CIFeR voting. An alternative to asset voting


I used “Asset” to refer to a class of voting systems that have
also been called "candidate proxy. Asset has been discussed at
great length on various election methods mailing lists. There
could be many variations of how the basic principle is
implemented. Warren gave the method its name, if I’m correct:

  What we now call Asset Voting was first described, to my

knowledge, by Charles Dodgson (Lewis Carroll) in the 1880s. It was
an STV tweak, so the method was STV, with the tweak that exhausted
ballots could then be used by the most-favored candidate, making
it practical to vote for one, reducing the risk of a wasted vote,
and Carroll was aware that most voters would not have the
knowledge to rank candidates in their own best interest (or in the
interests of social benefit), so he hit on making the votes
transferable by the candidates. The method was then proposed in
the 1990s by voting systems theoriests, and was called “Candidate

  My contribution was to realize that what Asset could actually do

is create an Electoral College that fully represents all the
voters. The primary, initiating purpose of that College would be
to convert votes – which are now votes for electors, not for
“candidates,” into quotas for seats. No votes are wasted, and
voter strategy becomes extremely simple. Vote for the candidate
you most trust to elect a representative for you. (Or to serve, if
the elector receives or is assigned enough votes.) I would expect
that in a large public election under Asset, few, if any, seats
would be directly elected.

  Before I ever became involved with election methods, my concerns

were broader, thinking of how large groups of people could
efficiently communicate, cooperate and coordinate.

  Warren was thinking about an election method, adding the

requirement that the election complete by some deadline (or
deadlock by some definition). I realized that – and this was
shown in the actual election we ran – that this was unnecessary,
that the electors could create a functioning representative body,
still representing at least a majority, even if some seats were
not yet assigned. Haste makes waste, in this case, wasted votes.

Warren’s rules:

Asset Voting is a simple multi winner
voting system. Suppose we’re talking about an election with W

  1.         Voters provide scores to every candidate. The scores are

non-negative numbers that add up to 100. In Simmons’ simpler
version, your vote is just “name one candidate” (who gets
100 from you, the others get zero).

  1.         A candidate's score-total then is the amount of an "asset"

he then “owns.” Candidates then negotiate, transferring some
or all of their assets to each other.

  1.         After negotiations end, the W most asset-rich candidates

are the W winners.

  1.         (Optional) we could have "weighted congressmen" where each

has a number of votes in congress proportional to his assets
when elected. This would cause even-finer "proportionality.
Warren being Warren, he used what we now call score voting. Warren
did not have clear ideas about the function of representation, he
was attempting to determine the “best W candidates” from ballot
information. So he tweaked Reweightable Range Voting to become
Asset. The core Asset invention is assignability of votes by those
holding voting power., having been awarded that by voters
directly. His system allows voters to cast 100 votes. Notice that
he mentions “Simmons’ simpler version.” It is possible to use
Approval Voting or Score voting with Asset, but to avoid giving a
voter excess voting power, one would reweight the votes to sum to
one full vote. This introduces ballot and canvassing complexity,
and Vote-for-One Asset is likely quite enough to establish full,
fair, proportional representation, with very high proportionality.
Years ago, w discussed Asset at great length, and Clay Shentrup
created an Asset election for the first steering committee for our
working group. It was vote-for-one, and Clay had a vague idea of a
deadline. The actual election trumped those ideas. Basically,
electors are given formative power and can, in fact, alter the
rules ad hoc, with enough support. As it happened, I had received
a majority of votes, as I recall. (Somewhere there is a record of
that election. If I’m correct, Clay later deleted that list, or
maybe my memory is defective.) Clay was runner-up. Warren was

  So I was considered elected, and gave enough excess votes to

Warren to elect him by the Droop quota, again as I recall. Clay
had not established a clear quota to be used. This left me some
excess votes, still. (Droop quota would be 1/4, Hare quota 1/3.
Thus with a majority, I had enough votes by Droop, maybe not
enough by Hare. But I could declare Warren elected and did – and
nobody objected. I could have declared Clay elected. Either way, I
was able to create a majority of seats in the committee, which
could then rule on any voting system issues. Asset can lift itself
by its own bootstraps.

