Candidate Proxy

[Forest Simmons]

Dear Steven,

thanks for your insightful and thoughtful response.

An explanation of my apparent dismissal of Approval in the context of
that message is given below, along with some other related thoughts.

Best Wishes,


Forest


On Sun, 8 Dec 2002, Steven J. Brams wrote:

> Dear Forest,
>
>       It's an interesting proposal, but I have a problem with it in
> principle, especially in light of a paper I'm writing with Remzi
> Sanver entitled "Voter Sovereignty and Election Outcomes."
>
>       I believe voters should be sovereign--their preferences should
> be determinative, whatever the method of vote aggregation.  Your
> proposal shifts the choice to some candidate, which compromises the
> voter's sovereignty even with a grading system.
>
>       The simplest example I can think of to illustrate the problem is
> two centrist voters whose second choices diverge (one prefers the
> left candidate, the other the right, if the centrist turns out to be
> out of the running).  Isn't it simpler that the voter be able to
> express his or her preference rather than hope there is a proxy that
> expresses this preference as well?
>
>       Best wishes,
>       Steve

I agree that Approval is simpler, especially for you and me, as well as
for others who are confident in their assessment of the strategic value of
approving their preferred of the front runners.

But the primary audience of my message was a group of IRV supporters many
of whom had already rejected my various sales pitches for Approval, so I
put that phrase ("simpler than Approval") in to keep them from saying,
"Here goes the Approval broken record again," followed by deleting the
message without reading further.

Even so, I vacillated back and forth before finally putting it in.

I guess what I really meant (and I believe this) is that Candidate Proxy
would seem simpler to THEM than either Approval or IRV.

These IRV supporters persist with amazing stubbornness in the following
two superstitious objections to Approval:

(1) They think that it violates "one person, one vote."

(2) They think that it violates "majority rule."

Despite many clear, eloquent, and ingenious explanations from many points
of view from members of the Election Methods lists trying to draw these
folks out from their unreasonable loyalty to their narrow interpretations
of these two concepts, none, to my knowledge, has ever admitted being
persuaded.

They also have two other objections to Approval that are insuperable for
them:

(3) They don't feel qualified to make the strategic judgment about when
they should approve the "lesser evil" candidate.

(4) Approval doesn't allow them to give any support to a compromise
candidate without giving him full support.

They believe that IRV is superior on all four of these crucial issues, and
that no other issue (e.g. criterion, such as Monotonicity, Participation,
or Consistency) is nearly as important as any of these.

To them the complexity of the IRV ballots and election process is a small
price to pay for its (perceived) superiority over Approval in these four
areas.

On issues (3) and (4) they deny that there is any need for strategy in
IRV, in spite of the many examples of plausible elections that would
require large groups of voters to rank compromise ahead of favorite in
order to avoid being stuck with their last choice as winner of the
election.

How does Candidate Proxy stack up on these four all important issues?

On (1) and (2) it passes with flying colors:  (1) Each voter makes
one mark on a good old fashion American ballot. (2) If one candidate gets
more than fifty percent of the marks, that candidate wins.

For (3) consider the following:

As my reference to Australia's experience with the Hare/STV system bears
out, most voters in public elections trust their favorite candidate's (or
party's) judgment regarding candidate rankings more than they trust their
own judgment (or else they don't think the research effort is worth it).

Celebrities, including professional politicians, are considered to be
experts with inside information and special insights unavailable to the
rest of us common folk.

So most voters consider it simpler to let their favorite candidate worry
about what to do with their vote in case there is no candidate with more
than fifty percent of the vote.

As to item (4), Candidate Proxy allows them to give support to a
compromise candidate vicariously, and actually extends that support only
if it turns out that it is appropriate AFTER Favorite has already received
full support.  Favorite takes clear precedence over compromise.

In the eyes of IRV supporters, Candidate Proxy rivals IRV, and definitely
beats Approval, on all four of these points.

Since the message that you responded to was my first overture to IRV
supporters on Candidate Proxy, it is too early to tell if the simplicity
of Candidate Proxy compared to IRV will win them over. The first response
was from the FairVoteOR list, and it was positive.

By the way, I believe that Candidate Proxy (with some version of Cardinal
Ratings as the Election Completion Procedure) is much better over all than
IRV, not just on simplicity.

Candidate Proxy is not as good as Approval over all, but for the above
reasons might be accepted more readily by the public.

