A Warning on Asset

#1

The version of Asset I and others have touted around (use regular ballots, candidates get a week to trade votes, single-winner) has convincingly been argued to allow major parties to draconianly restrict ballot access, making it FPTP instead. Consider Abd’s time-unlimited version instead for PR, though I don’t know how the major parties would react to that.

#2

Restricting ballot access will prevent any election method from returning fair outcomes. There’s a reason why it’s a popular tactic among dictators running show elections.

Anyway, I think that in setting rules for an Asset system, permanent vs revocable transfer is another important distinction, not just time limits. By permanent vs revocable transfer, I mean: if candidate A gives her assets to candidate B, but B lacks the support to win, who decides where the assets that originally went to A go next? If transfers are permanent, then B decides when to give up and who gets the assets he received from A. If transfers are revocable, then A decides when to give up on B, and where the assets she gave to B go next. Temporary transfers limit the number of degrees of separation between the voter and the person deciding where their assets go, so transfer decisions should more closely resemble the original intent of the voter. However, when transfers are permanent, the negotiation process gradually simplifies as the number of people making decisions decreases, which may be more practical.

A key feature of Abd’s rules for Asset is that transfers are hyper-revocable: the person who originally received assets (the person whose name was actually on the ballot; Abd called them electors) that they later used to support a different candidate could revoke them even after the candidate they supported had already been seated in the legislature and was in the middle of serving their term. This would cost the candidate who lost the assets their seat, and they would need to make up the difference by gaining assets from another elector to get the seat back.

#4

The obvious advantage to hyper-revocable transfers is accountability: to stay in power, legislators have to keep their supporters happy (or rather, their supporters’ proxies).

However, the disadvantage is instability, especially since Abd’s rules require candidates to have Hare Quota of assets to be seated. Of course, a Hare Quota is the maximum amount of support a candidate could be required to have to win a seat that still makes it possible to fill every seat. Abd’s purpose behind this was that to completely fill a legislature, every vote would need to be allocated to a candidate. He described this as requiring unanimous consent to elect the entire legislature. (Note: as this would probably never actually happen, the likely result would be that the last couple seats would always be empty. Votes in the legislature requiring a majority or supermajority of support would base the number of votes required for the vote to succeed off of the theoretical maximum number of seats.)

In such a setup, a legislator could lose their seat if they lose any assets. So you could have cases where very small defections (e.g. one elector representing 4 voters) lead to headaches where even popular legislators have to replace a couple of defectors every month or so (it’s hard to estimate how frequent this would be in practice, though it’s clear that the larger someone’s coalition, the more often this would happen.) If it is too frequent, electors may spend a lot of time rebuilding their coalition after micro defections. Legislators may be able to make these instances rare by trying to get slightly more than one quota so that they can survive a couple of defections, but if any one legislator does this, one seat will always be empty. If the last seat is always empty, there will always be some unused assets, so there will be no reason for a seated candidate not to try to exceed the quota slightly.

In practice it will be as though the legislature has one fewer seat than intended, legislators need a Hagenbach-Bischoff quota to be elected, and supermajority thresholds, which are often arbitrary anyway, are slightly larger. (The threshold for a majority would also be slightly larger only if the legislature was intended to have an odd number of seats. Otherwise it would not change.)

#5

Anyway, requiring candidates to exceed the Hagenbach-Bischoff quota to be seated (this is the smallest possible quota which forbids more candidates meeting quota than there are seats) and having an odd number of seats ensures that a legislative majority requires a majority of assets, and that a majority of assets can command a legislative majority. For example, if there are 99 seats, the Hagenbach-Bischoff is 1% of the votes, and a legislative majority requires the support of 50 legislators. Each has more than 1% of assets, so in total they have more than 50% of the assets.

While this won’t work for supermajorities or unanimity, as noted earlier, most supermajority thresholds are arbitrary, so slight differences wouldn’t be terrible. Thresholds requiring a unanimity of assets would be impossible in practice, so the legislature would not use that as a threshold for any decision (e.g. a legislature elected via Abd’s Asset might allow the use of the consent calendar when there is no more than one unfilled seat.)

