Democracy is the theory that the common people know what they want, and deserve to get
it good and hard. (H. L. Mencken, Sententiae, A Book of Burlesques, 1920)

The many attempts at defining democracy over the years might well give grounds for pessimism that it can ever be defined. The remark by Lijphart that democracy might be undefinable perhaps reflects this pessimism (Lijphart 1977, 4). Over the years, the conceptual waters have been thoroughly muddied, and we must approach the problem afresh. Democracy we now understand it is essentially a modern concept, and requires a modern analysis. The problem of producing a conceptualisation that may be useful analytically is compounded by the fact that the term has become an important part of ordinary discourse, and so has become more an emotional symbol than an analytical concept.

There is clearly little hope that there can ever be a generally 'agreed'

Resumé

This paper attempts to provide a modern, universal, conceptualisation of democracy. J. D. May's 'responsive rule' approach is analysed. It is argued that his approach, although on the right lines, is not satisfactory as it stands. Democracy should be seen as referring to the principles which underlie the political process for a given regime, and is logically independent of the detailed institutional practices. Following Easton's analysis of a regime in terms of authority structure, values, and norms, democracy is analysed in terms of three principles of upward control, political equality, and norms defining acceptable policies, procedures, and behaviour. Democracy is not a dichotomous concept: given regimes differ in the extent to which they embody the principles of democracy in the operation of their institutions. In practice it will be hard, perhaps impossible, to find any regime anywhere which does not embody some elements of democracy to some degree. This vitiates the almost universal practice of using democracy and non-democracy as underlying concepts in a system of categorisation of regimes. Such categories become wholly arbitrary. Because of the subtle ways in which the democratic principles may work in different contexts, and because measures of these various manifestations of democracy can only be combined on a purely arbitrary basis, statistical measures of 'democracy' also become arbitrary. It is concluded that, although facets of the political process may be investigated using statistical techniques, ultimately the main thrust of empirical studies of democracy must be qualitative rather than quantitative.

definition of democracy, in the sense that we may obtain agreement on the definition of the speed of light, or even on the nature of 'the political system. Neither concept has ever formed an important part of the discourse of social or political action, and so there have been few normative axes ground in the course of defining them. This does not mean, however, that we should succumb to the pessimism of Lijphart. While there will never be a generally agreed definition, the very fact of its use in empirical research demands that we attempt to clarify the concept and put forward a sensible formulation that may then provide a solid starting-point for empirical analysis. This will involve stripping away as much of the emotional veneer as possible and exposing the essential nature of the concept. This process is likely to result in a formulation which is somewhat removed from that used in everyday political discourse, but which nevertheless underlies this usage. Until we do produce such a formulation, empirical research on democracy must be intellectually suspect.

John D. May has put forward a formulation of democracy which attempts to be intelligible, universal, a non-sectarian. I shall argue in this paper that his approach, which he calls 'Responsive Rule', is not satisfactory as it stands, although it is on the right lines. I shall suggest that one of the best routes by which we may approach the concept is through the normatively neutral doorway of systems analysis. I shall further argue that there can never be a simple clear-cut definition of democracy in operational terms, in the sense that we might have a clear operational definition of socioeconomic because democracy is a more complicated principle, or set of principles, that may be worked out in many different ways in different political contexts, and a fortiori there is no particular set of institutional arrangements that is essential to democracy. I argue that the central thrust of useful research into democracy must therefore be ultimately qualitative, rather than quantitative.

The General Character of Democracy

Part of the problem with the literature on democracy has been a tendency to concentrate too much on the precise institutional and legal aspects of a democratic system, and there often seems to be an assumption that democracy' some specific set of arrangements.

However, it is important to remember that democracy is a regime type, just like monarchy or totalitarianism, and as such it specifies the general principles which underlie the political process, and not the detailed institutionalpractices. talking about democracy, or monarchy, or whatever, therefore, we are talking about regimes that embody certain principles which underlie the political process, though with a less complicated concept

such as monarchy there is unlikely to be as much variation in how the
principles underlying the regime are worked out in practice.

Now a regime is that set of rules, conventions, and norms that govern the operation of the political process. It is a set of constraints upon the actions of the individuals, groups, and institutions in a political community. Easton in his systems framework (1965, 193) suggests that this set of constraints may be broken down into three components - (a) authority structure, (b) values, and (c) norms.

For Easton, the authority structure specifies the formal and informal patterns in which political power is distributed and organised within the community, values provide the boundaries within which day-to-day policy can be formulated without violating community beliefs, and norms are the kinds of procedures and behaviour that are expected, and are acceptable to the community.

In an earlier paper, Keith Dowding and I used this perspective to sketch a conceptualisation of democracy (Dowding & Kimber 1985). It is part of my objective in this paper to modify and rework this approach in more detail in the light of my criticisms of May's approach. Dowding and I argued that democracy involves the application of three fundamental principles. The first is that the authority structure underlying the institutional pattern should be that of what we called 'upward control'. By this we meant that ultimate sovereignty resides at the lowest level of the authority structure. The second principle underlying the democratic ideal embodies the value of political equality. The third aspect of democracy is that of norms. These are perhaps more difficult to specify. The most important norms of democracy are those which specify what behaviour is acceptable; that is, they constrain members to behave democratically. In other words, we argue that in a democracy, individuals' norm-governed behaviour must include an acceptance of the idea that individual and institutional behaviour is limited. This limitation must be in accord with other principles of democracy; that is, behaviour must be consistent with, and limited by, the principles of upward control and of equality.

