Wednesday, 10 March 2010

Quotas in Parliament

The upper house of India's Parliament has recently passed legislation reserving a third of the seats in the national Parliament and state legislatures for women. Which seats were reserved would vary between elections - they intend to keep a strict constituency link. It's excellent news, and should very quickly raise them from joint 127th (I count placings differently to the IPU) to at worst 16th and possibly into the top 10 in the proportion of women in Parliament rankings.

One of the issues raised in Shakesville comments on the news is the possibility that the places will go disproportionately to higher caste women. A similar complaint - that the places go disproportionately to white middle-class university-educated women - has been raised around the all-women shortlists used by Labour in the UK.

The obvious solution to this (as Melissa suggests) is just to add some extra quotas, but this gives some tricky problems with most electoral systems. They aren't (mostly) insurmountable, though, if you're willing to make sufficiently large adjustments to do so.

Single-member seat electoral systems

Multi-dimensional quotas reallly require multi-member seats. Otherwise, in a sufficiently diverse society (but one in which the default people have enough over-representation to make quotas useful or necessary), the total quotas may add up to more than 100%.

If the quotas are being applied in the Indian model - all candidates for seats A to H must be female, all candidates for seats I to K must be disabled, etc. - then if the quotas add up to more than 100% (or even close to 100%, if you want to leave a few seats open for default people to possibly get), then you (as the chief electoral administrator) have to do one of two things:

  • Decide that one or more of the quotas is less necessary than the others, and drop it entirely or reduce its size until everything fits, and hope that the people elected in the reserved seats are sufficiently diverse that the originally wanted quotas are met anyway.
  • Assign more specific intersectional quotas - this seat must be filled by a disabled woman, this seat by an LGBTQIA working-class person - controlling the overlap but in a potentially contentious way.

Another option, and one that is used in several countries, is to require that parties' candidate lists meet the quotas, but not place restrictions on where these stand. If you have a problem with discrimination among the electorate, as opposed to discrimination in the selection process, this is likely to end up with the set of elected candidates (as opposed to the set of all candidates) being less diverse than the quotas required.

A third option, as used successfully in Rwanda, is to create non-constituency seats for specific groups. This runs the risk of the constituency seats being far less diverse than they would otherwise have been (though in Rwanda, this does not appear to be the case), but also requires either expanding the constituencies significantly, or expanding the total size of the legislature. Rwanda's Parliament, with the reserved non-constituency seats, is still a very manageable 80 seats total. Adding sufficient extra seats to make a difference to the UK Parliament would mean increasing its size by a few hundred MPs or significantly increasing the constituency sizes.

Multi-member seat electoral systems

Multi-member seat electoral systems are generally better anyway, because they can be proportional to the political views of the electorate, allowing smaller parties more influence, and requiring the governing party or coalition to have the support of at least half the population. There have also been suggestions made that they directly improve representation and diversity, by giving electoral incentives to field a relatively diverse group of candidates to appeal to the broadest possible part of the electorate (rather than needing merely to get a bigger minority than anyone else).

While many electoral systems can be adjusted to use multiple seats, only two that I know of are actually proportional: list systems and Single Transferable Vote.

In these systems, it is easier to apply quotas, as rather than reserving specific seats, it's possible to state that X% of all those elected (rather than all those standing) must be (e.g.) women.

The usual way in which quotas are applied is that the election count proceeds normally until the point (which may never come, depending how the votes are cast) where it would be the case that electing another man (for instance) would make it impossible to elect another woman. At that point you declare that all the remaining men failed to make quota, and carry on the count with whoever is left (transferring votes as necessary in STV).

Where this goes wrong is with overlapping quotas. You can end up in a situation where you technically meet all the quotas, but go significantly against the spirit of them1. Imagine a situation where at least 10% of those elected have to be LGBT people, at least 10% BAME people, and at least 10% disabled people. It's quite possible, applying quotas in the usual way, to have the first 90% elected be non-disabled white heterosexual people, with the remaining 10% having to simultaneously be LGBT, BAME and disabled. As more quotas are added - especially if you break down the quotas into finer detail, the situation can arise where the final places have to be filled by people who weren't standing in the first place despite there being enough diversity in the candidates to theoretically meet those quotas easily.

In STV, this problem is especially severe, because the manipulations to meet quota can skew the original proportionality of the system. Parties can take advantage of this to get more seats for themselves by promoting their default candidates more heavily (because they'll be elected early or not at all) and if there are "suggested preferences" as some STV-using countries have, putting their default candidates further up. It's even possible, especially in a close election, for the order in which the quota effects are triggered to make a difference to who gets elected.

This, of course, as well as breaking proportionality, makes the intersectionality problem above even more likely, and even if it doesn't hardly encourages the parties to promote their candidates.

In practice, therefore, a list system must be used if multiple quotas are needed. I'm not as keen on list systems compared with STV, because they do disadvantage independent candidates significantly, as well as making it impossible to register a lack of support for a particularly bad candidate while still supporting their colleagues in the same party, but as a long-term temporary measure while society is equalised enough not to need the quotas and we can switch to STV, it's not a bad system - and at any rate, is significantly better than most of the electoral systems currently in use in the UK.

A list system with quotas

In a list system, parties put forward lists of candidates, and then are allocated seats in proportion to their number of votes. Exactly how the proportion is determined varies. The method used in UK elections to the European Parliament (except in Northern Ireland which uses STV) is D'Hondt. The first person on each list is given a value of "total votes for that list". The second person on each list is given half that, the third a third of that, and so on. The candidates with the highest values are then elected.

