STV for party candidate lists?

[Mike Ositoff]

Regarding the use of Borda in those elections, of course there 
are better rank-balloting methods. But until quite recently, really
good rank-balloting count methods weren't being proposed. 
In fact, I claim that, with the exception of Bucklin, rank-balloting
methods better than the nonranked Approval method hadn't been 
proposed till the early 1990s. Bucklin is just a little better
than Approval, because if you're one of the voters who put the
Condorcet winner over a majority, then you lose nothing by
voting less-liked alternatives in your ranking. Same for voters
whose votes would have arrived at the Condorcet winner later than
yours, had the count not been stopped. In other words, then, some
voters aren't penalized by voting too long a ranking.

Probably, if Borda was adopted for that system some time ago,
there'd been much less discussion of single-winner methods, and
very likely Borda & IRO were the only methods at all well-known.

Borda is at least probably better than Plurality, and that would make
it better than IRO.

Like IRO, Borda can force strategic voters to abandon their favorite,
and rank something else in 1st place. But Borda's strategy is far
simpler than that of IRO: Just rank the alternatives in order of
their strategic value (as I've defined it earlier on this list).
Of course, lacking information about tie-probabilities, you'd just
sincerely rank the alternatives.

The reason I say it seems better than Plurality is that at least
you can give _something_ to alternatives with 2nd best & 3rd
best strategic value, etc. 

One thing I especially dislike about it is that it's sometimes
(usually?) defined so that you have to rank all the alternatives
or candidates on the ballot. If you despise candidate X, but there
are several that you despise more, then you're forced by that
rule to place X high in your ranking. In that form, maybe it's
worse than Plurality. If you're allowed to vote
a short ranking, still giving top points to your top candidates,
then it's surely better than Plurality.

Predictably, if it were desired to replace Borda, for that 
application, I'd suggest Schulze's method, Smith//Condorcet(EM),
or Condorcet(EM).

Demorep suggested Approval. Though Approval isn't as good as
the above 3, it's still better than Borda. It's natural to believe
than any rank-balloting method must be better than a nonranked
method. But Borda limits the number of points that you can give
to candidates, in comparison to a flexible point assignment
method that leaves that choice to you (within specified bounds).
But, as we discussed earlier, with those flexible point systems,
your optimal strategy involves giving maximum points to some
candidates, and minimum points to the rest. So with the Approval
0-1 point system, you don't lack for anything that you need
in a point system.

And, unlike Borda, Approval doesn't ever require abandoning one's
favorite.

But Demorep again suggested limiting the number of candidates that
voters should be allowed to vote for. I asked him what the
advantage of that limitation would be and didn't get an answer.
The authors who've studied & discussed Approval don't propose
such a limit, and I'm not aware of one being used, except in the
election to choose stamp designs, now being conducted here--and
in that instance, the purpose of the limitation is probably to
reduce the number of votes that must be counted. That wouldn't
be a major consideration anymore in computer-counted public
elections. (A voter may only vote for up to 3 stamp designs in each
category).

Mike Ossipoff