June 5 Poll Report

Warren is ROLLING

1.     Yesterday’s Poll

Yesterday’s Poll of three tail-enders was "won" by John Delaney.  Delaney outperformed expectations, although not by a large enough margin to improve his standing in the larger system:

Delaney got 50% in a three-person poll, taking 2½% each away from Messam and de Blasio.    I kind of like Delaney; I think he seems like a good guy and I could support him, although I know he is unlikely to be anywhere on a ballot that I’ll have a chance to vote for.   The largest reason that Delaney doesn’t move up is that de Blasio had only been polled two other times, so his position in the polls is relatively fluid compared to those who have been polled more often.   Thus, when de Blasio under-performs, the system in essence says "Oh, I guess de Blasio wasn’t quite as strong as we thought he was," and lowers the number for de Blasio, rather than raising it for Delaney.   Not that the system thought de Blasio was "strong" before; he’s in 27^th place among the 29 people who are being polled. 

2.      Removal of the Old Poll

Removed from the study was the Poll of April 15, which was a very odd and atypical poll, and I’m glad to finally get rid of it; it was kind of screwing up the numbers.  In the April 15 poll John Hickenlooper got 21%, Bernie Sanders 22%--obviously not a normal or accurate impression of the relative strength of those two candidates.  I don’t know what happened in that poll.  Anyway, Elizabeth Warren got 45% in that poll, Tim Ryan 12%.    The removal of that old poll pushes Warren another 100 points ahead of everybody else, and also takes a big whack out of Tim Ryan’s position; that was one of his best polls, getting 12% against two strong Democrats.   That knocks him backward from 132 to 105—a BIG step backward for him.  105 is a low point for Ryan.    Hickenlooper, whose numbers have been dropping very steadily, drops another 40 points there to a new low of 231. 

3.      Warren is Pulling Ahead

I am not saying that my poll means anything more than my followers, but Elizabeth Warren is starting to steamroll people in my polls.  She is gaining strength very regularly, and is pulling well ahead of everybody else. 

And it appears that she is going to pull FURTHER ahead tomorrow.  Today’s poll is the four Democratic front-runners—Warren, Biden, Buttigieg and Harris.   So far, in that poll, Warren is EASILY outperforming expectations, taking votes away from all three of the other contenders, which will push her score even higher tomorrow. 

AND.

And in addition to that, another "stale" poll is going to be removed from the data, and that removal should ALSO help Warren.  In the poll of April 16, Warren got 67%, but 67% against relatively weak opponents.   The current standings suggest that she is STRONGER than that, so the removal of that poll should push Warren even FURTHER ahead tomorrow.  I don’t know what her number will be tomorrow, but it looks like it could be 1600 or more. 

4.      Kasich out, de Blasio, Bullock in

I did  a lot of background work since yesterday, taking John Kasich out of the group and adding in Steve Bullock and Bill of the Blasios.   In yesterday’s poll Kasich was not displayed, but he was there in the background; his position was still being used to help assess the relative positions of everybody else.   Now he is gone; those position indicators have been removed from the data. 

            While I was doing that, I re-centered the system at 10,000 total "points", 10,000 theoretical voters.  Elizabeth Warren’s current score (1437) actually means 14.37% of the vote.   But when I added Stacey Abrams to the poll in Mid-May, she came in at 570 or something.  Rather than take those 570 points away from the other candidates, which would cause a little tremor in the data, I made the system work on 10,570 points, then started taking a few of those points away every day, so that you would never notice it—10,560 one day, 10,550 the next, etc.  

            But then I realized that if I took Kasich OUT of the poll and added DeBlasio and Bullock back in, the cumulative effect of that would be to re-set the system almost exactly to 10,000 voters.   So I took advantage of that coincidence to re-set the system to 10,000 voters, which was its original intent. 

            This seems to have made no difference at all; it was a certain amount of work, and it involved a lot of "cells" in the spreadsheet, but it all seems to have been so consistent with the rest of the data that the change doesn’t have any impact.   Bullock and de Blasio have been added to the polls because they have now been polled three times each, so I have a reasonable degree of confidence in where they stand.   All of the real changes in the standings since yesterday are because of the removal of the April 15 poll.   The other things—yesterday’s poll, the removal of Kasich, the additions of de Blasio and Bullock, the re-centering at 10,000—have very, very small effects on where people stand.

