[EM] Why Condorcet should be more popular than IRV (was: Why IRV is popular)

[Dave Ketchum]

On Sat, 12 Mar 2005 15:00:36 -0800 (PST) Alex Small wrote:

> I mostly agree with Russ.
> 
>  
> 
> Another reason why IRV is popular is that it's basically an expanded and 
> automated version of an election method already used for many local 
> elections in the US, as well as elections to some higher offices in at 
> least 2 states (Nebraska and Louisiana):  2-step runoff.  People are 
> familiar with 2 step runoff, and it makes intuitive sense.  Yes, I know, 
> the majority generated by runoffs isn't as legitimate as being the 
> Condorcet Winner, but at first glance it sounds good.
> 

2-step runoff is another snake that bites with poisoned fangs occasionally 
- ask the French what they remember of it.

IRV, itself, uses the same ballot as Condorcet, and usually agrees as to 
winner.  But IRV does not look at all of every voters' rankings and thus 
can be bitten by IRV's spoiler problem, though not as often as Plurality.

IRV does have vocal backers, and is used some places because of their 
efforts.  Here we hear from Condorcet backers.

While those into history know of Condorcet as inventing this voting 
method, it uses ALL that each voter says about ranking to produce an 
array/matrix of what the voters collectively say about the candidates.  In 
counting each voter's ballot:
      For each pair of ranked candidates, the ranking indicates which won 
(or that they tied, if the voter is allowed to say that, and chooses to).
      Each unranked candidate does not play, and thus loses to each ranked 
candidate, while not winning any games.
      Note that, just as ballots from multiple voters in a precinct are 
summed in one array, arrays from multiple precincts in a district can be 
summed in one district array.

Usually the array will show one candidate winning over each other 
candidate, and thus winning the election.

However, we can have a cycle of near ties such as A>B AND B>C AND C>A. 
Here the MAJOR need is to have a rule for breaking the cycle, though there 
are endless debates as to what rule might be best (remember that the array 
shows each other candidate is a loser, while each of these have some claim 
toward winning).  Possible rule:
      The largest vote count among the cycle pairs wins.

     If a tie, break it by awarding to smallest vote count among tied winners.

      If still a tie, flip a coin or draw straws.

Painfully, there are two classes of such rules:
      Rules that depend only on the array Condorcet is expected to create, 
such as the above, are interchangeable.
      Rules that require additional information such as for Approval 
require all involved to understand and act their parts.

>  
> 
> If cycles were impossible then I think Condorcet would have a real shot 
> at public acceptance.  But since cycles are always a possibility, we 
> have to include them in the discussion.
> 

ASSUMING we use the words that we use within EM, we can expect trouble - 
seems to me the wording I use above should need little editing for reading 
by average voters.

>  
> 
> On the other hand, I don't follow sports at all but a friend was telling 
> me that in some tournaments the teams will start off in groups of 3, and 
> will play 3 games among themselves.  The undefeated team in each group 
> will advance, and if there's no undefeated team in the group then 
> they'll do something with the margins of victory to figure out who 
> advances.  That sounds awfully similar to some Condorcet methods.
> 

What they do is living with human limits as to how many games to play. 
What they say might be comparable to what we might say.

>  
> 
> Does anybody know anything about these methods and whether sports fans 
> find them too confusing?
> 

Almost certainly they will be doing descriptions to make themselves 
understood.

>  
> 
> Alex
> 
> Russ Paielli wrote:
> 
>     Date: Sat, 12 Mar 2005 12:35:22 -0800
>     From: Russ Paielli <6049awj02 at sneakemail.com>
>     Subject: [EM] Why IRV is popular
>     To: election-methods at electorama.com
>     Message-ID: <4233528A.7080509 at sneakemail.com>
>     Content-Type: text/plain; charset=ISO-8859-1; format=flowed
> 
>     Eric Gorr recently posted a link to a one-page document by Jim Lindsay
>     explaining why he and many political activists prefer IRV to Condorcet,
>     Approval, and other methods.
> 
>     One of Jim's "criteria" was "system easily explained." Surprisingly, he
>     put "somewhat" for both IRV and Condorcet.
> 
>     IRV is much easier to explain than Condorcet, and I believe that is the
>     primary reason that it is more popular.


DISAGREED!  IRV backers put more effort into selling, and being quiet when 
they have nothing to brag about.

> 
>     An IRV promoter simply explains that the voter ranks the candidates,
>     and
>     the first choices are counted. So far the counting procedure it is
>     identical to our current plurality system. Then the promoter explains
>     that if nobody gets a majority, the candidate with the least votes is
>     eliminated, and any voter who had that candidate at the top of his list
>     has his next choice bumped up to the top. The counting is then
>     repeated,
>     again as if it were a plurality election. The process repeats until
>     some
>     candidate gets a majority.


I  question whether that is a correct use of the word "majority".  Said 
that way, the listener can think it is a true majority of all the voters, 
rather than what it is - a majority of what is left after discarding 
exhausted ballots (not questioning what they do - Condorcet is equally 
willing to declare a winner without a true majority, but not into the same 
claims).

> 
>     My point is that the counting procedure is very similar to our current
>     system. The only twist is the elimination of candidates and transfer of
>     votes, which most people can grasp fairly quickly.
> 
>     The fact that IRV is non-monotonic and non-summable never even
>     occurs to
>     perhaps 99% of those who hear about it, and the activists who know
>     about
>     these deficiencies don't consider them important. IRV is popular
>     because
>     its rules are simple and the basic counting procedure at each round is
>     identical to plurality.
> 
>     The same cannot be said for Condorcet, particula rly when it involves
>     dropping of defeats. Just explaining the pairwise races and matrix is
>     already more complicated than IRV. Then when you start talking about
>     dropping defeats, people start wondering what sort of crackpot
>     scheme it
>     is. By the time you get into the actual *rules* for dropping defeats,
>     the game is over. Get into "margins" vs. "winning votes" and the poor
>     listener wishes he were late for a root canal.


If you are determined to commit suicide with confusing words, so be it. 
My claim is that, with a little effort, you can say it in the English your 
  intended audience understands.

> 
>     "Traditional" Condorcet with dropping of defeats may be appropriate for
>     organizations of people with a special common interest, but it will
>     never be accceptable for highly contentious public elections. The Bird
>     Watchers of America may be willing to agree to use defeat-dropping
>     Condorcet, but a large public jurisdiction never will. And think about
>     what would happen if they did: the losing side would immediately become
>     biased against the method. It's just too complicated and too arbitrary.
> 
>     The general public expe cts the rules of a public voting system to be
>     simple and easily understood. The public acceptability of a method goes
>     down by perhaps nearly an order of magnitude for each additional
>     sentence required to explain it. That's what I think, anyway.
> 
>     --Russ

On Sat, 12 Mar 2005 21:53:11 -0500 James Green-Armytage talks of getting 
those who can to use Condorcet and make it visible.

Great idea, except in messy elections it is hard to do without a computer 
reading the ballots.

Still, a bit of thought:
     Have a program available for the messy ones.
     Stack the ballots according to top rank - usually can call winner from these.
     Tell the program how many ballots of each pattern voted, and let it 
calculate winner.

He later talked of sequential dropping - words too technical for me to want 
to mention them.

-- 
  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.