Abstract and Keywords
It was argued earlier that active civil rights require agencies to formulate, maintain, and harmonize them; the question arises then, whether there are any kinds of governmental agencies that would be apt to produce and enforce rights. It could plausibly be argued that democratic institutions—universal franchise (on a one‐person, one‐vote basis), regular and contested voting operating at two distinct levels (the level of parliament and the level of general elections), and majority rule—can effectively perform this job and thus provide the setting required by civil rights. Democratic procedures are a stable and relatively reliable way of identifying, and then implementing, laws and policies that serve interests common to the voters or to a large number of them, presumably, at least a majority.
This key argument is deeply ambiguous; it covers several disparate options. We do not want to eliminate any of them from the list altogether; the best solution, then, would be to try to rank these options in some definite order. The chapter concludes by laying out the idea of a ranking procedure that would be acceptable to all the voters: the theme advanced here is that we should go to the basic practice itself (as outlined in the three democratic institutions and their various rationales, as found in Condorcet and Duncan Black and others) to try to establish an internal ground, one that can be located in the practice itself, for deciding on a ranking; the resultant ranking of options, if it could successfully be achieved, would thereby become part of the very justification for having and relying on democratic institutions.
In the previous chapter I argued that the set of universal political or civil rights in a given society is important because it includes all human rights in that society and, more generally, because the set includes many, perhaps most, of the high‐level commitments in that society for the well‐being of individual persons as individuals. Now, we cannot assume that civil rights will be a feature of every society. So the question naturally arises: What would it take for there to be active universal political rights in a given society?
Clearly, it would take a coordinated set of agencies to formulate, maintain, and harmonize rights. Thus, it would take a government of a fairly specific sort. But what sort? of what form? We are asking, then, whether there are any kinds of governmental agencies that could reasonably be expected to formulate and promote civil rights, that could be expected as part of their characteristic operation to produce and maintain such rights.
It would not be implausible to suggest in this regard that democratic institutions are somehow conducive to rights, for many people see a connection here. But the exact relationship, though often thought to be positive, is peculiarly elusive and has remained quite unclear. In what follows I want first to spell out more fully in the present chapter what might be meant by democratic institutions in such a claim and then to show in the next, as precisely as I can, what the crucial relationship is between these institutions and having civil rights.
More specifically, in the present chapter I want to begin by characterizing democratic institutions. Here democratic institutions are identified under three main headings: (1) universal franchise, on a one person, one vote principle, (2) competitive voting, (3) majority rule (understood, in the focal case, as decision making by a majority of first‐place votes). Then, I will consider what might justify democratic government (in particular, majority rule) so understood. There is, however, as we will note, a deep ambiguity in the proposed justifying ground insofar as majority rule is concerned, for a variety of disparate kinds of outcome are all supported there. Some plausible move to dispel this ambiguity is required. Accordingly, a decision apparatus is constructed (by reference to the other features of democratic institutions) so as to remove this ambiguity (or order these kinds of outcome) and, hence, complete the justification. The leading idea here, which (p.128) entirely animates this chapter and the next, is that that which justifies majority rule—indeed, justifies the whole set of democratic institutions—is what makes democratic government peculiarly conducive to the production of civil rights. Or, to put the leading idea more precisely: coordinated agencies which follow a program of justified democratic procedures can be expected, when staying in character, to tend to the formulation and maintenance of civil rights.
1. A Characterization of the Three Democratic Institutions
The three institutional headings identified in the previous paragraph—for simplicity I will call them the three institutions—do not define practices that would be found in all societies that call themselves democratic today. This is partly because democracy is an essentially contested concept, and there is rivalry about how democracy is properly to be understood.1 So I hesitate to describe these particular institutions, without qualification, as democratic ones. It may be that they are institutions only of one form, one preferred interpretation of democracy. And even where we restrict ourselves to just the one historic form, as found today principally in Western parliamentary democracies, we learn that these institutions have come to be emphasized, as primary institutions, only in comparatively recent times.
We can see this last point readily by looking at eighteenth‐century parliamentary democracy in the United States in contrast to its presentday version. Two main differences would be immediately noted: in the extent of the franchise and in the degree to which the staffing of the offices of government (here restricted to those involved in lawmaking) is to be decided by contested elections. Let me say a word about the first of these.
1.1. First Institution: Universal Franchise
Voting in colonial and in early constitutional America was highly restricted. All women were excluded, as were slaves and, of course, children and resident aliens. In short, the situation was not significantly different from what it had been in Athenian democracy during the fourth and fifth centuries B.C. Moreover, where some property qualification was in operation (as it was in most places in the United States), the restriction was quite possibly even greater than it had been in classical Athens.
In contrast, the voting privilege (or franchise) in the United States today is virtually universal. Women now have the vote (a phenomenon largely of this century); there are now no slaves and property qualification has disappeared. For practical purposes, the franchise is coextensive with the class of adult citizens.
(p.129) Children (that is, all young people below the age of 18) and resident aliens are, admittedly, still excluded. But young people are not excluded once they reach a certain age; then they are included along with everyone else. And resident aliens can have the vote if they decide to become citizens (an option open to most, indeed most all, of them).
For our purposes, we can say that the franchise is universal when all those normally subject to the laws in a given country are able to vote; and where they are not, the exclusion is not invidious and not permanent, for it is not set on grounds of unalterable traits—that is, traits unalterable over time—with respect to any given person subject to the laws. The idea of universal franchise, understood in this way (as 100 percent of those subject to the laws, at least in principle and over time), defines one important institution in contemporary parliamentary democracies.
The idea of one person, one vote is to be understood within this context, for it serves here merely to qualify the notion of universal franchise. It means that each participating citizen has the same vote—one and only one vote—in any given voting round (e.g., a general election) on any given topic (e.g., a particular contested election between two candidates or a yes/no vote on a single proposition, such as a specific amendment to the state constitution). Thus, on this principle any voter (chosen at random) in any given voting district will have a vote‐casting capacity identical to that of any other voter in any given voting round etc. for which each is eligible.
The principle of one person, one vote is to be contrasted with plural voting: that is, with the situation as described above except that some voters do have more votes than others—three votes, for example, as compared with someone else's one. Some actual voting situations have come close to plural voting in effect. For instance, Britain ended—but only in 1948—a system of voting (available to graduates of Oxford and Cambridge) which allowed each of them two votes for parliamentary representatives: one as residents of a constituency and one as members of a university corporation that elected its own representative.
Plural voting, though it has been advocated by influential people such as John Stuart Mill, never really caught on.2 Instead, the notion of one person, one vote has taken hold in all parliamentary democracies today. One can say, loosely, then, that in such a democracy everyone has the voting privilege and they have it equally (no one of them having more votes to cast than anyone else).
There are, of course, other ways to interpret the one person, one vote principle.3 But I will continue to emphasize this interpretation, in particular, in the account as it proceeds, for it identifies the central case.
In any event, what distinguishes contemporary parliamentary democracy from other forms prevalent today is not the institution of universal franchise, but rather that of contested voting. Indeed, universal franchise is important in our account largely as a backdrop to this institution.
Contested or competitive voting, on a basis of a universal franchise, is a rather complex idea. For our purposes, three features can be identified as central: (1) there is fairly frequent periodic voting, (2) competition occurs between alternatives in the vote, (3) the alternative selected is the one with the most votes.
Different models can satisfy the idea of contested voting. Let us consider first a simple “direct” model (for example, voting in a club or in a department of a university). Here the voters, acting mainly for themselves and not as agents for someone else, meet on a fairly regular basis to decide issues by ballot. The ballot is itself shaped by preliminary discussion and sometimes by preliminary voting. At each stage people can speak for or against the alternative proposals that are to be voted on. And when the vote is held the alternative with the most votes is the one selected, either to go on to some further stage of voting or to be the final choice.
We might allow the preliminary voting to proceed on a somewhat different principle. For example, in my department a vote is taken on candidates nominated for an annual lectureship. Usually the list is long and names are systematically eliminated by dropping the name with the fewest votes in each round. It occurs, though, that in the final round the alternative selected is the one with the most votes. What is crucial in competitive voting is that this be so. The final choice must go to the alternative with the most votes.
Voting on this model is called competitive or contested for several reasons. Alternatives are on the ballot (and we presume them to be genuine alternatives in the sense that they are different and only one can be chosen). Relevant alternatives can be added to the ballot on the initiative of members. The merits and demerits of the different alternatives are debated and various points of view heard (or at least there is opportunity for them to be heard).4 And, finally, the alternative with the most first‐place votes is selected and the others are defeated. Or, if some sort of weighted vote is involved (taking account of first‐, second‐, third‐place votes, etc.), then the one with the highest weighted total is selected and the others are defeated.
This means that the winning proposal is put into effect or kept in effect (for now at least) and the losing ones are not. The qualification (“for now at least”) is added because the decision now made could always, in principle, be overturned or significantly modified at some later point on the same basis, that of contested voting. For it should be clear that the resort to contested voting, as a way to make decisions, is not a one‐time thing but recurs instead on a regular basis.
