[EM] correction about (& discussion of myths re) later-no-harm / nursery effect

Warren Smith wds at math.temple.edu
Tue Oct 18 08:52:53 PDT 2005


To change my words slightly:
I do not want voters "always" to have incentive to provide more information,
but I do "usually".  I just do not want to get into a situation where most
voters plurality-vote or provide little info.

I do not think the LNharm condition expresses what I want.  It has
several deficiencies.

1. it unfortunately is absolute rather than "usually".

2. LNharm is predicated on the idea that voters provide must provide info in top-to-bottom
order with truncation at bottom.  I am one voter who, if I were to
provide incomplete info, for sure would not want to do it in that style.
I really am annoyed at the whole idea that votes of that form are a "good thing,"
as opposed to "an annoying artificial creation."

And I think it is important that voters provide info even about the bad or little-known
candidates.  If voters automatically "zero out" those candidates, the result is a failure
of the "nursery effect."  Specifically, with Range Voting it is experimentaly found that
a lot (not all, but a goodly number) of voters do give nonzero scores to a lot
of lesser-known low-scoring third party candidates.  That causes "infant"
small third parties to receive the benefits of voter honesty, enabling the better ones
to grow over time.  In a system like MDDA (probably) or Approval Voting (definitely),
I do not believe that will happen.
(With AV, we know it does not happen - huge vote reduction experimentally
found versus Range Voting for third parties.)  Consequently in these systems there will
be no "nursery" for the "infant" parties and they will suffer huge "infant mortality"
perhaps still leading to 2-party domination and freezing out new ideas from
government.

That could be a very bad thing.

3. LNharm is deceptive: in IRV, voters *DO* have incentive sometimes (probably quite often)
to truncate their ballots DESPITE the fact IRV obeys later-no-harm.
That is because causing harm to your favorite, is not the same thing as
causing harm to the expected election result.

This is not to say that I have a better condition to replace LNharm with.
But I think these are important points both about LNharm and about
systems with truncated ballots.

wds




More information about the Election-Methods mailing list