Reforming the Welsh Assembly: how do you choose an electoral system?

A nine-month inquiry by a specially convened expert panel has culminated today in the publication of a report that sets out the case for a substantial increase in the size of the Welsh Assembly. In this post, Constitution Unit Deputy Director and panel member Alan Renwick offers a personal reflection on the inquiry and its findings. He focuses particularly on the aspect of the Panel’s remit that is closest to his own research: the appropriate electoral system for an enlarged chamber.

The Expert Panel on Assembly Electoral Reform has today published its report. Set up last February by the Presiding Officer of the Welsh Assembly, the Panel was charged with investigating and making recommendations on three issues: the number of members that the Assembly needs to perform its role effectively; the electoral system through which it is elected; and the minimum voting age for Assembly elections. The Panel’s work fits into a wider agenda of Assembly reform – including a proposal to rename it the Welsh Parliament – to ensure it can exercise effectively its increasing powers and responsibilities.

Core recommendations

The Panel’s main recommendation is that the number of Assembly members should rise from the present 60 at least to 80, and preferably closer to 90. We examined compelling evidence that this change is essential – however difficult it may be politically – if the Assembly is to remain able to perform its functions properly.

Increasing the size of the Assembly in this way inevitably requires some change in the electoral system. We concluded that the simplest possible change – retaining the existing Mixed-Member Proportional (MMP) electoral system (also somewhat misleadingly known as the Additional Member System, or AMS) and increasing the number of list seats – would be defensible, but not optimal. Most crucially, it would make any increase in the size of the Assembly beyond 80 members – the very bottom of the range that we think necessary – unfeasible in 2021. Rather, the Panel recommends that, if the Assembly adopts gender quotas, the optimal system would be the Single Transferable Vote (STV). If the Assembly does not accept gender quotas (or concludes that it lacks the power to enact them – there is some legal uncertainty in this area), the best option would be a Flexible List system of proportional representation.

Regarding the voting age, meanwhile, the Panel comes down firmly in favour of a reduction to 16, accompanied by measures to ensure that young people are properly taught in schools and other places of learning about politics, including about the choices available at elections and beyond.

Reaching recommendations on the electoral system

The Panel had little difficulty in reaching many of these conclusions: the evidence on the need for more AMs is overwhelming; the case for gender quotas and for votes at 16 is also unambiguous.

The question of which electoral system is best is, however, much more contested. If there is any point on which everyone involved in this debate agrees, it is that there is no perfect system: any system requires some unavoidable compromises between competing desiderata. Indeed, there is even a mathematical proof – known as Arrow’s Impossibility Theorem – of the fact that no electoral system can guarantee that four features, all of which seem entirely worthy, will always be achieved. As an expert on electoral systems, I am frequently asked which system I think is best, but I genuinely have no general answer: it all depends on what is wanted and needed in each individual case.

But that could raise a quandary. The Panel was tasked with making recommendations. It would contribute little of value if it failed to give policy-makers a clear steer as to what would be best. Yet it is not the job of an expert panel to stray beyond judgements that can be clearly grounded in a combination of agreed values and firm evidence. Experts have been criticised – not without justification – for blurring the line between personal preferences and expert judgements in the debate over Brexit. If we are to retain public trust as providers of impartial guidance, it is important that we avoid this temptation.

Fortunately, a two-step process allowed us to derive clear, specific, and unanimous recommendations. At the first step, we identified a set of principles for judging electoral systems that are broadly uncontroversial in the Welsh Assembly context, through which we could narrow considerably the field of options under consideration. For example, the electoral system should facilitate effective and accountable government. It should deliver results that are no less proportional – and ideally more proportional – than does the status quo. It should encourage the election of a group of representatives who reflect the diverse composition of the electorate. It should allow voters a choice among individual candidates.

We found three systems that broadly satisfy these principles, which are the three systems already mentioned: the existing MMP system adapted for a larger Assembly; or either STV or Flexible-List PR using districts that are large enough to maintain proportionality and encourage diverse representation, but not so large as to risk excessive fragmentation of the party system or render AMs too distant from their constituents.

Any of these systems would be defensible. How we rank them in general depends on the weight that we attach to our various principles – which is a matter of taste as much as of expert judgement.

Generating a ranking for our report therefore required us to take a second step, which was to move from the general to the particular and consider further the specific needs of the Welsh Assembly. One consideration, as noted above, is whether an electoral system is capable of delivering an Assembly of around 90 members at the 2021 elections. STV and the Flexible List system would have no trouble in doing that, but it would not be feasible under MMP: major redistricting is impossible within that timeframe, so the number of constituency seats could not be increased beyond 40; and there are good reasons for saying the number of list seats should not exceed the number of constituency seats, leading to an overall cap of 80.

