In the shadow of direct democracy: A quantitative analysis of decision-making processes in...

33
In the shadow of direct democracy: A quantitative analysis of decision-making processes in Switzerland 1 Sarah Nicolet, Pascal Sciarini and Alex Fischer Graduate Institute of Public Administration (IDHEAP) 21, rte de la Maladière CH-1022 Chavannes-Lausanne Email: [email protected], [email protected], [email protected] Tel: + 41 21 694 07 63 Paper presented at the “Drei-Länder-Tagung”, workshop “Governing under the constraints of the referendums”, University of Bern, 14-15 November 2003 Abstract Despite the emphasis put on the consensus-building capacity of the Swiss decision-making system, the institutional mechanisms favoring compromise are still little explored. In this contribution we develop a simple model of the determinants of conflict and consensus that accounts for both conflict-conducive factors such as an act’s importance and consensus- facilitating practices, such as pre-parliamentary consultation procedures. Our results based on data from the 1995-99 legislature show that consensus among the parliamentary elite strongly facilitates the adoption of an act in the referendum phase, whereas the effects of pre- parliamentary procedures are more ambiguous and vary depending on the phases of the decision-making process. 1 The authors would like to acknowledge financial support from the Swiss National Science Foundation (grant no 5004-058512). We also would like to thank Max Bergman (SIDOS), Thomas Christin, Fabrizio Gilardi and Simon Hug for their helpful comments and suggestions on earlier drafts of this paper. The data-gathering for this study would not have been possible without the help of experts of Swiss politics and numerous officials in the Swiss administration and parliamentary services as well as Michael Hermann (University of Zurich) and our colleague Sandra Villiger. We express our gratitude to them.

Transcript of In the shadow of direct democracy: A quantitative analysis of decision-making processes in...

In the shadow of direct democracy: A quantitative analysis

of decision-making processes in Switzerland1 Sarah Nicolet, Pascal Sciarini and Alex Fischer

Graduate Institute of Public Administration (IDHEAP)

21, rte de la Maladière

CH-1022 Chavannes-Lausanne

Email: [email protected],

[email protected], [email protected]

Tel: + 41 21 694 07 63

Paper presented at the “Drei-Länder-Tagung”, workshop “Governing under the constraints of

the referendums”, University of Bern, 14-15 November 2003

Abstract

Despite the emphasis put on the consensus-building capacity of the Swiss decision-making

system, the institutional mechanisms favoring compromise are still little explored. In this

contribution we develop a simple model of the determinants of conflict and consensus that

accounts for both conflict-conducive factors such as an act’s importance and consensus-

facilitating practices, such as pre-parliamentary consultation procedures. Our results based on

data from the 1995-99 legislature show that consensus among the parliamentary elite strongly

facilitates the adoption of an act in the referendum phase, whereas the effects of pre-

parliamentary procedures are more ambiguous and vary depending on the phases of the

decision-making process.

1 The authors would like to acknowledge financial support from the Swiss National Science Foundation (grantno 5004-058512). We also would like to thank Max Bergman (SIDOS), Thomas Christin, Fabrizio Gilardi andSimon Hug for their helpful comments and suggestions on earlier drafts of this paper. The data-gathering for thisstudy would not have been possible without the help of experts of Swiss politics and numerous officials in theSwiss administration and parliamentary services as well as Michael Hermann (University of Zurich) and ourcolleague Sandra Villiger. We express our gratitude to them.

1

Introduction

Switzerland is traditionally referred to as a “democracy of concordance”, that is a system in

which decision-making rests on the search for amicable agreements and consensus widely

accepted among the main political actors. There is an abundant literature studying

Switzerland’s institutional features. To quote only a few examples, Lijphart (1974; 1984;

1999) has analyzed Swiss institutions as representative of a case of consensus democracy in

which the aim is to restrain majority rule and encourage the sharing of power between the

majority and the minority. Steiner (1974) for his part has underscored the importance of

“amicable agreements” among Swiss elites. The notion of political integration was also at the

heart of Deutsch’s study of the Swiss system (1976). However, the above mentioned studies

mainly concentrate on a macro-analysis of the consensus-building capacity of the political

system. That is, they identify key features likely to favor the emergence of compromise

among elites but they stop short from examining the effects of specific institutional

mechanisms on conflict reduction in legislative processes. The few contributions delving into

the analysis of the influence of specific decision-making institutions on consensus-building

(see for instance Neidhart 1970; Ossipow 1994) underscore the importance of “institutional

redundancy” (Bendor 1985; Landau 1969): The multiplication of institutional veto points

pushes political elites to cooperate in order for an act to go successfully through the decision-

making process (Ossipow 1994). Hence the search for consensus that is at the heart of the

Swiss decision-making system.

