[EM] Ruminations on strategy issues in IRV and Condorcet (was possible improved IRV method)

Jonathan Lundell jlundell at pobox.com
Fri Jun 30 07:36:47 PDT 2006

At 9:28 AM -0400 6/30/06, raphfrk at netscape.net wrote:
>What about a method with determines the IRV winner and the Condorcet 
>winner and then selects one of them using a random ballot.  I assume 
>that someone has already suggested it.
>It seems to me that if you included a reasonable number of election 
>methods, then strategic voting would become considerably more 
>difficult.  This is esp. true if the voters must use the same 
>ballots for all methods. Also,
>ballots would be required to be consistant (or it collapses back to 
>each voter voting strategically for each individual election).
>A ballot could be a range score ballot with an approval cutoff 
>score. This allowes a ranked ballot and an approval ballot to be 
>generated. Also,
>it ensures consistancy.  The IRV would have to handle equal rankings 
>though (or require that 2 candidates cannot be given the same range 

It's an interesting possibility. However...

An advantage of IRV and of Condorcet when there's a Condorcet winner 
is that they're (relatively) simple to explain to voters. Maybe not 
ab initio, but my experience is that voters quickly "get it", and 
generally approve of the process. Condorcet methods are a bit more 
complicated to explain in the face of multiple-member Smith sets, but 
let's assume that it's manageable.

Complexity of voting systems is a major political problem when we try 
to introduce them to naive voters, though. I'd resist (or at least 
defer) features like ratings or range scores that make the system 
more difficult for naive voters to grasp.

The major problem with STV systems, including IRV, is that they don't 
find compromise candidates of the "everybody's second choice" 
variety, since those candidates get excluded prematurely. I'm not 
entirely convinced (obviously) of how serious a problem that is in 
the real world.

We talk a lot about left/right/center candidates, in part, I think, 
because our advocacy of ranked ballots (whatever the counting 
mechanism) encourages the illusion that candidates fall along a 
one-dimensional scale.

And of course they do not. As voters, we evaluate candidates on a 
variety of scales, objective and subjective, conscious and 
unconscious. And each voter must then map their multivariate analyses 
of the candidates to a one-dimensional preference scale. But it's not 
justified to leap from that mapping to the implicitly conclusion that 
there's anything like a one-dimensional ranking of candidates that 
comes close to capturing the real complexity of positions and issues 
and subjective judgements.

/Jonathan Lundell.

More information about the Election-Methods mailing list