[EM] Re: The Allure of IRV

[Adam Tarr]

Dave wrote:

>I claim that approval is both weak enough for many of these to notice, and 
>more difficult than its backers are willing to admit:
>      Weak because, if I approve of two candidates, I CANNOT indicate my 
> preference between them.

True, but it can hardly be considered a step back from plurality.  I don't 
have to vote for the second candidate if I don't want to.  Furthermore, our 
discussions of equilibria in approval voting have suggested that despite 
this apparent weakness, the Approval winner will tend to be the Condorcet 
winner if one exists.

>      More difficult because where, in the scale between like and dislike, 
> is the point where I should stop approving candidates?

That has been the subject of some vary complex threads on this list.  They 
are really getting far more technical than they need to, in my 
opinion.  The simple strategy that works fairly well is to vote for your 
favorite of the front-runners (just as you would in lone mark plurality), 
and also vote for any candidate you like more than that 
one.  Strategically, this is no harder than voting in the current system.

So, in summary, approval voting can be done as easily as current voting, 
and it strongly tends toward the election of the Condorcet winner.  I 
disagree with the sentiment that approval is either weak or difficult.

>Condorcet backers claim advantages over IRV by Condorcet's looking at 
>whole ballots while IRV only looks at the currently top name in each ballot:
>      Usually they select the same winner, having similar ability to 
> ignore fringe candidates, this process usually leaving only two final 
> candidates to choose between.

They usually produce the same winner in the case of two strong candidates 
or factions, and relatively complete ballots.  Not the rest of the time.

>     Here IRV will see the 40% C, 25% PL, 20% XL, and 15% visible L; 
> discard those L votes, discard the XL votes and the L votes behind them, 
> and declare C the winner.  Condorcet will see the 60% L, discard XL and 
> PL as each less liked than C or L, and correctly declare L the winner.

This is an example of how IRV can break down, even when there are only two 
strong factions.  Even more trivial is showing IRV breaking down in the 
three-faction case.  Say the left candidate has 35% of the vote, the middle 
candidate 30%, and the right candidate 35%.  IRV will drop the middle 
candidate, and end up with an extremely tight race between left and 
right.  Condorcet will recognize that while the middle candidate has 
slightly less first-place support, middle is the second place choice of 
everyone, so middle beats both of the other candidates by 30% each.

-Adam

----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em