The centrist problem that you bring up is real, but is mitigated by
several considerations, including the inordinate trust of the average
voter in the judgment of his favorite candidate; some centrist supporters
might have a strong divergence of opinion on compromise candidates, but
(typically) most of them will trust their favorite's judgment.

A compromise that I like is this:  use Approval ballots, but those ballots
that approve only one candidate are interpreted as delegating proxy power
to that candidate.

Of course, in the public arena this compromise would bring us up against
superstitions (1) and (2) again.

Another mitigating factor is this.  Generally a candidate's strong
preferences among the front runners will correlate strongly with his
supporters' preferences.

Let's consider the example that you suggested in more detail.

If the centrist's supporters are about equally divided between left of
center and right of center, then this centrist candidate typically doesn't
have a strong preference between either of the two front runners, and
should rate them equally in the pre-election "reality check" rating.

Then during the Election Completion Convention this centrist has lots of
leverage to apply to both of the potential compromises.

[It doesn't violate the consistency requirement to pass one and fail the
other. I worded that requirement carefully to allow this possibility.
The requirement is violated only if a candidate is approved without those
rated/ranked higher also being approved.]

Furthermore, if the issue space is approximately one dimensional, as your
left/center/right language would suggest, then if neither of the extremes
(in a three candidate race) has a majority, then the best strategy for the
centrist is to hold tight and approve only self, and the best strategy for
the weaker of the other two is to approve center as well as self.

In other words, for this scenario the only Nash equilibria are those for
which Center is the winner of the election.

Consider the following representative left/middle/right scenario in which
middle has weak first place support:

42% left>middle>right
11% middle>left>right
10% middle>right>left
37% right>middle>left


If IRV were used to decide the issue, right voters would have to vote
middle over favorite to keep left from winning.

Under Approval, with accurate polls, right would support middle without
betraying favorite, and middle would know to hold out for favorite only.

With inaccurate polls, some middle supporters might be persuaded to
approve left, and some right, giving the election away in either
direction, depending on the bias of the polls, and relative strengths of
preference.

Candidate Proxy has a perfect built in poll; each candidate knows how many
votes they have to play with.  Middle could not be fooled by inaccurate
polls.

In other words, in this situation Candidate Proxy is less manipulable than
standard Approval ( not to mention IRV, Borda, Plurality, etc. ) .

Finally, I agree that Approval is better with respect to Voter Sovereignty
than Candidate Proxy (especially when pre-election polls are accurate),
but I believe that voters have at least as much effective control (when
not more) of election outcomes under (the best versions of) Candidate
Proxy as they do under IRV.

One extreme of Voter Sovereignty would be the "Direct Democracy"
proposal that we hear about from time to time, which would entail constant
voter referenda via the internet in place of decisions by traditional
representative bodies.

At the opposite end of the voter sovereignty spectrum we have the
Electoral College, the Speaker of the House, and the Supreme Court
variously choosing the President in various situations.

Perhaps Candidate Proxy falls within the reasonable range of compromise on
this issue; after all, the Election Completion Convention would have one
of the purer forms of Proportional Representation, much more so than our
current house of representatives, and much more so than any parliament
elected by the rules of any PR voting method in current use.

The pre-election "reality check" consistency requirement is specifically
intended as a form of voter sovereignty insurance; it limits the
candidates from making capricious or whimsical use of the votes.

This consistency requirement could be strengthened: the candidate includes
a tentative approval cutoff rank or rate on the pre-election rankings or
ratings. During the Election Convention proceedings the candidate must
pass all candidates strictly above that rank/rate and fail all candidates
below that rank/rate. The candidates that share that rank/rate can go
either way in the final grading.

If (in the pre-election reality check posting) a candidate ranks/rates
very many candidates at the cutoff level, the voters can take it as a sign
that this candidate is up to something fishy.

I confess that as near as I know, nobody has made a serious study of
Candidate Proxy methods.  The idea was inspired by a posting of an EM list
member whose email handle is "Demorep,"  who mentioned that in some
Proportional Representation schemes the elected representatives have
(permanent) residual weights attached to their votes to fine tune the
proportionality.

He called it Proxy PR. The Election Completion Convention idea (for both
single winner and multiwinner elections) first occurred to me while
reading that posting.

The evolution of the idea is documented by the EM list archives.

Do you think I should apply for an intellectual property rights patent?

Forest

----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em