Monroe Selection
-
Clay is advocating for a Monroe selection on RRV
https://medium.com/election-science/sequential-reweighted-monroe-score-voting-a410d1d2727a
The wolf committee explored this option a little and it is in the simulation code
https://github.com/endolith/Keith_Edmonds_vote_sim/I do not remember what the results of this system are so it is likely we did not spend much time on it. In any case it seems like a bit of a weird mismatch since a Monroe selection sort of implies quota based exhaustion not reweighting. In principle there is nothing "wrong" with this combination of systems. My only caution is that there is a connection between the quota and the reweight formula. Look at the table here
https://electowiki.org/wiki/Single_distributed_vote#Proportionality_Threshold
-
@keith-edmonds
Do you know if the quota updates between rounds? The RRV reweighting scheme doesn't decrease the total number of votes by a quota each time, so it could change.My initial reaction was that this system would discourage the use of the middle of the range, because if you give some points to a candidate but are not in the quota, then you pay for that candidate but don't contribute to their election. However, there seem to be strategies that involve deliberately giving candidates non-maximum scores to delay their election. I am not sure how viable they are in practice (likely with a 5 point scale), although with a very granular scale, I suspect they would be.
Consider the following 5-winner election.
20 A1[5] A3[4]
20 A2[5] A3[4]
10 A3[5]
50 B1[5] B2[5] B3[5]
There are equal numbers of A and B supporters. The B voters vote "normally", while the A voters try some sort of vote management scheme.
The first candidates elected are B1, B2, A1, and A2.
The weighted ballots after that are
50/3 B1[5] B2[5] B3[5]
20 A1[5] A3[4]
20 A2[5] A3[4]
10 A3[5]
Regardless of whether the quota is updated, A3 wins.
(B3 will get 250/3 points. A3 would get 90 with the original quota of 20, or 130 with an updated quota.)There might be ways that this could go wrong for A. The main one I can think of is that if A3 gets too many votes and ends up getting elected before A1 or A2, then the A votes become extremely weak instead of extremely strong. But party A could always put a divisive candidate unlikely to attract outside support in the third position to avoid this. Although party B could tell their voters to vote for A3, which would counter this strategy if they can convince enough of their supporters to do so.
Edit: come to think of it, there is no good reason for party A to split up A1 and A2 supporters. They're better off aiming for:
40 A1[5] A2[5] A3[4]
10 A3[5]
A3 ends up with more support if the strategy goes as planned, and party B takes more of a risk by trying to get their voters to vote for A3. -
Do you know if the quota updates between rounds?
I do not believe in this method that was the intent but perhaps there is something more slick that could be done incorporating the ideas of Sequentially Shrinking Quota.
The RRV reweighting scheme doesn't decrease the total number of votes by a quota each time, so it could change.
The RRV system has more issues than that as is pointed out in the Single distributed vote page. So the potential system here is a way to combine the changing quota idea in Sequentially Shrinking Quota, the balanced reweighting from Single distributed vote and the selection from Sequential Monroe. It may not be possible but if I find some time it might be straight forward. The issue with Sequentially Shrinking Quota was always the calculation of quota size but reweighting gives some insight there.
My initial reaction was that this system would discourage the use of the middle of the range, because if you give some points to a candidate but are not in the quota, then you pay for that candidate but don't contribute to their election.
As I pointed out in my first post a mismatch between selection and exhaustion/allocation is likely to cause strategic vulnerabilities. The vote management things you highlight could sink the system. However, there are many strategies people can use and it remains unclear (at least to me) how their effectiveness balances out. A weak exploit which is easy and obvious may be better or worse than a strong but rare exploit. These are things we cannot really simulate. Things like vote management require a group effort and that relies on the organizational abilities and cohesiveness of factions. I suspect the worst exploits are those which are obvious to a single voter and unlikely to back fire.
Independent of if it is viable in the real world it would be a great extension of the field to have a system like the one I propose. It is good to compare and contrast.
