Up: Issue 16 Previous: Paper 2 Next: Paper 4

Voting matters - Issue 16, February 2003

What would a different method have done?

I D Hill

Following an election, the question is often raised of what the result would have been had a different electoral method been used. In general, no reply can be given to this question not only because sufficiently detailed information is not available on the votes, but also because voters can be expected to behave differently if a different system is used.

In comparing one STV system with another, however, rather than totally different systems, it seems unlikely that there would be very much difference in how voters behave, and a reasonable reply is possible provided that the full voting pattern is divulged. It is very welcome that it has been divulged for the three constituencies counted by computer in the recent general election in Eire. Such openness is to be commended. Too often, though, the full voting pattern is regarded as confidential, and the only information is a result sheet, which is quite insufficient for the purpose.

As an example, the question might be whether the result of the 2002 ERS Council election would have been different had the Meek system been used. Working solely from the result sheet (the only information available) I have constructed a voting pattern in which some votes have the character # inserted within their preferences. Before running such data on a computer the # characters have to be replaced, either by a number representing a candidate, or by a space which is then ignored by the STV program.

If the # characters are all replaced by a space, and ERS97 rules used, the actual result sheet is reproduced. If Meek rules are used the same candidates are elected, following a similar order of events.

However, if the # characters are all replaced by the number that represents any one of the defeated candidates, and ERS97 rules used, the same result sheet appears, identical in every particular, but if Meek rules are used, that defeated candidate is elected, at the expense, of course, of one of those who was actually successful.

There is no suggestion that this artificial voting pattern is anything like the true one. I am absolutely sure that it is not, but it is somewhat remarkable that it is possible to devise such a voting pattern with no effect at all on the ERS97 result sheet. The fact that it is possible shows the extent to which the information available is totally inadequate to answer the question. I believe it to be impossible to do the reverse, leaving the Meek result unchanged while varying the ERS97 result.

The artificial voting pattern can be supplied on request.

Up: Issue 16 Previous: Paper 2 Next: Paper 4