These, then, are the three aspects of democracy with which I shall be concerned, aspects which, it is important to realise, may actually be implemented in different ways in different systems. Like May, we are seeking an approach that is universal; it should not tacitly assume the presence either of a large-scale association or of a face-to-face group, nor should it confine its attention to civil associations at the expense of private associations (May 1978, 3).

The Responsive-Rule Approach

May's definition is that democracy constitutes: 'necessary correspondence

between acts of governance and the wishes with respect to those acts of the persons who are affected' (1978, 1). This formulation has several components which require comment: (1) necessary correspondence with the wishes, (2) acts of governance, and (3) the persons who are affected.

(1) May seems to take the idea of necessary correspondence as selfexplanatory. the correspondence is taken to be necessary as opposed to contingent: that is, a correspondence with the wishes that just happens to occur does not constitute democracy; an act of governance has to take place because it is the wishes of the people affected. This is close to my own view - though, as we shall see later, I shall offer a more explicit formulation of the idea. The main problem with this element of the definition as it stands is that it seems to have no regard for the problem of how we recognise that there is a correspondence - or indeed, whether it even makes sense to ask whether the two do 'correspond'. If groups of affected persons have different wishes, how can the relevant act of governance to the wishes of those affected? May suggests that 'if each person exerts exactly the same amount of influence on each issue, and if each exerts influence on behalf of her policy preference, then every act of governance probably would comply with what is preferred by the largest number of persons' (1978, 11).

Ultimately, the problem here is that highlighted in the work of Kenneth Arrow (1951), namely that, in a non-dictatorial society, there is no rational means of consistently aggregating individual preferences into a social pattern may be said to be the 'societal preference'. Thus, at least something more needs to be said about what it means for an act of governance to correspond with people's wishes: in fact, it seems to me that the idea of 'correspondence' is basically misleading and needs to be replaced.

(2) May has concentrated on acts of governance as the objects of popular control. However, I would suggest that this approach is too narrow. Robert Dahl (1982) has argued that, for the people to be said to govern themselves, they must have final control not only over the acts of governance but also over the question of what matters are to be the content of acts of governance; is, they must have ultimate control over the political agenda. As we shall see, I shall not wholly embrace Dahl's position, but he is surely correct to draw attention to the importance of the political agenda in a democracy.

(3) Of the three main ways of defining the subject population - territorial, legal, and functional - May opts for the functional approach; that is, he writes of 'all affected persons', rather than of 'all inhabitants of a domain', or of 'all citizens'. This, he admits, 'seems to entail that the controlling population must vary in composition with almost every issue, and that on

some issues . .[it]. . must include in its ranks infants, lunatics, felons, and
foreigners' (1978, 8).

There are several problems here. First, the definition seeks to exclude those who are unaffected by the act. Thus, those Britons who never will take a holiday in South Africa are not considered part of the subject population in connection with controlling the Government's act of banning holidays to South Africa, were it to invoke such a ban. However, only considering 'those who are affected' seems to negate a generally held view that decision-making should not be the prerogative solely of those with a personal interest in the matter. Of course, one could argue that there is a sense in which all British subjects are affected by all decisions of a British Government, because what is being done is being done in their name. However, that would make the concept of 'persons affected' an empty one, and would effectively bring in the legal definition by the back door. Thus, it would seem that we have to say that the subject population should include at least those who are affected; but this, for the moment, leaves open the question of where the boundary should come. Second, considering everyone who is affected would mean that in deciding whether to go to war on country X, country Y would have to incorporate the wishes of country X into the decision for it to be democratic. Unless we have a category of policy decisions which we hold can never be arrived at democratically, we must, given the nature of the State, limit the subject population by some means other than the functional approach.

May's argument about why any group might reasonably not be enfranchised a democracy is obscure, and - in so far as one may make sense of it - absurd. Infants need not be enfranchised, he says, unless (a) they possess preferences, and (b) they are not already being represented, and (c) they lack non-electoral channels of influence, and (d) they would use their vote rationally. It seems to me that these criteria rule out almost any and every group one cares to choose.

I wish to argue that May's rejection of the legal approach to defining the subject population is misguided. He says that most of us regard extensions of the franchise (to females, to all adults, and so on) as gains for democratic rule, and that this 'exemplifies a rejection of legalistic approaches to defining "the people" ' (1978, 7). This is not so. Surely it is merely a matter of getting the legal definition right. Indeed, rather than viewing the problem from the point of view of extending the franchise to more and more groups (as has happened historically in most countries), there should be a general presumption that all individuals are included, with arguments being made to exclude only specific categories. Constructing such an argument to exclude the insane, or babies, would not be hard, though doing so for criminals, or teenagers, for example, might be more difficult.

A third problem is that awkward questions are also raised, as Dag Anckar

has pointed out (1982, 223), relating to the intentions of governmental acts and their outcomes (do we include the groups which the government intended to affect but did not, or which it did not intend to affect but did?); and there is also the question of direct and indirect effects (how much does one have to be affected in order to count as a member of the population?)