Fictional election.
PartyVotes for listHalfThirdQuarter
Liberal Democrat5100265017001325

Candidates in this election would be elected in the order: Con-1, Lab-1, Con-2, LD-1, Lab-2, Con-3, Lab-3, Con-4, LD-2. List systems need a good number of seats for proportionality to work best - if only three seats were available, the Lib Dems would be a few hundred votes short of their first seat.

In a "closed list" system, the parties decide what order their lists are in. In an "open list" system, candidates are voted for as individuals by the electorate, and their order on the list depends on the number of individual votes they get, and the list gets their combined votes. "Open list" is far more democratic, which is why the UK uses "closed list". For the purpose of applying quotas, it doesn't matter, though - one way or another, we get an ordered list at the end with the top candidate on each party list being the most preferred (either by the party or the electorate), and each list having a number of votes for it that will determine how many candidates in total it gets.

  • Start by excluding from consideration entirely all candidates on a list that didn't get enough votes to have even one candidate elected.
  • Then continue by picking the smallest quota. Go down the candidate list in order of election until someone who meets that quota is found. Declare them elected. This may involve going past the normally elected candidates - if it was an eight seat election on the above figures, but the first candidate to meet this quota was Lab-4, they would be elected, and one of Lab-1 to Lab-3 (probably Lab-3, at this stage) will miss out.
  • Keep going until this quota is filled, and then pick the next smallest quota. Hopefully some of the already elected candidates will have partly filled this already. If at any stage sufficient candidates have been elected from a party to meet its entitlement, exclude the remaining candidates. So, if meeting the quotas so far required electing Lab-1, Lab-3 and Lab-4, then Lab-2 cannot be elected even if they'd meet a future quota.
  • Your final quota is "100% of elected candidates must be candidates", which neatly fills the remaining spaces from the highest people on the candidate list.

Two things can go wrong with that simple method and need dealing with:

  • The quotas applied add up to over 100%, and the people initially elected weren't diverse enough to fill them so the current quota is unmeetable. This is far less likely if the smallest quotas are started with, but could still happen. In this case, I think the fairest thing to do is to find the first unelected person who meets at least two quotas (including from parties which are already at their limit), promote them up their party list to the lowest possible position which would have resulted in them being elected earlier, and restart the count.
    With particularly overlapping quotas, you might need to move up to "at least three quotas". Applying this enough will do one of two things: it will either give you the filled quotas with the most preferred candidates possible to do so, or it will show that it is not possible to meet these quotas at all. In the latter case, treat it as the situation below.
  • There are insufficient candidates standing in the parties that got at least one candidate to meet a particular quota. If any independent candidates remain who should be elected but haven't been yet, elect them anyway.2. The remaining places on the party list are vacant pending an appointment by the party of someone who fits the quotas needed for that place.
    This essentially pretends that the party had put that person on their list at the bottom (or with no individual votes) all along, and hopefully encourages them to do this in the first place next time.

Because we're using a list system, the party proportionality is preserved (though the people elected from each party may not be those originally preferred by the party or electorate). This is much less of a problem than the political skew that STV could introduce.

Quotas, categorisation, and the inevitable mess

One of the problems with quotas that a properly-designed electoral system can't solve is the categorisation problem. People do not generally come in neat categories; marginalised and unprivileged people especially often do not fit the kyriarchy-approved boxes.

Policing of identity is a problem. The seats reserved for trans* people, for instance - are they reserved for anyone who identifies as trans*, or only for those who've jumped through the relevant government-approved medically-certified hoops first? It should be the first, but it'll probably end up being the second, because the rules are set by the existing legislature, who are probably all cis and almost all lacking in clue. Similarly, are all trans women going to be eligible to be elected in the "women" quota? They should be, but it probably won't end up like that in the legislation.

Missing out a quota entirely is another problem. If, for example, disability is missed out as a quota, then disabled people count as "default" for the quota system, which potentially gives them greater difficulty being elected than they would if there were no quotas. (The quota system does this for genuinely default people as well of course, the point of having one being to negate their existing advantage). This can also apply to sub-quotas. A quota based on gender is a good idea, but no actual national implementations so far explicitly reserve spaces for non-binary people.

These problems are not a reason not to use multi-dimensional quotas3, but they are a reason to make sure that the quota definitions are set with full and representative input from the people they're trying to serve. How this is done when by definition if you need the quotas your legislature has too many default people is a tricky practical problem: the legislature doesn't know what "representative" looks like well enough to assemble the group of people that could tell it. Hopefully it can get close enough to approach a solution incrementally, such that at least some people are elected who can fix the problems with the previous definitions and quotas enough to get a properly representative legislature.


1 Probably less likely is that everyone elected under the quotas differs from the default in exactly one way, and intersectionality is mostly ignored. It's unlikely to be quite that extreme, though, with a reasonably large constituency size. It may also not be a problem in practice - the ratio of "female BAME Labour MPs":"female Labour MPs" is similar to the ratio of "BAME Labour MPs":"Labour MPs". It's not a very representative ratio in either case, but the partial use of all-women shortlists at least doesn't seem to have made BAME representation in Parliament worse.

2 Variant: elect the independent candidates at the start of the count since you'll have to elect them anyway, and so potentially make meeting the quotas easier.

3 These problems also potentially apply to single-dimensional quotas - the "missing quota" problem was the start of this post - but managing them becomes far harder as the quota space fills, especially if it exceeds 100%.