SYMPOSIUM ON EXPLANATION AND

JUSTIFICATION IN SOCIAL THEORY

Democracy and Social Choice

Jules Colemanand John Ferejohn

The problem ofjustifying democracy arises when a society sees the need
for cooperative, collective action. Collective action may be needed to solve
coordination problems, public goods problems, Prisoner's Dilemmas, and
other structural problems of human interaction or perhaps to realize
common ideals of justice in concrete political institutions.
Once the need for collective action is established, one must ask how
the goals of a cooperative endeavor are to be secured. Cooperation requires
that individuals, at least over a specified range of activity, pursue a joint
strategy. To assure voluntary compliance in a joint venture, cooperation
must be to each person's advantage. However, each person may have an
incentive to induce others to cooperate and to defect from the joint
strategy in the hope of enjoying the fruits of cooperation without incurring
the opportunity costs of compliance. This is the essence of the Prisoner's
Dilemma, and considerations of this sort suggest that if collective action
is to succeed, policies or strategies formed collectively must be coercively
enforceable.
Coercion, however, requires justification. Because solving the problem
of rational noncompliance requires that collective decisions be coercively
enforceable, the rules by which collective decisions are reached require
justification. Douglas Rae neatly puts the problem of justification that
emerges when political solutions are coercively enforceable as follows:
"Once a political community has decided which of its members are to
participate directly in the making of collective policy, an important question
remains: 'How many of them must agree before a policy is imposedon
the community?"" This is essentially the question to which the principle
of democratic rule provides an answer: by what process are collective
decisions to be made?
Answering Rae's question requires a normative framework. We could
say that a procedure for making collective decisions is justified if and

This symposium consists of the revised versions of papers and commentaries originally
given at the Weingart Conference on Explanation and Justification in Social Theory held
at the California Institute of Technology, March 25-27, 1985. It was organized with funds
from the Weingart Foundation.
1. Douglas Rae, "Decision-Rules and Individual Values in Constitutional Choice,"
AmericanPolitical Science Review 63 (March 1969): 40-56 (our emphasis).
Ethics 97 (October 1986): 6-25
? 1986 by The University of Chicago. All rights reserved. 0014-1704/87/9701-0003$01.00

6
 Colemanand Ferejohn Democracy 7
only if the procedure is one that the parties in the collective-that is,
those who will from time to time be required to act contrary to their own
preferences-would choose. Let's call this the contractarian or rational
choiceapproach to thejustification of collective-decision rules.2 Alternatively,
we could say that a collective decision-making procedure is justified if
and only if the procedure is one that promotes efficiently the goals of
the collective enterprise. Let's call this the instrumentalistor consequentialist
approach to the justification of collective decision. Finally, we could identify
a set of ideals with which any collective decision-making procedure ought
to comply. We might think of these as procedural ideals, and a process
of collective decision making would be more or less justifiable depending
on the extent to which it satisfies them. Let's call this the proceduralist
approach to the justification of collective decision.
Characterized this way, proceduralism is inadequately distinguished
from both instrumentalism and contractarianism. The instrumentalist
could be taken as holding that the correct procedural ideals are those
that identify a decision-making procedure that produces desired con-
sequences; and the contractarian can be understood as claiming that the
correct procedural ideals are those that identify a decision-making pro-
cedure that would be agreed to ex ante by members of the collective.
Indeed, any theory of the justification of a decision-making procedure
could be construed in proceduralist terms.
What, then, distinguishes proceduralism from other modes of jus-
tification? Proceduralism holds that whatjustifies a decision-making pro-
cedure is strictly a necessaryproperty of the procedure-one entailed by
the definition of the procedure alone. In other words, the justification
of a procedure depends on the nature of the mapping from the original
preference orderings of the collective to the outcome. This characterization
distinguishes the proceduralist from the instrumentalist and the con-
tractarian. For both instrumentalists and contractarians, the justification
of a procedure is logically independent of its definition. In one case,
justification depends on the consequences that will arise from following
the rule while in the other it depends on how these consequences would
be evaluated ex ante-two contingent properties of decision-making pro-
cedures.
The modern theory of social choice contains a number of attempts
to develop a defense of particular voting or collective-decision procedures
by appeal to axioms aimed at characterizing one or another aspect of
procedural fairness. Social choice theorists are, in general, committed to
the view that the appropriate project for democratic theory is to develop
and justify the constraints that can serve to represent our fundamental
notions of fair or legitimate procedures. Social choice theory attempts
this essentially proceduralist project through axiomatic constructions.
2. For a discussion of the distinctions between rational choice and autonomy-based
characterization of contractarianism, see J. L. Coleman, "Libertarian and Rational Choice
Contractarianism: Consensus and the Market" (Yale Law School Library, 1985, typescript).
8 Ethics October1986
Perhaps the best attempt to examine proceduralist justifications of
democratic institutions drawn from the theory of social choice was pro-
pounded by William Riker in his recent book, Liberalismagainst Populism.3
What makes Riker'sbook so important is that, among social choice theorists,
he is the first to attempt self-consciously to explore the implications of
social choice theory for fundamental problems in political theory, in
particular, the justification of democracy.
Riker argues that the project of social choice theory is hopeless.
Roughly the argument is this: every voting system that satisfies the re-
quirements of minimal procedural fairness necessary for democratic choice
produces arbitrary or meaningless social choices or outcomes, outcomes
that are essentially unrelated to the preferences of the members of society.
Because the outcomes are arbitrary in this sense, they can be given no
justification other than the purely procedural one that points to the
institutional processes that produced them. But those institutions are
themselves inadequate to preclude arbitrary outcomes, and because they
are, procedural modes of justification, at least so far as they can be
rendered axiomatically, are themselves inadequate. Riker argues, then,
not only that attempts to justify decision rules on procedural or axiomatic
grounds fail, but also that proceduralism is itself an inadequate basis for
evaluating collective decision-making institutions.
Having abandoned proceduralism, we are led, therefore, to employ
either contractarian or instrumentalist justifications for democratic in-
stitutions. By moving to either a contractarian or instrumentalist approach,
the choice of a particular family of democratic institutions can bejustified,
and another rejected. According to Riker, the uniquely defensible insti-
tutions are those associated with constitutionally limited representative
democracy, what he calls liberal democracy; those he rejects are majoritarian
or direct democracy unfettered by constitutional guarantees of regular
elections and individual rights, what he thinks of as populist democracy.
Crucial to Riker's argument are the claims that all voting procedures
produce arbitrary outcomes and the implicit contention that constitutional
limitations restrict the range of arbitrariness of outcomes in a way that
populist institutions do not. Neither of these claims can be sustained,
however. On the first point, a number of recent results in social choice
theory suggest that fair procedures are likely to produce constrained
outcomes-while there may be some ambiguity, its range is small. If this
is so, the rejection of populism in favor of liberal democracy on the
grounds that populist institutions produce arbitrary outcomes may be
premature. On the second point, there are as yet no established theoretical
grounds for the belief that choices made by liberal institutions are any
less arbitrary than those produced by populist ones. In short, it is unclear
that constitutional limitations on voting eliminate arbitrariness in social

