List Procedures, Contd.

Mike Ositoff ntk at
Thu Jul 30 10:53:06 PDT 1998

My motivation, for Rank STV(MO), wasn't so much about results.
It was more an aesthetic matter. How to choose what _reason_ to
use to determine which count to give precedence to, when 2 
successive counts conflict.

What attracted me was to say, for that nonmonotonicity conflict,
to give precedence to the count whose N is closer to the current
number of seats, that N being more likely to be the next fractie size.
I liked the sound of that as a reason behind a precedence rule.

Then I'd say "To determine which of the current new winners gets
the next list position (when giving precedence to the previous
count), or to order all of the new winners for addition to the
list (when giving precedence to the new count) use the
1st choice vote totals. Or one of the better single-winner methods
(use it repeatedly when ordering several candidates)."

Unavoidably, that paragraph has to be longer than a corresponding
paragraph for Rank STV(NL), because of the 2 possibilities.

Usually brevity of definition is important when proposing a
new electoral procedure, and that consideration probably takes
precedence here. A briefer rule to to have the precedence one
way is easier to propose.


Hand Count:

The easiest system for a hand count would be SNTV. Each voter votes
for 1 candidate, and they're ordered in the list according to their
vote totals.

Compared to Borda, SNTV is very much easier to count. And it's
a form of PR--one which doesn't need a quota or a pre-known 
number of seats. It's proportional only to the extent that factions
use good, well-informed strategy.

At a meeting of the leaders of a faction, someone might say:
"This party will probably win about 14 seats. Our faction has about
1/3 of the party's voters. So let's run only 4 candidates, just to
be on the safe side, and ask our faction's voters to each draw one
of the 4 names out of a paper bag, & vote for him/her."

Alternatively, the factions voters could be asked to vote for
a particular candidate if they live in a particular geographical
district, or if their name is within a certain part of an
alphabetical list of names. I prefer the random choice of
drawing a name from a paper bag, because then no one's telling
the voter whom to vote for--he's randomly determining it for
himself. Easier to administer too.

Bottoms-Up seems a clear improvement over SNTV, and is most likely
the best that could be done with one quota-less count. Though
the count labor is much more than SNTV, it's still feasible.

But, considering that computers are now available, there's no
reason to not use successive STV counts, Rank STV. That's the
best recommendation for best proportionality no matter how
many seats the party wins.

So I suggest Rank STV (successive STV counts with increasing N).
The version of Rank STV with the briefer definition may well be
the one that will go over better with the party members.

Mike Ossipoff

More information about the Election-Methods mailing list