A proposal for debate-inclusion voting
James Green-Armytage
When you have debates prior to a public election, there is a question as to who should be included. If the
number of candidates running for a single office is
very large, then you might want to hold some debates that do not include all of
them, in the interest of giving more debate time to those who are included. If
so, then you need to find a fair method that enables you to choose who to
include and exclude.
Let's
assume that we are able to take some kind of direct vote, or at least a poll,
where voters rank the candidates in order of preference. This is the
information which we will process and use to determine inclusion and exclusion.
When
choosing who to include in a debate, we might be interested both in
"breadth" and "depth". That is, we would be interested in
having the kind of breadth of perspective that a proportional representation
method would afford, and yet we might also be interested in affording more
slots to those who are close to being Condorcet winners. Especially if the
voting method for the final election is Condorcet efficient, we want to be
careful to include people at the the debate stage who are likely to go on and
win the office.
Thus,
it makes sense to use a single transferable vote / Condorcet hybrid for
debate-inclusion voting. Here is a proposal for such a system:
0. The number of participants in the
debate will be T.
1. Ranked vote.
2. Beatpath/ranked pairs tally, producing
an ordering of the candidates. Identify the top C candidates in the ordering.
These can be called the members of the "C set".
3. Modified CPO-STV (comparison of pairs
of outcomes by single transferable vote) tally, to fill all T slots in the
debate. The tally is modified such that it only considers outcomes which
contain all members of the C set.
You
can calibrate T and C as you like, to get the desired size of the debate and
the balance between breadth and centrist depth / inclusion of likely winners.
For example, T=10, C=4. That's a ten person debate which necessarily includes
the top four candidates in the beatpath/ranked pairs ordering. If you want a
smaller debate you could do T=5, C=2. Or T=2, C=2, in which case the
beatpath/ranked pairs ordering is all that matters. Or you could do T=15, C=0,
in which case there is no C set and the ordinary CPO-STV tally is all that
matters.
For
very important elections (such as presidential elections), it might make sense
to start with larger debates and then follow with progressively smaller
debates. However, exclusion from the smaller debates should not mean that a
candidate is out of the running.