Documente Academic
Documente Profesional
Documente Cultură
To cite this article: Owen Temby (2013): What are levels of analysis and what do they
contribute to international relations theory?, Cambridge Review of International Affairs, DOI:
10.1080/09557571.2013.831032
Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [University of Cambridge] at 05:24 31 August 2015
Cambridge Review of International Affairs, 2013
http://dx.doi.org/10.1080/09557571.2013.831032
Owen Temby
McGill University
Abstract The objective of this article is to clarify the significance and usefulness of levels
Downloaded by [University of Cambridge] at 05:24 31 August 2015
of analysis, a central IR concept, but one often used unproblematically. I argue that a level
of analysis should be defined as a social structure that is examined for its effects on
another social structure, or on the same social structure. Therefore, levels of analysis are
also relational, meaning that one is defined, in part, in terms of its associated unit of
analysis. Because this definition conceptualizes levels of analysis as methodological tools
rather than ontological postulates, it is consistent with a wide range of positions on the
agent-structure debate. More specifically, I show that the methodological issue of which
levels of analysis a researcher employs is separate from the ontological issue of whether the
theoretical lens is atomistic (reductionist) or holistic at any given level. One implication of
this definition is that researchers need not view their ontological commitments as overly
methodologically constraining. This article also addresses some questions raised by this
conceptualization, among them the possibility of multiple social structures existing at
a single level.
Introduction
Since at least as early as 1961, when J David Singer published his famous article
on the topic, levels of analysis have been a prominent analytical concept in
international relations (IR) discourse. Neorealists, neoclassical realists and
democratic peace theorists all define their theoretical approaches in terms of the
levels of analysis they employ; Martin Hollis and Steve Smith (1990) use the
concept as their primary means of distinguishing IR theories, and Wendt-inspired
constructivists rely on the distinction between micro- and macro-levels of
structure to provide an account of ideational structural change (Wendt 1999, 2003;
Checkel 2005; Mabee 2007). Despite their common usage, however, there seems
to be little agreement about what levels of analysis actually are, or how they
should be used. It should come as little surprise, then, that some scholars have
insisted that the concept has led to more confusion than it has granted in analytical
precision (Walker 1993; Moravcsik 2003), and that it should be eliminated
altogether (Patomaki 2002).
A previous draft of this article was presented at the 2011 annual conference of the
International Studies Association – Northeast in Providence, Rhode Island. I thank Patrick
Thaddius Jackson, Arne Ruckert, Brian Schmidt, Alexander Wendt and the anonymous
reviewers for the many helpful comments on earlier drafts.
1
See Onuf (1998) for a discussion of the indispensability of levels in social inquiry.
Onuf also provides an alternative historical overview of levels of analysis in IR to what is
presented in this article and expresses concerns similar to mine about conflating ontology
and methodology when conceptualizing levels.
2
For an overview of constitutive reasoning, see Wendt (1998); for an overview of
critical realism, see Dessler (1999), Patomaki and Wight (2000) or Jackson (2010).
Levels of analysis and IR theory 3
international system levels (and apparently eliminate the individual level) is not
arbitrary, but rather flows logically from the decision to study the international
system and the acknowledgement that states are its constituent elements by
definition. This definition of levels of analysis is consistent with Wendt’s (1999)
distinction between micro- and macro-levels of structure, and as such does not
replace Waltz’s (1959) three levels, but instead compliments them. Singer’s state
level is not the same as Waltz’s as the latter’s is an independent variable and the
former’s is a subset of the international system level which portrays that level
from the standpoint of the states—what Wendt (1999) and Buzan et al (1993)
refer to as the interaction (or micro-)level.3
Understanding the distinction between the second image state level and the
interaction level is key here. As stated above, the second image uses properties
of states such as their domestic institutions, production processes and
Downloaded by [University of Cambridge] at 05:24 31 August 2015
social identities to explain behaviour. The interaction level, on the other hand,
posits a causal mechanism at the level of the international system, but at the micro-
level, which depicts the international system from the states’ point of view and
explains state behaviour with reference to the relationship between states (Wendt
1999). This level has been conceptualized in recognition of the phenomenon of
emergent properties resulting from the interactions of states in accordance with no
properties other than their desires and the beliefs they have about how to meet
these desires when they take each other into account when making choices. These
types of structures are characteristic of micro-economic theorizing and game
theory where, in the latter case, units may bargain and exhibit common
perceptions about the activity that they are engaging in.
Similarly, Singer’s (1961, 84 – 85) only posited causal mechanisms are
state ‘goals, motivation, and purpose in national policy’, which ‘are consciously
envisaged and more or less rationally pursued’.4 He also discusses actor
‘perceptions’, in terms of whether scholars need to account for them
when modelling their behaviour. This appears to be a clear example of the use of
the interaction level since Singer abstracts away all properties of states except those
relevant to their strategic interaction. To understand or predict state behaviour,
then, he need only specify the content of these properties and position states in
relation to one another in the form of a game-theoretic model.
