Truncating Condorcet's Ballots

[Blake Cretney]

Donald E Davison wrote:
> Greetings list,
> 
>      This question of truncated ballots can best be answered if we had
> copies of some real ballots.
> 
>      I do not know of any election in which Condorcet is used, but some on
> this list have mentioned elections in which Condorcet was being used.
>      If one of those persons was able to obtain copies of ballots from a
> series of Condorcet elections, we would have much to study and maybe the
> question would be answered.

I think that the best common example is Robert's Rules of Order.

Assume a bill is before the house, and a sigle ammendment has been
suggested (or is the only one outstanding).

There are now 3 alternatives before the house
A-No change
B-Ammended bill
C-Original bill

If the ammendment makes the bill more moderate, the true preference
might be:
40 A>B>C
12 B>C>A
13 B>A>C
35 C>B>A

Of course, in this example, IRV chooses A, although B is the
Condorcet winner.  

If all voter vote sincerely in Robert's rules of Order, a majority
will vote for the ammendment over the original bill:
B>C 65-35
Then, a majority will vote for the ammended bill over no change
B>A 60-40
So, the Condorcet winner wins.

Notice, however, that a majority will favour the ammended bill, but
not the original.  That is, a majority wants the moderated bill but is
not willing to go as far as the original.  So, the A-1st voters could
use the strategy of voting against the ammendment, in the hopes that
this will prevent any bill from being passed.  This is equivalent to
falsifying a preference of A>C>B instead of A>B>C.

Then,
40 A>C>B
12 B>C>A
13 B>A>C
35 C>B>A

Now, the ammendment fails
C>B 75-25
And so does the original bill
A>C 52-48

Voters could also choose to abstain from the vote on the ammendment. 
If they are actually in favour of the ammendment, this could cause the
ammendment to fail, resulting in the failure of the bill as a whole. 
Of course, this isn't as effective as actually voting against the
ammendment.  This is equivalent to falsifying a preference of A>B=C.

40 A
12 B>C>A
13 B>A>C
35 C>B>A

In this case it works, the ammendment fails
C>B 35-25
The original fails
A>C 53-48

Of course, just as in all Condorcet methods, the A-1st voters are
taking the risk that their worst outcome (C) may be caused to win.

So, my point is that exactly the same strategy problems that you
criticize in Condorcet are present in Robert's rules of Order, as used
by many clubs and organizations including the American Congress.  In
fact, they are arguably worse, as the retaliatory strategies that have
been mentioned on this list do not work.  However, I don't see much
evidence of these strategy problems resulting in massive strategic
abstentions (abstentions usually come from convenience or divided
loyalty).  There certainly is no movement to replace this method with
IRV in the hopes of reducing strategy problems.  In fact, as I'm sure
you're aware, IRV has strategy problems of its own.  

---
Blake Cretney
See the EM Resource:  http://www.fortunecity.com/meltingpot/harrow/124