Friday, July 16, 2010

On proportional representation (1): the case for

This week and next I'll be writing a two-part series I've been intending to write for a while on proportional representation. This first part presents the case in favour of proportional representation, while the second part will present the case against.


Some background assumptions. Any theoretical project has some assumptions working in the background; I hope that the ones I'm making here are reasonable. If not, let me know in comments and I can back up a step and try to defend them.

I assume that some form of democracy is defensible. I'm reading "democracy" quite broadly, though, to include a broad spectrum ranging from a system where everyone has a say in determining the identity, nature, etc. of all state authority figures (to the extent that there may be no "state" in any recognizable sense), to a system where at least one person has a role in determining at least one feature of at least one state authority, and that person is not awarding that feature of that authority to him or herself. This is just a simplifying move.

I also assume that we're dealing exclusively with representative systems. So, I'm not trying to address the possibility that direct democracy might be preferable to proportional and non-proportional systems. I don't think it is. I think the practical difficulties are almost insurmountable, and I think there are strong reasons to favour assigning the complex task of government to someone other than the citizenry as a whole. But that's a different argument; here, I just assume my way around it.

I'm also not trying to address the possibility that a system of government based on an assembly of delegates would be prefereable to a system based on an assembly of representatives. Now, a delegate is assigned to act in one's stead, while a representative is assigned to act on one's behalf. So, a delegate is someone who carries entirely and only the authority that one does, while a representative carries his or her own authority. A delegate cannot go beyond what one has decided or would decide. In contrast, a representative can, through using his or her own judgement, experience, and so on. I think that rejecting the delegate theory is a difficult and involved task; certainly a task for another day. So, again, I'm just assuming my way around it.

So, what's being assumed here is that we want some system for electing people to some sort of assembly who act on behalf of, but not in the stead of, the citizenry. (Citizens rather than residents because citizens are those who are both responsible for and responsible to the state. Residents are only responsible to the state, not responsible for it.)

SMP vs. PR

I'm limiting my attention to SMP and PR for a few reasons. First, while I could spend some time considering them in detail, straight majoritarian or random selection methods strike me as deeply problematic, for obvious reasons, when applied to elections in complex multicultural societies. A small monoculture might work for simple majoritarianism; but, in such a society, would a representative system really be preferable to a form of direct democracy? As for random selection, it might make sense if there were no relevant differences between potential representatives, but I'm not sure how that could be determined to be the case in an actual electoral contest. Second, I probably should consider preferential systems in some depth. However, the reasons to reject preferential systems differ in kind from the reasons to reject SMP. I'll probably deal with that contrast in a later series.

So, what are the two systems under consideration? On the one hand, we have First Past The Post (FPTP) or Single-Member Plurality (SMP). I like the latter name better than the former, and will use it from here on. The salient reason, in my view, is that FPTP doesn't really identify where the "post" is -- it could be plurality, majority, or a very low threshold. SMP, by contrast, explicitly states that we're dealing with a system based on plurality. On the other hand, we have Proportional Representation (PR). There are, technically, many different forms of PR -- mixed member, transferrable vote, and so on -- but I think I can generalize over them, in a fruitful way, as follows.

What distinguishes SMP and PR, I think, are three features. SMP has:
  1. Single-member districts
  2. Winners determined by plurality
  3. No strong connection between expressed preferences of voting population and representatives
By contrast, PR has:
  1. Single- or more member districts
  2. Winners determined by threshold, plurality or majority
  3. A strong connection between expressed preferences of voting population and representatives
I'm deviating a little from standard accounts here, but not without reason.

With regard to (1), I don't think a PR system has to hold that districts need multiple members. Rather than electing multiple members in order to assure proportionality of representation to expressed preferences of votes, it would be possible to appoint each elected representative as the representative of a gerrymandered district containing just those who voted for that representative. This would be very impractical, of course, but impracticalities should be ruled out a posteriori not a priori. So, when it comes to (1), SMP and PR can be consistent.

