[EM] (28): Steve's 28th dialogue with Richard
VoteFair
ElectionMethods at VoteFair.org
Sun Jun 12 18:43:23 PDT 2016
Steve, here is my answer to your latest question (see below) about
whether a candidate deserves to win in a situation where that candidate
is not a popular first choice. In a moment I'll provide an example that
involves U.S. Presidential candidates, but I'll start with an example
that involves anchovies, pineapple, and salmon.
Let's suppose that nine people are choosing one topping for a very large
pizza they will share. Let's suppose they have the following
first-choice preferences:
4 people want anchovies (and cheese)
3 people want pineapple (and cheese)
2 people want salmon (and cheese)
And let's suppose that nobody likes the other two choices.
Specifically, suppose the anchovy fans strongly dislike pineapple and
salmon, suppose the pineapple fans strongly dislike anchovies and
salmon, and suppose the salmon fans strongly dislike anchovies and
pineapple.
Now suppose that each person expresses their full order of preference on
a 1-2-3 ballot, and suppose the ballot includes italian sausage (and
cheese) as another possible choice. And suppose that all 9 people mark
italian sausage as their second choice.
Hopefully it's obvious that the best "winning" choice would be to order
an italian sausage (and cheese) pizza, even though that choice is not
anyone's first choice.
This means that when voters have unpopular first-choice preferences --
and there are three or more of those unpopular choices, and none of
those choices gets a majority of first-choice votes -- then those
unpopular choices typically do not deserve to win. Instead, the choice
that deserves to win is the choice that is clearly more popular than any
of the unpopular choices.
Now let's shift to U.S. Presidential candidates. Suppose that, somehow,
the same nine voters were being asked to elect a U.S. President. And
suppose the four people who want anchovies prefer Donald Trump, the
three people who want pineapple prefer Bernie Sanders, and the two
people who want salmon prefer Ted Cruz. And in this hypothetical
election let's suppose that everyone's second choice is Hillary Clinton.
Using instant-runoff voting Clinton would be eliminated first, then Cruz
would be eliminated. At that point let's assume that the two Cruz fans
are split such that one of them marked Trump as their third choice and
the other one marked Sanders as their third choice. This gives Trump 5
votes and gives Sanders 3 votes, so Trump would win.
But notice that a majority of voters -- five (out of nine) -- prefer
that Trump not be elected. And notice that only one of those five
voters ranks Trump as high as third place, while the other four voters
rank him in fourth or fifth place.
As stated above, when voters have unpopular first-choice preferences, it
is not reasonable to allow any of the unpopular choices to win.
Now let's consider pairwise counting methods -- in which each pair of
candidates is considered one pair at a time.
Obviously in this hypothetical election the pairwise comparison between
Clinton and Trump has 5 voters giving support for Clinton and 4 voters
giving support for Trump, which means Clinton wins this pairwise
contest. Clinton also would win pairwise contests against Sanders (6 to
3) and against Cruz (7 to 2). So with this hypothetical voting pattern
the Condorcet winner would be Clinton, and she is preferred by a
majority against her closest competitor.
Again notice this pattern that a choice can be the most popular even if
none of the voters prefer that choice as their first choice.
Some people who support instant-runoff voting would argue that a
Condorcet winner (someone who wins all their pairwise contests) is not a
good compromise, but that claim is not backed up by mathematics, as
demonstrated above.
You refer to measurements of "preference intensity" but such
measurements are only possible when voters are completely honest about
how strongly the voter likes or dislikes each candidate.
None of the voting methods currently available gives a positive
incentive to be "honest." Instead all current voting methods are
vulnerable to strategies that give the voter less influence (over the
result) if the voter is completely honest in expressing how strongly the
voter likes or dislikes each candidate.
I see that you recently learned about a voting method called "majority
judgment" that some of its advocates claim is supportive of voters being
honest about how they mark their ballot. Part of the reason that some
voters would honestly reveal the strength of each like or dislike
ranking is that the majority-judgment vote-counting method is difficult
to fully comprehend, so it's difficult to identify useful strategies.
