Democracy counts. Should rulers be numerous ?

Tuesday 18 May 2010, by David Estlund

Thèmes : démocratie

Version imprimable fontsizeup fontsizedown

In this essay [1] I want to consider the moral significance of numbers in normative democratic theory. In Part I, I suggest that the feature of generality in liberal approaches to political justification leaves open many questions about how numerous the rulers should be. Even once democracy itself is assumed, the question of numbers is far from settled. All democracies employ subsets of the citizenry for much of the work of legislation and policy. Even supposing that they must be somehow democratically authorized that says nothing yet about how numerous they should be. In Part II, I consider a handful of expanding and shrinking factors in order to make several observations about how these countervailing considerations might be thought to interact. There are no practical conclusions in the offing. My modest aim is only to exhibit some of the complexity about numbers that remains even when the principle of democracy, or inalienable popular sovereignty, is taken for granted.

I. From Liberal Republicanism to Representative Democracy

1. Sovereignty and Numbers

The idea of democracy is partly about number and partly about class. The term demos in ancient Greece, referred to the class of ordinary people with no special qualifications to rule. Those with special qualifications were thought to be small in number as well, but that’s a separable point. The main appeal of rule by the experts was certainly not that they were few but that they were expert. The appeal for democrats of rule by the demos, the many who, by definition, have no special ruling expertise, can easily look puzzling framed in this way. What do they have going for them other than their numbers, and yet how could that possibly be a reason for them to rule? Plato raised this question repeatedly. [2] In the Protagoras he considers the suggestion that no elite should rule because in matters of political wisdom no one is much better than anyone else, but this is highly implausible (and Plato obviously does not himself accept it). [3] It conflicts directly with the obvious fact that some are so much worse than others. It’s doubtful that this could have been behind many people’s democratic convictions. So perhaps there is some-thing to the numbers themselves.

But Athenian democracy, while direct in its way, was small in comparison to modern states, provinces, and even many cities. Not only did it restrict participation to an elite subset of subjects; even if it had enlarged the franchise to cover all resident adults the numbers would not have been very large. Assuming, for simplicity, that enfranchising non-residents is out of the question, if democracy is preferable to oligarchy partly because of the larger number of rulers (we don’t yet know why this is an advantage), we might expect partisans of democracy to recommend that political communities be larger rather than smaller. Of course, traditionally they have recommended just the opposite. Rousseau is exemplary when he insists that even though every subject ought to share equally in sovereignty, democracy requires that political communities remain small, like Athens or Geneva. And yet, Rousseau unequivocally opposed reducing the number of active legislators by having some smaller set represent the others. Should the rulers be numerous or not?

Rousseau’s deeper principle, of course, was generality, not numerosity. Society, he insisted, must be governed in accordance with a general will, roughly the shared contents of every member’s particular will. Here generality is the crucial thing, not numbers. The principle which should guide government must be in accordance with the will of each citizen (at least so long as their will respects the same status for each other person’s will; call this constraint reasonableness). I’ll call this principle of justifiability to each, “liberal republicanism.” I will call the status of being owed such a justification the status of citizen.

Liberal republicanism involves a kind of general will. The requirement that an arrangement must be justifiable to each citizen, insofar as they are reasonable, will typically rule out a lot of arrangements. An arrangement meets this standard only if some part of what is justifiable to me overlaps with what is justifiable to you, and this overlap also overlaps with what is justifiable to the next person, and so on for every single citizen. If there is anything that meets this demanding standard it will be a certain shared content of every citizen’s “acceptability set,” or a kind of general will. Without saying a lot more than can be said here about the content of the proviso by which rejection is only counted if it is “reasonable” it is impossible to say whether there is any hope of any arrangement ever meeting the standard set by liberal republicanism. On the other hand, the word “reasonable” should not be assumed to mean what it means in ordinary language. It is a place holder, with its content to be filled in by substantive moral argument about what sorts of grounds of rejection ought morally to be decisive in scuttling a proposal’s legitimacy. A reasonable rejection, as I’m using the term here, is simply any rejection with this moral status. When I say, in what follows, that some certain ground of rejection is reasonable, then, I should be taken only to be asserting that it is one that should be taken to violate the liberal republican standard of justifiability to each citizen. This is always a substantive moral claim, not a derivation from some prior (let alone complete) specification of what makes a ground of rejection reasonable. So far, all I have said about the content of reasonableness is the vague and indeterminate proposition that a ground for rejecting an arrangement is unreasonable whenever it is incompatible with recognizing the equal citizenship of all other citizens—their own right to be given an acceptable justification. Proposals simply to use certain others for our own benefit as if they were tools clearly violate this requirement, but beyond a few simple cases like this the requirement would need more interpretation.
Liberal republicanism precludes justification in terms that any citizen could reasonably reject. I will call the set of considerations acceptable to all reasonable citi-zens as counting for or against a law or policy the domain of public reason. [4]

I believe, and will simply assume here, that Plato’s claim that certain people ought to be sole sovereign owing to their superior wisdom is precluded by public reason. [5] There is no criterion or indicator of the moral wisdom required of the sovereign that must be accepted by all who recognize the equal citizenship of others. Epistemic arguments for political arrangements are not necessarily all ruled out, but invidious comparisons (as I will call them) are. [6] This leaves what we might call “structural” epistemic arguments, since at least these do not involve invidious comparisons. By a “structural” argument, I mean an argument that sup-ports the epistemic value of some institutional arrangement in a way that does not depend on the use of any criterion by which the wiser citizens can supposedly be identified. For example, institutions protecting and encouraging freedom of expression might be held to promote the epistemic value of democracy in certain conditions, but this does not imply any invidious comparisons. [7]