#6

If you are going to allow politicians to be unseated in the middle of the term, at least make it so you need a Droop quota to win a seat, so there is breathing room.

With the HB quota, you could still have W+1 candidates exactly meet the quota. Droop is better because it uses floor(V/(W+1)) + 1, which at most W can satisfy.

Honestly, if we are going to go any farther than Citizens vote, candidates trade, and winners go to their term then we would need a constitutional amendment, and at that point we may as well upgrade to Liquid Democracy. (The nice thing about LD is that it constitutionally forces the gov to ensure every citizen has education, housing, and internet access!)

2 Likes
#7

I proposed requiring candidates to exceed the HB quota, not meet it. It’s possible W+1 candidates exactly meet the HB quota for W seats, but then they won’t have exceeded it. The only support requirement where an asset majority is both necessary and sufficient to gain a legislative majority is to require candidates to exceed the HB quota to fill the seat (and this only works when the total number of seats is odd.)

#8

OK, although exceeding HB is logically equivalent to meeting Droop.

#9

It’s important to realize that Asset is a family of methods. The path to the implementation of Asset is through adoption in NGOs, and it can be kept simple. Once the electoral college exists, it is a body which fully represents the electorate, with no compromises. However, as a variable-vote assembly, which may suffer from participation bias, I suggest an advisory role for the College, rather than a sovereign one (which for NGOs means decisive for the organization, binding it). Recall becomes possible, but I prefer not to overspecify Asset. Simpler than recall would be an assembly allowing electors to vote on issues, where, if their voting power was assigned to a seat, the vote would, for a fully-enumerated count, be subtracted from the seat holding the vote. If an elector’s vote was not assigned (what I have called the dregs), then the vote simply counts directly.
My sense is that in a functioning Asset Assembly, that direct votes would make a difference would be rare. I would not allow electors to introduce motions or have the right to formally address the assembly. That’s what seats are for. Efficiency is a very important consideration, and part of the concept is recognizing the importance of a peer assembly. So these ideas are tweaks, not core.

I have suggested setting the quota at the maximum number of seats, and then the possibility of defining a sovereign majority in the Assembly as being a majority of that number. There is no need to manipulate the number so that there is an odd number of seats, because an even vote fails, and there is nothing wrong with that. Assembly rules may provide that an elected chair can vote in that event. My own opinion is that 100 seats may be too large, but this depends on context and the needs of the organization. Asset can work for electing a relatively small body that is highly representative.
Asset can work for single-member elections, but it is better, my opinion, to use a parliamentary system, to let the elected Assembly elect officers and make decisions like that.
(The actual quota size is not crucial as long as it is the same for all seats, rather than electing some seats with fewer votes. The idea is not to elect some representatives who are the most representative, but a set who are each the chosen representative of a quota. The possible rules defining an effective majority then ensure that all decisions will be approved of by a majority of members, directly or through chosen representatives.)

#10

There is no harm in using the Droop quota, as long as we recognize that there are unrepresented votes unless the elector put that last seat together. By conceiving of fully-voluntary representation, absolute representation (though it may be delegated), the need of the Droop quota vanishes. There is no reason to set a deadline for the last votes to arrange cooperation. It is also possible that more than one seat will remain vacant for a time, or even for the entire time until the next election. This is where the understanding of a majority in the Assembly would guarantee full democracy, that all decisions have the approval of a majority of those who voted in the Asset election. The difference will not generally be great.

I do not expect political parties to welcome Asset. Eventually, though, a party will use it for its own organization. Asset will, I expect, remove the power of money to influence elections. Lobbying will be more direct. It will not be necessary to raise funds to be elected. The entire present electoral system arose from archaic understandings and competition rather than cooperation.

#11

Why? Why force the election of candidates who appeal only to their narrow base?

#12

There is no “forced election.” There is representation by choice, rather than by winning a contest. Winners mean losers, and why should anyone lose representation by choice?

What most methods do is to decide, by some algorithm, what set of winners best represents the electorate. The voters don’t get to choose, an algorithm does, based on something very different from chosen representation. Unclear on the fundamental concepts of democracy. To be sure, we have never been clear. “Representatives” in current systems, as in the U.S., do not represent people, they represent districts.