With regard to (2), I don't think that PR has to commit itself to any particular stance on what proportion of the vote a candidate needs to achieve in order to be elected. Obviously, there has to be some rule. But that rule will be determined largely be practical considerations. In larger, more populous countries, fewer representatives is probably better, if for no other reason than to make it possible to get anything done. Requiring a majority would limit numbers most strongly, plurality less strongly, and a lesser threshold even less. So, again, this issue looks empirical. So, when it comes to (2), SMP and PR can be consistent.

Now, it might seem that (3) under PR entails the negation of (2) under SMP. This is false, though. We could try to track expressed preferences amongst the populace and elect as representatives those who achieve a plurality of expressed preferences -- as when voters can vote in, say, three rounds. This would be odd, but not impossible. (I'm not sure this exact example would wind up being proportional, but something like that seems possible to me.) So, objections to majoritarianism are not well-founded here, I think. There may be reasons to reject majoritarianism, but this is not where to have that debate.

What follows, then, is it's (3) that's really crucial for distinguishing SMP from PR.

Why does (3) refer to expressed preferences of voters? There's two parts which could be questioned here. Expressed preferences could be replaced by real preferences, or by idealized preferences (such as real preferences counterfactually modified by perfect information and perfect rationality). Similarly, voters could be replaced by citizens. (Preferences I'm not interested in replacing as any candidate replacement for the term -- say, needs -- would move us away from a democratic system, violating one of my assumptions.)

The reason, in both cases, is basically epistemic. I acknowledge that it'd be best, probably, to match the idealized preferences of citizens. This would be best possible form of PR as we'd be addressing everyone who was responsible for and responsible to the state, in terms of what they would want under reasonable and ideal conditions -- so, their aims, goals, purposes, etc., corrected for errors in knowledge, reasoning, etc. would serve as the grounds for selecting the governing representatives.

Unfortunately, it's not possible, from an epistemic standpoint, to work with anything other than expressed preferences of voters. Nothing else can readily be measured. That is, we can't know what the preferences of the citizens are except insofar as they tell us. This means that we have to rely on voters rather than the citizens as only voters have bothered to tell us what they prefer. (Relying on a non-electoral poll has the same problem. We'd have to only deal with the people who had bothered to answer the poll.)

Similarly, we have to rely on expressed preferences rather than real or idealized ones because, while real preferences would probably be better, and idealized preferences even better, we can't get either off the ground without some information. And the best information we have is what people tell us they prefer.

(Although I would support doing away with birthright citizenship and only extending citizenship to those who demonstrated some threshold level of committment to and knowledge of their country. Which might get rid of the voters who sincerely express preferences that have nothing to do with their real preferences, and also bring real preferences closer to idealized preferences. But, again, that's a separate point. Starting to notice how quickly this discussion snowballs?)

The Default Position

The question of whether to favour SMP or PR thus becomes whether we should favour matching expressed preferences of voters in a strong sense, rather than in a weak sense. Under SMP, we do have some matching of preferences, in that those who vote for the eventual winner do have their preferences matched by a representative. It's not strong, though, as an SMP election can (and often does) result in the majority of expressed preferences being entirely dismissed. Votes above the threshold of plurality for the winner count for nothing, and votes for any candidate but the winner count for nothing. Under PR, by contrast, matching expressed preferences is part of the point of the system. After all, as I've been arguing, that's what makes PR interestingly different from SMP.

If you're with me to this point, then I think you should see why the strong connection is superior to the weak one. If matching expressed preferences matters, and it seems it does, then we should only favour the weak connection if the strong connection leads us into problems. If the connection matters, then we should aim for the best connection we can reasonably have. This means, though, that the argument in favour of PR rather than SMP is fundamentally very simple. PR is the default position that a representative system should adopt, at least in comparison to SMP. It's just a historical accident that we wound up with SMP first and would have to move from it to PR. The only way to block this conclusion is to identify some pressing and serious consequences of ensuring the strong connection.

So, stay tuned for Part 2 where I consider attempts to do just that.


Wilf Day said...

Given that about 55% of MPs in Canada were elected from single-MP communities, rather than metropolitan areas large enough to have several MPs, community representation is a serious consideration. The answer, of course, is mixed-member representation, proverbially the "best of both worlds."

ADHR said...

There's a step missing in your argument. Why is community representation important?