However, if majority judgment were used in a high-stakes election such
as for U.S. President then it becomes worthwhile for experts with
different political positions to use computer simulations to figure out
how "their" voters should strategically mark ballots to increase the
influence of those voter groups. As a result, honest preference
information is lost, and it becomes difficult to meaningfully measure
"preference intensity" values.
At a future time there might be a vote-counting method that makes it
possible to collect honest preferences that include information about
how much of a gap exists between each preference level (which is another
way to say "how strongly a voter likes or dislikes each candidate")
without any strategic advantage for "dishonesty," but such a method does
not yet exist. (When it does, it is likely to involve voting influence
over multiple races/contests or multiple elections, rather than the
current situation of each race/contest being counted independently.)
In the meantime the best we can do is to ask a voter to rank candidates
from most favorable to least favorable, without any information about
the size of the gap between each ranking level. That is what 1-2-3
ballots do.
Although 1-2-3 ballots can be counted using instant-runoff voting, the
above examples reveal why that approach does not deserve to be adopted
in situations where the outcome is very important. (It would be fine
for use by a small group of people gathered in the same room and voting
on a new logo, but that's not a high-stakes election.)
The advocates of approval voting would claim that approval ballots do
get marked honestly. However, notice that an approval ballot -- where
the only two choices are "approve" and "disapprove" -- does not allow
the voter to indicate any preference difference among the multiple
candidates who are marked as "approved" and the voter cannot indicate
any preference difference among the multiple candidates who are marked
as "not approved." So approval ballots do not provide enough
information to measure what you call "preference intensity."
Hopefully this answers your latest questions.
Thank you for recognizing that my book "Ending The Hidden Unfairness In
U.S. Elections" contains valuable insights about Presidential elections.
I wish more people would learn these insights.
The insight that will become relevant soon is the fact that pairwise
counting does work with the electoral votes that are an important part
of U.S. Presidential elections. To briefly clarify, we don't need the
"electoral college" part where people are given the power to cast votes
if a majority outcome cannot be achieved on the first round of voting,
but we do need to use electoral votes because they protect against: the
unfairness of weather differences on election day, biases in which
districts have intentionally long waiting lines, and Oregon having the
turnout advantage that everyone here in Oregon votes by mail.
I hope this answer helps you on your journey to better understand voting
methods.
Richard Fobes
On 5/30/2016 4:32 PM, steve bosworth wrote:
> 1. Re: (28): Presidential Election: Steve's 28th dialogue withRichard
>
> Fobes (VoteFair)
>
> Richar wrote:
>
> [….]
>
> If I should find time to do any writing about election-method reform,
>
> I'll use that time to write an article about what's going on in the U.S.
>
> Presidential elections ….
>
> This would be a great time to write about the link between single-mark
>
> ballots and the crazy Presidential primary results, and the need for
>
> better ballots and better vote-counting methods….
>
> Richard Fobes
>
> +++++++++++++++++++++++++
>
> Hi Richard,
>
> In your last reply, you again express your interest in writing an
> article on ‘the need for better ballots and better vote-counting
> methods’ for electing the President.Perhaps your article would start
> with the suggestions you offer in Chapter 10 (‘Presidential
> Smokescreen…’) and in Chapter 11 (‘Get Real ….’) of your book (Ending
> the Hidden Unfairness …).However, before discussing those more complex
> suggestions, I need some clarification of your related but simpler
> suggestions in Chapter 17 (‘What’s Up Gov…’)… for electing mayors and
> governors.
>
> On page 5 (of Chapter 17), you say that ‘VoteFair popularity ranking
> identifies the most popular candidate’.Please correct me if I am
> mistaken in saying the following:By using VoteFair (i.e. Kenemy), this
> ‘most popular’ winner might not have been expressly preferred even by a
> plurality of all the voters.This conclusion results from the example
> below in which 7 candidates are running for governor.