Thus, although the functional approach may appear at first sight to have the advantage of clarity, this is not so, and what clarity there is is achieved at the expense of reasonableness and of empirical utility. On May's approach, the population whose desires must be taken into account will frequently not coincide with any recognisable political object. The territorial which May also considers and rejects, is also unsatisfactory since decision-making units often overlap with one another and partly because its connection with decision-making is purely contingent. Thus, the legal approach to defining the subject population is the only sensible one. It clearly defines the population for whom the acts of governance valid. It possesses at least as much clarity as the functional approach, and does not have the latter's operational difficulties.

May's approach cannot be accepted as it stands: the notion of correspondence unsatsifactory; we need to say something about things other than just acts of governance; and the functional approach to the definition of the demos, to follow Dahl's usage (Dahl 1982), needs to be replaced by a legal one. The basic idea behind his approach, however, is sound, and is similar to that underlying my view of democaracy. Before spelling this out more clearly, I need to turn my attention to the three components of democracy identified earlier.

Upward Control

At first glance, it may seem that the term 'control' is not the best one we could have chosen - individuals can hardly be said to 'control' the authorities in a system, at least in the hard sense of the term 'control'. However, there is control in the sense that the demos, through its possession of ultimate sovereignty, is able to direct and place restrictions upon the actions of the authorities. What is important in a democracy is what the relationship of the individual is to this process; that is, what channels are available to individuals for the expression of their preferences about public policy, and what restrictions there are on the actions of the authorities. The means of expression of individual preferences may take many forms, though they may conveniently be summarized using the categories of Moyser, Parry, and Day (1986). They identify six dimensions of participation: voting, campaigning, group activity, contacting, protesting, and violence, though

we should not automatically assume that all of these are compatible with
democracy.

Underlying these dimensions of participation is the notion that individuals' are being input into a process which produces a societal decision. Thus, we require here that individuals have the capacity, and are actually able, to construct preference schedules relevant to the matter under consideration and, on the basis of that, are able to make a choice about the course of action to be followed. While it would not be reasonable to apply a strict test of rationality to this (it would probably exclude too large a proportion of the potential demos), this does imply rationality in a weak sense (perhaps, being able to recognise and rank the possible outcomes, able to make a choice, choosing the highest ranked outcome - but not necessarily being consistent, or having a transitive preference ranking, which a strong sense of rationality would require).

It follows from this that there should be unfettered political communication a level of understanding appropriate to the matter being decided. However, it should not be thought that this implies that we all have to be nuclear physicists in order to decide whether we want to build nuclear power stations. It is not reasonable to say that democracy requires certain specific criteria to be applied by individuals in evaluating a proposal; it only requires that a relevant evaluation can be made on some basis or other, and that the process is weakly rational. People's reasons for holding particular positions need not be taken into account. Only people who are unable to function at this minimal level need be excluded from the demos.

The process of actually inputting the preference also has certain entailments for example, that the appropriate civil liberties exist (freedom of association and so on), and that the appropriate gateway for inputting the preference is accessible (accessible polling stations, special arrangements for the infirm, freedom to campaign, protest, and so on).

Clearly a given individual cannot expect that his or her preferences will be the ones that directly affect the authorities and become the collective policy. What individuals can expect is that they have the right to be able to input their preferences into the process that determines the final view on the question of authority control, and that things are administered in such a way that they are actually able to exercise that right should they choose to do so. Their individual preferences may be formed in many ways, and the concept of democracy does not require us to specify how these are formed, except to say that they should be uncoerced; nor does it require individuals to express them. Indeed, we should allow non-participation as a legitimate preference, and where there is no formal or informal mechanism abstention, it would be wrong to insist on compulsory participation. means, for example, that compulsory voting is acceptable only if the individual is able to submit an invalid or blank ballot.

A point I wish to emphasise is that, from a conceptual point of view, the particular process by which individuals input their preferences is unimportant, that this process is subject to the necessary constraints implied by the principles of democracy. This means that, when concerned with the essential nature of democracy, one should not elevate voting above the other methods of participation. Neither particular methods of voting, nor the voting mechanism itself is necessary for democracy, though I shall qualify this statement later. The situation in which a tribe reaches its decisions by having, say, an elder take the sense of the meeting after a long discussion in which everyone has his or her say, qualifies as a democracy (other things being equal) every bit as much as the situation in which the decision is the outcome of a vote. What matters is not the particular mechanism by which the collective policy is found, given the individual preferences, but that there should be an entailment of a particular policy, given the particular set of individual preferences and the particular mechanism Of course, that mechanism, too, would have to be the subject of collective agreement, in the same way. This may seem to open up the possibility of an infinite regress, but this is not a problem for the position for which I am trying to argue. However many steps back one is forced to go, each step must be subject to the operation of the same democratic principles for it to be regarded as a democratic decision.

