Electoral Reform at the Oscars

January 26th, 2012 | Posted by Alex in Democracy - (0 Comments)

(Warning: This is a long post about a mostly theoretical problem with the Oscars selection process. Viewer discretion advised.)

The Oscars are now only about a month away and the nominations are out, with 9 films nominated for Best Picture rather than 10. Being far more an elections nerd than film nerd, the change in the rules that allows this is more interesting to me than the actual results. The view of the academy that led to the change is  that it’s not necessarily the only job of the nomination process to do a good job of aggregating the preferences of it’s members but also that any film nominated should have some claim to be an outstanding film worthy of the honour (and in any given year there might be a different number of films that qualify). Given this we get a new Rule 2 for the Best Picture Nominations:

The pictures receiving the highest number of votes shall become the nominations for final voting for the Best Picture award. [a] There may not be more than ten nor fewer than five nominations; however, [b] no picture shall be nominated that receives less than five percent of the total votes cast.

Now this rule is really the reason for the post, because I’m not convinced that the system they use is capable of ensuring both a and b – but it’s going to take a while to get there.

In the first place a system that can have a variable number of winners poses a bit of a problem for designing an electoral system as typically these are mechanisms for distributing limited goods (seats, awards, etc) where the number of these goods is fixed – you may not know before the election who’s going to win, but you do at least tend to know how much winners there going to be. Before we look at exactly how they’ve achieved this it’s helpful to look at how the old system works because it’s been cannibalized to produce the new process. I’m building on Steve Pond excellent series of posts on how Oscar votes are counted (in both old and new methods), but I’m going to restate of what he said because some of his terminology is a little off and to be more precise about the process (which may or may not be the same thing as being clearer).  Basically Best Picture nominations used to be decided on a slight variation on Single Transferable Vote.

The Old Method

  1. Gather all ballots
  2. Eliminate films with less than 1 first preference vote.
  3. Calculate Droop Quota for number of seats remaining (start at 10).
  4. Sort votes by highest valid preference
  5. Elect any candidates above the quota.
  6. If a candidate’s votes are more than 20% above quota, revalue all votes with a new value of surplus/total x current value.  If a candidate is between 100% and 120% of quota, discard votes.
  7. If no candidates have been elected, eliminate the candidate with fewest votes.
  8. Repeat (3 through 8) until all seats are filled

The only atypical bits are Steps 2 and 6. Step 6 requires an extra 20% beyond the quota before redistributing votes, which I’m assuming this is a practical measure so they can count by hand and not have to bother moving hundreds of votes around with very small values, while Step 2 is a weak version of a requirement for initial support that’s the main purpose of the new system. Otherwise this is a pretty typical STV process, I just write it out because of how similar the new method is:

The New Method

  1. Gather all ballots
  2. Calculate Droop Quota for number of seats remaining (start at 10).
  3. Sort votes by highest valid preferences
  4. Elect any candidates above the quota
  5. If a candidate’s votes are more than 20% above quota, redistribute all votes with a new value of surplus/total.  If a candidate is between 100% and 120% of quota, discard votes.
  6. Eliminate any candidate with less than 1% of the vote and redistribute to their next highest valid preference.
  7. Eliminate any candidates with less than 5% of the original vote total (including votes spent through electing a candidate).
  8. If  Number of remaining candidates (including elected candidates)  <= 10 then
    1. Elect all remaining candidates.
  9. If Number of Candidates > 10  then
    1. Calculate Droop Quota for 10 seats
    2. Sort votes by highest valid preference
    3. Elect any candidates above the quota
    4. If a candidate’s votes are more than 20% above quota, revalue all votes with a new value of surplus/total.  If a candidate is between 100% and 120% of quota, discard votes.
    5. If no candidates have been elected, eliminate the candidate with fewest votes
    6. Re-calculate Quota.
    7. Repeat (9.a through 9.f) until all seats are filled

If putting it like that is confusing, John Young’s description might be a better read.

