[Election-Methods] Response to Schudy re Range vs Approval voting

[Abd ul-Rahman Lomax]

At 12:11 AM 8/8/2007, Juho wrote:
> > But that is what Range is!
> >
> > Does the method change based on how we describe it?
>
>In some sense yes. In practice the situation is much more complex but
>in theory one could say that: 1) Strategic voters will study the
>method in detail and they will find the most efficient strategic way
>to vote irrespective of how the method is described. 2) Sincere
>voters will vote in the way they were told to vote.

Look, this is an assumption that "sincere" is equivalent to 
"clueless." Absolutely, voters should not, on the ballot, be "told 
how to vote," except where there are rules that will cause their 
ballot to be invalidated if they violate them. For example, "Vote for 
one" on a plurality ballot must be there! However, "Vote for one" is 
actually inadequate. Some voters, I'm pretty sure, will think that 
the votes will be counted.

And voters know that even though it says "Vote for One," nothing will 
happen to them, lightning will not strike, if they don't vote for any 
at all -- and there are many blank ballots in real elections -- or 
they vote for more than one -- which can be a quick way for a voter 
to nullify a vote on a paper ballot, some of these are, I'm sure, deliberate.

What was being presumed here was that the ballot had defective and 
biased instructions on it. Why make that assumption?

Most voters, for example, are quite accustomed to voting 
"Strategically." They vote for one of the major party candidates, 
even if it means holding their nose. Are these not "sincere" voters?

However, if "sincere" voters do vote the min and max for 
frontrunners, they lose nothing, and if they vote a little less than 
max, say, for a frontrunner when they prefer someone else, they lose 
very little. As I've written, if the range resolution is sufficient, 
the loss of voting power by a lowered vote for a frontrunner is miniscule.

What Range does, like Approval, is to make the vote for any candidate 
independent of the vote for any other. You are casting multiple 
votes. So if it is important to you that a candidate get votes, most 
likely you will give the maximum you can give, or *maybe* just a tad 
less, if you want to preserve preference order. As long as you don't 
go *below* your actual rating for the candidate, you are quite safe 
as far as expected outcome is concerned. In a highly contentious 
environment, you will tend to give the max or max - 1.

And I'm sure that if we have Range in an election, it will be broadly 
discussed. People, most of them, will know how to vote for maximum effect.

>In Range this means that if you present Range as a method that elects
>the candidate with best utility sum after the voters have marked
>their personal utilities in the ballots, then sincere voters have
>been advised in a way that makes them vote differently than what the
>strategists will do.

The determination of the winner in Range has *nothing* to do with 
"utility." Personal utility is a *strategy* for how to vote. It 
happens to be one that will tend to maximize overall social utility, 
so it has a value entirely apart from the voter's personal goals. 
However, the choice of whether or not to seek to maximize personally 
desirable effect or overall SU, and whether or not to trust other 
voters to do likewise, is completely up to the voter. And it is quite 
debatable which is really best for the voter. What does seem to be 
clear is that harm from voting "sincerely," i.e., accurately 
according to personal utilities, is small.

The ballot instructions would not mention "utility." They probably 
would not mention "rating." They would mention "votes." What other 
people will say about the method is up to those other people. Some 
candidates will be saying, "Vote max for me, min for everyone else." 
They could easily lose my vote by giving me that advice! Except under 
some circumstances....

Range as a method optimizes overall expected voter satisfaction, if 
voters express their expected satisfaction on the ballots. If they 
don't, it doesn't. However, it tends to optimize overall satisfaction 
for the voters who express it, as long as they normalize, i.e, vote 
max for their favorite and min for the least liked, and it gets even 
better with somewhat more sophisticated voting strategies. Voting 
Approval style is *fine*. There is nothing wrong with it, and that 
Range allegedly may encourage many voters to vote that way is not an 
argument against it.


>  If you present range as a method where the votes
>are expected to promote the candidates with either min or max values,
>or with intermediate values if they don't want to use their full
>voting power for some reason, then strategic and sincere votes are
>closer to each others (some additional but more complex strategic
>options that sincere voters might not use may still remain).

