[EM] recommendations

[larrythecucumber at adelphia.net]

This is the same post I posted last night, except that I have amended it to make it less wide.

This is my first e-mail to this listserv, although I have 
been checking the monthly pages for interesting entries 
for over a year and a half.  I can’t believe I waited so 
long to join in.  For some reason I’m more nervous about 
joining listservs seems than joining web discussion 
forums.  Maybe it’s because the last time I joined a 
listserv I was chased off by another member because I 
wasn’t a member of the political party which operated the 
listserv.  Well, on to the topic.

As for which single-winner method is best (at least in 
elections with large electorates and possibly all 
elections where there is the possibility of more than two 
options), I currently support a Condorcet method I like to 
call Locked Preferences.  It is very similar Maximize 
Affirmed Majorities, but the number of votes against a 
defeat is not used as a tie-breaker when two defeats are 
supported by an equal number of votes and pairwise ties 
which need to be resolved are treated essentially the same 
as defeats of equal strength that cannot both be affirmed, 
if they are not already treated the same in MAM (this 
could make a difference if two opposite preferences and a 
third preference that would contradict one of them and an 
already affirmed majority are expressed by an equal number 
of voters; I say “essentially the same” since a lot 
tiebreaker with each of the two candidates having an equal 
chance of winning would produce no different results from 
Steve Eppley’s tiebreaker if one pairwise tie was the only 
tie that needed to be broken).

These are the only real differences between Locked 
Preferences and MAM, but they are made more natural by the 
following difference in heuristic: instead of determining 
the winner of each pair and sorting each pair by the 
method described above, you sort the preferences 
themselves, all of them (“defeats”, minority preferences, 
and preferences that are part of a pairwise tie), by the 
method described above.  Using this heuristic could 
perhaps win over some supporters of margins being used to 
sort defeats rather than winning votes, since it would 
show that votes for a majority preference overriding those 
for the opposite preference is no different than votes for 
two defeats overriding those for a smaller defeat which 
contradicts those defeats.  It also just seems more 
natural to me.  My reason for not using the number of 
votes against a preference as a tiebreaker is that non-
random secondary sort factors like that don’t seem natural 
to me.  If margins aren’t used in sorting defeats in 
general, they shouldn’t be used as a tiebreaker either 
(and the preference opposed by the least votes among 
preferences expressed by an equal number of voters is the 
same as the preference with the greatest margin).  That 
same sense of what’s natural, plus the concept of 
comparing preferences rather than comparing pairs of 
preferences, is why I favor treating all cases where 
preferences expressed by an equal number of voters the 
same way as long as the outcome hinges upon the resolution 
of those ties.

Moving on to multiple-winner elections, ideally I would 
support CPO-STV, in pure form except that equal rankings 
would be allowed (how that would work is a topic for 
another post).  Ideally I would not even support the use 
of outcome elimination shortcuts since that would deprive 
citizens of a chance to see the full results.  But I have 
to be reasonable.  Eliminating outcomes from consideration 
that do not include candidates who would meet/surpass the 
quota in an STV count without exclusions would save a lot 
of time and money in most multiple-winner elections 
without compromising the integrity of the method or even 
(except rarely) eliminating comparisons of pairs of 
outcomes that are worth looking at.  I’m a little less 
sure about excluding outcomes from consideration that 
include candidates who cannot possibly win given evidence 
apparent early on, unless those candidates are write-in 
candidates.  I feel that a candidate has a right to see 
how close he or she came to being elected, and at times a 
polarizing candidate who would be excluded in one of James 
Green-Armytage’s shortcuts ( 
http://lists.electorama.com/pipermail/election-methods-
electorama.com/2003-July/010316.html ) may actually have 
come close enough to winning that a recount would not be 
unreasonable.  I also don’t like the idea of local CPO-
STV, which is described in the link shown above, although 
I can certainly see why it was proposed since it can 
greatly reduce the number of comparisons necessary and 
will only very rarely yield a different result than pure 
CPO-STV.

In some aspects, however, I am willing to support a far 
less pure CPO-STV method.  In order to make voters’ number 
of options less overwhelming, thus making it easier for 
them to make an informed ranking of all the options, I 
propose allowing candidates for the same office to run in 
slates, including slates established by political parties 
by the same methods used to choose nominees today, 
preferably in primary elections where eligible candidates 
could also run in slates.  A first place vote for a slate 
would count as a first place vote for the first candidate 
on that slate, a second place vote for the second 
candidate on that state, and so on up to the number of 
candidates on the slate, which should not be allowed to be 
greater than the number of candidates to be elected.  A 
second place vote for a candidate or slate would in effect 
count as an n+1 place vote for that candidate or the first 
candidate on that slate, where n is the number of 
candidates on all slates ranked above it, and an n+2 place 
vote for the next candidate on that slate, and so on.  One 
might be able to use shortcuts in addition to those 
proposed by Mr. Green-Armytage without further 
compromising the integrity of the method.  That, in 
addition to the fact that the number of distinct outcomes 
would often be much less than in the same election under 
regular CPO-STV, would in most cases greatly reduce the 
number of comparisons of pairs of outcomes which are 
necessary.  One might think that if candidates are going 
to run in slates than the addition of a Condorcet 
component is not worth the additional effort on the part 
of both the voters and the ballot-counters and the 
additional cost that will likely result.  I feel, however, 
that the Condorcet component provides some immunity from 
complaints that a certain party or slate benefited or was 
hurt unfairly by the way votes were divided among other, 
more similar slates.

I will conclude my description of my favorite multiple-
winner election method by answering the three questions 
which must be asked of any method with both a Condorcet 
and an STV component.  Not surprisingly, I support my 
favorite single-winner method, Locked Preferences, as the 
Condorcet method used to determine a winning outcome was 
all the preferences of one outcome over another have been 
calculated.  I support the use of the Newland-Britain 
quota, especially in “party-list CPO-STV” where the Hare 
quota could potentially penalize a group of candidates for 
running in one slate, but in regular CPO-STV as well.  I 
agree that using the Hare quota would lead to more 
proportional results.  However, I feel, for example, that 
if there are only two groups, each with 100% common voter 
support, and one group receives more than twice as many 
votes as the other, then the first group should be awarded 
a second seat before the second group is awarded the first 
seat since each of two candidate’s equal shares of the 
first group’s votes is greater than one candidate’s whole 
share of the second group’s votes.  Using the Newland 
Britain quota assures that this will happen.  The way CPO-
STV works, you won’t even need to have the caveat about a 
candidate’s votes having to exceed the quota for that 
candidate to be elected, except in the preliminary STV 
count to determine which outcomes can be excluded from 
consideration.  As for what surplus rule to use when there 
are multiple surpluses, I haven’t thought about it or 
researched it enough to make an informed decision.

I haven’t thought much about voting in smaller groups and 
about direct democracy beyond where there are more than 
two options on referendum questions.  (Maine, where I 
live, has had a couple of notorious three-way referendum 
elections in the past decade.  One was on restrictions on 
forestry practices including clearcutting and the other 
was on a shift of some of the funding for education from a 
primarily property tax source, the municipalities, to a 
largely income and sales tax source, the state – pretty 
divisive issues in Maine.)  This isn’t much different from 
choosing a favorite single winner election method, 
although I’m not sure I’d choose the same one.  I lack the 
energy to delve into the new realms of smaller group 
voting and proxy voting.  Thank you though, James, for 
starting this topic which finally got me to join this 
listserv.

Sincerely,

Kevin Lamoreau