3
Concept Specification in Political
Science Research
Arndt Wonka
The need for reconstruction results from destruction, from the fact that
our disciplines have increasingly lost all ‘discipline’. Amidst the
resulting state of noncumulability, collective ambiguity, and increasing
incommunicability, it is imperative to restore or attempt to
restore the conceptual foundations of the edifice. This is not to say
that an exercise in conceptual reconstruction will restore consensus –
we are far too disbanded for that. However, if the exercise succeeds, it
will restore intelligibility – and, with intelligibility, an awareness of
the enormous intellectual waste brought about by our present-day
indiscipline (and methodological unawareness). (Sartori, 1984, p. 50)
Introduction
Political scientists seek to derive general statements from their empirical
observations. For that purpose they make causal and descriptive inferences.
The goal of inference is to produce reliable descriptive information,
to test existing theories, and to formulate new theories (King,
Keohane and Verba, 1994). The validity of empirical and causal inference,
however, depends crucially on properly specified concepts. First of all, the
clear definition of a concept allows others to understand the meaning of
what we write. In addition, the content of concepts determines the
content as well as the explanatory and the empirical scope of our theoretical
hypotheses. For other steps in the process of designing research,
unambiguous concepts are most obviously important in the design of
an empirical strategy and the subsequent development of adequate
measures (for a discussion of measurement, see Miller, Chapter 5). The
41
reason for this is obvious: How are we to evaluate a measure’s adequacy
if we are not sure what to measure in the first place?
A number of articles and book chapters have been devoted to theoretical
discussions of concepts and concept specification (Collier and
Mahon, 1993; Gerring, 2001; Sartori, 1970, 1984). My aim in this
chapter is much more modest and instrumental: I would first of all like
to draw the reader’s attention to the centrality of concepts in political
science research, by discussing how the quality of concepts affects the
clarity of theoretical arguments and the empirical scope of theories. In
line with the other chapters of this volume, in section three, I will then
provide practical guidelines for how to get most out of the concepts we
wish to apply in our research projects. The fourth section applies this
chapter’s recommendations to the concept of ‘supranationality’ which
I frequently struggle with in my own research on the European Union
(EU). A brief discussion concludes this chapter.
Design problem: concepts and concept
specification in political science research
There are three components which make up a concept and have to be
distinguished analytically (Gerring, 2001; Sartori, 1984): a term assigns a
name to a concept. Attributes which define a concept’s meaning fill the
term with substance. All attributes taken together constitute a concept’s
intension. A concept’s intension is not only important because it defines
its meaning. The intension of a concept should demarcate this concept
from other concepts. Otherwise overlaps in meaning will lead to confusion.
Finally, a concept’s defining attributes relate the concept to real
world phenomena. The empirical scope of a concept is regularly referred
to as its extension. Analytically useful concepts draw distinct boundaries
between the real world phenomena denoted by themselves and those
denoted by other theoretical concepts. Figure 3.1 summarizes the above
and puts it into the broader research design context: We start with a
theoretical statement, that is, a hypothesis (see De Bièvre, Chapter 11;
Dür, Chapter 10). To make sure that anyone is able to understand the
statement’s meaning, we specify the concepts contained in the statement.
This is done by explicating the concepts’ defining attributes.
Finally, the concept has to be operationalized in order to systematically
relate it to real world phenomena (Miller, Chapter 5) – and to test the
empirical plausibility of our theoretical statements.
Concept specification then is the process by which a researcher
defines and explicates the attributes of the concepts she uses in her
42 Concepts and Theory
research. As such, concept specification can mean both the adaptation
of an existing concept to actual theoretical or empirical needs and the
formation of a completely new concept, which has to be specified from
scratch. The (instrumental) goal of both kinds of concept specification is
the same: We want to have analytical instruments at our disposal, which
allow us an unambiguous theoretical discourse as well as the clear denotation
of the empirical phenomena relevant for our research project.
Therefore we have to be clear about what the terms mean with which we
operate and confront other researchers. In political science the term
‘democracy’, for example, refers to a specific form of organizing the
relationship between those that govern and those governed in a given
territory. The term could be specified by the following defining attributes:
(1) the guarantee of citizens’ basic rights; (2) the rule of law; and (3)
regular competing elections for political offices. This specification of the
concept of democracy then extends to all states whose organization of
political power exhibits these attributes.
Some political scientists may oppose this chapter’s postulate that
unambiguous concepts are a prerequisite for meaningful (empirical)
research. They take vague concepts and associative discussions as an
inspiration and thus deliberately operate with vague concepts. They
Concept Specification in Political Science Research 43
Figure 3.1 Concepts and concept specification in the broader research design
context
Theoretical Statement/Hypothesis
Real World Phenomena
Operationalization
Concept
Specification
Extension: Designated phenomena (= referents)
Concept (= term)
Intension: Defining attributes(= meaning)
mistake conceptual vagueness for creativity. I hope that the arguments
in this chapter convince the reader that, rather than promoting inventive
political science research, vague concepts inhibit it.
