Note that this page has three sections of increasing length:

  • The short explanation of AMS
  • The more detailed, but still reasonably succinct explanation
  • The reasons why we don’t count constituency votes as part of the proportionality, and why doing so would be a dreadful idea

You don’t need the last bit to understand AMS, it just fits best on this page, and has inflated the advertised read length!

AMS - the Quick Explanation

Unlike elections to the UK Parliament, the Scottish Parliament is elected by a system of proportional representation (PR), as is the norm in democracies across the world. Our particular version is known as the “Additional Member System” (AMS).

That’s means it’s a mixture between two different kinds of election, with two votes – the Constituency Vote and the Regional List Vote. Together, these elect a total of 129 Members of the Scottish Parliament (MSPs).

The Constituency Vote

  • Scotland is divided into 73 Constituencies.
  • Each Constituency elects one MSP.
  • The candidate with the most votes wins.

You might also hear this referred to as the “First Vote”, and it uses the familiar FPTP system to elect an MSP. It doesn’t matter whether the candidate wins by 1 vote or 10,000, whether they get 29% or 60% of the total vote, all that matters is they came first.

Although this is simple, it’s also a very unfair system. It means most voters aren’t represented. That’s partly why when the Scottish Parliament was being set up, it was decided not to use this system alone. 

The Regional List Vote

  • Constituencies are grouped into 8 regions, of 8 to 10 constituencies.
  • Each region elects 7 MSPs.
  • These seats are allocated proportionally, taking into account how many Constituency seats each party won in that Region.

This one might also be referred to as the “Second Vote”, and it ensures that the diversity of voters is more accurately represented than is possible under FPTP. Although the two votes are separate, the seats they elect aren’t. The list seats are allocated so as to deliver an overall balance across the whole region, including constituency seats won.

To do this we use something called the “D’Hondt method”. I promise this isn’t anywhere near as complicated as the fancy name suggests. You just divide each party’s regional vote by one more than the number of seats it has won so far in that region, including constituencies.

This basically means that the more constituency seats a party wins, the fewer list seats it will win. Remember, that’s what makes this system fair – if a party has already got lots of constituency seats, their voters are already represented. The list seats then make sure that other voters get their fair representation.

Depending on region and the results for other parties, the rough regional vote share required per seat is about 5-6%. So a party that wins 17% of the regional vote would probably be entitled to three seats in total. If they didn’t win any constituencies, that would be three list seats. If they won one constituency, it’d be two list seats, and so on.

You can also read a bit more about how the system works, and how to communicate it, in this little guidance paper.

So, how should you vote?

I’m occasionally asked to give some voting advice. As a non-partisan project, the only advice I will ever give is this:

Vote for the party (or candidate) you most want to.

Ahead of both the 2016 and 2021 elections, social media was swirling with chat about tactical voting. We know that it came to basically nothing last time – RISE went hardest on it and achieved 0.5% of the vote, 0.1% up on the SSP’s 2011 performance, whilst Green gains match extremely poorly to the SNP’s losses.

Let’s be clear and honest: Scottish politics social media is a bubble. Most voters will go out and do exactly as I have suggested, and vote for who they want to. Tactical voting, where it occurs, is far more likely to be on the constituency side of things, where long established patterns are much easier to predict. 

AMS - the more Detailed Explanation

Let’s take the Lothian region from 2021 as our example to really dive into the numbers. This region saw three different parties win constituencies, and two further parties win regional seats, so it’s nice and diverse for illustrating how the system works – no region saw any more diversity than that.

Lothian Constituency Votes

There are nine constituencies that make up the Lothian region, and the result of the constituency vote in each was as follows:

The SNP won 7 constituencies, whilst Labour and the Lib Dems each won 1 constituency. If this election was pure First Past the Post, the SNP would have three-quarters of the seats here even though it wasn’t anywhere near winning three-quarters of the votes. That would be fundamentally unfair, and that’s where the regional list comes in.

Lothian Regional Vote

Now we know who has won each of the constituencies, we can allocate the list seats, to make them more reflective of votes cast. To do so, we first need to see how many regional list votes were cast in total.

This is when we start using the D’Hondt method, by dividing each party’s share of the vote by one more than the number of constituency seats they’ve won.

  • The SNP won 7 constituencies, so their vote is divided by 8.
  • Labour won 1 constituency, so their vote is divided by 2.
  • The Liberal Democrats won 1 constituency, so their vote is divided by 2.
  • No one else won any constituencies, so their votes aren’t divided.

That then gives us this (removing the parties that clearly don’t have enough votes for clarity):

Since the Conservatives have the highest total here, they receive the first regional list seat. We add that to their total, giving them 1 seat overall so far, and therefore for the next round of allocation their share is divided by 2.

This time it’s the Greens that have the highest total, so they receive the second regional list seat. That gives them 1 seat overall so far, and therefore for the next round of allocation their share is divided by 2.

The Conservatives the highest total, so they receive the third regional list seat. That gives them 2 seats overall so far, and therefore for the next round of allocation their share is divided by 3.