Now this process is on the face of it a bit odd, it’s a hacked-open version of STV that retains a quota but has an escape halfway through – it runs one round of STV, then abruptly eliminates a number of candidates. The old requirement on at least one first preference has been beefed up and moved further down the process  - why not discard candidates with less than 5% of prefs right at the beginning and only bother with any preferences if you end up with more than ten candidates left? As it turns out while it looks messy, the opening part of the process is quite important.

The goal here isn’t just that a film should have a fair amount of initial support but also that there must always be between 5 and 10 candidates. The first part is quite easy but as it requires setting an absolute standard for ‘initial support’, it makes it hard to guarantee that there will be at least 5 candidates.  The problem can come from two directions: over-concentration – where four or less films receive so many of the first preferences that it’s impossible for any other film to reach 5%, and dilution – where the vote is spread over so many films that less than five reach 5% at all.

Over-concentration

Retaining the early stv-esque stages of the process is essential to preventing over-concentration. If a few films are locking up all the votes they’re going to be significantly over-quota and so keeping the quota and redistribute rules mean that their surplus votes will be spread around the other films, making it much more likely that more than five films will make it through. So while the official press release says a film must have 5% of first preferences this isn’t strictly true, a film’s 5% can come through a mixture of their own votes, redistributions from unpopular files and the surplus of very popular films. The take home point there is that second preferences might have been important even when less than ten films emerge, as the second preferences of front-runners or <1%ers might have tipped a few more over with their surplus.  What isn’t clear is if they run steps 4 to 6 in a loop until all possible candidates have been elected before proceeding to the 5% rule, as steps 5 and 6 could push films above quota. If they’re not doing this looping, when say three films dominate 90% of the first preferences, and those three all give their second preferences to a fourth film then you’d still end up with less than five films. This isn’t necessarily a ridiculous example, as at that point any preferences for films 1-3 don’t need to specify 4 as a second preference to cause the problem, 4 simply has to be the highest valid preference once 1-3 have been removed through election.

Dilution

From the other end is dilution, which is more unlikely but not at all negated by the system.  If we imagine a situation the first preferences are evenly split between 21 films then all are above 1% so no redistribution takes place and none are above 5% and there are no winners at all. This is obviously an extreme outcome and it’s really unlikely that voters will be that split on their film selection –  however the system itself creates voting incentives that could lead to the vote being more spread out than voters’ actual preferences might suggest. While Pond thinks that most Oscar voters aren’t clued up enough about the system to strategize their vote (and he’s probably right) for the purposes of exploring the system a bit, I’m going to go into ways this system could be played.

The 5% threshold is around half the initial quota that a front-running film would have to clear in the first round, so in this sense for a film to even reach the quota in the first round means that half its votes are wasted as they do nothing to change the result. If you’re a voter that has several films you want to be nominated (say two, Film A that’s a shoe-in and you like more, and Film B that’s chances are more dubious) and you want to maximize the value of your vote, listing your preferences honestly 1st and 2nd isn’t the way to go. If there are 1000 votes, and 125 voters vote A>B (these are the only voters who have preferences for these films) then A will be nominated and 25 votes will be redistributed to B – this falls short of 5% and B will not be nominated. However, as A can be nominated with 50 votes, then those same voters can split their first preferences 75 for A, 75 for B and they’d both be nominated. In real life if you were one of those voters you’re not really able to coordinate with all the others on the optimal distribution of your votes, which means your best course of action is to put B as your first choice and free ride on the other voters to elect A anyway. The trouble then comes is enough people do this for B (and  C, D, E, etc) then even an obvious and wide-supported nominee like A might fail to make 5%.

Now I think this is an unlikely outcome but given a clued up electorate it is at least possible that only four films might make it through. That possibility makes Rule 2 flawed as the requirement for 5% support is in conflict with the requirement for at least five nominees. Strictly speaking the rule should prioritize one requirement over the other, either making the 5% rule absolute (allowing less than five nominees to emerge) or the nominee number absolute (requiring that the old rules kick back into play if there are less than five over 5% as well as if there are over ten.)