Another reason why concepts often have an ambiguous meaning is
that the same term is used with different defining attributes. As a consequence,
many concepts lurk under the same term. In 1942, Baudin
made this observation with respect to corporatism: ‘The army of corporatists
is so disparate that one is led to think that the word, corporation,
itself is like a label placed on a whole batch of bottles which are then distributed
among diverse producers each of whom fills them with the
drink of his choice. The consumer has to look carefully’ (Baudin, quoted
in Schmitter, 1979, p. 10).1
As this example makes clear, scientific discourse
based on ambiguous concepts is at least confusing, more likely
unproductive and definitely not cumulative.
Yet another reason for conceptual ambiguity is that political science
concepts often originate outside the academic discourse. Researchers
might take them from politicians’ discourse or from everyday speech.
The already discussed concept of ‘democracy’ is an appropriate example,
as is ‘freedom’, ‘justice’, or ‘happiness’. Such concepts are regularly
laden with historical and political connotations and carry a backpack of
meanings. As a result, they are often highly ambiguous.
To avoid such ambiguity, one could argue that only ‘scientific’ concepts
should be employed in research. In the extreme case, such advice
would amount to the exclusive reliance on a strictly formalized artificial
language, out of touch with popular discourse. Is such advice reasonable
when taking into account the costs of such a proceeding? I do not think
so. Relying on formalization will definitely exclude problems which
arise from a concept’s popular use. In addition, it is unlikely that an
artificial concept created for a specific scientific purpose will acquire
different meanings over time. At least that is what we know from other
scientific disciplines. Think, for example, about chemistry’s table of elements.
Yet, the costs of proceeding in this manner are considerable:
Anyone who wants to use such concepts or interact with researchers
using them will have to resort to the formalized language. Existing concepts
would eventually have to be translated. It is unrealistic that political
scientists in general will acquire these skills and that formalization
therefore will be the cure for conceptual ambiguity in our discipline. In
addition, concepts which originated in popular discourse might be the
most interesting and socially relevant. Political scientists have the
professional skills to inform public discourse on popular concepts such
as, for example, ‘democracy’. Yet, the exclusive reliance on an artificial
44 Concepts and Theory
language would severely limit our ability to communicate our results to
a non-academic audience and to contribute productively to public discourse
(see Lehnert, Miller, Wonka, Chapter 2). The most important
reason, which speaks against the reliance on using artificial language to
arrive at unambiguous concepts, is much simpler, however: The sine qua
non for a concept’s unambiguity is the explication of its defining attributes.
Our natural languages are sufficient for this purpose, as the example
of ‘democracy’ above shows.2
The definition of concepts must take place in the early phases of
designing a research project. The effort that has to be invested in this
step of designing a research project might vary. If no specification of a
concept is available in the literature, we have to come up with our own
definition. This definition should not contradict the meaning usually
associated with the term, because such a proceeding would give away
the ‘everyday-analytical’ leverage of the term which it owes to its nonscientific
use. In most instances, however, we will be able to rely on a
concept’s specification in the literature. This is not least because we
usually choose to embed our research projects in specific scientific
discourses. The choice of the scientific discourse will most likely be
based on the fact that we deem the concepts it deals with appropriate for
our research purposes. In case we can rely on already specified concepts,
our task is to make the respective concepts’ definition explicit. If, for
example, one plans to work with the concept of ‘corporatism’, one
should be explicit about which of the many variants of the concept
(Siaroff, 1999) one is working with.
(Re-)specification of a concept has to be considered if the original
concept does not fully cover a researcher’s theoretical interest, or if
the concept will be applied to a different spatial or temporal context. If
the concept is applied to a different context, the fit between the cases
investigated and the original concept might be lost. Thus, to make the
concept fit for traveling to a new empirical context, it might be necessary
to respecify it (see Rathke, Chapter 6). Otherwise, concepts are
stretched to empirical cases which the concept does not cover due to a
misfit between the latter’s defining attributes and the former’s empirical
characteristics. This might eventually lead to an erroneous theoretical
classification of empirical cases. Descriptive as well as causal inferences
drawn from such a basis would then be erroneous as well.
Reducing the number of defining attributes is one way to re-specify a
concept to adapt it to a different empirical context: ‘The rules of climbing
and descending along a ladder of abstraction are thus very simple
rules – in principle. We make a concept more abstract and more general
Concept Specification in Political Science Research 45
by lessening its properties or attributes. Conversely, a concept is
specified … by augmenting its attributes or properties’ (Sartori, 1970,
p. 1041). Making a concept more abstract might allow for a valid application
to a wider empirical context. Yet, it does not come without costs,
because such a strategy is very likely to reduce a concept’s analytical
leverage. An abstract concept of corporatism defined as ‘a political system
in which private interest groups and public actors interact on a regular
basis’ is applicable to a large number of political systems. At the
same time it offers little analytical leverage, because it does not allow us
to empirically discriminate between systems of interest intermediation
with different structural and behavioral properties. Theoretically, such a
definition is nonsense since it blurs the differences between the theoretical
concepts of ‘pluralism’ and ‘corporatism’ and thereby causes ambiguity.
As an alternative to decreasing the theoretical distinctiveness and
the analytical leverage of ‘corporatism’ by increasing the level of abstraction,
one could revert to a different, even more abstract concept, which
satisfies the theoretical as well as the empirical needs. To capture interest
intermediation between private and public actors, one could for
example revert to ‘governance’. Now, whether such a strategy is reasonable
depends first of all on the theoretical interests pursued. If one is
interested in studying ‘corporatism’, operating with ‘governance’ is no
viable option. If the application of a more abstract concept is not precluded
for substantive reasons and if the abstract theoretical concept
gives the researcher analytical leverage nothing speaks against using it.
