[EM] Dave on approval, ranked ballots

[Dave Ketchum]

On Wed, 27 Jul 2005 16:01:16 +0200 (CEST) Kevin Venzke wrote:

> Dave,
> 
> 
>>Please give an example, but:
>>      No IRV - let that be a separate project.
>>      No cycles - likewise, unless you state that there is no problem 
>>without cycles being involved.
>>
> 
> When no cycles are involved, Condorcet is 100% perfect. So I have to use
> examples with cycles:


Fine, so the voters are offering cycles - at least 3 conflicting opinions 
- to become a problem.

> 
> --- Dave Ketchum <davek at clarityconnect.com> a écrit :
> 
>>>>My comparison of methods:
>>>>     Approval - its backers like to brag, but it is not that simple - 
>>>>unlike Condorcet and IRV, voters CANNOT say:  I prefer Nader but, only if 
>>>>I cannot have Nader, give me Kerry as better than Bush.
>>>>
>>>However, ranked methods won't always obey your request. Condorcet methods
>>>and IRV will both sometimes take the vote "Nader>Kerry>Bush" and elect Bush,
>>>when if you had instead not ranked Nader, the method would elect Kerry. That
>>>is, the method ignores the fact that you said "If I cannot have Nader, give
>>>me Kerry."
>>>
> 
> 77 A>B>C
> 57 B>C>A
> 145 C>A>B
> 143 C=B>A
> 80 A>B>C
> 
> WV methods elect C. But if the 80 A>B>C voters instead lower A anywhere below
> B, then the winner is B. So the methods ignore these voters' wish that B be elected 
> if they can't get A.
> 

Huh?  When they SAID they preferred B, they GOT B.

However, when they said they preferred A, they were a minority part of a 
cycle in which major desire for C won.  They did not get ignored - simply 
were too small a minority to win.

> 
>>>Condorcet methods will sometimes see the "give me Kerry as better than Bush"
>>>part and give the win to Kerry, when it could have gone to Nader otherwise,
>>>despite the fact that your vote says "ONLY if I cannot have Nader, give me
>>>Kerry."
>>>
> 
> On these ballots:
> 49 Bush
> 24 Kerry>Nader>Bush
> 27 Nader>Kerry>Bush
> 
> WV methods (actually, all Condorcet methods) elect Nader. But if the 24 voters 
> just vote "Kerry," with no preference for Nader, then the winner is Kerry. So
> the methods ignore the 24 voters' request that they get Nader ONLY if they
> can't have Kerry.


A simpler set of votes, with no cycles.

Conceded that if the 24 were part of liking Nader, even without liking 
Nader best, they got Nader.

Now try an IRV example:
49 Bush
24 Kerry>Nader
27 Buchanan>Nader

Now Bush wins, while liked less than Nader, for:
      The 24 ballots get looked at and discarded.
      Bush wins over Buchanan.
      IRV ignores the 27 votes for Nader, without noting their existence - 
while Condorcet would have seen that Nader earned a win.

> 
> My point is that, at least with approval voting, the voter knows the limits of
> how the method will use his ballot. With ranked ballots one can't be sure that
> the method won't pick and choose what information to use, or even (in the
> first example) use the information to elect the ballot's last choice.


With approval I cannot SAY I like Nader, Kerry, and Bush, in that order.

With Condorcet I can rank, and get my ballots read, though minority votes 
properly rank below majority votes.

With IRV I usually do the same as with Condorcet, though IRV can and does 
ignore parts of some ballots, and has a different solution when Condorcet 
would see a cycle in a near tie.

NOTE - perhaps we are stumbling over language.  If the voter simply lists 
the candidates according to preference the results will please many.

> 
> Kevin Venzke

-- 
  davek at clarityconnect.com    people.clarityconnect.com/webpages3/davek
  Dave Ketchum   108 Halstead Ave, Owego, NY  13827-1708   607-687-5026
            Do to no one what you would not want done to you.
                  If you want peace, work for justice.