[EM] [CES #4445] Re: Looking at Condorcet

[Juho Laatu]

On 10.2.2012, at 0.59, robert bristow-johnson wrote:

> On 2/9/12 5:19 PM, Juho Laatu wrote:
>> Condorcet is a natural extension to the multi-candidate case (still assuming competitive elections). Maybe not the only one though. In another mail I just addressed the possbility of having single-winner elections that aim at electing the winner from one of the major parties (or more accurately, from one of the top two most representative camps).
> 
> now why is that a good thing?

Some targets could be set so. (I don't say that tis approach would be particularly good, just one option and approach that is less common than the Condorcet approach.)

>  i thought that giving third and fourth parties (and independents) a level playing field is an intended consequence of the main virtue of ranked-ballot methods which is to not penalize the voter for voting his/her conscience and does not offer a reward (even if the reward is a consolation prize) for voting tactically.

Yes, those are some typical essential requirements for methods that allow more than two candidates to run and not be a spoiler. There could be differences on which candidate we want to elect. Is the target to elect a candidate that is good for as high number of voters as possible (with "least opposition"), or is the target to elect a winner that represents one of the two strongest coalitions (with no need to be good for both coalitions)?

> 
>> Condocet could however maybe be seen as the most natural extension and a natural fisrt proposal for typical / basic single-winner elections.
> 
> that's what i thought from the beginning.  i remember when we voted for IRV in 2005 thinking that the language of the rules were not this natural extension and thinking "why did they adopt this method for dealing with the ranked ballots?".  but it occurred to me right away that all the pairwise champion needs to do is get into the final round, and he/she will win it.  so i thought it was probably unlikely that IRV would elect someone else.

IRV is a method that loks good at first sight, but that has some problems. It clearly seems to respect the idea that a pairwise winner should win at the last round, and maybe be strong also in the other rounds. I think IRV has also another nature, i.e. the tendency to favour large paries. This property correlates with the othe approach to single-winner methods that I discussed earlier. But also here IRV is quite heuristic and approximate, and does not systematically implement that approach either. It is thus possible that some people see IRV as "a bad Condrcet-like method", and some see it as a "improved two-party method with some problems". The method that I described earlier could be seen to be a cleaner method that represents this other approach to IRV (while Condorcet represents another approach to it).

> 
>>>  if Candidate A is the best candidate to be awarded office, that means that Candidate A is better than Candidate B.  it also means that Candidate A is better than Candidate C.  if Candidate A is the best candidate, it means that no other candidate is better than Candidate A.
>>> 
>>> so, how do we determine who is better?  we could make them take an exam to show how much they know about job that the elected office entails.  or we could make the candidates arm wrestle.  but, in a democracy, the way we determine that one candidate is better than some other is that we ask the electorate.  sorta like Pilate asking the crowd to choose between Jesus and Barabbas.  the ranked ballot tells us who the voter chooses given any pair of choices.
>>> 
>>> it's simple.  when a Condorcet winner exists, to elect *anyone* other than the Condorcet winner is the same as awarding office to the loser in a simple Two-candidate, Simple majority, One-person-one-vote election and i cannot see a *single* justification for doing that.  the "weak CW" argument does not cut it at all.
>> For typical single-winner elections, yes. In special cases, like the one that I discussed above,
> 
> yeah, but i don't get why that special case is desirable.  why should rules be adopted that favors the top two parties as an end itself?

If you want to elect the best compromise candidate, Condorcet is good. If you want to keep one of the leading camps in power, use the other approach. I thus assumed that in some environments and some electiosn people want two leading parties to rule in the typical aalternating way, while giving also third parties the possibility to participate in the election without becoming a spoiler, and one day replacing one of the two leading parties. Just a choice on what kind of a society / election you want to have.

> 
>>  also other approaches may be possible.
>> 
>> So I agree that Condorcet methods are a good first assumption (for competitive elections).
> 
> i just don't see the alternative (which is that you elect the loser of some pair of candidates).

Ok, the pairwise loser problem needs an explanation. I'll try to find one.

There are two 50% parties, A and B. They alternate in power. Then a new minor party C emerges. In the political spectrum C is somewhere between A and B. Already in the next election A voters vote 49: A>C>B and B voters vote 49: B>C>A. C is a Condorcet winner. A will be elected althoug C is the pairwise winner. The reason is that the society wants the winner to come from an established party that can take care of the country/state/city well for the whole term in office. The winner needs support from numerous people from his own party. Therefore C would not be a good choice. We want people to show more support to C before we allow C to win. C would be a good ambassador that could represent the country/state/city well in some other role, but we want the the winner of this election to be figure with massive supporting troops behind him.

In the next election C has gained popularity and has lots of active supporters. Now A suporters vote 45: A>C>B. B supporters vote 25: B>C>A. C party has now grown bigger than B party, and as a result we allow C to win the election. In addition to beating the other candidates in a pairwise comparison, candidate C needed also massive (first preference or preference over A and B) support to win.

The reason for having such rules was that we like the idea of having two alternating strong camps in power, and we like the idea that those parties who alternate in power should be the two strongest parties or groupings.

Juho



> 
> -- 
> 
> r b-j                  rbj at audioimagination.com
> 
> "Imagination is more important than knowledge."
> 
> 
> 
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info