Thus, the individual preference may be taken into account in many ways; we cannot simply look at voting and elections; it may be done via other institutional and non-institutional arrangements (pressure groups, parties, the bureaucracy, the media, and so on) using other means (campaigning, contacting, protesting, and so on). Each method of participation is likely to have a different significance in different systems. However, we cannot allow the various possible methods of incorporating individual preferences into decisions to take any form; there must be some constraints. Essentially, the problem is a two-stage process of ascertaining, or measuring, individual preferences, and then of aggregating them to produce a decision. Precisely what criteria should be applied is a complex, and in part technical, matter which is beyond the scope of the present paper. However, I would argue that whatever solution is adopted, it must meet two requirements:

(1) The whole of the preference schedule of all individuals must be taken into account. (Note that this requirement is not necessarily equivalent to Arrow's condition of universal scope, since factors such as socialisation act so as to produce a degree of consensus in society, thereby restricting the combinations of preferences that exist to a subset of the logically possible set, so reducing the likelihood of failure of the aggregative rule.)

(2) Each individual schedule of preferences must be given equal weight.

Of course, the formal method of producing a decision is by voting, but providing whatever necessary criteria we decide upon are met, any method may be used. In non-voting situations, however, the outcome must be identical to that which would have emerged had the preferences been aggregated formally; that is, the result of 'taking the sense of the meeting', say, must be as ifa. vote had been taken, if it is to be regarded as democratic. Conversely, a decision in which the authorities acted upon the preferences of only a small group in society would be highly undemocratic.

What is important for democracy, then, is (a) the ability of the individual to be able to form and express a preference, (b) the translation of whatever set of individual preferences happens to be expressed into a collective policy, (c) the application of this collective policy to, or by, the authorities, and these (and the way in which they operate) are what need to be investigated empirically. Hence, the emphasis on rather limited measures of participation (usually voting) in many empirical studies of democracy can be misleading because they are only looking at one aspect of upward control. Voting is just the formal method of measuring and aggregating preferences - though this is not to say that voting is not an interesting area of study in its own right. It should also be noted that, as implied by (c), democracy is not just to do with deciding; how the collective decision is acted upon is also included. Clearly the policy as implemented must impinge upon all relevant individuals equally and must reflect the intentions implied in the decision.

So far, I have implicitly assumed that what is being controlled is a set of authorities and the policies they implement. Some writers, however, also put emphasis upon the need for the demos to be able to control the 'political agenda'. Robert Dahl, in his book Dilemmas of Pluralist Democracy, actually includes control of the agenda by the demos in his criteria for democracy (1982, 6). However, the 'agenda' is a complex composite of the general concerns of the demos, and specific matters which people feel ought to be the subject of decisions, and the problem with this is that there is no obvious mechanism by which the agenda may be controlled. Ideally, one might say that the agenda which emerges should be the one that would have emerged had it been possible to have a formal vote on it. More practically, perhaps, the best we can do is to say that there should be no barrier, or impediment, to the placing of issues on the agenda - i.e. there should be an absence of veto on the part of any individual or group. This question is an extremely subtle and difficult area to investigate empirically. Further, it is not only the authorities and the agenda we must consider: the contents of the decisions themselves must also conform to the democratic principles. Decisions such as those restricting the democratic rights of

certain groups, which clearly conflict with the principles of democracy,
must be ruled out too.

Thus, to conclude this section, while the election of government officials, whether national, regional, or local, offers a limited method of controlling those who wield political power (and, indeed, the desire for election and re-election is itself an importan dimension of control), other processes may also be important. How important they are will depend upon the nature of the particular regime in question, though it is clear that they become more and more important as the demos becomes larger, and as one moves from a system of direct democracy to a representative one. Thus, party processes, in both one-party and multi-party systems, could also offer methods of institutionalising forms of upward control, depending on the exact nature of the processes. There is no contradiction in principle between the singleparty and that of democracy. What causes single-party regimes to be largely undemocratic in practice is their failure to ensure the implementation all the principles necessary for full democracy.

Interest-group systems may also be one of the more important sources of upward control, though there is some doubt that the currently fashionable neo-corporatist ideas constitute a move in the direction of democracy. One of the characteristic features of neo-corporatism, as Crouch points out (Crouch 1983), is the way in which organised interests constrain and discipline their members both to promote their particular group interests and for the sake of the general interest. That the organisational form of neo-corporatist interest-group organisations, and the degree of control exercised through them, is in reality little different from that of pluralist interest-group organisations should not lead us to assume that, because we tend to identify pluralism with democracy, neo-corporatism is therefore democratic. Rather, we should regard those characteristic features of neocorporatism are echoed in pluralist interest organisations as tending to reduce democracy even in a purely pluralist context.²

Political Equality

The second principle of democracy is that of political equality. It is importantto that we are talking about political equality, and not any other kind of equality. Further, like 'democracy', 'political equality' is a term of uncertain meaning. In his analysis of the concept, Ware has pointed out that, although it is frequently used, the term is rarely analysed. He sees political equality as 'the distributive principle that is used to allocate values in a democracy' (Ware 1981, 392). This is some way from the sense in which I shall be using the term. The criterion that a democratic government might employ in evaluating policy alternatives (such as maximin or some

other principle) is something analytically quite separate from the concept of democracy. Indeed, the criterion may itself be determined by use of the democratic process, and I contend that there is nothing inherent in the concept of democracy which leads one to conclude that society should adopt particular distributive principles. What Ware calls 'political equality' other writers simply call 'equality' (cf. Barry 1965, 119). I am concerned with equality as it relates to how individuals and their preferences are treated in the context of collective decision-making, and wish to establish the principle that people should, for political purposes, be treated equally, in the sense of consistently and in the same way.

