Poll Report May 22, 2019

May 22, 2019
 

Poll Report May 22, 2019

 

None of the Above

 

            Every day, when I do my polls, a few people are kind enough to tell me that I need a "None of the Above option."   Well, no, actually, I don’t need a "None of the Above" option; I don’t need that, at all.  If you think that the poll needs that, that’s because you don’t understand what I am doing with these polls, and I would anticipate that you don’t.   It’s new and different and it has never been done before, so it is natural that people are going to confuse it with the things which HAVE been done for the last century or so, and which are still being done.   But I don’t need a "None of the Above" option, and let me explain why. 

            Because we already know.   We’re just polling four people at a time.  There are 30 candidates on my list now; I assume you’re not going to vote for 29 of them.  There might be three or four people that you COULD vote for, but there are probably 26 or 27 that you wouldn’t.  I know this.   The system knows this. 

            Each poll that I post is not a free-standing event, as it would be if it were being done by Quinnipiac, Marist, Gallup, or some other group of educated, reputable pollsteristas.   Each poll simply feeds more information into the network.   The network doesn’t need to be TOLD that 99% of you don’t want to vote for Wayne Messam, Marrianne Williamson, John Delaney, Mike Gravel, Francis Underwood or Everett Dirksen, because it already knows that.   Telling it that doesn’t add ANY information to what it already has.  

            Each polling group is used to re-establish the position of those four candidates relative to the other people in that poll.   It’s a point-conservation method; points are being moved around among the four people in the poll.  If it was a weak group going in, it will be a weak group coming out.  If it was a strong group going in, it will be a strong group coming out.   It doesn’t change how that group stands relative to the other 26 candidates.

            Well, it does, but not very much.   Mathematically, this is what we’re doing.   In yesterday’s poll, Amy Klobuchar went in at 410 points, and Eric Swallwell went in at 118.   We call these the "initial values".   What the initial value is doesn’t matter at all, at all, at all, for reasons that I will explain later, but Klobuchar and Swallwell come in at 410 and 118.  There are 528 points to be shared between them.   Klobuchar beat Swallwell in this poll, 42 to 6, which is a 7-to-1 ratio.  We thus divide the 528 points that Swallwell and Klobuchar hold between them in a ratio of 42 to 6, or 7 to 1.   528 times 42, divided by 48, is 462, so the 42 to 6 ratio is 462 to 66.   Klobuchar gets a "position point" for that matchup of 462, and Swallwell gets a position point of 66. 

            Klobuchar came in at 410 and Cory Booker at 453, which is 863 points.   Booker beat Klobuchar in the poll 43 to 42, which is 43-42 ratio (duh).   When you split 863 points in a 43-42 ratio, you get 437 points for Booker, and 426 for Klobuchar.  Klobuchar gets a score from that matchup of 426.

            In each poll of four candidates  there are six comparisons like that (A to B, A to C, A to D, B to C, B to D, and C to D.)  Each of those six comparisons results in two "position points", so there are 12 position points resulting from each four-person poll.   I have done 42 polls so far (the 43rd is running right now) so that results in 504 position points, each one of those estimating the position of one candidate relative to another, and, inferring from that, relative to the field as a whole.   

            If we used one of the four spaces in a twitter poll to collect "None of the Above" votes, that would cut the number of comparisons in the poll from 6 to 3 (A to B, A to C, and B to C), thus reducing the number of positioning estimates that result from each poll from 12 to 6.   That is ALL that it would do; it wouldn’t do anything at all to make the specific candidates being polled that day stronger or weaker; it would merely reduce the amount of information coming out of the poll.   Rather than 504 estimates of how one candidate is positioned relative to another, I would have 252.   That would make the current estimates dramatically less accurate, less meaningful, but otherwise the same. 

 

How the Math Works

            To finish the explanation of the math, Amy Klobuchar has now been polled eight times.  Since we get three "relative position points" from each poll for each candidate, we now have 24 different estimates of where she stands relative to the field of candidates.   Two of those 24 estimates give her scores in the 500s, 13 of them give her scores in the 400s, and the other 9 give her scores in the high 300s.   Adding these three from yesterday to the group (which makes 24), we then re-calculate her average, which was 410 before.   After one round of re-calculations, she goes up from 410 to 414, while Eric Swallwell goes down from 118 to 114. 

            In the second round of calculations Klobuchar has an initial value of 414, and Swallwell an initial value of 114.   This makes a total of 528, just as it did before, and we split those, just as we did before, in the ratio of 42 to 6, which (again) makes 462 for Klobuchar, 66 for Swallwell.   Sometimes, in a four-person poll, one person will go down 7 points, another person goes up 2, another up 2, the fourth person up 3.   Then the total (528, as here) won’t be exactly the same, so the candidate scores different in the second re-calculation than he/she did in the first.    You use the output calculations from the second re-calculation as the input scores for a third re-calculation, etc.   After about three rounds of re-calculation the scores aren’t really moving anymore, but I do 60, 70 rounds of re-calculation every day anyway; you reach the point at which the input scores are the same as the output scores through 14 or 15 decimal points.  

