52 Names need no Elimination

[Steve Eppley]

Donald opined:
>It is not proper to say there are too many choices - unless you are
>buying ice cream. Candidates are not ice cream. You have no right to
>just eliminate them because you cannot handle their numbers.

As I pointed out, the labor to manage the voting not only includes 
the pairwise tallying if it's a pairwise method, but also the rest of 
the data entry and summarizing, etc.  Donald makes a valid point that
IRO takes less tallying labor than Condorcet, but even if the method
is IRO I'm pretty sure the group will opt to first eliminate some of 
the choices to make it less labor-intensive.

Though someone (Tom? Marcus?) recently posted something in EM which
stated that Robert's Rules don't mention Instant Runoff, I found in 
a book on Robert's Rules a clear definition of a Preference Voting
ranked ballot alternative to the "vote for only one" method.  The 
procedure described was Instant Runoff (although it didn't specify 
what to do if a ballot ranked two or more choices as equals; perhaps
there was an implication that equal rankings shouldn't be allowed).

The iterative procedure was to arrange the ballots in separate piles
according to the highest ranked of the remaining candidates on each
ballot.  The candidate of the smallest pile is eliminated and that 
pile's ballots are redistributed to other piles according to the next 
ranked choice.  (Hopefully the variations in ballot thickness are 
negligible; otherwise some counting would be necessary when piles 
are of similar height.)  Without a computer, the attractiveness of
that procedure, compared to an exhaustive pairwise tally, is 
undeniable.

I'd like to hear a rational argument, though, why IRO would produce
a more democratic outcome than something like the two-ballot method
N_most_approved///Condorcet or the one ballot method Approval.  But
I won't hold my breath for a rational argument from Donald, who will
never, ever attempt to explain why it's right to eliminate candidates
IRO's way but not other ways.

I'm posting a separate message about a possible way to speed up 
manual Condorcet tallies.

>It is understandable that this condorcet person wants to eliminate
>candidate names. 
-snip-

It wasn't my idea to eliminate names, which is why I'm sure they'd 
want to eliminate names even if IRO or Approval were used.  But yes,
if I had to tally a few hundred ranked ballots with a lot of choices,
and especially if the voters average many choices per ballot, I'd 
balk at doing it manually.  This of course is no argument in favor of
IRO if computers are available to do the tallying, as there will be 
in the public elections which ER is interested in.

>Each voter would have to make 51 selections in order to keep the
>mathematics honest with condorcet - the voters are not going to do
>that.

Disorderly Discussion Alert: Repetition of an assertion which has
been rebutted, without ever rebutting the rebuttal.

>This is one time that Steve should put aside his dislike of Instant
>Run-off and use it because it is best able to handle these 52 names
>"..without compromising the integrity of the results".

Why is it democratically better than plain Approval, which takes even
less labor than IRO?  (Oh, of course: because Donald says it is.)

>Steve did not say how many people would be voting in this election.
>If the number is to be less than one hundred we can expect a lot of
>ties. How these ties are handled will be important. Instant Run-off
>can handle the ties.
>
>Example: Suppose the election results had a string of twenty
>candidate names at the end with one vote each. These twenty are all
>tied. 

It think the ties worth breaking are the ties for first place, not 
the ties for last place.  Perhaps Donald is referring to ties when 
only one choice per voter is counted, as in Plurality?


Now - we
>could drop the entire twenty if the one candidate name before this string
>had more than twenty votes - but this condition may not exist.
>
>So - we are going to need to look at the second selections of only these
>twenty candidate names in order to decide which of these twenty we are
>going to drop. The rule is that we drop the candidate name that received
>the lowest amount of votes from only the second selections of only these
>twenty. If more then one received the same low number - we drop all the
>ones with this same low number. (If in the event there is no lowest we go
>look at the third set of selections of only the tied candidates.)
>
>The ones that are dropped are to have their votes reassigned to their
>second selections. We now drop the lowest candidate name and reassign his
>votes to the next selections. Or - if we again have a tied condition at the
>end, we repeat the above routine. We continue to solve ties and/or drop
>candidate names until we have a winner.
>
>No problem!
>
>Hopefully the winner will have a majority of the votes cast - a rule of
>Instant Run-off. We should ask the voters to make at least ten selections
>in order to avoid a follow up election.
>
>Donald,
>
>
>
>
>

---Steve     (Steve Eppley    seppley at alumni.caltech.edu)