3. William Riker, LiberalismagainstPopulism:A Confrontation

betweenthe Theoryof Democracy
and the Theoryof Social Choice (San Francisco: W. H. Freeman & Co., 1982).
 Colemanand Ferejohn Democracy 9
choices, and it is also unclear that the range of arbitrariness expected
under populist democratic institutions warrants their rejection. Consti-
tutional limitations aimed exclusively at reducing the scope of arbitrariness
of outcomes may be unnecessary, unsuccessful, or both. At the very least,
we have no satisfactory theory establishing that they are necessary or
adequate. Lacking such comparative institutional knowledge, we think
that Riker's endorsement of liberal institutions is premature. Without a
theory of the performance of liberal institutions, we do find unconvincing
the claim that rational individuals would impose constitutional limitations
on democratic procedures. Thus, the move from proceduralist to con-
tractarian modes of justification may not warrant the endorsement of
liberalism and the rejection of populism.
This is not to say that there are no good reasons for preferring
liberal to populist institutions. Rather, such reasons would rest on con-
siderations of practicality, such as the cost of monitoring officials and
the size of a polity, not on the formal properties of choice mechanisms.
PROCEDURAL CONSTRAINTS
Social choice theory structures collective-decision problems by imposing
a set of constraints on collective-decision procedures. The question in
each case is whether there exists a procedure or mechanism that satisfies
those constraints. Social choice theorists tend, therefore, to be proceduralists
about the justification of democracy. In social choice theory, democracy
itself is analyzed as a subset of institutions for aggregating individual
preferences into social choices; that is, democracy is a kind or set of
voting rules. In aggregation problems, social choice theory treats individual
preferences as basic and as given. Voting, in this view, expresses "tastes"
or preferences; it does not consist in the formation of a collectivejudgment.
In Riker's view, what distinguishes democratic voting procedures from
alternatives is their commitment to the ideals of equality, liberty, and
participation. Commitment to these ideals gives rise to more particular
constraints on collective-decision rules. Satisfaction of the derivative con-
straints justifies a collective-decision procedure because the constraints
are deducible from or otherwise justificationally connected to the more
fundamental commitments to equality, liberty, and participation.
What do participation, liberty, and equality mean for the social choice
theorist, and how do these ideals give rise to concrete procedural constraints
on voting? "The crucial attribute of democracy," Riker writes, "is popular
participation in government."4 Participation, in turn, is analyzed in terms
of voting. "Although the institutions of participation have been many
and varied, they have always revolved around the simple act of voting."5
Democracy requires participation. Participation entails voting. But, "voting
is not equivalent to democracy, only voting that facilitates popular choice

4. Ibid., p. 5.
5. Ibid.
10 Ethics October1986
is democratic.... Or, voting is a necessary, but not a sufficient, condition
of democracy. To render them equivalent, voting must be surrounded
with numerous institutions like political parties and free speech, which
organize voting into genuine choice."6
Of the connection between liberty and collective decision, Riker writes,
"The historic purpose of ... fundamental democratic liberties has been
not to provide freedom as an end in itself, but to render effective both
political participation and the process oft choice in voting."7 In the same
connection, citing Rousseau, he writes that "liberty resides in participation
in government, not in rights distinct from government."8
Equality, Riker writes, "originated in some rough sense as an in-
strument of voting. Voting would not mean much if each person's vote
were not counted in the same way.... Its logical base lies in the instrumental
value of making voting work."9 According to Riker, democratic rules are
the only collective-decision rules that satisfy these three abstract criteria,
satisfaction of which is both necessary and sufficient to justify them. "In
a society characterized by democratic justice, people are free (by reason
of democratic liberty) and have the chance (by reason of democratic
equality) to seek self-respect and self-control (through some kind of dem-
ocratic participation). The democratic method that is supposed to achieve
this idea . . . is the process of participation, specifically through voting,
in the management of society, where voting is understood to include all
the ancillary institutions (like parties and pressure groups) and social
principles (like freedom and equality) that are necessary to render it
significant. 10
Riker argues that the essence of the commitment to democratic
choice is captured by its commitment to the abstract ideals of participation,
liberty, and equality. The problem is to work from these abstract com-
mitments to specific constraints on possible social choice rules necessary
and sufficient to justify them. These more particular constraints appear
to be of two distinct sorts: normative and semantic.
On the one hand, justified social choice (or voting) procedures must
be fair, provide for full and equal participation, and allow for autonomy
in the Rousseauean/Kantian sense, according to which an autonomous
agent complies with dictates or constraints of his own choosing. On the
other hand, autonomy in this sense presupposes full participation because
full participation is necessary to guarantee that the voting rule produces
outcomes that express the collective or general will. This line of analysis
brings us to the cognitive or meaningfulness criterion. In order to be
meaningful, the social choice must be unambiguously interpretable in
terms of individual wills, preferences, or voter profiles.

