polls since then, he is not actually QUITE that low, so removing that poll from the data helps Beto and Biden, and hurts Sanders and Buttigieg a little bit.

            Amy Klobuchar is up 8 points as a result of yesterday’s poll.

            Steve Bullock is down 6 points as a result of yesterday’s poll.

            Bernie Sanders is down 9 points as the result of the removal of the May 29th poll from the data. 

            And also Elizabeth Warren is up 30 points as a secondary adjustment, but I have decided that I am not going to comment on secondary adjustments anymore.   I have about 20 "position point estimates" from the various candidates, 20 for each candidate, based on where they stand relative to other candidates.  As the scores of the other candidates change, the previous polls are re-evaluated based on the best information we have about the other candidates.  This leads to constant small adjustments in the standings.  I have been commenting on these if they meet my two standards:  at least 6 points of raw movement, and at least 1% relative to where the candidate was yesterday.

            But I am going to stop commenting specifically on those adjustments.  They wash out over time, mostly.  If a candidate is making a significant movement in the polls, a meaningful movement, I can comment on that when he or she is polled or when an old poll is removed.  Elizabeth Warren moves up 30 point (0.3 percent), that’s not really anything on her scale.  Commenting on those things focuses attention on the process of the calculations rather than on the candidates, which was fine when I needed to explain daily the process of the calculations, but not really helpful at this point.  I’m not making those comments anymore. 

            These are the updated Support Scores:

Rank

First

Last

Support

1

Elizabeth

Warren

1972

2

Pete

Buttigieg

979

3

Joe

Biden

956

4

Kamala

Harris

814

5

Donald

Trump

575

6

Amy

Klobuchar

457

7

Andrew

Yang

423

8

Julian

Castro

423

9

Bernie

Sanders

398

10

Beto

O'Rourke

336

11

Cory

Booker

317

12

Kirsten

Gillibrand

269

13

John

Hickenlooper

266

14

Jay

Inslee

259

15

Bill

Weld

233

16

Tulsi

Gabbard

218

17

Michael

Bennet

177

18

Howard

Schultz

167

19

John

Delaney

122

20

Tim

Ryan

116

21

Bill

de Blasio

104

22

Marianne

Williamson

95

23

Steve

Bullock

92

24

Seth

Moulton

83

25

Mike

Gravel

75

 

            Thanks for reading, and for voting. 

 

 
 

COMMENTS (4 Comments, most recent shown first)

OldBackstop
Oh, followup from a RC post....why four? It is already beyond head to head....If you have a concern about spamming AND a pressing need for fresh data, why not five or 6 choices? It would wipe out the bottom 10 of your 25, but......wiping bottoms is good.

heh....just thought of that one.​
3:10 PM Jul 19th
 
OldBackstop
I see. I suppose another alternative would be weighting newer polls?.....

I apologize if I have been a....what was your term....angry Yorkie, but I don't tangle with things unless I am intrigued. I have listened, on behalf of candidates, to various polling systems over the years, evolving through contact eras, so I sort of kick the tires.

My main concern, obviously ad nauseam, is sampling, and I say that with a reason that might not be obvious. This is going to be a fog of war shttstorm sortout on the Dem side. To have a model react to this races virulent, subtle, fast-paced events, it has to look (for a campaign) like it can instantly reflected the nuanced reactions people have to all the crazy spinning tires being thrown into our monkey cage.

I can tell you....I would be very interested in employing a system like this if I were, say, a special interest group PAC in New Hampshire on the Dem side. I would hand you a representative demographic list....maybe even compensated.... and tell go you have at it in a very tight time frame....the surge results telling me where to invest my money amongst the B list. I would see that as a round hole in a round peg for this.

Of course, a twitter poll to an enormous group of largely NH Dems could largely swamp the sampling concerns. I get that -- 1500 responses is a huge number, you are to be commended.