The question is whether structural epistemic arguments within public reason support popular sovereignty. The move to structural epistemic argument by no means guarantees this. Suppose that there were publicly convincing arguments independent of any invidious comparisons among citizens that justice would be best promoted by a dictatorship. Hobbes’s view, for example, recommends unlimited non-popular sovereignty as the most conducive to justice, but not on the basis of any supposedly superior wisdom of certain potential rulers. The problem with Hobbes’ argument that is usually emphasized, following Locke, is that unlimited sovereignty poses such a threat to citizens that it might be even worse than the state of nature. [8] Hobbes is right, on the other hand, to notice the difficulty in enforcing limits on sovereignty, but suppose that this problem can be solved. So assume that sovereignty must be limited by some form of constitutional law. This does not yet show that the sovereign could not be a limited autocrat such as exists in a constitutional monarchy. The need for limits on sovereignty is Locke’s main step beyond Hobbes, but it goes no distance toward explaining the requirement of popular rather than elite sovereignty. Even if invidious comparisons are ruled out, there might yet be reasons for rule by one or a few rather than by many, all under the assumption of constitutional limits on the legitimate exercise of power. And while it is not Hobbes’s strategy, a Platonic view chastened by the need for constitutionally limited sovereignty, and also by the rejectibility of invidious comparisons, might yet propose structural epistemic arguments for elite rule. [9]

So the question of how many rulers there should be could be said to arise, in the first instance, at the level of determining the location of sovereignty. It seems, best, though, not to refer to this as a point about rulers at all. The reason is that it conceals the step from popular sovereignty to a principle of democracy, a step that requires argument of its own. More on that in a moment, but let me propose some more terminology that will be useful in focusing our attention on this crucial step. By the sovereign body I will mean the set of individuals whose actual acts of authorization are required for legitimate rule. Liberal republicanism, recall, is the requirement that justification be acceptable to every reasonable citizen (let generic republicanism mean the weaker thesis that justification requires appeal to the good of the governed, not necessarily one by one). It is important to note that:
Liberal republicanism does not logically imply popular sovereignty.

By popular sovereignty I shall mean that no citizen has a greater share of sovereignty than any other, and only citizens are sovereign. It is perfectly consistent (though entirely contestable) to say that the general will (as we might call it) implied by liberal republicanism might best be ascertained and promoted by some form of elite rule—rule by those most expert at ascertaining and implementing the general will. The set of citizens, understood as those to whom justification is owed, may or may not all be among the legitimate members of the sovereign body, understood as those whose actual wills are equally weighted in authorizing law and policy.

While the set of citizens can, in principle, be larger than the legitimate sovereign body, Rousseau, for one, did not take this view. He held not only (as a good liberal republican) that society must be ruled in accordance with the completely general will, but also, and separately, that the job of discovering the general will is not to be put in any elite hands but should be performed by extending the franchise to everyone. But why? The equal moral standing of persons in politics is recognized already by liberal republicanism. The next task is to discover what collective arrangements can meet the liberal republican requirement of no reasonable rejection by any subject. Why think that the general will is best ascertained by letting everyone equally participate in ascertaining it, even when it seems obvious that some are better at ascertaining it than others? Here it may be that the larger numbers are an advantage, though it is simply unclear from Rousseau’s text whether this was his reason. [10] If it were it would be harder to understand why he thought that political communities must stay small. (Perhaps this is a condition, for contingent reasons—such as the ways communities divide into factions—on a group’s having a general will at all. [11] If so he would be free to hold that so long as there is a general will, the more voters the better.)
It is a substantive question, not settled by the concepts, whether only popular sovereignty could be justifiable to every (reasonable) individual. Hobbes, for ex-ample, accepts the need for justification to each individual (and is protoliberal in this sense), but the justification he offers yields only elite sovereignty. [12] Plato, by contrast, is only a generic rather than a liberal republican and shows no sign of accepting either liberal republicanism or popular sovereignty. [13]

An important question, then, is whether liberal republicanism—the requirement of justification to each person who will be subject to the authority in question—can (contra Hobbes) only be met by popular sovereignty (even though they are two different things).

But the next important point is that, in any case,
popular sovereignty does not entail democracy.

The reason is that the bare idea of popular sovereignty would seem to allow each individual to autonomously alienate or transfer her share of authority on all political matters to some other person or group. This does not reduce the idea of popular sovereignty to nothing. The crucial question would constantly arise whether the authoritative act of transferring authority had actually occurred as the principle of popular sovereignty would, at a minimum, require. But democracy would be the stronger principle that no individual’s share of sovereignty can successfully be alienated in this global way. This is so far vague, since I assume the concept of democracy doesn’t preclude democratic authorization of representatives or delegates in certain circumscribed ways. I will suppose that alienation of one’s sovereignty is the irrevocable authorization of another person or body, whereas (certain kinds of) revocable or responsive authorization of another is not alienation of sovereignty. Government by representatives should not, in this day and age, be precluded by a definition of democracy, since it would be obtuse to hold that “representative democracy” is an oxymoron. The substantive questions are not about whether representative democracy is democracy, but when, if ever, governance through representatives is legitimate or advisable. If we see the requirement of democracy as the conjunction of popular sovereignty with inalienable sovereignty, and add that sovereignty is only alienated in the prohibited sense when the transfer of authority is irrevocable, then we have begged no questions about the legitimacy of all institutions in which there are representatives. The legitimacy and advisability of revocable representation remains open.

Who are the rulers? If sovereignty were alienable, then the members of the sovereign would not be guaranteed to be the rulers, since it is a status they could transfer irrevocably. If sovereignty is inalienable then members of the sovereign are also the ultimate rulers. If sovereignty is also popular—equally possessed by every citizen—then every citizen is equally and inalienably a ruler. It is a separate matter, and a problematic one, by the way, how “self-rule” might be respected by the principle of democracy. Rousseau, for example, hopes to explain how each can come under the authority of the sovereign and yet “remain as free as before,” by showing how popular sovereignty can respect individual self-rule. I am not here assuming that it can; each of us is obviously subject to the will of the whole ruling body (even if one is a member of it), which will often not coincide with our own will. But still, each could be an equal member of the set of ultimate rulers, which is a weaker requirement by far than individual self-rule.