It doesn't need to be that complicated. In Range 10, you may cast 
from 0 to 10 votes for each candidate. The winner is the one with the 
highest total of votes.

What should guide your voting? Read your party Vote This Way or We 
Won't Like You pamphlet!

Seriously, it is up to voters. I happen to believe that they are not 
stupid. I predict that most of them will vote just like some have 
said: they will give max to their favorite frontrunner and min to the 
other. But *many* will add intermediate votes for other candidates, 
thus giving rise to what Smith calls the "incubator effect." New 
parties are allowed some breathing space. Their support will be 
measured, it will not be invisible and suppressed by the need for 
true strategic voting.

It's ironic that a method which makes true strategic voting, which 
has generally referred to preference reversal, unnecessary, is 
accused of somehow fomenting it. *Smart* voting is still needed. If 
you have 10 votes to cast for each candidate, and you don't cast all 
of them for a frontrunner, and none for another frontrunner, you are 
obviously wasting some of your voting power!

>(A third quite common way to describe Range is to use normalised
>values.)

That's not a description of Range. Normalization is not part of the 
method description, unless there is a normalization stage in the 
counting (which has been proposed, but it complicates things, for little gain).


> > Absolutely, someone could describe Range, on a ballot, in a way
> > that would encourage voters to waste their vote. Consider it an
> > intelligence test. If you see through this defective advice, your
> > vote will count for more!
>
>Yes. This is something I would not like to happen. It is better to
>try to give all voters the same power, irrespective if they are less
>strategic or less analytical.

But they all have the same power no matter what we say. In a real 
election, in today's world, lots of people are trying to convince 
them to vote against their own interests. Nothing new about that! But 
none of these people are *taking away* the power of the voter. Unless 
they do something *else*, like preferentially "clean up" the voter 
registrations.


> > But I would never support such a description being on a ballot.
>
>Yes, it is too late to try to educate voters using the ballot. They
>should know beforehand.

The ballot description should be enough. Look, Range is *really* 
simple, if we leave out the "quorum rule" that is currently in some 
descriptions, and that is an argument for leaving it out. You get 10 
votes, and you can give anything from 0 to 10 votes to each 
candidate. The candidate with the most votes wins.

It's not hard to understand! So even if a voter has not been paying 
attention, most would get it.


> > The description should inform the voter how the voter's marks will
> > be used. One of the descriptions that has been suggested (by me?)
> > is precisely that "You have 10 votes, and you may cast as many of
> > them as you like for a given candidate, without any restriction on
> > how many you cast for another candidate. The winner will be the
> > candidate with the most votes. For each candidate, mark the
> > position on the ballot corresponding to the number of votes you
> > wish to cast for that candidate."
>
>Maybe some example votes would clarify to the voters what the typical
>voting patterns are. This description may still lead to different
>interpretations by them.

Like what?


> > That's Range 10.
>
>(This naming policy btw has the problem that nowadays I don't know
>what method people are talking about when they say "Range 2". You
>could mean the ability to give from zero to two votes or the ability
>to give two different kind of values.)

I realized that there was no convention for Range N. Cardinal Ratings 
10 has ten ratings, that was established. So I have defined it 
differently, because we have the opportunity, and the different 
definition is actually clearer (though not necessarily to someone 
familiar with the CR-N terminology.)

Juho, I invented "Range N." People can use it or not. We were 
informally using it, sometimes to mean CR-N, sometimes not. So I 
announced how I was using it.

Range N means that there are N preference steps. It means that you 
are like a voter with N votes. Approval is thus Range 1.

If I'm correct, few people ever referred to Approval as Range 2, 
without adding an explanation. I may have done it, it was always a nuisance....

I found that when I tried to keep it the same as the Cardinal Ratings 
measure, I was always writing (N-1)...

Simple, keep it simple. Range N, you have N votes. Plurality is a 
restricted form of Range 1, where you may only vote for 1 candidate. 
Range 0 means you get no votes!

Lots of people express range ratings as percentages. Quick, in Range 
100, you vote 50. What percentage is this?

With my terminology, it's 50%. With Cardinal Ratings, it is 50/99.

This is a Free Association, nobody is obligated....


It works beautifully.