[EM] Open Message to Foley, Maskin, and Other Tweakers

[Forest Simmons]

How can this one simple tweak improve on any method currently in use?

Because currently no Landau efficient method is in use, and each step of
this method takes you closer to the Landau plateau in the direction of
steepest ascent, so to speak.

It works like gradient ascent. The hyperplane that bisects the line segment
between W and C is roughly tangent to the level surface of the smoothed
local ballot mass density.

The number of ballots/voters on the W side of the hyper plane is smallest
(so pairwise opposition to C is minimal) when C is in the direction of
steepest  ballot/voter mass increase ... i.e. the vector from W to C is
roughly parallel to the mass density gradient.

The beauty of gradient ascent is that no matter where you start, each step
gets you closer to the Landau plateau supporting the single peak, if there
is one i.e. the region of local maximum average ballot density.

That's why this method is so robust!

Shouldn't there be a law that any method based on RCV ballots be perforce
retrofitted with this tweak, unless it has a built in  seamless way of
provably guaranteeing Landau efficiency? Who could rationally object to
this zero-cost Landau insurance "coverage"?

Any questions? Comments? Gripes?

-Forest

On Thu, Nov 3, 2022, 8:40 PM Forest Simmons <forest.simmons21 at gmail.com>
wrote:

> Here is a simple tweak that will vastly improve any election method that
> is not already Landau efficient:
>
> Initialize the variable W with the name of the winner of the imperfect
> method.
>
> Elect W if no candidate covers W,
>
> Otherwise ...
>
> While W is covered let C be the candidate among those covering W with the
> least pairwise opposition from W, i.e. the one minimizing the number of
> ballots on which it is outranked by W.
> Then update W with the name of candidate C. (End While)
>
> Elect the candidate named by the final value of W.
>
> Rationale: If C covers W, then geometrically C is between W and the
> (generalized) median voter. The more closely C covers W from the voters,
> the fewer voters will prefer W over C.
>
> It doesn't matter how horrible your imperfect method is ... IRV, Borda,
> Nanson, Coombs,, Copeland, Black, Random Favorite Candidate, Random
> Truncated Candidate, etc or even non-horrible like MJ, MinMax, Ranked
> Pairs, Schulze CSSD, River, etc.....
>
> If it can elect a covered candidate, then this tweak is needed!
>
> -Forest
>
> On Wed, Nov 2, 2022, 10:30 PM Forest Simmons <forest.simmons21 at gmail.com>
> wrote:
>
>>
>>
>> On Wed, Nov 2, 2022, 9:04 PM Ralph Suter <RLSuter at aol.com> wrote:
>>
>>> Thanks very much for the detailed comparison and clarification. I just
>>> haven't had time to follow the election-methods list well enough to nearly
>>> as knowledgeable about alternative methods as I would like to be.
>>>
>>> Given the shortcomings of the article, can the authors at least be
>>> credited for pointing out that IRV is seriously flawed in that it can
>>> result in a winner that is outranked by another candidate? (I prefer
>>> calling it IRV, because "ranked choice voting" misleadingly implies that
>>> there is only one form of ranked voting.).
>>>
>>
>> Yes, I'm glad they made that important point.
>>
>>> IRV also has two other major flaws: (1) It can take a long time before
>>> it's possible even to make an educated guess of the winner. In the Alaska
>>> special election, it took more than two weeks.
>>>
>> It's because every round requires a complete pass through the ballots to
>> execute the vote transfers. Their suggestion, Baldwin, has the exact same
>> problem. DMC does not because bothered the pairwise matrix and the "total
>> votes" implicit approval are precinct summable, with only one pass through
>> the ballots.
>>
>> (2) It makes useful polling, either pre-election or exit polling, all but
>>> impossible.
>>>
>>> Would DMC avoid both flaws? Also, do you think it's easily enough
>>> explainable for it to gain the support needed to replace IRV?
>>>
>> IRV is highly entrenched, but STAR and Approval have made some progress,
