[EM] IRV vs Condorcet

[Adam Tarr]

Dave Ketchum wrote:

>      Approval - actually more complex for the voter than ranked voting - 
> all that is needed for rank is better vs worse, Approval requires 
> judgment as to cutoff, but does not let the voter express the ranking 
> required to get this far.

An interesting point - approval may be easy to understand HOW to vote in, 
but voting intelligently is probably a little harder than in a good 
Condorcet implementation.

>      PR - debate this another day - for the moment I am staying with 
> elections with single winners, and not concerned with legislatures where 
> PR MIGHT make sense.

PR definitely makes sense.  There's a lot of debate here about what sort of 
PR is the best, but saying that means PR isn't a good idea would be like 
saying plurality is good because we can't agree on the best alternative.  I 
know that's not what you're arguing about PR, but I'm just making the point.

>      While IRV gets away from most of Plurality's spoiler problems, it 
> has a few of its own.  Condorcet simply DOES NOT DO spoilers.

While winning votes-based Condorcet lacks the conventional spoiler problem, 
it is still definitely present in margins-based Condorcet.  (Markus, here's 
an example of a Condorcet supporter using a clones-type spoiler.)

Bush 49%
Gore 51%

Obviously Gore wins (save the jokes).  Now add in Nader:

Bush 49%
Gore 24%
Nader>Gore 27%

This example looks pretty similar to the eye to the ABC one you used, only 
this one has the same spoiler as IRV when using margins-based 
Condorcet.  Winning votes Condorcet still gives the election to Gore by a 
wide margin (sorry).

Furthermore, one of the tried-and-true criticisms that IRVists use against 
Condorcet is "your second-pace vote can help defeat your first place 
vote."  This is virtually never true in winning votes, but it is frequently 
true in margins.  In the example above, if 2% of the Bush supporters 
expressed a sincere second preference for Gore, they would cause Gore to 
beat Bush.

-Adam