2. Representation

Suppose we accept liberal republican citizenship and also, on some so far unspecified grounds, popular but limited and inalienable sovereignty. (We are not arguing for all of these, but looking toward what would or would not follow from them. [14]) Now divide the choices faced by the collective of citizens into two categories: election and policy. Election means the (revocable or responsive) sovereign authorization, by voting, of a person or body to make certain choices on behalf of, or in the name of, the sovereign body (which we are supposing is the whole body of citizens). Distinguish appointment from election so that an appointment is the authorization of a person or body not directly by the sovereign body, but by some other person or body authorized to do so. Let policy mean all collective decisions that are not elections or appointments. (A special case is legislation. I’ll say more about legislation shortly.)

Once popular sovereignty is accepted there is no avoiding representation in some form. Even Athenian democracy could not entirely govern through the general assembly. [15] Even if all policy questions could be brought before the sovereign body, there must still be implementation, and the implementers must somehow be chosen. Either they are chosen by the sovereign body, which would be representation, or they are chosen by someone else so authorized by the sovereign body, in which case that person is the representative. Either way there is representation. (Nothing, by the way, is implied here about any relation between the representative and the represented other than that the one authorizes the other.)
If popular sovereignty is inalienable, as the principle of democracy says, then not only must there be representation but it must be revocable or responsive to the will of the sovereign body. It is tempting to call this “representative democracy,” but there is an ambiguity in this term that will come up shortly.

It may seem as though the implementers could, for all the principle of democracy says, be chosen without any democratic authorization, such as by a hereditary scheme, or by lot, or some other method provided for constitutionally. Constitutions are themselves limitations on the exercise of popular sovereignty, but they are not inconsistent with it so long as they do not create any alternative sovereign for any purpose. A constitution may have provisions that are not open to sovereign revision (or it might not) but there might still be no sovereign other than the body of citizens. If the constitution instituted a hereditary line for certain positions the holders of those positions would not be popularly authorized, and so would be parallel or alternative sources of sovereignty. Whether or not this is a decisive argument against heredity depends on whether there are decisive reasons for insisting on popular sovereignty, a question I am not taking up. I leave aside the question how far the constitution needs to be popularly authorized. My question is when an authority arrangement, possibly including a constitution, itself in-corporates popular sovereignty. The question becomes pointed when we concentrate on the inevitability of relatively small numbers of policy makers for at least a sizable fraction of policy decisions.

Some might say that the question of direct vs. representative democracy is not about whether every policy is somehow democratically authorized, but only whether legislation must be made by the whole sovereign body or whether it might instead be made by elected or appointed intermediaries. Perhaps then it would be unnecessary to resort to smaller numbers. In truth, though, the emphasis on legislation would not change matters much. Technology may well allow everyone to vote from their home computer or cellular phone, but the amount of time that would be required in order run a legislature that consists of all citizens would be wildly infeasible if it is even possible in principle. (Don’t think just of time spent voting when bills make it to a final vote. How could the whole body of citizens run the deliberations that would bring the bill to this point? And keep in mind the combined burden of all legislative activities that are constantly occurring at local, state, and national levels of government.)

Even if the time problem were solved, it is not clear that there is a relevant principled distinction, for these purposes, between law and all other policy any-way. It is unavoidable that much policy be made by intermediaries rather than by the sovereign body as a whole. And we are taking for granted (by the principle of democracy) that these decisions must be authorized at least indirectly by the popular sovereign body as a whole. Either they are made by authorized representatives, or by those duly appointed by representatives. Only a small fraction of the necessary decisions could possibly be made by the sovereign body as a whole, and it is a cheap route to direct democracy simply to define any decision made by representatives or appointees as non-legislative.

If the question arises whether representatives count as rulers, there is no reason to deny it. I do not mean only that, as citizens, they too have equal membership in the sovereign group. In addition to this their status as authorized representatives gives them ruling powers than other citizens lack, and so they are rulers in a further sense. But it would be important to distinguish between authorized rulers and ultimate rulers. The ultimate rulers under inalienable popular sovereignty would still be the set of all citizens, since any other ruler must be authorized by acts of theirs.

3. Regency, Representative Assembly, or Referendum?

Pure direct democracy, then, is wildly infeasible. So the question is not whether democracy can legitimately employ representatives, but when it would be good to do so and when not. Not all decisions can be made by direct democracy, but since some can, there is the question of which ones should and which ones should not. Compare two positions on this question, both available within the commitment to democracy: the first, pure representative democracy, is the view that the sovereign body ought not to make any decisions except the election of representatives. The second view, mixed representative democracy, holds that while election of representatives is necessary and legitimate (so long as the arrangements don’t amount to an alienation of sovereignty), the sovereign body may also make some non-elective or directly democratic decisions. Call directly democratic decisions other than elections referenda. (Consider here all and only referenda whose passage would constitute legally binding policy.)

Pragmatic reasons preclude pure direct democracy, but it is not clear that it would be desirable anyway. We are assuming popular sovereignty, but the connection between that and numerous rulers requires argument. We can grant inalienable popular sovereignty (democracy) and constitutional limits on its authority, and even grant that pure direct democracy is wildly infeasible and either pure or mixed representative democracy is required on at least pragmatic grounds. But none of this lends any support to the idea that the representative body must have more than one member, an authorized and legally limited representative legislator. (Hobbes had no trouble with the idea that a whole people could be represented by a single person.) Nothing about the idea of representation provides the rationale for large numbers of legislators rather than a single individual or a small ruling body. In an important sense, this would be a representative democracy, but it would be misleading to let that name, which invariably connotes a large number of representatives, cover it. I will refer to this possible office as “regent,” which connotes derivative authority, and also the possibility of either an individual or a small board. [16]

Carl Schmitt noticed this point:
If for practical and technical reasons the representatives of the people can decide instead of the people themselves, then certainly a single trusted representative could also decide in the name of the people. Without ceasing to be democratic, the argumentation would justify an antiparliamentary Caesarism. [17]

It may seem that this is precluded by the logistical problem of time that counted against pure direct democracy. No individual, even a professional, could do all that needs to be done by a legislature. However numerous, elected representatives are expected to closely attend to all legislation. They can only do so with staffs of their own and various appointed agencies to assist. But these devices of assistance could be made available to the regent too. She or they could have extensive powers of appointment constrained by a constitution.