Better methods?

So given these (largely abstract) problems, is there a better way to count? In terms of improvements to this system, the looping I talked about earlier should definitely be used if it’s not already and in step 2, a quota of 5% should be used instead of the droop quota as if even two films get in through quota in the first round that’s 10% of the votes unnecessarily removed from circulation, which only increases the risk of dilution. In terms of doing something completely different it’s worth thinking about what we’re actually trying to accomplish here. STV is a good way of handing elections with many candidates for many positions but the problem in this case is it can produce compromise winners, who are few people’s ‘great!’ choice but many people’s ‘alright’ choice – what we want is a way of discriminating against these winners and using first-preferences is a good way to that (and is used in other STV systems, for breaking ties in the ERS97 rules for example). As we’ve seen these rules aren’t doing that cleanly, as many second-preferences can also come into play in this ‘initial support’, is there a better way to do this?

My solution would be to run the old rules to produce 10 nominees – but then to remove any nominees who are not in the top 10 of first preferences received.  This seems to me accomplish the same basic goal (variable number of winners based on initial support) in a more straightforward way. To ensure that at least five nominees survive, the exact rule should be that ‘Nominees should be checked to see if they should be eliminated in inverse election order (and then within rounds, the nominee with smaller surplus checked first) for a maximum of five eliminations.’ My gut instinct says it’d be more likely to be around the high end of 8-10 rather than the low end but given that this year 9 nominees came out, that’s probably not that different in most cases and as it has the possibility of chucking candidates out and so accomplishes the same basic purpose.

In short, Rule 2 potentially contradictory, new process weird but accomplishes a purpose, might be better ways to accomplish that purpose, Oscars aren’t made of solid gold, and Oscar himself stands on a film reel which one day will be as wonderfully anachronistic as the floppy disc you click when you hit ‘save’. That about covers everything I think.

,

Scottish Independence and the Electoral Register

January 15th, 2012 | Posted by Alex in Democracy - (0 Comments)

The BBC reports that Westminster is claiming a Scottish independence referendum wouldn't be able to use the full electoral roll, which seems a bit strange:

Scotland Office Minister David Mundell said: "The UK government has set out very clearly that the Scottish government cannot legislate lawfully for a referendum with its current powers.

"The full electoral register can only be used for specific purposes such as the general or Holyrood elections.

"The legal position is clear – if the SNP attempts to go ahead with its own referendum then it will be nothing more than an opinion poll, with a million Scots unable to be asked for their view."

If true this means many potential voters could be excluded as they're not on the publicly available (reduced) register. As a layman I've had a quick skim of some of the acts related to the electoral register and according to the explanatory note of the Representation of the People (Scotland) (Amendment) Regulations 2002:

A copy of the full version of the register may not be supplied by a registration officer otherwise than in accordance with an enactment (which, by virtue of the definition in inserted regulation 92(2), includes subordinate legislation) including the principal Regulations (inserted regulation 94(3)).

The effect of which includes an enactment to be  'any provision of, or of any instrument made under, an Act of the Scottish Parliament' via the Representation of the People Act 2000.  Which, again I might be misreading, I take to mean the Scottish Parliament can in principle have referenda using the electoral roll. If they can use it for even a non-binding question about a reserved power is contested.

As Matt Qvortrup sort of argues in the Scotsman, the international recognition of self-determination (which I think is less strong than he suggests, but going with it) would likely mean in the event that Scotland holds a democratically legitimate but illegal referendum, it would likely have to be recognised by Westminster – which is why moves to de-legitimize the potential vote by arguing people will be excluded make sense, and as I've said before the SNP should be careful to avoid ambiguous questions that give rise to these kinds of challenges.

Update: Ah, clearer copy of SNP statement that seems to support my point.

However a spokesperson for the Scottish government rejected Mr Mundell's and noted that the referendum would be carried out after the appropriate legislation was passed at Holyrood, ensuring full access to the register.