            If you can establish the relative standing of A to B and the relative standing of B to C and the relative standing of C to D, then you can establish the relative standing of A to D, even though you have not put A and D in the same poll.   To explain why it doesn’t matter what the starting positions are. . . suppose that you have four candidates:  Dwight Eisenhower, John F. Kennedy, Lyndon B. Johnson, and Richard Nixon.   These are their starting values—100 points each:

 

 

Ike

JFK

LBJ

Nixon

 

100.0

100.0

100.0

100.0

 

            We have just four polls to place them relative to one another, and we have just two Presidents in each poll.   In Poll #1, JFK beats Nixon, 58 to 42.   There are 200 points to be divided; we divide them 58-42, that makes a score of 116 for Kennedy, 84 for Nixon:

 

Ike

JFK

LBJ

Nixon

 

100.0

100.0

100.0

100.0

Poll 1

 

116.0

 

84.0

 

              In Poll #2, Eisenhower beats LBJ, 59 to 41.  We split those 200 points, then, in a 59-41 ratio, which makes 118 to 82. 

 

Ike

JFK

LBJ

Nixon

 

100.0

100.0

100.0

100.0

Poll 1

 

116.0

 

84.0

Poll 2

118.0

 

82.0

 

 

 

              In Poll #3, John F. Kennedy is compared to LBJ, and beats him 68 to 32.  We split those 200 points 68-32, which makes 136-64. 

 

Ike

JFK

LBJ

Nixon

 

100.0

100.0

100.0

100.0

Poll 1

 

116.0

 

84.0

Poll 2

118.0

 

82.0

 

Poll 3

 

136.0

64.0

 

 

 

            In Poll #4, Eisenhower beats Nixon, 64 to 36.  We split those 200 points 64-36, which makes 128-72:

 

Ike

JFK

LBJ

Nixon

 

100.0

100.0

100.0

100.0

Poll 1

 

116.0

 

84.0

Poll 2

118.0

 

82.0

 

Poll 3

 

136.0

64.0

 

Poll 4

128.0

 

 

72.0

 

 

            We now have an average of the two polls for each of the four candidates—123 for Eisenhower, 126 for Kennedy, 73 for Johnson, and 78 for Nixon:

 

 

Ike

JFK

LBJ

Nixon

 

100.0

100.0

100.0

100.0

Poll 1

 

116.0

 

84.0

Poll 2

118.0

 

82.0

 

Poll 3

 

136.0

64.0

 

Poll 4

128.0

 

 

72.0

Average

123.0

126.0

73.0

78.0

 

 

            We then repeat the process, using these averages as the input numbers.   In the second round, we get output numbers of 122.1 for Eisenhower, 126.8 for Kennedy, 72.0 for Johnson, and 79.0 for Nixon:

 

 

Ike

JFK

LBJ

Nixon

 

100.0

100.0

100.0

100.0

Poll 1

 

116.0

 

84.0

Poll 2

118.0

 

82.0

 

Poll 3

 

136.0

64.0

 

Poll 4

128.0

 

 

72.0

Average

123.0

126.0

73.0

78.0

Cycle 2

Ike

JFK

LBJ

Nixon

 

123.0

126.0

73.0

78.0

Poll 1

 

118.3

 

85.7

Poll 2

115.6

 

80.4

 

Poll 3

 

135.3

63.7

 

Poll 4

128.6

 

 

72.4

Average

122.1

126.8

72.0

79.0

 

 

            We repeat the process a third time, using the output numbers (the averages) from the second round as the input figures for the third round:

 

Ike

JFK

LBJ

Nixon

 

100.0

100.0

100.0

100.0

Poll 1

 

116.0

 

84.0

Poll 2

118.0

 

82.0

 

Poll 3

 

136.0

64.0

 

Poll 4

128.0

 

 

72.0

Average

123.0

126.0

73.0

78.0

Cycle 2

Ike

JFK

LBJ

Nixon

 

123.0

126.0

73.0

78.0

Poll 1

 

118.3

 

85.7

Poll 2

115.6

 

80.4

 

Poll 3

 

135.3

63.7

 

Poll 4

128.6

 

 

72.4

Average

122.1

126.8

72.0

79.0

Cycle 3

Ike

JFK

LBJ

Nixon

 

122.1

126.8

72.0

79.0

Poll 1

 

119.4

 

86.5

Poll 2

114.6

 

79.6

 

Poll 3

 

135.2

63.6

 

Poll 4

128.7

 

 

72.4

Average

121.6

127.3

71.6

79.4

 

 

            We repeat the process a fourth time, using the output numbers from the third round as the input numbers in the fourth round.  