>
> As a result of using IRV in this extreme example, candidate F is elected
> with a 61% majority and with an average intensity of preference of 9.62
> out of 10.In contrast, by counting the same 100 ballots using MAM (I
> assume the result would be the same using VoteFair), candidate E is
> elected with the explicit support of only 39% and with an average
> intensity of preference of 9.28.
>
> If this is correct, your ‘most popular’ winner would have been
> explicitly support by only 39%.Therefore, would it not be better to use
> a form of IRV?Am I mistaken?
>
> I hope you can find the time to respond.
>
> Best regards,
>
> Steve
>
> Example:
>
> 100 citizens vote to elect one winner.They rank the 7 candidates as follows:
>
> *IRV COUNT*
>
> 100 CITIZENS RANK CANDIDATES EFGKMNP AS FOLLOWS:
>
> 38
>
>
>
> 23
>
>
>
> 25
>
>
>
> 9
>
>
>
> 5
>
> F
>
>
>
> G
>
>
>
> E
>
>
>
> M
>
>
>
> K
>
>
>
> F
>
>
>
>
>
> N
>
>
>
> P
>
>
>
>
>
>
>
> E
>
>
>
> E
>
> *COUNT USING IRV*
>
> FIRST
>
>
>
> 38
>
>
>
> 23
>
>
>
> 25
>
>
>
> 9
>
>
>
> 0
>
>
>
> 0
>
>
>
> 5
>
>
>
> F
>
>
>
> G
>
>
>
> E
>
>
>
> M
>
>
>
> N
>
>
>
> P
>
>
>
> K
>
> SECOND
>
>
>
> 38
>
>
>
> 23
>
>
>
> 25
>
>
>
> 9
>
>
>
> 0
>
>
>
> 5
>
>
>
> ELIMINATE
>
> THIRD
>
>
>
> 38
>
>
>
> 23
>
>
>
> 30
>
>
>
> 9
>
>
>
> 0
>
>
>
> ELIMINATE
>
>
>
> FOURTH
>
>
>
> 38
>
>
>
> 23
>
>
>
> 30
>
>
>
> ELIMINATE
>
>
>
> 9
>
>
>
>
>
> FIFTH
>
>
>
> 38
>
>
>
> 23
>
>
>
> 39
>
>
>
>
>
> ELIMINATE
>
>
>
>
>
> SIXTH
>
>
>
> 61
>
>
>
> ELIMINATE
>
>
>
> 39
>
>
>
>
>
>
>
>
>
>
>
> WINNER
>
>
>
>
>
>
>
>
>
>
>
>
>
> Using IRV, F wins with a 61% majority, and with a preference intensity
> of 9.62 out of 10.
>
> Note:Separately, I have already emailed to Richard a copy of the exact
> calculations which discovered E as the MAM winner having only 39% of the
> expressed preferences and having an average intensity of support of 9.28
> out of 10.These and other relevant passage are print in green within
> Appendix 4 (‘Comparing Rival System’) to my article: ‘Super Equality for
> Each Citizen’s Vote in the Legislature’.I would be happy to send this
> appendix (and/or article) to anyone who requests this
> (stevebosworth at hotmail.com).
>
> .+++++++++++++++++++++++
>
> On 2/19/2016 11:47 AM, steve bosworth wrote:
>
> > [EM] (27) APR: Steve's 27th dialogue on NUTS with Richard Fobes
>
> >>Date: Wed, 17 Feb 2016 23:48:29 -0800
>
> >>From: ElectionMethods at VoteFair.org
>
> >>To: election-methods at lists.electorama.com
>
> >>CC: stevebosworth at hotmail.com
>
> >>Subject: Re: [EM] (26) APR: Steve's 26th dialogue with Richard Fobes
>
>
>
>
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info
More information about the Election-Methods
mailing list