[EM] Shortlisting

[Toby Pereira]

 I wouldn't rule out using some sort of proportional mechanism. It wouldn't offer "perfect" proportionality, but you could simply use a sequential "greedy" proportional method that starts off by electing to the shortlist the candidate that would win in a single-winner election.
If you use approval voting for the primary election, you could elect to the shortlist the most approved candidate. Then for the second candidate, elect the candidate with the highest approval opposition to the first. So the candidate approved on the most ballots that didn't approve the first candidate. (I got this concept from Chris Benham.) For the third candidate, would it then be logical to pick the candidate approved on the most ballots that didn't approve either of the first two? And so on. Maybe.
This could be generalised to score voting by finding at the candidate with the highest total difference to the top-scoring candidate on ballots where they score higher. Then for the third you could find the candidate with the highest total difference to the higher scoring of the top two on ballots where they outscore both. And so on.
For ranked ballots, perhaps the equivalent would be to find the candidate who has the highest number of ballots where they are ranked above the winner. Then for the third, the candidate with the highest number of ballots where they are ranked above both the top two. And so on.
Toby
    On Saturday, 31 January 2026 at 20:10:03 GMT, Kristofer Munsterhjelm via Election-Methods <election-methods at lists.electorama.com> wrote:  
 
 On 2026-01-30 18:24, Joseph Malkevitch via Election-Methods wrote:
> Dear Colleagues,
> 
> The linked paper offers some interesting ideas related to elections, 
> voting and social choice:
> 
> https://arxiv.org/pdf/2601.21277 <https://arxiv.org/pdf/2601.21277>

That shortlisting is different from PR is a good point. It reminds me of 
a similar thought I've been having about runoffs vs single-winner 
(single-round) methods.

A single-winner method should find the best winner according to the 
information provided by the ballots; or at least find a winner that can 
be defended by some criteria and doesn't violate others, since "best" 
may be difficult to discern due to incommensurability, etc.

But a runoff method is somewhat different. It should, I think, find a 
set of candidates who could plausbly be the winner if the people were to 
investigate that set more closely.

In a way, it is a shortlisting operation, but instead of using a 
distinct group (the shortlisters), it instead uses noisy ballot data to 
determine a set of candidates each of whom, were the noise of a 
particular type, could plausibly be the winner.

The candidates continuing to the general/second round probably should 
include the winner of a good single-winner ballot applied to the primary 
ballots, because one model of noise is "no noise" or "whatever noise the 
single-winner method can most easily handle". This immediately makes it 
clear that if the single-winner method is Condorcet, then the general 
election candidate set can't just be a multiwinner PR solution, since 
Droop proportionality is incompatible with Condorcet.

OTOH, just picking the n highest ranked candidates according to the 
single-winner method has its flaws, too: something like block ranked 
pairs is really clone-dependent (just clone the RP winner n times to 
fill the general).

So presumably, the solution of finding a good candidate set for a runoff 
is distinct from both "just use PR" and "just use bloc results". But 
what is it?

(Also, such a runoff concept would critically depend on the voters being 
given more information about the candidates who win the primary so they 
can refine their opinions before the general. Thus it isn't useful for 
"instant" runoff methods, because there's no such period of deliberation.)

-km
----
Election-Methods mailing list - see https://electorama.com/em for list info
  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20260131/029c36e9/attachment.htm>