 

 

Ike

JFK

LBJ

Nixon

 

100.0

100.0

100.0

100.0

Poll 1

 

116.0

 

84.0

Poll 2

118.0

 

82.0

 

Poll 3

 

136.0

64.0

 

Poll 4

128.0

 

 

72.0

Average

123.0

126.0

73.0

78.0

Cycle 2

Ike

JFK

LBJ

Nixon

 

123.0

126.0

73.0

78.0

Poll 1

 

118.3

 

85.7

Poll 2

115.6

 

80.4

 

Poll 3

 

135.3

63.7

 

Poll 4

128.6

 

 

72.4

Average

122.1

126.8

72.0

79.0

Cycle 3

Ike

JFK

LBJ

Nixon

 

122.1

126.8

72.0

79.0

Poll 1

 

119.4

 

86.5

Poll 2

114.6

 

79.6

 

Poll 3

 

135.2

63.6

 

Poll 4

128.7

 

 

72.4

Average

121.6

127.3

71.6

79.4

Cycle 4

Ike

JFK

LBJ

Nixon

 

121.6

127.3

71.6

79.4

Poll 1

 

119.9

 

86.8

Poll 2

114.0

 

79.2

 

Poll 3

 

135.3

63.7

 

Poll 4

128.7

 

 

72.4

Average

121.4

127.6

71.4

79.6

 

 

            After ten cycles of repetition the scores are not moving, and they will never move again, no matter how many cycles you repeat the process:

 

 

Ike

JFK

LBJ

Nixon

 

100.0

100.0

100.0

100.0

Poll 1

 

116.0

 

84.0

Poll 2

118.0

 

82.0

 

Poll 3

 

136.0

64.0

 

Poll 4

128.0

 

 

72.0

Average

123.0

126.0

73.0

78.0

Cycle 2

Ike

JFK

LBJ

Nixon

 

123.0

126.0

73.0

78.0

Poll 1

 

118.3

 

85.7

Poll 2

115.6

 

80.4

 

Poll 3

 

135.3

63.7

 

Poll 4

128.6

 

 

72.4

Average

122.1

126.8

72.0

79.0

Cycle 3

Ike

JFK

LBJ

Nixon

 

122.1

126.8

72.0

79.0

Poll 1

 

119.4

 

86.5

Poll 2

114.6

 

79.6

 

Poll 3

 

135.2

63.6

 

Poll 4

128.7

 

 

72.4

Average

121.6

127.3

71.6

79.4

Cycle 4

Ike

JFK

LBJ

Nixon

 

121.6

127.3

71.6

79.4

Poll 1

 

119.9

 

86.8

Poll 2

114.0

 

79.2

 

Poll 3

 

135.3

63.7

 

Poll 4

128.7

 

 

72.4

Average

121.4

127.6

71.4

79.6

Cycle 5

Ike

JFK

LBJ

Nixon

 

121.4

127.6

71.4

79.6

Poll 1

 

120.2

 

87.0

Poll 2

113.8

 

79.1

 

Poll 3

 

135.3

63.7

 

Poll 4

128.6

 

 

72.3

Average

121.2

127.8

71.4

79.7

Cycle 6

Ike

JFK

LBJ

Nixon

 

121.2

127.8

71.4

79.7

Poll 1

 

120.3

 

87.1

Poll 2

113.6

 

78.9

 

Poll 3

 

135.4

63.7

 

Poll 4

128.6

 

 

72.3

Average

121.1

127.9

71.3

79.7

Cycle 7

Ike

JFK

LBJ

Nixon

 

121.1

127.9

71.3

79.7

Poll 1

 

120.4

 

87.2

Poll 2

113.5

 

78.9

 

Poll 3

 

135.5

63.7

 

Poll 4

128.5

 

 

72.3

Average

121.0

127.9

71.3

79.7

Cycle 8

Ike

JFK

LBJ

Nixon

 

121.0

127.9

71.3

79.7

Poll 1

 

120.4

 

87.2

Poll 2

113.5

 

78.9

 

Poll 3

 

135.5

63.8

 

Poll 4

128.5

 

 

72.3

Average

121.0

128.0

71.3

79.7

Cycle 9

Ike

JFK

LBJ

Nixon

 

121.0

128.0

71.3

79.7

Poll 1

 

120.5

 

87.2

Poll 2

113.5

 

78.8

 

Poll 3

 

135.5

63.8

 

Poll 4

128.5

 

 

72.3

Average

121.0

128.0

71.3

79.8

Cycle 10

Ike

JFK

LBJ

Nixon

 

121.0

128.0

71.3

79.8

Poll 1

 

120.5

 

87.3

Poll 2

113.4

 

78.8

 

Poll 3

 

135.5

63.8

 

Poll 4

128.5

 

 

72.3

Average

120.9

128.0

71.3

79.8

Cycle 11

Ike

JFK

LBJ

Nixon

 

120.9

128.0

71.3

79.8

Poll 1

 

120.5

 

87.3

Poll 2

113.4

 

78.8

 

Poll 3

 

135.5

63.8

 

Poll 4

128.4

 

 

72.3

Average

120.9

128.0

71.3

79.8

Cycle 12

Ike

JFK

LBJ

Nixon

 

