[EM] RE: [Condorcet] Truncation

Abd ul-Rahman Lomax abd at lomaxdesign.com
Wed Sep 14 13:55:08 PDT 2005

At 04:56 PM 9/13/2005, Kevin Venzke wrote:
>--- Abd ul-Rahman Lomax <abd at lomaxdesign.com> a écrit :
> >Kevin may have talked me out of Condorcet entirely (unless truncation
> >is better handled, and unless there is a decent consideration of
> >approval cutoff.)
>Truncated candidates are considered tied with each other, and ranked
>strictly below all ranked candidates. [examples given]
>How would you want to handle truncation instead?

That is the best way. The problem is the effect 
truncation can have. I'm not that familiar with all the Condorcet variants...

Without an Approval cutoff, it becomes impossible 
to tell whether a vote is a true approval or is almost a complete rejection.

Warren Smith has written quite a bit comparing 
Range Voting with Plurality, IRV, and Approval, 
and has made the claim that Range shows more 
"support" for third party candidates than the 
other methods; however, without some kind of 
definition of what the range numbers *mean*, 
interpreting the numbers as "support," unless 
they are very high or very low, this is basically speculation.

After all, a grade of 60% would typically be an 
F. In a study, Warren Smith compared a few 
percent plurality or approval vote with a 
twenty-some-odd percent Range vote, and asserted 
that the latter was "more support." The latter is 
more accurate a measure in some ways, yes, but 
whether or not it is support or rejection is a different question.

Approval cutoff is easy to add to ranked ballots. 
All it takes for full approval expression is to 
add a dummy "candidate" which is the margin between approved and disapproved.

Whether or not to *use* the generated information 
in determining the winner is another question, as 
well as how to use it if it is used. I argue 
below, however, that in initial reforms, only 
approval information should be used, not ranking.

As I have often written, there is a basic 
question that must be addressed before it is 
possible to intelligently pick an election 
method: supposing everyone votes sincerely, which 
is better for society: the Condorcet winner or 
the Approval winner? (Roughly, the most popular 
single candidate, or the candidate with broadest support.)

People, including election methods experts, don't 
agree on the answer to this question, so, of 
course, they are not likely to agree on the best method!

I have my own opinion: democracy is fundamentally 
government by consent of the people, and the 
broader that consent, the more democratic the 
society. Plurality violates a basic democratic 
principle, the will of the majority. It easily 
elects leaders who were *disfavored* by a 
majority. Condorcet methods don't guarantee a 
majority winner, though they make a minority 
winner less likely. Approval methods also come 
with no guarantee, but probably will produce a 
winner with the broadest acceptance. And this should be good for society.

There is no guarantee that basic Approval would 
be voted sincerely. But at least it would make it 
possible! (And the cost of basic approval is 
minimal, it is easy to understand, etc.) Approval 
voted insincerely is no worse than Plurality. 
Tossing the no-overvote rule is something that 
ought to be done regardless of what other reforms 
are undertaken. It was *never* a good idea.

(Allowing overvoting also turns standard ranked 
ballots effectively into approval ballots; basic 
Approval is a Range method with two possible 
ratings (0 and 1). A truncated IRV or Condorcet 
ballot where the voter only votes for approved 
candidates, and ranks them equally, is exactly 
the same as an Approval ballot. So one method of 
implementing, effectively, Approval with an IRV 
or Condorcet ballot is to allow overvoting: 
ballots with two or more candidates marked at a 
given rank should not be discarded!)

(High equal rankings would generally mean 
approval, low equal rankings would generally mean disapproval.)

What I'm suggesting is that simply repealing the 
no-overvote rule should be a baseline reform, at 
the worst harmless, and quite possibly much 
better than that. Then, if there is the political 
capital or energy for more complex reforms, 
typically involving ranked ballots, fine. As long 
as it is possible for voters to rank equally, it 
is democratically acceptable. I'll point out that 
if a Condorcet method does *not* ostensibly allow 
equal ranking (perhaps it is implemented on a 
lever machine that simply won't allow the entry), 
it still is allowed through truncation, but only 
at the bottom end. Why not allow it at the other?

One simple extension beyond basic Approval would 
be the addition of an extra option for each 
candidate. Instead of the single position to 
mark, there would be two, perhaps labelled:

(1) Preferred
(2) Approved

Rejection being assumed for any unmarked 
candidate. (I have elsewhere argued that even if 
a candidate is a genius and a saint, he or she 
should not be elected until and unless enough 
people *consent* to the election. So positive 
consent is a basic necessity, failure to consent *means* rejection.)

This is only slightly more complicated than 
simple Approval. Again, if overvoting is allowed 
for both positions, the voters can single out one 
or even a few candidates as Preferred, more as 
Approved, i.e., "acceptable." This answers the 
objection which has been made about basic 
approval that voters should be able to specify 
their favorite; this ballot would allow it.

Would that information be used in determining the 
winner? There are methods which would allow full 
use. As a Range method, one might assign 0 to 
blank, 0.5 to Approved, and 1 to Preferred. But, 
it will be noted, this will create strategic 
incentive to always mark the preferred 
frontrunner as Preferred, or else the vote is 
effectively diluted and the voters risk seeing 
the preferred frontrunner lose because of the 
half-votes instead of the full votes that would 
be received in basic Approval. Sincerity would be penalized.

So, at this time, until there is more experience, 
and even if the expanded ballot is used, I'd 
recommend that the election be counted as basic 
Approval. Preferred and Approved would have the 
same effect as far as determining the winner is 
concerned. And *this* allows voter *more* 
expression of preference than they would have if 
the preference information is used!

Public campaign financing would be distributed 
based on Preferred Votes. If a voter votes more 
than one as Preferred, the money would be split. 
But the vote itself would not be split. It would 
do its utmost to elect any of the Approved candidates....

(And if there is a tie, of course, then there 
would be a ready means of breaking the tie! 
Highly unlikely that if there is an Approval tie, 
there is also a Plurality tie.)

Later on, there would then be real data from real 
elections which could indicate where to go with 
further reform. And, immediately, supporters of 
third parties could express that support, 
something they cannot do now. Approval or Range 
(or the other methods) are not going to magically 
create third-party winners. Whether or not they 
will favor or disfavor one or both of the two 
major parties is really unclear; historically, we 
can readily think of elections where having had 
this system would have reversed elections, and in 
some it has been one party that unfairly won, and in some it was the other.

This really should be a nonpartisan issue, overall.

However, when there is a single strong third 
party, not a centrist party and therefore more 
closely aligned with one of the major parties 
than the other, election reform could possibly 
work against the other party, which has been 
benefiting by the spoiler effect. This changes 
with time, though. What might help a little today 
might hurt a little tomorrow, and vice-versa.... 
In the end, we either care about democracy or we 
care about getting our personal way. My vote is 
with democracy. Even when I don't like the outcomes.

More information about the Election-Methods mailing list