[EM] Please read Range Voting article on YourHub

[Jan Kok]

Thanks, Dave, I really appreciate your effort to help Denver get a
better voting method. You could post your own article, and I would be
happy to reply to it. Posting your own article would get more
attention, and you can format it better. The comment box doesn't even
allow paragraph breaks.

By the way, everyone, Rob Richie of CVD and I have exchanged a couple
emails. They are posted on the RangeVoting Yahoo group.

On 7/29/06, Dave Ketchum <davek at clarityconnect.com> wrote:
> I tried to register there - no luck, so here are my thoughts:

I just registered Thursday and had no problem. You could write to
support at yourhub.com, or there might be a phone number you could
call.

>
> We start with runoffs being seen as too expensive in various ways.
>
> Thus Plurality voting is unacceptable, for it only sees voters' first
> choices and, given three or more candidates, each could be first choice of
> a third of the voters - need to let the voters say more.  The three
> methods below promise needed capability without imposing excess complexity
> on voters.

In a standalone article for the public, you'll need to explain that
Plurality is our usual vote-for-one voting method.

>
> Councilwoman Kathleen MacKenzie proposes IRV, a ranked choice (RC) method.

Starting with San Francisco, I think, CVD has started to call IRV
Ranked Choice Voting. I think the original reason for that is they
suspected they wouldn't be able to report results on election night,
hence not "instant". But since they are now selling RCV, we can point
out that Condorcet is another kind of RCV. Maybe CVD will bite. I
really hope so.

> Jan Kok proposes Range Voting (RV), claiming "it is fully compatible with
> all existing voting machines".  I have to question this claim since RV,

See http://rangevoting.org/VotMach.html

> like RC, requires that the voter attach a number to each voted for
> candidate.  BUT, what does a difference of two in range mean?  Each voter
> has to decide intent; counting will decide meaning - let us stay with RC
> which provides needed meaning without excess complexity.
>
> I propose Condorcet which, given the same ballots as IRV but scoring as if
> reporting a tournament, sometimes does better at selecting a winner.

You don't want to give the impression that it will sometimes do worse.
You could say "often chooses the same winner as IRV but sometimes
chooses a better winner."

> So I see need for new voting machines - at least need for new/validated
> programming - since any above method will need this.

No, I don't think it's necessary for RV, unless all those bubbles bump
into a capacity limit.

>  Then I read the
> horror stories too many new voting machines have earned.  Seems to me that
> machines that have earned horror stories after acceptance by a testing lab
> demonstrate that such labs are unable/unwilling to notice and report
> problems and/or some governments ignore reported problems.   BTW - what
> computers can do successfully demonstrates that they have to be capable of
> a task as simple as voting, PROVIDED that becomes the goal of those in the
> doing.

I don't understand that last sentence.

> Finally, a deadline approaches if a needed ballot measure is to be on the
> ballot this November.  Perhaps the best that can be done is to propose
> moving ahead based on what I write above, while detail decisions are given
> needed time for more study (e.g., agreed to leave Plurality but studying
> where to get to).

Right now the entire voting procedure is specified in the Charter,
which MacKenzie and probably others don't like to change often. So,
one possibility is to change the charter to move the election rules
somewhere else that's easier to change. But that makes me nervous that
someone would mess with it while I wasn't watching and do something
awful, like that North Carolina 40% threshold IRV!

>
> In all three methods the voter selects from the candidates, the one or
> more worth positive rating:
>       For RV each rating is a range number.

I usually call them "ratings" or "scores".

>       For RC each rating is a rank number that shows only which is liked
> more, but does not attempt to be understood as to how much better one is
> liked than another.
>       For both RV and Condorcet, ties are permitted.
>
> For counting I ignore RV.  The two RC methods usually agree as to winner,
> but I show a set of 100 ballots for which they differ:  30 A; 22 B; 23
> C,B; 25 D,B:
>       IRV simulates a runoff, discarding what it sees as least liked until
> what remains indicates a winner, but does not look at more of any voter's
> ranking - here 30 A beats 25 D.
>       For each pair of candidates Condorcet will count how many voters
> have rated each as beating the other - here 70 B beats 30 A.  Normally one
> will win for beating every other, but such as M>N>O>M can happen with near
> ties and the rules must define how to decide the winner for such.

Believe it or not, those last paragraphs are probably too technical
for an editorial. I'd suggest saying something about how IRV can
sometimes make illogical choices of winner or fail to choose the
"most-preferred" candidate as winner. Then point to a web page for the
detailed explanation. I think it helps a lot to use real-life
examples, so the IRV gang can't say it's theoretical and will never or
almost-never happen in real life. See
http://rangevoting.org/FunnyElections.html for some real-life
examples.

Cheers,
- Jan