Condorcet pairs on the ballot

Hugh R. Tobin htobin at
Sat Nov 23 00:44:56 PST 1996

donald at wrote:
>      We should be able to prove that Condorcet Pairs on the Ballot is
> better than Run-off? Do you think this can be proved without using
> rankings? Condorcet Pairs on the Ballot has no rankings.
> Donald,

No doubt others will explain that no separate proof is required, as the
rankings in prior examples are only a shorthand, albeit slightly
constrained, means to cast votes in the pairwise races.  Voters with
transitive preferences among the candidates will vote the same way with
either method, and it is enough to show that a system is superior when
used by such voters.

I respond in order to admit that I was too harsh in my opposition to
Donald's pairwise ballot proposal, and to suggest that it has some merit
as a computerized option to the ranked ballot.  For three-candidate
races, where it is not burdensome, Donald's Condorcet ballot may be
better even without computers.

First, my earlier response omitted an advantage of Donald's proposal,
which I note in another posting today. It allows the partially
indecisive (but not necessarily irrational) voter to avoid the
compulsion inherent in the ranked ballot: that is, if one wants to vote
for A over B on a ranked ballot, one is forced to put C in one of five
positions (above both, equal to A and above B, between A and B, equal to
B and below A, or below both -- the last being the default option, as no
doubt it should be to reflect most voters' intent in leaving a candidate
unranked).  But maybe the voter has nothing at all to say about C. 
Donald's method does not force him to say anything.  The option to leave
a candidate truly unranked against all or a range of candidates also
could be accommodated on a ranked ballot, but the option could not be
quite as simple as omitting to check either of two candidates in a
pairwise box.  

Second, an advantage of Donald's approach is that it removes some of the
mystery from the system.  Seeing the pairwise races the voter would be
more likely to understand the statement, "a candidate who wins against
each other candidate is the winner," and would be less likely to assume
that the outcome would be based on some kind of point system or
weighting of the rankings, as occurs in certain elections to which some
voters may pay more attention (e.g. college football poll rankings).

Third, in more crowded races some voters who do not have a ranking
determined before entering the voting booth might find it easier to
check off pairs on a computer screen than to make the necessary pairwise
comparisons in their heads in order to create numbered rankings.   

Thus, I think it is worth considering not only whether a computer
program could offer the pairwise option without making the ballot too
complex, but also whether that program could make it easier for some
people to construct their rankings using pairs.  

Why not allow the voter, on a computer interface, to create a full or
partial ranking, or none at all, and then (if so desired) click a mouse
to display his vote in each pairwise race?  The voter could then fill in
any missing pairwise votes.  The computer could tell him when he has
filled enough boxes to create a full ranking (transitivity assumed),
could display provisionally further choices implied by transitivity,
could alert to nontransitive choices (or refuse to register them, if the
state is to enforce rationality), and no doubt could have other cool
features (example -- if one has made no ranking, then when one makes the
first pairwise selection one is asked whether one want to choose this
candidate in all races).  The voter could switch to and from the ranking
screen (partially completed) at any time.  The program could make it
hard enough to leave candidates wholly unranked (as opposed to ranked
equally last) so that this would not be done inadvertently, while still
making this possible.  If the voter did not want to see the pairwise
races he could just vote his ranking, or even a single (truncated) vote.
Such a program could be educational, in addition to offering the voters
the most options, both as to the substance of their ballots and the
method of completing them.  Obviously, the program could be widely
distributed so that many voters could become familiar with it before
using it in an actual election, and the ranked paper ballot could still
be allowed for the cyber-challenged voter.

-- Hugh Tobin

More information about the Election-Methods mailing list