When Political Philosophy Meets Political Science: Theoretical and Methodological Challenges in the...
-
Upload
independent -
Category
Documents
-
view
1 -
download
0
Transcript of When Political Philosophy Meets Political Science: Theoretical and Methodological Challenges in the...
1
When Political Philosophy Meets Political Science:
Theoretical and Methodological Challenges in the Study of a Philosophical Ideal1
Paper to be presented on the 3rd ECPR General Conference, Budapest, 8-10 September 2005
André Bächtiger & Dominik Hangartner
Institute of Political Science and Institute of Sociology, University of Bern
Lerchenweg 36, CH-3000 Bern 9
E-mail: [email protected], [email protected]
Reconnecting political philosophy with empirical or positive political science has become a
resounding call in the discipline (Rothstein 2004). As Alexander Wendt (2001) bemoans, much
positive political science is driving with the rearview mirror, by explaining what happened in
the past while neglecting the question where we should go from here. But if one is to say
something about the future, both “can” and “should” must be addressed. Yet, when normative
philosophical concepts are to be integrated in positive political science, researchers may
confront both theoretical and methodological challenges. Theoretically, they need to develop
reconstructions of normative theories that can satisfy both philosophers and positive theorists.
In other words, the theoretical bridges must be realistic enough without completely destroying
the utopian content of philosophical constructs. As we shall argue in this article, such
bridge-building activities can imply that metatheoretical debates – e.g. rationalism vs.
constructivism – have to give way to an approach that borrows strength from different (and
potentially opposed) metatheoretical perspectives. Methodologically, one may tap into
unchartered empirical territory requiring that researchers adopt unconventional research
designs while simultaneously devise measures of evaluation that allow putting the
philosophical concept in operational terms (Habermas 2005). Another methodological
challenge is that one cannot expect that standard social science data analysis techniques are
1 We thank Adrienne Héritier, Claudius Wagemann, Emanuel von Erlach, Marco R. Steenbergen, Christian Bolliger, Regula Zürcher, Anina Hitz, the participants of the departmental seminar at the European University Institute (EUI) (October 2005), and the participants of the workshop ”Methods” at the Annual Meeting of the Swiss Association of Political Science (November 2005) for extremely helpful and constructive comments on this project.
2
appropriate when empirically investigating concepts of political philosophy. Rather, since these
concepts frequently materialize as “rare events” in the real world, more sophisticated statistical
techniques are in order to correct for potential biases of standard techniques.
This article attempts to identify institutional contexts that make the democratic ideal of
“deliberative agreement” more likely in real world politics. By “deliberative agreement”, we
mean that political actors agree with counterarguments and/or demands of other actors
(whereby in case of demands, mere acceptance is not enough: actors must simultaneously
acknowledge them). We propose that only a broad understanding of institutions and
institutional effects can bring us closer to the Habermasian ideal. On the one hand, we build on
a rationalist approach which can explain the occurrence of “deliberative agreement” either by
the existence of specific incentives or – since it may be impossible to fully incorporate the
“reflective” and “transformative” aspects of deliberation in a rational incentives’ framework
(Müller 1995) - by the absence of a full-fledged strategic and calculus setting which opens up a
space for deliberative action (see Aspingwall/Schneider 2000). On the other hand, we build on a
constructivist or sociological approach constructivist which can explain “deliberative
agreement” as a product of specific guidelines or scripts inscribed in the very structure of
institutions. In concrete, we hypothesize that consociational, veto, and reflective institutions
enhance the chances of “deliberative agreement”. In this regard, consociational institutions
draw both from rationalism and constructivism and involve incentives, spaces, and scripts for
deliberative action; reflective institutions also draw from constructivism and rationalism and
involve scripts and spaces for deliberation and agreement; finally, veto institutions draw only
from rationalism and involve incentives for deliberative action. Our basic argument is that these
three institutional variables and their different mechanisms can work together to improve our
understanding when a philosophical construct such as “deliberative agreement” materializes in
real world politics. We test these hypotheses with 52 Swiss, US and German parliamentary
debates. These debates involve different types of issues and are conducted in different arenas
(public and non-public). To measure “deliberative agreement”, we build on the ”Discourse
Quality Index” (DQI) (Steenbergen et al. 2003) and focus on two of its dimensions: respect
toward counterarguments and respect toward demands. The DQI yields reliable data and allows
empirical researchers to peer into the deliberative dimensions of real world politics. Since we
are entering unexplored empirical territory, we cannot use random-sampling. Instead we will
rely upon a theoretical sampling whereby cases are selected purposefully so that we capture
important and comparable cases. While a theoretical sampling has the drawback of not
3
producing generalizable results, it will still permit us to obtain first important hints whether the
proposed institutional contexts indeed make a difference for “deliberative agreement”. To
analyze our data, we use a batch of techniques that match our problematic data structure.
Second, since our dependent variable is binary and heavily skewed - with “deliberative
agreement” representing a rare event - we check for the robustness of standard logistic
regression techniques by utilizing a number of specialized techniques: rare events logit and
gompit. Moreover, since our sample is finite we estimate all models with a non-parametric
bootstrapping procedure allowing us to obtain correct inferences under such conditions.
Making Philosophy Work in the Real World: Designing Institutions that Make Actors
Agree
How can philosophical concepts be integrated in positive political science? The major problem
of such an endeavour is that philosophical constructions and reconstructions occur at a very
high level of abstraction and that most philosophers have formulated philosophical ideals and
principals without considering what is possible in the real world. For many years, this
description also matched the state of art in the research on deliberation. As John Dryzek and
Braithwaite (2000: 242) put it: “the defenders and various critics of deliberation share one
belief: that matters can be resolved at the level of theoretical stipulation, with little recourse to
empirical evidence beyond illustrative anecdotes.” Certainly, one of the leading proponents of
deliberative politics, Jürgen Habermas, has directly confronted the relationship between his
social theory on the one hand and positive social science on the other:
[I]t is unclear how [my] procedural concept, so weighted with idealizations, can link up with
empirical investigations that conceive politics primarily as an arena of power processes…As I
understand it, this question does not imply an opposition between the ideal and the real, for the
normative content…is partially inscribed in the social facticity of observable political processes.
(Habermas, 1996, 287)
Thus, Habermas conceives of discourse theory as an abstraction from and refinement of broadly
empirical raw materials. But, as Michael Neblo (2005) holds, “even if deliberative theorists can
show that there is ultimately no alternative to discourse (because it is constitutive of our notion
of legitimacy)”, we would still require more concrete mechanisms how the theory can be
translated into workable institutions and practices for actually existing democracies.” Neblo’s
4
remark serves as a starting point for our attempt to reconstruct an empirical theory of
deliberation. One possible bridge between political philosophy and positive political science is
institutional. Institutions have always contained the hope that they can change actors’ behavior
in normatively attractive ways (see Ostrom 1998)2 . And as Bo Rothstein (1998) notes,
analyzing established institutions makes it possible for us to link the study of political
philosophy with an investigation of how “actually existing” political institutions operate.
However, contemporary “institutionalism” is not one thing but many (Goodin 1996: 2). Two
basic variants of institutionalism can be distinguished: one drawing from a rationalist approach
and one drawing from a constructivist or sociological approach. In order to preserve the
normative edge of deliberation, we propose that only a broad institutional conception including
both rationalist and constructivist components can bring us closer to the Habermasian ideal of
“deliberative agreement”.
Rationalism and Deliberation
Rationalism views institutions as features of a strategic context, imposing incentives (or
constraints) on actors which are treated as rational, goal-oriented and purposeful, maximizing
their utilities on the basis of given, fixed and ordered preferences. From a rationalist perspective,
deliberation could take place when institutional rules provide incentives for it. For instance, in a
situation of interdependent decision-making, deliberation might be necessary in order to
advance an argumentative consensus on the definition of the situation or for acquiring a
collective understanding of the underlying normative framework. Or strategic actors can
become argumentatively “entrapped” and thus be induced to switch to a deliberative mode of
decision-making (Johnson 1991; 1993; Keck 1995; Ulbert/Risse 2005). For instance, if
strategic action has badly failed to serve actors’ preferences – e.g., if bargaining does not result
in any acceptable result and cooperation is indispensable – then actors might be induced to
create an acceptable norm through communicative action that would then provide guidance for
2 Interestingly, recent developments in institutional theory seem to suggest otherwise. Two rational choice practitioners, Daniel Diermeier and Keith Krehbiel, claim that one should hold behavioural postulates fixed, since if one varies behavioural assumptions, comparative institutional analysis will fail to be cumulative and there will be no theories of institutions. Institutional scholars should specify from the outset with what behavioral postulate they operate: a rational choice logic with maximizing actors or a sociological logic with satisficing actors. What counts is that behaviour is kept fixed. Sure, at this stage of our knowledge about institutions there is merit to theoretical protectionism, taking one theory and seeing how far it can be pushed. But such a view on institutions and behaviour is deficient in several regards. On the one hand, it ignores the idea of classical institutionalists and philosophers that behavioral change can occur via institutions. On the other hand, it also neglects the basic insight of social choice theory that focusing on one behavioral assumption such as maximization can produce chaos under
5
the issue subject to bargaining. But also in situations of pure majoritarian aggregation –
misleadingly called the “institution-free” conditions - , deliberation might step in as a helper in
need. For instance, deliberation can induce preference structuration, thus facilitating the
disaggregation of a non-single peaked dimension into two or more single-peaked dimensions
and making cycles in decision-making less of a threat (Dryzek/List 2003).
