[EM] IRV in action

[James Gilmour]

> On Sun, 30 Mar 2003 22:38:15 -0800 Rob Lanphier wrote:
> > Very interesting.  It's a shame they didn't post the individual ballots
> > in a way that a Condorcet tally could be run.

Yes, it would have been interesting (to us), but this was the election Result
Sheet, so such information was irrelevant and hence, not published.

> > This looks like a really strange example of an election.  The
> > candidate/voter ratio was extremely high, and presumably, the candidates
> > had an opportunity to establish a somewhat personal relationship with
> > all of the voters (especially the incumbent pre-reform peers, who they
> > presumably worked with in the "good old days").  Thus, it's not hard to
> > imagine why the rankings were so deep compared with what you might
> > expect in another type of election.

Yes, it was a VERY strange election - for all the reasons you state.
In my experience of counting IRV and STV-PR elections, I have found that the depth
of rankings does vary very greatly among voters, even when the number of
candidates is small.  Unless we ask them, we cannot know the reasons why such
voters do not use all their preferences, but I suspect there are several different
reasons.  Repeated exposure to X-voting in FPTP elections may have some effect on
UK electors.

> > It's also interesting how they chose to whittle down the number required
> > to win as ballots dropped off.  There were 423 ballots, where the first
> > column lists "Votes needed to be elected: 212".  However, by the time
> > you reach column 42, the votes needed to be elected drops to 134.  Is
> > this standard practice in IRV, or was this done to avoid the
> > embarrassing situation where no candidate in the runoff receives a
> > majority, even after receiving 42 rounds of ballot transfers?

I don't know why this was done in this case, but I suspect it was simply intended
to be helpful to those reading the Result Sheet.

It is certainly not part of standard practice under the IRV Rules I codified for
the Electoral Reform Society in 1978.  Those Rules state, for first preferences:
"If the vote for any one candidate equals or exceeds the votes of all the other
candidates combined, that candidate shall be declared elected."
and for subsequent stages of the count:
"This process [exclusion - described in the preceding clauses] shall be continued
until the vote for some one candidate exceeds or equals the votes for all other
continuing candidates combined, and that candidate shall then be declared
elected."


> > I suppose this is an interesting problem for IRV in general.  I'd always
> > been under the impression that a big selling point of IRV is that the
> > winner would have "a majority", even if it's from a series of
> > transfers.  However, there's no guarantee that there will be a majority
> > even after all of the transferring is done.

As Dave pointed out in his comment below, you have to be careful with the word
"majority". It has also been discussed more extensively in other posts to this
list.

But the thinking behind the use of the word in this context is also instructive.
You have to remember that IRV is nothing more than a convenient method of
condensing an exhaustive ballot into one voting operation.  (It also avoids all
the horse-trading that typically takes place between the successive rounds of an
exhaustive ballot, but that's a different issue.)  Voters who drop out (stop
marking preferences) are saying to the Returning Officer: "If it comes to a choice
among the remaining candidates, I don't care who wins".  That is a free and
democratic decision of those voters and we must take care not to imply otherwise
in the language we use to interpret the election results.

> > In this particular
> > election, there's no knowing how many of the 115 ballots accumulated by
> > Montgomery of Alamein (the runner up) in the 42nd round actually listed
> > Ullswater (the winner) next (who won with 151 votes).  If less than 61
> > ballots (i.e. 212 minus 151) of the 115 Montgomery of Alamein ballots
> > had listed Ullswater, then there was indeed no majority winner.

> Dave Ketchum replied
> Here you need to understand the language IRV uses - they LIKE the word
> "majority", treading lightly on the fact that they are doing a majority of
> the ballots that remain to be used to determine a winner, and not a
> majority of total ballots.

See my comment above.  Whatever other defects it may have, IRV does ensure that
the winner has the support of half or more of those who are voting at the point
when the final decision is made.  If some voters choose to drop out before the end
point is reached, that is their choice.  No one should claim more for IRV than can
be justified, but equally those who don't like IRV should not misrepresent the
results.

> BTW - "majority" is a word that often needs qualification, such as "most
> of those who voted" or "most of the members".

"Majority" also needs further qualification when it is used with regard to
election results.  Here in the UK, the word has been distorted from its original
meaning ("more than half") to mean, in elections, "the winner's lead over the
second-placed candidate".  This distortion of the language goes well with our
placid acceptance of the distortions of FPTP in single-member districts!!


> With dozens of candidates (required to give 42 rounds something to work
> on), no voter should be required to rank every candidate, for it is
> unreasonable for voters to be expected to intelligently rank so many.
>
> Now, assuming A and B are liked better than the many Ms, and that some
> voters consider A and B to be less acceptable than any of the Ms they are
> prepared to rank, there will be ballots that are exhausted before we get
> to A vs B.  So the "majority" will simply be based on comparing A vs B,
> and be more than half of the ballots that included one or both of them in
> a voter's ranking.
>
> Oops, what I wrote above is not exactly right, but I choose to keep it in
> and correct it here:  Certainly ballots do get exhausted, so the majority
> to win is likely less than half the total ballots - it just has to be more
> than half the non-exhausted ballots.  So, if A has such a majority, then B
> and the remaining Ms, together, must have less, meaning that whatever
> might happen to the remaining Ms, A will still have a majority - and
> therefore the election can be terminated at this point without determining
> how decisively A won over B.
>
> There was interest above as to whether some Montgomery votes might have
> listed Ullswater as a lower choice.  I see no value in this as these
> voters voted against Ullswater by listing Montgomery first.
>
> True IRV weakness that I see:
>       44 candidates got some first place votes and can, possibly, win -
>           even with two to get started with (candidates with only one
> could list them next).

I don't understand this comment.  What would the candidates list?

>       Some of the above may have got only one - here the IRV counters
> decide which of these die now and which, if any, get a chance to advance.

What else can you do when you encounter a tie?  In the House of Lords by-election
election, the Returning Officer would appear to have made the decision to exclude
the tied candidates in alphabetical order.  Our regulations for public elections
require the casting of lots to determine which of the tied candidates should be
excluded.

>       37 got no first place votes.  These all die instantly, even if one
> of them got 400 second place votes - this happens because, whenever some
> of their second place votes get exposed, those will be discarded as lowest
> count among the remaining votes.

Are you seriously suggesting that electors would accept as the "winner" any
candidate who got NO first preferences  but lots of second preferences?  No matter
what social choice theory might say, I cannot see such a candidate ever being
accepted in a real election for public office.

James