, , ,

Progress of the Boundary Reviews

January 12th, 2012 | Posted by Alex in Democracy - (0 Comments)
I went to a talk by Ron Johnston, Charles Pattie and David Rossiter at the Constitution Unit yesterday on the in-progress boundary reviews (their slides at link). It was interesting in the sense that it confirmed with figures the kind of problems they’d previously predicted with the review and I’m just going to throw down a few notes while it’s fresh in my mind.

Building blocks and displacement

This review represents the triumph, though possibly only temporarily, of the mathematical principle (all constituencies should have equal numbers of electors) over the organic principle (constituencies should reflect real communities and locations) with a tighter than ever constraint on elector numbers. A basic problem of requiring constituency electors to fall within a certain range is that the building block used (wards) are too often too big to construct constituencies within the range.

Interestingly the English and Scottish commissions have different ways of dealing with this, the English method being to cross borders to make up the numbers, whereas the Scottish commission splits wards.  They pretty conclusively demonstrated the Scottish method is less disruptive, measured by relative numbers of ‘suboptimally placed (SP) electors’, who are electors displaced from the majority of their compatriots in their local authority, old constituency or ward. This does make sense intuitively, as crossing borders pushes the imbalance over into the next region and requires aggregate more slight movements of voters, but it’s good to have numbers on it.

The idea that SPs should be minimized is a good measure, but I wonder if the principle could be extended. If the proposed ‘one review a parliament’ schedule is kept up, in the worst case you could find voters who are moved around at every election – another possible measure would be to try and minimize the number of people who have been SPs more than once (TSPs?).

As the general trend is towards larger wards, they suggested using polling districts as a possible smaller building block. The other option they suggested in 2009 (with McLean) was postcodes (p. 464).

Public Hearings and political effects

It was interesting to hear that the public hearings (essentially curtailed versions of the public inquires) are still a vehicle for parties even if their ability to input has been reduced (limited to 40 minute slots). This isn’t entirely surprising as parties have the greatest interest in the result – I’m going to bring in here a bit here something I wrote at the time the bill was going through about the relationship between parties and the inquiries:

As boundary matters are decided in the UK by the apolitical Boundary Commission, what happens is whenever we have a debate about apportionment it rapidly becomes a debate about competing metrics of fairness, where each side just so happens to align with the position that's seen to result in a better deal for them. Natural population movement tends to lead to more boundaries that favour Labour over time and so in general Labour has been opposed both to frequent redistributions and now, very clearly, stricter equalisation. It's not a coincidence that it's the same party that postponed the acceptance of new boundaries until after the election in 1969 is now dragging it's feet in the Lords but equally the Conservative position isn't much deeper, if situations were reversed I've no doubt we'd see the same in the other direction. However, that a position has political advantages doesn't stop it being a correct one.

…

It's interesting that one of concessions given in the Lords seems to be the return of public inquiries as I thought that removing them was one of the more interesting elements of the plan. Public inquiries have a number of issues, they make the process very time consuming (which is an issue if we're going to have to have a review every parliament), tend to increase deviations from the quota size, but most importantly public inquires are one of the few ways for political parties to influence the decision of the Boundary Commission. What you sometimes get is that political parties will show up at the Inquiry with alternate plans, which as long as they fulfil the commission's criteria better than the Commission's own plan it has no grounds to reject on the basis that it's more partisan1 . Being apolitical means it can't consider the political consequences of it's work, this is far from saying there are none. Given the inherent political nature of their work, Richard Morill makes the interesting argument that political ends should be directly considered by apportionment bodies and that they should set out to create certain numbers of safe and marginal seats rather than pretending it's work has no impact on these questions, but it's hard to see this view gaining much currency in British circles.