120.9

128.0

71.3

79.8

Poll 1

 

120.5

 

87.3

Poll 2

113.4

 

78.8

 

Poll 3

 

135.5

63.8

 

Poll 4

128.4

 

 

72.2

Average

120.9

128.0

71.3

79.8

Cycle 13

Ike

JFK

LBJ

Nixon

 

120.9

128.0

71.3

79.8

Poll 1

 

120.5

 

87.3

Poll 2

113.4

 

78.8

 

Poll 3

 

135.5

63.8

 

Poll 4

128.4

 

 

72.2

Average

120.9

128.0

71.3

79.8

 

 

            JFK winds up at 128.0, Ike at 120.0, Nixon at 79.8 and LBJ at 71.3; that’s all the information there is in those polls.  Actually, the numbers are still moving here, but they are moving on the second or third digit.   We repeat the process 60 or 70 times so that they are only moving on some very, very out-there digit.

            Suppose, however, that for the first cycle, rather than giving each President 100 points, you give all 400 points to Richard Nixon.  In the first round of the process, this seems to make a difference:

 

Ike

JFK

LBJ

Nixon

 

 

 

 

400.0

Poll 1

 

232.0

 

168.0

Poll 2

0.0

 

0.0

 

Poll 3

 

0.0

0.0

 

Poll 4

256.0

 

 

144.0

Average

128.0

116.0

0.0

156.0

 

 

            Although Nixon starts out with all 400 points, in the first round he loses to JFK, 58-42, so JFK gets 232 of those 400 points, Nixon the other 168.   After one round, Nixon is down to 156.0 points.  After two rounds, he is down to 108.2:

 

Ike

JFK

LBJ

Nixon

 

 

 

 

400.0

Poll 1

 

232.0

 

168.0

Poll 2

0.0

 

0.0

 

Poll 3

 

0.0

0.0

 

Poll 4

256.0

 

 

144.0

Average

128.0

116.0

0.0

156.0

Cycle 2

Ike

JFK

LBJ

Nixon

 

128.0

116.0

0.0

156.0

Poll 1

 

157.8

 

114.2

Poll 2

75.5

 

52.5

 

Poll 3

 

78.9

37.1

 

Poll 4

181.8

 

 

102.2

Average

128.6

118.3

44.8

108.2

 

            And after three rounds, Nixon is down to 90.2:

 

Ike

JFK

LBJ

Nixon

 

 

 

 

400.0

Poll 1

 

232.0

 

168.0

Poll 2

0.0

 

0.0

 

Poll 3

 

0.0

0.0

 

Poll 4

256.0

 

 

144.0

Average

128.0

116.0

0.0

156.0

Cycle 2

Ike

JFK

LBJ

Nixon

 

128.0

116.0

0.0

156.0

Poll 1

 

157.8

 

114.2

Poll 2

75.5

 

52.5

 

Poll 3

 

78.9

37.1

 

Poll 4

181.8

 

 

102.2

Average

128.6

118.3

44.8

108.2

Cycle 3

Ike

JFK

LBJ

Nixon

 

128.6

118.3

44.8

108.2

Poll 1

 

131.4

 

95.2

Poll 2

102.3

 

71.1

 

Poll 3

 

110.9

52.2

 

Poll 4

151.6

 

 

85.3

Average

127.0

121.2

61.7

90.2

 

 

            When we start out with all 400 points assigned to Nixon, it takes more cycles of repetition to get to the correct numbers, but after 15 cycles the numbers will arrive at exactly the same numbers we had before, when everybody started out even:

 

 

Ike

JFK

LBJ

Nixon

 

 

 

 

400.0

Poll 1

 

232.0

 

168.0

Poll 2

0.0

 

0.0

 

Poll 3

 

0.0

0.0

 

Poll 4

256.0

 

 

144.0

Average

128.0

116.0

0.0

156.0

Cycle 2

Ike

JFK

LBJ

Nixon

 

128.0

116.0

0.0

156.0

Poll 1

 

157.8

 

114.2

Poll 2

75.5

 

52.5

 

Poll 3

 

78.9

37.1

 

Poll 4

181.8

 

 

102.2

Average

128.6

118.3

44.8

108.2

Cycle 3

Ike

JFK

LBJ

Nixon

 

128.6

118.3

44.8

108.2

Poll 1

 

131.4

 

95.2

Poll 2

102.3

 

71.1

 

Poll 3

 

110.9

52.2

 

Poll 4

151.6

 

 

85.3

Average

127.0

121.2

61.7

90.2

Cycle 4

Ike

JFK

LBJ

Nixon

 

127.0

121.2

61.7

90.2

Poll 1

 

122.6

 

88.8

Poll 2

111.3

 

77.3

 

Poll 3

 

124.3

58.5

 

Poll 4

139.0

 

 

78.2

Average

125.1

123.5

67.9

83.5

Cycle 5

Ike

JFK

LBJ

Nixon

 

125.1

123.5

67.9

83.5

Poll 1

 

120.0

 

86.9

Poll 2

113.9

 

