[EM] breakdown of Oakland mayor ballots

[Dave Ketchum]

On Nov 13, 2010, at 10:07 PM, Kathy Dopp wrote:

> On Sat, Nov 13, 2010 at 9:46 PM, Dave Ketchum <davek at clarityconnect.com 
> > wrote:
>> Actually, "differs from plurality" is not a sure indication of  
>> failure for
>> ranked voting - either IRV or Condorcet.
>>
>> For Plurality, voter can name only ONE candidate.
>
> David,  Not sure why you are restating the obvious.    You also are
> mischaracterizing what I said. Let me try again so you can hopefully
> get my meaning:

We make differing statements.  Out of all that, I am trying to be  
correct:

Condorcet reads all the data in the ballot, making it practical for  
ballots to contain voters' desires.

IRV makes educated guesses as to what, in the ballots, is losers to be  
discarded.  If, as an extreme example, Good is the second candidate in  
EVERY ballot, when N Goods are exposed by discarding what precedes  
them, those N Goods will qualify for immediate discarding.

Plurality accepts one name per voter.  A likely loser who could, per  
voter desire, be top preference in Condorcet, does not belong here -  
voter better vote a possible winner.

Thus those that made sense as top ranks in Condorcet do not deserve a  
claim that a Plurality voter should have made the same choice.
>
>
> I said that for the first time that IRV/STV in San Francisco found a
> different winner from the winner who would have won using plurality
> methods (only counting voters' first choices). In other words the
> IRV/STV winner has been the voters' 1st choice winner until now, so
> that the extra costs of using IRV/STV produced essentially nothing
> until this election.

Seeing misuse of "plurality" does not make me disagree with your  
thought.
>
> My statement has nothing to do with other means of counting rank
> choice ballots such as Condorcet. SF does not use Condorcet.

It is partly their use of almost identical ballots that gets me to  
mention IRV and Condorcet in this discussion (any IRV ballot should be  
usable in Condorcet).
>>
>> For ranking, voter can vote for more than one, perhaps ranking most  
>> desired
>> over more gettable.
>>     Condorcet will use all that is in the ballot.
>>     IRV will discard some, use the top of what is left, and never  
>> see the
>> bottom of what is left.
>
> Not sure that anyone on this list doesn't already know that, so not
> sure why you're restating the obvious Dave.
>>
>> So, three different methods each seeing parts of the voter's  
>> ballot.  My
>
> You just mentioned two methods of counting rank choice votes, not  
> three.

I counted Plurality here even though it covers its whole ballot.
>
>> guess is that the top of the IRV ballot is what the voter is  
>> presumed to
>> have chosen in Plurality.
> What I said is copied below in case you're wondering what I said.
>
> Warren Smith has proven that IRV has a very high probability of
> producing vagaries in cases when it does not produce the same results
> as a plurality election would. Most often, as we've seen, IRV/STV
> produce the same results as a plurality election in practice, as was
> clear if you read the news articles on this SF election.
>
> When IRV/STV does not replicate plurality results, it has a high
> probability of producing results that are nonmonotonic or eliminate
> the voters' favorite candidate (Condorcet winner).
>
> I hope that clarifies for you. I apologize failing to include enough
> information for you to understand what I said.  Perhaps you had not
> read the two news articles on the election like I had, nor read Warren
> Smith's prior article that shows how this latest election dodged a
> bullet.

I cannot keep tack of all that I have read.
>
>
> Again, there will not be very many cases like this when IRV fails to
> elect the plurality winner and actually elects the Condorcet winner.
> I suggest you read Warren's article on that issue.

The ugly cases are those in which IRV discards the Condorcet winner -  
best we can have is for them to be VARY rare.

Dave Ketchum
>
>
> Kathy