It appears that Singer’s international system level, also, is a subset of Waltz’s
third image international system level, in this case the one which Wendt refers to
as the systemic (or macro-)level of analysis, which he defines as the level of
‘multiply realizable outcomes’ since many different combinations of micro-level
interactions will result in the same macro-level outcome.5 Here ‘the posited causal
3
One might reasonably claim that, in fact, Singer is not talking about two levels of
structure, but about agent and structure, another important dichotomy in structural theory.
I disagree with this on the basis that agent and structure is an ontological problem, and
Singer’s concern is unambiguously methodological. The question he addresses regards the
best way to study things, not the way things are.
4
Even though these are Singer’s only causal mechanisms (and thus define his state
level of analysis) I admit that he also treats his state as a dependent variable and, insofar as
he does, employs it as a ‘unit of analysis’.
5
Wendt’s language is confusing here, so I shall clarify. According to Wendt (1999),
every structure has two levels, the micro- and macro-level. The international system, as a
structure, is no different. In the case of this structure, the micro-level is called the
Levels of analysis and IR theory 5
mechanism operates at the level of the population of states, not the level of
individual or interacting states’ (Wendt 1999, 151, italics in original). This is what
Singer (1961, 80) is describing when he says that ‘the systemic level of analysis,
and only this level, permits us to examine international relations in the whole,
with comprehensiveness that is of necessity lost when our focus is shifted to a
lower, more partial level’. The crucial point here is that Singer’s two levels occur
entirely within Waltz’s third image, and that, as I discuss below, Waltz’s second
image can also be portrayed as having a micro- and macro-level.
Waltz (1979) reentered the levels of analysis debate with the publication of Theory
of international politics. In it, he offers a ‘structural’ theory of state behaviour which
represents a significant reevaluation of the views expressed in Man, the state, and war
(1959). Waltz’s new formulation of the level-of-analysis (and agent-structure?)
problem consists of two levels, the unit and the international system. The unit level
Downloaded by [University of Cambridge] at 05:24 31 August 2015
apparently refers to individual states, that is, the ‘second image’, whereas the
international system level, as Wendt (1999) has argued, is the macro-structural level
of the international system, that is, the ‘systemic level’. In other words, Waltz’s
international system level is essentially the same thing as Singer’s, but with greater
specification about the causal behaviour it exerts on states.
According to Waltz (1979), IR theories which employ the international system
level of analysis are ‘systemic’ (a methodological term) and theories which
employ the unit level are ‘reductionist’ (curiously, an ontological term; more on
this below). Since Waltz’s primary concern is to construct a systemic theory of
international politics, he does not spend much space specifying the contents of the
unit level. This is unfortunate, as it leaves a lot of ambiguity as to the definition of
the unit level and its relationship to the international system level. Indeed, as
Keohane and Nye (1987) and Buzan et al (1993) have noted, IR theorists have
treated the unit level as a dumping ground for political phenomena which are not
deemed to be properties of the international system (such as states, interaction
between states, individuals, interest groups and so on).
The fact that Waltz’s (1979) revised conception of the international system can
be expressed as a subset of his first is due to the narrow way he defines the
former in relation to the latter. His third image is a permissive condition, which,
analytically, gives us the latitude to fit two different levels of structure within it,
even though he did not conceive of these levels. It is the international system writ
large, since its effects and the way it causes these effects are largely unspecified.
However, in Theory of international politics, his international system level has a
specific effect—it describes specific behaviour (such as balancing) from the
standpoint of the system, and without any reference to the utility functions of the
states. As such, it precludes a priori the micro-level, which his initial formulation
permits.
Footnote 5 continued
‘interaction’ level, and the macro-level is called the ‘systemic’ level. His failure to more
clearly explicate this point has, arguably, led to a lot of confusion about how these two
levels of the international system contribute to our knowledge of levels of analysis on the
one hand, and agent and structure on the other. See Onuf (1998) for an expression of
uncertainty due to the ambiguity of Buzan et al’s (1993) conceptualization of the interaction
level.
6 Owen Temby
6
Hollis and Smith (1990) distinguish between causal (outsider) explanations and
interpretive (insider) explanations, and argue that holism and individualism are
compatible with both. What is important here is that holism and individualism are
placed on the same causal or interpretive continuum and allowed to be a part of either story,
which enables the difference between holism and individualism to be reduced to simply the
magnitude of their constraining effects.
Levels of analysis and IR theory 7
7
Although this was a novel insight in the levels of analysis debate, the proposition that
scientific inquiry presupposes ontology dates at least as far back as Friedrich Nietzsche. In
his words, ‘[s]trictly speaking, there is no such thing as science “without any
presuppositions”; this thought does not bear thinking through since it is paralogical: a
philosophy, a “faith,” must always be there first of all, so that science can acquire from it a
direction, a meaning, a limit, a method, a right to exist’ (1967, 151– 152).
8 Owen Temby
The debate between Wendt and Hollis and Smith continued among other
scholars (Carlsnaes 1994; Smith 1994; Jabri and Chan 1996; Hollis and Smith 1996;
Chan 1998), but did not continue to explicitly address the levels of analysis
problem. This is unfortunate since the initial debate left much unresolved, in
particular the specifics of how the ontological question of what is relates to the
methodological problem of how to study what is. How do we reconcile Hollis and
Smith’s point about an ontological problem existing at every level of analysis with
Wendt’s enriched conception of holism defined in terms of generative and
constitutive effects (rather than simply constraining and causal effects) of
structure? I shall attempt to resolve this matter below.