#13

I could accept this argument for something like a local election, where the voters don’t really know the candidates. But when you throw in polling, and the ability of a campaign to shift, can’t it be argued that utilitarian methods allow voters to choose, by adjusting their scores, which candidates rise or fall? In the end, social decisions are either made by majority or utility, and utility seems the better measure… I understand that majority is about personal choice, whereas utility is about aggregation. But don’t we live in a world where people can’t possibly know every problem others have, where sometimes, the best we can do is ask you how you feel on each choice, and aggregate it in such a way that it is most likely to be of benefit to the minority?

#14

The reason for the odd number of seats was to ensure both that a majority of seats always represented a majority of assets, and that a majority of assets was sufficient to secure a majority of seats. With an even number of seats, the former is true but not the latter.
When there are 2N seats, the smallest proportion of assets is needed to obtain a majority is (N+1)/(2N+1). When there are 2N+1 seats, you need to exceed a proportion of (N+1)/(2N+2)=0.5 of the assets; that is, earn a majority.
Granted, the difference is probably small enough to be of little practical concern, I just think it’s a nice property.

#15

If the base will not support anyone else, then this is a consequence of proportional representation in general. If 10% of the voters vote for the Nazis and only the Nazis, then 10% of the seats go to Nazis, or it isn’t PR.

Anyway, the Abd post you quoted expresses a Monroe interpretation of proportionality.
See: Thiele vs. Phragmen/Monroe: Two very different interpretations of proportionality
The criticism of Monroe is that in cases like:
10: Max support for A, B, C
10: Max support for A, B, D
(2 winners), AB and CD are considered equally good selections of winners, but AB Pareto dominates CD. Of course, if this were an Asset election, while a coalition for C (for example) could form, the electors who don’t support C would have a chance to talk them out of supporting C and instead supporting D.

If the C and D supporters each preferred C and D slightly more than A and B, but hated the other, they could force the election of their candidate (they can do so in any PR system), or they could negotiate a compromise not to elect C/D if the other guys don’t elect D/C. If you try this with a PR method, you wind up with a prisoner’s dilemma. So in a sense, Asset is only as Monroe as the electors want it to be.

Of course, if only one of C and D existed, one side would have no leverage, but if that’s a problem for you, then as noted before, nothing really can fix it.

#16

Thanks for the education!
The example is confusing for me, but that’s mainly because there are a lot of implicit concepts that I’m not familiar with. I feel AB is the best winner-pair, seeing as they have both consensus and good first-preference proportionality, but I understand that others disagree.
What I’m most worried about with Asset is hijackability. Seeing as it’s a system that empowers politicians, it is the most likely to be affected by ballot access/write-in “issues” (looks like we can’t use write-ins -> looks like we can’t have all these “vote-splitting” candidates on the ballot -> limited choice, in which case I think voter input is preferable. If you’re going to have an oligarchy of on-ballot candidates, might as well make them compete, rather than letting them “cooperate”, especially considering that when some candidates are politically ambitious, they may be incentivized to threaten politically suicidal moves with their votes if it scares other electors into picking them. This can either lead to politicians who don’t have to respond to public pressure (because they know they’re going to lose the Asset election, so why even bother honestly trading? Winning in the next election isn’t guaranteed, and ascending up the political ladder requires taking risks) i.e. our current system, or politicians who always back one of the ambitious, leading frontrunners out of fear of electing the other frontrunners i.e. our current system)
I think simple always wins when it comes to reform, so long as there aren’t fatal flaws (and as we’re seeing with IRV, even then.) This leads me to conclude that in the long run, Asset has some chance as a PR system, which I’m not happy about if it beats less hijackable, more voter-friendly systems (not to mention, more utilitarian multi-winner systems.) If it turns out to not be hijackable, it could yield better outcomes than FPTP, and give energy to voting reform at-large, but if not…
Edit: I don’t want to be cruel to Abd’s version. I’m still pondering over how it would be implemented/whether it’s hijackable before/after its implementation, and how. I don’t think we could know until it was tried out for public elections, and I don’t see the risk/reward ratio working out… in the meantime, it seems a safer bet to go with a deterministic voting system, even if it yields strictly worse performance, if we suspect it’s less hijackable/are more certain of its quality.

