STV for the Senedd ... Parliament for Wales' submission

One of the comments on last week's posts by Penddu and myself mentioned that the Parliament for Wales campaign group submitted very similar evidence to the Richard Commission in 2003. As people tend not to read comments after a few days, I thought it would be best to address this in a new post.

PfW's formal submission to the Commission and a transcript of their oral evidence is here:

     Formal Submission
     Oral evidence and Questioning

This is their PowerPoint presentation, together with a précis of what it says:


1. The first past the post system produces a noticeably lower turnout in seats where the outcome is not in doubt. These safe seats, nearly always Labour held, had on average a turnout of about 34%, whereas the turnout for seats where there was a close contest was 44%. This shows that many people do not vote if they think their vote will make no difference to the result, and it is therefore reasonable to conclude that turnout would increase by adopting STV.

2. The the difference between the percentage of votes cast and the percentage of seats won is called "disproportionality". Disproportionality would be reduced under the existing Additional Member System if the number of regional seats were increased relative to FPTP seats. Under the current system the disproportionality (in 2003) was about 10%, it could be reasonably be reduced to 5% but would never be completely eliminated because very small parties do not get any seats.

3. They then presented five different options for the number of multi-member constituencies in an STV system (some based on the existing Westminster constituencies and some on Local Authority boundaries) and compared them with AMS for proportionality. A 15 constituency division (the lowest number they analysed) came out best, but all STV divisions were better than AMS.

In terms of the division into multi-member constituencies, I would particularly note the point that fewer, slightly larger, constituencies tends to produce more proportionality. Therefore the 14 districts that Penddu suggested is going to produce better proportionality, though one or two more or fewer wouldn't make such a huge difference. Looking at all the options presented by PfW, I do lean more towards the Local Authority than Westminster constituency way of splitting things, and still like what Penddu has come up with.


As the person who gave the links commented, it does seem that there is a good deal of consensus about how STV might operate. The bigger question is to get it adopted at all. So in the remainder of this post I'm going to focus on one aspect of the various detailed proposals where I believe my view may be a little at odds with the consensus ... namely that I am in favour of slightly larger constituencies. The consensus probably stretches to six, and PfW said that more than six doesn't work so well ... though that conclusion does seem to be based on what works for Northern Ireland. So I'm going to make the case for the average to be six, but for individual constituencies to vary from four to eight. Perhaps others would like to give me some feedback.


There are two major arguments against larger constituencies: the first is that geographical size might make it difficult for somebody at one end of the constituency to think they had anything in common with people at the other end. The second is that a greater number of seats gives smaller, and sometimes undesirable, parties the opportunity to get elected.

Geographical size

If we look at the various splits suggested, the geographically large constituencies are Gwynedd/Môn, Powys and Ceredigion/Pembrokeshire. However none of these has a large number of members. These three districts are all at the bottom end of the range with 4 or 5 seats in an 80 member Senedd (3 or 4 in a 60 member Senedd).

The crux of the issue is that the districts do not all have to have the same number of seats. It is only if we insist on a model in which all districts return an equal number that we get a problem.

What constitutes too many seats?

The balance to be drawn in any system of proportional representation is between proportionality and an individual member's relationship to the people s/he represents.

However, systems with voting units that return a large number of members tend to operate a threshold, to prevent the very real problem of small, unrepresentative parties getting a seat in the legislature. To give two examples, both Germany and the London Assembly set the threshold at 5%.

In a constituency system there is no defined threshold, but the number of members does tend to set its own threshold. The largest constituency I proposed would return eight members, so it is worth looking at a parallel. The NW England European constituency returns eight members, and this was the result in June this year:

North West England, 2009

Conservative ... 25.6% ... 3 seats
Labour ... 20.4% ... 2 seats
UKIP ... 15.8% ... 1 seat
LibDem ... 14.3% ... 1 seat
BNP ... 8.0% ... 1 seat
Green Party ... 7.7% ... 0 seats
Others (total) ... 8.1% ... 0 seats

The final seat was won with 8% of the vote. Much of the argument about larger constituencies does not centre so much on the principle, but on the specific reaction to the possibility of a particular party (in this case the BNP) being elected. People tend to argue against larger constituencies not so much on the matter of principle, but because in practice a party like the BNP might be elected.

