Let's found an organization to oppose IRV

[Bart Ingles]

LAYTON Craig wrote:
> 
> I'm definately beginning to doubt the desirability of approval (vis a vis
> IRV) as well.  Despite the apparent simplicity of the system, electoral
> strategies, polling etc becomes much more of a problem.  Consider the effect
> of successive polls in a Bush-Nader-Gore approval election (I'm giving Nader
> extra support, and assuming that more Gore supporters prefer Nader over Bush
> than the reverse);
> 
> 1) Initial polling, and alot of media about the approval system encourages
> most voters to vote for at least one of these three candidates (ie, Buchanan
> voters mostly decide to vote for Bush as well).  Bush 40%, Gore 45%, Nader
> 35%.
> 
> 2) Nader voters decide to withdraw their support for Gore, because they
> appear to have a fairly good chance of winning.  Bush 40%, Gore 35%, Nader
> 35%.
> 
> 3) Nader and Gore voters realise that Bush will win if they don't multiple
> vote.  To improve their outcome, a substantial block of Nader and Gore
> voters trade preferences.  Bush 40%, Gore 45%, Nader 45%.
> 
> What is the optimal strategy now? It would seem (if it were possible to
> organise) that the best strategy for Gore's campaign is to pledge to support
> Nader in a recipricol arrangement, but instruct voters not to vote for Nader
> at the last minute (ie how to vote cards kept confidential until they are
> actually being handed out on the morning of the election, showing only a x
> next to Gore).  I don't suppose how to vote cards are particularly important
> with plurality, but once you introduce any other method, they become
> crucial.  Maybe the Gore and Nader campaigns get into a big fight & withdraw
> all vote sharing arrangemets (or the voters simply don't adhere to them
> because they're mad that the other party's stubborness has handed Bush the
> election).
> 
> The point is not so much that polling can be strategically abused by
> approval, but rather that the outcome of approval voting is so uncertain
> (any one could win, and this may or may not have any correlation to people's
> actual wishes).  If there was the same contest with IRV, Gore and Nader
> voters can vote for their own candidates first (ie vote sincerely) and
> maximise their outcome at the same time.
> 
> I certainly don't advocate IRV, and I'm not even saying that it is superior
> to approval, but I question whether it is so much worse.


Realistic elections would have some variation in voter preference
levels, which would tend to dampen out the strategy shifts like those
described below.  In other words, not all voters would be likely to
shift strategy at the same time.

But even if you did have three monolithic voting blocs, you have the
same problem with any method.  The main difference is that with approval
voting, it tends to occur when voters are ambivalent about the
compromise candidate (if the voter is strongly for or against the
compromise, polling data won't have much effect).  With ranked systems,
the strategy shifts are more likely when voters are stonily for or
against all candidates.

Condorcet example:
1) Polls show a low social utility [?] candidate as the likely Condorcet
winner:

45  A(10)    B(2)     C(0)
11  B(10)           A,C(0)
44  C(10)    B(2)     A(0)


2) The A and C voters realize that they can improve utility expectations
by eliminating B, placing them in a statistical dead heat with
one-another.  They plan to bullet-vote, choosing the AC lottery over B.

45  A
11  B
44  C


3) Further polling data shows that A is likely to win after all.  The C
voters react by ranking B.  A has no particular reason to vote sincerely
under a wv system (and scant reason under margins):

45  A
11  B
44  C, B


IRV has some bizarre strategies as well -- if there is any doubt about
Nader's ability to defeat Bush, then the Nader voters should
strategically rank Gore first.  Unless of course they think that Gore is
not much better than Bush, in which case they should threaten to bullet
vote (or even to rank Gore below Bush) in an attempt to coerce the Gore
voters into strategically ranking Nader first.

Changing strategy in response to polling data is not unique to approval
voting.

Bart