July 17 Poll Report

July 17, 2019
 

July 17 Poll Report

 

            Good morning everybody.   Yesterday’s poll went 92% as we expected it to go, based on previous polling.   We expected Julian Castro to win the day with an anticipated 44% of the vote, and he did in fact win the day with 44%.  The only thing that happened in that poll was that Kirsten Gillibrand took 3% away from John Delaney and 1% away from Michael Bennet:

Scores

Delaney

131

Gillibrand

261

Bennet

170

Castro

448

 

 

 

 

 

 

 

 

 

Predicted

Delaney

13

Gillibrand

26

Bennet

17

Castro

44

Actual

Delaney

10

Gillibrand

30

Bennet

16

Castro

44

 

            The discrepancies are so small that they don’t really have any meaningful impact on the standings.   Of more significance was the poll that we removed from the data considered relevant, which was the voting of May 27th, which was Buttigieg (70%), Gabbard (13%), Hickenlooper (11%) and Michael Bennet (6%).  The removal of that poll has larger effects.  Since yesterday:

            Michael Bennet is up 13 points, despite the 1-point underperformance yesterday, due to the removal of the May 27th poll from the data.   Bennet has been steadily moving upward in the poll over the last two weeks.  He was at 126 on July 2nd, got 41% against a weak group on the poll on that date, moved up to 133, moved up to 137 as a secondary adjustment on July 6, moved up to 143 as the result of the removal of an old poll on July 7, and moved up to 145 as a secondary adjustment on July 8.  He did well on the poll that was taken on July 8 (he got 27% in a field that included Bernie Sanders), and moved up to 157 as a result of that (July 9), moved up to 168 on July 11, when we removed Jeff Flake, Stacey Abrams, Eric Swalwell and Wayne Messam from the polling group and distributed their Support Points among the remaining candidates.  He went up to 171 on July 15, as a secondary adjustment, and is up to 183 today as the result of the removal of the May 27th poll from the data.  The 1% under-performance in yesterday’s poll doesn’t hurt him to any meaningful extent.  Anyway, he has moved up 57 points in 15 days, which puts him on the Green List for the first time.  I like Bennet, based on what little I know about him; he seems like a sensible person, and based on what I know about him, my impression is that I could vote for him.

            John Hickenlooper is up 11 points as a result of the removal of the May 27th poll.

            Kirsten Gillibrand is up 9 points as a result of doing well in yesterday’s poll.

            John Delaney is down 9 points as a result of under-performing in yesterday’s poll, and

            Pete Buttigieg is down 26 points as a result of the removal of the May 27th poll from the data.  He got 70% on May 27; that’s hard to sustain.  Buttigieg, now at 975 (9.75%) has fallen under 1,000 (10%) for the first time since May 15th.   This is consistent with the post-debate trend, discussed here several times in the past, of the leaders declining, giving away a little bit of support to a variety of lesser-known, less-supported candidates.   Warren is down 7% from her peak score on June 27, the date of the first debate. Biden is down 25% from his peak score, also on June 27, and Harris is down 5% from her peak on July 8th, and is up only 1% since June 27.  Meanwhile, a whopping ten candidates are now marked in green, meaning that each of them is up 33% in the last 30 days:

Rank

First

Last

Support

1

Elizabeth

Warren

1861

2

Pete

Buttigieg

975

3

Joe

Biden

915

4

Kamala

Harris

804

5

Donald

Trump

580

6

Amy

Klobuchar

481

7

Julian

Castro

449

8

Andrew

Yang

414

9

Bernie

Sanders

399

10

Beto

O'Rourke

320

11

Cory

Booker

319

12

Jay

Inslee

300

13

John

Hickenlooper

272

14

Kirsten

Gillibrand

270

15

Bill

Weld

237

16

Tulsi

Gabbard

234

17

Michael

Bennet

183

18

Howard

Schultz

167

19

John

Delaney

122

20

Tim

Ryan

119

21

Bill

de Blasio

104

22

Steve

Bullock

104

23

Marianne

Williamson

91

24

Seth

Moulton

85

25

Mike

Gravel

81

 

            I see what I am doing with these polls as being a logical continuation of what I have done all of my career.  There is a passage in the writings of Harvard’s William James in which he says, in essence, that it is the task of the thinking man to unite the questions of the earth with the questions of the sky; that is, to make a meaningful link between the small, every-day facts of life and the great eternal questions.  Although I didn’t discover that passage in his writings until I was past 50 years old and have now lost track of the citation because I am now way past 50, I think that that is exactly what I have always done.   The Runs Created method is in essence a way of uniting the common events of every game—singles, doubles, triples, walks—with the larger question of the game of baseball:  How are runs scored?  The Pythagorean Method is a way of uniting Runs with the yet larger question of how runs are related to wins.  The original Hall of Fame Monitor, from the 1970s or early 1980s, was/is a system of uniting the common events of a career—hitting .300, driving in 100 runs, etc.—with the greater question of who deserves to be in the Hall of Fame.  