Another strategy for adapting concepts to different empirical contexts
is based on the premise that a concept consists of one or a few central
attributes and a number of non-central attributes. Collier and Mahon
call such concepts ‘radial concepts’ (Collier and Mahon, 1993). While
the ‘prototype’ radial concept contains all defining (central and noncentral)
attributes, variants of the concept may contain only the central
attribute and one of the prototype concept’s non-central attributes. To
designate the different meanings of the respective variant terminologically,
an adjective is added to the prototype concept (Collier and
Mahon, 1993, p. 848). Consequently, not all cases covered by one of the
concept’s variants have to exhibit all of the prototype concept’s attributes.
All variants nevertheless share some of the prototype concept’s
attributes. Thus, to allow for empirical variations of the prototype of
‘corporatism’, many variants have been formulated by adding an adjective.
‘Sectoral’ corporatism, for example, has been introduced to apply
the concept to countries in which the structure of interest representation
in some economic sectors is corporatist while in others pluralist
46 Concepts and Theory
patterns might prevail. Theoretically such a refinement might be of
interest to reflect on how such ‘mixed’ systems perform macroeconomically.
Empirically the refinement of the concept is instrumental to
avoid overgeneralizations in describing countries’ structures of interest
representation. This strategy might preserve the basic meaning of the
(prototype) concept, while its substantial and terminological qualification
as a variant of the concept guarantees its valid empirical application.
Thus, radial concepts allow the flexible adaptation of a concept to
a wider context without necessarily having to make it more abstract or
eventually having to resort to the use of a different concept. Yet the
formulation of ever more variants of the prototype concept might
negatively affect this concept’s original appeal by reducing its distinctiveness
and its analytical leverage by blurring clear demarcations
between the different variants. Another lesson learnt from the dazzling
conceptual history of ‘corporatism’.
Which of the above discussed strategies one applies to adapt a concept
to its research employment very much depends on the researcher’s
theoretical goals and the theoretical state in the respective field. In a
conceptually highly fragmented context, the researcher might not only
strive for more abstract concepts for empirical reasons, but also to
achieve theoretical integration by cutting through the conceptual jungle.
A researcher with such an interest should follow Sartori’s advice and
define abstract concepts with relatively little properties or attributes. In
the opposite case – that is, if the concepts in a given field are highly
abstract and ambiguously employed – the researcher will very likely opt
for a radial concept, because increasing the level of abstraction would
run counter to the researcher’s intention. In addition, one might want
to preserve the original concept to the largest extent possible and simply
qualify it for an extended application, because it has thus far been
meaningfully applied in the literature and has gained considerable
prominence.
Finally, instead of respecifying existing concepts, there is always the
option of forming completely new concepts. However, the benefits of
forming new concepts should always be carefully weighed against
potential costs. If well-specified concepts exist which already serve our
theoretical interests, the temptation to form a new concept should be
resisted. A pragmatic reason for this is to save energy affiliated with promoting
a new concept and arguing for its usefulness. A substantive reason
is that the creation of ever new concepts bears the risk of ever
smaller research communities developing their own conceptual languages.
This inhibits intelligible discourse across those communities and
Concept Specification in Political Science Research 47
risks making intellectual cross-fertilizations among sub-disciplines
increasingly difficult to achieve. If the goal of a research project is the
innovation of a completely new theory, however, then concept formation
will be one of its central tasks. The same holds if one aims at replacing
a highly fragmented and ambiguous concept with a new one. This is
what Siaroff (Siaroff, 1999) did with his proposal to capture countries’
differences in their systems of interest intermediation through his concept
of ‘integration’ instead of the established concept of corporatism. If
none of the above is the goal of a research project, the formation of new
concepts should be treated with care.
To this point this chapter has made clear that in order to validly apply
a theoretical concept it must be properly defined in the first place, must
have an unambiguous meaning and, moreover, the empirical domain it
covers should correspond to the concept’s defining attributes. What has
not yet been addressed, though, is how to handle a concept validly in
heterogeneous empirical contexts (van Deth, 1998, Rathke, Chapter 6).
An extreme position would be to argue that no concept can be meaningfully
applied across different contexts. One could arrive at such a
position by arguing that real world objects designated by a concept
relate to and affect their social, political and economic environment differently
in different empirical contexts.3
Thus, they play different roles
in different contexts. For instance, the concept of parties designates reasonably
similar organizations in democratic political systems. Yet, with
respect to the role they play in democratic and autocratic political systems,
respectively, parties differ considerably. For empirical inferences,
this is not a problem, since the aim then is not to relate a concept to
others, but to describe its occurrence across contexts. Yet, for causal
inference it might be a problem. Thus, we may use the concept ‘party’ to
denote equivalent (in terms of their attributes) organizations across
democratic and autocratic political systems. We should not, however,
automatically make causal inferences on the equivalence of the role
parties play in all these contexts based on the label itself.
