On Tue, 14 Jul 2009, Simon Lyall wrote:
Following a discussion in the Internetnz list I've
made a little post on
tactical voting that some may find interesting:
Disclaimer: I'm noting a serious expert on voting systems so there might
be an error with my suggestions.
Simon's suggestion is based on the understanding that voters whose first
preference has been elected do no participate in any more voting rounds (all
their other preferences are ignored). This appears to be borne out by the
2008 election for councillors: the first to be elected was Jamie, who had 16
first preferences. After his election, 16 fewer voters participated in the
The description of the voting system at
some vital information. It states, at step 1:
"All 1st preference votes are counted for each candidate".
Nowhere does it mention anything about redistributing lower-ranked
preferences, only that lowest-ranked candidates are "temporarily" removed
The literal reading would lead one to conclude that there is no point voting
for any candidate other than your first preference, and that most certainly
is *not* what preferential voting is supposed to be about.
So one is left to infer that indeed preferences *are* taken into account,
despite the literal wording, but there is nothing to suppose that this works
any differently when candidates are removed on steps 3 and 4.
So I conclude that the methodology as implemented is not what is written in
And furthermore, it's broken, because
(a) it reintroduces "tactical voting", the lack of necessity for which is
supposed to be one of the great strengths of Preferential Voting; and
(b) it under-represents persons who vote for "popular" candidates
(especially during the first round).
I hasten to point out that the converse approach -- of leaving all voters
"in" for all rounds is actually worse in terms of fairness, because it
unfairly favours bloc-voting (a marginal majority can direct the election of
ALL the positions). But at least with such a scheme a voter wouldn't need
foreknowledge of the result in order to cast an optimal vote.
A more reasonable middle-ground system would be the (considerably simpler)
approach of eliminating lowest-polling candidates until there remains only
sufficient candidates to fill the available positions, and then declare them
all to be elected. (Such a system is "reasonably fair" provided that the
ratio of candidates-to-positions is smaller than the ratio of
voters-to-candidates, but does still slightly bias against supporters of
An alternative "fairer" middle-ground system is to eliminate successful
candidates in the current manner, but (a) set the electability threshold T
as Vt/N (where Vt is total of votes for all candidates and N is number of
positions still to be filled); and (b) don't completely discount a
successful candidate's supports' votes in subsequent rounds (as currently),
but rather reduce them by a factor that yields fair treatment of supporters
of popular candidates.
I've done up an explanation at
along with an example of how
this would have applied to last year's council election.