6. Ibid.
7. Ibid., p. 7 (our emphasis).
8. Ibid.
9. Ibid.
10. Ibid., p. 8.
 Colemanand Ferejohn Democracy 11
Only democratic choice procedures are capable, even in principle,
of satisfying both conditions. The problem social choice theory raises for
democratic theory is that any democratic voting procedure that is fair in
the appropriate sense will be normatively defensible but not meaningful,
that is, its outcomes will be arbitrary. Only voting that is meaningful and
fair can be justified. Unfortunately, no voting procedure can be both.
THE PROBLEM OF AMBIGUOUS INTERPRETATION
AND PARADOXICAL OUTCOMES
We can distinguish between two defects in voting rules, either of which
would be sufficient to render voting meaningless. If a voting rule could
generate a paradoxical ordering of preferences, it is said to be unstable.
Generally, we refer to theorems that establish paradoxical orderings or
cycles as instability, disequilibria, or chaos results. The voting paradox,"I
Arrow's Theorem'2 and McKelvey/Schofield results'3 are instability results.
However, an outcome which, in social choice theory, is itself a complete
ordering of preferences need not be paradoxical to be deeply troubling.
An outcome is also troubling if it is not unique. An outcome, OR, is not
unique if, given a profile of preferences, D, another outcome, Os, might
have obtained had fair aggregation rule S rather than fair rule R been
employed. In such a case, OR is said to be ambiguous.Ambiguity concerns
us only if there are no independent grounds for preferring R to S when
aggregating preferences in D. We refer to defects of the first sort as
instabilities and to those of the second sort as ambiguities. The bulk of
Riker's argument attempts to demonstrate the meaninglessnessof voting
by establishing generalized instability or ambiguity results.
In addition to the distinction between instabilities and ambiguities,
we can distinguish between different kinds of implications we might draw
from either. Riker employs the instability and ambiguity results to dem-
onstrate the meaninglessness of voting. One could just as well take these
results as bearing on the normative defensibility of voting rules because
both ambiguity and instability create opportunities for differential influence
over outcomes that is incompatible with the democratic ideal of equality.
Thus, if different outcomes result from different rules, then access to
the agenda increases-at least potentially-one's influence over outcomes.
We shall return to these problems in due course. For now, we want to
focus on Riker's claim that instability and ambiguity demonstrate the
meaninglessness of voting. We first address the ambiguity argument.

11. Marquis de Condorcet, Essai sur l'applicationde l'analysea la probabilitydes decisions

rendues a la plurality des voix (Paris, 1785).
12. Kenneth Arrow, Social Choice and Individual Values (New Haven, Conn.: Yale
University Press, 1963).
13. Richard McKelvey, "Intransitivities in Multidimensional Voting Models and Some
Implications for Agenda Control," Journal of Economic Theory 12 (June 1976): 472-82;
Norman Schofield, "Instability of Simple Dynamics Games," Review of EconomicStudies 45
(October 1978): 575-94.
12 Ethics October1986
TABLE 1
IMAGINARY ORDINAL AND CARDINAL UTILITIES FOR
FIVE ALTERNATIVES AND FIVE PERSONS

Rank Order Voter 1 Voter 2 Voter 3 Voter 4 Voter 5

(highest)
1 ....................... a (1.00). d (1.00) e (1.00) b (1.00) b (1.00)
2 ....................... d (.90) a (.61) c (.80) d (.90) e (.96)
3 ....................... b (.60) b (.60) a (.70) a (.75) c (.70)
4 ....................... c (.55) e (.59) b (.55) e (.74) a (.60)
5 ....................... e (.50) c (.50) d (.50) c (.50) d (.50)
NOTE.-Cardinal utilities are in parentheses.
SOURCE.-Reprinted from William Riker, Liberalismagainst Paternalism (San Francisco: W. H. Freeman & Co., 1982),
p. 37.

According to the ambiguity argument, the result of applying rule

R, namely, OR, is ambiguous if a different outcome, Os, might have
obtained had we applied rule S. Because OR is ambiguous, we do not
know what meaning to assign to it-what to make of it as a collective
choice. For example, given a profile of preferences among voters, four
methods of amalgamation or summation-the Borda count,14 cardinal
utility comparison,'5 multiplicative cardinal utility comparison,'6 and or-
dinal pairwise comparisons can result in different social choices. Each
voting procedure satisfies a general, minimal conception of fairness, and
there is no theoretical basis either to assert that only one of these rules
is capable of accurately revealing public sentiment or deny that any one
of them can. Riker argues for this claim by examining the example in
table 1.18
It turns out that, among the five alternatives in table 1, a is the
Condorcet or pairwise winner, b is the plurality winner, d is the ordinal
utility winner, and e is the multiplicative cardinal or Nash winner. All
four methods are plausible choice rules.
The same profiles of preferences can yield several quite distinct
"winners" depending upon the aggregation procedure employed. So

14. In a Borda count, each individual voter assigns points to alternatives; for example,
five points to the first choice, four to the second, and so on. The points for each alternative
are summed and the winner is the alternative with the highest total of points.
15. Under the "Bentham rule," each person assigns cardinal values to his choices;
e.g., ten for the first choice, but only five for the second if he prefers the first alternative
twice as much as the second. Then the total for each alternative is found by adding up
the points each individual has assigned to it. The winner is the alternative with the highest
total. In Borda counting, each voter must use the same scaling, i.e., award the same number
of points to his first, second, and third choices. Different voters, under the Bentham rule,
can assign different values to similarly ordered choices.
16. In "Nashian" voting, the cardinal utilities assigned by each voter are multiplied
rather than summed. Again the winner is the alternative with the highest total.
17. In Condorcet voting, there is a winner if and only if it defeats all alternatives when
paired against them individually.
18. Riker, p. 37.
 Colemanand Ferejohn Democracy 13
even if outcome a in Riker's example wins in pairwise comparisons against
each alternative, there is no reason to interpret a as the choice embedded
in the preferences of the voters. Had another choice procedure been
employed, another outcome would have been the winner. The question
is, What interpretation can we plausibly give to this argument? By his
own account, Riker has "been very much impressed with this deep am-
biguity" in the "process of aggregating preferences.'9
There are at least two ways to interpret the argument. One inter-
pretation is consistent with the concerns voiced by Riker, the other is
not. In one sense, that a social choice is a function both of the profiles
of preferences and the aggregation rule does not imply that the choice
has ambiguous meaning or significance. The meaning of a social choice
can be reconstructed in terms of the interplay of the data base and the
decision rule. That other winners may have emerged under different
decision rules is simply not relevant either to interpreting the outcome
or to assessing its merits. Consider the problem of aggregation in another
context. In American football, teams are awarded six points for a touch-
down, three for a field goal, one for an extra point, and two for a safety.
The outcomes of football games depend on these assignments of points
for events. A different assignment of points for events, or a different
way of aggregating them, might well lead to different outcomes: some
wins would become losses and vice versa. (Different scoring rules would,
of course, lead to changes in strategy,but changes in aggregation procedures,
if known, would also affect voting strategies.) Even though we could
score football games differently-according to different rules-it does
not follow that the outcomes of football games are ambiguous or that
the score does not reveal what occurred in the game. Just showing that
different outcomes result from applying different procedures, all of which
are minimally fair or plausible on other grounds, is not enough to establish
a troubling result for democracy.
There is, however, a different sense in which the fact that, given
the same profiles of preferences, different outcomes result from different
social choice rules can be problematic. The problem is not that the outcomes
lack meaning or significance. Instead, the problem is that it is impossible
unambiguously to interpret an outcome as uniquelyrepresentingthepopular
will. We can make sense of this interpretation of the argument only if
we assume that in every set of voters' preference profiles there "exists"
a unique popular "will" defined and revealed by voting, and that voting
is not meaningful or significant unless it uniquely identifies this will. In
this view of collective decision, voting is important because it specifies
or defines the popular will, and the ambiguity problem is troubling because,
although voting is supposed to reveal a unique general or popular "will,"
the "will" that is revealed by a social choice is as much a function of the
method of choice as of the initial profile from which the choice emerges.

19. Ibid., p. 36.

14 Ethics October1986
Thus, one can never know that the outcome of an election reflects the
popular or collective will because, in general, one always has good reason
to believe that a different outcome would have been produced by applying
a different, equally plausible aggregation procedure. But if a voting
procedure, in particular a democratic one, is defensible only if its results
meaningfully reflect a unique. public disposition or will, then it is natural
to be skeptical about the legitimacy of any voting procedure.
The revised ambiguity argument is as follows: voting is meaningless
if its outcome is ambiguous; the outcome of voting is ambiguous if it
does not uniquely reveal the popular or general will; because different
rules produce different outcomes, voting cannot unambiguously define
the popular will. We note that this argument relies upon the claim that
democratic theory is committed to (1) the existence of a unique popular
will in all electoral matters and (2) the constraint that that will is defined
by voting. Are these conditions plausible constraints on democratic theory?
For Riker they appear to be. Recall his claim that voting is democratic
only if it expresses popular will or sentiment. Democracy grounds the
requirement of full participation, the constraint of fairness, and the pro-
tection of speech, association, and the like. Democratic institutions, the
constraints they impose, and the freedoms they create are alleged to
organize choice meaningfully, that is, to insure that a social or public
choice reflects public sentiment. If Riker is right about this characterization
of democratic political institutions and particularlyabout the role constraints
on voting procedures play in the theory of democracy, then the ambiguity
problem is serious. For, as he sees it, democraticvoting must express a
unique public sentiment or preference. But the fact that different voting
rules express different sentiments, given the exact same data base, means
that no voting rule can uniquely determine the public will.
Riker appears to treat these two constraints-that in any election
there exists a unique popular will and that that will is defined or specified
by voting-as part of the meaning of democracy. Otherwise, the argument
from ambiguity would fail to establish the meaninglessness of democratic
voting. Since only democratic voting rules are adequately fair, their
meaninglessness suffices to shed doubt on all collective-decision procedures.
At this juncture in the argument, Riker's distinction between two
conceptions of democracy becomes relevant. In the end, Riker does not
claim that all forms of democratic voting are indefensible. What he pre-
viously treated as commitments of all democratic theories, he now claims
are commitments of only one sort of democratic theory: democratic or
constitutional populism. Populist democracy is then distinguished from
constitutional liberalism or liberal democracy. According to populism,
each election reveals a popular or general will. The point of elections is
to define and uncover the general will so that, by complying with laws,
individuals act in accordance with their own will. The liberal sees elections
in an entirely different way. For the liberal democrat, for example, James
 Colemanand Ferejohn Democracy 15
Madison, the point of elections is the possibility of removing officeholders.20
The claim Riker ultimately advances is that the results of social choice
theory undermine populism, but not liberalism. Liberal democracy alone
withstands the challenge of social choice theory.
POPULIST DEFENSES
Several avenues of defense are open to a populist who seeks to meet the
challenge of social choice theory as it is construed by Riker.
First, a populist can deny that a general will must exist in all electoral
matters. Following Rousseau, a populist might accept the existence of a
general will over a certain domain of electoral issues (such as those con-
cerned with the provision of public goods) while denying that there is a
general will in all matters that may be brought up for a vote. Next, over
the domain of issues in which it is alleged that a general will exists, a
populist can deny being committed to the claim that the general will is
unique. Rather, just as welfare economists need not subscribe to the idea
of unique efficient outcomes, the general will may only specify a set of
acceptable outcomes. Populists may, therefore, be committed to a general
will but they need not assert its uniqueness or its existence in all electoral
matters. Against this defense, Riker's criticisms of populism are telling
only if he can show that the range of ambiguity is so large that the
general will is essentially vacuous.
As Riker implies, a number of recent results in the theory of collective
choice have been construed as ruling against this defense.21 Under quite
general conditions, the range of instability of any voting rule is large
enough to encompass all the feasible alternatives. A voting procedure is
liable to pick any alternative and so the general will must be entirely
unconfining.
Whether the general instability results preclude a populist from
pursuing this line of defense depends on the proper interpretation of
those results, and that is a matter of some dispute. Our view is that the
implications of the instability results are somewhat overdrawn. While
voting procedures are able to produce any feasible alternative as an
outcome, they may be very unlikely to do so.22
Second, a populist might argue that in those electoral matters in
which there exists a general will, it is not specified by voting. This line
of defense relies upon specifying the relationship between voting and
the general will in a nonsemanticor noncriterialway. One alternative is to
view the relationship between voting and the general will as epistemic.

20. James Madison, The FederalistPapers, no. 52 (New York: Mentor Books, 1961), p.
327.
21. Riker, pp. 181-200.
22. For a further discussion, see the section entitled Liberal Democracy and nn. 28-
32 below.
16 Ethics October1986
(Again, this appears to have been Rousseau's view.) Moreover, such a
view requires treating voting over issues involving a general will as consisting
in judgments-which can be either true or false-rather than in expressions
of preferences-which are neither. Like an advocate of the first line of
defense, the epistemic populist need not hold that there is a uniquely
good or most reliable voting procedure for all electoral matters, nor need
he hold that there exists a general will in all matters. Rousseau, for one,
appears to have held that the range ofissues in which there is a general
will is limited; that the relationship between voting and the general will
over those matters is epistemic; and that no particular voting rule is
optimal in all cases.
The epistemic populist begins, then, by distinguishing between those
electoral issues in which there exists a general will and those in which
there is no reason to believe that there is a general will or disposition.
For all those electoral issues in which no general will exists, the ambiguity
argument of social choice theory poses no special problems for a populist.
Where no general will exists, electoral outcomes can be meaningful even
if they are not unique.
We therefore restrict the remainder of this discussion to those cases
in which the populist is committed to the existence of a general will.
Over these issues there is a correct judgment about what a collective
ought to do. By voting, individuals express judgments about what ought
to be done. One's judgment about what ought to be done by way of
providing public goods need not coincide with one's own preference; in
any case, one's vote is not intended to express one's preference. Voting
brings together judgments. Different voting rules amalgamate those
judgments in different ways. Voting rules are then to be justified in terms
of their capacity to provide evidence of the general will. And, as in other
contexts, different rules will be more or less well suited to this task.
The epistemic populist denies that there is a necessary connection
between the results of voting and the general will. He holds that the
general will is not entailed by the result of a vote, and so he escapes the
implications of social choice theory. However, he accepts, along with the
criterial populist, that the results of a vote can give evidence of what the
general will is. This presents the epistemic populist with a problem not
faced by the criterial populist. The problem is explaining why the results
of voting provide evidence of the general will. This explanation is simple
for the criterial populist. Voting provides evidence of the general will
because voting defines the general will. No such explanation of the ev-
idential connection between voting and the general will is open to the
epistemic populist. What sort of account of the evidential relationship
between voting and the general will can a populist offer?
One explanation, but by no means the only one, open to the epistemic
populist is to argue that voting provides evidence of the general will if
there exists a reliabilityconnection between voting and the general will.
The desirability of a voting rule will then depend on its reliability-the
 Colemanand Ferejohn Democracy 17
extent to which the collective judgments it generates converge with what
is in fact the correct judgment. The key features of a voting rule would
then relate to its capacity to discriminate among various levels of voter
competence in making judgments and in its capacity to encourage voter
competence in forming the correct judgment.
One might be tempted to object to the reliabilist analysis of the
evidentiary property of voting as follows: in any epistemic account, the
general will transcends voting. There is a correct answer in regard to
the issue at stake whether or not a vote is taken. The voting procedure
is a device to help citizens discover what the correct thing to do is.
However, if there exists an independently specifiable general will, then
either we can uncover what it is on a particular issue or we cannot. If
we cannot determine what the general will is, then voting, understood
epistemically, is uninteresting. If, however, we can discover the general
will, then we can do so by voting only, in some other way, or both by
voting and in some other way as well. If, on the one hand, we can identify
what is in the general interest over a domain of issues without regard to
voting, then it cannot be part of the justification of voting that it is
necessary to reveal the general will. On the other hand, if the general
will can be revealed only by voting, then we appear to lack an independent
criterion for assessing the reliability of various voting mechanisms. The
problem, for an epistemic interpretation of voting, is that we need either
an independent means for identifying the general will-in which case
voting, for this purpose, is otiose-or independent criteria with which
to assess the reliabilityof various voting procedures which do not themselves
rely upon the existence of an alternative mechanism for identifying the
general will.
These remarks contain two different objections to this particular
form of epistemic populism. Both rely on the following claim. The reliability
of voting rules depends on our having independent access to the nature
of the general will. Either we have such access or we do not. If we do,
then voting is otiose; if we do not, then we cannot determine its reliability.
Both objections can be met. First, it need be no part of an epistemic
theory of voting that voting should be the only way of gaining insight
into the nature of the general will. What recommends voting is neither
its epistemic uniqueness nor its privileged status. It may lack both features.
Instead, voting may give evidence of the general will in a particular way-
by public forum, discussion, and participation. And, in matters concerning
the coercive authority of the state, these features of voting are not in-
significant. Second, while it is true that, if we have no independent access
to the nature of the general will, then we may be unable to determine
whether or not a voting rule is in fact reliable, the epistemic populist is
trying to specify the conditions under which a voting rule could be justified,
not the conditions under which we would know that it was. The voting
reliabilist would deny that in order for a voting rule to be justified we
must also know that it is a reliable guide to the nature of the general
18 Ethics October1986
will. It simply must be reliable, independent of our knowledge of its
reliability.
Nonuniqueness and epistemic populism are not the only ways in
which a populist might attempt to meet the challenge of social choice
theory. A populist might seek to avoid the implications of the instability
and ambiguity results by arguing that there exists a uniquely defensible
voting rule. In this way, even if the outcome of an election is thought
to define the general will, and not merely to relate to it evidentiarily,
problems of ambiguity will not arise because only one rule is admissible.
Perhaps the best-known such attempt to meet the challenge of social
choice theory along choice theoretic lines is Kenneth May's characterization
of majority rule with two alternatives.23 If the implications of May's
theorem are accepted, admissible choice on larger sets of alternatives
would be made by voting rules that are extensions of binary majority
rule.
Leaving aside for the moment what would constitute an extension
of simple majority rule, we must ask if May's defense of his axioms are
in fact sufficient to preclude all alternatives to simple majority rule. In
our view, May's formal statement of his axioms is stronger than the
defense that may be offered for them. One example is sufficient to
illustrate this and to undercut this line of argument.
In the presence of May's other axioms we can restrict our argument
to the case of special majority rules. Within the class of special majorities,
only simple majority satisfies May's strong monotonicity condition;24 yet
special majorities, including unanimity, are weakly monotonic. If we drop
the strong monotonicity condition in favor of weak monotonicity, the
uniqueness result evaporates. Maintaining strong monotonicity as a con-
dition on voting rules requires an argument that only it expresses some
more fundamental demand of collective decision. The plausible candidate
for this deeper demand would be that only a strongly monotonic function
could specify a general will. But as we have seen, a populist wants a
choice rule to do two things: (1) rule out perversity and (2) specify a
general will. Weak monotonicity suffices to rule out perversity. Special
majorities are weakly monotonic and, moreover, they are more likely
than simple majorities to constitute what we would normally think of as
expressions of popular or general sentiment. The strong monotonicity
requirement is inadequately motivated; weakening it, however, precludes
uniqueness.25

23. Kenneth 0. May, "A Set of Necessary and Sufficient Conditions for Simple Majority
Decision," Econometrica20 (October 1952): 680-84.
24. A rule is monotone whenever, in moving from one preference configuration to
another, if one alternative becomes more attractive relative to the other alternatives, for
some individual, it does not move down in the social judgment. Strong monotonicity
requires, in addition, that, if two alternatives are tied in the social decision and one alternative
"moves up" relative to the others, the tie is broken in favor of that alternative.
25. Roughly, the same remarks are in order in regard to the anonymity condition.
While anonymity may seem, on its face, to express the requirement that voters be treated
 Colemanand Ferejohn Democracy 19
Something like the uniqueness argument is required by a populist
who is committed to a criterial conception of the relationship between
voting and the general will. Unfortunately, it appears that no such argument
can be sustained.
If May were right, simple majority voting would be the preferred
social choice procedure in binary choice situations. As Riker notes, however,
there are almost always more than two natural alternatives. Simple majority
over more than two alternatives is unstable. Consequently, to avoid insta-
bility, the alternatives must be reduced to two. But every method of
performing this reduction is subject to strategic manipulation and this
fact reintroduces ambiguity into the process. In this sense, then, the
uniqueness of simple majority voting is either mythical or useless.
The responses we have thus far offered each assumes that social
choice theory presents a serious problem for populism. The problem is
this: voting is meaningless. If the outcome of a vote is meaningless, how
can it be thought of as defining anything, let alone the general will?
Populism, even in its modest forms, is committed to a general will at
least over some electoral issues. Populism is committed to nonsense, or
worse, is itself nonsense.
Taking the results of social choice theory at face value, as deeply
problematic for populism, we have suggested how populists might yet
defend their position by advancing what we call epistemic populism. We
have also suggested that a criterial populist might deny the uniqueness
of the general will. In short, populism might be defended by accepting
that the relationship between voting and the general will is definitional,
denying uniqueness, or by denying that the relationship between voting
and the general will is definitional.
We have not fully considered two other related alternatives. The
first is that perhaps the results of social choice theory are as problematic
for liberalism as they are for populism. The second is that the implications
of the social choice literature are subject to dispute and, in particular,
that neither liberal nor populist political institutions behave as poorly as
the instability theories would have us believe. We turn now to considering
each of these alternatives.
LIBERAL DEMOCRACY
We begin by summarizing Riker'sargument for liberalism. Only democratic
institutions can satisfy minimal standards of fairness. But democratic
institutions produce arbitrary social choices. A radical skepticism about
voting is therefore warranted. Three things appear to follow from this
skepticism about voting. First, no purely proceduralist defense of democracy

equally, it is, in fact, much stronger than this. By requiring that every pair of voters be
interchangeable, anonymity rules out a variety of procedures that accord everyone the
same influence over the collective choice but which are not anonymous. An example of
such a rule would be a "representative" procedure in which, for a proposal to prevail, it
must receive a majority vote in each of the majority of districts.
20 Ethics October1986
can be sustained, for whatever the virtues of participation or procedural
fairness might be, one cannot defend the morality of what is either
paradoxical or otherwise unintelligible. It hardly counts in favor of a
procedure that participation is fair, full, and equal if the outcomes of
every application of the procedure turn out to be unintelligible. Second,
the argument for autonomy and democracy is undercut since democratic
voting procedures do not necessarily yield expressions of the general
will; so, complying with measures that result from democratic voting
procedures is not tantamount to acting in compliance with rules of one's
own making. Thus, radical skepticism about the meaningfulness of voting
requires that democracy be defended on nonprocedural grounds having
nothing to do with the relationship between democratic collective choice
and the aims of collective policy. "All elections do or have to do is to
permit people to get rid of rulers. The people who do this do not themselves
need to have a coherent will.... The liberal interpretation of voting thus
allows elections to be useful and significant even in the presence of cycles,
manipulation and other kinds of 'errors' in voting.... The kind of de-
mocracy that thus survives is not, however, popular rule, but rather an
intermittent, sometimes random, even perverse, popular veto.... Liberal
democracy is simply the veto by which it is sometimes possible to restrain
official tyranny."26
The virtue of liberal democracy is that it promotes liberty by veto.
"Suppose freedom is simply the absence of governmental restraint on
individual action. Then the chance to engage in vetoing by rejecting
officials and the chance that the rejection actually occur are the very
essence of this freedom."27
Voting outcomes are either deeply ambiguous, unstable, or both.
That turns out not to matter, however; voting does not have to be mean-
ingful. It simply has to be effective in a well-defined way. The voting
process need only make it possible for an electorate to remove officials
from office. In doing so, democratic voting is the guardian of liberty.
Liberty is the absence of governmental constraints or coercion. Liberty
is maximized when coercion is minimized. Voting helps to reduce coercion
by increasing the costs on legislators of imposing coercive legislation,
that is, by making real the possibility of their removal from office. Voting
as the guardian of negative liberty, not as the expression of popular will,
is the essence of liberal democracy.
In fact, we can flesh out a bit more the possible connections between
liberty and democracy and identify exactly which relationships between
the two the liberal democrat, at least for Riker, is committed to. Riker
has in mind three ways in which the ideals of democracy and liberty
might connect in a liberal theory of democracy. The first concerns the
relationship between voting and those concrete liberties of, say, association

26. Riker, p. 244.

27. Ibid., p. 245.
 Colemanand Ferejohn Democracy 21
and expression that render voting possible and judgments over alternatives
informed. The next concerns autonomy, or Kantian liberty. Here, liberty
consists entirely in complying with prescriptions of one's own making.
The third sort of liberty is negative liberty, the absence of governmental
constraints. In the full theory of the state that ultimately emerges from
the social choice theory of democracy, democracy as popular will, as
autonomous agency, does not survive. What does survive is a theory of
the liberal state in which negative liberty is maximized or insured through
popular elections and where the liberties necessary for the franchise to
be meaningfully exercised are constitutionally guaranteed.
The liberal democrat, it appears, is committed to a skeletal version
of classical liberal constitutionalism. A constitution exists that specifies
not only the terms of collective decision making but guarantees a set of
fundamental political liberties. These concrete, political liberties, for ex-
ample, of association and speech, derive from the need to make political
decisions informed and reasoned. Whatever rule of collective decision
making emerges in the constitutional process is necessary to reduce the
possibility of political tyranny and to maximize, thereby, negative political
liberty. The choice of voting rules is then to be made exclusively in the
light of this goal. For these reasons, liberal democracy, so conceived,
survives the challenge of social choice theory. But does it?
Let's begin by noting that this defense of liberal democracy is based
not on a proceduralist criterion of assessment but on both instrumental
and contractarian ones. There is no claim made that liberal democratic
collective-decision procedures are justified because they satisfy a certain
set of ideals or constraints. Quite the contrary. Liberal democracy appears
to emerge precisely because no voting procedure-at least in Riker's
view-satisfies any plausible set of such constraints. Instead, liberal dem-
ocratic institutions have the desired effect of allowing individuals to remove
officeholders, and this maximizes (negative) liberty. That is their instru-
mental value. Moreover, each potential voter, thinking through the choice
of collective-decision rules ex ante, would prefer liberal voting institutions
precisely because they have this property.
But do the ambiguity and instability results of social choice theory
distinguish between populist and liberal democracy sufficiently to render
the rejection of populism and the defense of liberalism plausible? Consider
the ambiguity arguments first. To the extent that the outcomes of social
choice are ambiguous, they render liberal democracy just as incoherent
as populist democracy. If it is impossible to interpret the outcomes of
elections in a reasoned way, officials removed from office under election
system P might not have been removed under process Q and so on. How
then can we interpret the removal of officials as an expression of dis-
satisfaction with their performance? How can we expect officials to take
account of such ambiguous signals in deciding how to behave?
The problem with Riker's argument goes even deeper than this. If
he is right about the implications of the social choice theory, namely,
22 Ethics October1986
that all voting is meaningless, then what follows is that no set of collective
choice institutions-liberal or populist-can be defended on proceduralist
grounds. Recall that, as Riker interprets proceduralist modes ofjustification,
to be justified, a voting rule must be both fair and meaningful. If he is
right about social choice theory, no voting rules are meaningful. No
institutions, liberal or populist, are therefore justifiable on proceduralist
grounds. Justification must rest on instrumentalist or contractarian ap-
proaches. But if this is so, it is not enough to give an instrumentalist or
contractarian defense of liberal democracy; he must also show that liberal
institutions are to be preferred to populist ones on these accounts. In
other words, if it shows anything, social choice theory demonstrates that
neither liberalism nor populism may be defended on proceduralist grounds.
It leaves open the possibility that both can be defended on some other
grounds. But Riker considers nonprocedural defenses only of liberalism,
not of populism. He appears to believe that populism entails a particular
form ofjustification, namely, proceduralism, whereas liberalism also entails
a unique form of justification, namely, instrumentalism. Surely this is
wrong. When Riker says that the meaninglessness of voting is compatible
with liberalism, he should be saying that even if voting is not meaningful,
it is still possible to justify its application on other grounds-that is, its
good consequences. But then, a populist might make precisely the same
argument. Voting, though meaningless in the sense of social choice theory,
still strengthens allegiance, increases competence, develops a sense of
community, and the like. All of these are virtues of electoral systems that
populists have generally endorsed. Elections have good populist conse-
quences.
In fact, once the line of justification switches from proceduralism to
instrumentalist, it is not at all obvious that the liberal democrat, as he
is construed by Riker, does as well as Rousseau's populist. For it surely
cannot be a good consequence of voting that it randomly, that is, non-
rationally, removes officeholders. So does sudden illness. But nonreasoned
removal from office is precisely what follows if Riker is correct in inter-
preting the instability results of social choice theory as demonstrating
the meaninglessness of voting. If outcomes are arbitrarily connected to
the preferences of the electorate, we cannot infer from his removal from
office that an officeholder's conduct was in fact disapproved of by the
voters. This is hardly the ideal of officeholders being put at risk by
elections that we associate with liberalism. In short: if we take social choice
theory's results as conclusive as Riker does, then what we must give up
is proceduralism, not necessarily populism. Moreover, in moving from
proceduralism to instrumentalist, it may be that populism is more plausible
than the sort of liberalism Riker advances.
Next, in our view, Riker overstates the implications of the instability
results of social choice. The problem of ambiguity arises at two levels.
As Riker first raises it, the problem is simply that the outcome is always
relative to the collective-choice rule; had a different rule been employed,
 Colemanand Ferejohn Democracy 23
a different outcome might have been reached. Understood in this way,
ambiguity is troublesome neither for the liberal nor for the populist. As
long as there is a well-defined set of rules in place, the outcome of an
election is interpretable within the given electoral framework. That an
alternative framework could be thought of that would produce a different
outcome is simply irrelevant.
The ambiguity problem resurfaces, however, in a systemic way. Con-
sider that more than two alternatives are almost always available. The
application of any nonunanimity rule to the set of alternatives may therefore
lead to unstable outcomes. As all methods for reducing the set of alternatives
to two are subject to manipulation, there is no guarantee that the electoral
outcome will be related to the distribution of preferences. This is the
sort of systemic ambiguity that is in theory at least problematic for both
liberal and populist theories.
Putting the two ambiguity problems together suggests the following.
As long as the electoral framework is well defined, its outcomes are
interpretable and only uninterestingly ambiguous. The problem is in
deciding upon an electoral framework. The aim of any such framework
is to generate outcomes that are systematically and coherently related to
the distribution of preferences. If there are more than two feasible al-
ternatives available for choice, the nature of this connection will depend
on the institutional details of the electoral system. Rather than inferring
from the ambiguity argument that electoral outcomes are arbitrary, we
ought instead to infer that institutions matter. They matter because their
particular content and contour determine the extent to which electoral
outcomes reflect popular preferences. From this perspective, the real
intellectual need is for a workable theory of institutional behavior.
It is true, as Riker says, that if there is a large enough number of
available alternatives, any voting rule that requires less than unanimity
will produce large cycles for most configurations of popular preferences.
But, while such cycles usually exist, and all outcomes are unstable in the
sense of being liable to defeat by some alternative, this does not imply
that such cycles will be realized, or that the actual performance of the
collective-choice institutions will be chaotic in the ordinary sense of that
term. While cycles may be ubiquitous, and all outcomes unstable, some
alternatives may be much more vulnerable than others, and the less
vulnerable alternatives may turn out to be quite nicely connected to the
distribution of preferences.
In fact, two recent lines of argument have been made in support of
this claim. The first, due to Gerald Kramer28 and to John Ferejohn,
Richard McKelvey, and Edward Packel,29 shows essentially that, even in

28. Gerald Kramer, "A Dynamical Model of Political Equilibrium,"Journal of Economic

Theory 16 (1977): 310-44.
29. John Ferejohn, Richard McKelvey, and Edward Packel, "Limiting Distributions
for Continuous State Markov Voting Models," Social Choiceand Welfare 1 (1984): 45-68.
24 Ethics October1986
the presence of a global cycle, majority-rule "paths" or "trajectories" are
well behaved in the sense of exhibiting stable, limiting properties. In-
tuitively, the arguments may be stated as follows: while alternatives that
are disliked by nearly everyone may be reached in the majority-rule cycle,
there are relatively few paths leading to these alternatives from any given
alternative. Most paths lead toward alternatives that are attractive to large
numbers of people so that such paths tend to stay among relatively
attractive alternatives. For this reason, majority-rule paths converge on
relatively small sets of alternatives which are "centrally" located in the
sense that they are moderately attractive to large numbers of people.
The second argument is due to Nicholas Miller30 and is developed
further by Kenneth Shepsle and Barry Weingast3' and Richard McKelvey,32
and it shows that even when there is a global cycle, if the members of
the electorate take full account of their strategic opportunities under the
institutional rules governing consideration of alternatives, only a small
subset of alternatives may arise as outcomes of the process. Thus, while
each alternative may be defeated, only a small subset of them may arise
as equilibria. Moreover, this set is "close" to the set found in the paper
by Ferejohn, McKelvey, and Packel. While neither of these lines of ar-
gument has been formulated in as general a setup as have the instability
results, it seems likely that similar properties will be found in general
models.
From the present point of view, the important thing about these
results is that they apply to a wide variety of implementations of voting
rules, liberal as well as populist. Thus, on the basis of this work, there
seems to be no reason to separate the likely performance of liberal and
populist institutions. If the one is unstable, so is the other, and the extent
of instability seems likely to be similar. The instability results may be
inadequate to preclude either liberal or populist institutions, or for that
matter, proceduralist attempts to justify either.

CONCLUSION
Social choice theory is replete with theorems demonstrating the instability
and systemic ambiguity of voting. Riker argues from these results to the
incoherence of populist democracy and to the defensibility of liberal
democracy. But the instability and ambiguity results do not distinguish
between liberal and populist institutions. Whether these results are com-

30. Nicholas Miller, "A New Solution Set for Tournaments and Majority Voting,"
AmericanJournal of Political Science 24 (1980): 68-96.
31. Kenneth Shepsle and Barry Weingast, "Uncovered Sets and Sophisticated Voting
Outcomes with Implications for Agenda Institutions," AmericanJournal of Political Science
28 (1984): 49-74.
32. Richard McKelvey, "Covering Dominance, and Institution-free Properties of Social
Choice," Social Science Writing Paper no. 494 (Pasadena: California Institute of Technology,
1983).
 Colemanand Ferejohn Democracy 25
patible with the defensibility of both or establish the incoherence of both
remains the central question. Though we are concerned by the implications
of instability theorems, we think it premature to see these results as
establishing the arbitrariness of collective decision making. Rather, these
results demonstrate the importance of gaining a fuller understanding of
the likely performance of democratic institutions.