I will finish this section with a short remark on concepts which cannot
be directly observed. It is a truism that something which cannot be
observed constitutes a particular challenge in empirical research. Should
political scientists therefore stop using concepts such as power, influence
or legitimacy? Of course every empirical researcher wishes to operate with
observable concepts, which can be validly and reliably measured. If we
have the choice between a concept which is directly observable and one
which is not and both fit our purpose, we should definitely go with the
former. In addition, if a particular specification of a concept allows us to
48 Concepts and Theory
observe it directly, we should specify the concept accordingly. This greatly
facilitates the operationalization and valid measurement of the respective
concept and allows for a straightforward decision whether a particular
case can be subsumed under a concept. If our theoretical and substantive
interests lead us to use concepts which are not directly observable, everything
which has been said so far applies as well: Specifying a (unobservable)
concept properly prevents unobservability equaling unintelligibility.
Particular effort must then be spent in arguing for the adequacy of a particular
way of measuring a given concept (Miller, Chapter 5).
To summarize: Concepts are the building blocks of our theories. For a
productive and cumulative scientific discourse among the participants
of a research community it is necessary for the meaning of concepts to
be clearly understood. The whole purpose of concept specification is to
define and explicate a concept’s meaning to avoid ambiguity with
respect to a concept’s (theoretical) meaning. At the same time, a concept’s
specification analytically demarcates it from other concepts. In addition,
a properly specified concept allows the researcher to deal with the
empirical vagueness inherent in any theoretical concept and constitutes
a proper base on which empirical operationalization and measurement
can take place (Collier, Brady and Seawright, 2004b). Thus, a concept
properly specified is not only imperative for intelligible theoretical discourse
but also to design an inter-subjectively comprehensible and
methodologically justifiable research strategy, which serves as a sound
bridge between social science theory and the ‘real world’:
Let it be stressed, therefore, that long before having data which can
speak for themselves the fundamental articulation of language and of
thinking is obtained logically – by cumulative conceptual refinement
and chains of coordinated definitions – not by measurement.
Measurement of what? We cannot measure unless we know first what
it is that we are measuring. (Sartori, 1970, p. 1038)
… It should be understood, therefore, that operational definitions
implement, but do not replace, definitions of meaning. Indeed there
must be a conceptualization before we engage in [empirical] operationalization.
As Hempel recommends, operational definitions
should not be ‘emphasized to the neglect of the requirement of
systematic import’. This is also to say that definitions of meaning of
theoretical import, hardly operational definitions, account for the
dynamics of intellectual discovery and stimulation. (Sartori, 1970,
p. 1045)
Concept Specification in Political Science Research 49
Of course, the presentation of this way of proceeding is ideal-typical. Each
empirical researcher will and should think about the (im-)possibilities
of operationalizing a concept, while working on the theoretical specification
of the concept (Gerring, 2001). Otherwise she risks finding out
that, after having invested considerable effort in the specification of her
theoretically highly interesting concepts, the project does not fly empirically.
Yet, concept specification and the operationalization of concepts
must be treated separately! If, during the research process, we recognize
that it is difficult to measure a concept, it is hardly conceivable to go
back and change the concept’s specification with a view on data availability.
Concept specification is foremost guided by a particular theoretical
interest. This theoretical interest is unlikely to change due to
measurement problems. Thus, when facing difficulties in measuring a
concept, instead of re-specifying it to make it fit the data, one should
rather discuss potential problems with the validity of the measure
applied (Miller, Chapter 5). The presentation in this chapter is intended
to remind researchers of the often neglected but nonetheless constitutive
function well-defined concepts have in our research.
The consequences of improper concept specification can be summarized
in three distinct points. Concepts whose meaning is ambiguous
due to insufficient specification hamper the collective and cumulative
effort of (political) scientists by leading to:
1. Theoretical infertility. Lack of precision in the meaning of theoretical
concepts renders an intelligible and critical theoretical discourse
impossible and constrains improvement on purely intellectual
grounds.
2. Empirical arbitrariness. Not clearly specified theoretical concepts have
a vague empirical denotation and their operationalization and measurement
is vulnerable to criticisms of arbitrariness.
3. Invalid (empirical and theoretical) inferences. Concepts without a clearly
defined meaning risk being stretched to empirical objects with overly
heterogeneous properties and/or extended to temporal and/or spatial
contexts, in which the roles of the objects covered by the respective
concept are far too different to subsume them meaningfully under
one concept.
Practical guidelines: six rules of concept specification
In the previous paragraphs I have argued that concepts are of prime
importance for doing empirical social science research. Unfortunately,
50 Concepts and Theory
however, social science concepts often leave the reader puzzled about
the concept’s exact meaning and its theoretical status in the author’s
argument. In this section I provide six practical guidelines which help to
avoid conceptual problems.
1 Search the literature for specifications of the concepts
you will use in your research project!
The early phase of the research process should be spent searching the
relevant literature for specifications of the concepts which will eventually
be used in the research project. We might recognize that we work in
a field with properly specified concepts. If applying one of these well
specified concepts in our own research, we have to make sure that the
specification fits our theoretical and empirical purpose – and that we
explicate the definition of the concepts (Rules 2 to 6). Yet, a researcher
might as well discover that her field is full of dazzling concepts which do
not come with specifications. She then has to come up with such definitions
herself. In order to avoid contradictory terminology, the definition
should be in line with the meaning normally associated with the
concept.
