[EM] Re: Testing 1 2 3

Kevin Venzke stepjak at yahoo.fr
Mon Jan 12 14:54:13 PST 2004


If I may...

 --- Dgamble997 at aol.com a écrit : 
> This sounds like quite a good idea. I do see one or two potential problems 
> with it though:
> 1/ How from a CR ballot do you determine the ranked ballot truncation point ( 
> the level of utility at which voters cease ranking candidates).

This is certainly a problem, especially in that different methods require
different strategies.  In Condorcet methods, ranking only those candidates
that you would approve is not typically the best strategy, for instance.
I would note that you can't get away from this problem simply by starting 
with rankings that are already truncated, since the question remains of
what exactly it means when a ballot doesn't rank certain candidates.

> 2/ Lack of real world comparison data. I adjusted the votegenerator part of 
> the spreadsheet as a result of looking at the data sets it was producing and 
> comparing them with the South Australia state election of 1997 ( 3 parties- 
> Liberal/National, Labor and Australian Democrat made a reasonable showing ) and 
> the Queensland state election of 1998 ( Liberal/National, Labor and One Nation). 
> The distribution of second preferences in the votegenerator data sets was 
> much more variable than those in real elections and votegenerator was adjusted as 
> a result of this.

This sounds like you're forcing your generator to assume a one-dimensional
spectrum.  If it were me, I wouldn't do that, because it limits the applicability
of any conclusions you draw.

> I know that you've made several posts on the subject of converting CR ballots 
> into Approval ballots could you tell me where they are?

Unfortunately, those methods repeatedly convert the CR ballots into Approval
ballots based on the results of each iteration.  I dare say the point of them
is do the cutoff-placement work for the voter.  Not much help to you, though, 
I think.

Off the top of my head, I might use some crude method to find two front-runners
from the ballots, and then form the Approval ballots as though every voter
believes that only those two candidates have a chance at winning.  Given voter
utilities for each candidate, optimal Approval ballots can be found very

> Just a quick point, comparing different 'flavours' of Condorcet is actually 
> very boring as Condorcet cycles are fairly rare unless you produce data sets 
> based on unrealistic preferences within the electorate for example:

You don't have to produce the sets, just the matrices.  Open up "mspaint" and 
plot four points, ABCD, and draw different colored arrows (representing defeats 
of different strengths) among them, and see what winners are picked by RP, Schulze, 
Raynaud, etc.  It's interesting, I swear.

Also, cycles are not so rare if voters give a short ranking.  Take the favorite:

49 A
24 B
27 C>B

That's a cycle on paper, though maybe not in actuality.

Kevin Venzke
stepjak at yahoo.fr

Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français !
Yahoo! Mail : http://fr.mail.yahoo.com

More information about the Election-Methods mailing list