[EM] Argument for Approval Primaries

[Bart Ingles]

It occurs to me that one place where ranked ballot methods are entirely
unsuitable is in party primary elections.  Here neither the CW nor the
SU maximizer are necessarily winnable choices, which seems to moot those
criteria.  A similar statement applies to any determinate system.

Of course this is not to say that there is no workable strategy which
could be used with Condorcet, but here you would have to rank candidates
based in part on your estimate of their viability in the general
election.  I doubt that you could expect a voting population to do this
reliably, and in any event once you start incorporating these kinds of
estimates into your ranking strategy, it no longer makes sense to argue
for a system on the grounds that it "allows you to rank sincerely."  A
voter may find the need to betray an ideological favorite to be just as
unpleasant as with Plurality Voting.

In an Approval primary, voters have a face-saving way of supporting the
winnable candidate, although ideal strategy is slightly different from
what it would be in a general election.  In a primary, a not-too-extreme
voter who would have insincerely voted for a winnable moderate should
probably continue to bullet vote for this candidate, at least so long as
it is palatable to do so.  The reason is that only one candidate can
advance to the general election-- once you have decided on a winnable
choice, you don't want your "favorite" to screw things up.

A more extreme voter, who could not have brought himself to betray his
ideological favorite in a plurality election, may find it palatable to
approve both candidates.  Approval gives this voter "an out"-- he can
still say that he voted for his ideological choice.

Of course, the more theoretically sound solution is to do away with
primaries entirely, but this may not be likely any time soon.  And there
are some fairly good pragmatic arguments for keeping primaries in place,
which I'll leave to others to debate for now.

Bart



Bart Ingles wrote:
>
> Anyway, it seems to me that there are probably situations where either
> of Approval or Condorcet are better suited than the other.  To me it
> makes the most sense to start with the simpler of the two, and introduce
> the other wherever the need has been actually demonstrated.