While a rationalist approach is in a pivotal position to examine institutional situations where
actors use deliberation (or elements of it) in a goal-oriented and strategic way, the problem is
that rational choice theory ultimately cripples other forms of behavior than goal-oriented and
strategic behavior and thus cannot exploit the full spectrum of deliberative action with its stress
on reflection and transformation (see Müller 1985). Furthermore, in a rational choice
framework deliberation represents more something of a “helper in need” for coping with
exceptional situations and should hardly inform whole decision processes. As Christian List
(forthcoming) – a leading social choice theorist - holds, agreement at a substantive level is not
only empirically unrealistic, but also logically unnecessary for meaningful collective
decision-making. Thus, if deliberation produces agreement at a meta-level, this may be “good
enough” deliberation from a social choice perspective.
However, there is more to deliberation and agreement from a rational choice perspective.
Applying a domain of application approach to rational and strategic action, we can identify
specific contexts where deliberative action has better chances – and contexts where this is less
the case. Our claim is that when stakes are high and defections from expected standards costly,
actors will follow a course of action that is highly compatible with the predictions of rational
choice theory. But if stakes are lower and defections from expected standards less costly, actors
have more space to behave differently: they may remain strategic but they can also become
truly discursive actors and genuinely search for common understanding (see
Aspingwall/Schneider 2000: 27; March/Olsen 1998: 952-953). This is also backed by
experimental research. Thin models of rational choice have proved to be powerful engines of
predictions for competitive situations in which selective pressures screen out those who do not
maximize external values, such as profits in competitive market. But they have been far less
successful in explaining or predicting behavior in situations where such pressures are not
present (for a concise overview see Ostrom 1998).
specific institutional rules (such as majoritarian aggregation). If this is true, then real-world actors might be well
6
Constructivism and Deliberation
Institutions do not only structure incentives, they also have constitutive and cognitive
dimensions (Powell/DiMaggio; see also Finnemore 1996: 326). In the constructivist or
sociological view, social structure is ontologically prior to and generative of agents, and human
action is more context-driven than goal-driven. A sociological approach to institutions builds
on two key concepts: a “logic of appropriateness” (March/Olson 1989; 1998) according to
which actors are guided by collective understandings of what constitutes proper, i.e. socially
accepted behavior in a given rule structure. On the other hand, it builds on ”institutional
isomorphism” suggesting that actors who frequently interact, who are exposed to each other or
are located in a similar environment, develop similarities over time in practices and meaning
structures (DiMaggio and Powell 1991; Scott/Meyer 1994). From the sociological perspective,
deliberation (or strategic behavior) would be inscribed in the very structure of institutions. For
instance, if the institutional goal itself would be deliberation, then actors operating within this
institution would learn and internalize the concept, turning deliberation into appropriate
behavior and making it part of actors’ scripts and identities. Harald Müller (2004) has presented
such an argument and claims that arguing (and bargaining) are fully legitimate and sanctioned
modes of behavior under the right circumstances. For instance, it is normatively acceptable and
even prescribed to bargain in international trade negotiations; in negotiations, it is appropriate
for actors to pursue their self-interest unless it collides with a valid norm that prescribes
different behavior such as arguing. That such deliberative norms or guidelines can make a
difference for deliberation is also backed by experimental research. Shawn Rosenberg (2004)
has tried to model the ideal discourse condition in a field experiment about public education. To
achieve this goal, he had to provide guidelines to participants by reiterating the goals of
deliberation, namely that deliberation means civility, polite discussion, reasoned argument,
considerations of basic justice, and productive conclusions. Rosenberg’s experiments underline
that one cannot simply bring people together and then expect that they would automatically
deliberate in a Habermasian fashion. Rather, the context needs to be structured in a
“constructivist” fashion in that people are provided with guidelines that emphasize reflection
and agreement.
While the constructivist or sociological approach has the advantage to unravel the social
underpinnings of deliberative action, the major problem with the sociological approach is – as
many have stressed - that it frequently underplays the role of agency. If institutions determine
advised to change their action logic under conditions if stable outcomes are to be achieved (see below).
7
preferences, then how can one explain why agents acting in the same type of institutions
sometimes hold different preferences? This is aggravated by the distaste of sociological
institutionalists for calculation which most empirical researchers would not lightly dismiss.
While the economic approach may represent an undersocialized view of how agents’ establish
preferences, the cultural view represents an over-socialized one (see Granovetter 1985). In
other words, while rationalism in itself may be too weak to theoretically capture the
philosophical ideal, a sociological and constructivist argument is too strong and ignores that
there may be strong counterincentives that work against a normative script.
However, any theory of institutional design which tries to integrate rationalist and
constructivist approaches confronts a meta-theoretical caveat. As to the integration of
rationalist and constructivist approaches, some argue that the individualist perspective of
rational choice institutionalism and the societal perspective of constructivism can only be
integrated by adopting a comprehensive theory of action that builds on a superior logic since
theories based on contradictory ontologies would be incoherent (Müller 2004). Others argue
that the metatheoretical debate is over and argue for a more pragmatic approach that sees the
two approaches as commensurable while building on a careful theoretical integration of
rationalism and constructivism combined with empirical testing (Jupille et al. 2003). We opt for
the second, pragmatist strategy. As described above, any subsumption approach – be it
sociological or rationalist -, has explanatory shortcomings and empirical blindspots that
preclude its straightforward use as a superior logic of action.
While many pragmatic integration attempts utilize a sequencing strategy that views rationalism
and constructivism as guiding logics in different phases of the political process (March/Olsen
1998; Börzel/Risse 2000), we adopt a combination strategy that views the different action
logics as additive components. Such a combination model assumes that actors are both sensitive
to costs and benefits and can be simultaneously influenced by guidelines or scripts. Our major
argument is that in order to achieve “deliberative agreement”, actors need the right incentives,
the right scripts, and an action space. However, we are not fully agnostic to any hierarchy of the
various metatheoretical approaches. We always start from a rationalist foundation and add
sociological and constructivist complications only where necessary. In other words, when the
rationalist approach is fully sufficient to explain a certain type of event (in particular, the
non-occurrence of “deliberative agreement”), we shall not consider sociological or
constructivist explanations anymore.
8
But what type of concrete institutions enhance the chances of “deliberative agreement”?
Surprisingly, the question of favorable (and unfavorable) institutional contexts for deliberative
action is a topic which has been greatly neglected in the literature, both by institutionalists and
deliberative theorists. There are only few and sporadic hints at favorable institutional settings
(e.g. Scharpf 1997: 165-66; Risse 2001: 20ff.). We propose that three institutional contexts
further “deliberative agreement”: consociational institutions, veto institutions, and reflective
institutions.
Consociational institutions
At the cradle of consociationalism lies deliberation. The pre-modern German philosopher
Johannes Althusius (1563-1638) developed a political theory revolving around the covenant or
“consociatio”. The covenant was an arrangement in which people would live together in mutual
benevolence, striving for consensus by way of open communication. Althusius became
forgotten after the French Revolution, and only in the 1960’s was the idea of “consociatio”
readopted by David Apter and Arend Lijphart and built in a new empirical type of democracy:
consociational democracy, and its later variant, consensus democracy (see Andeweg 2000).
The “deliberative” conception of the Althusian convenant is not alien to the consociational or
consensus model of democracy (see Steiner 1974). As Lijphart (1999: 294) notes, there is an
intimate connection between his consensus democracy and Jane Mansbridge’s “unitary
democracy” (Mansbridge 1980). In unitary democracy, people come together with their friends
to find agreement; this democracy is consensual, based on common interest equal respect, and
persuasion. Mansbridge contrasts this with an adversary ideal which has dominated Western
democratic thinking since the 17th century. In adversary democracy, there is no common good
and the mechanical aggregation of conflicting selfish desires is the very core of an adversary
system. Interestingly, the adversary vision has become so dominant in positive political science
that most scholars tend to equate the consociational or consensus model of democracy with a
mere bargaining model of politics (e.g. Colomer 2001: 193; Lehmbruch 1967). While we do
certainly not deny that there is a lot of bargaining in consociational democracies, we think that
this is a limited understanding of the actual functioning of consociational institutions. As we
shall argue below, consociational institutions also involve a deliberative dimension, whereby
we identify both a rationalist and a sociological path to deliberation and “deliberative
agreement”.