            My criticism of what I call Old Fogey polling is exactly the same:  that it has so little focus on the small questions that it cannot possibly create an accurate portrayal of the bigger picture.  The Old Fogey polling method completely fails to distinguish the support of Marianne Williamson from that of Tim Ryan or John Hickenlooper.  They poll 800 people, mostly people who, pardon my French, don’t give a shit about the candidates, and 790 of the 800 will recall the name of Joe Biden or Bernie Sanders or Elizabeth Warren or Kamala Harris or maybe Amy Klobuchar, with just a trace of the vote left to be divided among Jay Inslee and Bill de Blasio and Steve Bullock.  They don’t have any idea how these candidates stand with regard to one another, who is moving up, who is moving down, who is stronger, who is weaker. 

            My method draws 1,000 votes a day divided among four candidates, so that the distinctions between relatively weak, quite weak, relatively strong, moderately weak but gaining, moderately weak but losing. . .so that these distinctions become clear, and I actually do know how Michael Bennet stands compared to Tulsi Gabbard.   To the best of my knowledge, mine is the only poll that does that, and the people who read my poll reports are the only people who actually know that.  This is what I have done all of my career:  focus on the exact answers to the small questions, in such a manner that they help us to understand the bigger picture.   Thanks for reading. 

 

 
 

COMMENTS (11 Comments, most recent shown first)

MattGoodrich
I agree with OldBackstop. I've taken a few statistics classes over the years and find they spend 99% of the time on 'mathematical mechanics'. Standard deviations, means testing, sample sizes, confidence levels, margins of error. The general assumption is that the data is valid. But the data is never perfect and often it isn't even very good. A 'random' phone poll gives you a sample of people who are home, people with phones, people who answer their phones, people who will talk to a pollster. That's not random. It's probably disproportionately old people.

I think the methodology of this poll is interesting, but it probably won't be particularly useful in predicting the actual voting.




11:17 AM Jul 18th
 
OldBackstop
Well, as to the Yang Wave, maybe 85% of Bill's followers are the Yang Family. You just don't know.

But in the real world of politics, he is at 1.4 percent, which is within the margin or error of having died in 2011.

https://www.realclearpolitics.com/epolls/2020/president/us/2020_democratic_presidential_nomination-6730.html
12:02 AM Jul 18th
 
bertrecords
I think the major criticism is that polling isn't representative of some well defined group such as Iowans, likely to vote.

I think the major strength is that, especially for less well known candidates, trends can be detected.

Why this is important is that typically running for president is done on a national stage. And, typically, if Group A (Twitter poll) is trending one direction then Group B (Iowans likely to vote) is trending that same direction.

Recognizing the limitations of the polling-- I think everyone here does reognize limitations-- I nonetheless find the polling informative.

For example, this methodology chronicles the rise of Yang, which based on the world where I live seems to be a real phenomenon. I am not sure that "old fogey polls" do that, or if they have, I didn't detect it.
10:18 PM Jul 17th
 
MarisFan61
Actually I can see that part of it fine, if we’re only looking at this as an indication of the population in these polls — and, if the method were used on a more general population, it would be fine for that. Of course this leaves the other issues....
9:31 PM Jul 17th
 
OldBackstop
Seeding based on opinions of evolving anonymous random laymen?
7:57 PM Jul 17th
 
Gfletch
The so called Old Fogey polls assume everyone is running in the same race at the same time. Bill's method is sort of like seeding a tennis match without having the players already seeded or at risk of being knocked out of the tournament.

Or like any other sporting playoff system. One could say the old polls are like the regular season, Bill's method are like the playoffs, or like being able to see how candidates might do if paired up against each other, or at least within small, easier to handle groups.

But...I may be wrong.?
6:01 PM Jul 17th
 
OldBackstop
Everything Maris said.

I've been commissioning political polling for 30 years. I'm, frankly, saddened to see such a huge amount of your talent going toward something that....sorry, looks a little tinfoil hatty.

You either have better data on your twitter following or are vastly overestimating them as a representative group of.....anything. They are probably 90% white, 99 percent male, and 50 percent Massachusetts-based (take that as a compliment to your historic work with the Sox.)

A good poll to estimate support in the run up to an election first focuses on whether the people they are polling are going to actually be involved in the election. The best way to do this is go to likely voters, established by either phone interviews or their actual voting records (i.e.: have voted in three of four of the last elections.) That is easily done. People tend to be voters or not. Ask Bernie Sanders, where his alleged popularity wound up out on the quad in a drum circle and forgot to vote. Or registered as a Dem. Or even know if they are.