The ideal of equality may be institutionalised in two respects. First, there may be an equality of treatment of individuals in respect of the actual processes of upward control; that is, all members of the demos are treated as having political preferences of equal weight. May seems to argue (1978, 10) that this is the only intelligible sense of 'equality' in this context.

However, I would wish to add a second respect in which there must be equality. The condition of equality may also be taken to imply that all individuals have an equal right to perform those actions which are necessary for the particular forms of upward control that exist, together with an equal opportunity to exercise that right. How this equal opportunity is to be created may be open to argument. Some may wish to argue that it entails a system of welfare payments to ensure that, in practice, everyone has the opportunity to exercise his or her rights. I would argue, however, for an administrative solution - that is, the processes of upward control should be organised such that the disadvantaged are not deprived of the opportunity of exercising their rights. Of course, I am not arguing that there should necessarily be particular rights and liberties, such as the rights to assemble, to speak freely, to have equal access to information, and so on: I am simply saying that whatever relevant rights there are should be accorded to citizens equally. In most situations, of course, such rights would apply.

Douglas Rae and others have emphasised that,

Equality splits itself into many distinct notions, each an element in its grammar. Equality's subject may be individual-regarding, bloc-regarding, or segmental; its domain may be straightforward, marginal, or global; the idea of equality may be applied directly (equal results) or may be a version of equal opportunity (which in turn may equate means or prospects); equality may be based on uniform lots or on lots equally accommodating differences; it may be absolute or relative (and, if relative, based on any of several distinct notions of relative equality). (Rae et al. 1981, 132-133)

He argues that these distinctions must be combined with one another if equality is to be made flesh, and that it is meaningless simply to speak of 'equality for all js' without specifying how the distinctions apply to the particular situation. Accordingly, I would say that, in Rae's terms, the equality I am referring to is simple individual-regarding equality (one class

of equals, each the equal of the other), with the domain of account being the rights necessary for the pursuance of upward control. The equality is global in the sense that all individuals end up with the same rights, and straightforward in the sense that there is a coincidence between the rights which are accorded and those which are required for upward control. The opportunity for participating in upward control is equal in the meansregardingsense all individuals have the same rights to participate in this process, and also in the prospect-regarding sense that circumstances must be such that they have the same probability of being able to exercise their rights should they so choose. Since the rights accorded to each individual are exactly the same, the equality is lot-regarding; since every pair of individuals who are supposed to be equal at all are fully equal, the equality may be said to be absolute.

Norms

In his book on democracy, Cohen (1973, 200) distinguishes between the origins and the effects of actions in relation to democracy. A given action may be undemocratic in origin, but may be democratic in effect, and vice versa. A military government, for example, may arbitrarily impose a democratic system, or a democratic society may choose to restrict the voting rights of those citizens with IQs of less than 120. Cohen argues that the question 'Can a democracy act undemocratically?' must be given an unconditionally positive answer.

Of course, there may seem to be a sense in which this is so; clearly all sorts of decisions may emerge from a given decision-making system. However, if we ask the more precise question 'Can a regime embodying democratic principles act undemocratically (that is, perform actions which negate those principles) and still be intelligibly referred to as a democracy ¹, I would argue that the answer is clearly not. To the extent that a system makes a decision, however arrived at, which negates the democratic principles, however defined, it may be said not to be exhibiting the characteristics democracy. In other words, as we have already seen, in a democracy the subject matter of decisions which may be taken democratically also constrained. The subject matter of democratic decisionmaking not have an unrestricted domain.

This is where the third principle of democracy comes in. I have rather loosely referred to it as involving a requirement to behave democratically, and this is meant to be interpreted in the sense that everyone should accept that his or her behaviour, and the decisions resulting from that behaviour, are limited by the first two principles. Of course, the only way of institutionalisingthis through the legal system of the regime; but the legal

system is not itself enough. There must also exist accepted norms to the
effect that decisions and behaviour are limited.

As we have seen, even within a so-called 'democracy', laws may be promulgated that discriminate politically, intentionally, or unintentionally against sections of the society. Thus, perfectly legal rules may have antidemocratic and implications; that is, they may just be antidemocratic, they may produce anti-democratic results under certain conditions. Hence, under certain conditions individuals may be under a democratically inspired moral imperative actually to break those laws. Such actions, though law-breaking may still be held to be democratic and not anti-democratic (Dowding & Kimber 1985, 28).

Further, no individual behaviour must go beyond the limits implied by the principles of democracy, even if such behaviour falls within the institutionalised and legal rules. Essentially, as I have stated above, this second reason implies a general acceptance of the first two conditions (upward control and equality) and constitutes a kind of meta-rule. We may regard these behaviour rules as meta-rules over and above any actual legal rules. When the meta-rules prohibit some action, even where the institutional legal rules allow it, that action must be said to be antidemocratic.