2 Explicate clearly and exhaustively the attributes
you ascribe to the concept(s) used in your
theoretical framework!
Concepts are important in empirical research, because they allow for a
systematic look at the objects of investigation. To exploit a concept’s full
potential, the attributes one ascribes to a concept must be explicated.
The explication of a concept’s attributes makes sure that the researcher
herself as well as potential readers becomes fully aware of the concept’s
meaning. We thereby avoid ambiguities, which again is a prerequisite to
having a meaningful scientific discourse. When a concept’s intension is
explicated, existing definitions in the literature must be taken into
account. This saves you energy and helps to avoid terminological babble
as well as concepts that increasingly lose their distinctive meaning and
analytical leverage due to cross-cutting specifications.
When I state that corporatist arrangements have a positive effect on a
country’s macroeconomic performance, I have to state what I mean by
corporatism. This is important to allow others and myself to understand
the exact content of the concept and the causal hypotheses for which it
is used. It also allows us to clearly demarcate an argument from other
arguments which operate with the same concepts. Corporatism is
defined, among other things, by interest group concentration. It will
Concept Specification in Political Science Research 51
make a difference whether concentration is located at the sectoral or the
national level. Knowing this will greatly help us make sense of an
author’s causal hypotheses. If various studies using the same concepts
come to different conclusions, checking their respective concept
specifications will eventually tell us why this is the case.
3 Think hard about how the attributes of the
concept relate to each other – and to the concept’s
overall meaning!
Since the attributes ascribed to a concept define its meaning, it is important
that they add up to a coherent meaning. A concept is coherent if a
logical relationship between its attributes is clearly discernible. In other
words: If it makes intuitive sense to group these things under one label.
Thus, it makes sense to define corporatism as a ‘system of interest
representation in which the constituent units are organized into a limited
number of singular, compulsory, noncompetitive, hierarchically
ordered and functionally differentiated categories’ (Schmitter, 1979,
p. 13). Adding ‘networked’ to this list of attributes would be confusing
and adding ‘competitive’ would be plainly contradictory. Adding attributes
whose relationship with other attributes is confusing or even contradictory
decreases the coherence of a concept’s meaning and thus
limits its analytical purchase.
4 Try to keep your concept’s level of abstraction
low (if it is theoretically reasonable)!
A concept’s level of abstraction should correspond to the theoretical and
empirical needs of a research project. It should not be formulated in an
unnecessarily abstract way. An important reason for avoiding abstract
concepts in empirical research is their empirical vagueness. As the level of
abstraction increases, concepts loose direct reference to a concrete set of
real world phenomena. As a consequence, it becomes much more difficult
to choose and argue for an indicator which allows for the valid measurement
of the concept. If an abstract concept is needed, the problems of
operationalization and measurement are traded against the concept’s theoretical
generalizability. When facing such a trade-off, we should keep in
mind that a concept which cannot be reasonably operationalized and
measured does not allow for empirical and causal inferences.
5 Relate the concept’s attributes to the units of analysis you
empirically investigate!
The attributes of a concept have to be explicitly related to the units of
analysis to which the respective concept refers and which are the objects
52 Concepts and Theory
of the empirical investigation. The clarification of the relationship
between a concept’s attributes and its empirical referents not only prevents
us from inadequately stretching concepts to objects for which
they are inaptly suited. The explication of the units of analysis also contributes
considerably to the comprehensibility of a concept’s meaning.
It allows the reader to form a concrete idea about that to which the
author refers with his concept. If, for example, someone hypothesizes
about a public actor’s supranational preferences in EU decision-making,
it will help our substantive understanding to know whether the person
talks about decisions in general, legislative decisions or constitutional
decisions. In addition, explicating the units of analysis to which a
concept refers facilitates the assessment of the validity of a concept’s
empirical operationalization. A researcher who argues about the European
Commission’s supranational preferences in EU legislative decisionmaking
and empirically investigates the European Commission’s behavior
in the bargaining on the European Constitution will not be able to
convince us with her empirical results.
6 Be aware that in empirical research any concept eventually
needs to be operationalized and measured!
Before one can think about the operationalization of a concept, the concept
needs to be properly specified. Only after the full meaning of the
concept has been determined is it then possible to make an adequate
choice about a concept’s operationalization and measurement. Yet,
when designing a research project, one should from the very beginning
take into account potential problems: Data might not be easily available
and it will almost certainly prove difficult to get an empirical grip on a
theoretical concept. Thus, while the proper theoretical specification is
the fundamental precondition to theoretical intelligibility, the practical
problems in measuring a concept (Miller, Chapter 5) will finally determine
whether an empirical research project is feasible.
Application: underspecified and overextended?
The concept of supranationality in European
Union research
When reading academic as well as journalistic accounts on European
Union (EU) politics, one frequently comes across the concept of ‘supranationality’.
Most often in its adjective form ‘supranational’. The term
‘supranational’ is usually used in opposition to national. Applied to EU
decision-making, which will be the main subject of this section, the
‘supranational scenario’ (Tsebelis and Garrett, 2000) conceptualizes
Concept Specification in Political Science Research 53
political conflicts in EU decision-making as struggles about more
‘national’ or more ‘supranational’ policy solutions with actors positioning
themselves according to their national or supranational policy
preferences respectively.