The fact that levels of analysis have been used in so many different ways indicates
a demand for language that will give expression to these various related concepts.
To grant us sufficient leverage, a definition of the term should be able to clearly
specify how levels of analysis, micro- and macro-structure and agent and
structure fit together. It should be consistent with how the concept has been used
historically since its inception, but it should also enable scholars with different
ontological and epistemological commitments to converse about the nouns which
constitute IR, even if their views about the ontological status of these nouns differ.
Broadly stated, therefore, I define a level of analysis as a social structure which is
examined for its effects on another social structure, or on the same social
structure—such as when examining the effects of an anthropomorphized state’s
regime type on its social identities.8 If we wish to limit these effects to causal
effects, we can define a level of analysis more narrowly, as an antecedent social
structure whose properties are examined to explain the behaviour or properties of
a contingent structure.9 Following Nuri Yurdusev’s (1993) and John Gerring’s
(2004) example, I call the contingent structure a ‘unit of analysis’, thus
differentiating it in definition from a level of analysis. An implication of these
definitions is that levels of analysis are also relational, which means that one is
defined, in part, in terms of its associated unit of analysis. This is because a level of
analysis is not the social structure itself, but a social structure (however it is
understood) as it is employed in analysis. Kenneth Waltz’s state level is different
than Paul Wapner’s (1995) because Waltz’s unit of analysis is the state, whereas
Wapner’s units of analysis are corporations and consumers.
8
I will employ Kyriakos Kontopoulos’s (1992, 389) definition of a social structure
which, following Anthony Giddens’s lead, is ‘rules and resources recursively implicated in
the reproduction of social systems,’ such as humans or other social kinds. The vital question
regarding individuals is not whether they are social structures, but what the ontological
status of this structure is. Are individuals ontologically privileged entities that mediate
between different impulses and, through an act of will, make choices, or are these structures
reducible to the behaviour of their constituent brain cells, atoms, subatomic particles and so
on? This specific question is outside the scope of this article but, as I explain below, its
answer affects the types of question (causal or constitutive) we are able to ask at the
individual level.
9
For a useful typology of causal effects that can be attributed to levels of analysis, see
Buzan (1995). Buzan’s text also provides an alternative account of the history of the levels of
analysis concept to the one presented in this article.
Levels of analysis and IR theory 9
with the implicit assumptions of closed systems and Humean causality, resulting in
the reification of the hypothesized entities. A corollary of these mistakes is to think
that there are different ‘levels of analysis’ in the study of IR. Social being is reduced
to statistically measurable ‘factors’, with the implicit assumption that the world
is not real, nor differentiated, layered and structured. There are only levels of
analysis and hence perceptions or observations that can be measured and analysed.
(77 – 78, emphasis in original)
For Patomaki, in other words, positivism,11 atomism/individualism and, indeed,
levels of analysis, are coterminous. Although he is possibly the first IR scholar to
explicitly link the latter to the other two, this dubious claim, I contend, is a
manifestation of the more general and common error of conflating the ontological
and the methodological; that is, questions of what is and practical matters about
how to pursue political inquiry. Levels of analysis are not necessarily based on
atomism or individualism, nor must they be used exclusively with causal
reasoning.
First, to put it in simple terms, the ontological question, ‘who and what are the
actors?’, is different from the methodological question, ‘what level of analysis are
we using?’. This is itself a contentious claim since ontologies are often expressed
as methodologies which contain ontological postulates. For example, consider the
ontological positions of reductionism and social atomism. Based on the history of
the usage of the terms social atomism and reductionism, they refer to the same thing.
Reductionism is rooted in René Descartes’ (2003) view that objects of inquiry are
best understood as the sum of empirically observable parts. Thomas Hobbes
(1998, 102) represents this line of thought, as well, in On the citizen, with the
admonition that to understand society we must ‘look at men as if they had just
emerged from the earth like mushrooms and grown up without any obligation to
each other’. Although this is expressed as a methodology (that is, how to look at
10
The reason that holism is a necessary condition for mutual constitution is that the
latter presupposes the existence of social facts which are not properties of ontologically
primitive individuals.
11
By ‘positivism’ Patomaki and I are employing the narrow meaning of the term, as a
philosophy of science consisting of Humean causation, the ‘covering-law’ model of
explanation, an instrumentalist treatment of theoretical terms, and the operationalization of
scientific concepts. See Wight (2002) and Jackson (2010) for an overview of the meaning of
this and other philosophies of science used in IR.
10 Owen Temby
the individuals of which they are composed. However, this does not mean that
individuals themselves must be the object of analysis, since structures can be
examined in terms of the distribution of properties of the individuals or other
constituent parts—just as Waltz (1979) suggests in his definition of structure as the
‘distribution of capabilities’ among state agents.