5.      Klobuchar may have caught O’Rourke

Amy Klobuchar has been gaining ground in my polls steadily, while Beto O’Rourke has been losing support.   It appears that Klobuchar, who caught and passed Cory Booker just a day or two ago, has now caught or passed O’Rourke.  I have her at 464.8, Beto at 464.6.    That could easily go the other way tomorrow, but both her gains and O’Rourke’s losses have been steadily mounting for the last month. 

6.      The Current Standings

This is where we are at the moment:

7.      Tango’s Question

I had a question from Tom Tango on Twitter. . .let me see if I can copy it from Twitter:

Bill, if we look at the current scoring (see below), DJT would get 29% of the votes, a near-perfect match to your poll from a few weeks ago (see above). However, if it was head-to-head DJT against ANY of these 3, I'd expect he'd get less than 50% (i.e., vote-splitting). 1/2

          (Snapshot of polls goes here.. . .pretty much the same as above, positions 9 through 17, with highlights on the names and totals of Abrams, Trump, Yang and Gabbard.)   

 Would it make more sense to keep the two lists separate, so that the voting can indeed be predicted based on the scoring, whether it's a 2-, 3-, or N-person voting system?

Me:  What two lists are you referring to? 

 Tom:  Sorry, I meant a Dem list and a Repub. list, with their own points.  Right now, the implication of a DJT v Yang would be 58/42.  But if you were to run a poll with just them two, you’d get 70/30 in favor of Yang. 

            Well, let’s think about it.   I think of this as the asymmetrical warfare problem.   You may have, on the one side, a vast army equipped with hundreds of thousands of big bombs and powerful rifles and big scary weapons of all kinds, and, on the other hand, a bunch of pissed-off locals who don’t have any of those things, but are just tired of being pushed around.   In theory, the rich, powerful army should win every time, but history shows us many, many examples of wars like that that were won by the local guys with pitchforks and rocks. 

            The same problem occurs in sports competitions.   In football, for example, it may be that LSU can beat Arizona State and Arizona State can beat UCLA, but nonetheless UCLA can beat LSU because UCLA may have (theoretical example) UCLA may have three receivers who can run like bats out of hell, and Arizona State may be able to cover those three receivers adequately but LSU might not be.   It’s a rock/paper/scissors problem, in a sense.  

            Your question suggests that this is uniquely a problem of Democrats vs. Republicans, but I would argue that it is a much broader problem than that.  There are in my voters a certain number of people who would strongly prefer to vote for a woman, a certain number of people who would strongly prefer to vote for a man, and a certain number of people who are equally happy to vote for a man or a woman based on other factors.  If you put Amy Klobuchar in a poll with Elizabeth Warren, Kamala Harris and Bernie Sanders, it may well be that the "woman’s vote" would have little effect, so that Sanders would beat Klobuchar (let us say) 30-15, but if you put Klobuchar and Sanders head to head in a two-person poll, it might be that Klobuchar would win 60-40. 

            This is the same problem.   Or it could be that there is a "black" vote and a "white" vote, or it could be that there is a "young" vote and an "old" vote, or it could be that there is a "rural" vote and an "urban" vote.   Let’s call them Asymmetrical Biases, or ABs for short.   There is an asymmetrical bias caused by the fact that there are more Democrats in my polling group than Republicans. 

            If there was only ONE of these AB problems in the data, then we could solve the problem, as you suggest, by isolating the Democrats from the Republicans.   But if there are MANY such AB problems in the data, then removing one of them simply makes the others, that you have not removed, more significant than it was before.   I’m sure you’re familiar with that syndrome from doing research—when you remove confounding influence #1 from the study, you suddenly find that you are overwhelmed by confounding influence #2.  

            What I am interested in is the overall strength of each candidate relative to each other candidate, despite ALL of the confounding influences, despite ALL of the asymmetrical biases. 