Rule 2 of this chapter’s practical recommendations states that the use
of a concept should always be accompanied by the clarification of the
concept’s meaning. ‘Supranationality’ as used in the literature is defined
by two dimensions: An institutional and a motivational. The meaning
of the institutional dimension is straightforward: The EU’s institutionallegal
system is supranational, since it is a system independent of the
institutional-legal orders of the member states. In addition, its supranational
quality results from its superiority to member state law in the
sense that in cases of conflict between national and EU law, the latter
dominates the former. A supranational public actor, accordingly, is one
which is constituted by the EC treaty and draws its competences from it.
The meaning of the institutional dimension of supranationality is thus
clearly defined by the institutional-legal quality of the political system
of the EU. Its defining attributes are independence and superiority vis-à-vis
member state law.
The motivational dimension of supranationality, on the other hand,
denotes actors’ interests or ideas driving their actions in EU politics. The
meaning of this dimension is much less clearly defined. Usually it is
taken to mean that actors with supranational preferences want ‘more
Europe’ (cf. Pollack, 2003, p. 36). Yet, how can we conceive of an actor’s
preferences for ‘more Europe’ in EU politics, that is, when the units of
analysis are individual policies (Rule 5)? For constitutional politics the
answer is clear: Supranational constitutional preferences are defined by
the desire to further integrate policy areas, for which the member states
have exclusive or predominant policy-making competences up to this
point. When bargaining about further institutional integration during
EU Treaty negotiations, an actor with supranational constitutional
preferences wants more competences to be transferred from the member
state to the EU level. In EU constitutional politics, supranational constitutional
preferences thus have a clear meaning and can in principle be
unambiguously identified in empirical research (Moravcsik, 1998).
At this point it becomes obvious why the two dimensions of supranationality
I explicated above, should be separated analytically: The British
government might have a strong preference for the (constitutional)
integration of further policy areas in the EU Treaties. Yet this does not
make it a supranational actor in institutional terms. At the same time,
the European Commission might strongly oppose the integration of a
54 Concepts and Theory
policy area, while being a supranational actor in institutional terms.
Automatically inferring from the institutional to the motivational
dimension of supranationality, and vice versa, would lead to false inference.
Rule 3 of the practical recommendations shall help us avoid such
false inferences by clarifying the relationship between the attributes of a
concept. It is of course possible to define supranationality in terms of
both dimensions – in other words, to extend the concept from the
institutional to motivational dimension. The analytical usefulness and
validity of such a definition, however, must be empirically established.
During the last decade, EU scholars increasingly turned their attention
to the everyday politics of EU legislative decision-making. Thus, the
unit of analysis changed from constitutional to legislative decisions. As
stated in the practical recommendations, when extending a concept to
new units of analysis, the validity of such a strategy and its meaningfulness
should be carefully considered (Rule 5). According to the ‘supranational
scenario’ the pro-integrationist European Commission and
European Parliament (EP) side with pro-integrationist governments to
legislatively realize ‘more Europe’ (Hörl, Warntjen and Wonka, 2005;
Tsebelis and Garrett, 2000). Thus, in the case of the European
Commission and the EP the institutional and the motivational dimension
are said to coincide. According to the above definition, this is clearly not
the case with respect to the so called pro-integrationist member state
governments.
How then can we conceive of supranational legislative preferences? If
there has not been any EU legislation and a decision will establish such
legislation, those actors preferring EU legislation could be conceptualized
as having supranational preferences: they want to replace national
with European legislation. Such a definition would then be a direct analogy
to the constitutional preferences discussed above. But how can legislative
decisions be conceptualized along the national-supranational
continuum, which are embedded in a policy area for which extensive
European legislation already exists? In these policy areas the question is
not whether to replace national policies with EU policies, but rather
which form and content further EU regulations shall take. The meaningfulness
and accordingly the empirical usefulness of conceptualizing
the EU political space along the national-supranational continuum can
be seriously questioned in such a context. Thus, when the concept of
supranational preferences is extended to cover legislative politics,
researchers should explicate the meaning of the concept applied to the
new subject matter (Rule 5). Otherwise, concepts which proved highly
meaningful and useful in one context – that is, supranational
Concept Specification in Political Science Research 55
preferences in EU constitutional politics – are stretched to contexts in
which their meaningfulness can be questioned and should be seriously
debated by the research community. The ex post imposition of old concepts
to new contexts might otherwise seriously handicap rather than
facilitate our understanding of decision-making processes in the EU
(Hörl, Warntjen and Wonka, 2005).
Empirical results and theoretical interpretations generated by two
papers affiliated to the most encompassing and systematical large-n
empirical research project on EU decision-making help to point out the
conceptual problems with the application of ‘supranationality’ in the
study of EU legislative decision-making (Kaeding and Selck, 2005;
Thomson, Boerefijn and Stokman, 2004). Both studies operate with the
concept of supranationality and both employ similar methodological
tools with which they identify comparable empirical patterns.