Relationship to electoral system

Something that was danced around in the Q&A is the way in which the boundary commission is related to the voting system. In many way the boundary commission is the dark and complicated back-end of the ‘simple’ FPTP system (or any single-member system like AV) in that single-member constituencies are smaller than multi-member constituencies, and so are more vulnerable to population shifts (and the problem of building block size mentioned earlier), plus the all-or-nothing nature of the systems make the partisan implications of boundaries a bigger issue than they might otherwise be. This isn’t a hugely live issue (as was pointed out the boundary proposals might be unpopular among MPs but changing the voting system is even more so) but it’s worth remembering it all fits together.

While district magnitude and electoral system have effects on the importance of boundaries, something that’s thought about less is that ultimately the conflict between the organic and mathematical criteria are the result of the fixing representative voting weight at ‘1’.  Trying to express both through the same measure (boundaries) inevitably requires one to be prioritized over the other, what we need to do is open up another stream to get the information into the chamber.

What Toplek proposes2 in the context of the US is rather than endlessly debate methods of dividing seats among the states on the basis of their population, use any method but then give a state’s reps a voting power that is equal to the population of that state. Using representative voting weight as the means of expressing mathematical equality frees up the boundary process to deviate in size to allow more ‘natural’ constituencies (even extreme deviations in size could then be permitted within the system without being exceptions).

This doesn’t resolve all the problems of boundaries (safe seats, etc) but it does resolve the current debate in a much easier (and cheaper) way.  Beyond that fix, [laying with representative voting weight opens up a whole new realm of possible electoral structures, including methods of dynamic apportionment which reduce the need for boundary reviews even further and more exact methods of proportional representation. But that’s a matter for future posts.

  1. Lots of interesting discussion of this in D. J Rossiter, R.J Johnson, and C. J Pattie, The Boundary Commissions: Redrawing the UKʼs map of Parliamentary Constituencies, (Manchester University Press: Manchester, 2009), pp. 246-262 ↩
  2. Link takes you to pdf, but ref is: Jurij Toplak, ‘Equal voting weight of all: Finally "One Person, One Vote" from Hawaii to Maine?’, Temple Law Review, Vol. 81, 2008, pp. 123-178 ↩

The Revolution will not be Amplified: Uses and Abuses of the Human Mic

November 23rd, 2011 | Posted by Alex in Democracy | Protests - (0 Comments)

The idea of the 'human mic' was one born out of a practical necessity, the police prevented the first OWS protests from having electronic amplification so to compensate the mechanism of the HM emerged, where the person currently with the 'mic' would say a short phrase of their statement, which those around them would shout, before the speaker moved onto the next phase and so on – there's a good example here:

In addition to solving the initial amplification problem Mark Lance and Sydney Levy at the New APPS blog point out that this has some quite interesting effects on the nature of discourse through the HMs:

What is most interesting about the HM is the way that it shapes the time and space of a discourse.  Of course the most obvious feature is that it slows things down.  Everything is said twice.  This is a mixed blessing, no doubt, stretching out already long consensus meetings.  But at the same time, we must remember that we are dealing with a huge, diverse crowd of people, many of whom are completely new to the very idea of directly democratic decision-making, and none of whom have much experience with this group.  So slowing things down and repeating everything that is said before it is possible to respond also functions to draw people in and lends an air of seriousness to the comments of others.

Similarly, the practice eliminates side-conversations.  If you participate in the HM, you have to listen so that you can repeat.  You can't be chatting snarkily with your neighbor, or even focusing on a text.  Neither can folks talk over one another as they could if, say, everyone was mic-ed electronically.  The HM can only repeat one thing at a time.  So there is a built in technological imperative of waiting one's turn.

…

HM as a practice seems to function to "decenter" conversation in large groups.  You don't have to be on the stage, at the mic, or in front to be a part of the conversation of the whole.  And even in the most inclusive groups, the process of coming up to the front will be disproportionate intimidating to some and,  in the case of some disabilities, physically difficult.