We have now canvassed, if only briefly, the central features of competitive voting. The notion of contested voting is to be understood as defined by all of these features. Let us take one further step. When many decisions, including (p.131) important ones, are made by the procedure of contested voting (as defined), we can say that this particular procedure has been institutionalized. In our account we are interested only in institutionalized contested voting.
So far we have a model, a simple one (that of direct voting on issues or proposals), for institutionalized competitive voting. It is not, however, the model that best fits the case we actually began with, the case of contemporary parliamentary democracy. For one of the distinguishing features of this particular form of democracy is that most of the decisions made by voters, as participants in a universal franchise, are not decisions on issues or proposals per se but, rather, decisions on representatives, who then do the bulk of the job of voting and thereby deciding on issues. Even when the voters do decide directly on issues (such as a school closing or a state constitutional amendment or some issue of public policy, e.g., a state lottery or a municipal nuclear‐free zone) it is often the case that the issue has already been framed and to that extent determined through debates and votes by those representatives.
In any event, it is correct to say that the voters in a parliamentary democracy exercise their franchise largely by electing or not electing persons as representatives. Thus, in order to accommodate the fact of “indirect” or representative democracy—a fact characteristic of contemporary parliamentary democracies—we need to qualify our simple model.
We could do so by identifying a model with two distinct stages, in the order indicated: (1) the electoral stage, in which voters, on a basis of universal franchise, engage in institutionalized competitive voting in order to select their lawmaking representatives, (2) the lawmaking stage, in which the duly elected representatives, themselves acting on a basis of universal franchise within a representative body, engage in institutionalized contested voting in order to organize their business and to decide policy proposals, public laws, and so on. Insofar as each stage reproduces the leading features of institutionalized contested voting, in a way appropriate to that stage, the elaboration of the simple model to cover the special case of lawmaking by elected representatives does not seem arbitrary.
Of course, we realize that some changes might be made, or thought desirable, once we had this expanded model in view. For example, the electoral stage (if it involved competing parties) might be organized to reflect, in the body of elected representatives, the proportion of votes each party had in the election that chose that body. Thus, the principle that the alternative with the most votes wins (and the others lose) would be modified to say that the alternative with the most first‐place votes ends up with the most representatives, and so on, down to the alternative with the least number of votes (which loses out altogether or, if not that, has the least number of representatives). Or proportionality could be allowed to range not over the first place votes but over weighted‐vote totals (where preference (p.132) orderings had been expressed as well). Or some method of selecting individual representatives in individual electoral districts might be chosen, on the principle that the most votes win in any given district. And here some care would have to be taken to see to it that the individual districts were more or less equal in size (so long, that is, as the districts were being conceived of as electoral districts and not as something else, as ethnic blocs, for example, or as geographic entities or as “sovereign” states).
But the leading idea in all the cases mentioned is that the lawmaking stage—once the representative body was constituted by election—would duplicate the main features of the simple model (for contested voting) in its decisions about policy and public law. There might, even so, still be important differences between representative lawmaking and direct lawmaking by voters. For example, in a direct democracy voters might stay more involved or become better informed. And their characters might be improved by such active participation (or at least Aristotle and Mill thought so).
In any event, the inclusion of a model for indirect democracy is intended to accommodate the fact that such a model is prevalent in parliamentary democracies today. My aim here is simply to show what would count as competitive or contested voting in such a setting and to show why people in these democracies are inclined, by and large, to regard parliamentary government, when it conforms to the two‐stage model I have outlined, as democratic.
1.3. Third Institution: Majority Rule
Or, at least, they are so inclined when the institution of majority rule is incorporated into the two‐stage model. For it is one of the signal features of contemporary parliamentary democracy to require that the winning alternative in the lawmaking stage is to be decided, not merely by the most votes, but by a majority vote—by over 50 percent of the first‐place votes of those voting.
I realize, of course, that other patterns of majority—or plurality—voting are possible (for example, using preference scales rather than first‐place votes, with first‐place votes used to break ties), but I have chosen this one, to fix attention, because it is the one particularly emphasized by advocates of parliamentary democracy. And it does define, to some extent, one of the main strands in the actual pattern prevalent today in parliamentary democratic states.
The important thing to note here is that majority rule, like universal franchise, is intended to qualify the notion of contested voting. Specifically, it identifies and makes more precise the idea of a winning vote. It says that—in a sequence of votes—the final vote, the winning vote, the vote that is put into effect (for now) is to be a majority vote.
And if we followed through on our earlier two‐stage analysis of contested (p.133) voting, we could insert into the electoral stage as well the requirement that the winning vote is to be a majority vote. Thus, we would have majority rule in that two‐stage model when (1) the voters select their lawmaking representatives under the requirement that each representative selected is to have a majority of the first‐place votes cast in the particular contest in which he or she was involved and, then, when (2) the elected representatives, a working majority of whom have the support of a majority of the total election vote cast, subsequently organize their business by majority vote and decide on the final form of laws and public policies, again, by majority vote (by over 50 percent of first‐place votes in every case). Thus, the idea of a majority electoral base for any given elected representative translates (by the qualification introduced in ) to the representative body as a whole.5
Now, quite obviously, it is not the case that important decisions regarding law or policy are always going to be decided within the context of the two‐stage sequence just identified. For unelected courts (e.g., the U.S. Supreme Court) and unelected administrators (e.g., the U.S. Federal Reserve Board) often do make important decisions. In order to talk of majority rule in such a situation, we would have to say that most important decisions of law and policy are made in accordance with the model of the two‐stage sequence, by majority vote, and that other decisions (by courts and administrators) are subject to these majority decisions.
The notion of ‘subject to’ here is quite vague. It could mean that the decision of a court could be overturned by a majority decision (e.g., a modified version of a law struck down by a court is subsequently passed, this time with success) or that an administrative agency or a court acts within the confines of a law itself passed by majority vote—or could be so constrained if such a law were passed.
One thing does seem clear, though. The decisions of courts and administrative agencies could not be subject to the majority votes of elected representatives unless there was an appropriate publicity about these decisions. The legislators would have to know, or be able readily to know, what these administrative or judicial decisions were and know—in at least a general way—what was being done by court or administrative officials in order to carry out these decisions. By the same token, if we took seriously the extension of majority rule to include (as it does in the two‐stage model) a majority electoral base, then the notion of appropriate publicity would itself have to be extended in that direction. Voters too would need to be in a position to learn about and discuss administrative or judicial decisions and to know or feasibly come to know, in reasonably timely fashion and at least in general outline, what policies and actions had been taken pursuant to such decisions.
Of course, it would be difficult, in advance of all contingencies, to spell out and especially to institutionalize the notion of appropriate publicity; but we often are in a position to know that the notion has not been well served in (p.134) an individual case. More important, we are always able (given the two‐stage model for majority rule) to insist on the importance of the principle of appropriate publicity. I will return to this point again, later in this chapter and in Chapter 9.
In the account I am suggesting, then, we have an ideal model of majority rule—or rule consistent with the principle of majority rule—when (1) the final stage of many important decisions is decided by more than half of the first place votes cast, (2) where those votes are cast by elected representatives who have engaged in institutionalized contested voting at the lawmaking stage and who themselves have been selected, in such a way that the working majority of the representative body reflects a base of majority electoral support, by (3) universal franchise electors who have themselves engaged in institutionalized contested majority decision voting in a prior electoral stage, and when (4) decisions about law and policy not reached as in (1) through (3), for example, by courts and administrators, are subject to those that are.
1.4. Problems With Majority Rule
There are many problems with the account of majority rule I have offered, even after we set aside its vague, almost impressionistic character. Let me mention two problems in particular.
First, I have specified that better than half of the first‐place votes actually cast is sufficient to count as a majority decision. Why not say over 50 percent of all eligible votes?
Take a legislative body with 100 members (e.g., the U.S. Senate). If we specified a vote of better than half of all eligible voters, then every winning vote in this body would require at least 51 votes. But often fewer than 100 votes are cast in a given contest. (Senators are frequently away or they abstain from voting.) It does not seem reasonable to count these nonvotes as, in effect, votes against. Rather, the option chosen has been to subtract the nonvotes from the voting pool and to require a decision only by a majority of the votes actually cast. So long as no significant inhibition on voting is imposed on those who take the nonvote route, there seems to be no issue of principle involved. For a vote of more than half of the first‐place votes of all those eligible to vote was always a viable possibility.
By the same token, a winning vote of less than half of those eligible to vote can be tolerated at the electoral stage if no significant inhibition on voting was imposed on those who chose or took the nonvote option. For, again, a vote of over 50 percent (of first‐place votes) by all those eligible to vote was always a viable possibility; and the decision, then, to go with better than half of the votes actually cast poses no problem of principle.