Comparing the theoretical inferences of these two studies allows for a
discussion of the importance of investigating the fit between a concept
and the empirical units of analysis to which it is applied. Michael
Kaeding and Thorsten Selck analyze positional data4
of 70 EU legislative
proposals which comprised 174 controversial legislative issues (Kaeding
and Selck, 2005; Thomson, Boerefijn and Stokman, 2004). Their goal is
to find out how the European Commission, the EP and member state
governments position themselves in EU legislative decision-making.
Having identified their empirical pattern, Kaeding and Selck conclude
that: ‘the supranational institutions seem to be largely ideological actors
taking extreme positions outside the clusters. Their policy positions are
significantly different from the member states … The Commission and
EP are much more pro-integrationist than any member state, a fact
which our three-dimensional solution demonstrates more clearly than
the existing two-dimensional studies by Selck (2003) and Thomson et al.
(2004)’ (Kaeding and Selck, 2005, pp. 282–3). From their empirical
results the authors infer that the Commission and the EP are indeed
institutionally and motivationally supranational actors. They thereby
support the two dimensional concept of supranationality commonly
used in the literature to characterize the EU’s supranational institutions
(Pollack, 2003, p. 36; Tsebelis and Garrett, 2000, p. 16).
Robert Thomson and his colleagues analyze exactly the same data.
They employ very similar methodological tools and also rely on the
concept of supranationality. With respect to the positions taken by
the member state governments, the European Commission and the
European Parliament in EU legislative decision-making processes, the
56 Concepts and Theory
authors conclude as follows:
Two dimensions on which the preferences of the actors can be placed
were identified. On the first dimension, the Commission and
European Parliament’s position are located at one end, and the reference
point (the outcome if no decision is taken) at the other. The
Member States are clustered at the centre of this dimension. This clustering
of the Member States indicates that there are no Council members
that are consistently closer to the Commission’s position than
others. Their support depends on the proposal at stake at any particular
time. (Thomson, Boerefijn and Stokman, 2004, p. 256)
Again, the common assumption about the Commission’s and the
European Parliament’s supranational policy preferences seems to be
supported.
Yet, in scrutinizing the empirical pattern they identified and in searching
for a theoretical interpretation of their empirical finding, Thomson
and his colleagues did not only stretch the supranational scenario to
their empirical results. Instead they checked whether the concept of
supranationality can be meaningfully extended to their empirical cases.
After having done this, they elaborate that ‘[a]lthough the ordering of
actors on this dimension resembles that posited in the IntegrationIndependence
dimension [i.e., the ‘supranational’ scenario, AW], we
found that this ordering is neither confined to, nor even concentrated
in, issues that contain choices between European harmonization versus
national solutions’ (Thomson, Boerefijn and Stokman, 2004, p. 256).
Yet, the units of analysis of a legislative decision-making process must
allow for a choice between a more ‘European’ and a more ‘national’ policy
solution in order to meaningfully apply the (motivational dimension
of the) concept of supranationality. If they do not, an actor’s
actions in the respective decision-making process cannot be driven by
her preference for a more ‘European’ solution. Such an interpretation
would amount to saying that someone who has chosen between two
dishes – fruit salad and potato salad – and went for the latter was driven
by her preference for meat. Specifying what the units of analysis in this
case were – in other words, fruit and potato salad – excluded the ‘meat
preference’ as a logically possible explanation for that actor’s action as
well as the outcome, that is, her eating potato salad. Applied to the less
tasty topic of EU legislative decision-making this means that to interpret
an actor’s behavior in EU decision-making as supranational when ‘more
Concept Specification in Political Science Research 57
Europe’ is not on offer in the policy decision under investigation, one
infers a wrong motive from an actor’s action and erroneously proposes
this motive as the cause for the political dynamic leading to the
observed outcome. Such an inference is theoretically misleading.
Had Thomson and his colleagues (2004) applied the concept of the
‘supranational scenario’ to their empirical findings without checking
the quality of their units of analysis, they would have extended the
concept to referents in a theoretically misleading way. Following this
strategy, they would have concluded that political dynamics in EU legislative
decision-making are decisively shaped by the EP and the
Commission taking ‘supranationalist’ positions – just as Kaeding and
Selck concluded in their paper. Yet analysing the same empirical data
and employing similar data analysing methods, the two studies drew
quite different theoretical inferences. These differences result from
Thomson and his colleagues’ careful examination of their study’s
empirical referents. In order to interpret the pattern they identified as
supranational, they checked whether the decisions in their sample
involved a choice between more or less European harmonization. If
it did, the expectation of the ‘supranational scenario’ is that the
Commission – which is conceptualized to want ‘more Europe’ (harmonization)
in the supranational scenario – positions itself on the harmonization
end of the scale. Yet what they do find is that ‘[o]f the
remaining 130 issues, we find the Commission and the reference point
at opposite ends of the issue scale on 60 (46 percent) of the cases. Most
importantly, these 60 issues are not concentrated in the group of 40
issues classified as harmonisation issues. Of the 60 issues on which we
find the reference point and Commission at opposite extremes of the
issue scales, only 16 (27 percent) referred to such harmonisation issues.