As to the choice between STV and Flexible Lists, this was finely balanced. Flexible List systems are not currently used for any elections in the UK, whereas STV is used for most elections in Northern Ireland and local elections in Scotland, and the Welsh government proposes that local authorities should be allowed to adopt it in Wales too. Familiarity and consistency across electoral systems are important, leading to our recommendation that STV should be the preferred option for Assembly elections. But we also concluded that, in the absence of gender quotas, the balance tips in favour of Flexible Lists, which would be more likely to foster diverse representation.

Summing up

Serving on the Expert Panel has been a fascinating experience. I have learnt a great deal about optimal design of legislatures, measures for promoting inclusive representation, and a host of other issues. Most enlightening for me has been the process of whittling down alternative electoral systems to a final recommendation. Were I asked to complete a similar exercise for another legislature or another time, I would not necessarily end up favouring STV. For the Welsh Assembly in its current context, however, our advice is clear. The Assembly urgently needs to grow. An STV system coupled with gender quotas would be the best way to achieve that and deliver effective, democratic governance in the interests of the people of Wales.

Download the full report of the Expert Panel on Assembly Electoral Reform.

About the author

Dr Alan Renwick is the Deputy Director of the Constitution Unit, and was a member of the Expert Panel on Assembly Electoral Reform.

Here are a few comments on a very interesting report.
1) Germany does not have by-elections. Whether a constituency seat or a list seat becomes vacant during a Parliament, it is filled by the next person on the Party’s list.
2) Flexible list is open to the same disadvantage as compulsory voting. In Australia, there is the so-called “donkey vote” where those who don’t want to vote simply vote for the candidates at the top of the list. Here, if most people just tick the Party box rather than voting for an individual, which candidate from a party is elected may be the result of very few voter preferences and who gets elected depend sensitively on the threshold for counting individual votes. We don’t know how this will work out and it seems dangerous to propose a system where something unexpected might happen. In contrast we understand STV elections quite well.
3) There is a question that, perhaps, should have been asked about constituencies; which is should their geographical size as well as their population be taken into account. The most blatant examples are Powys and Ceredigion & Pembrokeshire in the 17-constituency option. Should there have been a option to split the latter from one 6-member constituency into two 3-member constituencies and could Powys have borrowed a member from two of the adjoining authorities to allow a similar split.
4) Zipping lists in principle is a good idea but there is a danger if a party is very strong in one area of the country and weak in another part. In the part where the party is very strong it will be almost equally likely to return an odd or even number of AMs; in the part where it is weak it may return either 0 or 1 AM. Who is top of the list matters a great deal in areas where the party is weak, but much less where the party is strong. It may be useful to check using data from previous list elections and see whether using zipped lists with various pairing criteria makes a difference or not. It would not look good if zipping was enforced but gender differences remained.

2 thoughts on “Reforming the Welsh Assembly: how do you choose an electoral system?”

Pingback: Why it’s time to reduce the voting age to 16 in Wales | The Constitution Unit Blog
Laurence Cox says:

December 12, 2017 at 2:42 pm

Here are a few comments on a very interesting report.
1) Germany does not have by-elections. Whether a constituency seat or a list seat becomes vacant during a Parliament, it is filled by the next person on the Party’s list.
2) Flexible list is open to the same disadvantage as compulsory voting. In Australia, there is the so-called “donkey vote” where those who don’t want to vote simply vote for the candidates at the top of the list. Here, if most people just tick the Party box rather than voting for an individual, which candidate from a party is elected may be the result of very few voter preferences and who gets elected depend sensitively on the threshold for counting individual votes. We don’t know how this will work out and it seems dangerous to propose a system where something unexpected might happen. In contrast we understand STV elections quite well.
3) There is a question that, perhaps, should have been asked about constituencies; which is should their geographical size as well as their population be taken into account. The most blatant examples are Powys and Ceredigion & Pembrokeshire in the 17-constituency option. Should there have been a option to split the latter from one 6-member constituency into two 3-member constituencies and could Powys have borrowed a member from two of the adjoining authorities to allow a similar split.
4) Zipping lists in principle is a good idea but there is a danger if a party is very strong in one area of the country and weak in another part. In the part where the party is very strong it will be almost equally likely to return an odd or even number of AMs; in the part where it is weak it may return either 0 or 1 AM. Who is top of the list matters a great deal in areas where the party is weak, but much less where the party is strong. It may be useful to check using data from previous list elections and see whether using zipped lists with various pairing criteria makes a difference or not. It would not look good if zipping was enforced but gender differences remained.