>> and they seem to be more open minded.
>>
>> STAR is instant top two score runoff. Once you have the candidate scores
>> along with pairwise defeat scores ready for the instant runoff, why not pit
>> the top score candidate against the champion of the 2nd and 3rd score
>> candidates.  Better yet take this to its logical conclusion ... do
>> SequentialvPairwise Elimination starting at the bottom of the score list
>> and working up, so that the final matchup is between the top score
>> candidate and the champion from below.
>>
>> I'm good at explaining methods, but I am not an organizer or influencer.
>> I'm in my mid seventies so my energy, sight, and other faculties are not
>> what they once were.
>>
>>> Finally, do you have any thoughts about the best way to conduct open
>>> primaries? Alaska's top four primary followed by a ranked choice general
>>> election seems clearly superior to California's top two.primary followed by
>>> a majority winner general election.
>>>
>>> Would either or both top two or top four be improved by using approval
>>> voting or some other method to choose the final two or four? Also, would
>>> top five be even better than top four, or would top three be nearly as good
>>> as top for?
>>>
>> Someone with more open primary experience could advise you better on
>> this. However...
>>
>> If possible I would say "Top k approval" with k between one and four,
>> determined by how close the other approval scores are to the leader's.
>>
>> If no challenger gets 70 percent of the leading approval, then it seems
>> that a runoff would be a waste of time and money.
>>
>> On the other hand,  if four candidates were within five approval points
>> of the leader, all four should be allowed in the runoff (it seems to me).
>>
>> In any case, the standard for runoff participation should be delineated
>> clearly well in advance.
>>
>> IMHO
>>
>>> Thanks again for your really helpful feedback.
>>>
>> Thanks for caring about democracy enough to inform yourself and others!
>>
>>> -Ralph
>>> On 11/2/2022 3:28 PM, Forest Simmons wrote:
>>>
>>> "Single most important"?
>>>
>>> Not anywhere close!
>>>
>>> The "tweak" proposed by Maskin and Foley is either Baldwin (if the total
>>> votes are recalibrated between rounds as in IRV) or Benham applied to the
>>> Borda order the same way DMC is Benham applied to the implicit approval
>>> order ... without recalibration between rounds.
>>>
>>> Neither one is appreciably better than IRV. Baldwin retains IRV's
>>> non-monotonicity and its procedural mess of vote transfers, while the
>>> monotonic version more fully retains Borda's clone dependence and burial
>>> incentive.
>>>
>>> A much better "tweak" is to interpret "total votes" as implicit
>>> approval, the number of ballots on which a candidate is voted ahead of one
>>> or more other candidates.
>>>
>>> This definition of total votes is especially apt when the ballots only
>>> allow ranking of a few candidates, as is the case of all RCV ballots in the
>>> USA wherever RCV has been publically adopted so far. Voters have to
>>> consider carefully which candidates are worth ranking. That gives extra
>>> value to implicit approval.
>>>
>>> After tallying the total votes (i.e. the ballot totals of votes of any
>>> rank) of the candidates, in each round eliminate the candidate with the
>>> fewest total votes, until among the remaining candidates there is an
>>> undefeated candidate ... (total votes not recalibrated between rounds).
>>>
>>> This method is monotonic, clone free, and Condorcet efficient.
>>>
>>> It has a name: Implicit Approval DMC, Definite Majority Choice, the very
>>> method proposed by a consensus of the EM list to Toby Nixon twelve years
>>> ago after careful consideration of the merits of all of the competing
>>> proposals (including BTR-IRV, another tweak better than Baldwin) ... not
>>> just an off the cuff tweak.
>>>
>>> DMC is a streamlined procedure for determining the common winner of
>>> Ranked Pairs, Schulze CSSD (Clone free Schwartz Sequential Dropping), and