It follows from this third, normative, principle of democracy that not all methods of resolving political problems and not all possible forms of participation are acceptable. Of Barry's seven categories of means of resolving disputes - coercion, debates, voting, bargaining, arbitration, contests, and chance - we must rule out coercion immediately (Barry 1965, 92). The others would seem to be acceptable, providing that their use is sanctioned by democratic means, and that their implementation is consistent the principles of democracy (for example, in a bargain there must be a transfer of assets of equal value, or if they are unequal such an arrangement must be entirely uncoerced). Similarly, with the categories of participation of Moyser et al. (p. 204, above), we must rule out the use of political violence in a democracy.

The Definition

We are now in a position to draw the various threads together in order to
say more positively what democracy is.

Firstly, we need to be explicit about the criterion for citizenship - that is, to specify which individuals are entitled to have their preferences incorporated into the collective decision. Dahl's definition (1982, 61) that the demos should include all adults subject to its laws, except transients, is a sensible approach, though there is room for argument about the age of adulthood, and whether the insane should be included.

Insanity is a particular problem here. At a theoretical level, it may be possible to reach agreement that the insane should be excluded on the grounds that they are not capable of rational thought, that they have no realistic perception of the political questions being asked of them, and that they are unable to form intelligible preferences about such questions. However, in practice, the problem is one of how we test for these abilities, and whether we can ensure that any test is applied justly.

With the young, the question simply seems to be: at what age can they, as individuals formulate intelligible preferences about policy independently? ability to do this clearly varies from individual to individual, and probably generally occurs at a lower age than is allowed for in most political systems. The two problems are not different in principle, and it may be possible to devise some method of testing for the ability to formulate such preferences, but most societies would at least regard it as impractical to have young people becoming part of the demos at different ages, and would opt for an arbitrarily chosen age. My guess is that such testing might show that 14 or 15 is about right.

Democracy, then, is that type of regime in which the demos, as defined, exercises control over collective decisions in the sense that a given set of freely expressed individual preferences, together with a given mechanism for producing a collective decision from them, entail the outcome, which is then applied, where relevant, to all members of the demos.

The mechanism used in translating preferences into policy must treat all individuals' preferences as having equal weight. Further, all citizens must be permitted to exercise those rights which are necessary for participation in the processes of upward control, and they must actually be able to exercise those rights if they so choose. Both the subject matter of decisions, and the behaviour of individuals and groups, including the authorities, must be constrained to the extent that they do not breach the rights and opportunities all citizens have to contribute to the exercise of upward control. There is no reason to think that any particular set of legal rules is required to ensure this; indeed, cultural norms alone could suffice. However, is most likely to be achieved by a mixture of legal provisions and supportive cultural norms. Table 1 summarises these points.

It should be clear that with this approach it is neither necessary nor possible to give a list of individual rights, such as the freedom to assemble, freedom of speech, and so on, as is common in discussions of democracy. This is so for two reasons. First, the necessary rights are implied by, and thus follow from, the nature of the particular decision-making mechanism^)which Second, since there is no particular institutional mechanism for arriving at the collective decision which is necessary for democracy, there is not set list of rights which is appropriate for all manifestations of democracy. In democracy individuals must be accorded

those rights necessary for the achievement of upward control, but need not
be accorded others. In practice, of course, it is likely that the usual rights
of free speech, free association, and so on, will be required.

One of the problems in defining democracy is distinguishing between those considerations which form part of the meaning of democracy, and those which are prerequisites for democracy. May's definition of democracy as Responsive Rule implies that some of the considerations here regarded as part of the meaning of democracy should really be considered to be prerequisites. However, it should be remembered that the approach here derives from the prior question of what constitutes a regime, and the particular Eastonian interpretation of the characteristics of a regime that has been adopted.

Conclusion

I have argued that democracy, should not be seen as entailing a particular,
fixed pattern of institutional arrangements but is, rather, a set of principles,
worked out in practice in different ways in different contexts.

If follows that it is not useful to talk about absolute pre-conditions for democracy (other than logically necessary ones); it depends on which processes of democracy control are being used. In so far as there are common patterns in the way democracy is manifested among a group of systems, then it might be possible to identify certain pre-conditions for democracy to flourish in those contexts. However, we should be cautious about giving an exhaustive list of pre-conditions for democracy, in some

generalised sense, in the way that Cohen (1973, Part Four) seems to (see Table 2). We may only usefully talk about the pre-conditions for a particular manifestation of the principles of democracy (or for a set of similar manifestations);that we may talk about the pre-conditions for particular processes of upward control to operate effectively in similar contexts.

Of course, some of Cohen's categories are somewhat trivial, and others are contentious. What, for example, is the role of the economy? Empirical studies, as Sanders shows, are less than unanimous in this area (Sanders 1981). Care needs to be exercised in interpreting some of the statistical evidence: if X is correlated with democracy, is it a causally necessary precondition, merely an effect? Further, even if we grant that X is a precondition, need to be clear about what aspect of democracy it is a pre-condition for: the birth of democracy, perhaps, or the stability of democracy. The latter then opens a whole series of questions including what we mean by 'stability'.³

The general approach offered in this paper differs from conventional approaches partly in it putting more emphasis on democracy as a principle underlying general political activity, and partly in the emphasis on the need for norms limiting the behaviour of members in the system. This is absent, for example, from Robert Dahl's list of five ideal criteria for democracy in his book Dilemmas of Pluralist Democracy (1982).

