[EM] SciAm article. Advantages of many candidates in poll.

[MIKE OSSIPOFF]

When I commented yesterday about the Sci. Am. article, it was incorrect to 
say that circular ties were only introduced in the paragraph that defined 
Copeland. They'd mentioned circular ties earlier, but it was in that 
paragraph that they brought up the matter of solving them.

They didn't call Copeland's method by name. It seems to me that they 
described it in 24 words. As I said, they didn't advocate Copeland, but 
merely said that it perhaps is the simplest circular tie solution.

As I said, their name for pairwise-count, "Majority Rule", is a good one, 
and a bold one. And, as I said, it's a deserved name, because pairwise-count 
methods are the proposed & well-discussed methods that have the potential 
for best honoring majority rule. For pretty much guaranteeing majorilty 
rule.

I should emphasize that it apparently isnt' _only_ pairwise-count methods 
that can do that, since apparently NES & DSV act very similarly to wv 
Condorcet.

Sure, the article should have mentioned Approval, which would only have 
taken a few words.

But that omission is understandable because many advocates of  Approval 
don't advocate pairwiise-count, and many advocates of pairwise-count don't 
advocate Approval.

Overall, in spite of not mentioning pairwise-count methods other than 
Copeland, the article is still a good thing. For pairwise-count methods to 
be advocated in a national magazine, even if only in its online edition, is 
valuable.

I hope tha that article is in the print issue of Sci Am too.

I like the Majorilty Rule name, but wv continues to live up to that name 
even in circular ties, even with truncation or offensive order-reversal. 
That's because the whole notion of "majority" is a wv notion.When you speak 
of a majority for A over B, you're saying that a majority have voted A over 
B.
You're talking about the number who have voted A over B. You're talking 
about winning-votes.

It's therefore no surprise that wv is the method that really fully honors 
majorilty rule. Not margins. Not Copeland. Copeland and Margins forget about 
majority rule if there isn't a BeatsAll winner.

Why did the Sci Am authors only mention Copeland? They may have been running 
out of space, as suggested by the brevity of their description of it. Also, 
Copeland and Black are the only pairwise-coiunt methods that have gotten 
much acceptance and discussion among the academics.

Since Condorcet wrote about pairwise-count methods in the 18th century, and 
there's been a fair amount of discussion of it among academics, no one could 
say that mention, and even advocacy, of pairwise-count isn't 
academically-sanctioned. And the releatively more-discussed Copeland is the 
more academically-permissible circular tie solution for pairwise-count.

The value of the article isn't llimited to the value of Copeland. The value 
of the article is that it advocates pairwise-count, and brings it to 
people's attention as a solution to the problems of currently-used voting 
systems. Sure, Copeland itself isn't a solution, but the article still 
accomplishes a lot just by introducing and recommending pairwise-count to 
Sci Am's wide audience.

The article is "Chapter One", the first installment and big breakthrough in 
useful voting system discussion in a big publication. Obviously Chapter Two, 
the essential matter of how to solve circular ties in a way that continues 
to protect majorilty rule, remains to be published. But the article is still 
a valuable start, and is a lot better than nothing.

By the way, another academic author surprised me by advocating Condorcet's 
method in an article in the Op-Ed section of the Washington Post, for June 
21, 1992. The title of the article, which, it seems to me, was at the top of 
the page in that section, was "The Last May Be First: Factoring Perot". Or 
maybe "Factoring Perot: The Last May Be First". That was remarkable. 
Academic advocacy of Condorcet's method to the publc is something that I've 
never heard of before or since.

***

It's a good thing to have so many candidates in the poll. It makes it more 
likely that we'll get a circular tie, which can better bring out differences 
among methods. (Not that it takes a circular tie to make IRV fail). 
Especially since, with 46 candidates, lots of truncation is inevitable.

By the way, it would be interesting to also count the rankings by Bucklin. 
Of course Bucklin's defensive strategy involves truncation. Because rankings 
are so much better-counted by Condorcet wv, Bucklin isn't an important 
enough method to nominate or to justify asking people to indicate a Bucklin 
cutoff. But, for the purpose of an unofficial Bucklin count, one could use 
Approval cutoffs as Bucklin cutoffs, making the assumption that, for 
Bucklin, only the Approval-voted candidates are ranked.

I should add that I realize that the rankings and ratings in my ballot are 
inconsistent. When I changed my ballot, I changed some candidates' ratings 
but neglected to change their rankings, and changed some candidates' 
rankings but neglected to change their ratings. I'll fix that by re-voting 
again within a few days.

With the existing candidate lineup, if it were a real presidential Condorcet 
election, I'd rank all the candidates, however many there are. But, if it 
were a real Approval presidential election,  even with the 2004 candidate 
lineup, I wouldn't extend my Approval votes any lower than I did in this 
poll. With candidate lineups like those of previous decades, I'd probably 
not rank anyone for whom I wouldn't vote in Approval.

Mike Ossipoff

_________________________________________________________________
Click, drag and drop. My MSN is the simple way to design your homepage. 
http://click.atdmt.com/AVE/go/onm00200364ave/direct/01/