My point is not to recommend the institution of a regent, but to ask what, if anything, is wrong with it from the standpoint of democracy. It is representative, democratically authorized without any alienation of sovereignty, and constitution-ally limited. Citizens would not vote for laws, but nor do they do so if the legislature has 500 members in a society of 200 million citizens.

I believe it is a critical challenge for normative democratic theory to explain what, if anything, would be wrong with a single member legislature.

Part II: Shrinking and Expanding Factors

4. No Expansion Without Quality

In the remainder of this essay we will look at several factors that might be thought to press toward larger numbers, and several that might press toward smaller numbers. The aim is only to make a beginning toward comparing the significance of these factors when taken together. There are bound to be other factors that bear on the question of numbers, and other points to be made even about the factors I consider here. I emphasize this in order to guard against construing any of the arguments here as intended support for any practical proposal. The points I make here are far too preliminary and incomplete for that purpose.

Before looking at particular expanding or shrinking factors it will help to lay out a simplified picture of how these countervailing considerations might be use-fully compared. The primary simplifying device I propose is to imagine only three sizes, to be vaguely construed as regency, representative assembly, and referendum. This framework yields two places where an expanding or a shrinking factor might apply: between regency and representative assembly, or between assembly and referendum. We can leave the numbers involved fairly vague for our purposes, but let us assume that the referendum size is in the millions, and the representative assembly size is between, say, a hundred and a few thousand. The following graphic summarizes these points.

Figure 1

Some factors count against rule by regents and toward larger numbers. Other factors count against rule by referendum, toward smaller numbers. It might seem that these balance out in favor of rule by middle numbers, such as a representative assembly. However, a closer look at the several shrinking and expanding factors makes this far from obvious.

Before looking at the specific shrinking and expanding factors, notice that countervailing factors have no general tendency to resolve in a middle space between the polar extremes. In Figure 1, Consider range A, the choice between small and medium numbers, or between regency and assembly. Suppose the shrinking and expanding factors resolve so as to push to the medium numbers. But then what would prevent them from resolving in the same direction in range B, the choice between medium and large numbers? What would stop the expanding factors from outweighing the shrinking factors? The same question can easily be put starting from the other direction. If shrinking factors outweigh expanding factors in range B, pushing toward assembly as against referendum, why wouldn’t they continue to prevail in range A, pressing all the way to regency?

Of course, we will have to look more closely at the factors to see how they might resolve. The point here is just an introductory one: there is no general reason to assume that countervailing factors will resolve somewhere between the poles. Sometimes they do. If I have some reason to bring a large suitcase, in order to fit more clothes, and some reason to take only a small suitcase, which is easier to carry, these might resolve in favor of taking a middle-sized suitcase, compromising between the countervailing factors. But in other cases, the stronger factor might simply prevail. The stronger team in a tug of war might be only slightly stronger, but that will not prevent them from pulling the other team all the way over the goal line. Countervailing reasons can operate this way as well: there might be some reasons to write a long book, but some reasons to write a short book. It is by no means guaranteed that these will resolve in favor of a middle-length book. The reasons in favor of greater length might still prevail in the choice between large and medium, even though there remain some reasons militating against greater length.
Consider, then, a list of several considerations that might be offered in favor of larger numbers.

Enlarging factors:

Fair deliberation: fairness to all views
Apart from any tendency to produce good deliberations or outcomes, each person is owed a hearing.

Coverage: representation of all interest perspectives
Each person’s interests should be entered into deliberations if voters are to make an informed decision.

Condorcet: aggregative knowledge
Other things equal, if voters are better than random on the choice at hand, under majority voters will make correct decisions more often if they are more numerous, apart from any benefits of discussion.

Collective Wisdom: inclusion of all reason perspectives
Different people bring different perspectives to public discussion, enhancing the epistemic quality of the process.
Three of these four, excepting only fair deliberation, suggest that larger numbers of participants are supported by the higher quality this would lend to the out-comes or at least to the content of the deliberation. The first, fair deliberation, is put forward by Waldron as a non-epistemic account of the rationale for large numbers in the assembly, and I criticize his argument shortly.

Incidentally, there is no institutional necessity in the voters and deliberators being the same set of people, and some interesting questions arise if we contemplate distinguishing them. But here, in order to focus on certain points, we will assume that they are identical.

It is hard to think of expanding factors that do not rely on quality considerations. The familiar idea of procedural fairness to all participants makes no reference to quality of deliberations or outcomes, but nor does it have any apparent tendency to support larger numbers. There’s no reason a regency could not be procedurally fair to all members of the larger polity; suppose the members were chosen by lot for example. Who does this unfairly favor? If that is a fair procedure but an objectionable one, then the objections must be on grounds other than procedural fairness alone.

It might be argued that larger numbers than regency can be supported on the grounds that they are required in order to get a statistically representative assembly, and that this rationale requires no appeal to the idea that this improves the content of the deliberations or the outcomes. But if that is so, then what is the reason for having a statistically representative legislative body? To ensure that no quality considerations are moving us, suppose that we happen to know that, as it happens, even in the statistically representative body no outcomes or arguments will arise beyond those that would have arisen in a regency. Now, with those quality considerations out of the picture what reason is there to have a representative assembly? At this point is natural to say that people have a right to have their point of view represented, even apart from whether this improves the content of the deliberations or the outcomes.

This, I take it, is Waldron’s rationale, a kind of fairness to views. I think it’s hard to see why this is a valuable kind of fairness unless epistemic considerations are brought in. We don’t care, after all, about having a fair say if no one can understand us, or if the decision is made prior to the discussion. It may be a kind of procedural fairness, just as a random choice of policies is a kind of procedural fairness. But fairness itself can’t choose between them, and so fairness can’t ex-plain any need to give people a fair say. [18]

5. Quality Grounds for Expansion

I doubt, then, that there is any plausible account of why the numbers should be larger than a regency that refrains from any claims about the salutary effect of the larger numbers on the deliberations or decisions. Without defending the more quality-based rationales (coverage, Condorcet, and collective wisdom), I will simply suppose for expository purposes that they succeed in supporting a preference for larger numbers than regency. We will turn shortly to the question whether there are shrinking factors that countervail this support.