This is an important consideration when, as is consistently the case in many Western parliamentary democracies, large numbers of eligible voters (p.135) do not vote in any given election; for here the winning vote, in any electoral contest, is never or only rarely going to be better than 50 percent of those eligible to vote. Accordingly, we will simply specify from now on that any vote—at the electoral or at the lawmaking stage—will be acceptable, under the stipulation of no significant inhibition, if it constitutes over 50 percent of the first‐place votes actually cast.
This brings us to the second and more serious problem. For the real problem comes, not in choosing to go with a vote of better than 50 percent of those actually cast, but in making a decision with a mere plurality vote—that is, a vote of less than 50 percent of the votes actually cast. Yet this, in fact, is what the Western parliamentary democracies routinely do at the electoral stage. Thus, to take a typical case, candidate A might have 48 percent of the vote cast and B, the nearest rival, 44 percent. On the principle of contested voting followed in most parliamentary democracies, candidate A would be selected—though on a principle of majority rule A ought not be.
The thing I just described is not exceptional but is, rather, typical. And, multiplied many times over, it can lead to a mere plurality base, at the electoral stage, for the working majority in representative bodies. Let me be more concrete now.
All contemporary parliamentary democracies operate with political parties. Three main patterns have emerged: disciplined parties in a two‐party framework (e.g., Great Britain), loose parties in a two‐party framework (e.g., the U.S.), and a multiparty system (e.g., Italy).6 The problem is that, in each of these cases, events have transpired to make majority rule at the electoral level difficult to achieve. Thus, the party that comes to power through electoral victory in a two‐party framework, a frame designed to secure majority rule, is often too loosely organized to function as an instrument of majority will. And, more important, there often is no discernible majority will (since Presidents in the U.S. can be elected, and the majority party in Congress too, with only a plurality of the total vote in an election). Or the party that comes to power may be a disciplined party (as in Great Britain) but quirks of the electoral system and the gradual weakening of the two‐party framework has led in these cases as well to mere plurality government, not government that reflects majority support at the electoral level. It may well be that disciplined parties (or, to take another case, ideologically oriented ones) in a situation of universal franchise and contested voting at the electoral stage, especially in a class‐divided society like Great Britain, naturally tend to a multiparty framework. And multiparty systems are notorious for the fact that no one party has, or even realistically could have, a majority electoral base. The tendency in multiparty systems, then, is to form a coalition government. But the coalition itself does not have—since it is typically formed after the election and was not voted on as a coalition—majority electoral support. And, more important, it is possible for coalitions to be formed, as governing coalitions for purposes of making law and policy, (p.136) even though the parties involved had only a plurality of the total election vote and not a majority.
One might be tempted, given these facts, to see plurality rule in elections as an acceptable alternative to electoral majority rule (as most parliamentary systems do in practice). And it might well be acceptable, practically speaking, if the winning vote was high enough (say, over 40 percent) or runoffs were allowed or parliamentary or presidential elections were held fairly frequently.
Thus, one might be tempted to revise the earlier account of majority rule so that it now reads (1) plurality rule at the electoral stage, (2) majority rule at the lawmaking stage. But such an amendment would run directly counter to the reigning ideology (as distinct from the practice) of all contemporary parliamentary democracies—for that ideology embeds majority rule, specifically, at the electoral level and regards representative lawmaking as peculiarly justifiable for this very reason, that it rests on a foundation of majority electoral support.
This, then, is the crux: the ideal of majority rule, to which all these democracies subscribe, is belied by their consistent inability to put this ideal into play. What we have here is a serious gap between ideology and practice.
Whether this is a cause for concern—and, if so, what should be done about it—is something we cannot yet say, however. For we have not yet determined that majority rule is an ideal that matters, that it is a reasonable or desirable procedure to follow.
Clearly, if it is, then practice would need to be reformed at a number of points—through equalizing electoral district sizes, through changes in ballots and in the way tellers assess returns, through use of runoff elections between, say, the top two vote getters, through submitting coalition governments to electoral ratification, what have you. Even then it may be, since the set of constitutional institutions in any society is very complex, that no perfect coordination of democratic institutions with the other institutions is feasible. So we should aspire, at a minimum, to incorporate improved democratic institutions into the constitution in such a way that majority rule (as originally defined, as functioning at both the electoral and lawmaking stages) is more likely, rather than less likely, to occur in a given parliamentary democracy. But we need not, as I have already noted, turn to this question of institutional design until we have resolved the prior question of the intrinsic desirability of democratic institutions.
In this chapter I am discussing one issue only: what might make contested majority rule voting on a basis of universal franchise desirable. I will not be, except incidentally, engaged with questions of institutional design.
I will assume for now, however, that the instinct expressed in democratic ideology is sound—the instinct that majority rule at the electoral level is, arguably, an essential institution among those that go to make up the form of government we call parliamentary democracy. On this conception, then, (p.137) the set of democratic institutions would define one important aspect, perhaps the most important aspect, of the political constitution in such a society. Here (1) voters, on a basis of universal franchise, (2) engage in institutionalized contested majority decision voting to elect representatives, who (3) subsequently engage in institutionalized contested voting, in a way wholly consistent with the principle of majority rule, to organize their business and to decide on important matters of law and public policy. We would, of course, add the proviso that decisions not reached as in (1)–(3) would be subject to those that are.
This schema, designed to identify the main democratic institutions in a so‐called indirect or representative democracy, would have to be modified to be applicable to direct democracy. Here we would simply say that (1′) people, on the basis of universal suffrage, (2′) engage in institutionalized contested voting in order to decide, but only by majority vote, on important matters of law and public policy, with the proviso that (3′) decisions not reached as in (1′) and (2′) would be subject to those that are.7
Thus, if electronic direct democracy ever replaced representative democracy—thereby allowing people (by using home computers or telephone buttons) to vote directly on laws and policies—this is the form it would take.8 But in either case—that of direct democracy or that of representative democracy—the same institutions would be present in a recognizable and appropriate form: universal franchise (on a principle of one person, one vote), contested voting, and majority rule.
2. The Desirability of Democracy: A Plausible Account
Now the problem we have set is to determine why these institutions, in particular, the institution of majority rule, would be thought reasonable or desirable ones for a society to have in its political constitution. Or to pose the issue slightly differently, we could ask, Why might these constitutional institutions commend themselves to people who were already committed, or fated, to living together in a single society in which there would be, inevitably, some rules binding on all? Our concern in such a setting is the rationale for making decisions, under conditions of universal franchise, by contested voting on a principle of majority rule—regardless of whether these decisions are made in accordance with a simple direct democracy model or with a two‐stage representative democracy model.
One thing seems clear here: a decision procedure based on majority rule is not designed to secure the interest or the advantage or the maximum benefit of some one particular person, known to us by proper name, or some relatively small groups of such persons. For it is surely too much to expect that, as a matter of course, majority rule voting would routinely support or advance the prospects of that one person or that one group in particular.
(p.138) More generally, a policy endorsed by majority decision, though it might offer some advantage to any one person A or to persons in any relatively small group A, would probably be less in the interests of A than policies directly tailored to A's preferences or than policies which might be developed if that person or group were acting on its own, in relative isolation or with a relatively free hand. For we expect majority voting to identify and advance interests that are widely, rather than narrowly, shared in the voting population, and to advance these more general interests even where, for any person or group A (taken at random), there was some other policy that was more in A's interests.
Majority rule is a mechanism, not of individual, but of social choice. It is a mechanism designed to secure general interests, the interests of many people, not the preferred interests of a given individual or of a relatively small group. The question is, Why should a group of people install this decision procedure, rather than some other, for choices involving matters of social or general interest?
2.1. The Basic Rationale for Democracy
Rule by a majority of first‐place votes is thought desirable, presumably, because it is a relatively stable way of coordinating interests of people in the overall group. In short, the main rationale for contested voting, on a principle of majority rule combined with universal franchise, is that this is a reasonably reliable way of identifying policies or laws that serve interests common to the voters or to a large number of them and, then, of implementing such policies.9
The argument in support of this rationale sets out from the claim (presumably intuitively sound) that each voter is, more likely than not, able to judge that a policy or law is in his or her own perceived interest. Now, we have no reason for thinking that, when some policy also happens to be in the perceived interests of others, voters would be any less able so to judge. Accordingly, we can extend the claim to cover these cases explicitly. Indeed, that claim might seem even more acceptable when it was focused specifically on shared or coordinated interests, especially those perceived interests that are widespread among the voters.
For here there is no need—or considerably less—for the individual voter to estimate and discount the contrary interests of others. And there is no need, or considerably less need, to calculate over a heterogeneous and potentially divergent set of interests and to engineer complicated majority coalitions from among them (and to continue to effect the tradeoffs required to keep such coalitions together). Here, instead, the discussion of issues would tend to focus on the various widespread interests being served and on the connection of policy with them, and this connection would be brought home to each voter—with the result that the voters would be better (p.139) informed about the connection in their own case and the identifiable interest being served. Further, we might be inclined to think that the judgment of any given voter is more perspicuous or more reliable when it is able to discern some single policy that served an interest common to all, or one shared at least by a great many.