9
From a rationalist perspective, one mechanism of consociational institutions may be
particularly important for the production of “deliberative agreement”: the existence of grand
coalitions. Grand coalitions generate, first of all, a logic of joint decision-making where
argumentative rationality might be necessary for the “problem-solving” or “production” part of
policy negotiations in order to successfully cope with different situation definitions, cognitive
problems, factual disagreements, as well as standards of appropriateness, fairness and justice
(Risse, 2000; Holzinger, 2001). Moreover, a coalition arrangement creates collective agency
and shared responsibility with parties being judged at least in part by their government record
(Goodin 1996: 331). Coupled with the motive to maintain the stability of the coalition, this
provides incentives for actors to adopt a cooperative attitude vis-à-vis their coalition partners
which is a precondition for deliberation to develop.
Two other mechanisms enhance further the deliberative potential in consensus systems. First,
grand coalitions obscure policy making clarity and this reduces the possibility that coalition
parties will use political success to compete with other members of the coalition (Powell and
Whitten 1993). Second, when grand coalition arrangements are made permanent and parties
obtain relatively secure power positions (e.g. by obtaining a fixed seat in the government) the
importance of electoral competition among coalition partners is reduced. These factors
facilitate the development of a cooperative interaction orientation (Scharpf 1997) and therefore
contribute to the creation of a “deliberative space”.
The situation in competitive systems with a majority-minority relationship is quite different.
Here, not only can the majority ignore the arguments of the minority, the fact that the stakes are
losing or winning power positions is conducive to competitive interaction orientations among
the parties. Moreover, a minority which is “cooperative” must fear that political successes are
assigned to the majority actors (Fiorina 1981; Lewis-Beck 1988). Therefore, common problem
solving, search for genuine consensus, and “deliberative agreement” may constitute a loss in
the overall competitive game (Scharpf 1997: 166), thus reducing the incentive for deliberation.
In other words, while consociational institutions are stake-reducing and allow for less
calculative and strategic policy-making that makes “deliberative agreement” more likely,
competitive institutions are stake-enhancing and strongly conducive to strategic and adversarial
behaviour that tends to undermine “deliberative agreement”.
10
However, the rationalist path to deliberative agreement in consociational institutions is a
relatively minimal one. It basically stipulates a “deliberative space” and some incentives, but
neglects a concrete script for “deliberative agreement”. Therefore, we complement the
rationalist path with a constructivist and sociological path.
The constructivist path draws from classical consociationalism with its stress on the willingness
of actors and on respective scripts. According to Lijphart, consociational actors should also
have the willingness to transcend the borders of their own group, be receptive to the claims of
others, and accommodate the divergent claims and demands of the segments (Lijphart 1969:
216). In our view, this willingness can be encoded in organizational scripts stressing mutual
understanding and conciliation (see Bächtiger/Steiner 2003). If scripted this way, then actors
operating in consociational institutions should be more open to “deliberative agreement”.
It is important to point that these predictions refer to “ideal realizations” of consociational
institutions; they may not always fully hold. In particular, it would be misleading to think that
grand coalitions prevent competitive interaction orientations. Not only must coalition partners
face their constituents on their own account, parties might also have an interest in strengthening
their positions within a coalition arrangement. This can set an important constraint on parties’
cooperative and deliberative capabilities. As such, “deliberative agreement” will remain a “rare
event” even in the context of consociational institutions. Our only claim is that the respective
chances are higher than in competitive settings.
Veto institutions
Veto institutions are defined as “structural incentives providing windows of influence” (Kaiser
1997: 437); one can think of them as places in the decision-making process where the consent
of one or more players is required to leave the status quo. As an explanatory construct, veto
institutions are not independent from the distinction between consociational and competitive
institutions discussed previously. Indeed, one way to conceptualize consociational democracies
is in terms of the presence of numerous veto players. By contrast, competitive systems are
characterized by far fewer veto players. The correlation between veto institutions and
consociational/competitive institutions is not perfect, however. In a grand coalition, not all
parties may be always essential for the passage of legislation or the survival of the coalition,
which severely limits the credibility of any veto threats by those parties (see Strom 2000: 280)
In competitive systems, certain types of legislation may require super-majorities, which may
11
give opposition parties a veto position. Therefore, it is necessary to expand the veto argument
and generalize it to all decision-situations.
In the philosophical discussion, veto institutions have served as a potential bridge between
constitutional contractarianism and Habemasian discourse ethics. For the economist James
Buchanan, the right of veto at the constitutional level is crucial since it protects the autonomy of
the individual. If a solution is not in the interest of an individual, she can say no and forego
agreement. The primary agreement form is compromise, but there is an opening toward with
Habermas’s consensus-based discourse notion of agreement. According to Vanberg and
Buchanan (1989), preferences have not only an “interest-component” but also a
“theory-component”. While the former may be unaffected by deliberation, the latter can be
affected when actors seek and share information about the likely effects of policies. Hence, the
outcomes of constitutional debates may not only represent a mere aggregation of interests, but
can also entail communicative or “deliberative” forms of agreement.
Similar to Vanberg and Buchanan, we hypothesize that veto institutions form an “enabling
constraint” for deliberative agreement. As already mentioned in the context of consociational
institutions, veto power generates a logic of interdependent or joint decision making where
“deliberative agreement” might be a result of the “problem-solving” or ”production” part of
policy negotiations. As negotiation theorists stress, actors might have incentives to find
mutually agreeable solutions on such fundamentals in order to make subsequent bargaining
efficient and to secure the overall success of negotiations. When veto institutions are absent,
there is no need for participants to achieve deliberative agreement on fundamentals; the
majority can forego serious arguing and quickly proceed with voting. Thus, veto institutions
involve a rationalist, yet functional path to the philosophical ideal of deliberative agreement.
But when controlling for veto power, will there be any differences in the frequency of
“deliberative agreement” between consociational institutions and competitive institutions?
Following George Tsebelis’s (2002) veto player theory, partisan veto players in consociational
institutions are functionally equivalent to institutional veto players in competitive institutions.
Consequently, the frequency of deliberative agreement should vary according to the presence
or absence of veto players, but not between consociational and competitive institutions.
However, we propose that the two types of veto players differ in terms of deliberative capacity.
While in a consociational setting, actors have incentives in finding a common solution (see
Bolleyer 2004), this is not true in a competitive setting. Here, actors have a prime incentive to
12
attack the political opponent and block policies (as the resulting policy successes will provide
the majority party with a greater benefit). Certainly, there may be occasions where the majority
and the minority have strong pressures to find agreement (for instance, when the majority and
minority face a broad public consensus on an issue that effectively precludes competition on
that issue), and consequently, we expect the functional dimensions of veto power to set in for
the production of agreement. But the absence of collective agency does not provide incentives
for problem-solving; rather we expect it to further tough bargaining. This should decrease the
frequency of deliberative agreement, and consequently, controlling for veto power should not
wipe out the differences in deliberative agreement between consociational institutions and
competitive institutions.
The veto argument can be contested, however. First, and very basically, Avio (1997) holds that
there is no way to blend contractarianism and discourse ethics in a manner faithful to the root
theories. In his view, the strategizing bent of actors in contractarianism cannot be reconciled
with “social” and “transformative” bent of discourse ethics. Second, David Austen-Smith and
Timothy Feddersen (2002) challenge the veto argument fundamentally and claim that it is
actually majoritarian voting rules – and not veto institutions - that promote deliberation.
Although unanimity rule creates incentives for supporters of the status quo to reveal
information, it likewise creates incentives for others to conceal information favoring that status
quo. This in turn generates an externality rendering information from all members of the
committee suspect. Consequently, the deliberative process tends to break down under
unanimity rule.
These are important objections that solicit a response. First, whether constitutional
contractarianism and discourse ethics are irreconcilable is a question that needs to be
investigated empirically, given the fact that negotiation theorists have presented a plausible
functional argument in favour of veto power. Second, the problem with Austen-Smith and
Feddersen’s argument is that it involves at least one assumption which is highly problematic.
They assume that actors value private interests as much as the common good. This assumption,
however, is frequently spoiled in real world settings: in many cases, actors have clear
preferences over different demands and the relative weight they put on private or constituent
interests. Under majoritarian decision making, this may create a serious motivational problem
for deliberation, as actors may simply not be willing to listen to minority demands or
13
counterarguments. Put differently, only veto institutions guarantee that other actors are
“forced” to listen and deliberate.