For a definition of holism, I use one provided by Lewis and Weigert (1985, 455),
who define it as a doctrine that ‘attempts to account for social order by reference to
assumptive or emergent properties of collectivities that are independent of, and
antecedent to interaction among particular individuals’. This is an ontological
stance since, regardless of how the methodological question ‘to which level of
analysis do we ascribe explanatory importance?’ is answered, a choice has been
made to recognize the emergent characteristics of the object under analysis and an
at least implicit position on the agent-structure problem has been established.
The methodological/ontological distinction is shown in Figure 1 where levels
of analysis, as structures, are conceived of in both holistic and atomistic terms. In
the left-hand column, the heading ‘Agent and structure’ refers to the fact that
holism and atomism/reductionism represent solutions to the agent-structure
problem. Hollis and Smith (1990) and Wendt (1999), instead, place holism and
individualism in opposition to each other. This is a mistake, as individualists are
holists at the individual level of analysis; that is, they believe that brain cells and
Levels of analysis
Individual State International system
Agent and structure
12
However, I contend that this definition does not go far enough in expressing the
meaning of atomism, which, by the word atom, connotes a commitment to reductionism
that extends deeper than the individual, such that the individual itself is ontologically
problematic.
Levels of analysis and IR theory 11
other body parts have emergent properties, and as such humans are not reducible
to them. But if an agent-structure problem exists at every level of analysis, then
this is true at the individual level too, meaning that even the ontological status of
individuals is problematic. Therefore, following Taylor (1989), I place social
atomism in opposition to holism, and treat individualism as a special case of
atomism in which the individual is privileged.
Under the ‘Levels of analysis’ heading of Figure 1, I have confined the example
to Waltz’s three images, but this is for convenience only. Arguably, and given my
definition of the term, there are as many potential levels of analysis are there are
social structures. At the state level, the structure may be conceived of as an actual
entity, or as a ‘necessary fiction’, to use EH Carr’s (2001, 137) term, reducible to its
voters, interest groups, government structures and so on. The key point here,
however, is that claiming the state is a real thing, or even a person or ‘unitary
Downloaded by [University of Cambridge] at 05:24 31 August 2015
actor’, does not in any way preclude the use of other levels. If the state is a person,
this does not mean that domestic actors do not matter or do not have agency, but
rather that their properties as domestic actors are generated and constituted by the
structure that they instantiate (Wendt 2004). Thus, as critical realists have
observed, attempts to reduce the state to domestic actors will fail since these
efforts will result in a bulky conception of individuals forced to carry with them
social facts which really exist externally (Archer 1995).
This is also true at the individual and international system levels. The holistic
ontological stance at the individual level is characterized by folk psychology,
which maintains that individual beliefs and desires are irreducible to the
biological elements of human beings. Eliminative materialism, on the other hand,
is atomistic because it maintains that beliefs and desires are reducible in this
way.13 Holism at the individual level is a defining feature of liberal theory in
general, as it is arguably a necessary logical condition for individual rights.
At the international system level, the holistic position maintains that states are
constructed by the international system, whereas the atomistic position maintains
that only their behaviour is affected. The words ‘the distribution of . . . ’ have been
placed under atomism since this is the language used by both Waltz (1979) and
Moravcsik (1997), and since it captures the idea that the structure is defined in
terms of the ontologically prior states that the international system is reduced to.
Second, if levels of analysis are compatible with constitutive reasoning, the
grounds for Patomaki’s (2002) claim that they are based on positivism and should
thus be eliminated is undermined. This is because empiricism, that is, the
reduction of ontology to observable phenomena, is an element of the positivist
philosophy of science by definition. However, constitution presupposes the
existence of unobservable and emergent ideational facts that are irreducible to
agents. The reason is that what changes when individuals, states or the system are
constituted are not materials (guns, brain cells and so on), but emergent social
facts such as shared ideas about appropriate behaviour or the cognitions agents
have about who they are, which are endogenous to structured interaction. These
emergent properties are, by definition, irreducible to agents since they are not
13
See Jackson and Pettit (1990), Paul Churchland (1981) and Patricia Churchland (1986)
for descriptions and defences of folk psychology and eliminative materialism, and
Connolly (2002) and Wendt (2010) for explorations of the link between the study of politics
and the study of the brain.
12 Owen Temby
parts of the agents to begin with. Therefore, to employ constitutive reasoning with
levels of analysis, both the level and the unit must be conceived of holistically. This
arguably violates prominent understandings of positivism (or ‘empiricism’) in IR
and instead presupposes some form of ontological realism.
Conceiving of levels of analysis with constitutive effects is complex, so I will
provide three brief examples: (1) one in which a level of analysis structure
constitutes a different unit of analysis structure, (2) a special case of this in which a
level of analysis structure constitutes a unit of analysis structure which is also its
agents and (3) an example in which the level and unit are the same structure. In
these instances, where both structures are understood holistically, the designation
of which structure represents the level of analysis and which represents the unit is
flexible since each is potentially iteratively examined for its effects on the other
and neither is temporally prior.