79.2

 

Poll 3

 

130.1

61.2

 

Poll 4

133.5

 

 

75.1

Average

123.7

125.1

70.2

81.0

Cycle 6

Ike

JFK

LBJ

Nixon

 

123.7

125.1

70.2

81.0

Poll 1

 

119.5

 

86.6

Poll 2

114.4

 

79.5

 

Poll 3

 

132.8

62.5

 

Poll 4

131.0

 

 

73.7

Average

122.7

126.2

71.0

80.1

Cycle 7

Ike

JFK

LBJ

Nixon

 

122.7

126.2

71.0

80.1

Poll 1

 

119.6

 

86.6

Poll 2

114.3

 

79.4

 

Poll 3

 

134.1

63.1

 

Poll 4

129.8

 

 

73.0

Average

122.1

126.9

71.3

79.8

Cycle 8

Ike

JFK

LBJ

Nixon

 

122.1

126.9

71.3

79.8

Poll 1

 

119.9

 

86.8

Poll 2

114.1

 

79.3

 

Poll 3

 

134.7

63.4

 

Poll 4

129.2

 

 

72.7

Average

121.6

127.3

71.3

79.7

Cycle 9

Ike

JFK

LBJ

Nixon

 

121.6

127.3

71.3

79.7

Poll 1

 

120.1

 

87.0

Poll 2

113.8

 

79.1

 

Poll 3

 

135.1

63.6

 

Poll 4

128.9

 

 

72.5

Average

121.4

127.6

71.3

79.7

Cycle 10

Ike

JFK

LBJ

Nixon

 

121.4

127.6

71.3

79.7

Poll 1

 

120.2

 

87.1

Poll 2

113.7

 

79.0

 

Poll 3

 

135.3

63.7

 

Poll 4

128.7

 

 

72.4

Average

121.2

127.7

71.3

79.7

Cycle 11

Ike

JFK

LBJ

Nixon

 

121.2

127.7

71.3

79.7

Poll 1

 

120.3

 

87.1

Poll 2

113.6

 

78.9

 

Poll 3

 

135.4

63.7

 

Poll 4

128.6

 

 

72.3

Average

121.1

127.9

71.3

79.7

Cycle 12

Ike

JFK

LBJ

Nixon

 

121.1

127.9

71.3

79.7

Poll 1

 

120.4

 

87.2

Poll 2

113.5

 

78.9

 

Poll 3

 

135.4

63.7

 

Poll 4

128.5

 

 

72.3

Average

121.0

127.9

71.3

79.7

Cycle 13

Ike

JFK

LBJ

Nixon

 

121.0

127.9

71.3

79.7

Poll 1

 

120.4

 

87.2

Poll 2

113.5

 

78.9

 

Poll 3

 

135.5

63.8

 

Poll 4

128.5

 

 

72.3

Average

121.0

128.0

71.3

79.7

Cycle 13

Ike

JFK

LBJ

Nixon

 

121.0

128.0

71.3

79.7

Poll 1

 

120.5

 

87.2

Poll 2

113.5

 

78.8

 

Poll 3

 

135.5

63.8

 

Poll 4

128.5

 

 

72.3

Average

121.0

128.0

71.3

79.8

Cycle 14

Ike

JFK

LBJ

Nixon

 

121.0

128.0

71.3

79.8

Poll 1

 

120.5

 

87.2

Poll 2

113.4

 

78.8

 

Poll 3

 

135.5

63.8

 

Poll 4

128.5

 

 

72.3

Average

120.9

128.0

71.3

79.8

Cycle 15

Ike

JFK

LBJ

Nixon

 

120.9

128.0

71.3

79.8

Poll 1

 

120.5

 

87.3

Poll 2

113.4

 

78.8

 

Poll 3

 

135.5

63.8

 

Poll 4

128.4

 

 

72.3

Average

120.9

128.0

71.3

79.8

 

              These are exactly the same numbers we arrived at when everybody started out with 100 points.   Those are the only numbers that you can arrive at, if you’re splitting 400 points between those four candidates based on those polls.   The process of the analysis drives the numbers to the only possible conclusion that it can arrive at.  

            In this example we are positioning just four candidates based on four simple polls.   In my daily Presidential polls I am positioning 30 candidates based on 42 polls, but the process is the same.   All of the polls (so far) count toward the standings.   After I have done 50 polls I will start removing the oldest polls out of the data, but so far they all count.

