[EM] A procedure for handling large numbers of candidates using scorevoting with primaries and runoffs.

Kristofer Munsterhjelm km_elmet at lavabit.com
Tue Apr 10 11:32:30 PDT 2012

On 04/10/2012 03:57 AM, ⸘Ŭalabio‽ wrote:
> ¡Hello!
> ¿How fare you?
> We require candidates to get the squareroot of the number of voters
> to get on the ballot with writeins allowed.  Let us look at an
> hypothetical single-winner election:
> Let us suppose that 420 people make the ballot.  ¿How can the voters
> vet this many candidates?  Well now, we have a score of parties and a
> score of independents.  That is a score of candidates per party and a
> score of independents.

Or this could also be an option:

Have the first round be Approval. Tell people to be generous, since it's 
only a primary. The idea is for the voters to do a very quick judgement 
of the candidates, and to err towards admitting a candidate rather than 
rejecting him.

Then, once the first round is done, remove everybody who is approved by 
less than a certain fraction of the voters. Here, Approval is used as a 
kind of probabilistic Range.

Even with a threshold on the order of 1%, that will deal with most of 
the candidates. If it doesn't, remove least-approved candidates until 
you have narrowed the field down enough (or have more Approval rounds if 
you can afford them).

Once you've narrowed down the field to, say, 20 candidates, use your 
favorite voting method (Range, MJ, Condorcet) to determine the actual 
winner. If you feel like it, have a top-two at the end, as it may deter 
certain forms of strategy and, if Abd's right, impart some "utilitarian 
weighting" with people not bothering to show up if they don't care for 
either candidate.

I think the "primary" rounds should be party- and independent- agnostic, 
so that if the voters disregard the parties altogether and just vote for 
independents, they should get multiple independents.


I see a problem there, however... it's not as great with 20 candidates, 
but say you wanted only four in the second round and had a plurality for 
your party. Then you could use that to fill all four positions and the 
subsequent rounds wouldn't matter. Note that your party restrictions 
wouldn't help: the strategic party could just make decoy lists (four in 
this case) to fill all the party slots. If you expanded the number of 
places for the subsequent rounds when additional parties joined, then 
independents would be foolish to stay independent; instead they should 
organize mini-parties of their own, even if that means clogging up the list.

So it'd seem that you'd ideally want some kind of proportional 
representation method to select from the first list. To keep the 
simplicity and speed of the first round, that method should be 
Approval-based, something like PAV, and with 20 candidates, you'd get 
good diversity in the second round.

Perhaps you could have a dynamic seat size. If the PR method was ranked, 
you could have it like this: if a set of candidates of cardinality k 
gets more than p Droop quotas, p >= q, then consider that set to hold p 
seats. If p >= q, that means some seats will be unused. So if a 
candidate got all but one Droop quota, you'd go right to the second 
round with him as well as whoever got the final one. For Approval, I'm 
not entirely sure how you'd do it.

More information about the Election-Methods mailing list