[EM] ironclad pro-Condorcet argument?

[Steve Eppley]

Dave K wrote:
> On Tue, 24 Aug 2004 01:06:48 -0700 Steve Eppley wrote:
>> Dave K wrote:
>>> On Mon, 23 Aug 2004 11:29:48 -0700 Steve Eppley wrote:
>> -snip-
>>>> Another positive argument for Condorcet-consistency 
>>>> uses the single-elimination pairwise voting procedure 
>>>> recommended by Robert's Rules,
>> -snip-
>> 
>>>> Most of the people reading this, I assume, are aware that 
>>>> under the Robert's Rules procedure, the Condorcet winner 
>>>> (when there is one) will be chosen, assuming either 
>>>> that every voter votes sincerely or that every voter is 
>>>> strategically sophisticated and knows the preferences 
>>>> of all the voters.
>>>
>>> HUH???  The Robert's example is IRV.
> 
> AND, they express dislike for Condorcet by their example voting
> procedure for preferential voting - the procedure shared by IRV
> and Condorcet. 

I doubt their omission of Condorcet was an expression of 
dislike for Condorcet.  Remember, that section was written
long before the computer age that made it practical 
to exhaustively tally all the pairings given voters' 
preference orders.  And at the time, not a lot of people
were aware of Condorcetian procedures.

>> No, I wasn't referring to the Robert's Rules IRV example
>> of "preferential voting" that they reluctantly recommend 
>> when the members are scattered, as in a mail-in vote
>> (when better methods are impractical, or so they thought
>> when that section was written long ago).  I was referring 
>> to their main method, recommended for use when the 
>> members are assembled together and hence it's practical
>> to use a method that requires multiple rounds of voting.  
>> This method is sometimes called "agenda voting" and 
>> sometimes called "sequential pairwise voting."  It
>> doesn't ask the voters to express orders of preference.
>> And it's like a single-elimination tournament, not 
>> a round-robin tournament.  
> 
> Please tell me exactly where they, inconsistently, 
> express approval for Condorcet.

I never claimed they do.  Don't get this discussion of
"an ironclad pro-Condorcet argument" mixed up with the
recent discussion of the "definition of preferential 
voting."  What I've been saying is that the voting 
procedure that Robert's recommends for _normal_ usage 
within assemblies will elect Condorcet winners.  It's 
a shame that Robert neglected to mention his procedure 
will elect Condorcet winners, but that omission isn't
critical.

It might be worth searching their text for the 
phrase "best compromise."  Robert criticized IRV
because it can easily defeat the best compromise.
Maybe he used the phrase elsewhere?  Maybe he even
defined it somewhere?

> What I see is repeated balloting of the entire question - 
> not even deletion of weakest candidates, as would earn
> the label "runoff".
> 
>> Here's a simple example:  Someone proposes a bill and 
>> someone else proposes an amended version of the bill.  
>> In the first round of voting, those two alternatives 
>> would be pitted against each other.  The loser of 
>> that vote would be eliminated.  The winner of that 
>> vote would go on to the next round of voting, 
>> a vote between it and the status quo.
> 
> This example is not Condorcet - in Condorcet all the 
> versions of the bill would contend in a single election.

Right, we're defining Condorcet as a family of voting
procedures that accept preference orders from the voters
and elect the Condorcet winner, if there is one, given
those votes.  

But this Robert's procedure nevertheless satisfies the 
Condorcet criterion, which requires the election of the 
Condorcet winner if there is one.  It will elect the 
sincere Condorcet winner given either sincere voting 
by everyone in each round or strategically-sophisticated 
voting by everyone in each round.  And it also satisfies 
the top cycle criterion.  

The "ironclad" argument is about requiring the election 
of Condorcet winners when they exist, and within the
top cycle in general, since that's what happens in 
democratic assemblies, where methods better than IRV 
have long been feasible.  The argument is not about 
the mechanics, such as whether the votes must be orders 
of preference.

--Steve