James Green-Armytage
Recommendations
Single-winner elections (large electorates)
Multiple-winner elections (large electorates)
Single-winner elections (large electorates)
This
is when you have a large electorate (for example, several thousand or million
people), and you are voting between a series of mutually exclusive options. For
example, an election for president, governor, mayor, or another executive
office.
I
argue that Pairwise comparison is the best way to find a majority
winner, and that the
best definition of majority rule is minimal dominant set efficiency (a.k.a. Smith efficiency). That is, a method that
aims for majority rule should always choose a member of the minimal dominant set
(a.k.a. the Smith set). The
minimal dominant set is the smallest set of candidates such that every candidate
within the set has a pairwise victory over every candidate outside the set. If
the minimal dominant set only has one member, this candidate is called the
Condorcet winner.
My
primary recommendation for single-winner elections is to use the cardinal pairwise
method. This method uses ordinal (ranking)
information to determine the direction of pairwise defeats, and cardinal
(rating) information to determine their strength. I
think that it is preferable to Condorcet methods which only
use ranking information because
If cardinal pairwise is not a possibility, there are several other possible strategy-resistant pairwise methods. If none of these are possible either, then I stress that one should at least measure defeat strength in terms of winning votes rather than margins, because margins probably cannot provide strategic stability.
I argue that the best "base methods" (which can each be used in conjunction with various defeat strength definitions (winning votes, margins, cardinal pairwise, etc.)) are beatpath, ranked pairs, and the river method.
If the electorate is unable to accept pairwise count methods, but is amenable to instant runoff voting methods, then CWO-IRV might make for an interesting compromise, as it tends toward minimal dominant set efficiency while having a relatively moderate level of strategic vulnerability. IRV methods in general should allow equal ranking of candidates.
If ranked ballots are absolutely not a possibility, then approval voting is still clearly preferable to plurality, although it is difficult to evaluate in comparison with ranked ballot methods, and should probably be thought of as a transitional method rather than a goal in itself.
Multiple-winner elections (large
electorates)
Again
we are dealing with a large electorate, but now not all of the options being
voted on are mutually exclusive. The main example of this is when you are
electing more than one seat in a council or legislative body. In most cases I
think that it is desirable to use proportional representation to fill seats in
legislative bodies. In cases where you are not aiming for proportional representation,
you can usually make small modifications to single-winner voting methods in
order to make similar multiple-winner methods. However, to achieve proportional representation, you need fundamentally different methods.
I argue that the basic principle of effective proportional representation methods is the single transferable vote (STV). In particular, I recommend CPO-STV, which stands for "comparison of pairs of outcomes by single transferable vote." This is an effective hybridization of the STV and Condorcet principles, where complete outcomes are compared to one another using modified STV tallies. It is also possible to apply the principle of cardinal-weighted pairwise to CPO-STV.
In
this case, your electorate is smaller, hopefully small enough to fit in a single
room. For example, the voting that goes on within a legislature or council.
Because
the group is smaller, it should be relatively inexpensive to hold more than one
balloting. Hence I recommend the use of an iterative pairwise voting procedure.
In a nutshell, the procedure for single-winner voting is
1. to hold provisional ranked votes, and
tally them by a good Smith-efficient method.
2. The winner of these provisional
tallies, then, should be subject to a simple yes/no vote, such that if a
majority votes no, another balloting is taken, and so on until a
provisional winner is subsequently confirmed by a majority "yes"
vote.
This
procedure should have substantial benefits in terms of providing a strong line
of defense against strategic manipulation, and providing authentic majority
rule.
A
similar procedure can be used to achieve proportional representation (by
replacing the Smith-efficient tally with a CPO-STV
tally or an STV
tally
If the group is not contentious, and majority rule is not the goal, then methods like ratings summation and approval voting might be appropriate.
I
have written in detail about delegable proxy systems of direct
democracy. Voters would be able to vote directly on issues, but they
would also be able to designate proxies, such that if they do not have the time
to vote on a given issue, the weight of the vote will be carried by the proxy
of their choice.