This time Labour have the highest total, so they receive the fourth regional list seat. We add that to their total, giving them 2 seats overall so far, and therefore for the next round of allocation their share is divided by 3.

The Conservatives have the highest total, so they receive the fifth regional list seat. That gives them 3 seats overall so far, and therefore for the next round of allocation their share is divided by 4.

Labour have the highest total, so they receive the sixth regional list seat. We add that to their total, giving them 3 seats overall so far, and therefore for the next round of allocation their share is divided by 4.

The Greens have the highest total, so they receive the seventh and final regional list seat. We add that to their total, giving them 2 seats overall. Therefore, the final distribution of seats across the Lothian region is:

7 SNP (all Constituency)
3 Conservative (all Regional)
3 Labour (2 Regional, 1 Constituency)
2 Green (all Regional)
1 Liberal Democrat (Constituency)

We can then compare this with the total number of votes cast on the regional list, to see how proportional it is:

And the answer is, pretty proportional! It’s not perfect, as we can see the SNP have a bit of an overrepresentation, whilst the other four major parties are slightly underrepresented, and the smallest parties aren’t represented at all. However, that’s a natural consequence of there only being 16 seats up for grabs, and it’s substantially better than just using FPTP alone.

Why don't Constituency votes count towards proportionality?

You may see figures circulating on social media, in various contexts, that effectively combine the constituency and list vote into a “total” vote, and then use that “total” to pass comment on proportionality. However, such approaches are fundamentally misleading. Beyond determining who wins each constituency, where it doesn’t matter if a party wins by 1 vote or 10000 votes, the constituency vote does nothing. It’s a mathematical curiosity. It plays no part in the process of allocating seats.

That’s pretty standard for this voting system. New Zealand, a close comparator for Scotland in many ways, uses MMP and aims for overall proportionality according to the list vote. So too does Germany at Federal level, and at State level for 14 of the 16 Länder – the two exceptions are Bavaria and Baden-Württemberg, which we’ll get onto later.

The question is, why? Why just use one vote? Let’s dig into that a bit.

Not everyone stands everywhere...

The most obvious stumbling block to using both votes, and why you should be extremely cautious of people using the 2021 results this way, is that not every party standing on the list puts up a candidate in every constituency. In fact, if you’re an Independent candidate contesting the list (as Margo Macdonald successfully did three times) you physically and legally can’t stand in every constituency in a region!

This therefore allows some parties to get double counted, as voters can give them both votes, whilst supporters of other parties (and Independents) can only get that support counted once. It’s important at this point to remember that we should be aiming to represent voters, not simply votes, proportionally. Everyone may have two votes, but they are still just one voter.

The four Westminster parties (SNP, Conservatives, Labour and Lib Dems) all have a tradition of standing in every constituency. The other Holyrood party, the Greens, notably do not, contesting just 12 of 73 constituencies in 2021 – though that was by far the most they’d ever stood in.

Since even the most committed Green voter couldn’t give both votes to the party in most constituencies, pretending their “total” vote is comparable to e.g. the Lib Dems’ would be comparing apples and oranges. Similarly, nobody who voted for Alba or All for Unity had the option to give both votes to those parties, and most other parties contesting the list had few or no constituency candidates.

With an eye to the Greens in particular, some commentators have cried foul, and argued parties shouldn’t simply be able to “opt-out” of one part of the system. Some have gone so far as to say parties should be compelled to stand in every constituency if they want to contest the list. You could do that, but it’d be a very democratically dodgy idea.

... because it costs a bomb...

There’s a very obvious reason for why the Greens, and poorer, smaller, and newer parties, didn’t contest many constituencies – the cost of deposits. It’s £500 a pop to stand a constituency candidate, so £36,500 to stand in all 73 seats. Even if you’re confident you’ll get that money back, it’s not available to spend during the campaign. It’s therefore a perfectly sensible strategic choice to contest a limited number, or indeed no, constituencies.

The Greens deciding to give 61 seats they’d never win a pass gave them £30,500 more to spend on things that would actually win them votes and seats. Meanwhile, the Lib Dems ended up pouring £25,000 straight down the drain in lost deposits across 50 constituencies, all for the privilege of being able to say they got 1.4% in Glasgow Provan, 1.9% in Ayr and so on. That’s a lot of money just for the sake of status.

Compelling parties to contest constituencies whilst deposits exist would be outrageous. As I noted in my briefing on ballot access, wealth inequality exists amongst parties just as it does amongst people. Expecting parties to stump up tens of thousands of pounds for unwinnable seats would be deeply unfair, handing an immediate campaign advantage to larger, wealthier, and/or longer-established parties.

Of course, we could abolish deposits. But still there would be issues in terms of candidate availability. 

... and you need loads of candidates.