            I’m not saying your wrong, really, that that approach might be better.  In the next round of polls I am polling candidates by strength, but in doing that Bill Weld and Donald Trump, the only two Republican actual candidates in the race, would be pitted against one another.  When you pit them against one another you get most of my left-leaning audience going to one of the Democrats, so the foreseeable result of that is that it would get an unsupportable number for the strongest Democrat in that group of four.  I don’t want that to happen, so I break my own rules and move Weld out of the Donald Trump heat, into a separate heat.  This is consistent with what you were saying; it is a way to dodge one effect of the Asymmetrical Bias problem. 

            You are, in a sense, attempting to apply to my system an expectation derived from how other polls work, or, to be frank, don’t work.  I would note that neither my system nor the traditional polling methods gives a reasonable or credible answer to many proposed one-on-one matches, as you suggested for Trump vs. Yang.   In the latest Quinnipiac Poll, Joe Biden is at 38%, Andrew Yang is at 1%, and Kirsten Gillibrand is at ---, which is reported elsewhere as .2%, don’t know whether that is right or wrong.  

            But if you polled Joe Biden against Andrew Yang, do you REALLY think that Biden would win 38-1?   My poll says that Biden would beat Yang 1127 – 240, or 82% to 18%; Quinnipiac says it would be 38 to 1, or 98% to 2%.   Which of those two do you think is more accurate?  

            My answer seems to be one hell of a lot more credible, to me.   I’m not saying that my polling is RIGHT; I am saying that it’s a hell of a lot closer to being right than the National Polls are.  

            Sorry; I kind of lost the thread of my thought there.  I was saying that you are, in a sense, attempting to apply to my system an expectation derived from how other polls work.  There is a sort of assumption that these polls are intended to predict an election, as if there was going to be an election tomorrow that pitted Andrew Yang against Donald Trump or Andrew Yang against Joe Biden.   What I am really saying is that there is not going to be an election tomorrow of Andrew Yang against Joe Biden OR Donald Trump, so what is the point of speculating on how that vote would go?  

            What we CAN do, and what we should be doing at this point, is measuring the relative strength of the candidates, one against another—and this is where I believe that my method is vastly superior to what the national polls are attempting to do.   Quinnipiac has Kirsten Gillibrand at "—" and Andrew Yang at 1%;  Gallup is dodging the issue entirely by running polls about Biden’s "favorability" and has Biden I think at 32%.  Harvard-Harris has Biden at 36%, Yang at 1%, Gillibrand at "—".   Real Clear Politics average of multiple polls has Biden at 34.8%, Yang at 0.8%, Gillibrand at 0.3%.   I have Biden at 11.3%, Gillibrand at 2.7%, Yang at 2.4%. 

            I would argue that mine is the ONLY poll around that has reasonably accurate measurements of the relative strength of the minor candidates—not that Yang might not actually be ahead of Gillibrand or anything like that, but just reasonably accurate numbers for the minor candidates. 

            And, at this time, THAT IS WHAT ACTUALLY MATTERS—not how the candidate would do in some vote tomorrow which is not going to happen, but how much strength does this candidate have at this point in time, relative to the field?    And it matters for this reason.   Suppose that there are, in a field of 25 candidates, 16 candidates who are somewhere around 150 in the rankings, or 1.5% of the vote each.    This is sort of close to the actual situation; there are in fact somewhere around 16 Democratic candidates who are somewhere around 150 in the standings, counting anything between 30 and 300 as somewhere close to 150.

            But suppose that, in the period of two months, half of those 16 move up to 250, and the other half drop down to 50.   Those who drop to 50 are basically dead; the other 8 are still weak compared to the Buttigiegs and the Warrens, but they’re alive.

            Suppose that, in the following two months, 4 of those at 250 move up to 400, while the other four drop back to 100.   Now we are down to four candidates at 400, and the other four are basically dead.  

            Suppose that, in the following two months, 2 of those move up 600, while the other two drop back to 200.

            My point is, history shows many, many, many examples of minor candidates in a crowded field who came on to win a nomination by moving up a little, and up a little, and up a little.   The Democratic Party shot themselves in the foot in 2016 when they straight-jacketed the process to prevent that from happening. What we really SHOULD be watching, right now, is which ones of those minor candidates are growing stronger, and which ones are fading?  

            And mine is the only polling system which can answer that question for you.  That is the point of it; that is the value of it.  I don’t think you can create that kind of value by doing either two-person polls, or 15-person polls.