             

Yesterday’s Polls

 

            The candidates in yesterday’s poll were Cory Booker, Amy Klobuchar, Eric Swallwell and Steve Bullock.  The expected results of this poll, based on previous polls, were:

 

Booker

42

Klobuchar

38

Swallwell

11

Bullock

9

 

              This was the first time I had polled Governor Bullock, who just jumped into the pool, so actually his expectation was not established by previous polling; it was just a guess, although I happened to guess about right.   The actual results of the poll were:

Expected

Booker

42

Klobuchar

38

Swallwell

11

Bullock

9

Actual

Booker

43

Klobuchar

42

Swallwell

6

Bullock

9

 

            So Swallwell did not do swell in this poll, with 4% of his support fleeing to Klobuchar, and another 1% to Booker.   These, then, are the updated standings:

Rank

First

Last

Current

 

1

Elizabeth

Warren

1229

 

2

Pete

Buttigieg

1033

 

3

Joe

Biden

940

 

4

Kamala

Harris

890

 

5

John

Kasich

710

 

6

Stacey

Abrams

550

 

7

Beto

O'Rourke

536

 

8

Bernie

Sanders

494

 

9

Donald

Trump

467

 

10

Cory

Booker

459

UP 7

11

Amy

Klobuchar

419

UP 9

12

John

Hickenlooper

346

 

13

Bill

Weld

275

 

14

Kirsten

Gillibrand

248

 

15

Andrew

Yang

238

 

16

Julian

Castro

219

 

17

Howard

Schultz

195

 

18

Tulsi

Gabbard

194

 

19

Jeff

Flake

181

 

20

Jay

Inslee

176

 

21

Michael

Bennet

157

 

22

Tim

Ryan

139

 

23

Eric

Swallwell

110

DOWN 8

24

Mike

Gravel

79

 

25

Seth

Moulton

71

 

26

John

Delaney

67

 

27

Marrianne

Williamson

38

 

28

Wayne

Messam

32

 

 

 

What the Hell You Doin’ Here, James?

 

            A couple of weeks ago I had a outbreak of posts from annoyed readers who were either (a) wanting to know what I was doing with these polls, or (b) demanding that I stop doing it, depending on one’s viewpoint and on the specific annoyed reader.  You can better understand what I am doing by observing what I am doing than by having it explained to you, but I’ll try to help. 

            The pivot point in the history of American polling, as I would guess many of you know, is the Literary Digest Presidential poll of 1936.  The Literary Digest, a popular magazine with a large circulation, conducted straw polls to predict the winner of the Presidential contest in 1916, 1920, 1924, 1928 and 1932, and they were correct every time.  They were the Gold Standard of polls.   They predicted who would win state by state, and they got most of the states right, always.  

            In 1936, the Literary Digest not only predicted that Alf Landon  would beat Franklin Roosevelt; they predicted that he would beat Roosevelt by a wide margin.  Public faith in the Literary Digest poll was such that many, many people simply assumed that Landon was going to win. In fact, Landon lost 46 of the 48 states.   They had polled 2.4 million people, which is about 3,000 times as many as a modern pollster would reach—yet they got it completely, massively wrong.  A young pollster, George Gallup, polling a much smaller number of people, predicted correctly that Roosevelt would win a landslide victory.   This launched the career of George Gallup, literally killed the Literary Digest, and charted the course which polling has followed to the present day. 

            In retrospect, the Literary Digest made a long series of polling mistakes.  First, their outreach sample was biased toward those with money.   They polled their own users first; that was a relatively high-income sample.  They built on that by lists of phone users and registered automobile owners, but in 1936 not everybody had a telephone or an automobile.  Those who did tended to be more wealthy than those who did not, and they tended to vote more Republican.

            That, however, was not the big problem.  The bigger problems were (1) that they allowed self-selection in their voters, the return of mail-in cards, (2) that their poll was way too far out in front of the vote, and (3) that the widespread belief that Landon had the election in the bag caused Republicans to take their foot off the pedal.  When you allow self-selection of the voters—as my Twitter polls do, for example—you allow a campaign to take advantage of that by driving their supporters to participate.  Second, they were polling people in July about how they would vote in November.  And third, the Republicans’ belief that they were going to win the race was the exact same mistake the Hillary made in 2016, when she didn’t go to Wisconsin because she thought she didn’t have to.  It created an imbalance in voter motivation. 

            For the last 83 years, polling has been marching down a narrow path—a path defined and restricted by the mistakes that were made in 1936, and have been repeated countless times in different disguises since 1936.  One by one, mistake by mistake, we have learned things to do and not to do—make sure that you have the right mix of Democrats and Republicans, the right mix of men and women, the right proportion of minorities, the right mix of young and old and in-between.  Make sure you have got mostly registered voters, the right number of "likely" voters.  They call this "scientific" polling, although that is not a good name for it.  A set of guidelines is static.  Science is not static; it evolves rapidly.

            Scientific polling, frankly, is not all that damned good.  The reputable, upstanding, Skienterific pollsteristas are currently trying to tell us that 31% of Democrats or 38% of Democrats or some such number are currently supporting Joe Biden.   Frankly, I think they’re full of shit.   I think these are stupid estimates.  I don’t believe them; I don’t 50% believe them, I don’t 10% believe them.   They’re just wrong. 

            It’s not that these polls are not professional, or that the practices they believe in are not valid practices.  It is that they are using a simplistic, outdated polling model that blurs the truth more than it reveals it.  It has two huge problems.  One is that, that by their own rules, they are required to poll people who are not prepared to give them a thoughtful answer.   The other is, in essence, that they are using an out-of-focus camera.  I am trying to show them how they can focus their camera. 