Thus, we might emphasize common or widely shared interests as a favored case. Here the focal claim would be that any given voter and hence each voter is, better than half the time, reflectively able to judge correctly that a policy or law is in the voter's own interest (as the voter perceives that interest), in particular when the interest is common to all the voters or is shared by indeterminately many of them or can be coordinated with such common or shared interests.
Putting the claim this way seems closer to the main rationale for majority rule as initially set out. And the claim does accurately reflect the democratic ethos, which reposes great faith in the judgment of individual voters in precisely such cases. Whether this faith is misplaced has been a matter for much discussion, for the claim is certainly not empirically true of literally everyone in the actual political world. But we need not enter into that discussion here, except to note that the claim in this form—and taken for the most part—is not inherently implausible, especially where the standard of appropriate publicity has been met and voters are reasonably well educated and attentive.
Now that we have these preliminaries in hand let us turn to the main consideration. I will lay it out as a single argument, with two parallel formulations. In the first formulation we assume a probability, for each voter, that the voter is correct more than half the time about whether or not policies or laws are in the voter's own interest (as the voter perceives that interest). For simplicity of presentation we will assume as well that each voter has an equal ability to make such a judgment. It follows, then (for it can be shown to follow), that there is an even greater probability that the majority of voters will be correct about whether or not policies or laws are in their interests—that is, in the interests of a whole group of them, presumably at least a majority.
Certainly, some of the things in the interest of any such majority of voters will be things common to all the voters or widely shared by them. In accordance with the central claim, then, we will focus on precisely such common or shared interests. We take up this particular standpoint—that of the central claim—because it is probably the most plausible way to state the contention at issue. But we should be clear about the point being made here: it is not that the probability value of the social decision is higher in the case of common interests than in the case of a mere majority interest, say; it is, rather, that the claim which emphasizes them (that is, the so‐called central claim) is more likely to be true than any parallel claim which speaks simply of majority interests, without further qualification.
This should be obvious for the reason already given, at the beginning of (p.140) the previous paragraph, that things in the majority interest are often things common to all voters or widely shared by them. It is also obvious because all mere majority interests are actually included within the central claim, as but one of several available special cases that it is intended to cover. Thus, as more likely to be true, the central claim is a more acceptable claim; so we reformulate our argument in accordance with it. And this reformulated argument becomes, then, the basic argument.
In this revised formulation, we conduct roughly the same argument, this time starting out from what I have called the central claim. Here we focus, when we make the assumption that the voter is able to judge correctly more than half the time respecting policies or laws that are in the voter's perceived interest, on those interests common to the voters or shared by them or by indeterminately many of them. And we utilize the same simplifying assumption we made in the initial formulation, of an equal probability in each voter's case. It follows, then, that a contested majority vote, on a universal franchise basis, would exhibit an even greater probability than any given voter on his or her own of being correct about policies or laws in the voter's perceived interest, whether that interest is an interest common to all the voters or an interest shared by indeterminately many of them (presumably a majority) or an interest that can be coordinated with such common or shared interests.
Thus, we reach the rationale that majority rule is a reliable and stable social decision procedure to follow when we want to identify, in order to implement, policies or laws that are likely to serve just such widely shared interests, the interests of a lot of people. Namely, interests that are common to all voters, etc.10
The point with which I began this chapter—that there is a positive connection between the creation of civil rights laws, on the one hand, and the use of democratic procedures, on the other—is seemingly supported by this analysis. (More, of course, would need to be said, though, to make this particular contention perspicuous and to bring it to conclusion.)
Before we proceed, though, two things about this rationale should be noted. First, it draws on the direct democracy model in justifying rule by a majority of first‐place votes. And, second, it justifies such rule on epistemic grounds. Let us consider these points in turn, briefly.
2.2. The Compatibility of the Two Models
Even though the direct democracy model was used in the basic argument, it is not clear that laws and policies would be better under direct rather than indirect lawmaking (each on the same principle, that of rule by a majority of first‐place votes) or even that such laws and policies would necessarily be more representative of the public will, or the preferences, of voters. For we can readily conceive a majority of elected representatives, in accordance (p.141) with the central claim, as exhibiting an even greater probability than any given representative on his or her own of being correct about policies or laws in their and the voter's perceived interest, especially when that interest is an interest common to all (voters and representatives) or an interest shared by indeterminately many of them (presumably a majority) or an interest that can coordinated with such common or shared interests. Here, in this indirect democracy version of the argument, the voters are conceived principally as exercising a check on representatives, removing some through electoral defeat and helping others to form legislative majorities through election and reelection.
The same considerations are seemingly at play in elections as were present in voting on policies (at least as regards those considerations that were captured in the direct democracy argument). For it is not implausible to say (especially where the assumptions of appropriate publicity, an educated populace, reasonably attentive voters, and so on, hold good) that individual voters do in fact correlate their interests with policies, as one factor in their decisions about whom to vote for, and do in fact make a rough‐and‐ready calculation about which candidate is more likely to act or vote in favor of preferred policies, should the occasion arise. Thus, a rough parity between laws and policies approved by a majority of representatives and those approved by a majority of the voters, the benchmark goal, is a viable outcome—indeed, a not unlikely outcome under institutions well designed for achieving that particular goal (and assuming appropriate publicity, etc.).
It is essential to make this very last point, for otherwise any preference for representative government rests simply on practical grounds or reasons of economy. The point is, representative government can be preferred on such grounds over so‐called direct democracy, within the confines set by a theory of democratic institutions, only where the benchmark goal is a viable outcome, or can be made so through refinements in institutional design.
Of course, there is one special interest legislators have that may be incompatible with the interests of voters. I mean the desire of legislators (and administrators and judges) to stay in office, to gain special perquisites, and to get big pensions at the public's expense. Many legislators will, in effect, conspire to give advantage to incumbency and to preserve themselves in their positions. The voters need to offset this special interest in incumbency, as best they can, by setting limits on campaign financing (or even by replacing private with public financing), by requiring fair exposure of the main parties and positions and candidates in the media, by setting limits to tenure in office, and so on. The basic incompatibility between legislators' and voters' interests here cannot be wholly resolved; it can only be controlled (to a degree). Institutional design is needed to accomplish this. Though the incompatibility is serious and chronic, it need not disrupt, in any fundamental way, the benchmark goal just described.
We turn next to the point that, in the basic argument, majority rule was justified on epistemic grounds. For, as we saw, the main move was from a presumed correctness in judgment on policies, of at least better than half, by individual voters to a demonstrated greater likelihood of correct policy judgment by them collectively, when they use a contested‐voting procedure based on majority decision and on universal franchise. And correctness here means that voters—including here voters in an election or representatives in a policy vote—can perceive when a policy is in their own interest and, presumably, when it is in the interest of others as well. For we do not assume that the voter is any less likely to perceive which policies or laws are in the voter's interest when the voter has an interest in common with, or shared with, many other voters.
This epistemic understanding of the justification of majority rule is in line with the classical rationales offered by Aristotle and Marsilius and Rousseau.11. And it is quite different from saying that majority rule is justified because it gives more people, rather than fewer, what they want or desire.12 Or from saying, as did Locke, that the body politic can stay intact only if it moves that way “whither the greater force carries it, which is the consent of the majority.”13
It is important to be clear also on the argument that supports this particular epistemic rationale or justification for majority rule. The argument is probabilistic in form. The argument, put starkly, is that a probability of correctness, for each individual voter, of greater than half (and for simplicity we assume an equal probability for each voter) would yield a majority rule social decision having an even greater probability of correctness (as to which policies were in the perceived interests of an indeterminately large number of these voters). And it could be shown, further, that the probability value of the social decision will increase with an increase in any (or in two or in all) of three factors: (1) the absolute size of the difference between the number of majority voters and the number of minority voters, (2) the probability of correctness of each individual voter, (3) the size of the majority vote, expressed as a percentage of all votes.
Take just (1) and (3). To illustrate, Brian Barry says:
One final point of clarification is in order. The supporting probabilistic analysis here is best understood—properly understood—as ordinal in (p.143) character. Thus, we should say merely that where the probability of correctness of each individual voter is presumed greater than half—i.e., p(c) >1/2 (and, for simplicity, we assume that the value p(c) is the same for each voter)—then a majority vote yields a social decision p(s) with an even greater probability of correctness, i.e., p(s) > p(c). And, moreover, the probability value p(s) increases here, p(c) remaining constant, with an increase in either (or both) of the factors identified in the illustration just quoted. The important point here is that we need not plug in precise amounts to give a cardinal value for p(s), such as 69 percent or 99.97 percent, or for p(c), such as 51 percent (the percentages given in the illustration above).
