Voting matters - Issue 6, May 1996

A form of STV with single-member constituencies

B A Wichmann

One over-riding concern that appears in the Plant report is the desire to retain single-member constituencies in any reform of the electoral system for the House of Commons. A natural question is if STV can be adapted in some way to retain single-member constituencies, but avoiding the non-proportionality of the Alternative Vote (AV). This paper presents such an adaptation.

The basic idea is to use a two-tier system in a similar manner to the German system by having single-member constituencies augmented by members elected in a more proportional manner. The second tier is a group of constituencies which, for convenience, we call a county. The electors provide two 'votes' by giving the usual preferences to the candidates in their constituency and then also providing a preference vote to all the candidates in their county.

The election proceeds in two stages, firstly each constituency is considered individually using STV (which degenerates to AV). However those votes which have not been used to elect a candidate here are forwarded to the county vote (or second stage). The county vote first eliminates those candidates already elected at the first stage, and then uses STV to fill the county seats.

The main parameters of this voting system are the number of single-member constituencies used to form a county, and the number of seats available at the county level. It appears that about 5 (or more) constituencies should be grouped into a county in order to provide reasonable proportionality and that the number of county seats should be not less than 2 for the same reason.

This system is quite different from conventional STV for a number of reasons:

1. This system, like FPTP has safe seats, whereas STV has no such equivalent. For instance, in the Irish elections, almost every constituency has a Fianna Fail or Fine Gael candidate who is not elected. My reason for concluding this is that I believe that the main parties, even for safe seats, would not propose more than one candidate since this would appear to present a divided party.
2. The elector's ability to select within a party is restricted. If you are a Conservative party supporter in a safe Conservative constituency with a male candidate, you could not select a woman candidate (given the restriction noted above of a single candidate). On the other hand, if you were in a Labour constituency, your vote would be wasted, allowing you to select a woman candidate from the county list as your first preference.
3. Minority interests would be represented at the county level. These interests would be accumulated as wasted votes and hence would have a good chance of representation, depending upon the number of county seats.

There are some technical details to resolve. I have based my proposal on the use of the Meek algorithm for STV, although this is not strictly essential. However, it is clearly important to compute the fraction of each vote which is wasted (from the first stage) in order to conduct the second stage. This is straightforward since for each voter who contributes to the elected candidate, the percentage wasted is simply the percentage of votes above the quota. This implies that about ½ of the votes would go forward to the second stage. This might imply that about half of the seats should be at county level, but a smaller number is probably satisfactory.

My belief is that this proposal would be quite easy to implement, at least using the Meek algorithm. However, since we have no similar system, it does not seem possible to construct realistic data with which to do any serious study of its suitability.