Downloaded by [University of Cambridge] at 05:24 31 August 2015
First, if our level of analysis is the state, and our unit of analysis is a network of
actors in global civil society (GCS), the two entities cannot coconstitute unless they
are actually ontologically privileged. If our state is just voters, members of interest
groups and the government structure and so on, then the properties of this state
can be understood to some extent cause the behaviour of our holistically
understood network of GCS actors, and vice versa, but the state’s mutual
constitution with the structure of GCS actors has been ruled out since the state’s
constitutive social facts (for example, mutual understandings identifying the state
and distinguishing it from other social structures) are nonexistent or instead are
properties of its elements. This is not to say that members of domestic interest
groups and government officials on the one hand, and participants of global civil
society policy networks, on the other hand, cannot be understood to coconstitute
through an iterative process of policy learning. Rather, it means that if they are
analysed as doing so, it involves the use of a level and unit of analysis in which
both structures have been conceptualized holistically.
Second, if our level of analysis is the international system and the unit of
analysis is the state, we are arguably examining a mutually constitutive agent-
structure relationship (since states are the international system structure’s agents
or elements). Analyses of this sort include those that use Finnemore and Sikkink’s
(1998) norm life cycle (NLC) approach to explaining how the normative context
among states shapes and is shaped by states and their behaviour. For instance,
Matthew J Hoffmann (2005) used the NLC approach to examine the changing
norms of participation in international efforts to address environmental problems
(in particular, ozone depletion and climate change). He argues that ozone
negotiations were initially characterized by North-only participation, but that,
following the suggestion of a ‘norm entrepreneur’, Southern states become
involved in subsequent amendment negotiations. As many Southern states
became involved, this altered the normative context within which Northern states
operated, and which constrained their own behaviour regarding ozone
negotiations. Through the process of acting in accordance with these norms,
Northern (and Southern) states internalized them, so that the norms caused their
identities to accord with them and they considered universal participation
legitimate. As is typical for a constructivist analysis in which state agents and an
ideational superstructure are understood holistically and interact in a mutually
constitutive way, the level of analysis and unit of analysis are fluid and changing
depending on the focus of the analysis.
Levels of analysis and IR theory 13
Third, if our level of analysis is the state, and our unit of analysis is also
the state, their constitutive effects could be thought of as what David Campbell
(1998, 24) calls the ‘performative constitution of identity’, the ‘reiterative and
citational practice by which discourse produces the effects that it names’
(Butler 1993, 2). Continuing with the environmental theme above, a state that
perceives a particular environmental problem as damaging to its narrowly
defined self-interests may act to ameliorate the problem and, in doing so,
recursively create an identity consistent with these actions, thus redefining
its interests. Robyn Eckersley (2004, 103) argues that the ‘green identity’ of
several European nations, as evidenced during the 1997 Kyoto climate change
negotiations, is due in part to the long-term realization by these states that
greenhouse gas abatement measures are economically beneficial. Here, a state’s
identities and interests are mutually constituted through process.
Downloaded by [University of Cambridge] at 05:24 31 August 2015
Levels of analysis
Individual State International system
Macro-international
Macro-individual Macro-state
(‘systemic’ level)
Holism
Micro-international
Micro-individual Micro-state
(‘interaction’ level)
Macro-international
Macro-individual Macro-state
Atomism
(‘systemic’ level)
Micro-international
Micro-individual Micro-state
(‘interaction’ level)
Figure 2. Micro-macro as distinguished from agent and structure and levels of analysis.
14 Owen Temby
international level is conceived of atomistically, and Fearon and Wendt (2002) and
Wendt (1999) point out that micro-structures are compatible in principle with
constitutive effects, and in such cases are irreducible wholes.14 While constitutive
effects rule out atomism a priori, the macro-structural methodology of explaining
broad patterns of behaviour without reference to the attributes of particular agents
does no such thing if the structure is defined in terms of the attributes of the agents
to begin with, and if it is only granted the capacity to have causal effects (as with
Waltz’s neorealist delineation). Therefore, nothing in the selection of which level
of analysis to employ, macro or micro, connotes a solution to the agent-structure
problem.
As mentioned above, the macro-international level is often referred to as the
systemic level, and the micro-international level as the interaction level. It has
become commonplace to regard theory based on either of these levels to be
Downloaded by [University of Cambridge] at 05:24 31 August 2015
14
This points to a contradiction in Wendt’s work. On the one hand, he makes the
argument just mentioned, that micro-structures can generate properties and have
constitutive effects, and can therefore be holistic. On the other hand, he argues that
micro-level dynamics are rooted in individualism, and macro-level dynamics are rooted in
holism (Wendt 2003). Therefore, in this instance, he arguably commits the common error of
conflating ontology with methodology.
15
Waltz (1979, 18) defines reductionism an approach by which ‘the whole is
understood by knowing the attributes and the interactions of its parts’. It maintains that
‘[o]nce the theory that explains the behavior of the parts is fashioned, no further effort
is required’ (60). Note that Waltz’s definition is virtually identical to mine above. This is
Levels of analysis and IR theory 15
Levels of analysis
Individual State International system
Classical realism
intergovernmentalism
Holism Constructivism
Classical realism Institutionalism
Footnote 15 continued
an ontological postulate, one which maintains that structures do not have emergent
properties and are therefore reducible to their parts. Waltz (18) defines systemic IR
theory as a class of theories that ‘conceive of causes operating at the international level’.