Just realize...political social media wonks and their orcs can smell blood in the water from five miles away. If this gains visibility you will attract disproportionate vertical bursts of interests....like, suddenly Buttigieg gets 8000 votes in a day. I have no idea how you would handle that, but it is a virtual certainty if it gains attention...kind of like your pinned article about analysis bringing destruction.

3:03 PM Jul 19th
 
bjames
OldBackstop

Curious.....why 50 days (is it 50?) in dropping data out? It seems like an awful frenetic time of the Dem timeline right now to be going back that far. It also seems like it causes big swings forcing you to daily explanations.....jes sayin'.

It is too long a time frame, and I am planning to change it to 40 days when I can. You absolutely need at least three polls--9 positioning estimates--to have a clear sense of where everybody is and to let little arbitrary things like listing position in the polls and posting the poll at such an hour that it is seen in California more than it is seen in New Jersey. . . you need multiple polls to let those things even out. Four is much better than 3, and 5 is meaningfully better than four.

My judgment is that if I were to post more polls per day, I would irritate readers to such an extent that I would lose followers and lose poll respondents, so I don't think that is practical.

As candidates drop out of the race those remaining are polled more frequently, and this will allow me to shorten the polling window, reduce the number of polls considered relevant. I think that I need to maintain at least 18 positioning estimates per candidate to have solid numbers, so I need 6 polls per candidate. In about ten days to two weeks I will have enough data available that I will be able to begin closing the window. What I will do is eliminate two polls a day each day, rather than one, so that the number of polls included drops from 50 to 49, 48, 47, etc., down to 40. Probably later, when Cory Booker and Beto O'Rourke and Kirsten Gillibrand and others realize that this isn't going anywhere and they're better off getting out before they get humiliated in Iowa, then I'll be able to shorten the window even more. But I'm still a couple of weeks away from the point at which I can do that.
2:13 PM Jul 19th
 
OldBackstop
"And also Elizabeth Warren is up 30 points as a secondary adjustment, but I have decided that I am not going to comment on secondary adjustments anymore."

My cardiologist thanks you :-)

Curious.....why 50 days (is it 50?) in dropping data out? It seems like an awful frenetic time of the Dem timeline right now to be going back that far. It also seems like it causes big swings forcing you to daily explanations.....jes sayin'. Seems like a canary in a coal mine.

A number of times you have commented on Warren's bright future, and she is up more than 2 to 1 in your scale. But her numbers in the RCP average poll have been basically flat....in fact, only polling 9 percent in the most recent national poll (The Hill) poll, released Wednesday.

Something commented in Reader's Posts....why not ping your base with a simple Trump approval poll from time to time? That would give some hand grenade depth sounding on the political makeup of your respondents.

Final thought.....I think I can crystallize my knee jerk reaction to twitter by wondering why people with apparently no national name recognition are even showing up. My assumption is that if you put four choices in front of a twidiot, and one is "Hickenlooper", they will make a fart noise, giggle, and click it. (Same for Buttigieg and Bullock.) Or vote for Howard Schultz because their childhood dog was named Snoopy. Or....okay, I'm done there.

538 had an interesting (different from you) article this week that sort of weighted polling averages to name recognition.....which I've never seen expressed that way. Name recognition, of course, cuts two ways.....48 percent of Americans wouldn't pull for Hillary with two engines out on the campaign plane.

But in attempting this tallest midget thing in a poll of unvetted respondees, name recognition is the biggest issue at this nascent phase of the campaign, and the sampling numbers are just too tiny. I would be tempted if I were you to sneak in Hank Greenberg or Mordecai Brown and see how the dummy hand plays.

Here is the article if you didn't see it last week:

https://fivethirtyeight.com/features/what-weve-gleaned-about-the-democratic-primary-from-6-months-of-polls/


1:34 PM Jul 19th
 
 
©2019 Be Jolly, Inc. All Rights Reserved.|Web site design and development by Americaneagle.com|Terms & Conditions|Privacy Policy