[I]f we have a voting body of a thousand, each member of which is right on average fifty‐one percent of the time, what is the probability in any particular instance that a fifty‐one per cent majority has the right answer? The answer, rather surprisingly perhaps, is: better than two to one (69%). Moreover, if the required majority is kept at fifty‐one per cent and the number of voters raised to ten thousand, or if the number of voters stays at one thousand and the required majority is raised to sixty per cent, the probability that the majority (5,100 to 4,900 in the first case or 600 to 400 in the second) has the right answer rises virtually to unity (99.97%).14
A formula, such as the one relied on in the illustration, that included precise evaluations of probability would, of course, support the ordinal claim; thus it can provide evidence for the truth of any such claim. But the ordinal claim does not reduce to any such precise formula nor does it require one. All that is needed to make the crucial probability analysis work is the idea that one probability value can be estimated as greater than another—e.g., that p(s) > p(c)—and that one such value can estimated as greater at one time than it is at another—e.g., that p(s) at tn > p(s) at t n − 1.
The main value of using the ordinal approach is that it captures the important relations but without commitment to precise probability values. For often we have only passable ordinal estimates, that this probability (say, p 1) is greater than that one (say, p 2), but are quite unable to say how much greater it is (as, typically, we are unable to say exactly what the probability of voter correctness is over against the probability of voter error in a given case). Thus, the use of an ordinal analysis is to be preferred to Barry's (and Rawls's) cardinal analysis for the crucial probability values.15
In sum, I have distinguished an ordinal from a cardinal form of probabilistic analysis and have indicated that the ordinal analysis is sufficient to support an epistemic rationale or justification for a contested‐voting decision procedure that incorporates universal franchise and majority rule. The rationale is that, assuming an equal and better than even chance of correctly judging by each individual voter what policies are in that individual's interest, we can say that a majority decision16 by such voters would be a reliable and stable social decision procedure to follow when a social group wants to enact policies or laws that are likely to serve the interests of its individual members taken overall, in particular when those interests are common to all the members or are shared by indeterminately many of them (at least a majority), or can be coordinated with such common or shared interests.
We do not, in conducting the argument in ordinal form, desert the fundamental standard of perceived interests; for we still start with the point that each voter is reflectively able to determine, more likely than not, when policies are in that voter's interest (especially when that interest is common (p.144) to all the voters, etc.) and we still end with the rationale that rule by a majority, as specified (with universal franchise, etc.), is even more likely to enact laws or policies that tend to serve widespread interests—perceived interests—of members of that majority, taken overall.
The rationale here does not involve saying that some independent standard (such as the public good or the voter's true interest) emerges in the majority vote. Rather, the same standard operates throughout (that of perceived interests, for each voter, in relation to policies); what changes is simply that majority rule, as specified, provides a higher probability of correctness in such cases—regarding common perceived interests, interests of indeterminately many, and so on—than the presumptive probability of correct judgment by each voter taken one by one. It is this higher probability of correctness, then, that lies at the heart of the rationale for majority rule developed in the present chapter.
3. A Deep Ambiguity in This Rationale
There is, however, a significant ambiguity in this rationale, for it covers a range of different possibilities. First, there is an ambiguity in the notion of common interests, interests shared in some common object by all the members. Does this ‘all’ here mean (1) in the interests, or in the overlapping interests, of each and all or does it mean (2) in the interests of all—that is, of the corporate group of which each is a member—though not necessarily in the interest of each?17 Then, second, there is a need to distinguish these common interests—as found in (1) and (2)—from (3) interests shared by indeterminately many of the members (presumably a majority) though not by all, interests that may indeed be harmful to the interests of some (presumably a minority). Finally, these interests—as in (3)—can be distinguished into (3a) those interests that are compatible and can be coordinated with the common or shared interests—in (1) and (2)—and (3b) those that are not compatible and cannot be coordinated.18 So, when we talk of a rationale for a majority rule, it is not evident whether we are relying on (1) or (2) or (3a)—or even (3b)—or on some combination of them. And there may be other ambiguities as well.
Clearly, then, if the rationale is to be meaningful we need some principled way to resolve the ambiguities we have noted. Or, at the minimum, we need some device for ordering these options as features of the rationale for majority rule.
3.1. Dispelling the Ambiguity: Preliminary Remarks
I would suggest that one way, perhaps the best way, to remove these ambiguities—or to order them—would be to generate a procedure for (p.145) assessing the main options from within the notion itself of contested voting on a basis of universal franchise and majority rule. Let me, briefly, indicate the merit of this suggestion.
We might decide to choose between the options by some random device—for instance, throwing dice. Then if, e.g., (3b) won the toss, we could say that simply settles it. This is what we are now going to intend with the rationale, and where the interests identified in (2), say, conflict with those identified in (3b) in a given vote, we should choose those in (3b). Or we could decide to choose among the options by using some property of the voter—for example, the ability to win relay footraces (assigning to each option a relay team representative of the voting population). One could always say about such devices (first) that, insofar as we were trying to refine the rationale for majority rule more precisely, these devices seemed wholly irrelevant or (second) that they simply would not afford a good reason for choosing among options at all—at least not as a general rule. Both ripostes seem wholly appropriate.
One could, responding in the light of these ripostes, say that we should decide among the options by using an acknowledged good reason for choices. Thus, we might turn to utilitarian calculation and presumably choose (2) or possibly (3a). Or we could let the Pope decide and probably get (2). Or we might introduce Kantian considerations, about respect for each person as an equal and never treating anyone as a mere means, and decide to choose (1) over (2). And so on.
But we could always say about these devices that they are good ones if one is a utilitarian or a Catholic or a Kantian in the first place (or, if the account were extended further, a Marxist, a Hobbesian, a rational choice theorist, what have you). In addition, and more important, it is not clear that any one of these devices is necessarily the device one should use in trying to decide about a rationale for majority rule in particular. For none of them has anything to do with majority rule per se. And it seems—if we are trying to determine something about what is best intended in the rationale for majority rule—we should find reasons for refining that rationale which are themselves somehow intrinsic or connected to majority rule.
Such a stance would be especially appealing to someone who had already accepted the basic probabilistic argument for the rationale and who wanted now merely to choose between the eligible options as a way of suitably clarifying that rationale. Such a person could regard the issue as one of determining more precisely the purpose or point of majority rule, once the probabilistic argument had set the main options in place. Now it should be clear that such persons cannot rely; in order to solve this problem, on the probabilistic reasoning that was originally employed. (For that reasoning incorporates the very ambiguities we are here being called upon to resolve—indeed, they were introduced at the very beginning, in the intuitive claim from which the main argument set out.) So, such persons might think it (p.146) reasonable as well as most likely to lead to resolution, in determining the principal object to be achieved in a practice, to go to features of the practice itself. In doing so they would try to establish the best internal rationale for majority rule.
Rationales of this sort are never conclusive, as we know, all things considered. But they are often a good place to start. And if we then go on to approve or reject internally rationalized majority rule by reference to other standards, we do have some confidence that the things we are here accepting or rejecting are pertinent to the rationale for majority rule in particular.
Thus, we reach again my proposal that the best way to resolve the issue, at least as a first step, is to see if there is something about the practice of contested voting conjoined with universal franchise and majority rule that would allow us to choose among the options or to order them. There is nothing circular about such a procedure, for the main options—as features of the basic rationale—have already been argued for and decided upon on other grounds, and these grounds are not being gone over again.
3.2. Dispelling the Ambiguity: The Construction of a Decision Apparatus
Our project, then, is to use the resources of the practice we recounted earlier in this chapter, use them to help us choose between options in the rationale for that practice (a rationale itself independently established). If this revision to our earlier account of democratic institutions is successful (so that the preferred option, or the ordering of options, can be explicitly mentioned in the rationale), we will have a resolution of the ambiguity that has so troubled us. And by building this resolution into the overall account, we will also have the most convincing rationale for majority rule that could be offered from within the perspective of majority rule.
In the earlier account of democratic institutions, two main models were developed: a simple direct model and a two‐stage indirect or representative model. Some contrast between direct and indirect democracy is probably in order, but I do not think (as I have already indicated) that the difference is sufficiently great to draw forth any significant difference in principle. For the point, in any case, is that each can be justified in the same way. This is something we will come to see more clearly as we go along.
From these two models we could plausibly construct a single model which abstracts the features that they have in common. For in each model voters, leaving it undetermined whether they are electors or policy makers, engage in a regular and continuing way in institutionalized contested voting on a basis of universal franchise and of majority rule. And the decisions so reached are regularly put into effect; such decisions are the dominant decisions in the example we are focusing on.
My proposal, then, amounts to this: that we take the arena or forum (p.147) identified by such a composite or abstract model as the framework for assessing and deciding about the main options in the rationale for majority rule. The justifying rationale that emerged, in accordance with this proposal, would consist of the original rationale as modified through the ranking of options, where some or all of them had been determined to be permissible.
