Electing an Enlarged Welsh Assembly

Simon King
University College, London WC1H 9EZ

 

2.   Simulations

In 2002, the Welsh Assembly government instituted a Commission to review the powers and electoral arrangements of the National Assembly. Part of its remit was to consider whether the size of the Assembly is adequate to allow it to operate effectively and whether the means of electing the Assembly adequately represents all significant interests in Wales. Commentators have suggested that if the size of the Assembly were to be increased that it should go up from 60 to 80 members. What follows is an exploration of what might happen under various methods of electing a National Assembly for Wales if it were expanded from 60 to 80 members. We take the election results of the two elections held so far under AMS, in 1999 and 2003 as well as the 2001 Westminster contest and simulate them under different electoral systems.

The calculations are based on three scenarios:
A.
B.
C		 Welsh voters divide just as they did in the 1999 Assembly contest.
Welsh voters divide just as they did in the 2001 Westminster election.
Welsh voters divide just as they did in the 2003 Assembly contest.

Several systems are considered.
  1. 80 single member seats elected by first-past-the-post (FPTP).
  2. (a) 40 double member seats with each elector having two votes (FPTP).
    (b) 40 double member seats with each elector having one vote (FPTP).
    (c) 40 double member seats with each elector ranking candidates by number and seats allocated using the Droop quota (votes ÷ (seats + 1)) (STV).
  3. 40 single member seats (FPTP) and 40 AMs elected on regional lists.
  4. 40 single member seats (FPTP) and 40 AMs elected on a national list.
  5. 60 single member seats and 20 AMs elected on regional lists.
  6. 80 members elected on national list exactly proportionately using the Hare quota (votes ÷ seats).
  7. 80 members elected on national list using the existing d’Hondt divisor (1,2,3,4…).
  8. 80 members elected on a national list using the modified Sainte-Lägue divisor (1.4,3,5,7….) as used in New Zealand.
Other systems are conceivable. The most obvious would involve multi-member constituencies (perhaps twenty 4-seaters) with election by the Single Transferable Vote as used for the Dáil in the Republic of Ireland and for the Northern Ireland Assembly. This would produce a moderately proportional Assembly but might provoke more individualistic politics and intra-party conflict. It is very difficult to model its exact consequences though it would rarely yield a clear majority for any one party.

As the tables below show, the outcomes can vary greatly both between the different systems and between the 1999, 2001 and 2003 hypotheses. The following calculations are subject to assumptions which are arbitrary and in some cases plainly absurd.

i. In systems 2(a), 2(b), and 2(c), where constituencies are divided to make 80 seats out of 40, it has to be assumed that the two halves are identical in partisan terms. This would in fact not happen; in marginal cases the runner-up party would certainly pick up one or two AMs and certain parties would have concentrated support in certain areas which would not make for an even distribution of the vote.
ii. In system 2b it is assumed that in certain constituencies the leading party would show sophistication in putting up two candidates and guiding half of its supporters to support one candidate and half the other. Therefore in seats where the winner’s majority is more than twice the second candidate’s vote the second seat is allocated to the winning party. However in 1999, in 2001 and in 2003 there were a few seats where the margin was only just over twice the second vote and the efforts to manage an exact division might have failed; in most of these cases Labour would have been the loser. Labour would perhaps win three to five seats fewer than these tables suggest. Otherwise, the figures have been calculated using single member constituency results only. This underestimates the performance of small parties such as the Greens or the UK Independence Party who concentrated their campaigns on winning regional list seats.
iii. If Wales were to get 41 or 42 seats in the ongoing Westminster redistribution, as is rumoured, the need to redraw boundaries would complicate any arrangements for the Welsh Assembly. This article assumes that Wales will continue to have 40 Westminster seats.
iv. The regions of Wales are based on the five MEP constituencies used in the 1994 European Parliament election. From 2004 there will be only four Welsh MEPs—but that need make no difference to the regions used for the Welsh Assembly. Where relevant it is assumed that the five regions will be maintained intact.
v. There is no way of knowing at all precisely how voters would order their STV preferences. The figures for system 2 (c) make the assumption that each party’s supporters share the same preference ordering. Two possible orderings are offered. (A) is based on an arbitrary set of suggestions made by diverse observers; (B) is based on the findings of the 1999 Welsh Election Survey.

Table 1. Hypothetical Voter Preference Ordering of Welsh Parties
(A) 1 2 3 4
Con Voters Con LibD Lab PC
Lab. Voters Lab LibD PC Con
Lib.D Voters LibD Con PC Lab
PC Voters PC LibD Lab Con
(B) 1 2 3 4
Con. Voters Con PC LibD Lab
Lab. Voters Lab PC LibD Con
Lib.D Voters LibD Lab PC Con
PC Voters PC Lab Con LibD
In practice in every area individual choices would be diverge widely from each of these schema and in some areas, the balance might be very different from elsewhere. Our three examples show how two assumptions produce very different outcomes.
vi. If, instead of 40 constituency and 40 list members, the system was based on 60 constituency and 20 list members, there would have to be comprehensive redrawing of constituencies. The calculations in scenario 5 assume that the 60 constituency seats would divide in the same ratio as the actual 40 in 1999, 2001 and 2003 and that the 20 list seats would divide as they did in 1999 and 2003 or might, hypothetically, have divided in 2001.
vii. I also explore the outcomes that could come from an 80 seat Assembly elected by national list on the d’Hondt system as used in the Netherlands and elsewhere. The d’Hondt system of allocating seats (used currently for the list seats in Wales) is recognised as being kind to the larger parties. It is notable that the adoption of the ‘fairer’ Sainte Lägue system actually makes little difference.
viii. Proportionality is calculated using the Gallagher Least-Squares index (LSI). It is calculated thus: square the vote seat differences for each party (ignoring parties with less than 0.5 percent of the vote); sum them; divide the total by two and then take the square root. The principal advantage of this index over others is that it is not so easily distorted by the presence of small parties; nor has it too many problems with systems containing large numbers of parties.
ix. The table for the 2003 elections also includes results for the Green and the UK Independence party. Both parties performed well in the election although neither won any seats.