Tuesday, May 3, 2016

Proposal for a reformed voting system : the Single Proxy Vote

A guest post by John S Sharp

We get a lot of speculation (including much self-interested misinformation) on the effect of different voting strategies under AMS for the Scottish Parliament; and beyond that I've seen a few people suggesting reforms. We indulge in endless discussion of voting strategies and outcomes when I think that a simple form of proxy voting - as you might do as do to authorise a delegate to vote on your behalf at a congress or AGM - could entirely solve just about all the problems commonly discussed. This might be termed Single Proxy Vote (SPV) - although it has a somewhat different context from what is normally meant by proxy. None of this can be an original idea (it's all too obvious, I think) but I have not been able to find any documents describing anything similar for an elected chamber.

There is only really one big idea (which admittedly might be a cultural shock to traditionalists) and it is that after a voter has cast his or her single electoral vote within a constituency, then the MSP so elected, instead of casting one vote in the parliament, will cast the number of votes that (s)he gained in the election, along with a further adjusted share of the votes for their party. With electronic voting it should be easy enough to administer this at the time of votes within the partliament.

How constituency and list voting could work

Starting with the votes cast at the polling booth, a voting form would show :

(a) a list of constituency candidates - one entry for each party / named person

(b) a list of parties that have put up some regional list candidate(s) - one entry for each party

Effectively this latter part (b) resembles the current regional list, but each entry, whether a constituency person/party, or a list party entry, has equal status - and the elector has only a single vote, to be placed against any entry.

A constituency MSP will be elected, taking into account only the votes cast for the constituency candidates. One may expect that most electors would cast their single vote directly for a constituency candidate, but the voter may alternatively choose to vote for a party with list candidate(s) only, if preferring this to any constituency candidate. Votes cast for losing constituency candidates, and for list-only parties, would be accumulated, according to party.

We might also start with the idea that the total number of list MSPs in the parliament will remain fixed at 73 constituency seats and 56 regional seats. We can assume that there will generally be more votes cast for losers than for winners (perhaps up to twice as many) and also note that there are fewer regional seats than constituency seats. Hence, if reallocated party votes were simply divided levelly to each of the regional MSPs (and this could be done either within each of the 8 regions, or nationally) then on average each one might end up with more votes than a constituency MSP - typically up to twice as many votes in parliament. We might want to stick to the principle that the constituency MSPs should have most of the voting power (73 / 129 = approx 57% as at present). It is not difficult to address this: once list MSPs have reached some threshold of votes, then the remainder would be spread back to constituency MSPs (still within party, of course - thus, the addition per MSP would vary between parties).

The Electoral Commision would have to set the broad principles of this in advance. For example, we might think that in principle, the average voting power of a Regional MSP should be the same as an average constituency MSP. In that case, only 56/(56 + 73) of the total votes cast should be allocated to the list MSPs, and the remainder then spread back to the constituency MSPs (all calculations being made within party). If we think that, in principle, constituency MSPs should typically have more voting power, the votes allocated to list MSPs can be proportioned down.

There would be some problems in trying to treat fairly parties with small shares of the vote. Firstly, a party polling a reasonably uniform, say, 2% across the country, would fail to get any constituency MSPs, and would also fall below the threshold needed to justify a "whole" list MSP in any of the Regions. However a party getting 2% ought to get some representation in a parliament of 129 MSPs. This could be addressed by having a "National List" to support the election of a list MSP from a party that has accumulated enough votes, in comparison to those counted on a genuine Regional basis. Secondly, a number of parties will not attract enough votes to justify any list MSP, even on a nationwide scale (in 2011 there were 16 parties which gained less than 1% of the total vote). It is perhaps an open question as to whether votes for these parties simply do not count; or whether they could be reallocated, either according to a binding declaration in advance of the election, or determined by the party post-election).

The Electoral Commission would have a job to do in setting actual thresholds after total votes were known; this should be uncontroversial, as the balance of voting power between the parties would not be affected by such decisions.

At this point, I'd say that the principles are well-defined and the rest is just a matter of detail. Before looking at this in some more detail (using example figures from the 2011 Scottish Parliament election) consider some of the advantages we might obtain from such a system.

Advantages and potential problems

Advantages :

1. Electors cast only one vote.

2. Every vote counts, and eventually has exactly the same weight in parliament, whether the constituency is large or small, and whether the candidates are running neck and neck or the result looks a foregone conclusion.

3. Each constituency is represented in proportion to its population (or rather its population x turnout rate) - which means that differences in size do not create a democratic imbalance.

4. If preferred, we could have single member, or multi-member constituencies, or a combination of types, even within one election.

5. There is little motivation for tactical voting - in the present sense of voting for a party candidate you don't really want in order to stop one you really hate succeeding in a FPTP election. Doing so will always damage your preferred party.

