[EM] Smallest Change for Greatest Improvement in Single-Seat Elections

[Bart Ingles]

Steve Eppley wrote:
> 
> Bart wrote:
> > My suggestion would be for a limited form of Approval Voting,
> > where the voter is allowed to vote for at most two candidates.
> > This could be called "Approval-With-Training-Wheels", and would
> > help mitigate any concerns that voters wouldn't understand the
> > ideal strategy and would vote willy-nilly for anyone they consider
> > halfway acceptable.
> 
> I don't see why Bart believes voters will understand the
> "ideal" strategy in that system, either.  A majority need to

I didn't say limited voting would be easier to understand, I just said
it would limit the effect of voters going crazy and voting for
candidates they don't need.  Of course, if would limit people voting for
candidates they need as well, but I believe two votes would be
sufficient for the majority of elections.

> vote for the needed compromise or something close to it,
> which takes strategizing and requires pre-election polling
> data.  The more alternatives compete, the more difficult it

A majority wouldn't nessarily need to vote for a compromise.  In
elections like most of our general elections, only 10-15% of the voters
would need to compromise, unless the political climate changes
drastically.

> will be to elect the best compromise, so alternatives will be
> deterred from competing unless they want to be spoilers.

True, with more alternatives the two-vote system will be inadequate, but
still better than plurality.
> 
> If the Bucklin system was dropped from use because voters,
> allowed to rank candidates, tended to rank only one (for fear
> that ranking a second choice would defeat their favorite)
> wouldn't any variation of Approval have an even worse problem?

I would like to see a complete account of why the Bucklin system was
dropped.  If voters ranking only one was the problem, was it replaced
with a ranked method that encouraged more complete ranking?
> 
> In Bucklin, the voters' second choices are counted too when no
> candidate receives a majority of first choices.  (Bucklin is
> an iterative method.  If no candidate has a majority when both
> first and second choices are counted, then third choices are
> counted too, etc.)  With Bucklin it's pretty easy for voters
> to see how they can defeat their favorite by ranking a second
> choice, and it will be easy for them to see this with Approval.

If voters are that unwilling to compromise, do you think that Condorcet
would fare any better?  Lower choices have a chance of causing the
favorite's defeat under that system as well.
> 
> So I think Bart has identified the wrong problem with Approval
> when he writes of a concern that voters will select too many.
> The real concern is that they will fail to include the best
> compromise(s) in their selection(s).

In that case, there is no real downside to Approval, compared with
Plurality.  Since it works with any voting equipment, is there any
reason not to support it, at least until suitable equipment is
available?  If voters will be as unwilling to compromise as you expect,
there should be no problem demonstrating the need for a more complex
system even after Approval is in place.

> 
> * *
> 
> It's easy for voters to miscalculate with Approval systems.
> Suppose there are three alternatives, and alternative C is
> the *last* choice of a majority.  But the majority is split
> on its first choice:
> 
>    30%:  A > B >> C
>    25%:  B > A >> C
>    45%:  C >> B > A
> 
>    I've used the symbol '>>' to indicate where voters might
>    be predisposed to draw the line between "approving" or
>    "disapproving."  This doesn't consider strategic voting.

As I have been saying for some time, ranked examples are not sufficient
to show what the best strategy would be under Approval.  You need
information about how similar the candidates are to one-another as well
(in other words, a rated scenario).  If the  A and B voters have a
strong dislike for one-another, and consider the other almost as bad as
C, there is not much incentive to compromise.  Why is it so important to
elect a candidate that 75% of the population dislikes?

> 
> Unless they have enough pre-election polling data and
> organization to strategize, some of the majority may fail to
> select a second choice for fear of unnecessarily defeating
> their favorite.  When that happens, both A and B can lose.

The need for polling data is even more critical under Plurality, yet we
all seem to know who to vote for.

> Also, even when the majority succeeds in defeating C, it's
> clear from the above that though B has more support than A,
> B may have a slim chance of being elected, unless the C voters
> have the ability and polling data to strategize (select B).

Again, it depends on whether the voters believe that B is an acceptable
compromise.  An example showing only rank positions doesn't provide
enough information to tell us this.

> 
> So I feel Bart is wrong to claim that an Approvalish system
> will produce great improvement.  Allowing voters to rank
> the alternatives is not a drastic procedural change, in

Not a drastic procedural change?  It's not even possible under most
voting equipment.
--Bart

> my opinion, and is justified by its greater improvement:
> It would permit more candidates to compete without fear of
> spoiling, and would elect centrist compromises without
> anybody needing to strategize.
> 
> ---Steve     (Steve Eppley    seppley at alumni.caltech.edu)