            Look I understand the problems with Twitter polls.   In all likelihood, despite your warm and generous efforts to explain the problems to me, I understand these problems a great deal better than you do.   I do not actually believe that Elizabeth Warren would beat President Trump by a ratio of 1229 to 467 in a public vote, any more than you do.  

            But on the other hand, I don’t believe that Joe Biden has 35% support among Democrats, either.   These polls walk up to people and ask them who they are supporting for President.  People haven’t thought about it; the first name thsy many of the come up with is Joe Biden, so a good many of them say "Joe Biden."   They haven’t yet focused on the issue.   You’re not really getting "support"’ you’re getting name recognition, which doesn’t actually mean anything in a contest of this nature, because, by the time people vote, everybody will know both of the names.  

            Sometimes they don’t walk up to people; sometimes they call them on the phone, but that’s really the same thing; they are polling people who are not prepared to answer the question.   The other thing they do sometimes is, they show people a list of all the Democrats running for President, and they ask which one do you support?   People look over the list; don’t know that guy, don’t know that woman, is that a real name or a made-up one?   They don’t know.   You could add "Yogi Berra" to the list of people being polled and he’d get 8% because people would recognize the name and not be able to place it.   

            And, by their logic, you have to poll 800 people to have "statistically significant" results.   Which is PURE BS; granted, it is BS based on misunderstood statistical theory, but it’s BS.   Under those circumstances, you don’t get any type of legitimate discrimination between one down-list candidate and another.  

            When we poll only 4 people at a time, each of those people has a fair chance to stand out from the other three.   The "norm" isn’t four or five percent apiece; it is 25%.   You have an opportunity to support the candidate that you have seen something in that you like. 

            You can believe it or not, but the people who respond to my Twitter polls know the difference between Wayne Messam and Tulsi Gabbard, and we know for certain that they do.   We know that they do because every time that we poll Wayne Messam, he gets a position point between 22 and 45, and every time we poll Tulsi Gabbard, she gets a position point between 162 and 256.   This is true of every candidate in the group; they go up and they go down, but they’re not getting random outcomes.  They’re getting predictable outcomes.   That can only happen if people KNOW who they are voting for. 

            Because of this method that I have developed here, my readers actually KNOW how Julian Castro stands in comparison to Michael Bennett, how the support for Stacey Abrams compares to the support for John Hickenlooper, and whether John Kasich is getting traction in the Republican Party or whether he is not (he is.)  Granted, they only know these things within the boundaries of my twitter following, but that is something.   Outside that poll, who has ANY answer to these questions? 

 

            It has always been the central goal of my career to contribute to the discussion—not to contribute to the discussion for a moment, but to contribute to the discussion in such a way that the discussion is permanently different than it was before.  In my view, there is a spin-your-wheels discussion, in which the same cycles repeat themselves over and over, generation after generation, and there is a knowledge based discussion which builds understanding over time.  I have always tried to contribute to the knowledge-based discussion, while participating no more than necessary to the spin-your-wheels discussion. 

            Regardless of what you think that I am doing here, what I think that I am doing here is changing the discussion—not changing it today, very  much, not changing it tomorrow, but in the long term, profoundly changing it.   A lot of the attacks on my polling system, from my readers, have been based on the assumption that what I am trying to do is merely a flawed version of what pollsters have been doing for a hundred years—but I don’t think it is a weaker version of what others have done before; I think that it is new, original, and that it is superior to what other pollsteristas are doing.  I believe that, in time, what I am doing in these polls will be imitated by other pollsters until some unforeseen derivative of this polling method becomes the standard, normal way that political polls are done.   I believe that that will happen because I believe that when you build a better mousetrap, the world will beat a path to your website.  I’m not saying that I’m right; you can believe what you want to believe.   But I believe in what I am doing. 

 
 

COMMENTS (15 Comments, most recent shown first)

bjames
Fireball Wenz
A poll is different from a voting booth in that no one can see how you voted at the polls.


But I was asking about the TWITTER poll.
4:31 PM May 24th
 
Fireball Wenz
A poll is different from a voting booth in that no one can see how you voted at the polls. It's not rational, but even a stranger on the phone can inhibit respondents. And people who tend to support candidates who tap into their paranoia may not trust the privacy of online polling either. George Wallace always polled poorly vs. his results, because people were ashamed or were sensitive to the hostility of many to him. I suspect Trump is the same way. If you are conservative, Kasich is a far more"respectable" choice.
8:16 AM May 24th
 
bjames
Yogi was from St. Louis.
9:15 PM May 23rd
 
MidnighttheCat
1. very interesting, I will keep checking in to see how it is going.