Here the choice and ranking of the main options has become, in effect, a feature of the justification of democracy. And the most plausible such justification is the one in which that choice and ranking is done within the forum—the framework and assumptions—of the contemporary idea of democracy itself. I will not attempt to construct this forum in every detail. I will, however, simply call attention to some of its more important features.
Now, the choice in question (the ranking) has to be done by someone. For reason, though it may be impersonal, is never disembodied. It requires voice and location. But who are these someones going to be? Who should they be?
They are representative citizens, people who will live under and with the democratic institutions and whose deep instincts have been informed by them. For just as we have specified the composite or abstract model so we can specify (if that is the word) the representative citizen. These are people who have traits that typify main features of that model and who will conduct their reasoning about the options (choosing some as preferred and ranking them) within the confines of those typical traits.
These participants (the assessors) in the arena formed by the composite or abstract model—for short we can call it simply the abstract model—would necessarily have a certain reflexive understanding of themselves. They know, or could come to know, that they are persons and fellow citizens. They know that all the members of the particular political society involved—given the stipulation of universal franchise—could be participants and are to be counted as assessors respecting the determination to be made. (I mean all the citizens except as noted at the very beginning of the chapter—excepting, then, for example, children.) The participants know as well that, in the deliberation about the permissibility and the ranking of the main options, the assessors have an equal status (an equality reflecting the stipulation of one person, one vote and the stipulation of interchangeability).19 And they know that all the assessors are reasonable beings—that is, reasonable within normal limits. For, according to the stipulation of institutionalized contested voting and of the interchangeability of voters, each participant (each assessor) is presumed able to contribute to the discussion, able to follow it, and able to reach a reasonable decision (as already specified, in the discussion of probabilities) about the main options.
We should note that no radical constraint on the knowledge of the assessors is imposed. Thus, we presume them to know that they have interests and to know, at least in a general way, what their interests are. We can also (p.148) presume them to know that others have interests as well, some common, some compatible, some incompatible with their own. But, although each is able (more likely than not) to make a reasonable estimate of whether a particular choice (in particular, a given policy) is in his or her own interest, no one is able to predict exactly what decision(s) would be reached by majority vote on a given occasion. Indeed, this is often very unclear. Moreover, matters grow progressively dim where we to put the vote and its ingredients further and further into the future.
The knowledge elements here identified precisely overlap those I identified earlier (in the basic argument) as elements each voter could be presumed to have. In the basic argument, individual voters were deciding, with a better than even probability of correctness, about particular policies, which had definite connections to determinate interests they had; and they were in usable possession (or could be, assuming appropriate publicity, etc.) of relevant local facts and the immediate historical backdrop attendant to the vote on particular occasions. The same could be said for the assessors, as I have just characterized them.
I did not add earlier (as I have now) that the voters might be less clear about the connection of their interests to policies in distant and hypothetical circumstances, where local facts could only be guessed at (or were far different than from those at present), where the policies themselves often took some getting used to (for they represented things the voters were unfamiliar with and, accordingly, did not readily understand), and where the historical background, leading from the present to a remote future, was simply missing. Nor did I add that the voters would have an even harder time predicting outcomes of votes in the distant future than do voters respecting outcomes of present votes (where, even so, the probability is presumably not high—perhaps below one‐half).
I did not add these things at the earlier point; nonetheless, they are so. If they were added, as they could be (to complete the record), then the voters as earlier characterized (in the basic argument) and the assessors as now characterized would be identically described.
I did not add these things earlier (about preferences in distant votes or about the voter's ability to predict voting outcomes, either now or in the future) because they did not seem relevant to the situation described in the basic argument. For the concern there was with an individual majority rule decision on a given occasion about a particular policy (where there was, presumably, a set of determinate options and of first‐place preferences and a fairly definite localized and historical setting in which these arose).
But these things do seem relevant to the situation we are here contemplating. For we are now concerned with majority rule as an ongoing practice—concerned, that is, with the institution of majority rule itself and with whether it should be restricted on some principle or, alternatively, designed to be wholly unrestricted. (And, for our purposes, majority rule will be (p.149) regarded as unrestricted if all the main options are allowed and no ordering is imposed on them.)
As noted, we do not assume here that the participants (the assessors) will always have a clear idea of the details of their preference scales in advance of the on‐site voting contests in which the elements in the scale and the preference order itself become determinate (for the situation we are now contemplating goes beyond the highly localized assumptions of the earlier basic argument). More important here, we now add explicitly that these assessors are unable to foresee the long‐term history of contested voting, and thereby are unable to tell what the outcomes will be and, for the most part, how these outcomes will relate to their individual interests. That is, they cannot foresee these outcomes or attendant effects on interests when no ordering constraint has been imposed on majority voting.
But each person involved in this assessment of options within the forum constructed—and this person could be any one of us—is able to imagine what effect a long‐term regime of unrestricted majority rule might have on his or her interests in particular, or on the interests of others. And to free each from dogmatic or unthinking slumber, we ask everyone to imagine that they do not know at what point they enter into the continuing sequence of such votes. Thus, each must carefully assess what tendencies a practice of unrestricted majority rule might have. Each might reasonably conclude here that such a practice is likely at some point to be inimical to the interests of any given person—certainly to some of that person's interests, perhaps to a great many. And each is able to communicate this assessment to fellow assessors and to expect a considered response. For all know that, unless things change, they must live their lives under the decisions made by a regime of contested voting on a principle of unrestricted majority rule.20
The knowledge conditions I have identified in this case are ones I take to be peculiarly characteristic of a continuing but unrestricted scheme of contested voting under democratic institutions. Thus, the abstract model, so like Rawls's model of the “original position” in other respects, is singularly unlike it in the very limited “veil of ignorance” here imposed.21
And, of course, the object of the two models is quite different as well. Rawls's original position constitutes an arena for deciding about the principles of distributing certain “social primary goods” through the basic structure of a society, principles chosen from a short list of principles of political justice. My abstract model constitutes an arena for deciding, in a society governed by democratic institutions, about the proper characterization of the rationale for majority rule—through choosing from among options that were ambiguously present in the original rationale for such rule.22
Specifically, the object of assessment in the abstract model is to decide about the permissibility and ranking of these options, to determine which options to embed within the rationale for majority rule and in what order. (p.150) The main options (and here I simplify a bit) are that the rationale for majority rule is to be understood as affording priority to choices, and ultimately to policies, which are (1) in the interest of each and all, (2) in the interest of the corporate group of which all are members, though not necessarily in the interest of each, (3) in the interests of indeterminately many (presumably a majority) even though harmful to the interests of some (presumably a minority). What has to be decided in the abstract model is which of these options should be preferred. (And if none should be preferred, though all these options are retained, that too is a decision. It is a decision in favor of unrestricted majority rule.)
Now, since the options are the ones described in (1)–(3), we need posit no special motivational assumptions about the participants. For any of the three options, absent the claim about priority, is consistent with what any participant could probably tell by introspection about oneself or others—or at least what anybody could tell from the past history of democratic voting itself. I mean each could learn that the participants are fairly self‐interested (outside the immediate circle of family, friends, co‐workers, and so forth), that the participants prefer to advance their interests rather than not to (especially when there is little or no cost in doing so), that they are willing to live with cooperative arrangements (where all benefit, though not necessarily equally) over Hobbesian state of nature arrangements, and so on.
It would, in fact, be desirable to let the idea of participant self‐interest remain sufficiently indistinct so that it directly attaches to no one of these options as preferred. For if it did so attach, then we do not have an arena for decision but, instead, a decision—one that has been built, perhaps arbitrarily, into the motivational assumption. Thus, that assumption should be realistic (for the model) but weak in that by itself it does not determine one of the options as the preferred one.
Indeed, no single feature—not even all the features—of the arena should decide the preferred option, as though by logical entailment; rather, it is the admissible arguments which could go on within the forum that should be determinative. Here the participants try to see if some arguments afford considerations that, on balance, decide the issue.23
4. The Assessment Procedure
Individual persons here are trying to determine what would be a reasonable ranking of the options and, if that could be decided, whether all are permissible as ranked. They make this determination about what is reasonable by assessing which option it is likely the participants would prefer, that is, prefer after reflection on admissible arguments within the constraints afforded by the abstract model, where all are identically situated.
Let me elaborate a bit on the decision procedure outlined here. Each (p.151) assessor, within the framework described, makes a subjective estimate of which institutional option it is most likely that an individual assessor would come to prefer in light of an estimate of what most others would come to prefer. It is an assessment made in the shadow of the uneasy uncertainty each feels about the success or even survival of the individual's own interests in the face of, or with the backdrop of, a long‐term regime of unrestricted majority rule. This estimate, then, comes within the general province of decision theory insofar as it concerns subjective judgments as to preferred outcomes under conditions of uncertainty.24
The estimate is, I have emphasized, based at least in part on arguments (though they are arguments yet to be heard). These arguments articulate the principal reasons that could be offered for selecting one option over the other two. When one finds any such argument convincing, it is (in the arena designed) convincing to all, or almost all. Thus, on the presumption that the argument for one option is likely on balance to be convincing, that argument regiments the estimates of the individual assessors into a concurrence of estimates—each estimate reflecting an individual's subjective judgment as to which institutional option it is most likely that that assessor and most others (indeed, almost all) would prefer.