Empirical contributions on the determinants of conflict and consensus are equally rare and are

usually limited to the analysis of some specific phases of the decision-making process (e.g.

Sciarini and Trechsel 1996; Trechsel and Sciarini 1998) or to a narrow set of legislative acts

(e.g. Lehner 1984; Trechsel and Sciarini 1998). The few studies attempting to give a complete

and generalizable assessment of the functioning of Swiss decision-making are hampered by

the weakness of empirical data on some of the key phases of the process (e.g. Kriesi 1995;

Poitry 1989; Sciarini 1999). In particular, despite its utmost importance in the decision-

making system (e.g. Kriesi 1980), the so-called “pre-parliamentary” phase is still very poorly

documented. Similarly, one lacks studies of decision-making in the upper house of

parliament. In addition, most of the studies rely on data from the early 1970s and are in need

of an update.

The aim of this paper is to overcome these weaknesses and to develop a simple model of the

determinants of conflict and consensus in the decision-making process in Switzerland. The

2

model includes both conflict-conducive factors, such as the importance of a legislative act,

and consensus-facilitating practices, such as pre-parliamentary consultation procedures. More

specifically, we rely on three main sets of research hypotheses. First, we test for the influence

of legislative acts’ intrinsic features (importance, juridical type) on conflict in the various

phases of the decision-making process. Secondly, a number of hypotheses are developed

regarding the effects of institutional devices on consensus-building. Close attention will be

paid in particular to the contribution of pre-parliamentary procedures to conflict reduction.

These pre-parliamentary procedures are strongly related to the existence of direct democracy:

According to a famous hypothesis in the literature (Neidhart 1970) the “referendum threat” is

one of the key mechanisms pushing elites to cooperate, from the pre-parliamentary phase on.

Finally, the influence of internationalization on consensus and conflict is also integrated in the

model in order to account for recent changes in the context in which the Swiss decision-

making system operates.

Resorting to a set of quantitative techniques (structural equation modeling and binary logistic

regression), we test our model on a unique dataset containing information on all legislative

acts voted by the Swiss Federal Assembly during the 1995-99 legislature, and subject to one

of the three direct democratic institutions (i.e. popular initiative, optional referendum,

mandatory referendum). This dataset presents the advantage of providing quantitative

information on all phases of decision-making (including the pre-parliamentary one) and of

containing the results of all votes held on these acts for both upper and lower chambers. These

features permit to empirically assess the level of conflict in the different phases of the

decision-making process and to evaluate the determinants of conflict and consensus.

Particular attention will be paid to the comparison of conflict level and conflict reduction

within the two houses, an aspect that is still understudied in the literature (Tsebelis and

Money 1995; Tsebelis and Money 1997).

The paper is structured in the following way. The first section offers an overview of the

phases of the Swiss decision-making process and of the main empirical studies on the topic.

Then, a model of the determinants of conflict and consensus is presented and the underlying

research hypotheses are developed. The third part describes the dataset used and the

operationalization of the variables. Empirical findings are presented in the fourth section

starting with a descriptive analysis of our dependent variable, that is the level of conflict in the

parliamentary and referendum phases. Then, the validity of the model is assessed using

structural equation model (parliamentary phase) as well as binary logistic regression

(referendum phase). In conclusion, the main research findings are summarized and discussed.