Reflective institutions
From the time of the ancient Greeks onward, the idea of a council of elders, who could promote
an element of “second” and longer-term thought, has been advocated by philosophers and
statesmen (Aristotle, Cicero, John Stuart Mill, James Madison). We shall call this type of
institution “reflective institutions”. Reflective institutions put a prime on reflection and
productive conclusions. Drawing from constructivism, we assume that actors operating in such
a “reflective” institutional context learn and internalize the concept of deliberation, turning
“deliberative agreement” into a form of appropriate behaviour that can become part of actors’
cognitive scripts and identities. This is enhanced by specific selection criteria: reflective
institutions are composed of actors who are older, have more political experience and are also
more moderate. Besides these sociological and constructivist components, reflective
institutions also involve a “space” component: longer terms of office and less electoral
pressures, which lowers stakes and creates an additional space for deliberation.
Similar to veto institutions, reflective institutions are nested within consociational and
competitive institutions. In the context of a consociational setting, we expect reflective
institutions to strengthen the deliberative assets of the former: not only do they enlarge the
deliberative space even further, the specific organizational guidelines also reinforce the scripts
of “mutual understanding”. In the context of competitive institutions, however, we expect
reflective institutions to enhance deliberative assets as well, but there will be important
crosspressures. Actors nested in competitive institutions always have strong incentives to
forego deliberative agreement. As with veto institutions, we do not expect reflective institutions
to equalize differences between consociational and competitive institutions.
Methodology
14
Dependent Variable: Deliberative Agreement
“Deliberative agreement” is a binary variable that tries to get hold of the Habermasian ideal of
argumentative change and consensus.3 To measure “deliberative agreement” we rely on the
discourse quality index (DQI) developed by Steenbergen et al. (2003). The DQI represents an
attempt to overcome the dearth of measurement instruments in the research on deliberation. It is
based upon broadly shared concepts of the theoretical literature on deliberative democracy and
allows for a quantitative content analysis at the level of speech acts of recorded debates. To
capture “deliberative agreement”, we focus on two indicators of the DQI: respect toward
counterarguments and respect toward demands. Thereby, we concentrate only on their highest
categories: agreeing with counterarguments and agreeing with demands or proposals (or not
doing so). Agreeing with counterarguments measures substantive agreement with arguments
put forward by actors with different opinions and qualifies as a “clean” indicator of the
Habermasian notion of consensus and agreement (see Goodin 2005). However, for assessing
the deliberative dimensions of real-world debates, this may be too strict a test. First, as James
Johnson (1999: 174) holds, a plausible argument for deliberation cannot be utopian in the sense
that it makes heroic assumptions about the deliberative capacities of actors and assumes that the
latter will massively change their opinions in normatively attractive ways. Following Johnson,
it seems sensible to include meta-agreement, productive conclusions as well as compromises
that rely on greater mutual understanding and respect for diverging standpoints
(Knight/Johnson 1994: 285). However, agreeing with demands involves a problematic aspect,
since we may confound the Habermasian conception of agreement with a bargaining
conception of agreement. Thus, mere acceptance of a demand is not counted as “deliberative
agreement”. To obtain this qualification, actors have to make a statement that they came to
value or got persuaded by a demand submitted by other participants. In the empirical analysis,
we shall consider two indicators of “deliberative agreement”: a combined “agreement with
demands/counterarguments” indicator and “agreement with counterarguments” indicator. The
first indicator counts instances where actors either agree with demands and/or with
counterarguments and represents the more lenient test for “deliberative agreement”. At the
same time, it is also provides an indirect test for the rationalist path to “deliberative agreement”.
If institutions only comprise an incentive-based mechanism for deliberative agreement – as is
3 ”Deliberative agreement” closely resembles Thomas Scanlon’s conception of ”reasonable agreement”. However, Scanlon’s conception places intersubjective justification at the center, while our prime interest is not the act of reason-giving but the act of ”agreement”. Put differently, we imply that actors have already deliberated on the merits of arguments and demands when they agree with them. In order not to confuse this kind of agreement with Scanlon’s conception of ”reasonable agreement”, we have coined the term ”deliberative agreement”.
15
the case with veto institutions – then we would expect only this indicator to make a difference.
However, this does not mean that institutions involving scripts and spaces – such as reflective
institutions would have no effect on this indicator. To the contrary, we expect these
mechanisms to affect the frequency of agreement with demands/counterarguments as well. The
second indicator counts instances where actors agree with counterarguments and represents the
stricter test for “deliberative agreement”. It provides an indirect test for the non-rationalist and
constructivist path to “deliberative agreement”. We expect that only institutions that involve
spaces and scripts as well – i.e. consociational and reflective institutions – have an effect on this
indicator, while incentive-based institutions – i.e. veto institutions – should not produce any
effects.4
The DQI has proved to be a reliable measurement instrument of deliberation. In a series of tests,
we could demonstrate that the “inter-coder reliability” of the respect category is very good (see
Steenbergen et al. 2003). Two different coders scored a subset of the debates analysed in this
study. The rate of coder agreement was 0.919 for a combined respect category that comprised
respect toward demands and counterarguments. Because of the possibility of inter-coder
agreement by chance we also computed Cohen’s kappa, which controls for this. Kappa yields
0.881 for the combined respect category. These values indicate a very high inter-coder
reliability.
Independent Variables: Selection of Cases
We will test for the proposed institutional factors with real-world debates in legislatures. But
why choosing legislatures – and not other, more benign organizational contexts? As James
Fishkin and Robert Luskin (2005) hold, “much of the discussion among political elites is
posturing or negotiation rather than deliberation. … Legislators are representatives, after all,
and elected to support or oppose certain things (and because they already do so). They are also
constrained by their parties. So even elite-level politics is not necessarily that deliberative.”
4 A further conceptual note is in order. The common and colloquial understanding of Habermasian notions of agreement is that actors agree with arguments (or demands) directly within discussion. But as Robert Goodin (2003) convincingly argues, argumentative change can also mean ”deliberation within”. ”Deliberation within” consists in the weighing of arguments in private, within the head of each individual. So, we may find ”agreement” that is not directly interactive, as actors come to agree with arguments and demands in private and not during actual discussion. Our conception of ”deliberative agreement” includes both types of agreement, interactive and non-interactive ones.
16
While we agree that deliberation and ”deliberative agreement” are necessarily constrained in
representative institutions, we still think that genuine and consequential deliberation is possible
in legislatures. First, and in line with the new institutionally-oriented research program on
legislatures (see Döring 1995), we propose that the occurrence of deliberative agreement is
largely dependent on favorable institutional contexts. Hence, generalizing from the competitive
United States Congress – as Fishkin and Luskin seem to do – can lead to a biased evaluation of
the deliberative capacity of legislatures. Second, from an institutional point of view,
parliaments offer a unique opportunity to directly compare the impact of different institutional
factors on actors’ behavior while holding the basic organizational setting – legislatures –
constant. And third, from a research-practical point of view, parliamentary debates are
protocoled and accessible, even in the non-public sphere (which, for example, is mostly not true
for interactions within the government).
The foundation of our research design is the idea of typological or theoretical sampling (e.g.
Lamnek 1995; McKeown 1999). But why choosing this relatively unconventional design and
not drawing a random sample of legislatures and debates allowing us to produce generalizable
findings? The reason for refraining from random-sampling is twofold. On the one hand, there are
no lists of parliamentary debates that permit to draw a random sample. Of course, one could
select debates from listings of bills, but that sampling frame is insufficiently detailed because a
bill can trigger multiple debates. On the other hand, the goal was to select debates that matter. A
random sample, however, might have yielded a selection of debates on relatively minute issues.
That might have been correct from a statistical viewpoint, but it would have been of little
theoretical interest. Despite the fact that a theoretical sampling does not yield generalizable
findings, we think that it can serve an important purpose. The goal of a theoretical sampling is to
generate a set of “critical cases” (Eckstein 1975), i.e. cases for which we would expect the theory
to hold true, if it is going to hold true at all. If we were to find that the theory does not hold true
for those cases, this would cast serious doubt on the this qualification, actors have to make a
statement that they came to value or got persuaded theory, regardless of the fact that the cases do
not constitute a random sample.
For a legislature with consociational institutions, Switzerland represents one of the best
empirical approximations (see Lijphart 1999). It has an institutionalized grand coalition which
structures party relationships in a less competitive fashion. Of course, this does not exclude
competitive relationships among parties, but it allows for other types of interaction orientations
and sometimes even for truly cooperative ones. Moreover, the constructivist path to
17
deliberation is also clearly present, with some actors stressing “mutual understanding” among
coalition partners. Switzerland also allows testing for decision situations with and without veto
power of the coalition partners. The distinction between a veto and a non-veto situation is made
according to the presence or absence of a referendum threat of one of the coalition partners.