Again, this is a level of analysis, not an ontological stance regarding the agent-structure
problem.
16 Owen Temby
reductionist nor systemic, as it is defined by its use of the state level of analysis, not
its ontology. There are certainly reductionists who theorize about the
phenomenon (Moravcsik 1997), but so do holists (Kahl 1998). Colin Kahl explains
democratic peace with reference to a ‘collective liberal identity’ among irreducible
state agents. Similarly, while Bruce Russett and John Oneal (2001) display no
explicit reductionist or holistic commitments in their effort to construct a theory of
democratic peace, they employ Wendt’s (1999) holistic argument about the
constitutive (and thus irreducible) effects of the international system, which, when
made more peaceful by democracies that believe they can trust each other, further
contributes to the absence of war. Democratic peace theory, then, as theory which
explains state behaviour with reference to the properties of states (conceived in
holistic or reductionist terms), is reduction neutral.
Several of the other classifications in Figure 3 merit clarification and
Downloaded by [University of Cambridge] at 05:24 31 August 2015
explanation. First, classical realist Hans Morgenthau et al (2005) uses the first and
second images to explain state behaviour, while Carr (2001) uses only the second,
and both argue that the structure at the second image does not exist. Second,
neoclassical realists and liberal intergovernmentalists use all three levels while
concurrently maintaining that individuals are the only ontologically privileged
entities (Moravcsik 1997; Schweller 2003). Thus, these approaches arguably
have more in common methodologically than is sometimes acknowledged. Third,
although world systems theory is generally thought to be international systemic, it
also uses the second image insofar as Wallerstein (2004) distinguishes between the
behaviour of ‘weak’ and ‘strong’ states. It is holistic at the international level due to
the deep generative effects it grants to the structure of the capitalist world system,
and atomistic at the state level since it accepts that states are reducible to the narrow
interests of capitalists. Fourth, it should come as no surprise that constructivists use
all three of these levels of analysis (and more), whereas institutionalists do not, as
the Fourth (rationalist-constructivist) Debate is as much about methodology as
ontology. Constructivists value breadth in their accounts of social life, and in
pursuance of this end are holists ontologically, methodologically imputing
causation to multiple structures. Rational institutionalists, on the other hand,
bracket the holistic effects of many structures (thereby not necessarily denying their
existence) and often use a limited number of levels of analysis, both for reasons of
parsimony.
Ambiguities
Although this article has thus far attempted to clarify many of the ambiguities
associated with levels of analysis, the definition proposed necessarily raises its
own. In this section I will clarify two of the most obvious: (1) multiple structures
existing at a single level of analysis, and (2) social systems or structures that cut
across levels of analysis. To this end, I provide one example of a theory in which
the concept of system and structure, at first glance, render the use of levels of
analysis questionable. However, as I explain, a relational understanding of levels
as methodological tools grants the researcher a degree of independence from the
ontological problem of specifying what counts as ‘the international’, for example.
And the fact that nearly all IR theories examine the effects of the empirical stuff of
Levels of analysis and IR theory 17
cuts across various levels, this does not preclude employing a state, some
dimension of global civil society, or the structure of the global political system as a
level of analysis. And, although MST does not agree with the ontology of agent
and structure, and does not view the state as an ontologically privileged entity
outside of the discourse that refers to it, the theory has a commitment to
constitutive theorizing, meaning that the ontological question at any given level is
essentially taken for granted in favour of a holistic understanding of social
organization. This is true of the state level of analysis as well since the state is a
proxy for a geographic instantiation of the political system—what Luhmann
(2000, 63) in a different context referred to as a ‘program strand’—which is
mutually constituted with other geographical instantiations, as well as with other
systems.
Downloaded by [University of Cambridge] at 05:24 31 August 2015
Conclusion
Ultimately, philosophical reflection in IR should serve the purpose of enabling us
to craft accounts of global social life which are coherent and trustworthy. Yet our
subject matter makes this inherently difficult, as a reasonable amount of ambiguity
exists over how to understand IR concepts and the relationships between them. In
this article I suggest that, separate from debating such ontological matters, IR
would benefit from adopting a flexible, yet consistent, understanding of levels of
analysis as methodological tools. More specifically, I have argued for a relational
conception of levels of analysis, as social structures that exist as levels in relation to
a unit of analysis whose behaviour or properties the level’s own properties are
examined to explain. As I have shown, this understanding is consistent with the
way in which the concept has been consciously conceptualized in IR discourse,
dating back to Waltz’s Man, the state, and war (1959) and extending through the
early 1990s debate between Wendt (1991; 1992) and Hollis and Smith (1990; 1991;
1992). Furthermore, this relational understanding of levels of analysis is consistent
with a range of ontological positions and theories employed in the study of global
social phenomena. This includes theories that explicitly or implicitly address
questions related to agent and structure, as well as at least one theoretical approach
(MST) that rejects such reasoning outright.