3

1. The decision-making process in Switzerland

Overview of the formal steps of the process2

As underscored by Linder (1987, p.173), the evolution of the Swiss decision-making process

is characterized by “the concern for taking widely into account the different interests and

values present”. This concern is constant across the different stages of the decision-making

process and is already visible at the very onset of the decision-making process. Next to a

federal state actor (Federal Assembly, Federal Council, administration), the impulsion for a

new legislative act can also be given by the “cantons” (through the “cantonal” initiative) and

by the citizens (through the constitutional popular initiative).3 During the pre-parliamentary

phase, a phase in which legislative acts are drafted by the administration, a number of

consultation mechanisms are available to favor the inclusion of relevant social, economic, and

cultural groups. A first draft of the text can be submitted to a so-called “pre-consultation” of

the interested organizations. The project can then be re-drafted by an expert commission.

These commissions do not simply consist of experts in the field but they include

representatives of interest groups, cantons and parties (Morand 1987, p.76). Relevant actors

can also be invited to give their opinion on the issues at stake in a written consultation

procedure organized by the government or the department in charge of the question.

The inclusive character of decision-making is further reinforced in the parliamentary stage by

the perfect bicameralism of the Federal Assembly (Morand 1987, p.77). The Council of States

(the upper chamber) in which each canton has two deputies regardless of its population has

equal power as the National Council (the lower chamber) that represents the people. When a

new project enters the parliamentary arena, the presidents of both chambers decide which

council is the first to treat the object. The project is then examined by the competent

commission of the chamber having priority, before being submitted to the parliamentary

groups that define the position to defend in the plenary session. Next, the object enters the

first chamber where the representatives decide whether the bill is to be accepted for further

discussions or not. This debate is followed by an article-by-article discussion of the project.

So-called “intermediary” votes take place at this stage. The deputies either formally vote on

2 We limit ourselves to a presentation of the “standard” procedure leading to the adoption of a federal law. Oneshould thus bear in mind that a draft law does not necessarily pass through all the steps mentioned here.Obviously, the following overview imperfectly describes the procedure that applies to urgent federal laws or tointernational treaties.3 By collecting 100’000 signatures in 18 months, any group of citizens can put an issue on the political agendaand force the elite – and then the electorate – to vote on a constitutional amendment.

4

an article or approve it tacitly (if no representative asks for a formal vote).4 At the end of the

first round of deliberation, members of parliament (MPs) vote on the overall project

(“ensemble” vote).5 The project is then transmitted to the second chamber, where it follows

the same procedure. If there are divergences between the two chambers, the project goes back

to the first chamber. This marks the start of the “navette” process in which the legislation

shuttles between houses in order to reconcile differences. Since 1992, the navette is limited to

two rounds. If the two chambers still disagree after that, a meeting of conciliation

(“conférence de conciliation”) is set up to break the deadlock between them. The conference

is composed of thirteen members of the two houses’ standing committees that originally

examined the legislation.

Once both chambers have adopted a legislative act, it enters the last stage of the process, the

referendum phase. The referendum offers citizens the possibility to vote on legislative acts

adopted by the parliament. In the case of an act submitted to the optional referendum, the

constituency has to vote on the object if 50’000 citizens ask for a referendum during the 100

days following its adoption by the Federal Assembly. If an act is submitted to the mandatory

referendum, the popular vote takes place automatically without citizens’ demand. If no

referendum is launched or if the constituency accepts an object, the legislative act is

definitively adopted. Its entry into force is fixed by the Federal Council or, depending on the

cases, directly into the law.

Overview of the literature

Switzerland’s complex and encompassing decision-making process has often been praised for

its capacity to overcome divergences and achieve consensus in a society characterized by

important linguistic and cultural cleavages (Kriesi 1995; Linder 1987; Linder 1998;

Papadopoulos 1997). A few empirical studies have attempted to analyze the determinants of

conflict and consensus in some specific phases of the decision-making process. Thus, some

contributions have studied the influence of the development of consensus in the parliamentary

arena on the acceptability of acts in the referendum phase. Based on all legislative acts

adopted by the National Council between 1947 and 1995, Sciarini and Trechsel’s study (1996;

4 In a few cases, intermediary votes take place automatically (without a deputy asking for). This holds, forexample, for intermediary votes on an “emergency” clause and for articles subject to the “brake to debts” clause.5 This vote allows a council to reject an act if it is no longer coherent or acceptable after the first round of debate(Trivelli 1975, p.262). In such a scenario, the process starts over. The project goes to the other chamber, where itis examined in its entirety. After a positive “ensemble” vote in this chamber, it comes back to the first chamber,which discusses the object again. In case of a second negative “ensemble” vote, the project is abandoned. Thissituation has very seldom occurred.

