Up: Issue 4 Next: Paper 2

Voting matters - Issue 4, August 1995

Progressive Elimination

P Dean

Peter Dean is a Trustee of ERS Ballot Services.

In my previous article [Issue 3, page 6] I took the Solent mock election of 1989 to show that electing 5 candidates from a field of 20 gave a different result to choosing 5 from the last 6.

It occurred to me that a computer used in a progressive elimination (19 from 20, then 18 from 19 and so on) could give a different result. Dr Hill proved this to be the case, though he did not favour this method.

Whereas all systems elected candidate Nos 1, 7, 9 and 18; the normal manual method elected No 2, but electing 5 from the last 6 preferred No 20. The progressive elimination finally elected No 19 with No 14 as the runner-up. An examination of the first 5 preferences on each ballot paper revealed that No 19 came 2nd (60), No 20 - 6th (45), No 14 - 7th (37), and No 2 - 8th (34).

This demonstrates that a candidate with considerable secondary support can easily lose out in such an election. No 19 was originally 9th to be eliminated, and No 14 was 13th to go out, being less than a vote behind his running mate - No 2.

Taking only the top 8 based on the first 5 preferences produced that same result as the progressive elimination process.


Up: Issue 4 Next: Paper 2