[EM] Runoff vs IRV

[Markus Schulze]

Hallo,

IRV has the property that it chooses the Condorcet candidate
at least in those cases where he is among the two finishers.
This suggests that the Condorcet efficiency of IRV is at least
as large as the Condorcet efficiency of Top-2 Runoff. The fact
that IRV also satisfies independence of clones and majority
for solid coalitions suggests that the Condorcet efficiency
of IRV is even larger than the Condorcet efficiency of Top-2
Runoff because these criteria mean that IRV is less vulnerable
to spoilers.

John B. Hodges wrote (26 Feb 2004):
> Looking at the question of filling a single seat when there
> are more than two candidates, Merrill used computer simulations
> to study the performance, measured by (1) "Condorcet efficiency"
> and (2) "average utility" , of seven different election systems,
> including plurality, top-two runoff, IRV-Hare, IRV-Coombs,
> approval, Borda, and Condorcet(defaulting to Borda in case of a
> cycle).  Under a variety of different circumstances, different
> assumptions etc., Plurality was consistently the worst and
> Top-two the second-worst.

Markus Schulze