#17

The point of the example is that AB is unambiguously the best winner pair, since it is the only one with unanimous support. Since Monroe only cares how a single quota of voters scores a candidate, it will see CD as an equally good winner set, because the 20 voters can be assigned to candidates C and D such that each candidate gets a quota of voters and all voters score their assigned candidate maximally.

#18

Wow, I really misunderstood the example. It seems to me that there should be no potential for consensus to be defeated, in which case I’m leery even of Asset. However, are there any strategic benefits to Monroe that you can easily identify? I know there are concerns with Hylland Free-Riding and consensus candidates not getting their honest support, so I’m wondering if there is anything useful in Monroe for solving that, or even other problems in PR.

#19

How about that (multiply everything by 1000):
A gets 6,667 votes.
B gets 6,666 votes.
C gets 3,333 votes.
D gets 3,334 votes.

A has a droop quota and is 100% guaranteed to win. If C and D are really bitter rivals, there would be no reason one would enthrone the other, and B would either get more votes or else nothing happens and B wins by default.

The worst that can happen is AC or AD winning, but voters would not like that. My opinion of Asset is that many potential problems can be solved by voters adapting.

#20

Asset requires more candidates than just the frontrunners - it requires electors who can represent the voters’ preference orders between the frontrunners. For that reason, it is less reliable as to whether a voter can change things among the frontrunners; first, the elector must take some of the frontrunner’s support, and that support must be interpreted in a way that lets the frontrunner adjust to the voters’ wills. This is where I find cardinal systems make it easier for society to know what it wants; marginally more popular policies can be represented as such with score polling.

1 Like
#21

Printed ballots with candidate names creates a system hijackable due to ballot access rules, and probably requires “campaigns” and all the rest. I see no way for blank ballot (which is how elections used to be run, were you aware of that?) to be “hijacked,” if the system makes every vote count. The problem with high candidate count is vote-splitting, and that is exactly what Asset was designed to address.
“Empowers politicians”? what kind of politician? Asset creates small-scale power, power to be represented, and it does it without what are ordinarily called “politicians.” Then the electors will deal with candidates, who may be professionals, i.e, “politicians.” They will, in general, be able to actually meet the candidates, and will be far better placed to make intelligent choices than ordinary voters ever will be, people who do not generally pay close attention to governmental issues.

Deterministic voting systems lead to Arrow’s Theorem and major issues with strategy, it is probably intrinsic. I’m not going to use a deterministic system to make decisions about my own life, why would I want this for public use? It dumbs down the system. The best voting systems in current use are runoff systems, with write-ins available, and that is inherently non-deterministic. And this is all well-known in deliberative process, deterministic systems are highly discouraged by Roberts’ Rules.

Public systems are different because of a belief that a decision must be made. Yet a good two-round system can probably beat every single-round system ever proposed, for single-winner elections. Two-round score has lower Bayesian regret in simulations than single-round score. And those simulations did not consider the reasons Robert’s Rules gives for repeating the election instead of using a ranked system: the value of the new look at the candidates, now informed about what choices are realistic, and with new candidates being possible.

Getting there from here is another question, and this is why I suggest Asset for NGOs.

“Quality” is one very fuzzy concept. If I can choose my representative, what does “quality” mean? In an Asset election, a seat represents a quota of voters who have agreed on that seat as representative. They may use polling systems to advise them, perhaps, but far better, my opinion, would be simple delegable proxy, with final assignment of votes still subject to elector approval.

Unanimous agreement of a quota on a seat, and consent of a majority of the electorate, as represented, for all decisions. This is so mind-bogglingly simple that it astonishes me that people argue about it and make up endless reasons why it won’t work; what is being said, really, is that basic democracy won’t work, because Asset hews to very basic democratic principles, obvious in voluntary organizations – but still rarely implemented fully – but public elections have always been different, because of the history. The people were not represented, districts were!