(1) Equality in voting.

(2) Effective participation.
(3) Enlightened understanding.
(4) Final control over the agenda.
(5) The inclusion of all adults.

Roughly speaking, (2) and (4) correspond to the criterion of upward control, while (1), (3), and (5) correspond to the second criterion of political equality. What is absent from Dahl's list is the idea that there should be a prevailing, accepted, norm to the effect that all behaviour should conform to the requirements of these five criteria, which I believe to be an important part of the meaning of democracy.

The three principles I have described suggest that democracy involves not just limited government, but universal limitation. Government is directly limited through the principle of upward control, but individuals are limited too, indirectly through upward control, and directly through the principle of equality, and the acceptance of the meta-rules.

Of course, the idea of limitation in this context is not at all original, but I do wish to make it clear how far I am willing to go in this respect. There is much talk in the literature about limitations on the majority, and thus of minority rights. Many authors seem to be willing to go much further than I am in this respect. Giovanni Sartori (1987), for example, in a recent reformulation of his classic study, asserts that: 'minority rights are a necessary condition of the democratic process itself (Sartori 1987, 33).

It is hard to dissent from this; but which minority rights? I would argue that the majority is only limited in respect of the political rights necessary for democracy, and not more than that. It may well be that a particular group always constitutes a majority on all questions; in such a situation I would require the minority to be protected in respect of its democratic rights, but certainly not in respect of all possible policy decisions by the majority, or even any non-democratically relevant decision. Of course, the majority may well treat the minority unjustly in these other respects, but this is a problem of justice and not of democracy. I would not wish to define democracy in such a way that by definition a democratic society had to be just, equal and so on, in all ways. As we have seen from history, majorities can behave abominably (though we should not assume that they

will always do so). Presumably we would want to allow the possibility of
asking the question 'How just is this society?' about different societies, all
of which meet the criteria of democracy to a given extent.

Of course, anything that intervenes between the preferences of the individual and the outcome, such that the individual's preference is unable to play its proper part (as discussed above) in the choice of the outcome, is undemocratic. This implies that indirect systems of decision-making violate the democratic principles to an extent (depending on their nature), and we can say that the fewer the number of levels intervening between the individual and the final decision, the more democratic the system. This would seem to imply that decisions should be reached at the lowest level possible. Further, the nature of the individual's representation at higher levels is relevant here; particularly, whether the system involves representatives delegates. Although in principle there need be no difference in democratic efficacy between the two, representatives (who have scope to exercise their initiative, and may even ignore the manifest wishes of their constituents) in practice are often less democratic than are delegates who are bound to follow the collective wish of those who have mandated them.

It follows from this that, unlike our conceptualisation of political stability (Dowding & Kimber 1983), democracy is not a dichotomous concept. In the case of democracy, it does make sense to speak of political objects being more, or less, democratic than others. Particular regimes will differ in the extent to which they embody the principles of democracy in the operation of their institutions. In practice it will be hard, perhaps impossible, find any regime anywhere which does not have some elements of democracy.

This interpretation runs counter to some influential views of the concept.
Sartori (1987) seems to see democracy as a system having certain properties,
for example:

Democracy is a system in which no one can choose himself, no one can invest himself with
the power to rule and, therefore, no one can arrogate to himself unconditional and unlimited
power. (Sartori 1987, 206, emphasis added)

While this parallels my own line of argument about democracy involving
universal limitation, it hints at a fundamentally different approach to the
basic nature of the concept. Sartori believes that,

Under the quantitative impetus we are no longer prompted to ask what is (or is not) a
democracy; we are, instead, required to ask to what degree, if any, a political system is a
democracy (p. 183, original emphases);

and he argues that,

variations within democracy or of democracy (relating to more and less democracy) require

that we first establish to what they apply, that is, that we first decide what is, and what is
not, a democracy, (p. 185, original emphases)

The kind of approach for which I am arguing would lead us to conclude
'that all the existing political systems are democracies, albeit to a lesser
and lesser degree' (p. 184).

These quotations make clear the difference between us: Sartori seeks to have the concept of 'a democracy' which he wishes to apply to various systems, and within this category there can be variations (more and less democracy). This leads to the USA and Britain, say, being seen as democracies presumably not to exactly the same degree), while Cambodia and Albania are clearly non-democracies. However, this sort of talk is easy with these sorts of examples - but what of Argentina, Singapore, or Nicaragua? The weakness in his position is ultimately revealed when he asserts that political systems are 'characterized by constitutive mechanisms and principles that are either present (albeit imperfectly) or absent (albeit imperfectly)' (p. 184).

If the same feature is imperfectly present in one system and imperfectly
absent in another, it is hard to see how that feature can form the basis of
an analytically useful categorisation.

