[EM] Election-Methods Digest, Vol 222, Issue 21

[Kevin Venzke]

Hi John,

> I looked at something similar when writing Vote::Count 
> [https://metacpan.org/pod/Vote::Count]. I couldn't find any literature 
> on it and it seemed like a great idea, I called it CondorcetVsIRV or 
> Redacting.
> 
> When I began to look at how it performed with data from elections where 
> Condorcet and IRV do not agree, it almost always chose the IRV winner, 
> demonstrating that in most cases where the two don't agree that there is 
> usually a later harm effect on the Condorcet side. I think the approach 
> could be  valuable as a tool for measuring later harm effects, but not 
> worth the complexity for use as a real world method.

Hmm, I can see the similarity. If I understand the page correctly:
https://metacpan.org/pod/Vote::Count::Method::CondorcetVsIRV

It looks like you find the actual IRV and Condorcet winners directly as a first step. I'm
fairly sure the latter part of that won't work for my purposes. It would be interesting for
me to check whether there is a way to use the original IRV winner with my approach,
although I guess it won't, since I already studied using top-two runoff and reached the
conclusion that while it can be done, the ballots then need to be truncated even more.

The page seems to say that some version of CondorcetVsIRV should satisfy Later-no-harm? Or
is that just relative to most other methods? I don't know any "Condorcet vs. ____" method
that satisfies it, so if it really can (or might) I'd take a look.

I think my ACP idea is simpler than IRV since there are not multiple rounds of the count.
But to advocate the method for use, one would want at least one clear reason why it's
*actually better* than IRV, and I don't really have that.

Kevin
votingmethods.net