What if? Post-war elections under the Single Transferable Vote

Author:
Dylan Difford, guest contributor. Opinions and research are solely the author's and do not necessarily reflect the opinions of the ERS.

Posted on the 28th April 2021

The Single Transferable Vote (STV) is one of the most popular voting systems among electoral reformers in the UK – favoured for its supposed ability to produce broadly proportional outcomes while retaining fairly localised constituencies. But STV isn’t a straight proportional system. As a preferential-proportional system, it isn’t as simple as X percent of votes will lead to X percent of seats. Voters are free to allow their vote to be transferred between parties in accordance with their preferences, preventing wasted ballots and affecting the allocation of seats.

To analyse the potential effects of STV, the ERS has, for each recent general election, created a model of how the vote may have gone under STV – using specially commissioned polls to see how voters would have cast their votes on a preferential ballot paper and then projecting this onto a system of 3-6 member constituencies. Building on the ERS’s work, I have gone back and modelled every post-war general election result under the same system, utilising high-quality data on voter’s preferences from the British Election Studies.

There are issues with doing this. Under a different voting system, parties and voters would likely behave differently and different governments may have been formed, enacting different policies. This isn’t an alternative history narrative, but a projection of each election in isolation onto the rules of STV. Hopefully, this can give us a strong indication of how STV would treat the existing British party system.

Table A: Summary of actual seats won in UK General Elections under FPTP and simulated STV seats

General Election FPTP Con Lab Lib SNP PC Oth STV Con Lab Lib SNP PC Oth
1945 210 393 12 0 0 25 256 326 37 0 0 21
1950 298 315 9 0 0 3 285 308 28 0 0 4
1951 321 295 6 0 0 3 301 312 8 0 0 4
1955 345 277 6 0 0 2 319 300 9 0 0 2
1959 365 258 6 0 0 1 332 281 15 0 0 2
1964 304 317 9 0 0 286 290 51 1 1 1
1966 253 364 12 0 0 1 261 330 33 2 1 3
1970 330 288 6 1 0 5 306 294 21 4 2 3
1974 297 301 14 7 2 14 243 251 110 17 2 12
1974 277 319 13 11 3 12 235 268 95 23 2 12
1979 339 269 11 2 2 12 293 254 68 7 1 12
1983 397 209 23 2 2 17 279 171 179 3 1 17
1987 376 229 22 3 3 17 284 184 159 5 1 17
1992 336 271 20 3 4 17 275 228 113 16 2 17
1997 165 419 46 6 4 19 184 327 112 15 2 19
2001 166 413 52 5 4 19 189 313 120 15 3 19
2005 198 356 62 6 3 21 196 254 160 11 3 22
2010 307 258 57 6 3 19 244 196 175 11 3 21
2015 331 232 8 56 3 20 273 236 28 34 4 75
2017 318 262 12 35 4 19 285 296 26 21 3 19
2019 365 203 11 48 4 19 308 225 59 30 4 24

From our results, we can see that STV is indeed fairly proportional, with an average disproportionality rating of 5.4% – comparable to levels seen in Ireland (also STV), Germany (MMP) or Finland (List) and a significant improvement on the average 15.5% score seen in the actual Westminster First Past the Post (FPTP) elections. There is a small bias in favour of the two largest parties, with their seat shares generally being inflated by 1-3% each. Single-party majorities are possible, four elections producing one outright and a further three seeing a party fall just short. A high level of party fragmentation is unlikely, with no more than five parties ever attaining more than 2% of seats and feasible two-party coalitions present in nearly all elections.

But some of these effects aren’t caused solely by STV. The slight favouring of larger parties is a by-product of the small constituency size. This is a factor across all proportional systems – smaller constituencies mean a lower level of proportionality which benefits large parties. But, of course, it is a trade-off. Larger constituencies would increase proportionality, but they would also increase detachment from local areas. And it isn’t a straight swap. Doubling the size of a constituency doesn’t halve the disproportionality, but it does halve its localness.

Where we can see the particular effects of STV is when we compare our results to the closest non-preferential proportional system – a Party List system in the same constituencies (using the Droop quota and the largest remainder method).