5

1998) for instance shows that reaching a consensus in the parliamentary phase is a winning

strategy in order to avoid a referendum. Similarly, the likelihood of acceptation of legislative

acts in a popular vote also increases with the level of consensus among the parliamentary

elite.6

Other studies have embraced a more encompassing view of the decision-making process and

have analyzed it in its entirety. Poitry (1989) has carried out an exhaustive, albeit

parsimonious, analysis of the decision-making processes of the legislative period 1971-75,

based on a selective number of indicators: origin of legislative acts, duration of the pre-

parliamentary, parliamentary and referendum phases of the process, degree of consultation in

the pre-parliamentary phase, level of conflict in the parliamentary and referendum phase, etc.

His work constitutes an important contribution to the understanding of the links between the

different phases of the process. Yet, the relationships are only examined on a bivariate basis,

which weakens the validity of his tests. Kriesi (1995) and Sciarini (1999) have both re-

analyzed Poitry’s database using multivariate models and qualify some of his findings.7

Finally, a few contributions push the analysis one step further and develop a complete study

of the accommodation of conflict in the decision-making process by examining the influence

of the different phases on each other. Lehner (1984) in particular measures in a single model

the effect of the level of heterogeneity of elite positions in the pre-parliamentary phase on

heterogeneity in the parliamentary and referendum stages. The same model is tested for both

houses. His findings show that the degree of support for a law in parliament, in the upper

chamber particularly, is important for the support it receives in popular votes (p.38).

Unfortunately, it is difficult to generalize on the basis of his study given that the data analyzed

is limited to 58 cases of decisions on economic and fiscal policy of the 1970s. In a recent

article focusing on the influence of internationalization on decision-making processes

Sciarini, Nicolet and Fischer (2002) show that conflict among the elite tends to diminish as a

consequence of a higher level of internationalization, and this both in the parliamentary and

referendum phases. Furthermore, their study casts some doubt on the consensus-building

effect of pre-parliamentary procedures. However, their contribution rests on regression

6 An analysis by institutional type of acts nuances this finding. While consensus in the National Councilincreases the probability of acceptation of an act in a popular vote for popular initiatives and objects submitted tothe mandatory referendum, this effect is very weak in the case of optional referendums.7 Thus, Kriesi’s work shows that once you control for an act’s intrinsic features (juridical type, perceivedpolitical importance, and political domain), the relationship between degree of consultation and parliamentaryconflict disappears almost entirely (p.185). Contrarily to Kriesi’s results, Sciarini finds that the resort to pre-parliamentary consultations tends to stir up conflict. However, as the author underscores, a key-variable ismissing from the model, that is the potential level of conflict of an act at the outset (pp.609-610).

6

models that measure the effects of determinants of conflict and consensus on a single

dependent variable but do not account for the simultaneous influences of a set of independent

variables on a set of dependent variables.

These various pieces of work have paved the way for the study of Swiss decision-making.

However, as this brief overview shows, most of them stop short from embracing a

comprehensive view of the decision-making system, or are limited to a handful of legislative

acts. In the next section we develop a complete but parsimonious model of the determinants

of conflict and consensus in the Swiss decision-making process, which we then apply to

recent data.

2. The model

Our model of the determinants of conflict and consensus focuses on three sets of variables,

that is the intrinsic features of a legislative act, the institutional setting of the decision-making

process, and the degree of internationalization of an act. Each research hypothesis is detailed

below.

Intrinsic features of an act

Conflict level of an act is first influenced by its intrinsic features, that is characteristics that an

act possesses before it even enters the decision-making process. Two different dimensions are

of particular relevance here.

First, as Kriesi (1995) and Sciarini’s (1999) re-analysis of Poitry’s data have shown, the

importance of an act affects conflict level. Highly important projects are expected to be more

conflictual than more trivial ones, which are unlikely to arouse much interest and opposition

(hypothesis 1a). This argument is in line with the widespread thesis that the most important

acts are more frequently attacked by referendum since the end of the 1960s (e.g. Germann

1994; Kriesi 1995).