>>> Jobst's River, the three great methods that reduce to one common method in
>>> the case where the strengths of the pairwise defeats are measured by the
>>> total votes (implicit approvals) of the respective winners of those defeats.
>>>
>>> The Maskin-Foley proposal is anything but the proposal to end all
>>> proposals.
>>>
>>> I'm afraid Maskin hasn't learned anything in the two decades since he
>>> was busy publically proposing ("The Fairest Vote of All") clone dependent
>>> Borda as a Condorcet completion method for clone dependent Copeland.
>>>
>>> Has he been too busy resting on his Nobel laurels to learn anything from
>>> the well informed feedback engendered by his Scientific American article?
>>>
>>> Some of these academics from the big name universities assume that they
>>> can come up with a method superior to any proposal from the rabble by
>>> merely shooting from the hip in an op ed.
>>>
>>> -Forest
>>>
>>> On Tue, Nov 1, 2022, 7:53 PM Ralph Suter <RLSuter at aol.com> wrote:
>>>
>>>> Rob,
>>>>
>>>> For some reasons, my last several messages, including the following
>>>> sent today, haven't been posted.
>>>>
>>>> The article mentioned below may be the single most important and
>>>> potentially ifluential one regarding election methods that has been
>>>> published for many years. It may finally get people to realize that instant
>>>> runoff voting is a seriously flawed ranked choice method and should be
>>>> replaced by a form of Condorcet voting.
>>>>
>>>> Will you please post it for me or copy and send it to the list?
>>>>
>>>> Thanks,
>>>> Ralph Suter
>>>>
>>>> -------- Forwarded Message --------
>>>> Subject: "Total Vote Runoff" proposed as better way to determine
>>>> ranked-choice winners
>>>> Date: Tue, 1 Nov 2022 13:13:21 -0500
>>>> From: Ralph Suter <RLSuter at aol.com> <RLSuter at aol.com>
>>>> To: election-methods-request at lists.electorama.com
>>>>
>>>>
>>>> In a Washington Post opinion article published today (11/2/2022),
>>>> election law scholar Edward Foley and economist (and Nobel laureate) Erik
>>>> Maskin propose a "tweak" to correct what they describe a flaw in how ranked
>>>> choice winners are currently determined. They call the resulting election
>>>> method a "total Vote Runoff".
>>>>
>>>> Would anyone like to comment? It appears they are essentially proposing
>>>> replacing instant run-off voting with Condorcet voting.
>>>>
>>>> -Ralph Suter
>>>>
>>>> ---------------------------------
>>>> Alaska’s ranked-choice voting is flawed. But there’s an easy fix.
>>>> By Edward B. Foley
>>>> <https://www.washingtonpost.com/people/edward-b-foley/> and Eric S.
>>>> Maskin
>>>> November 1, 2022 at 7:00 a.m. EDT
>>>>
>>>> https://www.washingtonpost.com/opinions/2022/11/01/alaska-final-four-primary-begich-palin-peltola/
>>>>
>>>> Excerpt:
>>>>
>>>> Alaska’s special election in August for the House of Representatives
>>>> was heralded as a triumph for ranked-choice voting, because MAGA favorite
>>>> Sarah Palin, a personification of polarization, could not attract enough
>>>> second-choice votes from moderate Republican Nick Begich’s supporters to
>>>> win.
>>>>
>>>> That’s true. But the way Alaska uses ranked-choice voting also caused
>>>> the defeat of Begich, whom most Alaska voters preferred to Democrat Mary
>>>> Peltola, the candidate who ended up winning.
>>>>
>>>> This anomalous outcome, contrary to the principle that the majority’s
>>>> preference should prevail, would be easily remedied by one small change.
>>>>
>>>> The key to ranked-choice voting is that a voter lists the candidates in
>>>> order of preference, starting with their favorite, rather than naming just
>>>> that favorite. The problem in Alaska — and other ranked-choice systems now
>>>> in use, from Maine to San Francisco — is the rule for eliminating
>>>> candidates when no one gets a majority of first-place votes. By tweaking
>>>> this rule, Alaska’s system would become more palatable to Republicans and
>>>> Democrats alike, and more likely to be adopted across the country.
>>>>
>>>> Begich was eliminated because he had the fewest first-place votes. That