HMs on the face of it seems to address the classic problem that people aren't really listening if they're working out what they're going to say next by keeping the audience tasked on repeating the current message. While someone with an electronic mic is addressing a passive (and non-tasked) audience, they might disengage and begin side-conversations if they find the speaker uninteresting, or challenge the speaker for control of the sound-space by shouting or booing. A tasked crowd is less able to do these things1. This raises a question: as the form of the human mic requires people to take part in other people's words, what do people do when they disagree with the message? Dan Kervick gives a possible answer:

 At least at the GA I attended, it seemed to be understood that you would try to repeat even those words you disagreed with. People were repeating words even as they were giving down twinkles of disapproval. However, even while repeating the words, you might not say them with much enthusiasm. Sometimes people would repeat the words fully, but in a "yada, yada, yada …" tone of voice. The force and enthusiasm element of the repeated words seems itself to be part of the consensus-building aspect of the assembly discourse. The speaker's rhetorical direction is influenced in some directions and away from others by the audible and continuous rising and falling of the assembly's support, and by the twinkles. [...] one guy got very vulgar and self-indulgent, and the repetition flagged in energy and took on a tone of "C'mon man, stop being an ass."

This might not be universal but in this case rather than cancelling or drowning out the original speaker, people who are part of the microphone instead modulated the original voice to express their opinion- the message and dissent to the message become one and the same thing. This means it isn't content-neutral as a medium, but also that people calling it "the people's mic" aren't being entirely silly, as there's an extent to which the crowd can imbue a statement with their own feelings about that statement (although it clearly can't apply past the people taking part).

In addition to the decentering point made above some in the comments at APPS make the point that this is a form of amplification that's slower and so friendlier to those who might not otherwise want to step up and speak, but at Occupy Patriarchy Lucinda Marshall argues that the human mic isn't necessarily that empowering:

Listening to people yell, “Mic Check!” at Occupy locations throughout the country, it is hard not to observe that those with the loudest voices are the ones who really get heard with this system, and those voices usually are male baritones.  Talking to women here in Washington and also reading reports from elsewhere, it is clear that many women find this system of having to yell at the top of your lungs to be one that is an uncomfortable way to communicate and participate.  Some women report being harassed when they speak, and even of mics being grabbed from them.

This doesn't seem to be a universal experience but it's a problem that should qualify descriptions of HM's democratic virtue –  for all its merits it is as heart a means of amplification and that leaves it vulnerable to existing disparities in voice. Still, it's a method that has potential to enhance the deliberativeness of public conversation and open up and democratize (to some extent) the sound-space, and so is far more than the simple yelling it might appear to be at first.

Given this understanding of what the HMs are supposed to do, it becomes clear that some groups aren't really using them correctly. Case in point,  a group of protesters at Cambridge University used what they claimed was a "people's mic" to shut-down a David Willetts event, but if it was it's a much poorer conception of the mic than the one emerging as their aim wasn't to conduct discussion but to close it down, not opening up the sound-space but filling with white noise to make it unusable. As the form of the human mic becomes better established it'll be interesting to see if it keeps its original deliberative spirit.

  1. Another possible factor is that HMs are imperfect methods of amplification and so can't tolerate as much other noise, and so people are obliged to be quieter to let it work at all ↩

,

Scottish Independence and Ballot Design

November 11th, 2011 | Posted by Alex in Democracy - (0 Comments)

ballot

This is interesting:

One of the other issues which Mr Mundell has been exploring is whether the Coalition Government might be able to amend the current laws to force the Scottish Government to ask a clear Yes/No question on independence.

Mr Salmond has made it clear he favours a three-option referendum with the Scottish people being asked to back full independence, the status quo or "independence lite" – a settlement which many unionists believe would leave Scotland independent in all but name.

There have also been suggestions from the Nationalists that Scots would be able to choose more than one option in the referendum – prompting critics to warn that the SNP could secure independence even if this was only the second-best supported option.

Peter Kellner of YouGov has a good run down of all the possible ways of asking this question apart from two-questions. Although the SNP seems to be backing down from a two-question route they're leaving a space open for it to return, and ERS Scotland is advocating a different form of the two-question format, making it worthwhile having a look at exactly what the problems are with it.