Table B: Summary of simulated seats won under Droop-LR and a comparison to STV

Election Droop-LR Con Lab Lib SNP PC Oth Preference Boost Con Lab Lib SNP PC Oth
1945 266 327 27 0 0 20 -10 -1 +10 0 0 +1
1950 291 315 16 0 0 3 -6 -7 +12 0 0 +1
1951 301 314 7 0 0 3 0 -2 +1 0 0 +1
1955 321 302 4 0 0 3 -2 -2 +5 0 0 -1
1959 326 290 11 0 1 2 +6 -9 +4 0 -1 0
1964 287 298 42 1 1 1 -1 -8 +9 0 0 0
1966 276 327 23 2 1 1 -15 +3 +10 0 0 +2
1970 309 299 13 3 2 4 -3 -5 +8 +1 0 -1
1974 243 249 109 20 2 12 0 +2 +1 -3 0 0
1974 239 272 86 23 3 12 -4 -4 +9 0 -1 0
1979 296 261 56 9 1 12 -3 -7 +12 -2 0 0
1983 284 179 165 4 1 17 -5 -8 +14 -1 0 0
1987 289 203 132 7 2 17 -5 -19 +27 -2 -1 0
1992 286 242 88 16 2 17 -11 -14 +25 0 0 0
1997 204 332 85 16 3 19 -20 -5 +27 -1 -1 0
2001 212 316 93 15 4 19 -23 -3 +27 0 -1 0
2005 216 255 142 10 3 20 -20 -1 +18 +1 0 +2
2010 253 209 154 13 3 18 -9 -13 +21 -2 0 +3
2015 264 226 23 35 4 98 +9 +10 +5 -1 0 -23
2017 295 288 22 25 2 18 -10 +8 +4 -4 +1 +1
2019 321 227 48 29 3 22 -13 -2 +11 +1 +1 +2

The most obvious feature of STV here is the consistent boosting of the Liberal Democrats and their predecessors. Historically the second choice of most other voters, it isn’t surprising that they would get a lift from a preferential system. But the lift doesn’t result in them being overrepresented by STV. The small constituency size in both proportional systems results in the Liberals generally being underrepresented, as they often won fewer votes than the average quota needed to elect an MP. STV’s boosting of the Liberals offsets the small constituency size and leads to our STV results being marginally more proportional over the post-war period than the supposedly straight proportional Party List system.

But preference voting doesn’t just affect the Liberals. Their gains are largely coming at the expense of the two main parties. There are clear periods where this is at the expense of Labour (the 1980s) and periods where the Conservatives are losing out (the New Labour years), but overall it is not particularly at the expense of either and the two generally remain slightly overrepresented. The transferable element of STV allows for a greater realisation of voter’s true preferences, rewarding parties with secondary popularity and punishing those without it, as well as reducing the likelihood of split or wasted votes.

While debates about proportional representation generally focus on headline proportionality, a constituency PR system, like STV, also makes parties more representative of their own voters. After the 2019 general election, 24% of Conservative voters did not have a Conservative MP, 51% of Labour voters had a non-Labour MP and 92% of Lib Dem voters were represented by an MP they had no hand in electing. These areas of non-representation are often contiguous, skewing each party’s geographic base and often leading to such areas being habitually ignored by both major parties – the holding party knowing they will keep the seat no matter what and the opposing party knowing they won’t win there. This can have devastating social and electoral consequences – with Labour’s recent experiences in Scotland and the ‘Red Wall’ being clear warnings of the perils of safe seats.

Under STV, major parties would have seats in nearly all constituencies, with 88% of all voters in 2019 being represented by an MP from their first-choice party – compared to 55% under FPTP. This would make parties a better reflection of their voters and would give all main parties (and governments) a stake in all parts of the country, producing caucuses of rural Labour MPs, urban Conservatives and a better balance between the SNP and unionists in Scotland. STV would create a more responsive and competitive constituency system than exists under FPTP.

The increasing number of voters opting to cast their ballot for smaller parties doesn’t really appear in most of our STV projections. Largely this is because these simulations are based on votes cast under FPTP – a system that sees alternative voices squeezed out, with larger parties preying on voter fears of wasting their ballot or allowing a worse alternative to win because of a split vote. These concerns don’t exist under STV, with voters being able to vote for their genuine first-preference without worry.

But STV is also partially responsible, with the small constituencies meaning that, in this model, there is an average quota of 20%. While the transferable nature of STV means that many seats are won from below the quota, winning seats from less than half is difficult – even for the most popular parties. Much like any constituency PR system, having 3% of the vote spread evenly across the country isn’t going to win you many seats under STV.

But, with an actual STV election, parties like the Greens (with a low, evenly spread vote) would likely win a higher vote share, as well as being helped by high levels of secondary popularity. Analysis of the 2019 general election suggests that, if the Greens merely doubled their share of the vote, they would be able to increase their number of MPs eight-fold from 2 to 16. STV can be hard for small parties that are just starting out. But for those that are able to build a support base, it offers room to grow and win seats – especially rewarding those that are able to gain secondary popularity, while punishing those on the extremist fringe.

In all, our simulations suggest that STV would result in broad proportionality in a definite constituency system, unlikely to incentivise fragmentation or to make government formation difficult, while offering benefits to voters and parties alike. Other proportional systems will offer greater levels of parliamentary proportionality, but they can only do so at the price of much larger constituencies. STV sits in the electoral system sweet spot, meeting nearly all of the ideal voting system criteria to a high degree. It is proportional, it is local, it maximises voter choice. It is perfectly understandable why it is one of the most popular voting systems among electoral reformers in the UK.

For more information on the methodology and findings of Dylan’s simulations, as well as an election-by-election breakdown of the results and an analysis of European Election results, a fuller summary of the simulations is available to download here.

Read more posts...