More info on "Just-in-Time Withdrawal" (JITW)

[Donald E Davison]

Steve Eppley,

     I am reading your paper on the JITW single winner method. I need more
time to study it all, but off hand I will say that your JITW seems to be a
bit better than Run-Off without Elimination.

     Could JITW be used in single seat districts???

     If so, what would be its value in these type of elections???

Donald Davison

  ------------ Forwarded Letter --------------
From: "Steve Eppley" <SEppley at alumni.caltech.edu>
To: ntk at netcom.com, DONALD at MICH.COM
Date: Mon, 15 Feb 1999 14:11:15 -0800
Subject: More info on "Just-in-Time Withdrawal" (JITW)
Priority: normal

Mike Ossipoff has mentioned JITW during this recent conversation.
His description of it isn't quite the same as mine.  Also I don't
think he attempted to rebut the argument made against JITW.

Here's some more info about JITW, which can be used in both single-
winner and PR elections.

JITW stands for "just-in-time withdrawal."  It is a feature which can
be added to any preference order election method.

   Just-in-Time Withdrawal
   -----------------------
   1.  After the votes are cast but before the final results are
       tallied, the voters' orders of preference are published
       (on the internet).  (To preserve the secret ballot, no
       info which could identify a voter is published, of course.)
   2.  Before the final results are tallied, candidates are given a
       short period of time, perhaps a week, to decide whether to
       voluntarily withdraw from contention.  (In the case of ballot
       propositions, the official sponsor of a proposition would have
       the authority to withdraw it.)
   3.  When tallying the results, the ballots are treated as though
       the voters had left withdrawn candidates unranked.

I prefer my description more than Mike's for a couple of reasons:
1) Mine makes it clear that the preference orders will be published.
2) Mike's makes it sound as though candidates can demand extra
tallying labor (a series of recounts).

I will mention a few JITW methods, beginning with the simplest.

The simplest "good" single-winner preference order method, in my
opinion, is JITW combined with Plurality:

   JITW//Plurality
   ---------------
   Each ballot counts for its highest-ranked non-withdrawn
   candidate.  The candidate with the largest count wins.

In my opinion, JITW//Plurality would be easier for voters to
understand than IRV.  It may perform better than IRV on important
criteria, but since its outcomes are not determined solely by the
ballots we need to examine incentives to understand it.  Candidates
will have an obvious incentive to voluntarily withdraw in order to
avoid being spoilers (the same as, given more primitive methods,
potential candidates have the incentive to not run at all and to
endorse a party's official nominee once the primaries are over).

I mentioned JITW//Plurality to Steve Hill when he visited the
Pasadena League of Women Voters in January, and he was concerned that
it would create the possibility of undesirable backroom dealing when
deciding who should or should not withdraw.  I have four responses to
that concern:

1) Such backroom dealing can take place anyway, in related forms.
All methods allow potential candidates to choose to run or not run,
for instance, and, where IRV is being promoted as a reform of
election systems which don't use partisan primaries, the backroom
effects could be quite similar to what they already have.

2) CV&D is already willing to endorse PR methods which have a
similar, if not worse, backroom deal problem: When cobbling together
the majority coalition needed to run the legislature and/or appoint
the prime minister, each MP has a full vote to sell any way s/he
wishes, whereas in JITW//Plurality each candidate can only step out
of the way of his/her voters' next choices.

3) Since the voters' orders of preference are published, there would
be considerable pressure on candidates to do the right thing.

4) During the period when candidates are considering withdrawal, the
media would present useful high-profile discussions about preference
order principles, which would eventually lead to the adoption of even
better methods such as those described below.

JITW can be combined with preference order methods which are better
than "plurality wins."  Combining JITW with better methods would
reduce the need for candidates to voluntarily withdraw, and thereby
reduce the opportunities for candidates to profit from backroom
dealing.  For example, given JITW//IRV or JITW//Condorcet most
candidates would not affect the outcome whether or not they withdraw;
they would have no bargaining chips.  (I wouldn't like JITW//Borda,
since Borda creates the incentive for a voting tactic which Gary Cox
calls "turkey-raising."  We can't be confident that a turkey will
gracefully withdraw, so turkey-raising could be effective in
JITW//Borda and the turkeys can cause spoiling.  Borda is really poor
on Arrow's IIA criterion.)

I'd have no trouble enthusiastically endorsing or JITW//IRV (or
JITW//Plurality), which I hope you will all agree would be only
slightly harder than IRV to explain to people.  It would be a
marvelous compromise if CV&D would agree to promote JITW//IRV.

Rob Richie wrote that Condorcet methods can allow the "tail to wag
the dog.  He also wrote that he deems it proper when PR parties form
a majority coalition for the coalition party with the most seats to
dominate the coalition, else it would be another case of tail wags
dog.  I mention this here not to try to persuade that the perception
of tail-wags-dog is inappropriate when the "tail" is actually the
center, but to point out that JITW//IRV would allow Rob's two big
"dogs" the freedom to choose in each election whether or not to "wag"
for more centrist candidates.

   45 ABC
   10 BAC          Given JITW//IRV:
   10 BCA          If C does not withdraw then A will win.
   35 CBA          If C withdraws then centrist compromise B will win.

Rob may believe the Dem and Rep parties would "collude" to not
withdraw when "appropriate."  That behavior would be legit with him.
As for me I have faith that the voters would successfully pressure
their candidates to withdraw when appropriate.  (Plus we'll have
gathered more info from the voters about their preferences than we
would with inferior methods, info which will be hard for those
seeking re-election to ignore.)  So I think JITW//IRV would satisfy
both Rob's criteria and my own.

Another reason CV&D promotes IRV, even though CV&D apparently
acknowledges that IRV would still leave the Dems and Reps all-
powerful, is that they view IRV as useful to educating people about
STV PR.  JITW//IRV would also serve to educate about STV.

I should point out that there is a PR analogy to JITW//Plurality
which is easier to understand than STV.  Should CV&D consider
promoting it?

    A preference order PR system which is simpler than STV
    ------------------------------------------------------
    1. After the votes are cast and before the final results are
       tallied, the voters' orders of preference are published
       (on the internet).
    2. Candidates may voluntarily place a cap on the number of
       votes they will accept.
    3. Each ballot is counted for its highest-ranked candidate who
       has not yet reached his/her cap.
    4. The candidates with the largest counts are elected.

I hope you will agree that this method is much easier to understand
than STV PR.

I believe it would behave much like the "random transfer" variation
of STV PR, since losers would have the incentive to choose a cap of
zero and winners would have the incentive to choose a cap around the
Hare quota.  A similar method could be devised which would behave
more like the "fractional transfer" STV method, but that would lose
the advantage of simplicity.

Just-in-time withdrawal may be a bit easier to explain than voluntary
cap, but either could be used in a single-winner method since
withdrawing is equivalent to choosing a cap of zero.  If the
Voluntary Cap PR method were being promoted, it might make sense to
use the voluntary cap terminology in descriptions of single-winner
methods as well.


---Steve     (Steve Eppley    seppley at alumni.caltech.edu)