For legislatures with competitive institutions, we choose the United States and Germany. The
reason for choosing two legislatures is that in the context of competitive legislatures,
presidentialism and lower party discipline can positively affect the quality of discourse (see
Bächtiger/Steenbergen 2004).5 Under presidentialism and lower party discipline, legislators
have more leeway to transcend party boundaries and can more open to argumentative change.
This creates an additional deliberative space for which we need to control. Both the United
States and Germany also allow comparing for the effects of minority veto. In the United States,
it is quite difficult to clearly distinguish between veto and non-veto situations, as low party
discipline frequently leads to negotiations among different factions of MPs as well as between
the two chambers. Hence, we focus on decision situations where there were direct and credible
veto threats (such as a presidential veto or a filibuster threat) as well as on decision situations
where it was ”clear” to the actors that a bill would pass with a simple majority. In Germany, a
veto situation arises when the opposition possesses a majority in the second chamber, the
Bundesrat, and the consent of the latter is required to pass a bill. Other bills can be generally
decided with the simple majority of the governing parties. Finally, we also use data from the
German Conference Committee (Vermittlungsausschuss), a body that is charged with
reconciling conflicts between the two parliamentary chambers. The reason for including the
German Conference Committee is the following: in the United States Congress, committee
debates are generally held in public and also have very few deliberative elements - neither
hearings nor legislative mark-ups involve sustained arguing among legislators; this confines
analysis to floor debates only. In order to explore the important constellation of a competitive
setting with weak party discipline behind closed doors - in fact, representing a standard
negotiation setting – the German Conference Committee is an extremely helpful addition: it
represents a non-public body located in a competitive system with its members partially
relieved from hierarchical directives with the expectation to develop innovative approaches and
overcome blockades to overcome legislative stalemates (Spörndli 2004).
5 We do not have consider a parliamentary consensus system. The reason is that the management of disagreements—the essence of Habermasian discourse ethics—is usually handled outside of the legislature, in the executive, in special committees, or in intra-coalition summits (Tsebelis 2002: 93, 109ff.).
18
Our three basic contexts also allow testing for the effects of reflective institutions. In the realm
of legislatures, second chambers can be considered relevant proxies. Not only does the
literature conceive of them as institutions that promote an element of “second” and longer-term
thought, also senators tend to view them as such. In the US Senate debate on partial birth
abortion (1995) Clairborne Pell (D-Rhodes Island) made the following remark: “The Senate has
a long and established tradition of careful deliberation ... for legislating such difficult issues
with thorough and adequate review.”6 Similar statements are also uttered by Swiss and German
senators.
The three proposed factors are not the only variables that raise the chances of deliberative
agreement in the realm of legislatures. While we have already mentioned party discipline as a
control variable, we use four additional controls: (1) arena, (2) issues, (3) roles, and (4) gender.
Non-public Arena. Non-public legislative bodies such as parliamentary committees
allow legislators to deliberate without external interference, lower the pressures of following
constituent demands, and make it easier for politicians to reflect, to show respect for the claims
of others, or even to change their opinions (see Checkel 1999). Conversely, in full sessions of
parliament, one can expect that most actors will primarily engage in ritualistic rhetoric and
strategic arguing. However, as Elster (1998) noted in his study on constituent assemblies, while
actors talked much more seriously behind closed doors, the non-public character
simultaneously enhanced bargaining. This could also reduce the frequency of “deliberative
agreement”.
Issue Polarization. Hereby, we distinguish between polarized and less polarized issues.
Less polarized issues mean that actors are ideologically closer, i.e. they share more values and
goals, while in polarized issues, ideological distances are larger. If an issue is not very polarized,
then we should expect more cooperative interaction orientations among the political actors.
This should promote more consensual decision making and, as such, should lead to a higher
frequency of “deliberative agreement”.
Roles. We also control for the presence of specific parliamentary roles. First, we expect
members of government – especially in consensus settings – to play a mediating role in
legislative interactions. This in, turn, should enhance the frequency of “deliberative agreement”.
In addition, we control for chairmen and experts. We expect these two type of roles to decrease
6 Partial Birth Abortion Debate, November 8, 1995, S16790
19
the frequency of “deliberative agreement” (but increase the frequency of neutral speeches).
While the major role of chairmen is to ensure that debates are conducted smoothly, the role of
experts consists of presenting legislators with information and facts. These role expectations
prime on implicit respect, but not on argumentative change and agreement.
Gender. Following feminist conceptions of democracy, we expect female legislators to
be more open to deliberative agreement. As Jane Mansbridge (1996: 123) has emphasized,
feminist politics involve “connectedness” and “mutual persuasion” instead of self-interest and
power politics.
As to the debates, we selected 52 debates, mostly from the late 1980s and 1990s,7 from the three
legislatures described before. The test is based on a total of 52 debates in these three legislatures,
involving polarized and less polarized issues in public parliamentary sessions and in non-public
parliamentary committees.
As our dependent variable, all the predictors in our analysis are measured as a dichotomy: either
a debate possesses the characteristic or it does not. While one could think of more precise
measures, using dichotomies has the advantage of casting the influence of characteristics in the
starkest possible manner. More precisely, Swiss debates score one on consociational
institutions, while debates in Germany and the United States score zero on this predictor. High
veto power is a predictor that takes on the value of one if there were strong veto players in a
debate and zero otherwise. In Switzerland, the distinction is made according to the presence or
absence of a referendum threat by one of the members of the grand coalition. In the United
States, we focus on situations where there was a direct and credible veto threat such as a
presidential veto or a filibuster. In Germany, a veto situation arises when the opposition has a
majority in the Bundesrat and its consent is required to pass legislation. As to reflective
institutions, debates in the Swiss Ständerat, the American Senate, the German Bundesrat, and
the German Conference Committee receive a score of one, whereas debates in the Swiss
Nationalrat, U.S. House of Representatives, and the German Bundestag receive a zero on this
predictor. Regarding our control variables, we first consider low party discipline. Initial coding
of this predictor was done at the systemic level, with debates in non-parliamentary/presidential
settings (i.e. Switzerland and the U.S.) receiving a score of one on this predictor and debates in
a parliamentary setting (i.e. Germany) receiving a score of zero. We then modified this coding
7 The exceptions to this time frame are the debates in the German Conference Committee which are drawn from the 1970s.
20
by going into the debates and considering their specifics. We found that party discipline had
been relaxed in some of the German debates, namely those on the abortion issue and all of the
debates in the Conference Committee. Those debates were therefore coded as a one.8 Our
second control variable is arena. In this regard, German and Swiss committee debates, as well
as debates in the German Conference Committee, are coded one on nonpublic arena, whereas
plenary debates are coded zero.9 Our measure of issue polarization draws heavily from Kriesi’s
(2001) analysis of German policy issues. In this analysis, socioeconomic issues and abortion
are considered highly polarized, whereas issues such as rights of the disabled, animal welfare,
and crime prevention are less polarized.10 We group the latter set of issues under the rubric of
low issue polarization. Finally, our actor-centered control variables are coded as follows:
female legislators are coded one, whereas male legislators receive a zero government members,
committee chairs and experts were coded as one, while ordinary legislators are coded as zero.
Statistical Methods
As mentioned before, deliberative agreement is a dichotomous dependent variable. The most
common approach of estimating equations with such dependent variables is to use logistic
regression (logit or probit). We shall use a logit as our baseline estimation. However,
philosophical concepts such as deliberative agreement tend to materialize as “rare events” in
the real world. When looking at the frequency distributions of our two dependent variables –
agreement with demands/counterarguments and agreement with counterarguments, the former
occurs in 7.8 percent of all observations, while the latter occurs only in 5.3 percent of all
observations. In such situations, standard logit and probit techniques can produce biased
coefficients that can lead researchers to under- or overestimate their size and significance (see
King/Zeng 2001). To overcome this problem, Gary King and Langche Zeng have developed a
procedure called “rare events logit”. As a further robustness check, we will use a gompit model.
Gompit (also called complementary log-log) is the inverse of the Gompertz distribution
function and frequently used when the probability of an event is very small. Another estimation
8 While members of the Conference Committee are partially relieved of partisan directives and have a lot of leeway, some might argue that it is too generous to declare party discipline low in this committee. We reran our analyses considering party discipline to be high and the results are consistent with those reported here. 9 We exclude the United States from the non-public versus public analysis because committee meetings are usually public and there are few other forums in which legislators deliberate behind closed doors. 10 We double-checked Kriesi’s (2001) classification using Party Manifesto data (Budge et al. 2001). In practically all cases, political parties took significantly different positions in their manifestos on the issues that Kriesi considered polarized, whereas positional differences were much smaller when issue polarization was low. One exception was crime, which Kriesi identified as polarized but we consider at a low level of polarization, based in
21
technique for skewed distributions of the dependent variable would be the skewed logit, or
scobit, model (Nagler 1994). Scobit relaxes the assumption inherent in probit and logit that the
maximum effect of the covariates is fixed at a probability of .5. Furthermore, rather than
assuming that positive responses are rare events, as the gompit model does, scobit allows the
direction and skewness in the sigmoid function to be estimated from the data. This is a flexible
function, but applied researchers may not find it practical for two reasons (see
Buckley/Westerland 2004). First, the scobit model can be difficult to estimate computationally.
