Peter Dean is a Trustee of the ERS Ballot Services.
I was surprised to see the existing manual system defended by R J C Fennel as being beyond reproach.
The basic flaw in the manual method is that it allows for the election of candidates receiving less than the quota. This has led to Tasmania requiring at least 7 preferences, combined with a rotated ballot paper since 1973. Even in our own elections some are elected with 4 fewer votes when there are 8 non-transferable votes.
There is a refinement which could easily be introduced in manual elections. This is that when the stage is reached when some candidate(s) fail to reach the quota, a recount takes place with only those remaining taking part. This means that votes previously wasted on candidates with no chance can then influence the result by being allotted to a lower preference for a candidate previously elected. The result will then be demonstrably fair. Taking an actual mock election in the Solent area in 1989 as an example, in which there were 20 candidates for 5 seats. The manual result gave a lead of 4.88 to the last elected - although short of the quota. The Meek system elected the runner-up instead by a margin of 2.01 votes. If a further 5 counts has been added the manual system would have come to a similar result, but by an even larger margin of 7.42. The new result is demonstrably fair - with the last candidate having 2.53 over the quota.
Sometimes the unfair result is even obvious to the public. Such a case occurred in Cork East in 1954. The two Fine Gael candidates received 153 more votes than the two Fianna Fail candidates (1162 non-transferable) yet Fianna Fail won 2 of the 3 seats. Such results discredit the whole system.
The current mechanised system is quite unsuitable for small elections. For instance, with 9 votes and 18 candidates for 3 seats it proceeds to eliminate 7 candidates with 1 vote completely at random. It is quite clear that a different order of exclusion would give a different result. Personally I favour a points method based upon the preferences expressed which would give some form of ranking order to be used instead of the random method.