Single- vs. Multi-winner: Jameson's rough notes on theory/philosophy


#1

I’ve had a few ideas about voting theory rolling around in my head for several months now. Rather than continue to wait until I can give them the treatment they deserve, I think I’ll put them out here in rough draft form. I’ll be quick and please trust that I have further sub-arguments that I’m skipping here.

  • Philosophically, voting theory is about collective decision-making.

  • I don’t buy utilitarianism 100%, but it’s clearly the best foundation for thinking about this.

    • Any deviation from utilitarianism opens the door to pathologies and/or impossibilities.
    • In most cases, the ways utilitarianism is arguably wrong end up going in effectively random directions in practice and thus canceling each other out. Only very rarely do they agree.
  • To me, a good overall model is that a society faces a series of many single-winner choices between courses of action. In many of these choices the options can be arrayed along a one-dimensional axis (continuous or discrete), such as “how much money should we dedicate to this end?”; and usually most voters’ preferences are single-peaked. In other choices, there is no natural spectrum/ordering, so Condorcet cycles are possible. Even in the latter cases, natural Condorcet cycles are relatively rare (probably ≤1/6 of the time).

  • In that situation, there are 3 simple possibilities (plus hybrids/compounds):

    1. Direct democracy: have separate votes on each choice, using some single-winner method.
    2. Majoritarian: have one vote, using a single-winner method, to elect a leader or leaders. Then they decide on each choice, for a period of time, until their term expires. Possible recall mechanisms, etc.
    3. Proportional: have one vote, using a proportional multi-winner method, to elect a legislature. They then vote on each choice using a single-winner method.
  • The downside of option 1 is that it’s time-consuming for the voters. Downside of 2 is that it’s hard to represent a multidimensional distribution of voters using a single point. Pretty clear to me that 3 is the best: allows the representatives to be deliberative enough to find non-obvious good solutions and/or reach consensus when possible, without wasting the voter’s time.

  • Also clear that ideally you’re looking for a system where the representatives are, um, representative of the ideological distribution (as a large-enough random sample would be), while being better-than-average in terms of non-ideological qualifications (intelligence, negotiating skills, knowledge, integrity, work ethic, etc.)

  • A random sample (sortition) would be representative in ideology, but also merely average in terms of qualifications.

  • The ideal for a voting method is to remain representative in ideology, while also selecting for the most-qualified representatives. It should also be relatively easy to use and trustable for the voters.

  • Asset voting would do pretty well at this, though having unequal voting weights within the legislature might impede the egalitarian give-and-take which helps deliberation. But let’s say for now that equal voting weight is a (constitutional?) requirement.

  • Pretty much any proportional method has free-riding incentives. Essentially, these incentives allow strategic voters to get extra voting power if they vote on an axis other than the principle component of variation of the population. For instance, in a two-party D/R scenario, you can’t use free riding to elect more Ds, but you can use it to get unfair extra power over which Ds are elected.

  • In this situation, I believe STV or reweighted approval actually give voters too much freedom to fine-tune their ballot. Free-riding potential is too high. Meanwhile, reweighted score is just reweighted approval for strategic voters, and less voting power for unstrategic voters.

  • I suspect that a partially-delegated, at-large system like PLACE is the sweet spot. That is, the representativeness of the ideology distribution of the winners is highest when the number of distinct ballots each voter could cast is about equal to the total number of candidates in the entire election. (This also goes for asset.) More than that is just an invitation to strategy; less than that is tying the voters’ hands and giving their power to elites or to chance.

  • Meanwhile, the system should have ways to simplify the task of voting for relatively low-engagement voters (such as the district-specific ballots, and weak incentives towards big-tent parties, in PLACE).

  • There should also be incentives for alliance formation/platform negotiation to be split between the period before the election (so that the voters can have a say on clearly-defined platforms) and after the election (so that platforms can react to the election outcome). This means that there the effective number of parties should not be too low nor too high; I think the sweet spot is around 3 (room for thesis, antithesis, and nascent syntheses). Thus, the 25% local-support threshold in PLACE.

I could go on for a page or more about any one of the bullet points above, but the purpose of this message is just to get them out where they can be discussed. So I’ll stop here.


#2

The main difference between PLACE and Asset is that PLACE is at large and optionally delegable, right? How do you think PLACE and Asset would differ in real world implementation, and which would be easier to use and/or pass?


#3

PLACE delegation is partial, but not optional. That is, candidates predeclare a 4-level rated ballot (party ally, party member, coalition ally, non-ally) and then within each of those categories transfers go in order of raw votes. So essentially, other voters from the same or allied parties help choose where your vote goes; it’s not purely up to the candidate.

The other major difference from Asset is that there are transfers of overvotes as well as of wasted “undervotes”, so that each winner ends up with equal voting power. I believe that this makes PLACE easier to pass — less of a change from current system. It’s also easier to use for voters because ballots highlight local candidates.


#4

Are you aware that there is a simple version of Asset where the top (number of seats to be filled) vote-getters get elected with equal voting power? How unrepresentative or bad do you think a lack of automatic vote transfers would make the winners under this version of Asset compared to PLACE?

Also, I’ve heard the argument that party based PR is unconstitutional because voters outside of a party help determine who its top nominees are or something like that, which could be considered a violation of right of association. I don’t see any obvious connections between that and this, but thought it might be relevant to anyone considering a more party oriented version of PLACE.


#5

That argument can’t be valid; it would mean that open primaries are unconstitutional.

PLACE does include a partisan aspect, in that the “party member” predeclared transfer rating is not chosen by the candidate but is simply the set of other members of the same party.


#6

That argument seems to apply more to blanket primaries, where a voter can vote for different parties’ nominees for different offices, whereas open primaries limit you to picking one party’s nominees for all offices (if I understand correctly). PLACE lets you pick multiple party’s “nominees” to some degree, but because the original choice is made by the party candidate themselves, I think it’s constitutional.