2. I wish Yogi were still alive, now that you mention it, I would vote for him, even over Elizabeth Warren or Bernie Sanders. It is high time we had someone from New Jersey running the country.
5:03 PM May 23rd
 
bjames
In what sense is the poll less private than a voting booth?
3:24 PM May 23rd
 
MattGoodrich
Polls are also problematic when the results are embarrassing. Even many Trump supporters think he is a moron, so they often find it difficult to publicly admit they voted for him. Trump will always do better in a private voting booth than in a poll.
12:39 PM May 23rd
 
tangotiger
Bill:

The main reason for a "none of the above" is that Twitter won't allow the user to see the results of the poll unless they vote. So, that's the point that people are making (or should be making anyway).

Now, if you believe that someone will either:
(a) not vote, and ignore that poll if they don't have an opinion on any of the 4 candidates or
(b) will vote randomly, and so, the voting simply adds random noise
then this is a non-issue, and having the 4 options is fine. I agree with your reasoning here.

As an example of where I definitely needed an extra option when I conduct polls is when I had users tell me which baseball movie they preferred, in a head-to-head. I was especially interested in movies that are more niche, like Eight Men Out, compared to movies that were contemporary, like 42. So, I *really* needed people to only vote if they saw both movies. I didn't want people to pick 42, because they liked 42 and never saw Eight Men Out. And I didn't want non-baseball people to randomly pick a movie just so they can see the results. I don't get enough votes to have random noise get washed away.

So, getting back to say Stacey Abrams. She is formidable, and you can tell the first time she speaks. Now, if she wasn't on center stage, nobody outside of GA would know her, and she wouldn't pick up much votes. So, your method is good because it reflects the "landscape". Once Stacey Abrams makes her presence felt, your method will quickly pick her up, just like Mayor Pete is quickly picked up.

12:29 PM May 23rd
 
bjames


willibphx
Bill, to your comment about none of the above. As I try and understand your method how do you compare a candidate getting 25% of a poll which received 300 responses to a 25% response to a poll with 5,000 responses? Is this not relevant to the question of the level of support?


Well, (a) I don't have any polls getting 5,000 responses, (b) I don't have any polls getting only 300 responses, and (c) no, it absolutely is NOT relevant.
10:18 AM May 23rd
 
meandean
I think Bill's method is a brilliant way to distinguish between candidates with equal name recognition. I think this is a major breakthrough in that sense. (And will make this process extremely interesting once all the no-hopers drop out and the general public does develop opinions about each remaining candidate.)

I also think that, to the extent that more informed voters drive the discourse, this methodology can yield moderately predictive trend information. There's presumably a correlation (albeit a weak one) between how candidates are doing with the most politically active voters (who are most likely to actually vote), and how they're doing with the general public. If we see the hardcores abandoning Beto for Buttigieg (actually not a hypothetical), that probably does tell us something.

I don't agree with the implication that the method, in and of itself, solves the name recognition issue. Bill's voter base, as stated, is extremely well-informed. His voters have strong opinions about whether Tulsi Gabbard is a better candidate than Wayne Messam. I don't see how we get from there to saying therefore, the people who have only heard of Joe Biden also have strong opinions about Gabbard vs. Messam, it's just that no one is asking them to compare them.

I don't know what would happen if you took the people who'd be tempted to vote for Yogi Berra because that's a name they've heard of, and asked them to choose between Hickenlooper, Castro, Bennet, and Williamson. I'd be really interested to see it. But I do feel pretty confident that it wouldn't reveal preferences that are nearly as strong and predictable as what we're seeing in these polls.
9:05 AM May 23rd
 
willibphx
Bill, to your comment about none of the above. As I try and understand your method how do you compare a candidate getting 25% of a poll which received 300 responses to a 25% response to a poll with 5,000 responses? Is this not relevant to the question of the level of support?
6:41 AM May 23rd
 
StatsGuru
I do try to look at the websites of each of the candidates before I vote, so I'm actually getting to know their positions. That's something I would have never done this early in an election before.
5:24 AM May 23rd
 
3for3
This method looks much better for order than an ordinary poll, of course it would be far more expensive to implement...
12:31 AM May 23rd
 
shthar
I thought "none of the above" was not taking the poll at all.

At least that's what I've been doing.


9:45 PM May 22nd
 
Steven Goldleaf
I appreciate your trying to explain your polling system to your great unwashed ignorant readers, of whom I am one. I'm still trying to figure out how this works, but now I think I should be able to. Thanks. I've been enjoying taking your polls, though I think from what I've done so far, you could now train an intelligent ape to cast my future votes for me.
8:19 PM May 22nd
 
MarisFan61
"no, actually, I don’t need a "None of the Above" option; I don’t need that, at all. If you think that the poll needs that, that’s because you don’t understand what I am doing with these polls, and I would anticipate that you don’t. It’s new and different and it has never been done before..."

Nothing here to add to what I've said, except that Bill seems more and more to be confirming my guess on what's his purpose in doing this.

(You'll have to pardon me, I'm just celebrating.) :-)

8:03 PM May 22nd
 
 
©2019 Be Jolly, Inc. All Rights Reserved.|Web site design and development by Americaneagle.com|Terms & Conditions|Privacy Policy