Moreover, on issues involving clear choices between more or less harmonisation,
the reference point and Commission were not significantly
more likely to be at opposite extremes than on other issues’
(Thomson, Boerefijn and Stokman, 2004, p. 253). By their close examination
of their empirical data and by clear concept specification, the
authors avoided the extension of the concept to units of analysis not
covered by the concept. Accordingly they conclude that ‘a more
detailed inspection of the actor alignments does not support the supranational
scenario’ (Thomson, Boerefijn and Stokman, 2004, p. 252).
The authors’ attentive application of the concept is in line with Rule 5
of this chapter’s practical applications and demonstrates how careful
investigation of the units of analysis helps to avoid concept misapplication
and misleading theoretical inference.
58 Concepts and Theory
Obviously, the empirical cases analyzed by the authors of both papers
show a difference in the positions taken by the Commission and those
taken by the member state governments. Yet the scrutiny of the units of
analysis showed that (at least in these cases) the concept of supranationality
should be restricted to the institutional dimension. Extending
it to the motivational dimension leads to erroneous or at least contradictory
theoretical inferences. The explication of both dimensions of
the concept at the beginning of this section (Rule 2, practical guidelines)
allowed us to identify this restriction in the concept’s applicability to EU
legislative decision-making. This opens the possibility to re-specify the
concept accordingly. It might as well lead to the abandonment of the
concept when studying actors’ behavior in EU legislative decisionmaking
processes and lead to the application of a different concept or
the formulation of a new concept – and thus to a potentially different
understanding of the political dynamics governing the successive
legislative integration of EU member states.
One conceptual option is to abandon the concept of ‘supranationality’
in order to avoid any confusion between its institutional and its
behavioral dimension. Thus, one could refer to a less EU specific and
more abstract concept such as ‘centralization’ in order to capture political
dynamics leading to the harmonization of policies across EU member
states. The use of this concept might invite theoretical discourse
with political scientists already working with this concept, yet dealing
with empirical objects other than the EU – such as for example international
organizations in general (e.g., Koremenos, Lipson and Snidal,
2001). Yet the decision to abandon ‘supranationality’ – instead of just
specifying it properly – and replacing it with ‘centralization’ also
involves considerable costs: Instead of using the established concept –
although perhaps with a slightly different specification – to relate one’s
argument to the rest of the literature, one first of all has to establish the
new concept and relate this to the rest of the literature. Whether one
takes on these costs might as well be influenced by a researcher’s general
interest in theoretical work. Again, these costs might be worth the effort,
if the expected payoff is high enough – that is, if the chances are good
that the respective scholarly community will welcome the new concept
because it provides them with superior analytical leverage over their
empirical field of interest.
Ideally, Kaeding and Selck had taken up the argument by Thomson
and his colleagues that the concept of supranationality, defined by the
institutional and the motivational dimension, cannot be meaningfully
applied to the legislative cases which they analyzed to kick off a debate
Concept Specification in Political Science Research 59
clarifying the concept’s limits. Both studies’ partly contradictory theoretical
interpretations, and the careful conceptual analysis by Robert
Thomson and his colleagues improve our conceptual understanding of
political processes in the EU. Taking this knowledge into account when
re-specifying the concept of supranationality to apply it to EU legislative
research or when forming new concepts will improve our inferences
from empirical analyses.
Conclusion
The aim of political science research is to add to our systematic knowledge
about political facts, events and processes. To be able to achieve
this we need appropriate theoretical and methodological tools. In this
chapter I have argued that to profit most from the theoretical and analytical
potential of concepts, their meaning has to be clearly specified. It
has to be specified with respect to a concept’s (unambiguous) meaning –
that is, its intension – as well as its empirical referents (extension). If
theoretical concepts are properly specified they serve as solid bridges
between social science theory and the ‘real world’ social processes in
which we are interested.
Ambiguous concepts not only hamper intelligible theoretical discourse,
thus frustrating the improvement of social science theory. Ambiguous
concepts also lend themselves to misleading theoretical ex post rationalization
of empirical findings. This merely confirms perspectives on a
given subject of which we grew fond, rather than revealing new insights.
Thus, they are also stumbling blocks on our way to valid empirical and
causal inferences. Having said all this, this chapter’s final comment on
ambiguous concepts is: Avoid them by all means!
Notes
1. Schmitter’s (1979) article is a brilliant discussion of an overly ambiguous
concept – that is, corporatism – and an outstanding illustration of how to
overcome such ambiguity by concept specification.
2. However, to express the relationship between concepts – that is, to formulate
causal hypotheses – formalization allows much more precise statements than
does a non-formalized language.
3. Note that what I discuss here is the theoretical equivalence of a concept in different
contexts. This is different from a discussion of the empirical equivalence
of indicators to measure the same concept. While my discussion
addresses the question whether the same concept has the same theoretical status
across different contexts, the latter discusses which indicators are most
60 Concepts and Theory
suitable to measure the same concept in different contexts (for such a discussion,
see van Deth 1998 and Rathke, Chapter 6).
4. In interviews, experts positioned the EU legislative actors on a continuum
from 1 to 100 on each of the 174 issues. The end points of the scale represent
the extreme solutions for a respective issue. For example, no animal fat
allowed in chocolate vs. 50 percent of animal fat allowed in a piece of chocolate.
The actors were located on the respective issue dimension according to
their preferences.
Concept Specification in Political Science Research 61