SP21 – Reforming AMS

In my very final analysis piece for last month’s Holyrood election, I ran through six possible reforms to the Scottish Parliament’s voting system, and how they’d have impacted the results. Five of these were relatively minor, changing just one aspect of the system to see what the effect would be. The last was a completely new system on the Scandinavian model, which long term followers of BBS will know is my preferred electoral system.

However, the five small changes would not individually go far enough to resolve the key problems with AMS. On the other hand the Scandi-style approach, as much as I may wish otherwise, is just too radical to be brought in anytime soon. I’ve therefore spent a bit of time since the election thinking about a middle ground – solid reforms to AMS that are nonetheless realistic enough to imagine happening.

Effectively, the problem with AMS as it stands is that it isn’t quite proportional enough. In that piece about various hypotheticals, I said there were two main reasons for that.

Current Problems

Too Many Constituencies

Since its inception, AMS has been tilted towards the constituency seats. 73 (57%) of the seats at Holyrood are elected via First Past the Post, and 56 (43%) by the proportional lists. FPTP is highly unproportional (that’s why we have the lists) so when you’ve got a party that is completely dominant in that element, as we do now in the SNP, they can win so many seats they knock the proportionality off. That can happen with any number of FPTP seats, but it’s particularly bad when those seats make up a clear majority of the chamber.

Indeed, this was true even when Labour were the dominant party. Though they never matched the SNP’s current seat or vote shares nationally, they were so strong across the Central Belt in particular that they regularly swept up almost every constituency seat there. Even in 2007, when the SNP first came to office, Labour had 37 of the 73 constituencies, including 9 of the 10 in Glasgow. That ensured they came just one seat behind the SNP in that election, when had AMS worked entirely correctly they’d have been four behind.

Self-Contained Regions

Compounding the issues of proportionality resulting from the FPTP element is the fact each of the eight regions is entirely self contained. The issue here is surprisingly enough basically the same as with FPTP – if you’re only allocating seats based on subdivisions of the whole, what is mostly proportional locally can become unproportional across the whole parliament.

Just like FPTP, winning seats is as much about where your votes are as it is the number of votes. With a rough effective threshold of 6% per list seat, anything outside of whole multiples of that is basically wasted. If you get 12%, you get two seats. If you get 15%, you still get two seats. This can mean parties piling on votes in places they don’t really need them, whilst falling short where they do.

We can see that with the Greens, for example. In South they came just 115 votes short of winning a seat, which would have given them representation in every region. Next door in the West region however, they had about 6000 votes more than they needed for an MSP. So, across this area, they had the votes for two MSPs but only won one. The same is true for the Lib Dems between the North East (short 2500) and Mid & Fife (excess 4000) regions. 

Closed Lists

This one isn’t about proportionality, but another common critique of AMS is that list MSPs are somehow unelected. The claim is that as their names aren’t on the ballot paper they don’t have their own mandate, and are beholden only to their party as a result. This is a largely nonsense claim – apart from anything else, parties also select their constituency candidates. There isn’t some special power voters have to refuse a constituency candidate that they don’t possess for list candidates, at least when we accept the reality that most voters are voting on a party rather than personal basis.

Nonetheless, it is something that nibbles away at the legitimacy of a large number of Scotland’s parliamentarians – and drives nerds like me up the wall. Even if just to spare Twitter all the terrible “UNELECTED BACK DOOR MSPS!!!!!!” patter, this could do with being fixed.

A Full Proposal for Reform

Equalise the Elements

If it’s problematic having more constituency than list seats, there’s a really easy solution – have an equal number instead. We could do that by reducing the current number of constituencies and converting the lost seats into list seats, but that’s actually quite a big change, and is likely to run aground of MSPs who don’t want to lose their seat in what would be quite substantial boundary changes. So, instead, we could equalise upwards, by adding a total of 16 new list seats to give a parliament of 145. This was actually one of the proposals in my previous piece, but doing it by itself isn’t quite enough.

I know folk might instinctively recoil at the thought of more elected representatives, but it genuinely does make sense. The Welsh Parliament is likely to expand from 60 to as many as 90 members in this term, partly to recognise the huge increase in workload since it was created as the Assembly in 1999. The Senedd has been on a more dramatic journey than Holyrood, as when initially set up it didn’t have full law-making powers, but the same principle applies here.

Holyrood has gained more powers, and as it has done so, it has become clear it needs more MSPs to do the tough work on committees to scrutinise the work of government, and indeed more MSPs to draw from to form that government. That makes expanding the number of list MSPs at Holyrood doubly good for democracy – improving proportionality and the work of parliament.

A National Component

There’s an equally simple, at least in principle, solution to self-contained regions – use the national vote to allocate seats. That said, I don’t really think there’d be much appetite for a single national list in the style of New Zealand. I’m also not a fan of such national approaches because whilst I think the need for local representation in a national legislature is exaggerated in this country, there is still some degree of sense to it. It is important that rural areas have a clear and distinct voice from urban areas, for example.

We could, as in another of my proposed tweaks, back-allocate the list seats to regions. This would mean going through the process by first identifying who was due the next list seat based on the national vote, then looking at the regions (with seats left to fill) to see where they should get it. Yes, that’s a more complex allocation process than at present, but normal people don’t care about the maths, just bores like me. 

