[EM] What is the ideal election method for sincere voters?

[Michael Poole]

Abd ul-Rahman Lomax writes:

> Okay, that rarity of rarities, an original post by me.
>
> A great deal of criticism of voting methods is based on how the 
> methods behave with "strategic voters," those who vote insincerely to 
> gain some presumably favorable outcome.
>
> However, if we assume sincere voters, what is the ideal election 
> method, or the best among the options we know?
>
> This list is about election methods, but it is often assumed that 
> these elections are taking place in environments where some people 
> will be highly motivated to be deceptive, if they think this will 
> benefit them, and some methods, more than others, "encourage" this.
>
> But not all election environments are like that. Further, if we have 
> an understanding of what is ideal with sincere voters, then we have 
> some means of approaching the problem with insincere voters, because, 
> presumably, we may want the same outcome and, in simulations, at 
> least, we can see if the outcomes match and thus have an 
> understanding of what methods are vulnerable to strategic voting.
>
> A few of us have stated that Range Voting is, of course, the ideal 
> method, but it is impractical because it is allegedly vulnerable to 
> strategic voting. I've challenged that conclusion, which is typically 
> based on considering bullet or Approval-style voting to be 
> "strategic," where, in fact, it is merely the expression of strong 
> preference, and, I argue, one will only vote Approval style in Range 
> if there is a strong preference, in which case it's not insincere! 
> (At least this is now sane people will behave, and most people, in my 
> opinion, are sane in this way.)
>
> However, is Range ideal with sincere voters? If not, why not?

I alluded to this briefly before, without specifically saying it:

In practically all non-trivial elections, given enough information
about voter preferences, reasonable people can disagree about the
ideal result.  The usual tradition instead is to define criteria that
capture what one thinks are important rules about the outcome.

We can argue all day about which criteria are most important, but in
my opinion, agreeing on that is less important on this list than being
able to agree on what "important" criteria mean, which methods satisfy
each important criterion (or which criteria one method satisfies), and
strategy-related issues such as how easy it is to maximize a vote's
impact and how likely that is to change the method's result relative
to sincere voting.

(To determine which criteria are "important", I think the most
productive approach -- and what this list has generally done -- is for
each person to identify that for himself, and then cite or share the
definition.  If others think a criterion is useful, they tend to test
their own favorite methods against it.)

> And, please, explain to me why a method that will work well for 
> selecting pizzas, with sincere votes, will not work well selecting 
> political officers, similarly with sincere votes. If you think that.

Setting aside the issue of how to define "will work well", there are
several factors that make it reasonable to make more approximations
for models of large-scale elections than for models of pizza
selection.  The root cause of the factors I can think of is rational
ignorance -- mostly on the part of voters, but also in the election
system.

One factor is understanding the candidates' past positions: It is a
large amount of work to accurately understand how (and why, when "why"
is important) the candidate behaved in the past.

Another is estimating future positions and situations: "Past results
are no guarantee of future performance."

Partly because voter ignorance is rational, voters in public elections
often substitute emotion or labels (often those provided by third
parties) for critical examination.

On the system side, functions like social utility are usually neither
one-dimensional nor linear in their inputs.  Pizza is a fairly trivial
example in terms of social utility; political office is not.  Thus, it
is harder to create a ballot that captures a significant fraction of
the voter's objectives for a political office.

> If we cannot agree on the best method with sincere votes, we are 
> highly unlikely to agree on the best method in the presence of 
> strategic voting, though I suppose it is possible....

I submit that having consensus on "the best method" is a luxury: it is
nice, but not necessary.  Elections methods are at least as much
social science as they are analytics, and there is a huge scope for
interesting and useful research on the topic.  Good research is very
often spurred by places where the community does not know the answer.

Michael