In our case, the ancillary parameter produced heavy convergence problems. Second, since
“deliberative agreement” classifies as a classical rare event, the gompit function provides
sufficient flexibility for our purposes. Besides skewness, there is a further problem to tackle. In
the context of our finite sample, the variance of the coefficient is unknown. Instead of drawing
inferences on the basis of a theoretical (and asymptotic) assumption, we rely upon the empirical
distribution. Bootstrapping allow us to do that. Second, bootstrapping also enables us to get rid
of an autocorrelation problem in our data. Not only are actors nested within specific debates, the
same actors can also speak up repeatedly. This violates the independency requirement in
statistical analysis, which again can lead to erroneously estimated standard errors. Replication
via bootstrapping allows us to address these problems.11 Hence, we will apply a non-parametric
bootstrapping procedure with 1,000 replications to all models – logit, rare events logit, and
gompit.12
A final comment is in order. Some might argue that the lack of a probability sample prohibits
any form of statistical analysis. We disagree since our goal is not to generalize from the
statistical results, which would be problematic based on a typological sample, but rather to
discover if the effects of predictors are reliable. Our predictors divide the legislatures and
debates into different groups; our goal is to determine if the variance across those groups is
sufficiently large compared to the within-group variance to consider the effects of the
predictors meaningful.
part on the Manifesto data. Inclusion or exclusion of this issue in the low polarization category does not affect the results. 11 We have rerun all models by using cluster-corrected standard errors at the level of debates and speakers (results available upon request). These analyses produce very similar results to those reported here. 12 In addition to these multivariate analyses, we also conducted a series of tests of each predictor by itself (results available upon request). Those tests have the drawback of providing less statistical control. On the other hand, they allowed us greater control in selecting debates that were truly comparable. The results from these analyses sketch a picture that is very similar to that presented here.
22
Results
Is there evidence that our three institutional variables enhance the frequency of “deliberative
agreements”? First, let us look at a batch of bivariate tests (tables 1-6). The bivariate
relationships indicate that our three institutional contexts further “deliberative agreement”.
Consociational and reflective institutions produce statistically significant differences both for
the lenient version of “deliberative agreement”, agreement with demands/counterarguments,
and for the strict version thereof, agreement with counterarguments. Veto institutions, in turn,
only produce significant differences in the context of the lenient indicator, but no statistically
reliable differences can be found in the context of the strict version. This corroborates our
theoretical expectation that institutions involving purely incentive-based mechanisms – such as
veto institutions - are not sufficient to produce the philosophical ideal in its “pure form”. Let us
now check whether these findings also hold in the multivariate model.
Table 1:Agreement with Demands/ Counterarguments in Consociational
and Competitive Institutions
Table 2:Agreement with Demands/
Counterarguments in Veto and
Non-Veto Situations
Table 3:Agreement with Demands/
Counterarguments in Reflective
And Non-Reflective Institutions
Comp.
Inst.
Consoc.
Inst.
Non-
Veto
Veto
Non-
Refl.
Refl.
Inst
Not
Agree
2427
96.2%
1597
86.6%
4024
92.2%
Not
Agree
454
95.6%
3570
91.8%
4024
92.2%
Not
Agree
2620
93.9%
1404
89.0%
4024
92.2%
Agree 96
3.8%
246
13.4%
342
7.8%
Agree 21
4.4%
321
8.2%
342
7.8%
Agree 169
6.1%
173
11.0%
342
7.8%
Total 2523
100%
1843
100%
4366
100%
Total 475
100%
3891
100%
4366
100%
2789
100%
1577
100%
4366
100%
Chi2 = 134.337, Pr. = .000 Chi2 = 8.596, Pr. = .003 Chi2 = 33.648, Pr. = .000
Table 4:Agreement with Counterarguments in Consociational
and Competitive Institutions
Table 5:Agreement with
Counterarguments in Veto and
Non-Veto Situations
Table 6:Agreement with
Counterarguments in Veto and
Non-Veto Situations
23
Comp.
Inst.
Consoc.
Inst.
Non-
Veto
Veto
Non-
Refl.
Refl.
Inst.
Not
Agree
2444
96.9%
1690
91.7%
4024
94.7%
Not
Agree
458
96.4%
3676
94.5%
4024
94.7%
Not
Agree
2664
95.%
1470
93.2%
4024
94.7%
Agree 79
3.1%
153
8.3%
232
5.3%
Agree 17
3.6%
215
5.5%
232
5.3%
Agree 125
4.9%
107
6.8%
232
5.3%
Total 2523
100%
1843
100%
4366
100%
Total 475
100%
3891
100%
4366
2789
100%
1577
100%
4366
100%
Chi2 = 56.590, Pr. = .000 Chi2 = 3.188, Pr. = .074 Chi2 = 10.621, Pr. = .001
When we look at our combined agreement with demands/counterarguments indicator in Table 7,
we see that consociational and reflective institutions enhance the chances of ”deliberative
agreement”, while there is no effect for veto institutions. Our bootstrapped logit model (with
1,000 replications) yields an average logit coefficient of 1.428 for consociational institutions
and an average logit coefficient for reflective institutions of .455. Both are significant in the
95% bootstrapped confidence interval (in the sense that the 95% confidence interval excludes
0). Consociational institutions increase the probability of “deliberative agreement” by .357
percentage points (maximum effect), while reflective institutions increase the respective
probability by .114 percentage points (maximum effect). For the control variables, there is – as
predicted – a significant and positive association of low issue polarization with deliberative
agreement, and a negative association of chairmen with deliberative agreement. No statistically
reliable differences can be established for low party discipline, non-publicity, gender,
government members, and experts. The specialized techniques – rare events logit and gompit –
produce virtually identical results and corroborate the findings of our baseline logit model (see
Table 9 in the Appendix). This gives us confidence that the differences we found are
meaningful.
Table 7: Predictions of “Agreement with Demands/Counterarguments”
24
Independent Variables Average Effect 95% Bootstrapped Confidence Interval
Consociational Institutions 1.428* (1.079, 1.777)
Veto Institutions .099 (-.382, .581)
Reflective Institutions .455* (.196, .715)
Low Party Discipline/Presidentialism
-.265 (-.774, .244)
Non-Public Arena .125 (-.173, .424)
Low Issue Polarization .767* (.520, 1.015)
Female Legislators -.097 (-.400, .206)
Members of Government .417 (-.082, .916)
Chairmen -.821* (-1.394, -.248)
Experts -.214 (-.828, .399)
Constant -3.650* (-4.305, -2.996)
Coefficients are based on bootstrapped Logit with 1000 replications. N=4364. * Indicates confidence
interval excludes 0.
Second, when we look at agreement with counterarguments, we only find one statistically
significant institutional effect, namely for consociational institutions; reflective and veto
institutions, on the contrary, do not yield significant differences anymore. Consociational
institutions increase the probability of “deliberative agreement” by .277 percentage points
(maximum effect). For the control variables, the pattern looks relatively similar to the first
model, with statistically significant differences in deliberative agreement for low issue
polarization, government members and chairmen, and no statistical differences for lower party
discipline, non-publicity, gender, and experts. Again, the specialized techniques corroborate
the findings of our baseline logit model (see Table 10 in the Appendix)
25
Table 8: Predictions of “Agreement with Counterarguments”
Independent Variables Average Effect 95% Bootstrapped Confidence Interval
Consociational Institutions 1.107* (.684, 1.530)
Veto Institutions .051 (-.490, .592)
Reflective Institutions .274 (-.032, .579)
Low Party Discipline/Presidentialism
-.235 (-.804, .333)
Non-Public Arena .100 (-.268, .467)
Low Issue Polarization .648* (.352, 944)
Female Legislators -.264 (-.628, .099)
Members of Government .611* (.052, 1.171)
Chairmen -.674 (-1.361, .013)
Experts -.194 (-.902, .514)
Constant -3.684* (-4.421, -2.947)
Coefficients are based on bootstrapped Logit with 1000 replications. N=4364. * Indicates confidence
interval excludes 0.
For both models, we also probed whether there are interactive effects among the three
institutional variables. In particular, we checked whether the effects of veto and reflective
institutions play out differently in the context of consociational and competitive institutions.
