At 05:26 PM 4/28/2010, Jameson Quinn wrote:
2010/4/28 Peter Zbornik <<mailto:[email protected]>[email protected]>
OK, thanks.
Please go on to propose the condorcet, if you think it is the best.
Approval voting was used in the French presidential election, first
round, where far-right nationalist Le Pen got to the second round.
Le Pen was hardly a centrist.
See
<http://en.wikipedia.org/wiki/Approval_voting#Effect_on_elections>http://en.wikipedia.org/wiki/Approval_voting#Effect_on_elections
Quote:"one study [16] showed that approval voting would not have
chosen the same two winners as plurality voting (Chirac and Le Pen)
in France's presidential election of 2002 (first round) - it instead
would have chosen Chirac and Jospin. To some, this seemed a more
reasonable result[citation needed] since Le Pen was a radical who
lost to Chirac by an enormous margin in the second round."
Peter
I think you're misreading Wikipedia there. Approval was not used;
the passage simply says that some suggest that if it HAD been used,
the results would have been better.
The study was of a test election conducted "in parallel," using
actual voters who voted in the real election.
http://rangevoting.org/FrenchStudy.html. This page on rangevoting.org
describes the test, and does link to the original paper. The
Wikipedia article links to the Wayback machine, and it wasn't
responding. I edited the Wikipedia article a bit.
The rangevoting.org paper considers the use of approval as the first
round in a two-round election, and suggests that this would have
chosen Chirac and Le Pen to go into the runoff. The landslide for
Chirac in the runoff, with increased turnout, shows that Chirac vs.
Le Pen was not a good runoff choice. The problem was massive
vote-splitting in the primary. Any advanced method in the primary
would have produced a better result, probably. Approval alone in the
primary, if used to finish the election, would likely have chosen
Chirac as well, based on the French study, but there was serious
majority failure, and thus a runoff would really be important.
(Trying to decide elections with *many* candidates using a single
ballot is difficult. Doing it with plurality in a primary and a
runoff is known to fail in exactly this way, this was not the only
well-known election to show this effect.)
Le Pen had very high "core strength," i.e., his supporters were very
exercised to elect him. But that was it; while overall turnout
increased in the runoff, Le Pen only gained a small number of votes,
whereas all other votes were turned to him, so this was the heaviest
landslide ever seen in a French Presidential election. Had it been
Chirac vs. Jospin, it would have been close. My guess is that turnout
would have been substantially lower, and that Jospin would have won.
But the proof would be in the pudding.
Bucklin in the primary, and with that many candidates, more Bucklin
ranks, possibly, though 4 ranks (3-rank traditional plus the default
No vote of a blank) in Bucklin-ER can handle a lot of candidates.
Would encourage a certain increase in the addition of approvals over
standard approval voting, which doesn't allow the specification of a
preference among approved candidates. In Bucklin-ER, one can
categorize candidates in up to three ranks, with standard 3-rank
Bucklin, and these are all approved ranks. Standard Bucklin had only
one unapproved rank, one placed a candidate here by simply not voting
for the candidate.
But a range ballot could be used to feed Bucklin just as well as a
ranked ballot. That ballot, if it has enough ranks, could allow
complete ranking; if voters simply rank all the candidates in
sequence of preference, the ballot becomes a Borda ballot, which is
often a good approximation of a Range ballot.
I believe that using Bucklin in a primary, with Range ballot input,
but only using the approved categories to determine a winner by a
majority, if that exists, and then using the ranking and rating
information to make better choices of runoff candidates, allowing up
to three, would handle a wide variety of election situations with a
voting method that is still very easy to count, it is just the sum of
votes in each rank or rating that is needed, it is precinct summable.
----
Election-Methods mailing list - see http://electorama.com/em for list info