Several questions arise if we look to range B, the larger range involving the choice between representative assembly and referendum. One question is whether there are new expanding factors at that level. I will simply assume that there are not. But the question remains whether the same expanding factors in the small range are still expanding factors in the large range. We are not yet asking whether they are countervailed by other factors, but only whether there is a pro tanto or de-feasible quality advantage in large over medium numbers. Another way of asking this is whether the advantages of large numbers somehow run out or precipitously fade away above the middle range.

Consider three quality-based arguments for middle rather than small numbers:

Coverage: Representation of all interest perspectives
The idea of a regency naturally prompts the objection that only a small subset of relevant points of view is given any political influence. Middle sized numbers (assembly size) will tend to represent a larger subset, which seems like an improvement that would tend to result in less injustice. Now the question is whether it would still be a further improvement of the same kind to move from the middle numbers of an assembly to the large numbers of rule by referendum (so far as feasible, in a mixed system).

Clearly no further improvement would be possible if all relevant interests were already represented by a body of the assembly size. This would not be guaranteed by size alone, and might require further arguments about the value of elections or of statistical representation. However, it is implausible in any case to think that the number of relevant interest perspectives is smaller than the number of people. The idea that, for some balance of reasons, we should settle for a representation of only a subset of relevant points of view is not at all the same as the far less plausible idea that the interest perspectives that are not represented are not relevant.

Suppose there is a diminishing marginal value of the following kind: for each representative added, fewer people get their relevant interest perspective thereby represented. This might be so under some arrangements, though it would require some mechanism whereby the largest constituencies get represented first. Both election and lot might be argued to have this effect, though I will not pursue the question here. In any case, if the number of relevant perspectives is equal or close to the number of citizens, the goal of representing as many relevant interest perspectives as possible does remain an expanding factor even in the high range. Whether it is countervailed by some shrinking factor is a question taken up later.

Condorcetian aggregation:
The expanding factor based on Condorcet’s Jury Theorem, [19] consisting in the epistemic value (under certain conditions) of larger numbers of voters has a similar profile. There is no reason to think of it as dramatically weaker in either the large or the small range. There is a diminishing marginal epistemic value of extra votes, but in our simplified model we cannot say that the marginal extra voter matters significantly less in the high range than in the low range. We can be sure it matters very little in the higher parts of the high range, since the closer the group competence gets to 1 the less difference there is to be made even by very large numbers of additional voters. It follows that the marginal value must diminish, but how fast it diminishes will apparently depend on how close the average individual competence is to .5. For example, if it is far above .5 then the group competence will get very close to 1 even at the moderate size we’re calling the assembly. Marginal votes cannot be worth a great deal above that size. But if the average competence is only slightly above .5 then the group competence might still be a long way from 1 even at a size far above the assembly level. I doubt there is any basis for supposing the average competence to be, say, .6 rather than .501. Since we are simplifying to only two ranges, we might summarize this uncertainty, combined with the fact that there must be some diminishing value, by saying this expanding factor is only modestly weaker in the high range. It remains an expanding factor.

Collective wisdom: deliberation among all reason perspectives
The expanding factor of the value of public reasoning among a large and diverse set of participants seems again to operate in the large as well as in the small range, leaving aside for now whether there are countervailing factors. The reason is similar to that in the case of coverage. Just as the number of relevant interest perspectives approximates the number of citizens, why think that the number of relevant reason perspectives is any smaller? Again, there may be some institutional arrangement in which the marginal representative brings to the table the reason perspectives of fewer citizens. Still, this remains an expanding factor unless above middle numbers marginal representatives failed to bring in any new reason perspectives at all. This is extremely doubtful. Whether or not there are good reasons on balance for sticking to the assembly size, it is hard to see what basis there could be for thinking there are significantly fewer reason perspectives than there are citizens. I count this as a (pro tanto) expanding factor even in the high range.

6. Shrinking Factors

We turn now to several candidates for shrinking factors: considerations that support smaller numbers. Something might be a shrinking factor in one range but not the other, so the two ranges require separate consideration in the case of each factor. First, though, we will see that most of the candidates fail to count as shrinking factors after all. In considering the shrinking factors I will not only consider whether they are indeed shrinking factors, but also, if they are, how they might weigh up against the quality-based expanding factors we have allowed above.

Pragmatic: Pure Direct Dem. fails utterly, on pragmatic grounds
We have seen that pure direct democracy is infeasible in its demands on citizens’ time and productive energy. This might seem to suggest that referendum sets the number or fraction of the citizenry who participate in legislation too high. But all it really shows is that referendum can only feasibly be used in a mixed system in which some fraction of questions is handled by other means. Once that is assumed, the pragmatic problem with deciding some appropriate fraction of issues by referendum evaporates. It turns out that the problem is not too many people, but too many issues. We will assume, then, that pure direct democracy is not an option, and that only a feasible number of issues will be put to referendum if referendum is used at all. The pragmatic shrinking factor disappears.

If we are assuming a mixed system, and supposing referendum were to be used for some fraction of issues, there would remain the other issues to be handled by some smaller number. The question would remain whether those issues should be decided by assembly or by regency.

Preeminence: some few are significantly wiser
The idea of preeminence reflects the familiar thought that a well selected smaller number of rulers is likely to rule more wisely than a larger number of ordinary people. It is surely the most important traditional anti-democratic idea, but even if democracy is assumed there is a need for representation on at least some issues for pragmatic reasons. Then the question can arise, without any anti-democratic implications, whether a small number of expert representatives will do better than a large number of less expert representatives.