>>>> seems logical at first glance. But the flaw in this outcome — and why
>>>> Republicans have reason to be resentful — is that a majority of voters
>>>> would have favored Begich had the race come down to a head-to-head matchup
>>>> against either Peltola (52 percent to 48 percent) or Palin (61 percent to
>>>> 39 percent). He lost only because it was a three-way race.
>>>>
>>>> Here’s how to fix the flaw. If Alaska eliminated the candidate with the
>>>> fewest *total* votes, rather than the fewest *first-place* votes, the
>>>> ranked-choice system would be sure to elect a candidate such as Begich who
>>>> defeats all rivals in one-on-one matchups.
>>>>
>>>> Call it a “Total Vote Runoff.” A candidate’s total votes in such a
>>>> system would be determined by the number of other candidates he or she is
>>>> ranked above. For example, when a candidate is ranked first on a ballot in
>>>> an election involving three candidates, then this first-choice candidate is
>>>> ranked above two other candidates and gets two votes from this ballot.
>>>>
>>>> When that same candidate is ranked second on another ballot, the
>>>> candidate is favored over only one other candidate and would receive only
>>>> one vote from that ballot.
>>>>
>>>> A candidate ranked last on a ballot, or not ranked at all, is not
>>>> favored over anyone and gets no votes from that ballot.
>>>>
>>>> Calculating the number of votes that a candidate gets on each ballot —
>>>> two, one or zero — and adding up the candidate’s votes from all the ballots
>>>> yields the candidate’s total votes.
>>>>
>>>> Using this method, we can identify the number of ballots on which each
>>>> of Alaska’s three candidates was ranked first or second and then calculate
>>>> each candidate’s total votes (there were only three candidates in the House
>>>> special election):
>>>>
>>>> Alaska House results using total vote runoff
>>>>
>>>> *First-place votes get counted twice because voters put their first
>>>> choice ahead of two other candidates.*
>>>> Column 1: Candidate
>>>> Column 2: first-place votes
>>>> Column 3: first-place votes, counted again
>>>> Column 4: second-place votes
>>>> Column 5: Overall Total
>>>>
>>>> Begich 53,810 53,810 81,253 188,873
>>>> Peltola 75,799 75,799 19,024 170,622
>>>> Palin 58,973 58,973 31,611 149,557
>>>> Source: Alaska official results, Alaska cast vote records, MIT
>>>> Election Data and Science Lab, Election Law at Ohio State, author
>>>> calculations
>>>> <https://www.elections.alaska.gov/results/22SSPG/RcvDetailedReport.pdf>
>>>>
>>>> Palin had the fewest total votes, so she would have been the first
>>>> candidate eliminated in a “Total Vote Runoff” tweak to RCV.
>>>>
>>>> With Palin eliminated, the race would have been between Begich and
>>>> Peltola. Because a majority preferred Begich to Peltola, he would have been
>>>> elected. Total Vote Runoff captures the will of the majority more
>>>> accurately than Alaska’s current elimination system does.
>>>>
>>>> Republicans should like Total Vote Runoff because its procedure would
>>>> help ameliorate the “candidate quality
>>>> <https://www.nbcnews.com/politics/2022-election/mcconnell-says-republicans-may-not-win-senate-control-citing-candidate-rcna43777>”
>>>> problem that plagues their party, as Senate Minority Leader Mitch McConnell
>>>> (R-Ky.) lamented. A candidate popular only with the party’s base would be
>>>> eliminated early in a Total Vote Runoff, leaving a more broadly popular
>>>> Republican to compete against a Democrat.
>>>>
>>>> Democrats, too, should welcome Total Runoff Voting to protect against
>>>> losses caused by excessively progressive candidates who are unacceptable to
>>>> a large portion of independent voters. Alaska-style ranked-choice voting
>>>> might keep in contention a left-wing candidate whose first-place votes
>>>> reflect enthusiastic but limited support, but Total Runoff Voting would
>>>> promote Democratic candidates whose wide appeal makes them more competitive
>>>> overall.
>>>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20221103/a0150b4d/attachment-0001.htm>