The list vote
In a Holyrood election you have two votes, the constituency vote and the list vote. The make-up of the Parliament is determined entirely by the list vote. If the parties have the following shares of the list vote, then they will have these percentages of Parliamentary seats - constituency and list seats combined:
The list vote asks the question, "Which is your preferred party?", and measures the results.
The electoral system converts list-vote shares into seats by topping up constituency seats with list seats. So if the list-vote percentages and constituency seats (in a mythical 100-seat parliament with 55 constituency seats and 45 list seats) are as below, the top-up will be:
The top-up in this case brings large numbers of seats to Con and Lab, but almost none to SNP, because that's what the list-vote shares dictate. The list-vote share is king.
"Splitting the vote"
"Splitting the list vote" loses no seats, because the split party will still get its original combined share:
Three considerations modify this transparent and practical arrangement. The first is that, for Holyrood election purposes, Scotland is divided into eight regions; the second is that, for reasons of arithmetic practicality, the method used to allocate seats to parties according to their list-vote share (the D'Hondt method) gives results that can't be predicted by simple arithmetic, and are not entirely proportional.
The eight Scottish regions are Central Scotland, Glasgow, Highland and Islands, Lothian, Mid Scotland and Fife, North East Scotland, South Scotland and West Scotland. Each has 8, 9 or 10 constituency seats (the constituencies lie wholly within regions) and exactly seven list seats.
- Regions are of different sizes: Highlands and Islands had 200,000 voters in 2016, South Scotland had 300,000. So the parties' percentages of Parliamentary seats may differ from their nationally-aggregated list-vote shares. But it still remains true that the sole determiner of a party's Parliamentary seats is the list vote.
- The ballot is evaluated exclusively within each region: the list-vote share that applies is the list-vote share within that region, the constituency seats that are counted are the constituency seats within that region, and the list seats that are allocated to the parties are the list seats within that region.
- The ballot within one region has no effect on any other region. Readers should therefore treat with extreme circumspection any commentators' remarks on voting patterns, polling patterns or predicted outcomes that don't include a regional differentiation, and most don't.
The arithmetic for allocating list seats to parties throws up two impossibilities:
- Seats can't be divided. If there are ten seats to be allocated, and three parties with equal vote-shares, the tenth seat can't be split three ways.
- All seats must be allocated: we can't just throw the tenth seat out of the window.
- These constraints have an impact, because seats are allocated within regions. A region has only 15, 16 or 17 Parliamentary seats. 10% of the list-vote therefore equates to 1.6 Parliamentary seats, which is a dog's breakfast under any allocation system.
The D'Hondt method solves these two problems by using a bidding system to allocate seats - an auction, but one where the bids are rigged in advance. Details are given on the D'Hondt method page. You cannot evaluate any election predictions without a D'Hondt calculator.
The third consideration, an important one for the current 2021 election, arises when a party's constituency seats overfulfill its list-vote share. This can be seen in the notional scenario below, where the SNP's vote-share has been halved by the split, so that it now has 17 more seats (from constituencies) than its list-vote share allows it. It'll therefore get no list seats, and the 45 list seats awaiting allocation will be split proportionally between the other three parties:
The split-off party SNP+ has no constituency seats, which gives it an advantage in the first stages of the D'Hondt bidding. It wins the first 6 seats without batting an eyelid, until its biddable votes come down to within spitting distance of those of the other parties. (You can see the bidding here.)
- This leaves the remaining 39 list seats to be divided between three parties.
- The result is a massive transfer of seats from Lab and Con to SNP+, raising the total of seats on the independence side from about 40% to about 60%.
This is the strategy that Alba is pursuing, and it (or something like it) will happen if their vote-share reaches an appropriate level.
Wikiscot D'Hondt calculator
Wikiscot has commissioned its own D'Hondt calculator from Derek Rogers, which can be found on the VoteHelp website. It's free of charge, always available, no ads, no sign-in, no cookies, and lets you enter your own data to see what outcome it will produce. It also includes full data from the 2016 Holyrood election, for you to play with.
We ran five analyses, using that D'Hondt calculator:
- The 2016 election data for the South Scotland region, but splitting the SNP vote with Alba at levels ranging from 50-50 (i.e. Alba gets half of the SNP vote) down to 90-10 (Alba gets 10 percent of the SNP vote). This region was chosen because it was the one where the SNP had its worst performance. The result was unchanged at every level of split.
- As above, but for the Glasgow region, where the SNP performed best. With a 50-50 split, the independence parties gained two seats; at 90-10, the result was unchanged. There were no losses at any level of split.
- The 2016 election data, but adding the Alba party in each region, and giving it, in each region, 6 percent of the total vote. This 6 percent was taken from the SNP's vote. The two independence parties between them took an extra seat in four regions; nowhere did they lose seats. So the split gained the independence parties four seats.
- As above, but giving Alba 10 percent of the total vote. This gave the independence parties 6 more seats, with no losses.
- As above, but giving Alba 15 percent of the total vote. This gave the independence parties 13 more seats, with no losses.
These five sets of results are on the VoteHelp website.