Secondly, and again in line with Kriesi and Sciarini’s findings, we assume that the level of

conflict varies according to the juridical type of an act. More specifically, we argue that

popular initiatives display a lower level of conflict than the other legislative acts (acts subject

to the optional or to the mandatory referendum). This basic distinction is justified by the fact

that popular initiatives are the only legislative acts that do not have their origin within the

political system but emanate directly from a popular impulsion. Usually challenging the

established legislative order popular initiatives are likely to result in a higher level of

consensus against them (that is a lower level of conflict) than other types of legislative acts

(hypothesis 1b).

7

While specific intrinsic features presumably make consensus-building easier or more difficult,

the decision-making system’s institutional design also exerts a key influence on the fate of a

project. Swiss political authorities have sought to develop a number of institutional devices

aimed at favoring the emergence of consensus.

Institutional devices

“Institutional redundancy” (Bendor 1985; Landau 1969) is considered as paramount for the

building of consensus and is at the core of the Swiss decision-making process (Ossipow 1994;

Papadopoulos 1997). As Ossipow puts it, “the Swiss constitutional history since 1848 can be

interpreted as a progressive development of a highly sophisticated system of institutional

redundancy” (p.28). In order to be adopted, a legislative act needs to transit through numerous

formal steps, resulting in a multiplication of the number of bodies having to agree for a

decision to be ratified (Papadopoulos 1997, p.28). To block the adoption of a legislative act, it

suffices to put a veto at one stage of the decision-making process. As a consequence,

decision-making is rendered very difficult. But it is precisely this difficulty that fosters

compromises among the elites. “Only cooperation between the political actors permits to

overcome the difficulties generated by the complexity itself of the system” (Ossipow 1994,

p.10).

In this perspective, the veto threat represented by the existence of direct democracy is one of

the key mechanisms forcing elites to cooperate. The fact that the electorate has the possibility

of rejecting unsatisfactory laws at the very end of the decision-making process introduces

uncertainty about the outcome of the process. Like a sword of Damocles it hangs as a

permanent threat over the political actions of the elite. According to Neidhart’s famous

hypothesis (1970), this has led to the development of the pre-parliamentary phase as a way of

preventing possible oppositions in the subsequent phases of the decision-making process.

Every political actor able to threaten in a credible way the process is invited to take part in the

pre-parliamentary phase. This is done through mechanisms of extensive consultation (expert

commissions, consultation procedures) favoring the finding of a compromise. As implicit in

Neidhart’s argument, the existence of inclusive pre-parliamentary institutions favors the

creation of consensus among the elite and decreases the likelihood of an act being attacked by

a referendum. This is summarized in the following hypothesis: The more developed the pre-

parliamentary procedures, the more the level of conflict is likely to be reduced in the

subsequent phases of the decision-making process. While Neidhart’s hypothesis primarily

applies to the impact of the pre-parliamentary consultations on the referendum phase

8

(hypothesis 2a) it can be extended by analogy (Kriesi 1995) to the parliamentary phase

(hypothesis 2b).

In addition, the level of consensus reached in the parliamentary arena is likely to affect in turn

the subsequent referendum phase. In agreement with hypothesis 2c, acts that were more

conflictual in the parliamentary phase are more susceptible to be attacked by a referendum. As

Trechsel and Sciarini have shown (1998, p.119), the existence of a large consensus among the

parliamentary elite “…significantly reduces the probability that a political actor shall attack a

bill, forcing a popular vote”. One expects thus a negative relationship between intra-

parliamentary consensus and the likelihood of a referendum being launched.

Internationalization

As evidenced earlier, most of the studies of consensus-building in Switzerland focus on the

early 1970s, that is on a time period in which the decision-making system was relatively

sheltered from the influences of the international arena. However, since the 1980s, one has

witnessed a speeding-up of the process of globalization/internationalization both from an

economic and institutional perspective (e.g. Busch 1998; Cerny 1997; Keohane and Milner

1996; Scharpf 1996).8

This evolution has also consequences for the functioning of the Swiss decision-making

process (e.g. Armingeon 1998; Häusermann, Mach, and Papadopoulos 2002; Sciarini,

Nicolet, and Fischer 2002). First, international law takes up a greater share of the domestic

legislative production (e.g. Linder, Schwager, and Comandini 1985). Secondly, internal and

external politics become more entangled. An increasing number of issues that used to be

domestic in nature are now co-defined at the international level and/or strongly influenced by

norms elaborated abroad (Sciarini 1999, pp.635-636). These changes affect in turn the

consensus-building capacity of the system. Scholars hold contradicting views with respect to

the impact of internationalization on the level of conflict among the domestic elite. On the one

hand, a number of authors argue that globalization/internationalization makes the adoption of

new domestic rules easier (Armingeon 1998; Grote and Schmitter 1999; Sciarini 1994).