One possible set of questions (Set 1) would be:

1: Do you prefer independence to the status quo?

2: Do you prefer devo-max to the status quo?

This question choice could lead to the situation where both independence and devo-max receive majorities (or even, for the purposes of argument, super-majorities) but devo-max receives a larger majority than independence – what then does the referendum say should be done about Scottish independence?

One answer would be that the Scottish government would at that point be well within its rights to declare independence, a majority has declared its support.  The other way to argue it is that when there are two possible consensuses on offer, a democracy should seek to create the largest possible consensus – which in this case is devo-max and it'd be wrong to leave the union when a strong majority is in support of it. As a convincing case could be made either way Set 1 creates the possibility of an ambiguous result, which is bad.

There is something seemingly nonsensical about the idea a majority can be in favour of both the union and independence which gives us a clue as to where Set 1 went wrong. The problem is that although it seems like we do, we don't actually have information on if people would prefer independence to devo-max or vica versa. If 'status quo' is A, 'independence' is B and 'devo-max' is C, we have information that B and C are superior to A (to differing degrees), but we can't from this deduce people's ordering of B and C. If you replaced all voters who preferred C to B to A with voters who preferred B to C to A the questions in Set 1 would still return exactly the same result. Given in the situation above we're trying to work out which winning claim is superior, that all the people who actually have opinions on this can change their mind without affecting the result is a problem.

If the SNP want to use this question they're really shooting themselves in the foot. It's highly likely that if independence can win a majority then devo-max will be able to win a higher majority and so make any resulting move towards independence appear illegitimate – which it may or not be because we have no way of knowing which of B or C voters prefer. One alternative set of questions (Set 2) would be:

1: Would you like a change to the status quo?

2: Would you prefer independence or devo-max?

This set was pretty much the format used for the 1997 Scottish referendum and one of the possibilities that was occasionally floated for an electoral system referendum. Willie Sulivan at ERS Scotland argues that:

“If there is only to be one question then the structure is straightforward; a simple Yes or No for Independence.

“If the consultation responses indicate that the Scottish people want the opportunity to vote on all three options, then the best way to do this would be to ask two questions on the same ballot paper: The first asking yes or no to any change and the second to determine what level of change. This approach was recently road-tested in New Zealand.

“This is by far the simplest and fairest way to ensure that everyone gets a real say but we are concerned that instead of ensuring that it’s the Scottish people who have a chance to determine Scotland’s future, the debate among politicians is becoming polarised and disingenuous.”

But I think especially in the context of this referendum, this question set can end up treating a particular group of voters unfairly – while at first glance this seems to solve the problem above by giving a clear choice between B and C, but it doesn't allow people to place A(status quo) as a middle preference. Say we get a majority in favour of change in question 1 and a majority in favour of independence in question 2, we have information that a majority prefer B or C to A, and that a majority prefer B to C. The problem is that the ballot design assumes A is either a voter's first or last preference. If question 1 could be swung by a single vote but question 2 already had a majority for independence, a voter who wanted devo-max, but doesn't want independence (and so has the status quo as their second choice) will vote yes on question 1 – giving independence victory despite the fact that that was not their intention.

From these we can see that it's very hard to get people to choose between more than two options using two yes/no questions and that there are grounds for restrictions on the kind of questions and not just the wording that should be asked in referendum (although this should probably be done by the Electoral Commission rather than a political stitch-up in Westminster). This leaves either a single yes/no question on a single issue, adding another yes/no question to make it a Condorcet vote as Kellner recommends, or preferential voting among multiple options using AV – of which the last for me seems the simplest method of better way of determining what people actually want. Obviously this was a non-starter in an AV referendum as conceding that voters are capable of dealing with multiple preferences would have been giving the game away, but many Scottish voters already use preferential voting with STV in local elections so there's no real reason why it couldn't be used – 'simpler' formats really aren't simpler.

(This post was updated on 26/01/2012 to include references to various groups as statements have been made and evolved).