6. Devious gerrymandering of boundaries has little reward, and so we can be relaxed about setting of boundaries.

7. Each elector has a link to a constituency MSP (a point that many adduce in favour of maintaining FPTP).


Some might think the following could be described as problems. I think that they are really just differences.

1. Commentators might have a bit more of a problem in speculating on resuts of a vote in parliament, but if they can't fix that with a spreadsheet, then they are not up to much.

2. Some MSPs might be said to be more important than others, measured by voting entitlement; but why should they not be, if they have drawn the electors' votes.

3. You might not like the idea of seeing votes in parliament recorded as numbers such as 12,306 (Sturgeon), 4,462 (Scott) etc. (using examples from the 2011 election as described below). We could make the numbers smaller by some formula such as 1 parliamentary vote for each 100 electors' votes (so 123 for Sturgeon, 44 for Scott). Or the MSP with the highest total of electors' votes would get 100 parliamentary votes, and all others indexed to that scale. I think that any of these "simplifying" schemes just make things more complicated and it's best just to stick to the original figures.

4. There could be slightly different incentives to commit electoral fraud. At present there is a significant gain from fraud only if it affects the FPTP result in that constituency. In many, probably most, constituencies the result is not close enough to be affected unless fraud is carried out on a large scale, and in any case there would be focus on any suspicious activity around a close result. Under SPV proxy voting, there would be some gain from any undetected fraudulent voting, or corrupt loss of valid voting, wherever it might occur. A few hundred votes added to a fourth place gives as much fraudulent gain in the parliament as would adding it to the winner, or adding it to a near second place to swing it it to a win. Overall, however, SPV should make it more difficult to achieve any significant scale of influence through fraud.

5. It becomes more difficult for any single party to get an absolute majority of votes in the partliament. You can firmly and exactly say that a party must exceed 50% to do it. However, I'd suggest that this voting system might have some direct effect on fragmentation of parties, as in next point.

6. This voting system might encourage parties to fragment, on the grounds that this could be done with no or minimal impact on the total strength of the party in parliament. For example, if some SLab members wanted to stand on Labour / Independence platform and drew some votes away from mainstream SLab (and SNP) they could argue that this would not reduce the total Labour-aligned vote in the parliament (I recognise however that there would be a lot more to be said about party whip). Similarly, any major policy splits within SNP, could lead to group(s) prepared to compete separately within the banner of independence.

7. All the calculation part sounds horribly complicated, and therefore would put people off.  However, once you convince yourself that it is all fair, I'd suggest that the only point at which this matters is in the few hours between the close of polls and the announcement of the new crop of MSPs, each with their numbered voting entitlements.

How might things have worked in 2011

To illustrate with results from constituencies in the 2011 election

- in Glasgow Southside, Nicola Sturgeon would have been elected for the Scottish National Party, with a basic entitlement to cast 12,306 votes in the parliament, plus some "spread back" top-up to be calculated as above. Unsuccessful candidates' votes would be reallocated as follows: Labour 7957, Liberal Democrats 612, Conservative 1733.

- In Shetland Islands, Tavish Scott would have been elected for Scottish Liberal Democrats, with a basic entitlement to cast 4,462 votes (plus spread back top-up) in the parliament. Unsuccessful candidates' votes would be reallocated as follows: Conservative 330, Scottish Labour 620, SNP 1,134, and Independent* 2,845. Obviously, the redirection of votes for an Independent is problematical in comparison with that for party allegiance, but I see no reason why an Independent could not be required to state in advance of election how (or if) reallocation would be applied to his/her votes in the event of an unsuccessful candidacy.

The overall result, based on constituency vote only, would have given SNP 902,915 votes available for their MSPs to cast in the chamber (45.39%); Labour would have had 630,461 (31.69%) etc.. For smaller parties such as Scottish Greens, we'd have to guess that their region vote (4.37%) is more representative of the votes that they might have aggregated under what I am describing as STV. Thus, in 2011, pro-independence parties would have fallen short of being able to assemble a pro-independence majority in the parliament.

But if we wanted a guide to how things might turn out in a future election, such as May 2016, then instead of looking back at 2011 votes, we'd be better just to look at current polling and guess that parties would poll somewhere between constituency intention and list intention figures returned by the pollsters.

Could this actually happen?

I'd assert that this style of SPV would provide excellent democratic accountability, and if by some chance we were already using it, very few would want to change it for FPTP or multiple-vote AMS. However, starting from where we are, is there any chance that we could move to SPV? I would think not. It's just too big a change in one step from what we do now. If we had a need to elect another assembly on a party political basis - i.e. not our main parliamentary chamber - then perhaps there might be more readiness to trial a break from our tradition of one vote per member in the chamber.