That’s because even if it’s free to stand, you still need to find candidates. Turning to the Greens, this is the point at which they’d be perfectly fine. They had 76 list candidates stand in 2021, so finding 73 constituency candidates would have been a breeze. As a point of interest, had they done so it’s likely they’d have then won about 6% of the constituency vote that year. That’s because across the 12 constituencies they did contest, their share was approximately 75% of what they got in the list vote, and nationwide their list vote was 8.1%.

But then we’re still left with other small and new parties. Nobody else, not any of the eight other parties who contested every regional list, stood 73 or more candidates in total. All for Unity came closest with their 56 list candidates, but that’d still have left 17 constituency gaps where they couldn’t get votes, and thus could only be single counted even if a voter wanted to give them both votes.

A new party aiming to stand nationally can do so currently by finding a minimum of 8 candidates, but 73 would be a nigh-insurmountable barrier. It’s not reasonable to expect parties that may hope to win one or a small handful of seats to find dozens of candidates, even if you abolish deposits. That’s true regardless of whether the pressure to stand in constituencies is purely from the fact they’d need to do so to maximise their vote share, or from legal compulsion as a condition of list ballot access.

That about covers things off as to why counting both votes is a bad idea from the perspective of it requiring parties to stand everywhere. But we’re still not done listing the ways in which count both votes for proportionality is a bad idea!

FPTP is self-distorting...

It isn’t simply that FPTP isn’t proportional that distorts the results. It’s the fact that voters are aware that it isn’t proportional, and vote accordingly. So not only do the seats not match votes, but the votes themselves aren’t an entirely true reflection of voter preferences.

For all the chat about tactical list voting, as I briefly state above, it’s still the case that FPTP is far more conducive to it. It’s been an established pattern for decades, voters understand how to do it, and even if not massively politically aware they do generally know who can and can’t win in their area. An absolutely prime example of this would be the Edinburgh Western constituency, where we can look at 2021 results:

The Lib Dem constituency vote here is absolutely massive. On the other hand, this is one of the lowest vote shares the Conservatives won anywhere in the country – on a par with their weakest parts of Glasgow. Edinburgh Western is a pretty affluent constituency, exactly the kind of place the Conservatives have done very well in as part of their revival. Meanwhile, the overall national result that election for the Lib Dems was their worst since devolution. 

Now, with all due respect both to the Lib Dems and their MSP Alex Cole-Hamilton, it simply isn’t credible to suggest that a decade into decline, they have the strong support of more than half the voters in any mainland constituency. Nor is it possible to believe that in the midst of a strong revival, the Conservatives are so weak here. But what the Lib Dems did have, and had in 2016 as well, was a strong target campaign, and the obvious sense that they were the only party that could beat the SNP here.

That’s borne out by their regional list vote – it’s less than half their constituency vote (29% less). That’s a yawning chasm between ballots, well beyond the norm. Meanwhile, the Conservative vote is over three times as large (14% more). The Labour vote is also abnormally low in the constituency, and about twice as high on the list (almost 6% higher).

That’s just one example, and of course the Lib Dems aren’t the only party to have benefitted in this way – and in other places, they’ll have suffered for it. It affects different parties in different places to different degrees. But summed up across the country, given the uneven spread of places various parties can win, the distortion can be substantial, especially to the disadvantage of smaller parties. So even for parties that make it onto every ballot, their constituency vote is going to be at least a bit off, and thus deviate from ideal proportional representation of voters.

... and Culturally Embedded

Going way back to the start of this section, I mentioned two of the German Länder were exceptions that don’t drive proportionality from the list vote. Instead, each of the two southern states takes a slightly different approach. Bavaria does what we’ve discussed at length above, which is drive the proportionality using the sum of both votes. Baden-Württemberg does away with the dual ballot entirely, and uses a single vote to allocate both the constituency and list seats.

Whilst Germany certainly won’t be immune to the issues I’ve outlined above, it does have one big difference – political culture. Germany uses PR at every level, and has done so for the entire duration of the modern Federal Republic, both of these states being in the former West Germany. Voters are extremely used to PR, and therefore less encumbered by the tactical mindset of FPTP.

Scotland, by contrast, has only had forms of PR for a couple of decades, and FPTP-centred Westminster is still the election with the highest profile and turnout. Especially if we moved to a single-vote model, it’s much more likely that some Scottish voters would tunnel-vision onto “I don’t want X party to win my local seat”, and not think “it’s fine if X party win my local seat, because PR means I’ll probably get at least one MSP I want at the regional level instead.”

Alright then, wrap it up!

This has ended up as a massive screed about the ins and outs of AMS/MMP, what forms of counting it should be used, and why some people are confused by it all. That alone is a sign that, at least in the Scottish context, perhaps what was adopted as a careful compromise system might just be a shade too confusing and contentious, especially in our current political climate.

Although the maths behind the system are slightly more difficult, I feel that’s another reason to look at my proposed Scandinavian style model. A purely proportional system, driven by a single vote, simply isn’t going to get bogged down in all of this chat about how people voted and what they meant by it. It doesn’t have, though you can never eliminate it entirely, the same drivers for tactical voting.