The whole discussion seems founded upon the mistake of regarding democracy as an entity. Is it really possible to give a sensible answer to the question, 'When did Britain become a democracy?' - presumably not in 1215, nor 1700; 1832? 1918? 1928? Some might argue that even now it is not 'a democracy', with the media forbidden from reporting that which may be reported in every other country in the world (including material already in the public domain).⁴ The variable way in which the principles of democracy are manifested should make us⁵ cautious about using phrases such as 'a democracy', and this raises the question of how suitable the term 'democracy' is as the underlying concept in the classification of regimes into categories such as 'democracies' versus 'non-democracies', and so on. We might conceive a scale of democracy along which regimes might be ordered, but the concept cannot provide a satisfactory means of making distinctions between regimes., except by means of purely arbitrary categories.

Further, although I have referred to the possibility of a democratic 'scale', I should make it clear that this is simply a notional concept - one which can never be developed. There is an important reason for this. Democracy has been conceptualised as a set of principles underlying the political process and which may be manifested in different ways in different systems. It is the context in which the principles operate that will determine the actual institutional arrangements. Individual preferences may have an impact on the outcome in many ways, even in the same system. In order to produce an overall 'democratic score' for a given regime, we would not

only have to be able to measure the degree to which the democratic principles were embodied in the voting system, group activity, campaigning, protesting, and so on (difficult certainly, perhaps impossible), but we would also have to be able to combine these into some kind of weighted index. This cannot be done, except purely arbitrarily. Thus, no meaningful single measure of democracy can be constructed statistically. The empirical study of democracy must ultimately be pursued in a more conventional and qualitative manner.

Of course, the arguments in this paper have nothing to do with the question of whether the most democratic system is actually practical or, indeed, desirable. It does not help the discussion of democracy that nearly everyone assumes that it is invariably 'a good thing'. In the choice of constitutional arrangements, many considerations come into play, and it is quite rational for a society deliberately to choose a less than pure democratic system as the one most suited to its needs and most likely to provide a stable decision-making framework.

---

NOTES

1. I should like to thank Elias Berg, Keith Dowding, Jan-Erik Lane, Juan Linz, Giovanni Sartori, and Brian Smart for reading and commenting on an earlier draft. This paper was given at the conference on 'Advances in Comparative Institutional Analysis' at the Inter-University Centre of Postgraduate Studies, Dubrovnik, October 1987, and is a revised version of one given at the workshop on 'Socio-economic Pre-conditions for Political Democracy' at the ECPR Joint Sessions of Workshops, Amsterdam, April 1987. I am grateful to the participants of both conferences for their comments.

2. See Dowding and Kimber (1985) for a discussion of pluralism and neo-corporatism in the context of democracy and political stability. I am pursuing a somewhat different line in this paper, since I feel that that part of the argument in the 1985 paper is incorrect.

3. See Dowding and Kimber (1983) for an analysis of this concept.

4. At the time of first writing this piece, the British Press and Television were prevented from reporting details of the Government's actions in the Australian courts relating to the book Spycatcher by Peter Wright, and from discussing the latter's contents, even though it was published abroad and copies circulated freely in Britain. While this restriction has since been lifted, the Government has now imposed limits on the BBC and ITV in their reporting of matters relating to Northern Ireland, using its very extensive powers under the BBC's Charter and the Broadcasting Act 1981. Section 29(3) of the latter states that '. . . the Secretary of State may at any time by notice in writing require the Authority to refrain from broadcasting any matter or classes of matter specified in the notice; and it shall be the duty of the Authority to comply with the notice". I am grateful to Graham Higgins for drawing my attention to the very broad wording of this provision.

5. The 'us' here refers to 'us' as analysts seeking rigorous description and explanation of political phenomena, and not to 'us' as dinner-table conversationalists, or even TV pundits. Hopefully, both these categories, and others, will come to see the pointlessness of attempting to construct generalisations on the basis of such broad, imprecise, and arbitrary categories.

REFERENCES

Anckar, D. 1982. 'A Definition of Democracy', Scandinavian Political Studies 5(3), 217-235.

Crouch, C. 1983. 'Pluralism and the New Corporatism: A Rejoinder', Political Studies 31,
451-460.

Dahl, R. A. 1982. Dilemmas of Pluralist Democracy: Autonomy vs. Control. New Haven:
Yale University Press.

Dowding, K. M. & Kimber, R. 1983. 'The Meaning and Use of "Political Stability"', European
Journal of Political Research 11, 229-243.

Dowding, K. M. & Kimber, R. 1985. 'Conceptual Problems in Research on Democratic
Stability'. Paper, ECPR Workshop on The Stability and Instability of Democracies, Barcelona.

Easton, D. 1965. A Systems Analysis of Political Life. New York: Wiley.

Lijphart, A. 1977. Democracy in Plural Societies: A Comparative Explanation. New Haven:
Yale University Press.

May, J. D. 1978. 'Defining Democracy: A Bid for Coherence and Consensus', Political Studies
26(1), 1-14.

Moyser, G., Parry, G. & Day, N. 1986. 'Political Participation in Britain: Images, Structures
and Characteristics'. Paper for the annual meeting of the Political Studies Association,
Nottingham.

Rae, D. et al., 1981. Equalities. Cambridge, Mass.: Harvard University Press.

Sartori, G. 1987. The Theory of Democracy Revisited. Part One: The Contemporary Debate.
New Jersey: Chatham House.

Ware, A. 1981. 'The Concept of Political Equality: A Post-Dahl Analysis', Political Studies
29(3), 392-406.