Friday 7 May 2021

Who will be happy with today's Senedd election results?

 Counting starts today in the parliamentary elections in Scotland as well as Wales. Both use a form of voting which attempts to reflect the true proportions of party preferences across the nation. However, for reasons detailed in the critique below, many will not be happy about the outcome. Unlock Democracy (the successor to Charter 88) and Liberal Democrats for Electoral Reform (LDER) are two groups who want to progress to a better system and will be holding a Zoom meeting on 19th May in the light of the results of the current elections.

The discussion will be conducted by Wendy Chamberlain, Lib Dem MP for North East Fife, vice-chair of the all-party parliamentary group for electoral reform and president of LDER, and Tom Brake, director of Unlock Democracy and former MP for Carshalton and Wallington.

Proportional representation in action?
LDER Chair Denis Mollison writes:

Elections for the Scottish and Welsh Parliaments, and for the London Assembly, will all use the Mixed-Member Proportional (MMP) system. It is an opportune moment to consider how well the system is working, in terms both of proportionality and voter empowerment.

How does MMP work?
MMP is often explained as a first-past-the-post (FPTP) system, topped up with additional members so as to make it reasonably proportional overall (hence its alternative name of 'Additional Member System’); where the numbers of additional members are calculated by an obscure sequential process.  But it is better understood as a proportional party list system invaded by a lot of FPTP cuckoos – the constituency members.  If the system is working as intended, these constituency members simply replace some from their own party’s list; but if there are too many of them from a particular party, they may also displace some from other parties, thus making the overall result disproportional: this situation is called an 'overhang’. Overhangs have never happened in the London Assembly, but are frequent in Wales, and the likelihood of them is crucial to understanding the forthcoming Scottish election.
 
The list-PR ('regional vote’) part of this system, while much better than FPTP, has a considerable disproportional bias in favour of larger parties: around 42% of the votes is sufficient to win 50% of the seats. But worse is to come when we factor in the constituency results. In these three UK elections, more than half the seats are allocated to the FPTP constituencies, and a dominant party with around 40% support may be able to win almost all of these, and thus an overall majority; if it does that, its regional vote no longer matters. The possible result is a cascade of disproportionality, because the rational choice for its supporters is then to use their regional vote to vote tactically for their second choice. Alex Salmond’s new Alba Party has been formed with the explicit intention of exploiting this flaw in the system. Also, it becomes impossible to say exactly how disproportional the result is, because we no longer know what the voters’ real preferences (supposed to be reflected in their regional votes) are.


Forecasts for the Scottish Parliament
Detailed predictions based on polls taken around the beginning of April can be found at www.macs.hw.ac.uk/~denis/mmp_elections/scotland2021.html.
These present figures showing for each region: (a) how seats would be allocated from simple list-PR, (b) a graphic of how constituency votes might change for each seat in the region, and (c) the resulting allocation under MMP.
 
The predictions show the SNP winning 63 constituencies but just 2 regional seats, giving them the barest possible majority in the 129-member parliament. This is of course subject to various uncertainties, even if party support were to remain unchanged on election day, but as it shows the SNP exactly achieving their target it makes a convenient central estimate on which to base discussion.
 
Before summarising their implications, note that the analysis presented is on a region by region basis, which (pace almost all press coverage) is the only way to understand an MMP election.  In particular, constituency results only matter where they either increase or decrease an overhang, and overhangs are a regional phenomenon.  In the estimates considered, 4 regions (C, G, NE and W) seem very likely to have overhangs, 2 may (Lothian and MSF), and 2 are unlikely (H&I, South).
 
The key result is that the SNP’s fortune rests almost exclusively on constituency contests, precisely the opposite of what is supposed to happen if the system is working proportionally. Constituency results are more difficult to forecast than regional seat allocations under PR: parties concentrate their campaigning on the small fraction of voters that live in marginal constituencies, with difficult to predict effect. There are 10 seats where the majority in 2016 was less than 7%, of which the estimate shows the SNP winning 5, so it seems reasonable to suggest a possible error of +/- 5 seats in the forecast, giving them a total in the range 60-70.
 
Second, the contest among other parties depends almost entirely on their relative regional votes. One of the oddities of the system is that if a party (here the SNP) loses a constituency that matters (i.e. in a region with an overhang), the party that benefits is not necessarily the one that wins the seat – it is whichever was the list runner-up in that region, which may be a quite different party. Thus the contest among other parties, particularly the question of which party comes second, does not depend directly on the constituency battles that will decide whether the SNP wins a majority. So the election really is a game of two halves: the constituency vote is crucial for the SNP, the regional vote for determining how the non-SNP seats are shared among the other parties. Because the latter depends just on their regional votes, it is much more predictable, with the dominant uncertainty being simply the accuracy of opinion polls, typically +/- 3%.
 
The final remark on Scotland here should be that these polls only represent one snapshot. As I write (1 May) the polls have been narrowing, and an overall SNP majority looks less likely, though their projected support is not yet low enough to threaten the great majority of their seats that they hold with majorities of 12% or more.


London Assembly
In the London Assembly, no party has held a majority, though Labour fell only just short in 2012 and 2016, winning 12 of the 25 seats on 41% and 40% of the regional vote respectively. There have been no overhangs, though the Conservatives came very close in 2004: they won 9 constituencies, and were only 1.3% behind Labour in another, while their regional entitlement of 9 seats could have fallen to 8 on a less than 0.5% swing to Labour.
 
A recent poll by Yougov  suggests that Labour have a chance of an overall majority in the current election (44%, 12-13 seats), with Conservatives on 29% (7-8), Green and Liberal Democrats each on 11% (just enough for 3 seats).


Wales
Wales has a significantly smaller proportion of regional seats in its system, only one-third as against 43-44 % in Scotland and the London Assembly. As a result, there have been regular overhangs, all in favour of Labour, in 4 of its 5 regions; in 2011 Labour won 50% of the seats despite having only 37% of the regional vote. In 2016 they fell back by one seat, but retained power by bringing the one Liberal Democrat, Kirsty Williams, into coalition as Education Minister.

Polls from April 2021 suggest it is very unlikely that Labour (ca. 34%, 26 seats) will achieve an overall majority. They are likely to have to negotiate an agreement with Plaid Cymru if they are to stay in power. 


Discussion
This round of elections is likely to confirm that MMP as a system, while a great improvement on FPTP, is at best not very proportional, and significantly worse when there are overhangs.  Overhangs can be eliminated by adding extra 'compensatory’ seats, as is done in Germany; but the number of such extra seats may be large – in their most recent election 111 extra MPs needed to be added, making the proportion of regional seats 58%.
 
MMP also rates poorly in voter equality and empowerment. It combines the marginal seat and tactical voting problems of FPTP with the party lists of list-PR. And it is poorly understood: very few voters realise that if the system is working as it should electing a candidate in a constituency simply replaces someone from the same party on the list.
 
For electoral reformers, the most interesting question in this election is whether Labour, Plaid and the Liberal Democrats can win 2/3 of the seats in Wales between them, as that would enable them to complete the reforms proposed in the McAllister Report of 2017 (http://orca.cf.ac.uk/125328/), which include replacing MMP with STV.


 


 

 


No comments: