Review 2

Taking Preferences Seriously:
A Liberal Theory of
International Politics
Andrew Moravcsik

This article reformulates liberal international relations (IR) theory in a nonideological and
nonutopian form appropriate to empirical social science. Liberal IR theory elaborates the
insight that state-society relations—the relationship of states to the domestic and transna-
tional social context in which they are embedded—have a fundamental impact on state
behavior in world politics. Societal ideas, interests, and institutions influence state behavior
by shaping state preferences, that is, the fundamental social purposes underlying the strate-
gic calculations of governments. For liberals, the configuration of state preferences matters
most in world politics—not, as realists argue, the configuration of capabilities and not, as
institutionalists (that is, functional regime theorists) maintain, the configuration of informa-
tion and institutions. This article codifies this basic liberal insight in the form of three core
theoretical assumptions, derives from them three variants of liberal theory, and demon-
strates that the existence of a coherent liberal theory has significant theoretical, methodologi-
cal, and empirical implications. Restated in this way, liberal theory deserves to be treated as
a paradigmatic alternative empirically coequal with and analytically more fundamen-
tal than the two dominant theories in contemporary IR scholarship: realism and insti-
tutionalism.

For detailed comments and criticisms, I am grateful above all to Anne-Marie Slaughter, who was there
from the beginning, and to Lea Brilmayer, rence Broz, Marc Busch, James Caporaso, Dale Copeland,
David Dessler, Jeffry Frieden, Martha Finnemore, Charles Glazer, Michael Griesdorf, Stefano Guzzini,
Ernst Haas, Stanley Hoffmann, Stephen Holmes, Ted Hopf, Alan Houston, David Lumsdaine, Robert
Keohane, Yuen Khong, Larry Kramer, David Long, Steven Lukes, James Marquart, Lisa Martin, Jonathan
Mercer, Henry Nau, Kalypso Nicolaı̈dis, James Nolt, Joseph Nye, John Odell, Kenneth Oye, Robert
Paarlberg, Daniel Philpott, Gideon Rose, Judith Shklar, David Skidmore, Allison Stanger, Janice Stein,
Andrew Wallace, Celeste Wallander, Stephen Walt, Alexander Wendt, Mark Zacher, Fareed Zakaria, Michael
Zürn, and three anonymous referees. I thank also two other critics: Peter Katzenstein encouraged a more
direct comparison with constructivist approaches and John Mearsheimer invited me to state the liberal
case vis-à-vis realism in a series of public debates. I am also indebted to participants in seminars at the
Program on International Politics, Economics, and Security (PIPES), University of Chicago; University of
Konstanz; University of Toronto; University of California, San Diego; Olin Institute and Center for Inter-
national Affairs, Harvard University; International Jurisprudence Colloquium, New York University
School; Fletcher School, Tufts University; and the European University Institute. For research support, I
thank Amit Sevak, Brian Portnoy, and PIPES. For more detailed and documented versions of this article,
see Moravcsik 1992.

International Organization 51, 4, Autumn 1997, pp. 513–53
r�1997 by The IO Foundation and the Massachusetts Institute of Technology

@xyserv1/disk3/CLS_jrnl/GRP_inor/JOB_inor619/DIV_619z01 kris

514 International Organization

Grounding liberal theory in a set of core social scientific assumptions helps over-
come a disjuncture between contemporary empirical research on world politics and
the language employed by scholars to describe IR as a field. Liberal hypotheses
stressing variation in state preferences play an increasingly central role in IR scholar-
ship. These include explanations stressing the causal importance of state-society re-
lations as shaped by domestic institutions (for example, the ‘‘democratic peace’’), by
economic interdependence (for example, endogenous tariff theory), and by ideas about
national, political, and socioeconomic public goods provision (for example, theories
about the relationship between nationalism and conflict). Liberal hypotheses do not
include, for reasons clarified later, functional regime theory. Yet the conceptual lan-
guage of IR theory has not caught up with contemporary research. IR theorists con-
tinue to speak as if the dominant theoretical cleavage in the field were the dichotomy
between realism and (‘‘neoliberal’’) institutionalism. The result: liberal IR theory of
the kind outlined earlier is generally ignored as a major paradigmatic alternative.

Worse, its lack of paradigmatic status has permitted critics to caricature liberal
theory as a normative, even utopian, ideology. Postwar realist critics such as Hans
Morgenthau and E. H. Carr took rhetorical advantage of liberalism’s historical role as
an ideology to contrast its purported altruism (‘‘idealism,’’ ‘‘legalism,’’ ‘‘moralism,’’
or ‘‘utopianism’’) with realism’s ‘‘theoretical concern with human nature as it actu-
ally is [and] historical processes as they actually take place.’’1 Forty years later, little
has changed. Robert Gilpin’s influential typology in international political economy
juxtaposes a positive mercantilist view (‘‘politics determines economics’’) against a
narrower and conspicuously normative liberal one (‘‘economics should determine
politics’’). Kenneth Waltz, a realist critic, asserts that ‘‘if the aims . . . of states be-
come matters of . . . central concern, then we are forced back to the descriptive level;
and from simple descriptions no valid generalizations can be drawn.’’2

Liberals have responded to such criticisms not by proposing a unified set of positive
social scientific assumptions on which a nonideological and nonutopian liberal theory can
be based, as has been done with considerable success for realism and institutionalism, but
by conceding its theoretical incoherence and turning instead to intellectual history. It is
widely accepted that any nontautological social scientific theory must be grounded in a set
of positive assumptions from which arguments, explanations, and predictions can be de-
rived.3 Yet surveys of liberal IR theory either collect disparate views held by ‘‘classical’’
liberal publicists or define liberal theory teleologically, that is, according to its purported
optimism concerning the potential for peace, cooperation, and international institutions in
world history. Such studies offer an indispensable source of theoretical and normative inspi-
ration. Judged by the more narrowly social scientific criteria adopted here, however, they
do not justify reference to a distinct ‘‘liberal’’ IR theory.

Leading liberal IR theorists freely concede the absence of coherent microfounda-
tional assumptions but conclude therefrom that a liberal IR theory in the social scien-

1. See Morgenthau 1960, 4; Keohane 1989, 68, n. 17; and Howard 1978, 134.
2. See Waltz 1979, 65, 27; Gilpin 1975, 27 (emphasis in original); and Gilpin 1987.
3. See Bueno de Mesquita 1996, 64–65; and Keohane 1986.

@xyserv1/disk3/CLS_jrnl/GRP_inor/JOB_inor619/DIV_619z01 kris

Liberal Theory of International Politics 515

tific sense cannot exist. Robert Keohane, an institutionalist sympathetic to liberal-
ism, maintains that ‘‘in contrast to Marxism and Realism, Liberalism is not committed
to ambitious and parsimonious structural theory.’’ Michael Doyle, a pioneer in ana-
lyzing the ‘‘democratic peace,’’ observes that liberal IR theory, unlike others, lacks
‘‘canonical’’ foundations. Mark Zacher and Richard Matthew, sympathetic liberals,
assert that liberalism should be considered an ‘‘approach,’’ not a theory, since ‘‘its
propositions cannot be . . . deduced from its assumptions.’’4 Accurate though this
may be as a characterization of intellectual history and current theory, it is second-
best social science.

I seek to move beyond this unsatisfactory situation by proposing a set of core
assumptions on which a general restatement of positive liberal IR theory can be
grounded. In the first section of the article I argue that the basic liberal insight about
the centrality of state-society relations to world politics can be restated in terms of
three positive assumptions, concerning, respectively, the nature of fundamental so-
cial actors, the state, and the international system.

Drawing on these assumptions, I then elaborate three major variants of liberal
theory—each grounded in a distinctive causal mechanism linking social preferences
and state behavior. Ideational liberalism stresses the impact on state behavior of
conflict and compatibility among collective social values or identities concerning the
scope and nature of public goods provision. Commercial liberalism stresses the im-
pact on state behavior of gains and losses to individuals and groups in society from
transnational economic interchange. Republican liberalism stresses the impact on
state behavior of varying forms of domestic representation and the resulting incen-
tives for social groups to engage in rent seeking.5

Finally, I demonstrate that the identification of coherent theoretical assumptions is
not simply an abstract and semantic matter. It has significant methodological, theo-
retical, and empirical implications. The utility of a paradigmatic restatement should
be evaluated on the basis of four criteria, each relevant to the empirical researcher:
superior parsimony, coherence, empirical accuracy, and multicausal consistency.

First, a theoretical restatement should be general and parsimonious, demonstrat-
ing that a limited number of microfoundational assumptions can link a broad range of
previously unconnected theories and hypotheses. This restatement does so by show-
ing how liberalism provides a general theory of IR linking apparently unrelated
areas of inquiry. The theory outlined here applies equally to liberal and nonliberal
states, economic and national security affairs, conflictual and nonconflictual situa-
tions, and the behavior both of individual states (‘‘foreign policy’’) and of aggrega-
tions of states (‘‘international relations’’). Liberal theory, moreover, explains impor-
tant phenomena overlooked by alternative theories, including the substantive content
of foreign policy, historical change, and the distinctiveness of interstate relations
among modern Western states.

4. See Keohane 1990, 166, 172–73; Doyle 1986, 1152; Zacher and Matthew 1992, 2; Matthew and
Zacher 1995, 107–11, 117–20; Hoffmann 1987, 1995; and Nye 1988.

5. For other such distinctions, see Keohane 1990; and Doyle 1983.

@xyserv1/disk3/CLS_jrnl/GRP_inor/JOB_inor619/DIV_619z01 kris

516 International Organization

Second, a theoretical restatement should be rigorous and coherent, offering a clear
definition of its own boundaries. This restatement does so by demonstrating that
institutionalist theories of regimes—commonly treated as liberal due to ideological
and historical connotations—are in fact based on assumptions closer to realism than
to liberalism. This helps to explain why IR theorists have found it difficult to distill a
set of coherent microfoundational assumptions for liberal theory.

Third, a theoretical restatement should demonstrate empirical accuracy vis-à-vis
other theories; it should expose anomalies in existing work, forcing reconsideration
of empirical findings and theoretical positions. This restatement of liberal theory
meets this criterion by revealing significant methodological biases in empirical evalu-
ations of realist theories of ‘‘relative gains-seeking’’ and constructivist analyses of
ideas and IR due to the omission of liberal alternatives. If these biases were corrected,
liberal accounts might well supplant many widely accepted realist and institutional-
ist, as well as constructivist, explanations of particular phenomena in world politics.

Fourth, a theoretical restatement should demonstrate multicausal consistency. By
specifying the antecedent conditions under which it is valid and the precise causal
links to policy outcomes, a theory should specify rigorously how it can be synthe-
sized with other theories into a multicausal explanation consistent with tenets of
fundamental social theory. This restatement does so by reversing the nearly universal
presumption among contemporary IR theorists that ‘‘systemic’’ theories like realism
and institutionalism should be employed as an analytical ‘‘first cut,’’ with theories of
‘‘domestic’’ preference formation brought in only to explain anomalies—a prescrip-
tion that is both methodologically biased and theoretically incoherent. In its place, this
restatement dictates the reverse: Liberal theory is analytically prior to both realism and
institutionalism because it defines the conditions under which their assumptions hold.

If this proposed reformulation of liberal IR theory meets these four criteria, as I
argue it does, there is good reason to accord it a paradigmatic position empirically
coequal with and analytically prior to realism and institutionalism, as well as construc-
tivism, in theory and research on world politics.

Core Assumptions of Liberal IR Theory

Liberal IR theory’s fundamental premise—that the relationship between states and
the surrounding domestic and transnational society in which they are embedded criti-
cally shapes state behavior by influencing the social purposes underlying state pref-
erences—can be restated in terms of three core assumptions. These assumptions are
appropriate foundations of any social theory of IR: they specify the nature of societal
actors, the state, and the international system.

Assumption 1: The Primacy of Societal Actors

The fundamental actors in international politics are individuals and private groups,
who are on the average rational and risk-averse and who organize exchange and
collective action to promote differentiated interests under constraints imposed by
material scarcity, conflicting values, and variations in societal influence.

@xyserv1/disk3/CLS_jrnl/GRP_inor/JOB_inor619/DIV_619z01 kris

Liberal Theory of International Politics 517

Liberal theory rests on a ‘‘bottom-up’’ view of politics in which the demands of
individuals and societal groups are treated as analytically prior to politics. Political
action is embedded in domestic and transnational civil society, understood as an
aggregation of boundedly rational individuals with differentiated tastes, social com-
mitments, and resource endowments. Socially differentiated individuals define their
material and ideational interests independently of politics and then advance those
interests through political exchange and collective action.6 Individuals and groups
are assumed to act rationally in pursuit of material and ideal welfare.7

For liberals, the definition of the interests of societal actors is theoretically central.
Liberal theory rejects the utopian notion that an automatic harmony of interest exists
among individuals and groups in society; scarcity and differentiation introduce an
inevitable measure of competition. Where social incentives for exchange and collec-
tive action are perceived to exist, individuals and groups exploit them: the greater the
expected benefits, the stronger the incentive to act. In pursuing these goals, individu-
als are on the average risk-averse; that is, they strongly defend existing investments
but remain more cautious about assuming cost and risk in pursuit of new gains. What
is true about people on the average, however, is not necessarily true in every case:
some individuals in any given society may be risk-acceptant or irrational.

Liberal theory seeks to generalize about the social conditions under which the
behavior of self-interested actors converges toward cooperation or conflict. Conflict-
ual societal demands and the willingness to employ coercion in pursuit of them are
associated with a number of factors, three of which are relevant to this discussion:
divergent fundamental beliefs, conflict over scarce material goods, and inequalities
in political power. Deep, irreconcilable differences in beliefs about the provision of
public goods, such as borders, culture, fundamental political institutions, and local
social practices, promote conflict, whereas complementary beliefs promote harmony
and cooperation. Extreme scarcity tends to exacerbate conflict over resources by
increasing the willingness of social actors to assume cost and risk to obtain them.
Relative abundance, by contrast, lowers the propensity for conflict by providing the
opportunity to satisfy wants without inevitable conflict and giving certain individuals
and groups more to defend. Finally, where inequalities in societal influence are large,
conflict is more likely. Where social power is equitably distributed, the costs and
benefits of actions are more likely to be internalized to individuals—for example,
through the existence of complex, cross-cutting patterns of mutually beneficial inter-
action or strong and legitimate domestic political institutions—and the incentive for
selective or arbitrary coercion is dampened. By contrast, where power asymmetries
permit groups to evade the costs of redistributing goods, incentives arise for exploit-
ative, rent-seeking behavior, even if the result is inefficient for society as a whole. 8

6. This does not imply a ‘‘pre-social’’ conception of the individual unencumbered by nation, commu-
nity, family, or other collective identities but only that these identities enter the political realm when
individuals and groups engage in political exchange on the basis of them; see, for example, Coleman 1990.

7. Kant 1991, 44.
8. Milgrom and Roberts 1990, 86–87.

@xyserv1/disk3/CLS_jrnl/GRP_inor/JOB_inor619/DIV_619z01 kris

518 International Organization

Assumption 2: Representation and State Preferences

States (or other political institutions) represent some subset of domestic society, on
the basis of whose interests state offıcials define state preferences and act purpo-
sively in world politics.

In the liberal conception of domestic politics, the state is not an actor but a repre-
sentative institution constantly subject to capture and recapture, construction and
reconstruction by coalitions of social actors. Representative institutions and practices
constitute the critical ‘‘transmission belt’’ by which the preferences and social power
of individuals and groups are translated into state policy. Individuals turn to the state
to achieve goals that private behavior is unable to achieve efficiently.9 Government
policy is therefore constrained by the underlying identities, interests, and power of
individuals and groups (inside and outside the state apparatus) who constantly pres-
sure the central decision makers to pursue policies consistent with their preferences.

This is not to adopt a narrowly pluralist view of domestic politics in which all
individuals and groups have equal influence on state policy, nor one in which the
structure of state institutions is irrelevant. No government rests on universal or unbi-
ased political representation; every government represents some individuals and
groups more fully than others. In an extreme hypothetical case, representation might
empower a narrow bureaucratic class or even a single tyrannical individual, such as
an ideal-typical Pol Pot or Josef Stalin. Between theoretical extremes of tyranny and
democracy, many representative institutions and practices exist, each of which privi-
leges particular demands; hence the nature of state institutions, alongside societal
interests themselves, is a key determinant of what states do internationally.

Representation, in the liberal view, is not simply a formal attribute of state institu-
tions but includes other stable characteristics of the political process, formal or infor-
mal, that privilege particular societal interests. Clientalistic authoritarian regimes
may distinguish those with familial, bureaucratic, or economic ties to the governing
elite from those without. Even where government institutions are formally fair and
open, a relatively inegalitarian distribution of property, risk, information, or organi-
zational capabilities may create social or economic monopolies able to dominate
policy. Similarly, the way in which a state recognizes individual rights may shape
opportunities for voice.10 Certain domestic representational processes may tend to
select as leaders individuals, groups, and bureaucracies socialized with particular
attitudes toward information, risk, and loss. Finally, cost-effective exit options, such
as emigration, noncompliance, or the transfer of assets to new jurisdictions or uses,
insofar as they constrain governments, may be thought of as substitutes for formal
representation.11

9. Representative political institutions and practices result from prior contracts and can generally be
taken for granted in explaining foreign policy; but where the primary interests and allegiances of indi-
viduals and private groups are transferred to subnational or supranational institutions empowered to repre-
sent them effectively, a liberal analysis would naturally shift to these levels.

10. Doyle 1997, 251–300.
11. North and Thomas 1973, 87.

@xyserv1/disk3/CLS_jrnl/GRP_inor/JOB_inor619/DIV_619z01 kris

Liberal Theory of International Politics 519

Societal pressures transmitted by representative institutions and practices alter ‘‘state
preferences.’’ This term designates an ordering among underlying substantive out-
comes that may result from international political interaction. Here it is essential—
particularly given the inconsistency of common usage—to avoid conceptual confu-
sion by keeping state ‘‘preferences’’ distinct from national ‘‘strategies,’’ ‘‘tactics,’’
and ‘‘policies,’’ that is, the particular transient bargaining positions, negotiating de-
mands, or policy goals that constitute the everyday currency of foreign policy. State
preferences, as the concept is employed here, comprise a set of fundamental interests
defined across ‘‘states of the world.’’ Preferences are by definition causally indepen-
dent of the strategies of other actors and, therefore, prior to specific interstate politi-
cal interactions, including external threats, incentives, manipulation of information,
or other tactics. By contrast, strategies and tactics—sometimes also termed ‘‘prefer-
ences’’ in game-theoretical analyses—are policy options defined across intermediate
political aims, as when governments declare an ‘‘interest’’ in ‘‘maintaining the bal-
ance of power,’’ ‘‘containing’’ or ‘‘appeasing’’ an adversary, or exercising ‘‘global
leadership.’’12 Liberal theory focuses on the consequences for state behavior of shifts
in fundamental preferences, not shifts in the strategic circumstances under which
states pursue them.

Representative institutions and practices determine not merely which social coali-
tions are represented in foreign policy, but how they are represented. Two distinc-
tions are critical. First, states may act in either a unitary or ‘‘disaggregated’’ way. In
many traditional areas of foreign policy, ‘‘politics stops at the water’s edge,’’ and
there is strong coordination among national officials and politicians. In other areas,
the state may be ‘‘disaggregated,’’ with different elements—executives, courts, cen-
tral banks, regulatory bureaucracies, and ruling parties, for example—conducting
semiautonomous foreign policies in the service of disparate societal interests.13 Sec-
ond, domestic decision making may be structured so as to generate state preferences
that satisfy a strong rationality condition, such as transitivity or strict expected utility
maximization, or so as to satisfy only the weaker rationality criterion of seeking
efficient means. Recently, formal theorists have derived specific conditions under
which nonunitary state behavior can be analyzed ‘‘as if’’ it were unitary and rational,
implying that much superficially ‘‘nonrational’’ or ‘‘nonunitary’’ behavior should
actually be understood in terms of shifting state preferences.14

Taken together, assumptions 1 and 2 imply that states do not automatically maxi-
mize fixed, homogeneous conceptions of security, sovereignty, or wealth per se, as
realists and institutionalists tend to assume. Instead they are, in Waltzian terms, ‘‘func-
tionally differentiated’’; that is, they pursue particular interpretations and combina-
tions of security, welfare, and sovereignty preferred by powerful domestic groups

12. The phrase ‘‘country A changed its preferences in response to an action by country B’’ is thus a
misuse of the term as defined here, implying less than consistently rational behavior; see Sebenius 1991,
207.

13. See Slaughter 1995; and Keohane and Nye 1971.
14. Achen 1995.

@xyserv1/disk3/CLS_jrnl/GRP_inor/JOB_inor619/DIV_619z01 kris

520 International Organization

enfranchised by representative institutions and practices.15 As Arnold Wolfers, John
Ruggie, and others have observed, the nature and intensity of national support for
any state purpose—even apparently fundamental concerns like the defense of politi-
cal and legal sovereignty, territorial integrity, national security, or economic welfare—
varies decisively with the social context.16 It is not uncommon for states knowingly
to surrender sovereignty, compromise security, or reduce aggregate economic wel-
fare. In the liberal view, trade-offs among such goals, as well as cross-national differ-
ences in their definition, are inevitable, highly varied, and causally consequential.17

Assumption 3: Interdependence and the International System

The configuration of interdependent state preferences determines state behavior.

For liberals, state behavior reflects varying patterns of state preferences. States
require a ‘‘purpose,’’ a perceived underlying stake in the matter at hand, in order to
provoke conflict, propose cooperation, or take any other significant foreign policy
action. The precise nature of these stakes drives policy. This is not to assert that each
state simply pursues its ideal policy, oblivious of others; instead, each state seeks to
realize its distinctive preferences under varying constraints imposed by the prefer-
ences of other states. Thus liberal theory rejects not just the realist assumption that
state preferences must be treated as if naturally conflictual, but equally the institution-
alist assumption that they should be treated as if they were partially convergent,
compromising a collective action problem.18 To the contrary, liberals causally privi-
lege variation in the configuration of state preferences, while treating configurations
of capabilities and information as if they were either fixed constraints or endogenous
to state preferences.

The critical theoretical link between state preferences, on the one hand, and the
behavior of one or more states, on the other, is provided by the concept of policy
interdependence. Policy interdependence is defined here as the set of costs and ben-
efits created for foreign societies when dominant social groups in a society seek to
realize their preferences, that is, the pattern of transnational externalities resulting
from attempts to pursue national distinctive purposes. Liberal theory assumes that
the pattern of interdependent state preferences imposes a binding constraint on state
behavior.

Patterns of interdependence or externalities induced by efforts to realize state pref-
erences can be divided into three broad categories, corresponding to the strategic
situation (the pattern of policy externalities) that results.19 Where preferences are
naturally compatible or harmonious, that is, where the externalities of unilateral poli-

15. Ruggie 1983, 265.
16. Ruggie 1982, 1983.
17. On the contradictions within Waltz’s effort to avoid these ambiguities, see Baldwin 1997, 21–22.
18. Keohane 1984, 10; 1986, 193. Note that these are all ‘‘as if’’ assumptions. The world must be

consistent with them, but need not fulfill them precisely.
19. See Stein 1982; Snidal 1985; and …