It is the concurrence of individual subjective estimates that we are looking for, then; for such concurrence decides the matter. It is the function of the arguments developed in the abstract arena to make for concurrence.
It is wholly reasonable, I think, to refer the assessment, as we do here, to individuals rather than to the collective group. For only individuals can assess. We know of no intelligence—no human intelligence or intelligence comparable to human intelligence—in a group. Thus, the consensus of individuals in a reasonable assessment can be our only standard.25
And if we polled people to see if such a consensus existed, we might have a basis for saying that such a consensus did exist, but that poll would be in no way itself a vote. Indeed, we could never reach the required consensus through voting. For no vote, even a majority of first choices or, beyond that, unanimity, could be considered satisfactory until we had first determined the order of considerations within the rationale for majority rule. And this is the very thing we are here called upon to decide.
So we go then to the relevant arguments. I will take these up in the next chapter.26
(1.) For the idea of an essentially contested concept, and for discussion, see Gallie (1968: 157–91, esp. 161 and 168). For democracy, in particular, as such a concept and for an interesting typology of contemporary competing versions, see Macpherson (1966, chs. 1–3). A different typology of competing models, this one ordered historically, is developed in Macpherson (1977).
(2.) See J. S. Mill, Considerations on Representative Government (1861), ch. 8.
(3.) The U.S. Supreme Court, for example, has used the principle of one person, one vote to determine that electoral districts should be roughly equal in size of population throughout a given state. (The principal cases here are Baker v. Carr, 369 U.S. 186 , on justiciability, and Wesberry v. Sanders, 376 U.S. 1  and esp. Reynolds v. Sims, 377 U.S. 533 , on point.)
In the latter case, one person, one vote is not so much a qualification on universal franchise as it is an attempt at equalization, across electoral districts, of individual votes (expressed as a fraction of the total number of eligible votes in respective districts of residence). Thus, if 1/a expresses that fraction in district A and 1/b in district B, then the numerical value of these two fractions should be roughly identical in a given state S. Where this standard is met, then farmers, for example, who predominate in district A (which is larger in land area than district B, though roughly the same in population) do not have a greater proportionate vote, in the state assembly or in the Congress, through their elected representative than do the urban dwellers of district B through theirs.
Now, there clearly is a problem with unequal electoral districts; such inequalities (where the districts are concerned with the election of representatives to the same lawmaking body) can lead to significant distortions in the electoral basis of law or policy making. The Supreme Court's view here is correct. Accordingly, this and other acceptable qualifications will be assumed throughout when I talk of the principle of one person, one vote. Nonetheless, the central understanding of that principle—which will be emphasized throughout—is to see it in the way it is represented in the text proper, as a qualification on universal franchise when voters are compared one by one.
(4.) Rawls, in the context of contested or competitive voting, has argued that individual citizens should be guaranteed “fair value” there, a rough equality of access to political participation (see Rawls 1982a, sect. 7, also pp. 75–8). “[T]his guarantee means,” he says, “that the worth of the political liberties to all citizens, whatever their social or economic position, must be approximately equal, or at least sufficiently equal, in the sense that everyone has a fair opportunity to hold public office or to influence the outcome of political decisions” (1982a, 42). And this equalizing of the opportunity to count politically is to be achieved through such measures as (1) limits on campaign spending or (2) government funding of campaigns or (3) public subsidy (or a mandate for ready and inexpensive availability of air time) to candidates and political parties and other relevant points of view.
Again, as I said in the previous note (about the Supreme Court), Rawls's notion here of “fair value” or “equal worth” has considerable merit. His point is, there must be allowance not only for real contestation, real opposition between options, but also for the real input of individual persons in the political process if the standard of competitive voting (with universal franchise and one person, one vote) is to be a meaningful one. Accordingly, this and other acceptable qualifications will be assumed throughout—or at least left open for further deliberation—when I talk of contested voting. Nonetheless, the central understanding of that institution, as a real contest between genuine alternatives, as presented in the text proper, is the one I will continue to emphasize throughout.
(5.) We should also stipulate that in all these votes (in both electoral voting as in  (p.368) and in votes in the lawmaking body as in ) the requirement of interchangeability—or anonymity—is met (see May 1952: 681; Dahl 1956: 65–6). By interchangeability is meant, roughly, the replaceability at random of any one of the individuals that help constitute the losing portion (in a vote) by an arbitrarily chosen different voter who also holds (by hypothesis) the same view on the desired outcome of the vote as the individual replaced. Or, to make the same point, we can imagine a switch of places between any two voters, one originally in the majority and the other in the minority (or vice versa); after the switch the hypothesis in the previous sentence is said to operate, as if a magical change of mind had occurred in the crossover. The intuition in either of these formulations is that if voting really is decided by majority rule, the substitution will make no difference to the original outcome.
In the discussion that follows in this chapter, I will presume that both the stipulations—the one introduced into (2) in the text and the one in the first paragraph of the present note—hold and are satisfied. However, it is sometimes difficult or awkward to build the stipulations in explicitly; so I say this now to cover those cases.
(7.) My account of democratic institutions in this chapter draws on the work of Karl Popper (1950, ch. 7, sects. 2 and 3; ch. 17, sect. 7; and ch. 19, sects. 2 and 5), Joseph Schumpeter (1950, part 4), and Robert Dahl (1956, ch. 3). All three writers have emphasized the notion of institutionalized contested voting, in particular (see, e.g., Schumpeter 1950, ch. 22).
There are, however, important differences between these accounts and my own. For instance, Schumpeter restricts his account to representative government, while I have allowed for the two conventional models—direct and indirect. Nor does Schumpeter's, unlike my account, have any commitment to the notion of universal franchise (see 1950, ch. 20, sect. 4). None of the three authors, again unlike my account, has emphasized the institution of majority rule; indeed Dahl (1956, ch. 2) criticizes the main forms that the institution has taken historically and clearly opts (1956, ch. 3) for a plurality model at the electoral stage. And, as befits a professor of political science, Dahl (1956: 71, 84) puts especially strong emphasis on what he calls the “interelection stage” (as equivalent, roughly, to “the proviso” mentioned in the previous paragraph of the text in my account).
In none of these cases, however, including my own account, is any serious attempt made to explain how democratic institutions have arisen. And the focus in none of these cases is put on how these institutions are best to be maintained or coordinated with other institutions.
(9.) “Democratic laws generally tend to promote the welfare of the greatest possible number; for they emanate from the majority of the citizens, who are subject to error, but who cannot have an interest opposed to their own advantage” (Alexis de Tocqueville, Democracy in America , ch. 14).
(10.) The pioneer works in the theory of social choice are by Duncan Black (1968) and Kenneth Arrow (1963). For discussion and elaboration, especially as regards voting, and for a helpful bibliography, see Michael Dummett (1984).(p.369)
The argument I have developed (in two parallel formulations) is adapted from Brian Barry (esp. 1967 and 1965: 292–3). See also Rawls (1971, sects. 36, 37, 54, esp. p. 358). The point about stability is taken from Dummett (1984, ch. 4).
Ultimately, the argument in Barry's case and my own is based on a probabilistic analysis developed by Black (1968: 159–80, esp. 163–72). Black in turn was here explicating the theory of Condorcet (a mathematician and social reformer, member of the French Academy and of the Academy of Sciences, and permanent secretary of the latter organization). Condorcet's main treatise here is his Essai sur l'application de l'analyse à la probabilité des decisions rendues à la pluralité des voix (1785). This essay has never been translated, in its entirety, into English. A useful translation of a good chunk of the nonmathematical part of the essay, however, is found in Condorcet: Selected Writings, ed. Keith M. Baker, Library of Liberal Arts (Indianapolis: Bobbs‐Merrill, 1976), 33–70.
Marsilius of Padua is probably the most distinguished proponent of the theory of democracy in the ancient or medieval world. Unfortunately, his main argument here, which follows lines laid down by Aristotle, is clumsy (in that it ultimately depends on the inappropriate claim that the whole body of voters is analogous to a geometrical whole and thus that any such whole is obviously “greater than” any of its parts—in wisdom, for example); the argument is probably unsound to boot (because, from it, one could derive both p and not‐p). Marsilius's arguments are found in his Defender of Peace (1324). See Discourse (Dictio) 1, esp. ch. 12, parts 3, 5–7; also chs. 8, part 3, and 13, part 3. The objectionable discussion of geometric wholes is found in Discourse 1, ch. 13, part 4.