*  *  *

Just a reminder, because people often seem to forget what's written at the top of the post by the time they get to the bottom : this has been a guest post by John S Sharp, and the proposal outlined in it is his own.


  1. MSPs could apportion their votes to different sides of a parliamentary decision, depending on how they felt - eg 60% aye, 40% nay for something they kinda supported but not entirely :)

  2. It has interesting repercussions for lobbying, too.

    I'm not afraid of complex numbers, but expecting parliamentary votes to regularly add up to 2 million is startling to say the least. Not really a problem though, I believe Holyrood already uses electronic voting - this system would still be vastly quicker than Westminister's divisions.

    A question, does SPV allow the possibility of a minority of MSPs with a majority of votes? Something like 60 MSPs with 1.2 million votes between them and 69 MSPs with 1 million between them?

    1. "a minority of MSPs with a majority of votes" - Yes, I think that might well be the result. I don't see that as a problem though. The MSPs with a higher number of votes to cast got them by being more successful personally in a constituency and/or representing a more successful party and/or having contested a larger constituency.

    2. It is a problem, especially for depopulated areas (cf Scotland vs SE England) however it is easily dealt with. You could require two types of majority - members and voting strength. In a way you get a house of reps and a sentate in one.
      The way to deal with the problem of stalemate is to give one side the upper hand so that the members thing is more of a veto and might require more than a simple majority (don't you just hate the abuse of the term absolute majority for slender simple majorities e.g. 50.1%).

  3. Sounds a bit like the wonderful model of the US party conventions in terms of weighted and/or split voting.

    I would point out a couple of things though.

    Firstly, representatives elected from more populous constituencies are far more likely to have a larger voting mandate than those from smaller and, maybe typically, insular constituencies so there will likely always be a perceived bias of the urban over the rural.

    Secondly there is the potential, albeit ever so infinitesimally unlikely, scenario that nobody votes in a particular electoral category. For instance a list race could be regarded as so critical that nobody risks voting on the constituency level even on a high turnout. Or the other way around with constituencies eclipsing lists. Highly unlikely of course but we have to consider all the potentialities.

    My own feeling is that this proposed system is a very minority flavour. However I will defend John’s right to suggest this and any other model in terms of creating debate. My personal preference is for d’Hondt in the Israeli model where the country is one huge multi-member constituency and seats are allocated to match the vote in a truly proportional manner as ably explained by James in his recent film. I get the impression that in our modern interconnected world the “local” representative is becoming less important but that’s probably just my spin having lived in Estonia for 20+years where the country is spilt into 12 multi-member constituencies and the place I live has 8 MPs. That’s d’Hondt as well! 🙂

    1. Since drafting this article article it had occurred to me that another (omitted) advantage would be that if you were, say an SNP voter, but had an absolute personal objection to the SNP constituency candidate, you should be able to cast your single vote against the SNP list rather than the constituency candidate - i.e. you can make your point against the individual, while the party still benefits by exactly +1 vote.

      I hadn't thought about whether this might occur on a large scale. Suppose a constituency was pro-SNP but had a very unpopular SNP candidate, we might we get SNP constuency 10,000, SNP list 15,000, Labour constituency 15,000, Labour list 5,000. The constituency would return a Labour constituency MSP, but would still add more voting power in the parliament to SNP. That doesn't appear to me as a problem. Everyone gets what they deserve out of the vote.

      I don't follow why you think a "list race" might emerge. There is no benefit/disbenefit to the party whether the vote is constituency or list, no magic %age threshold to hit, not advantage to heading off another party by +1 vote other than another 1 vote (out of 2 million) in the parliament.

    2. @John
      My intention is not to attempt to arbitrarily dismantle. Instead it is to “what if...” The idea of a “list race” comes down to perceptions. If you only have one vote but also a choice of where to cast that vote then polarising interests or circumstances could end up mitigating certain voting patterns.

      With the benefit of a day or two of reflection I believe that Scotland is certainly too large geographically for a single constituency d'Hondt à la Israel. Instead I would probably favour a single vote, single tier election with a regional list d'Hondt, roughly on the terms of the current list regions with a population weighted number of mandates.

      Currently each region returns between 14 and 16 MSPs but that would need adjustment for population weighting. However the allocation of seats could remain fluid dependent upon the electoral roll prior to an election i.e. the number of seats per region would only be finally settled when the number of voters eligible for any given election has been settled at the registration cut-off date.

  4. I very much welcome a discussion about a better electoral system, but I must admit I'm a bit boring – I'd rather adopt a tried and tested system from elsewhere.

