[EM] Random thought on Range Voting

[Brian Olson]

On Jan 3, 2005, at 9:41 AM, Rob LeGrand wrote:

> It has been claimed that Range Voting might be an easier sell than
> Approval as a voting reform, which could be true.  And I understand
> that some Range advocates see the fact that many voters would vote
> sincerely as a good thing.  But since strategic voters would have
> more power in a Range election and might be seen as "cheaters" by
> the sincere voters, I think there would likely be a public demand
> for restrictions on voting candidates at the extremes, turning
> Range into something more like Borda.
>
> When I advised the Free State Project (www.freestateproject.org) on
> voting systems for their choose-the-state-to-move-to election, they
> initially wanted to use cumulative voting.  I managed to convince
> them that cumulative reduces to plurality when voters are
> strategic, ...

I think it's been shown that the optimum strategy is not to 
vote-for-one (plurality) on a ratings ballot, but to vote max-rating 
for any choice above some threshold internal to you, and min-rating for 
the rest. Thus straight cumulative vote degenerates to Approval under 
strategy (not plurality).

> but then they offered to add restrictions such as "you
> can't give more than half of your votes to any one candidate",
> which would make the system worse.

I think the answer is to tinker with the counting on the back end, not 
make limits on how a voter can vote. Systems like "Instant Runoff 
Normalized Ratings", or James Green-Armytage's ratings-based Condorcet 
cycle-break method solve the immediate shortcomings of straight 
cumulative vote, and still provide the satisfaction of an expressive 
ballot.

> I believe restrictions for
> Range Voting such as "you can't give any two candidates the same
> rating" (when the number of allowed ratings is finite and fairly
> small) would be intuitively appealing to many voters who would like
> to vote sincerely and want to force others to do so.

Everyone wants to limit what someone else can do so that they can't 
cheat? Maybe it's just me but that sounds somehow socially cynical.

> Approval
> Voting makes it obvious that it is natural and acceptable to vote
> at the extremes and so would offer no such temptation to tinker
> with the system.
>
> How could Approval be tinkered with after adoption?  Although I see
> it as unlikely, some voters might want to limit the number of
> allowed approvals.  But allowing n approvals in a race would allow
> n + 1 parties to compete fairly in that race, which is still a
> strict improvement over plurality.

Hmm, recast Approval as not "some number of yes votes", because that 
violates some people's sense of one-person-one-vote, but instead "a 
yes/no vote per choice". I think that should cancel any desire to limit 
the number of yes votes; which is a degradation and not an improvement 
on Approval vote.

If you want to improve Approval, you're going to need a more expressive 
ballot, at which point you may as well just move to rankings or 
many-valued-ratings (as opposed to the two-valued-rating of Approval) 
ballots.

Brian Olson
http://bolson.org/