The Federalist papers relied heavily on the idea that legislators would be the best and brightest among us. [20] This preeminence argument (along side the pragmatic and professionalism arguments) for elected representatives provides a reason to stem the expansion from representative assembly to direct democracy. This raises a difficult question. Since, as I have argued, invidious comparisons of the political wisdom of citizens are precluded by liberal republicanism’s constraint of public reason, we must decide whether the claim that election produces an epistemically superior set of representatives involves just such invidious comparison. (All we need to mean by “superior” here is that they have epistemic features that compensate for the presumed epistemic loss involved in the drastically smaller number of participants as compared with direct democracy.) Is the preeminence argument available within public reason? Public reason precludes our identifying any given features that make them superior, but it does not obviously preclude our saying that voters will tend to select superior people. Still, we would need some basis for this that did not appeal to any invidious bases of comparison.

Some argue on empirical grounds that election tends to choose social superiors. [21] But the idea that social superiors have superior moral and political is a central example of the kind of invidious comparison that I have claimed is open to reasonable rejection.

Manin argues on a priori grounds that election produces representatives who are perceived by voters to have, to a rare degree, some set of features that are deemed by voters to be desirable in a representative. [22] He calls this an “aristocratic” implication of elections that distinguishes them from certain other methods such as lottery. But the argument gives us no reason to suppose that the features being sought have any epistemic value, and in any case it only claims that elected representatives will be perceived to have the rare and desirable features, not that they will actually have them, as Manin clearly notes. So Manin’s argument would be no basis for thinking that election produces representatives with sufficient epistemic advantages to offset the epistemic loss involved in the smaller numbers of participants in a representative body.

On the other hand, if, as we are supposing, the point of having a numerous assembly is that collective deliberation has some epistemic value, and this claim is supposed to be available to public reason, then it would be initially plausible to suppose in parallel fashion that collective deliberation about who would be an epistemically superior representative is itself epistemically effective, at least under certain favorable conditions, and available to public reason. This avoids identifying any non-epistemic feature of candidates which is selected for by elections and then claiming, invidiously, that people with that trait are morally or politically wiser. Yet it would still suggest that elections have some tendency to select the better candidate. However, this idea leads to a puzzle that reduces its plausibility.

If a set of people can select a subset that is epistemically better despite the loss in numbers, then the subset could presumably do the same. They could select a smaller subset yet. This sub-subset could do better than the subset, but they could also select a further subset (a sub-sub-subset) that could do even better, and so on. In the end, there would result a regent with a strong claim to epistemic superiority. Call this the problem of the shrinking assembly.

The idea that deliberative bodies have epistemic value cuts both ways: they have some epistemic advantages over smaller groups or single individuals, but they presumably would then also have the ability to select epistemically good smaller groups or individuals. Which epistemic effect is greater so far as can be determined within public reason? Notice that an argument that prevents the shrinking assembly will also tend to count against using an assembly at all rather than the larger body of sovereign citizens as a whole. It is hard to see, then, how a preeminence argument for elected representative bodies could be contained to prevent it from justifying the unattractive institution of regency instead. There may be other considerations that allow the first shrinkage but not the second, but so far it is not clear what they would be. [23]

If things have the general structure I have described then we should think either that electoral preeminence favors regency, or we should conclude from the implausibility of the supposition that the successively elected regency would rule best, that there was no effect of electoral preeminence even at the first stage, shrinking from referendum to assembly. It is important not to confuse preeminence with professionalism, the very different factor that emphasizes the epistemic benefits of being a legislator as a full time job. I turn to that factor next. But if we put professionalism aside by supposing that everyone would have the same time and experience to devote to legislation, is it so clear that election would select significantly wiser people than a random selection? That is what electoral preeminence, by itself, supposes. Since, so far as I can see, it would also give rise to the shrinking assembly, I am tentatively inclined to doubt that there is any shrinking factor in the idea of electoral preeminence. Non-electoral grounds for preeminence are bound to rely on invidious comparisons of a kind that are precluded by the liberal conception of political justification.

Professionalism: time and experience promote quality
Professionalism is the idea that those who can develop experience in the job and devote themselves full-time to it will rule significantly more wisely as a result. This could be a shrinking factor since it is plain for pragmatic reasons that not everyone could fully acquire the epistemic advantages of professionalism. Leaving legislation to those who can might more-than-compensate for the epistemic loss from smaller numbers by the increased ability of the smaller number of professional legislators.
There is no problem of a shrinking assembly in this case, since the maximal benefits of professionalization would seem to be available to numbers in the middle range (or more) rather than only to some very small number. So professionalization might be a shrinking factor in the high range, though apparently not in the low range. In the high range it is countervailed by the quality-based value of larger numbers. The question, then, is whether the lost value of numbers is outweighed by the increased value of professionalization.

On the other hand, there are difficult questions about whether the supposed epistemic value of professionalization relies on invidious comparisons of the kind that are precluded by a liberal conception of justification. The analogy is quite close to the suggestion that those with more education (of some specified kind) are more capable of ruling wisely, a consideration that led no less of an egalitarian liberal than John Stuart Mill to recommend giving extra votes to the educated (among others). [24] The reasons for rejecting Mill’s scheme apply reasonably directly to the claim that professional rulers are sufficiently wiser to outweigh the disadvantage of their smaller numbers. I can only sketch those reasons briefly here.

Consider a literacy criterion for voting: from the set of the literate, pull a demographically representative sample, removing the sample error with respect to race and class. Now give double voting power to everyone in the repaired sample, and so half as much to all illiterate citizens (and also to others who were excluded as a consequence of repairing the sample with respect to race and class). In this case, the cognized and demonstrable sample biases are removed, and the beneficial trait of literacy remains. The scheme strikes me as objectionable, but on what grounds? My contention is that objections to this scheme on the grounds that there may remain important sample errors of which we are unaware are not so unreasonable that they should be disqualified. That is, they are decisive against the appeal to a supposed overall epistemic benefit, from the standpoint of the wisdom required for good political rule, of being among the literate portion of the general citizenry. If this seems right in the case of a literacy criterion, why not also for any educational criterion? And if education can not publicly be held to ground superior ruling wisdom, is professionalization different in any way that would allow it to escape the problem? On what grounds would we put such concerns beyond the pale? Given the actual history of ruling arrangements that privilege some citizens over others, such a worry needn’t be crazy, or based on ill will. More argument is required on this point, but for reasons of this kind I doubt that professionalization is available as a public basis for attributing greater political wisdom. In that case professionalization is not available as a shrinking factor at all.