That said, existing usage of levels of analysis, and the conceptualization of
their relation to agent and structure, has lacked analytical precision. In a recent
review of the literature on the study of the state, one observer argued that ‘it is
inappropriate to treat the state as a unitary actor’ (Levi 2002, 53). Her point is that
doing so precludes a priori the examination of domestic politics as an independent
variable on state action. However, as has been shown in this article, this popular
view is based on a conflation of two different issues, one ontological and one
methodological. What the state is need not determine our decision regarding
whether or not to examine domestic actors for their effect on state action or any
other unit of analysis. But since an agent-structure problem exists at every level of
analysis, it is still important to engage the ontological question when making a
methodological choice. This is because, as I have argued, our ontological
postulates determine some of the possibilities open to us when engaging in
scientific inquiry. If we choose to be atomists we cannot engage in constitutive
theorizing because atomism discards the existence of the requisite social facts.
Levels of analysis and IR theory 19
Holism, on the other hand, rejected by positivists at all but the individual level of
analysis, is perfectly compatible with both causal and constitutive theorizing.
Thus, if levels of analysis and agent and structure do not offer something to
everyone, we should at least recognize that their utility is broader than is often
understood. Nearly every theory of IR uses levels and units, and every theory
solves the agent-structure problem at the levels and units it uses, whether
explicitly or implicitly. The purpose here has not been to advocate the use of
certain levels or a particular ontological view of them, but rather to encourage
their consistent use, in a way that respects their conscious historical development,
so that we can be clear about the relatively distinct issues of agent and structure
and level of analysis and, maybe, have one more common way of expressing
our ideas.
Downloaded by [University of Cambridge] at 05:24 31 August 2015
Notes on contributor
Owen Temby (PhD, Carleton University) is a postdoctoral fellow at the
Department of Natural Resource Sciences, McGill University, and a research
associate at the Loyola Sustainable Research Centre, Concordia University. His
current research examines domestic and transnational environmental policy
networks in Canada and the United States. Email: owen.temby@mcgill.ca
References
Albert, Mathias (1999) ‘Observing world politics: Luhmann’s systems theory of society and
international relations’, Millennium: Journal of International Studies, 28:2, 239– 265
Andriole, Stephen J (1979) ‘The level of analysis problems and the study of foreign,
international, and global affairs: a review critique and another final solution’,
International Interactions, 5:2 – 3, 113 – 133
Archer, Margaret S (1995) Realist social theory: the morphogenetic approach (Cambridge and
New York: Cambridge University Press)
Ashley, Richard K (1984) ‘The poverty of neorealism’, International Organization, 38:2,
225–286
Berkowitz, Bruce D (1986) ‘The level of analysis problem in international studies’,
International Interactions, 12:3, 199– 227
Bull, Hedley (2002) The anarchical society: a study of order in world politics (New York:
Columbia University Press)
Butler, Judith (1993) Bodies that matter: on the discursive limits of sex (New York: Routledge)
Buzan, Barry (1995) ‘The level of analysis problem in international relations reconsidered’
in Steve Smith and Ken Booth (eds) International relations theory today (University Park,
Pennsylvania: Pennsylvania State University Press), 198– 216
Buzan, Barry (1996) ‘The timeless wisdom of realism?’ in Steve Smith, Ken Booth and
Marysia Zalewski (eds) International theory: positivism and beyond (Cambridge and New
York: Cambridge University Press), 47 – 65
Buzan, Barry (2004) From international to world society? English School theory and the social
structure of globalisation (Cambridge and New York: Cambridge University Press)
Buzan, Barry and Mathias Albert (2010) ‘Differentiation: a sociological approach to
international relations theory’, European Journal of International Relations, 16:3, 315– 337
Buzan, Barry, Charles A Jones and Richard Little (1993) The logic of anarchy: neorealism to
structural realism (New York: Columbia University Press)
Campbell, David (1998) National deconstruction: violence, identity, and justice in Bosnia
(Minneapolis: University of Minnesota Press)
Carlsnaes, Walter (1994) ‘In lieu of a conclusion: compatibility and the agent-structure issue
in foreign policy analysis’ in Walter Carlsnaes and Steve Smith (eds) European foreign
policy: the EC and changing perspectives in Europe (London: Sage Publications), 274– 287
20 Owen Temby
Carr, EH (2001) The twenty years’ crisis, 1919– 1939: an introduction to the study of international
relations, ed Michael Cox (Basingstoke, United Kingdom and New York: Palgrave)
Chan, Stephen (1998) ‘An ontologist strikes back: a further response to Hollis and Smith’,
Review of International Studies, 24:3, 441– 443
Checkel, Jeffrey T (2005) ‘International institutions and socialization in Europe:
introduction and framework’, International Organization, 59:4, 801– 826
Churchland, Patricia Smith (1986) Neurophilosophy: toward a unified science of the mind-brain