Your twitter following is not a random sample. You saying last week that your sampling is becoming more conservative due to a friend retweeting demonstrates that I am not introducing that concept to you. But you need a few black female friends to do so. And even then it is just a guess.

And, by the way, mixing them daily with HoF polls is just further tightening your myopic sampling to baseball fans, because people wander away from constant polling in which they have no interest..

Like Maris, I was hoping were working on a methodology rather than feeling your real results had real world value.

So I guess I don't get it....so what? White male Red Sox fans like Massachusetts Senator Elizabeth Warren 2 to 1? No sheet. Polling methods can be brilliant, but it is all about the sample.....garbage in, garbage out.

It's as if......you are spending 90 percent of your baseball team's practices finetuning that they all know how to keep a proper scorebook. But you aren't doing the hard work, the fielding and batting.

And that is all about valid samples.


5:31 PM Jul 17th
 
shthar
"people, mostly people who, pardon my French, don’t give a shit about the candidates"

It's a good thing so few of these people actually vote...
4:13 PM Jul 17th
 
MarisFan61
At risk of being seen the way I'm often seen here :-) I'm going to question some of the key aspects.

As I've implied since early on with my comments about "what is Bill really doing here," in which I concluded that it's probably mostly to try out a methodology without much actual interest in the current subject matter, I figured it was unlikely to be about the exact subject matter because it has seemed pretty clearly not to be doing the kinds of things you're saying here that it does.

Like:

"The Old Fogey polling method completely fails to distinguish the support of Marianne Williamson from that of Tim Ryan or John Hickenlooper."

So?

"They don’t have any idea how these candidates stand with regard to one another, who is moving up, who is moving down, who is stronger, who is weaker."

Let me say first that the "They" is unclear in whether it refers to the fogey polling methods or to the people in their polls, but assuming it's the former: Sure they do, in the largest respect, which is whether they're the #1 picks of anybody -- and it seems to me that even with regard to the more detailed aspects, such results are fairly indicative too, which is related to what I'm going to say about the next thing...

"My method draws 1,000 votes a day divided among four candidates, so that the distinctions between relatively weak, quite weak, relatively strong, moderately weak but gaining, moderately weak but losing. . .so that these distinctions become clear, and I actually do know how Michael Bennet stands compared to Tulsi Gabbard"

I question the meaningfulness of what is shown about those distinctions, and I don't think at all that the distinctions become "clear." First of all, let's remind ourselves (I know that you don't need to be reminded; this is just 'for the record') that the results here are only from the sample of who's voting on your Twitter account, which doesn't represent the broad public and which is unrepresentative in various ways, some unclear but others very clear. An example of the latter is that the people in these polls of yours are more knowledgeable than average about the candidates; I think a likely additional way is that they're probably more apt than average to go and look a little into who Bennet and Gabbard and Hickenlooper really are. Plus, as we've seen as as you have noted many times, the political tendencies of your group is pretty far off from what's out there. So, I don't see how what these people are saying about those sorts of candidates says anything clear about how they stand or how they're moving.

If you mean something a little different -- like, not that the "clearly" thing is about where the candidates stand and how they're doing in terms of "out there" but rather that it would show it clearly if this method were used for the major polls, I'd say it brings us back to what I said for the above excerpts. I think the essences of the relevant movements are already shown by the candidates' basic poll numbers, and that when any of the lower-tier candidates may start breaking into more moderate territory, then they would be given greater attention in general and they'd be put to more meaningful tests -- and, I don't see that these candidates' showings among your group, knowing more than average about the candidates and probably looking into them more, have any relation to how any of these candidates would fare and would be seen if they do happen to break out and be put under a bigger spotlight. If they don't break out from just tiny numbers in the fogey polls, it doesn't matter how they are seen vs. one another, unless we are to believe that the kinds of inner things shown in your polls (either here or in the hypothetical application of them to general polling) are indicative or predictive for breaking out or not breaking out, and for the reasons I've stated, I wouldn't at all think that they are.

Cliff's Notes: I don't see that the fogey polls majorly fail in the ways that you say they do, and I don't see how this method "clearly" shows the things that you say it does.

Don't get me wrong: It's a very interesting method and I think it could be well applied to other kinds of things -- I was guessing that you were aiming mainly to use it for some baseball stuff and that this candidate stuff was mostly just a convenient platform for it -- and the results you've been getting are very interesting. But it leaves me with the doubts I've stated.
3:20 PM Jul 17th
 
DavidTodd
agree, last three paragraphs are good
2:21 PM Jul 17th
 
Gfletch
Thank you for those last three paragraphs.
1:02 PM Jul 17th
 
 
©2019 Be Jolly, Inc. All Rights Reserved.|Web site design and development by Americaneagle.com|Terms & Conditions|Privacy Policy