Nonetheless, I’m discounting that option too, because it’s more likely to give some weird outcomes. For example, there’s one region below where purely national apportionment would have meant zero SNP seats. Although winning seats in that region would tip them over their fair national share, it would also be noticeably bizarre to voters in that region not to have any SNP representatives. Instead, a sensible compromise would be to do most of the list seats per region directly, then fill the last seat in each region using that national back-allocation approach.

As ever, I’m applying a 3% threshold for eligibility here. A party could still win a seat or seats with a share below that if they were strong enough in a specific region, they just wouldn’t be in the running for any of the nationally allocated seats. This is quite a generously low threshold by European standards – of countries with national elements to their vote and formal thresholds, only Denmark (2%) and the Netherlands (“one full seat in Parliament”, 0.67% in their case, 0.69% in this model) are lower than that.

Ensuring Accuracy of Small Party Representation

Given we’re allocating a portion of the seats nationally, there isn’t actually a huge difference in terms of using D’Hondt versus Sainte-Laguë when it comes to the overall share of seats per party. Where it would make a difference to use the latter, however, is in ensuring that smaller parties definitely win seats in their strongest areas, rather than being squeezed into often odd locations via the levelling seat process.

More Natural Regions and Constituencies

A benefit that arises from the use of Sainte-Laguë and a national element to apportionment is that we definitely no longer need to have roughly equally sized regions. Instead, all we really need is a roughly equivalent number of voters per MSP. Apart from possibly Glasgow and the Highlands & Islands, our current regions don’t really reflect natural divisions of Scotland. Why does Lothian not include most of East Lothian? What is Central Scotland, and why is only a small part of it the same as the Central Scotland Regional Council that existed from roughly 1974 to 1994?

Instead, we can do more natural regions like Ayrshire, and we can break up the geographically huge Highlands & Islands region. As ever my proposal mostly maps to what I propose for local government regions in my New Municipalism project. These aren’t perfect, but they are much better than what currently exist. 

The same logic actually applies to constituencies, which have a little bit more flexibility to be under or oversized when there are more list seats to balance out the FPTP element. That’d be most useful in reinstating the Angus-Aberdeenshire border that doesn’t exist with current constituencies, and leads to my odd “Tayside and Mearns” region in this proposal.

We could also do things like have Clackmannanshire become its own constituency rather than tie in with a bit of Stirling. Though it doesn’t map to my local government regions, I also pondered whether a “Dumfries, Galloway and Borders” region might make sense if the Borders seats were redrawn not to include that chunk of Midlothian.

In apportioning list seats to regions, I’ve maintained the current situation where the Highlands & Islands have a slightly inflated share of seats. It has 15 of 129 right now, and since I’ve added the equivalent of 2 list seats per current region but removed Moray from the area, I settled on 16 seats between the two new H&I regions. I also set it such that all regions have at least three list seats, with the remaining seats then being apportioned to regions on the basis of total electorate.

(Eagle-eyed Twitter followers may notice a shuffling of seats in those H&I regions below versus what I tweeted recently – it was only after that I thought that the three list seat minimum should be brought in.)

Open Lists

For folk who believe list MSPs lack a personal mandate, the obvious way to resolve that is to use Open lists. This would give voters the ability to vote for a specific candidate from their preferred party, with the general principle being that if a party wins N seats, their N most popular candidates will be elected.

There is an argument though for allowing some degree of party influence over the order their candidates are in, that it helps to combat bias (unconscious or otherwise) amongst voters that might affect women and ethnic minority candidates. We could strike a balance by including a threshold here too, requiring a candidate to win at least one quarter of a seat’s worth of personal votes in order to bypass party order. This would effectively mean that if voters show relative ambivalence as to which candidates should take up seats, the party decision stands, but clearer preferences take precedence.

That threshold would be much higher in the smaller regions – around 5% of the total vote in Dumfries & Galloway for example, versus 1.5% in Lanarkshire. However, there are also likely to be fewer candidates in the smaller regions, thus a party’s votes may be expected to be spread less thinly.

Final 2021 Outcome of Proposed Reforms

Overall, if these reforms had applied to May’s election, they’d have halved the disproportionality of the outcome – from 7.2 under the Gallagher Index down to 3.6. That’d be a massive improvement, that more accurately reflects how voters used their list vote. Remember, much as some on both sides of the constitutional divide would wish otherwise, that is the vote which the overall result is intended to be proportional to – the total constituency vote is nothing more than a mathematical curiosity, in electoral terms.

Notably, although there are 16 more seats up for grabs, the SNP wouldn’t have any more than their current share. That’s because they were pretty heavily over-represented under AMS as it stands. They are still a bit in excess of what they are truly due here, but not by quite so much. (Bear in mind that the votes percentage is only for those parties crossing the percentage threshold.)

Although I view this as a much more credible and likely reform to Holyrood’s voting system than my preferred Scandinavian-style alternative, that fact could be the main stumbling block. Reform requires a two-thirds majority, and even if it didn’t, the SNP of course have half of the voting seats in the chamber as it stands. They might not, on a purely partisan basis, be very keen on a reform that would slightly weaken their hand.

Nonetheless, these proposals are fair, preserve most aspects of AMS, and tackle some of the most commonly raised complaints with the system. They improve proportionality, resolve the issue of list MSPs being accused of lacking a mandate, and ensure almost every vote cast counts. Any MSP with a genuine belief in democratic fairness should find these very easy to accept.