(Cambridge, Massachusetts: MIT Press)
Churchland, Paul M (1981) ‘Eliminative materialism and the propositional attitudes’,
Journal of Philosophy, 17:2, 67 – 90
Connolly, William E (2002) Neuropolitics: thinking, culture, speed (Minneapolis: University of
Minnesota Press)
Cox, Robert W (1987) Production, power, and world order: social forces in the making of history
(New York: Columbia University Press)
Descartes, René (2003) Treatise of man, ed Thomas Steele Hall (Amherst, New York:
Downloaded by [University of Cambridge] at 05:24 31 August 2015
Prometheus Books)
Dessler, David (1999) ‘Constructivism within a positivist social science’, Review of
International Studies, 25:1, 123– 137
Eckersley, Robyn (2004) ‘Soft law, hard politics, and the Climate Change Treaty’ in Christian
Reus-Smit (ed) The politics of international law (Cambridge and New York: Cambridge
University Press), 80 – 105
Fearon, James and Alexander Wendt (2002) ‘Rationalism v. constructivism? A skeptical
view’ in Walter Carlsnaes, Thomas Risse and Beth Simmons (eds) Handbook of
international relations (London: Sage Publications), 52 –72
Finnemore, Martha and Kathryn Sikkink (1998) ‘International norm dynamics and political
change’, International Organization, 52:4, 887– 917
Gerring, John (2004) ‘What is a case study and what is it good for?’, American Political Science
Review, 98:2, 341– 354
Hobbes, Thomas (1998) On the citizen, ed Richard Tuck and Michael Silverthorne
(Cambridge and New York: Cambridge University Press)
Hoffmann, Matthew J (2005) Ozone depletion and climate change: constructing a global response
(Albany, New York: State University of New York Press)
Hollis, Martin and Steve Smith (1990) Explaining and understanding international relations
(Oxford and New York: Oxford University Press)
Hollis, Martin and Steve Smith (1991) ‘Beware of gurus: structure and action in
international relations’, Review of International Studies, 17:4, 393– 410
Hollis, Martin and Steve Smith (1992) ‘Structure and action: further comment’, Review of
International Studies, 18:2, 187– 188
Hollis, Martin and Steve Smith (1996) ‘A response: why epistemology matters in
international theory’, Review of International Studies, 22:1, 111 –116
Jabri, Vivienne and Stephen Chan (1996) ‘The ontologist always rings twice: two more
stories about structure and agency in reply to Hollis and Smith’, Review of International
Studies, 22:1, 107– 110
Jackson, Frank and Philip Pettit (1990) ‘In defense of folk psychology’, Philosophical Studies,
59:1, 31 – 54
Jackson, Patrick Thaddius (2010) The conduct of inquiry in international relations: philosophy of
science and its implications for the study of world politics (New York: Routledge)
Jaeger, Hans-Martin (2007) ‘“Global civil society” and the political depoliticization of global
governance’, International Political Sociology, 1:3, 257– 277
Kahl, Colin H (1998) ‘Constructing a separate peace: constructivism, collective liberal
identity, and democratic peace’, Security Studies, 82, 94 – 144
Keohane, Robert O (2005) After hegemony: cooperation and discord in the world political economy
(Princeton, New Jersey: Princeton University Press)
Keohane, Robert O and Joseph S Nye (1987) ‘Power and interdependence revisited’,
International Organization, 41:4, 725– 753
Kontopoulos, Kyriakos M (1992) The logics of social structure (Cambridge and New York:
Cambridge University Press)
Levels of analysis and IR theory 21
Levi, Margaret (2002) ‘The state of the study of the state’ in Ira Katznelson and Helen V
Milner (eds) Political science: the state of the discipline (New York and London: Norton),
33– 55
Lewis, J David and Andrew J Weigert (1985) ‘Social atomism, holism, and trust’, Sociological
Quarterly, 26:4, 455– 471
Luhmann, Niklas (1995) Social systems (Stanford, California: Stanford University Press)
Luhmann, Niklas (2000) The reality of the mass media (Stanford, California: Stanford
University Press)
Mabee, Bryan (2007) ‘Levels and agents, states and people: micro-historical sociological
analysis and international relations’, International Politics, 44:4, 431– 449
Mearsheimer, John J (2001) The tragedy of great power politics (New York: Norton)
Moravcsik, Andrew (1997) ‘Taking preferences seriously: a liberal theory of international
politics’, International Organization, 51:4, 513– 553
Moravcsik, Andrew (2003) ‘Liberal international relations theory: a scientific assessment’ in
Colin Elman and Miriam Fendius Elman (eds) Progress in international relations theory:
Downloaded by [University of Cambridge] at 05:24 31 August 2015
Wendt, Alexander (2004) ‘The state as person in international theory’, Review of International
Studies, 30:2, 289– 316
Wendt, Alexander (2010) ‘Flatland: quantum mind and the international hologram’ in
Mathias Albert, Lars-Erik Cederman and Alexander Wendt (eds) New systems theories in
world politics (Basingstoke, United Kingdom: Palgrave Macmillan), 279– 310
Wight, Colin (2002) ‘Philosophy of social science and international relations’ in Walter
Carlsnaes, Thomas Risse and Beth Simmons (eds) Handbook of international relations
(London: Sage Publications), 23 – 51
Yalem, Ronald J (1977) ‘The level-of-analysis problem reconsidered’, Year Book of World
Affairs, 31:3, 306– 326
Yurdusev, Nuri (1993) ‘“Level of analysis” and “unit of analysis”: a case for distinction’,
Millennium: Journal of International Studies, 22:1, 77 – 88