Pathologies of Deliberation: large numbers disrupt rational deliberation
The final candidate for a shrinking factor is the idea that the epistemic benefits of public political deliberation are lost when the number of participants gets too large. Space prevents our looking at any specific pathological mechanisms or processes in any detail. It is commonly pointed out that in direct democracy voters face too many issues to be able to deal with them rationally. That objection is out of place here, though, since we are assuming the system must mix some direct democracy with some rule by smaller bodies, precisely to limit the demands on the time and rational attention of voters. So the number of issues they have to face can be modulated pragmatically.

One more pertinent common complaint about large numbers of participants is the claim that civility diminishes with the social distance between interlocutors, so it will be very thin indeed in large mass deliberations. Another worry about large numbers is that the expense of communicating to the whole electorate gives an unwarranted advantage to those who control wealth or media access. I return to this point shortly. There are also worries about “group-think,” and so on.

Stepping back from the specific pathologies that might accompany large numbers, consider the stakes. Suppose that pathologies of deliberation outweighed the expanding factors in the high range, and so on balance the assembly size is to be preferred to the referendum size. This would count against having a mixed system, since it counts against having any referenda at all.

Elections are, in effect, referenda themselves on the question of who shall be a representative. If the balance of considerations shows that referenda are seriously impaired by deliberative pathologies then the lesson must be applied to the con-text of election as well. These considerations might, then, offer some support for a pure representative system with the assembly chosen by lot, to avoid the pathologies in the deliberations around election contests.

Elections of members to representative assemblies have a mitigating feature: typically, the larger polity is divided into numerous sectors or districts so that voters in one district do not face the same slate of candidates as voters in other districts. As a result the number of participants in any single election is much smaller than the community at large. It might seem, then, that it has been misleading to associate direct democracy with large numbers. Representatives can be directly chosen with only a fraction of the overall citizenry participating in any single choice.

The idea of districting is certainly important to our topic since it has the potential to mitigate any pathologies of large numbers even in a direct democratic procedure. Still, it does not automatically inoculate assembly elections from whatever deliberative pathologies come along with very large numbers. In modern states there are many millions of voters. [25] The numbers involved in each election to an assembly seat is roughly equal to the number of voters divided by the number of assembly seats. We are, somewhat arbitrarily of course, considering the assembly size to be no larger than around 1000 members. Simplifying again, suppose the deliberative pathologies are not significant at that size, which is why assembly is initially favored over referendum with respect to deliberative pathologies. But suppose the pathologies are serious in larger sizes. This means that the pathologies would still be serious in districted assembly elections in polities with more than a million voters: 1000 assembly races decided by 1000 voters each. A larger community than that has too many voters per race, but dividing such a community into more districts raises the size of the assembly to levels where the deliberative pathologies are triggered. These numbers themselves are not to be taken seriously, since we have no clear account of the alleged deliberative pathologies or the numbers of participants required to trigger them, and to what degree, and under what institutional conditions, etc. But the more abstract point does not depend on the actual numbers: the fact that assemblies are chosen from districts is by no means a way of avoiding whatever deliberative pathologies are involved in large numbers. They would have that value only in political communities that are very small by modern standards. Otherwise, the districts would be either too big or too numerous.

It would be natural, next, to consider nested levels of decision so that small groups select representatives to join with other representatives, still partitioned into small groups, to select the next level of representative. Just to take one example: Suppose 1,000,000 voters are carved into 10,000 districts, yielding 100 voters per district. This in turn yields 10,000 representatives. But divide them into, say, 100 groups of 100 representatives per group. Have each of these 100 groups select several of their members, say 3, to serve in the final assembly. This yields an assembly size of 300, without any race having been decided by more than 100 voters. Such a nested arrangement has obvious interest for avoiding any pathologies of races involving too many voters. On the other hand, there are innumerable other issues about how such a scheme would operate, and so this single virtue is not any significant basis for recommending such an arrangement.

The idea of nested levels of deliberation suggests a possible reply to the very idea that there are pathologies of deliberation at large numbers. It might be argued that actual deliberations in large communities informally assume a nested structure. Very few participants are actually on the national stage. The image of each citizen deliberating with all other citizens on national questions is misleading in this context. Some citizens do speak to all others, as when they publish their opinions in a nationally accessible way. Still, only a tiny fraction of citizens participates in public deliberation at this level. Most citizens do deliberate with others about political matters, but in contexts ranging in size from a small circle of friends, to neighborhood or community groups, to publications at the level of a city newspaper. Above this size the number of participants is very small, very possible falling short of a size at which deliberative pathologies are serious. And at any of the lower levels, too, the numbers may be small enough to avoid the pathologies.

Again, this point is of obvious importance when it comes to evaluating the pro’s and con’s of large numbers. But even if it is true as far as it goes, it leaves out an important sense in which these informally nested deliberators are all in the same boat. In the districting case the smaller nested groups in districts were not deliberating on the same issue as voters in other districts, since the candidates would be different across districts. In the informal nesting case we ought to be assuming that the issues are common to all the voters across the nation. After all, the informal nesting point is offered as a reason to deny that there is any need to break the issues up in the way that districting does. So we are imagining national referenda addressed in common by all citizens, even if the number of people ad-dressed by most citizens is far smaller than the number of voters in the nation. The implication, I think, is that while some deliberative pathologies may yet be avoided, others will still be in play. This is a good place to turn to a few specific candidates for pathologies of deliberation in large numbers.

