The Alaskan Top Four Model & iEBs

Toby Pereira

@gregw said in The Alaskan Top Four Model & iEBs:

@toby-pereira said in The Alaskan Top Four Model & iEBs:

I'm not sure in practice how the results of the Excess Method would differ from SPAV, but I think its complexity counts against it. And considering this is just to narrow down the field, rather than get the best proportional result, I don't think it's probably necessary.

Thank you for the advice, your point is well taken.

For a Round 1 vote, I am leaning to a Webster version of SPAV and STAR for Round 2, the general election.

I saw that SPAV and STAR are listed as failing Participation on the Comparison of Electoral Systems wikipedia page.
Could those failures be a problem in these two applications?
Am I paying too much attention to the criterion lists?

Participation is a very difficult criterion to pass. SPAV fails it, but I think non-sequential PAV passes it. However, it might be computationally unfeasible, and probably not the best method for this particular application anyway. I think sequential PR methods pretty much all fail participation (apart from perhaps non-deterministic ones), so you might just have to accept it.

For the final round, score and approval both pass participation, so you could select one of them. Approval is generally considered a good method by most people, whether or not is it their absolute favourite.

GregW

@toby-pereira
Thank you for your help.

Lime

Notes:

Exact PAV or Harmonic Voting are infeasible, but very good approximation is trivial—integer linear programming is hard in theory but easy in practice. (There are better approximation algorithms than SPAV/SPSV, so I'd argue these are obsolete except as educational tools.)
I'm not actually sure proportional voting systems should satisfy affine invariance. Especially for the case of constant-sum invariance, it could be a negative property. I'm writing up the argument in another thread.

Lime

Although, I will say SPAV with Jefferson is probably a better system for a primary than something like PAV, because a primary should always select the candidates with the most votes; the goal is to maximize the probability that the best candidate will make it to the runoff, rather than optimizing the average overall representativeness.

Also, I think SPAV is fairly simple, but might be too complicated for a simple primary compared to cumulative voting. (Also also, any Condorcet winner should probably be guaranteed a spot.)

GregW

@lime said in The Alaskan Top Four Model & iEBs:

Although, I will say SPAV with Jefferson is probably a better system for a primary than something like PAV, because a primary should always select the candidates with the most votes; the goal is to maximize the probability that the best candidate will make it to the runoff, rather than optimizing the average overall representativeness.
Also, I think SPAV is fairly simple, but might be too complicated for a simple primary compared to cumulative voting. (Also also, any Condorcet winner should probably be guaranteed a spot.)

Yes, the idea is to nominate the best candidate, average overall representativeness could result in a boring general election.

I think voters would like cumulative voting. There would be a slight possibility of a strategic block voting campaign. If a party had that enough support to pull that off they would probably win the general election in any event.