This, of course, is a very tricky issue. In terms of democracy we want all parties to be treated equally, no matter how obnoxious we think one of them might be. So we must find a way of making sure that all parties are treated scrupulously fairly, but at the same time deal with the problem of parties that polarize opinion such as the BNP. However the answer is not the size of the constituency per se, but the electoral system itself.

Going back to the NW England constituency, this video gives a very neat explanation of how the current voting system (it applies to the regional element of the Assembly elections as well as the European elections) works:


In essence, each person only has one vote, and must choose one party (or person) and one party only. However this does not equate to the way that people actually think. I would suggest that each of us has a range of views something like:

• what I really want
• what I'd be quite happy with
• what I'd put up with
• what I'd be unhappy with
• what I definitely don't want

In my opinion a fair voting system must reflect that range of opinion. However as the author of BlogMenai said here, adopting STV deals very effectively with the problem of the BNP. But it does this not by arbitrarily singling out the BNP and making it a special case, but by relying on the fact that although 8% of voters in NW England voted for the BNP, the big majority of people who didn't vote BNP would put the BNP in the "what I definitely don't want" category. I would venture to suggest that very few people would put other parties in that category, so that STV would almost certainly have resulted in the Greens picking up second, third or fourth choice votes, and therefore getting the final seat ahead of the BNP.


Finally, on the subject of large constituencies, we effectively already have constituencies of between 11 and 13 members in Wales. North Wales comprises 9 FPTP and 4 additional regional seats. In 2007 the final seat was won with 6.64% of the regional vote. The BNP got 5.37%. In other words it would not take much for the BNP to get a seat in North Wales under AMS, but if we adopted STV it would be harder for them to get a seat because the majority of voters would not give them any subsequent choice votes.

Bookmark and Share


Cai Larsen said...

It's certainly true that an extremist party would find it more difficult to get a candidate elected under STV than under the present system (unless you had a very large number of candidates in a single constituency).

Although I like STV & would want to see it adopted in Wales, the idea of accepting or rejecting a system because it's likely to exclude certain parties (however unpleasant) is, in my view, problematic.

Unknown said...

While I do not condone the BNP or similar racist parties, you have to accept that if they can get sufficient votes they deserve to win seats Rememeber it is not that long ago that Plaid was in this position with a national vote share in single figures.

On the subject of constituency (district??) size, as well as considering limiting the size for distance considerations - there is also the question of homogenity - larger seats would mix up rural and urban communities with little in common, whereas smaller districts should be able to be more reflective of local needs.

MH said...

I must admit to being a bit surprised by both these comments. Perhaps I didn't make myself clear. I was trying to say that the BNP must be treated in exactly the same way as other parties ... and that STV would do this.

I was trying to counter an argument that I have heard from many who are against proportional representation; namely that large multi-member constituencies make it possible for small parties such as the BNP to be elected, and that they are therefore a bad thing.

STV has the advantage over other proportional voting systems because it allows voter to set out an order of preference. The ability to express a ranked preference means that a party which polarizes opinion to the extent that a majority would say "over my dead body" will be less likely to be elected.

Unknown said...

MH - My apologies if I misunderstood your point. I agree that BNP are less likely than other parties to pick up second preference votes.

Incidentally, with 4 member districts you would be guaranteed a seat with 20% of the vote, but would have a reasonable expectation of a seat with say 16% of the vote and a reliance on picking up second preference votes from smaller eliminated parties.

In a 6 member seat you would need 14.3% of the vote but could get a seat on say 12%.

I would forsee that in a STV system there would be a proliferation of smaller fringe parties spinning off from the main parties, who would vote for their first choice fring party but return their second preference vote to the 'parent' party.

So we could see say a Cymuned party and a Welsh Republican party both standing against Plaid in the initial vote, but then giving the second prefernce votes back to Plaid.

Fascinating for anoraks like myself!!!

James Dowden said...

There are various quirks that make STV undesirable in my opinion:

1) STV has in practice resulted in parties not putting up a full slate of candidates for fear of knocking themselves out. This problem gets worse as you increase the number of members per district.

