[EM] RV comments

[Abd ul-Rahman Lomax]

At 08:17 AM 7/21/2007, Juho wrote:
>I also think that Range is a good method in non-contentions polls and
>elections. But in the statement above a competitive election was of
>course the assumption.

Why?

*Assumptions should be stated.* Election methods are used in all 
kinds of situations. If there is an ideal election method for use 
where it would be expected that people will vote sincerely, shouldn't 
we know it?

And *then* we could look at what happens when some people don't vote 
sincerely? It's clear that there is harm done, under some 
circumstances, but how much harm and who is harmed the most? Is it 
the sincere voters? Or is it those who did not provide accurate 
information to the voting method, so it can't possibly optimize their 
satisfaction?

I don't find the answers to these questions obvious. Apparently some do.

>And there are some minor things in the description, e.g. Range is
>typically defined as having only a fixed number of possible ratings,
>not any value between 0 and 1.

A completely general definition of Range would be that votes are 
allowed anywhere in a Range of real numbers, and the votes are summed 
or averaged (it makes a difference, depending on how blanks are 
handled, and there are proposals one way and proposals another).

>But isn't t so that the description above is a quite valid
>description as a main rule for competitive elections (where we want
>all voters to cast votes of similar strength).

No. Who is "we"? I don't want all voters to cast votes of equal 
strength in all pairwise elections, because it patently damages the 
function of any election method to do so. What Approval style voting 
amounts to is only allowing the voter to vote in certain pairwise 
elections, with full strength, while totally abstaining from others. 
Range simply allows the voter an expanded range of intermediate possibilities.

Approval is a constricted Range method.

Lost in all this is the fact that the general consensus among Range 
advocates is that Approval is an excellent first step. It's cheap, 
it's simple, it's easy to understand. And it *is* Range, just the 
maximally constricted version.

Then, we suggest, there will start to be pressure for refinements. 
Those refinements can go in two directions: ranking or rating. 
Approval, because of the restriction to Yes/No on each candidate, can 
also be considered a ranked method with only two ranks. And, so, 
another rank could be added, and the vote analyzed pairwise as a 
truncated Condorcet method, or analyzed in Range style, perhaps as 
sum of votes.

Taking the first step of Approval does not prejudice us with regard 
to the next step, which could go in the ranked method direction or 
the higher-resolution Range direction. I *am* a Range advocate, which 
means that I have, so far, concluded that Range is generally a better 
method than ranked methods, but it should be noted that I don't 
consider Range the be-all and end-all of election methods, not 
simple, single-step Range.

And, in fact, I would have the Range ballots analyzed pairwise as 
well, and when the Range winner would be beaten in a straight-on 
contest by another candidate, I'd hold a runoff. A real runoff, not a fake one.

Once again, this takes us outside what many election methods 
aficionados are accustomed to, for there is no way to clearly predict 
what will happen with such a runoff. But if we *do* have such a 
runoff, we have done something quite significant, which is to make 
our Range method compliant with the Majority Criterion (and almost 
certainly would select the Condorcet winner, if there is one and that 
preference remains through the runoff). To be more than a joke, the 
Range method would have to have sufficient resolution that voters 
could indicate a preference between the top two rated candidates on a 
single ballot, with minimal damage to voting strength between both of 
these candidates and all other candidates. With low-res Range, I'd 
actually propose a Favorite marker, counted the same as the 
other-wise max rating for Range results, but used in the Preference analysis.

>  If someone has as
>exact and compact formulations (to fix this one or to propose a new
>one) on where and how Range works please put them forward.

Wikipedia currently has:

>Range voting is a <http://en.wikipedia.org/wiki/Voting_system>voting 
>system for one-seat elections under which voters score each 
>candidate, the scores are added up, and the candidate with the 
>highest score wins.

Now, looking again at what had been written:

>I think Warren Schudy put it well in a  July 2007 draft paper:
>
>"Range voting is a generalisation of approval voting where you can 
>give each candidate any score
>between 0 and 1. Optimal strategies never vote anything other than 0 
>or 1, so range voting
>complicates ballots and confuses voters for little or no gain."

This is from a draft paper, and we don't have that paper, so we don't 
know the context. I objected to a number of things about this statement.

(1) "Range voting is a generalization of Approval Voting." Maybe it's 
not important, but this implies that Range was developed as some kind 
of extension of Approval Voting. I don't think that is true. Range 
Voting actually comes, I think, from utility analysis. If one is 
trying to maximize utility over a series of "instances," i.e., 
voters, for a set of choices, one collects, for each choice, the 
utilities from each instance and adds them.

Approval Voting could be considered a simplification of this, but, 
again, that's not where it came from.

It is more accurate and complete to say, if you are going to say 
anything about Approval in defining Range, to note that Approval is a 
special case of Range where the only two allowed votes are 0 and 1. 
And that is what Wikipedia does.

Secondly, the statement that "optimal strategies never vote anything 
other than 0 or 1," doesn't appear to be correct, which certainly 
would be a problem in a Range definition!

Thirdly, that there is "little or no gain" from allowing intermediate 
votes does not follow from the comment about optimal strategies, 
unless we were considering only the gain of the individual voter, 
rather than gain to society overall.

Thus the entire statement is controversial, and controversy like that 
doesn't belong in an "exact and compact formulation."

Ballot complexity is a red herring. Approval is Range 1 and it is 
minimally complex, the simplest possible ballot, in fact.

Range 2 only adds one option; this has been used in MSNBC polls 
(-,0,+), and it is quite interesting to compare it to Plurality 
polls; it clearly provides much better information about how voters 
feel about the candidates.

In any case, there is minimal additional complexity here, and whether 
or not it provides a gain is *quite* controversial. I think it's 
pretty clear that it would, and simulations confirm this. Got any 
better method of predicting gain or loss?