[EM] Vote Management

[James Gilmour]

Olli asked (10 April 2003):
> > Could you please give us an example how it works? As far as I
> > understand STV this is impossible. As long as you vote in a solid
> > block (having the candidates of your party above all other
> > candidates) you get the seats you are going to get. It's a different
> > thing if the supporters of the other parties vote in a less
> > disciplined way.
> >
> > If you average the votes, your candidates may reach the quota without
> > many transfers. If you don't, the votes will just get transferred
> > until all the candidates you can elect have reached the quota. If you
> > don't cast secondary preferences you can't help other candidates of
> > your party and they may be unsuccessful. If you cast preferences only
> > for the candidates you expect to elect, your opponents may be elected
> > with less than the quota, which is not serious, you've only decided
> > not to have an influence in the order they are elected.

Markus replied:
> In an STV count, a candidate can win additional votes only as long
> as he is neither elected nor eliminated. Therefore, when a given
> party wants to win as many votes as possible during the STV count,
> this party has to take care that its candidates are as long as
> possible neither elected nor eliminated. The party achieves this
> by "averaging" the first preferences over its candidates so that
> each candidate is just below the Droop Quota.

Markus is absolutely right.  Imagine a party (party "A") that expects to get
one-and-one-half quotas of first preferences and is popular enough to pick up
substantial numbers of transferred preferences from supporters of other parties at
later stages in the count.  So the party is sure of winning one seat, but could be
in line for two.  The party puts up two candidates.  If most of the preferences go
to the more popular candidate, that candidate is elected immediately and the
surplus is transferred to the running mate (assuming strong party loyalty among
the supporters).  But the small number of votes the running mate then has may not
be enough to prevent the exclusion of that candidate at an early subsequent stage.
So there is no continuing candidate of party "A" to receive any transfers when
such transfers become available as other candidates are excluded.  So party "A"
wins only one seat.

Now suppose party "A" succeeds in persuading its supporters to average their first
preferences over the two candidates.  Neither candidate is elected on first
preferences, but both have enough votes to keep them in the running while others
below them are successively excluded.  So both can pick up transferred votes and,
if there enough transfers, both are elected.  The parties in Northern Ireland
certainly work hard to put this into practice where the voting patterns suggest it
will be worthwhile.

James