lved is enormous. Yet this was the plan
first advocated by Professor Nanson, who wrote, in reply to a definite
inquiry how the Gregory principle was applied:--"I explain by an
example. A has 2,000 votes, the quota being 1,000. A then requires only
half the value of each vote cast for him. Each paper cast for him is
then stamped as having lost one-half of its value, and the whole of A's
papers are then transferred with diminished value to the second name
(unelected, of course). The same principle applies all through. Whenever
anyone has a surplus all the papers are passed to the next man with
diminished value." Now, the effect of this extraordinary proposal would
be that the whole of the papers would have to be kept in circulation
till the last candidate was elected, with diminishing compound
fractional values. In a ten-seat electorate a large proportion would
pass through several transfers, and would towards the end of the count
have such a ridiculously small fractional value that it would take
several millions of the ballot-papers to make a single vote! It is no
wonder that this method was abandoned when the complications to which
it would lead were realized.

A simple method of avoiding this complexity would be to treat
transferred surplus papers as if the preferences were exhausted. It must
be remembered that in all transfers a certain number of papers are lost
owing to the preferences being exhausted, and the additional loss would
be small. Thus at the first Hobart election 206 votes were wasted, and
this number would have been increased by two only. Every surplus would
then be transferred by reference to the next choice, wherever expressed,
on both original papers and papers transferred from excluded candidates.

It might be provided, however, for greater accuracy that all papers
contributing to surpluses on the first count only should be transferred
in packets. Should these contribute to a new surplus, it should be
divided into two parts, proportional to (1) original votes and votes
transferred from excluded candidates, and (2) the value of the packet in
votes. Each part would then be distributed proportionally to the next
available preferences wherever expressed. To divide the packets into
sub-packets is a useless complication. The loss involved in neglecting
them would usually be less than one-thousandth part of the loss due to
exhausted papers.

Having now dealt with the main features of the different variati