But we found no such interactive effects. All interaction terms were non-significant (the
confidence intervals were large and always included 0), so we dropped the interaction terms out
of the models.
The results of the multivariate analysis differ from the bivariate findings. They show that is
primarily consociational institutions – involving incentives, spaces and scripts for deliberative
action - that bring about “deliberative agreement”. Competitive institutions, to the contrary,
26
involve powerful incentives that tend to undermine deliberative agreement. The effects of the
other institutional factors – veto institutions and reflective institutions – tend to be absorbed by
consociational and competitive institutions. This is particularly true for veto institutions.
Looking at the raw data, the overall rate of agreement with demands/counterarguments – the
critical indicator for an incentive-based institutional factor - is at 3.80 percent in the context of
competitive institutions, with a rate of agreement of 3.86 percent in veto situations and 3.54
percent in non-veto situations. In the context of consociational institutions, differences between
veto and non-veto institutions seem to be more marked: the overall rate of agreement with
demands/counterarguments is at 13.35 percent in the context of competitive institutions, with a
rate of agreement of 13.55 percent in veto situations and 6.86 percent in non-veto situations.13
For agreement with counterarguments, there are extremely tiny differences between veto and
non-veto situations: 0.74 percent in the context of consociatonal institutions, and 0.42 percent
in the context of competitive institutions. These results suggest that institutions thriving on
purely incentive-based mechanisms - such as veto institutions - do not perform well when it
comes to explain a philosophical ideal. As critics have suggested, there seems to be indeed no
straightforward reconciliation of contractarianism with Habermasian discourse ethics.
Reflective institutions, in turn – involving scripts and spaces for deliberative action – perform
better and produce an effect on “deliberative agreement” more broadly conceptualized
(including meta-agreement, productive conclusions and respectful compromises). But they fail
to produce statistically reliable differences for the stricter indicator, agreement with
counterarguments. This is a clear refusal of our theoretical expectations. The scripts and spaces
promoting deliberative action in reflective institutions are either absorbed by concurrent scripts
and spaces in consociational institutions or, in the context of competitive institutions, are
undermined by strong incentives to forego agreement with counterarguments. The raw data
corroborate such a conclusion. In the context of consociational institutions, the overall rate of
agreement with counterarguments is at 8.3 percent, with a rate of agreement of 9.79 percent in
second chamber debates and 7.04 percent in first chamber debates. In the context of competitive
institutions, the overall rate of agreement with counterarguments is at 3.13 percent, with a rate
of agreement of 3.29 percent in second chamber debates and 3.07 percent in first chamber
debates.
13 However, the number of observations in the non-veto situation is low (N=79) and the number of agreements under these conditions dwindles to 7; this precludes a strong interpretation of this finding.
27
In conclusion, we would like to stress two points. First, the powerful and robust effect of
consociational institutions on “deliberative agreement” underlines that consociational
institutions indeed involve a deliberative dimension and cannot be simply equated with
bargaining, as many scholars have claimed. Otherwise, we should not have found substantive
and significant differences for both types of deliberative agreement. Second, while the
insignificant effects of veto institutions in all models are a clear blow for our hypothesis, it does
not provide support for the critics’ argument either. Austen-Smith and Feddersen assume that
veto power is strongly counterproductive of deliberative action, which should produce a
negative association of this predictor with deliberative action. This, however, is not the case.
Conclusion In this article, we proposed and empirically probed three institutional design factors –
consociational institutions, veto institutions, and reflective institutions - in the production of the
Habermasian ideal of “deliberative agreement” in the context of legislative debates in
Switzerland, the United States, and Germany. The three institutional factors premise on
different mechanisms – incentives, spaces, and scripts - which can be assigned to different
metatheoretical traditions, namely rationalism and constructivism. Our main theoretical
expectation was that neither rationalism nor constructivism alone is capable of producing the
philosophical ideal. While rationalism in itself is too weak to theoretically capture the
philosophical ideal in its pure form, a constructivist argument is too strong and ignores that
there are may be strong rationalist incentives that run counter to its realization. Our findings
lend some support to these theoretical expectations. Our analysis suggests that we need
institutions – such as consociational institutions - comprising the full metatheoretical menu
with the right incentives, the right scripts and action spaces - to produce the philosophical ideal
in its pure form. Institutions having the wrong incentives – such as competitive institutions –
are strongly counterproductive of the philosophical ideal. Under such conditions, favorable
mechanisms – be it incentives, spaces, and scripts - are not much of a help, especially when it
comes to capture the philosophical ideal in its pure form.
Methodologically, one major challenge was to estimate causal effects from a problematic data
base, involving both “rare events” and a finite sample. We used a bootstrapped logit as our
baseline estimation method but checked for robustness with specialized techniques (rare events
28
logit, and gompit). In our case, the bootstrapped logit models performed well and did not
produce biased estimates and standard errors. But this may not always hold true: the more
extreme analysis becomes (in terms of skewness and number of observations) the more
necessary it becomes to perform thorough robustness checks with sophisticated techniques.
In sum, our analysis shows that certain types of political institutions indeed provide a meeting
point for positive political science and philosophy. But in the realm of real world politics, it is
fair to say that this meeting point is not at the main concourse of New York’s Grand Central
Station, but rather at a remote and exclusive “gentlemen’s club”. The Habermasian ideal of
“deliberative agreement” is a rare event even under propitious conditions.
29
References
Andeweg, Rudy B. 2000, “Consociational Democracy”, Unpublished manuscript, Leiden
University.
Aspingwall, Mark and Gerald Schneider 2000, “Same menu, separate tables: The
institutionalist turn in political science and the study of European integration”, European
Journal of Political Research 38: 1-36.
Avio, Kenneth L. 1997, “Constitutional Contract and Discourse Ethics: The Agreement
Theories of James Buchanan and Jürgen Habermas”, Journal of Theoretical Politics 9,
533-553.
Bächtiger, André and Jürg Steiner 2004, “Switzerland: Territorial Cleavage Management as
Paragon and Paradox”, in: Amoretti, Ugo and Nancy Bermeo (ed.), Federalism and
Territorial Cleavages, Johns Hopkins University Press, pp. 27-54.
Bächtiger, André and Marco R. Steenbergen 2004, “The Real World of Deliberation. A
Comparative Study of Its Favorable Conditions in Legislatures”, IUE Working Paper,
Florence.
Checkel, Jeffrey T. 1999, “Norms, institutions, and national identity in contemporary Europe”,
International Studies Quarterly 43: 43-83
Colomer, Josep Maria. 2001. Political institutions: democracy and social choice, Oxford:
Oxford University Press, 2001.
DiMaggio, Paul J. and Powell, Walter W. 1991, “Introduction”, in: Powell, Walter W. and Paul
J. DiMaggio (eds.), The New Institutionalism in Organizational Analysis, Chicago/London:
The University of Chicago Press, 1-38.
Döring, Herbert 1995, “Time as a Scarce Resource: Government Control of the Agenda”, in:
Döring, Herbert (ed.), Parliaments and Majority Rule in Western Europe, Frankfurt:
Campus Verlag.
30
Dryzek, John and List, Christian 2003, “Social Choice Theory and Deliberative Democracy: A
Reconciliation”, British Journal of Political Science 33: 1-28.
Eckstein, Harry 1975, “Case Study and Theory in Political Science”, in: Greenstein, Fred I. and
Nelson W. Polsby (eds.), Handbook of Political Science, Vol. 7, Strategies of Inquiry,
Reading, Mass.: Addison-Wesley.
Elster, Jon 1998, “Deliberation and Constitution Making“, in: Elster, Jon (ed.), Deliberative
Democracy, Cambridge University Press: 97-122.
Fiorina, Morris P. 1981, Retrospective Voting in American National Elections, New Haven,
Conn.: Yale University Press.
Fishkin James S. and Robert C. Luskin 2005, “Experimenting with a Democratic Ideal:
Deliberative Polling and Public Opinion”. Acta Politica 40/5 (Special Issue “Empirical
Approaches to Deliberative Democracy”): 284-298.
Goodin, Robert 1996, “Institutionalizing the Public Interest: The Defense of Deadlock and
Beyond”, American Political Science Review 90: 331-43.
Goodin, Robert E. 2003, Democratic Deliberation Within”, In: Fishkin, James S. and Peter
Laslett (eds.), Debating Deliberative Democracy, Oxford: Blackwell, 2003), pp. 54 79.
Goodin, Robert E. 2005, “Sequencing Deliberation”, Acta Politica 40/2 (Special Issue
“Empirical Approaches to Deliberative Democracy”): 182-196.
Habermas, Jürgen 2005, “Concluding Comments on Empirical Approaches to Deliberative
Politics”, Acta Politica 40/5 (Special Issue “Empirical Approaches to Deliberative
Democracy”): 384-392.