One sort of pathology that arises when a large number of voters is addressing the same choice is that it is very expensive and yet very effective to advertise one’s point of view to the whole large group of voters. This can tend to give an unwarranted advantage to individuals or groups that control either great wealth or access to large communications media. There are empirical complexities about this, since groups with broad support might turn out to be the best at raising money. Still, other things being equal, it seems clear that groups with support among rich people have an additional advantage. I leave aside the empirical questions in order to consider what the implications of this would be if it were true. One implication is that the informal sort of nesting described just above would still present one large mass audience for advertising one side or the other on the issue of the national referendum. The advantage of wealth, if there is one, would not be diluted by that kind of nesting. It would be somewhat diluted by the kind of nesting characteristic of assembly districts, since there is no national contest but only hundreds of local contexts. On the other hand, the existence of political parties blurs this distinction somewhat, since assembly elections are often partly framed as con-tests between national parties, in which case a mass audience for mass advertising of the party’s point of view reinstates the advantages of controlling wealth or media access.

7. Conclusion

I do not want to impose too tidy a conclusion, since the exercise has not been of that nature. Instead, let me put forward two propositions that I hope receive some support, albeit inconclusive, from the miscellaneous points in this paper.
First, the (attractive) principle that no individual is morally more important than another from a political point of view is at some argumentative distance from the (dubious) principle that none ought to have more formal political power or influence than another (evidently precluding even rule by assembly). And neither of these is the principle of democracy, which holds (as I understand it) that political power is to be authorized by (not necessarily always exercised by) a sovereign body in which each citizen has an equal and inalienable role.
Second, democratic citizens face complex choices about how numerous the authorized rulers should be, though pure direct democracy is and always has been out of the question. If we wish to criticize the institution of, say, regency, we will need to do without the indefensible charge that it is simply undemocratic.

by David Estlund

Version imprimable fontsizeup fontsizedown
Pour citer cet article :


[1] Previous versions of some of these ideas were presented and discussed at Colloquium on Law, Economics, and Politics, convened by Lewis Kornhauser and Lawrence Sager, and at a conference on “Referenda and Direct Democracy: Moral and Legal Dilemmas,” Tel Aviv University, June 5-6, 2000, and at a workshop at Australian National University in 2002.

[2] Cite Plato ***

[3] Hobbes also thought people were relatively equal in political wisdom, and so his undemocratic views have a very different basis from Plato’s.

[4] This is approximately the same as Rawls’s use of the term in Political Liberalism.

[5] I argue for this in “Making Truth Safe For Democracy” in The Idea of Democracy, edited by David Copp, Jean Hampton, and John Roemer, Cambridge University Press 1993, pp. 71-100, “Waldron on Law and Disagreement,” Philosophical Studies vol. 99, no. 1, May 2000, pp. 111-128, and Democratic Authority: A Philosophical Framework, Princeton University Press 2008, Chapter 11.

[6] I develop the idea of this sentence elaborately in, Democratic Authority, op. Cit..

[7] I discuss structural arguments further in “Beyond Fairness and Deliberation: The Epistemic Dimension of Democratic Legitimacy,” in Deliberative Democracy, James Bohman and William Rehg, eds., MIT Press, 1997.

[8] Cite Locke ***.

[9] Choosing these rulers in a constitutional regime would not have to be done by anything resembling voting, but might proceed as monarchies have, by inheritance; or the ruling office might be sold to the highest bidder.

[10] Mention Grofman and Feld on Jury Th. ***

[11] Grofman and Feld suggest this interpretation of Rousseau, since larger communities may be more prone to factions, which can seriously harm the preconditions of the Condorcet Jury Theorem.

[12] Cite Hobbes. ***

[13] Cite Plato. ***

[14] In particular, no adequate argument against what I call “epistocracy,” or the limitation of sovereignty itself to a wise elite, has been given here. I have alluded to the argument that invidious comparisons are not legitimately available, but more than this would be required to rule out epistocracy, since, as we’ll see below, there might be structural (non-invidious) grounds for thinking that certain institutions would identify the wise. These are important matters, but for another occasion. They arise again briefly below in the discussion of professionalism.

[15] Cite Manin ***

[16] “Regent, 1. One who rules or reigns; a governor; a ruler. —Milton. 2. Especially, one in-vested with vicarious authority; one who governs a kingdom in the minority, absence, or disability of the sovereign. 3. One of a governing board; a trustee or overseer; a superintendent; a curator; as, the regents of the Smithsonian Institution.” Webster’s Revised Unabridged Dictionary, 1998.

[17] The Crisis of Parliamentary Democracy, (MIT Press 1988) p. 34 Schmitt’s reference to “Caesar-ism” shouldn’t distract from the fact that the office that I prefer to call “regent” wouldn’t be a dictator, but would be constitutionally limited as well as elected.

[18] I critique the “fair say” idea at more length in Democratic Authority.

[19] For explanations of the basics of the Jury Theorem and some applications to democratic theory see Estlund ***, Grofman and Feld ***, Goodin ***…

[20] cite Federalist. ***

[21] Cite Manin ***

[22] The Principles of Representative Government, 132-149

[23] At stage 1 a large group A will get better laws if it delegates legislation to a smaller wiser group B. B is smaller, but still, by hypothesis, collectively wiser than A. The reduction in numbers does not produce a reduction in collective wisdom; it is outweighed by preeminence. Now at stage 2 group B can either legislate or delegate. If A’s delegation abilities exceeded their legislation abilities, why wouldn’t that be so for group B as well? If there is already little difference among us then little improvement would be possible after a certain point, but it is at least as plausible to suppose that there continue to be significant differences in the preeminence of individuals all the way down to a very small, regency-sized group.

[24] Cite Mill, Considerations on Representative Government ***

[25] I leave it vague as to whether the relevant category is actual voters or eligible voters. It’s not clear that we should our view to actual voters here since even citizens who do not actually vote might participate in political deliberation, raising the numbers that matter from the standpoint of the deliberative pathologies. This is just one of the interesting complications that arise if we drop the assumption that the voters and the deliberators are the same people.


© 2009 | Toute reproduction des articles est interdite sans autorisation explicite de la rédaction.

Motorisé par SPIP | Webdesign : Abel Poucet | Crédits