  Clay saw the situation and gave his votes to another, who was

then elected by Droop, not by Hare. But I then asked the remaining
candidate, with the fewest votes, if he would consent to the
election, and he was happy to do that, so the result was an
election with unanimous approval (as delegated).

  Assemblies generally have the right to determine their own rules,

it’s basic democracy. (Hence the “nuclear option” in the U.S.
Senate, which is based on standard democratic procedure and the
power of an absolute majority to Do Whatever The Hell It Wants, if
push comes to shove, and there are very sound reasons for this,
and for U.S. voters to be very concerned about loss of
collegiality and concern for common welfare and the value of
consensus. The U.S. has not returned to civil war for a long time,
and if we don’t address this problem, it may. With existing
election methods, populism is very, very dangerous, as was long
understood by American leaders. Instead of appealing to the
collective intelligence of the people, it appeals to knee-jerk,
reactive emotion, which, for humans, can lead to major social
disasters, examples abound.

  (But populist views should be fairly represented, and Asset would

do that.)

  So as long as a majority of the "full assembly" has been elected,

it can, representing a majority of the people, make any decision
it chooses. As is in the news recently, the House can refuse to
accept an election result, by simple majority vote. (I’m not sure
of the actual procedure, it might be that there is a routine vote
to accept election results, and accept new members, and so that
vote could fail. The basic democratic principle is that no binding
decision is made without majority support, i.e. over 50%, ties
failing to create a decision.)

  There is a possible deficiency in that, but only a deliberative

deficiency. I.e., there may have been voices not fully represented
in discussions. So I expect that an Asset-elected assembly would
not act to frustrate the democratic intentions of the voting
method, it would simply address how to supply what might be
missing. I’ve mentioned one tweak: using the Droop quota and then
allowing a single seat, with a half vote, to be elected, which can
only be a deciding vote in a otherwise-tie. That’s procedurally
very simple! And obviously fair. If standard procedure is followed
and the chair (presumed to be a member) votes to break a tie, this
extra seat could then reverse that or affirm it. It could reduce
the unrepresented voters to a half quota at most.

  (Other systems, more complex, could guarantee full representation

without damage to assembly efficiency.)

** “Asset” is thus a class of systems, not just a single set of
specific rules, but the common feature is delegated voting


I have a question about Asset used outside of electoral contexts. It concerns me that after all the great proportionality and utilitarianism of Asset in the elections, the legislature still pushes things through with an extremist majority. I had the idea that legislators should actually vote on each other, Asset style, and the winner would pick the legislation to be used by the full body, with perhaps a majority vote on that, perhaps not. The thinking here is that the winning legislator will produce more utilitarian legislation, rather than bowing to the largest group. Does this work?


There is a difference between, say, 80% extremists and 20% centrists, versus having 51% of people extremists for the same extreme viewpoint. With the former, Different disagreeing extremists will have to find some common ground.


So your point is that there would be no improvement by having a centrist between the factions pass the legislation, versus having the extremists compromise by themselves?


I think the basic idea of letting candidates remove the approvals for themselves is interesting (though probably impractical). But the final quality metric is problematic, because it doesn’t distinguish between approvals which are “spread thin” and those which aren’t. For instance, imagine a situation where there are 2 parties, orange and purple, and two genders, male and female. Voters can be “partisan” or “gender-centric”, while all candidates are “partisan” first and foremost. This system breaks down pretty badly.

I suspect a squared load quality metric a la Phragmen would fix this. Since this method is already pretty impractical, I don’t think that the downside of Phragmen — too math-heavy — applies.


Any quality metric would fix it. I’d prefer harmonic voting since it preserves most of the elegant properties of this method.


Would this still involve delegation? For this to work, the strategy has to be easy for campaigns to manage.


Running a CIFeR voting election requires tallying up every possible combination in which a voter can rate all the candidates. That means, in order to run a CIFeR election, you need to tally each of the 2^C possible ballot combinations where C is the number of candidates in the race.

That number is C! (pronounced “C Factorial”) = C * (C-1) * (C-2) * … * 3 * 2 * 1 . There are C ways to choose the 1st candidate, C - 1 to choose the 2nd, down to what’s left at the end.


This uses approval ballots, not ranked ones.


Yep. It’s 2^C, which is better but still pretty bad.