Kaiser, André 1997, “Types of Democracy. From Classical to New Institutionalism”, Journal
of Theoretical Politics 9: 419-444.
31
Keck, Otto 1995, “Rationales kommunikatives Handeln in den internationalen Beziehungen.
Ist eine Verbindung von Rational-Choice-Theorie und Habermas’ Theorie des
kommunikativen Handelns möglich?”, Zeitschrift für Internationale Beziehungen 2: 5 -48.
King, Gary and Langche Zeng 2001, “Explaining Rare Events in International Relations”,
International Organization 55: 693-715.
Holzinger, Katharina 2001, “Verhandeln statt Argumentieren oder Verhandeln durch
Argumentieren? Eine empirische Analyse auf der Basis der Sprechaktheorie”, Politische
Vierteljahresschrift 42: 414-446.
Johnson, James 1993, “Is Talk Really Cheap? Prompting Conversation Between Critical
Theory and Rational Choice”, American Political Science Review 87: 74-85.
Jupille, Joseph, James Caporaso and Jeffrey T. Checkel 2003, “Integrating Institutions:
Rationalism, Constructivism, and the Study of the European Union”, Comparative Political
Studies, 36, 1 & 2: 7-40.
Knight, Jack and James Johnson 1994, “Aggregation and Deliberation: On the Possibility of
Democratic Legitimacy”, Political Theory 22: 277-96.
Kriesi, Hanspeter 2001, “Die Rolle der Öffentlichkeit im politischen Entscheidungsprozess.
Ein konzeptueller Rahmen für ein international vergleichendes Forschungsprojekt”,
Veröffentlichungsreihe der Arbeitsgruppe Politische Öffentlichkeit und Mobilisierung,
Wissenschaftszentrums Berlin für Sozialforschung.
Lamnek, Siegfried 1995, Qualitative Sozialforschung. Band 2: Methoden und Techniken,
Weinheim: Pychologie Verlags-Union.
Lewis-Beck, Michael 1988, Economics and Elections: The Major Western Democracies, Ann
Arbor: University of Michigan Press.
Lijphart, Arend 1969, “Consociational democracy”, World Politics 21: 207-25.
32
Lijphart, Arend 1999, Patterns of Democracy. Government Forms and Performance in
Thirty-Six Countries, New Haven, Conn.: Yale University Press.
Mansbridge, Jane 1980, Beyond Adversary Democracy, New York: Basic Books.
Mansbridge, Jane 1996, “Reconstructing Democracy”, in: Nancy J. Hirschmann and Christine
Di Stefano (eds.), Revisioning the Political: Feminist Reconstructions of Traditional
Concepts in Western Political Theory, 117-38, Boulder (Co.): Westview.
McKeown, Timothy J. 1999, “Case Studies and the Statistical Worldview: Review of King,
Keohane, and Verba’s Designing Social Inquiry: Scientific Inference in Qualitative
Research”, International Organization 53: 161-190.
March, James G. and Johan P. Olson 1998, “The Institutional Dynamics of International
Political Orders”, International Organization 52: 943-969.
Müller, Harald 1995, “Spielen hilft nicht immer. Die Grenzen des Rational-Choice Ansatzes
und der Platz der Theorie kommunikativen Handelns in der Analyse internationaler
Beziehungen”, Zeitschrift für Internationale Beziehungen 2: 367-79.
Müller, Harald 2004, “Arguing, Bargaining and All That: Communicative Action, Rationalist
Theory and the Logic of Appropriateness in International Relations”, European Journal of
International Relations 10: 395-435.
Nagler Jonathan 1994, “Scobit: an alternative estimator to logit and probit”, American Journal
of Political Science 38: 230-55.
Neblo, Michael 2005, “Thinking through Democracy: Between the Theory and Practice of
Deliberative Politics”, Acta Politica 40/2 (Special Issue “Empirical Approaches to
Deliberative Democracy”): 169-181.
Ostrom, Elinor 1998, “A Behavioral Approach to the Rational Choice Theory of Collective
Action”, American Political Science Review 92: 1-22.
33
Powell, Bingham G., Jr. and Whitten Guy D. 1993, “A Cross-National Analysis of Economic
Voting: Taking Account of the Political Context”, American Journal of Political Science 37:
391-411.
Risse, Thomas 2000, “Let’s argue! Communicative Action in World Politics”, International
Organization 54: 1-39.
Rosenberg, Shawn 2004, “Examining three conceptions of deliberative democracy: A field
experiment”, Paper presented at the Conference on “Empirical Approaches to Deliberative
Politics”, European University Institute, Swiss Chair, Firenze 21-22 May 2004
Rothstein, Bo 1998, Just Institutions Matter: The Moral and Political Logic of the Universal
Welfare State, Cambrige: Cambridge University Press.
Rothstein, Bo 2004, “Is Political Science Producing Technically Competent Barbarians?”
European Political Science.
Scharpf, Fritz W. 1997, Games Real Actors Play. Actor-Centered Institutionalism in Policy
Research, Boulder, CO: Westview Press.
Spörndli, Markus 2004, Diskurs und Entscheidung. Eine empirische Analyse kommunikativen
Handelns im deutschen Vermittlungsausschuss, Wiesbaden: VS Verlag für
Sozialwissenschaften.
Steenbergen, Marco R., Bächtiger, André, Spörndli, Markus and Steiner, Jürg 2003,
“Measuring Political Deliberation”, Comparative European Politics 1: 21-48.
Steiner, Jürg 1974, Amicable Agreement Versus Majority Rule. Conflict Resolution in
Switzerland, Chapel Hill: University of North Carolina Press.
Strom, Kaare 2000, “Delegation and accountability in parliamentary democracies”, European
Journal of Political Research 37: 261-289.
34
Tsebelis, George and Money, Jeannette 1997, Bicameralism, Cambridge University Press.
Tsebelis, George 2002, Veto Players. How Political Institutions Work, Princeton University
Press.
Ulbert, Cornelia and Risse, Thomas 2005, “Deliberately changing the discourse: what does
make arguing effective?”, Acta Politica 40/3 (Special Issue “Empirical Approaches to
Deliberative Democracy”): 351-67.
Vanberg, Viktor and Buchanan, James M. 1989, “Interests and Theories in Constitutional
Choice”, Journal of Theoretical Politics 1: 49-62.
Wendt, Alexander 2001, “Driving with the Rearview Mirror: On the Rational Science of
Institutional Design”, International Organization 55: 1019-49.
35
Appendix
Table 9: Predictions of “Agreement with Demands/Counterarguments”
Independent Variables Rare Events Logit Gompit
Consociational
Institutions
1.420* (1.074, 1.763) .488* (.373, .602)
Veto Institutions .081 (-.390,.552) .017 (-.137, .172)
Reflective Institutions .453* (.194, .712) .172* (.074, .270)
Low Party Discipline/
Presidentialism
-.269 (-.772, .235) -.069 (-.227, .090)
Non-Public Arena .122 (-.174, .419) .063 (-.045, .172)
Low Issue Polarization .765* (.519, 1.012) .259* (.164, .353)
Female Legislators -.093 (-.393, .208) -.038 (-.148, .072)
Government Members .430 (-.060, .920) .194 (-.023, .410)
Chairmen -.793* (-1.346, -.238) -.326* (-.525, -.127)
Experts -.189 (-.781, 403) -.075 (-.284, .134)
Constant -3.610*(-4.253, -2.968) -1.326*(-1.527, -1.125)
N=4364. Table entries are average effects (estimates 95% bootstrapped confidence
interval in parentheses). * Indicates confidence interval excludes 0.
36
Table 10: Predictions of “Agreement with Counterarguments”
Independent Variables Rare Events Logit Gompit
Consociational
Institutions
1.010* (.680, 1.520) .333* (.212, .455)
Veto Institutions .028 (-.498,.554) .013 (-.146, .172)
Reflective Institutions .273 (-.031, .576) .087 (-.015, .189)
Low Party Discipline/
Presidentialism
-.239 (-.801, .323) -.042 (-.204, .119)
Non-Public Arena .096 (-.269, .461) .061 (-.053, .175)
Low Issue Polarization .646* (.352, 940) .199* (.098, .300)
Female Legislators -.255 (-.614 .104) -.081 (-.197, .035)
Government Members .630* (.086, 1.175) .231* (.011, .451)
Chairmen -.630 (-1.277, .017) -.246* (-.454, -.037)
Experts -.158 (-.830, 514) -.055 (-.274, .163)
Constant -3.635*(-4.355, -2.915) -1.344*(-1.553, -1.135)
N=4364. Table entries are average effects (estimates of 95% bootstrapped
confidence interval in parentheses). * Indicates confidence interval excludes 0.