The essence of Jean‐Jacques Rousseau's argument is his contrast of the “will of all” with the “general will.” See The Social Contract (1762), esp. book 2, chs. 3–4, and book 4, ch. 1. See also (for further elaboration of the general will and its properties) book 1, chs. 6–7; book 2, chs. 1–2; book 4, ch. 2.
All these arguments are epistemic, in that they judge the reliability of claims to correctness and opt for one claim over others. In the end they all support the claim that the citizens deciding as a whole (in accordance with institutions recognizably democratic) are more likely to reach a sound decision in public matters than any individual citizen or small group acting alone or on its own.
There is, it should be noted, no logical connection between the epistemic rationale itself and the probabilistic analysis that supports it in my text. Thus, the epistemic rationale per se does not require the probabilistic analysis. And none of the authors mentioned in this note appear, at least at first glance, to use that analysis in their arguments. Some have suggested, however, that Rousseau can be explicated on the model provided by the probabilistic analysis (see here Grofman and Feld 1988, esp. 568, 571).
One final point. As Aristotle's argument, in particular, makes clear, expert opinion is not to be despised. But who is the expert? In some matters the individuals affected are the experts; in others (in mores, in matters of right and wrong so conceived) virtually any normal “freeborn” citizen is an expert. His point, though, is that specialized expert opinion (the opinion of “the wise,” for (p.370) example), when voiced in the context of democratic institutions, is likely to help yield a majority decision having a greater probability of correctness than that of the expert on his or her own (see Politics, book 3, ch. 11, at 1218a 15).
(12.) As Robert Bork, for example, says. Thus, “Every clash between a minority claiming freedom and a majority claiming power to regulate involves a choice between the gratifications of the two groups.” Bork continues, “Courts must accept any value choice the legislature makes unless it clearly runs contrary to a choice made in the framing of the Constitution” (1971: 9–11). For “political truth is what the majority decides it wants today” (1971: 30–1).
According to Waldron a roughly similar idea (as to political ‘truth’) lies behind what might be called the Benthamite conception of democracy (see Waldron 1990). Here, presumptively, the democratic or majoritarian result reflects the preponderant weight in the overall balance or aggregation of individual desires.
(13.) See John Locke, Second Treatise of Government (1690, sects. 95–7; the passage in question is found in sect. 96).
(14.) Barry (1967: 122). The cardinal formulas indirectly referred to in my text, and relied on by Barry in the illustration there, can be found in Black (1968: 164–5). These formulas are expansions from Bernoulli's theorem. The crucial formula used by Barry is one of several cited by Black; it is found in Barry (1965: 293), and reads:
More precisely, if each member of the group is right in proportion v of the cases dealt with, and wrong in proportion e, where v + e = 1, then if in a given instance h members of the group give one answer and k members the other answer (h > k), the probability that the h members are right is:
(15.) Black is careful to distinguish the ordinal form of the argument (1968: 170–2) from the cardinal form (1968: 164–5) and to indicate that Condorcet's argument (despite initial appearances, as in the formula in n. 14, from which the argument starts) is of the ordinal form. Black's point here is that Condorcet did not rely on the formula in any simplistic or straightforward way, applying it mechanically to all cases; rather, he relied on the complex ordinal claim—and here we use the notation introduced in the text—that p(s) > p(c) and thus (1) in a contest between two options the decision should go to the majority and (2) in a contest among three or more options, that option which was supported by a majority over all the others (if there was any such option) should be selected. (See Black 1968: 163–4, 170–2; also 57.)
One advantage of the ordinal form, in addition to the one already mentioned in my text, is that the cardinal form—unlike the complex ordinal form introduced in the previous paragraph—will give an unwanted result where, among three options, one of the pairwise majority votes is exceptionally large relative to the other two. (See Black 1968: 168–70.)
Of course, one of the most interesting issues in the Condorcet–Black theory is to apply the ordinal analysis to the case where different majorities within the (p.371) same voting population can be established for three different options in the following cyclical pattern: A > B, B > C, and C > A. (See Black 1968: 171–3.) I will return to this particular matter in the next chapter.
(16.) By majority decision I mean (here and at further points) a decision in accordance with the complex ordinal form specified in n. 15: p(s) > p(c) and thus (1) in a contest between two options the decision should go to the majority and (2) in a contest among three or more options, that option which was supported by a majority over all the others (if there was any such option) should be selected.
(18.) For the language of compatible/incompatible interests I am indebted to Robert Hall (who calls them “consistent/inconsistent”). See Robert H. Hall, “J. S. Mill's On Liberty and Freedom of Thought,” Ph.D. dissertation, University of Kansas, Lawrence, Kan., 1986, 6–7 and passim.
(19.) The notion of interchangeability, discussed in n. 5, requires, roughly, that who is doing the voting makes no difference to the outcome; all that counts is the number of first‐place votes, with the decision going to the majority of these.
Children are excluded from the vote, not because they have no interests, but because they lack the relevant capacity for judgment. For it is true, presumably, of children—below the age, say, of 18—that in their case (and for the most part) p(c) < 1/2.
(20.) My account of the role of uncertainty in the abstract model has many points in common with the discussion of “constitutional choice,” as it is called there, in Buchanan and Tullock (1965: 77–80). My account differs from theirs, however, at other points. They stress that constitutional choice requires unanimity (Buchanan and Tullock 1965: 110–11, 188, 251–2, 256); mine requires, as we shall see, only a more or less universal subjective consensus. They give no special role to majority rule in everyday legislative or “operational” choice; indeed, they make a number of dismissive comments about such rule (see, e.g., Buchanan and Tullock 1965: 81–2, 89–90 [esp. 89], 92, 95–6, 186, 188, 194, 201, 302, 311, 339); mine, of course, gives majority rule an essential place. I should add, however, that materials are present in their argument for giving majority rule a more central role in operational choice than they do (see, e.g., Buchanan and Tullock 1965, ch. 6, pp. 70ff. [esp. 83] and pp. 126–8, 166–7).
(21.) Rawls's own summary of his “original position” model is given in Rawls (1971: 146–7; see also 126–7, 137). The metaphorical phrase “veil of ignorance” is, of course, his; see Rawls (1971: 12, 19, 169, and esp. sect. 24, pp. 136–42). For further discussion, see my book Rawls and Rights, ch. 1, sect. 2, and Ch. 13 of the present book.
((a)) First, the basic liberties as given by a list, for example: freedom of thought and liberty of conscience . . . ;
((b)) Second, freedom of movement and choice of occupation against a background of diverse opportunities;
((c)) Third, powers and prerogatives of office and positions of responsibility, particularly those in the main political and economic institutions;
((d)) Fourth, income and wealth;
((e)) Finally, the social bases of self‐respect.
One obvious additional difference between Rawls's account and mine here is that I generate the abstract model (analogous to Rawls's original position) out of a generalized account of democratic institutions; whereas Rawls does the very opposite. He generates the various political practices which collectively add up, roughly, to those included under the heading of democratic institutions out of features of the original position (see Rawls 1971, sects. 36, 37, 54, all cited earlier).
In his most recent writings (1985, 1987, 1988, 1989a), Rawls stresses that his primary concern now is to develop, not a moral theory, but a “political conception” of justice. Among the features of this political conception is the crucial claim that all the background elements (the original position, and so on) that ground the justification of the substantive principles of justice are themselves latent or implicit in the practices and attitudes of a contemporary democratic society. (For this last point see, e.g., Rawls 1985: 231 n. 14; 1987: 6, 8; 1988: 252; 1989a: 240. For general discussion see the last chapter of the present book.)
It follows, then, in this new political conception of justice, that the original position model can no longer be used to generate the so‐called democratic institutions or to justify them; for the original position model necessarily presupposes these very institutions in the work of constructing the political conception of justice in the first place. This same problem does not, however, arise for my account. For there democratic institutions continue to be justified on independent grounds (as afforded by the resources of the Black–Condorcet theory); they are brought in to justify a particular interpretation of the rationale for democracy only after that rationale has been independently established—a point made clear earlier, I hope.
(23.) See Rawls and Rights, 16, 20.
(24.) Accessible discussions of the theory of decision making under uncertainty can be found in many of the standard introductions to or surveys of rational choice theory; see, e.g., Resnik (1987, ch. 1). Prakash Shenoy has suggested to me, as particularly useful here, the essays by North (1990) and by Shafer (1990) and, for more advanced and technical discussion of this branch of decision theory and for further bibliography, Shafer and Pearl (1990).
(25.) I concur in the spirit of Duncan Black's observation that “the individual values; the group does not; it reaches decisions through some procedure in voting” (1968: 167); see also Dummett (1984: 15), and Buchanan and Tullock (1965: 32, 35–6, 332, and app. 2 n. 13 [p. 359]).
(26.) I want to acknowledge here the help and advice of Krister Segerberg, Prakash Shenoy, and many of my colleagues in philosophy at Kansas, all of whom read and commented on an earlier draft of this chapter.