    My preference would be one of the following two options: (1) Fix the AMS system used for Holyrood by adding "overhang seats" (extra seats when a party has got too many constituency seats) – this is what happens in Germany and New Zealand, where AMS has been used for many years. (2) Introduce the system used in Denmark, which is basically d'Hondt (like we know it from European elections) with personal votes – instead of voting for a party, you vote for a person within a party; your vote then gets used twice (once to determine the number of seats each party gets, and again to determine who gets elected within each party).

    I also think we should ideally use the same voting system for all elections – it's barking mad to use four different systems within one country (AMS for Holyrood, STV for councils, d'Hondt for the European Parliament, and FPTP for Westminster).

  5. My own preference is to introduce electronic voting for the electorate. Vote often and regularly on many issues, not just once every four years, and gradually reduce the need for politicians at all. It requires a well informed and motivated electorate, but electronic voting has the potential to be the biggest advance in democracy since universal suffrage.

    1. Unless you can secure it, electronic voting has the potential to allow the easiest coup d'etat ever.
      No electronic system can be the way you want it as it must have a voter verified paper trail (and I mean real paper, not a virtualized one).
      Not to mention that the secrecy of the ballot box is also in question.
      It certainly has to be open source, cos do you really want to trust the system to a potentially biased corporation like Diebold?

    2. @ Graeme

      The Estonian system of voting allows for remote e-voting by digital signature using the voter's encrypted personal ID card. However rather than increase the risk of fraud the opposite is true. But the key to secure voting must always start with positive identification of the electors unlike the nonsense in this country which has been discredited by election monitors many times over as unfit for purpose.

      What e-voting dispenses with is the need for – and risk of manipulation with – postal or proxy ballots as the voter can participate in an election from absolutely anywhere at home or abroad as long as he/she has internet access.

      Whilst Britain struggles with the shadow of suspected postal vote manipulation over 30% of votes cast in the 2015 Estonian General Election were done so electronically.

      Curiously it is also possible to alter one’s vote in that one can log in and re-vote as often as one likes but only the final choice counts towards the electoral total.

    3. @ Scottish economic...
      The Estonian system has a lot going for it and shows that the fears of an ID card need not be overblown if the system is designed for aiding the citizen rather than the bureaucracy. For example I love their ability to review who has been looking at their records.

      Now when it comes to elections, smart card auth is good but you have to bear in mind that not all cards may be up to the latest specifications for signing. For example in 2015 they decided (correctly) to retire certificates that were hashed with SHA-1 as this has known weaknesses. It will take time to replace the old cards. The new standard will be the ESTEID-SK 2015, a 4096-bit RSA key and use SHA-384. how long that will last is not known - presuming that the TLA agencies will have quantum computing long before the rest of us, you can be sure that they'll have the capability to steal elections before organised crime does. I'm sure that Russia will attempt this against Estonia. Without a paper back up, they will never even know. There's also the possibility that compromised software can be installed that uses the genuine cert but tricks the user into thinking that they'll be voting for who they think. This would be detectable if suspected but would it be suspected?? Then there are numerous potential problems with the infrastructure and processing power needed - so DDOS.

      There's ways to reduce this with block chain technology - there's a barcelona company trying just that - but it will always be an issue without the paper trail. it has the potential to undermine trust in the ballot.

      Now this is not a call to not modernise the system but to be careful about putting your trust in technology that few understand.

      Finally the idea of increasing the voting and moving to a more direct democracy. There are fundamental issues about this - e.g. people might not wisely consider their vote. There is also the question of what democracy is - majoritarian rule or a system to shuffle elites and veto unpopular policies...

    4. @Graeme

      Excellent and valuable technical info. As you rightly point out the Estonian authorities are pro-active in improving encryption. And you are right about Russia, there is undoubtedly an entire FSB directorate dedicated to undermining the Baltic States. However it is no coincidence that the NATO Cooperative Cyber Defence Centre of Excellence is based in Tallinn.

      We must not be afraid of technology and shy away from it in our threadbare analog fortresses, not that you suggest that. Instead we need to be aware of what technology offers and leverage it to our advantage.

    5. My suggestion is that the state pay for and/or extend an open source voting system. They should pay several teams of academics to audit the code. This could be done by a consortium of states. They should specify the type of machine hardware to be used and apply the appropriate physical tamper-proofing. The could lead to local companies providing the hardware, lead to diversity of systems and help local economies.

      The system should be touch screen and simple to use and internationalised (multiple languages) and also accessible for the disabled.
      At the end of the voting process (cryptographically secured of course) there should be a two part print out. The voter then verifies that the paper matches what they choose and keep one part and place the second part in the ballot box. The second part should be both human and machine readable so that verification can be both automated and done by hand.

      This makes voting easier for voters, leads to quick results, allows for complicated voting systems but most importantly of all - can be challenged.

      The USA experience of electronic and automated voting is not a happy one.