2) STV forces people to make non-existent preference judgments. Say that you had enough political knowledge to rank four Plaid Cymru candidates in order, and wished to make your next four preferences the four Liberal Democrat candidates, but had no way of distinguishing between them. Ranking them all as "5" would spoil your ballot at that point and fail to prefer them to the Labour, Conservative, Green, UKIP, BNP, and Monster Raving Loony candidates. Properly numbering them 5-8 would conceal the opinions of those who actually had a genuine one.

3) People really don't like massive arbitrary electoral districts (this is why STV got rejected in British Columbia). And one where one of the two largest population centres is Aberystwyth is almost inevitably going to create some sort of awkwardness.

I should far rather see a variant on the current system, with the constituency vote switched from FPTP to the Schulze-Condorcet method, which is a fairer ranking method that allows tied preferences. It is possible to use that method on a multi-member basis, but I don't really think that's generally desirable in most of Wales (with the possible exceptions of Cardiff, Swansea, and Newport). I believe that this would be the surest way of eliminating the "can't win here" problem.

Perhaps we should also move from regional lists to a national list (to make it easier to approximate proportionality with fewer AMs), and populate it with best constituency losers (per party) rather than dubiously-selected nominees, but those are possibilities that are independent of other changes.

James Dowden said...

Forgot to include something:

4) There's no good way of holding a by-election under STV. You either end up with gradually decaying proportionality (with non-replacement or single-seat by-elections), party appointees, or the need to have all the other AMs in the constituency seek re-election whenever one of their number dies.

MH said...

Pen, we're anoraks together on this. What you say about allowing smaller groups a voice is one of the advantages of STV, it should bring more people into the political process.

Perhaps the main reason for any confusion was that I was trying to move the debate away from those of us who are already convinced about STV, to try and address the concerns of others.

MH said...

James, I guess you are among those others. Thanks for the comments.

1. I'm not quite sure what you mean by a "full slate". The generally accepted wisdom is that each party puts up as many as it thinks it will be likely to win, plus one just in case their wildest dreams come true. So in a very Labour dominated area (with say six seats) where they might expect 60% of the vote, they would only put up four candidates. If you expected to get 20% of the vote you would tend to put up two candidates. Of course there are no rules to stop a party putting up more, but it tends to be counter productive to "flood" the ballot paper.

2. I understand your point here. I think a lot of it might be solved by the way the ballot paper is designed. I think candidates from each party should be grouped together, and ordered in the preference determined by the party. Each voter then has a "default" but is free to over-ride it and put their 1,2 and 3 (or 11, 12 and 13) in the order they choose. The order of parties (or independents) on the paper would be determined by lot.

I also think it would be good to publish a facsimile of the actual ballot paper at least a week in advance of the election, to minimize the chance of confusion on the day itself.

3. That's why I think Penddu's split into local authority areas is so good. People identify with these far more than they would with a more mathematically precise electoral boundary. It's also why I think constituencies should vary in size rather than all being the same size (as in NI).


I'd agree that Schulze-Condorcet is better than FPTP, but having only one winner per seat does not do very much to ensure proportionality. The other end of the scale is a national list, which produces very good proportionality, but at the expense of a relationship between an AM and their local area. STV is a compromise between them. I'd say it combines the best features of both.


4. I think STV works fine for by-elections (it happens in Scottish Local Authorities, for example) and in fact returning one member by STV is quite similar to Schulze-Condorcet. Of course it would affect the overall proportionality compared to the original election, but it's only going to happen a few times in any term.

I think it goes without saying that the above is better than "the next on the list" (which is what Wales has at present for the regional element). STV doesn't have party lists, which is a one of its huge advantages. It would also be very unfair for all AMs from a constituency to have to stand again in the event of one of them dying/resigning. I'm sure some AMs would be tempted to deliberately resign on some pretext in order to try and unsettle another AM in the constituency.

Denis Mollison said...

Ref one of the comments by James D -

It is possible to allow voters to give equal preference to groups of candidates in STV. The John Muir Trust have been using this variant of Meek's method (in many people's view the conceptually simplest and best version of STV, its only drawback being that it does need counting by computer) for 10 years for their trustee elections. It's maybe a step too far for immediate implementation in major political elections, but allowing equal preferences does provide a natural way for voters to put (e.g.) all candidates of one party equal. This is much better than allowing the party to choose an order for its candidates, as happens in Australia, but is quite contrary to the spirit of STV, namely that the voter is free to choose between individuals.

Post a Comment