[EM] Issues, Condorcet, and IRV (was: IRV vs. plurality)

[Eric Gorr]

>Here is a concrete example where Tideman's ranked pairs method
>violates the Participation criterion in a very drastic manner.

(In a form for my site)

03:A>C>B>E>D
12:A>C>E>B>D
07:B>A>D>C>E
02:C>E>B>D>A
03:D>A>B>C>E
09:D>C>B>E>A
01:D>E>B>A>C
05:E>A>C>B>D
10:E>B>D>A>C
01:E>C>B>D>A
03:E>D>C>A>B
04:E>D>C>B>A
03:A>D>B>C>E - the extra votes - call this line 'L'

If L was <= 2, A still wins
If L is 3 or 4, E wins
If L is 5, 6 or 7, B wins
If L is 8 or 9 there is a tie between A & B
If L is >= 10, A wins

But, why should it surprise anyone that discovered votes could change 
the winner?
It seems to me that in every election system this would be true.

What seems clear, based on votes, is that the voting population is 
quite confused on who should win.

In any of these cases, once all of the sincere votes were tallied, 
did someone win who, unambiguously, should not have won? This answer 
would appear to be no, whereas, in the case of IRV, the answer is 
often yes.