[Election-Methods] Is this Condorcet method reasonable?

[Steve Eppley]

Hi,

In Diego Santos' example (below) the A voters who strategically 
downranked B had a strong incentive not to do so, assuming they were 
aware the B voters intended to truncate.  This highlights some 
advantages of the truncation strategy used by the supporters of the 
sincere Condorcet winner (B):

   1. It does not need to be kept secret, and benefits by being announced. 
      Announcing their intention to truncate creates a group strategy
      equilibrium (more powerful than a Nash equilibrium) in which
      the sincere Condorcet winner is elected, assuming the other
      voters believe the announcement is credible.

   2. It encourages sincere voting by other voters.

That assumes the voting method satisfies the Minimal Defense criterion 
(similar to Mike Ossipoff's Strng Defensive Strategy Criterion).  For 
more information about the truncation defensive strategy, follow the 
link to Minimal Defense at my website, which is at:
    http://www.alumni.caltech.edu/~seppley
The website is about the Maximize Affirmed Majorities (MAM) voting 
method, which other people often refer to as "Ranked Pairs (Winning Votes)."

* * * * *
A simple way to minimize the need for defensive strategy is to allow 
candidates to withdraw from contention after the votes have been cast.  
In other words, at the end of election day the votes would be published 
(without identifying information that would disclose how individual 
voters voted, of course) and then the candidates would be allowed a 
period of time (days, maybe weeks) to decide whether to withdraw.  After 
that period of time has elapsed, the votes would be tallied to determine 
the winner, with withdrawn candidates omitted from each vote.

In the example, if we assume candidate C's preferences are similar to 
the preferences of the 44% who rank C on top, C has an incentive to 
withdraw if necessary to defeat A.  If the 46% who rank A on top attempt 
to defeat B by strategically voting B at the bottom, candidate C can 
neutralize their strategy by withdrawing. 

Note that some non-Condorcetian methods, if coupled with the withdrawal 
option, would favor sincere Condorcet winners.  Instant Runoff would.  
Borda not so much, I think, due to the egregious way Borda fails the 
clone independence criterion.

* * * * *
Another technique to make preference order voting much simpler for the 
voters is for each candidate to publish an ordering of the candidates 
before election day.  On election day, each voter selects one candidate. 
(Nothing could be simpler.)  This would be tallied as if the voter had 
expressed the ordering published by the candidate s/he selected. 

Such votes can be quickly tallied, would ease the voters' 
information-gathering burden, and would reduce the best compromise 
candidates' need for campaign money. 

Australia's STV proportional representation elections provide the voters 
a similar (optional) shortcut: Before election day, each party publishes 
a ranking and on election day each voter may either select a party or 
rank the candidates.  If the voter selects a party, her vote is tallied 
as if she had expressed the ranking published by that party.  Most 
Australian voters use the shortcut. (Some folks in Australia want to 
eliminate the shortcut option and force the voters to rank the 
candidates.  Gary Cox, professor of political science at UCSD -- 
University of California at San Diego -- and author of Making Votes 
Count, likes the Australian shortcut.  He says it makes STV PR behave 
more like closed list PR, which tends to reduce the pork promised during 
election campaigns. In pure STV PR, similar candidates compete against 
each other for the same segment of the voters, creating the incentive to 
promise pork for that segment.)

If we assume candidate C's preferences are similar to the 44% who rank C 
on top, C has an incentive to publish C>B>A, to help defeat A. 
(Similarly, A has an incentive to publish A>B>C to help defeat C.)  Many 
of C's supporters would want C to publish C>B>A, so if C publishes any 
other ranking it would attract media scrutiny and probably reduce C's 
vote total. (That is, some of C's supporters would vote for B if 
necessary to defeat A, or for some 4th candidate similar to C, say C', 
who publishes C'>C>B>A, if there is such a candidate.)

It might be desirable to allow candidates to change their published 
ordering before election day.  If this is allowed, and if voters may 
vote before election day by mailing in an "absentee ballot," the 
absentee ballot should probably be treated as if it expressed the 
ordering published before the absentee ballot was mailed.

Given adequate technology in the voting booth, the voters could be given 
more flexibility.  Each voter would begin by selecting a candidate. This 
would immediately cause that candidate's published ordering to be 
displayed to the voter.  Then the voter would be able to rearrange the 
ordering, if desired. (Perhaps by drag & drop.  The NetFlix interface 
now allows each customer to rearrange his movie queue using drag & drop, 
by the way.)  When the voter is satisfied with the displayed ordering, 
she submits it as her vote.

Regards,
Steve Eppley
------------------------------
Diego Santos wrote:
> There were many discussions in this mailing list about advantages of winning
> votes as counterstrategy against order reversal. But sometimes truncation is
> risky. Consider this example:
>
> 46: A > B > C
> 44: C > B > A
> 10: B > A > C
>
> B is CW.
>
> Offensive strategy by A voters:
>
> 46: A > C > B
> 44: C > B > A
> 10: B > A > C
>
> A wins under RP(wv) or margins.
>
> If truncation would be used:
>
> 46: A > C > B
> 44: C > B > A
> 10: B
>
> C, the sincere Condorcet loser, wins.
>
> Winning votes induces truncation. Voters should feel free to express
> complete preferences.
>
> I was thinking in something similiar to "automatic truncation", i. e.,
> pairwise stregth in ranked pairs should be measured by plurality. If
> approval is used, the method becames DMC. Maybe approval cutoffs are not
> needed, then RP(plurality) is sufficient.
>
> RP (plurality)  or pairwise sorted plurality offers weak burial resistance
> and is summable, opposite to Smith,IRV.
>
> Diego Santos
>
>   
> ------------------------------------------------------------------------
>
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info
>