Indeed, decisions taken at the international level help policy makers to reform domestic

policies without being blamed for the unavoidable domestic redistributive effects (Armingeon

1998, p.108). In that sense, decisions that would be highly controversial in a domestic setting

are more likely to be accepted under international pressure. This vision is compatible with

8 For a summary of the debate on the newness or not of the phenomenon of internationalization/ globalization,see Hefeker (1997).

9

Katzenstein’s (1984; 1985) argument that the economic openness and vulnerability of small

European states favor the development of a sense of common destiny among the elite and

push them to adopt consensual solutions. Hence the hypothesis that the more internationalized

a legislative act is, the lower its level of conflict (hypothesis 3a).

On the other hand, legislative acts negotiated at the international level can usually no longer

be modified at the domestic level. This strong “take-it-or-leave-it” character severely reduces

the elite’s room for maneuver and is likely to result in a more conflictual decision-making

process. The time pressure associated with the ratification of international agreements further

increases the risk of conflict. The necessity to ratify within a fixed period of time prevents the

elite from postponing its decision, a strategy that is sometimes used in domestic politics (Hug

and Sciarini 1995, p.65). According to this line of arguments, internationalized acts are

expected to be more conflictual than those that are more domestic in nature (hypothesis 3b).9

We can now summarize our hypotheses:

1a. The more important a legislative act, the higher the level of conflict in the decision-

making process.

1b. Popular initiatives display a lower level of conflict than the other types of legislative acts.

2a. The more developed the pre-parliamentary procedures, the lower the level of conflict in

the referendum phase.

2b. The more developed the pre-parliamentary procedures, the lower the level of conflict in

the parliamentary phase.

2c. The higher the level of consensus in the parliamentary phase, the lower the level of

conflict in the referendum phase.

3a./3b. The more internationalized a legislative act, the lower/the higher the level of conflict

in the decision-making process.

Control variable

As highlighted by Sciarini (1999, p.610), to measure the influence of different features and

institutional devices on conflict and consensus, one needs to take into account the level of

potential conflict of an act at the onset, that is the level of conflict an act is likely to produce

given its intrinsic features. This variable so far neglected in analyses of the Swiss decision-

making process is integrated in the model as a control variable.

9 While it is not included in our dataset, the agreement on the European Economic Area (EEA) is a case in point.It was rejected by a very small majority of the Swiss people in 1992 (Kriesi et al. 1993).

10

3. Operationalization

To empirically assess the validity of our research hypotheses, we rely on a new dataset

containing information on all legislative acts voted by the parliament during the 1995-99

legislature and subject to one of the three instruments of direct democracy, that is either

optional or mandatory referendums or popular initiative. This represents 210 legislative acts.

The dataset includes, among others, extensive information on all phases of the decision-

making process (pre-parliamentary, parliamentary and referendum). For each object we

collected information on its impulsion (timing and origin), the different pre-parliamentary

consultation mechanisms used (within and outside of the administration), and the fate of the

objects in the referendum stage (launching of a referendum, results of the popular vote). In

addition, the dataset also comprises the results of all votes (intermediary, ensemble and final)

that took place in both chambers of the parliament for all the objects. This dataset permits to

characterize the different objects in a quantitative way for each of the three main phases of the

decision-making process.

The intrinsic features of an act are measured in the following way. To assess the importance

of a project, we use two different measures. First, we rely on experts’ assessment of an

object’s importance on a scale from 0 to 4 (see appendix for a definition). This piece of

information was collected through a written questionnaire sent out to 80 experts of Swiss

politics (high-ranking civil servants, professors, journalists, politicians) and asking them to

evaluate a number of features of the projects under study. 44 questionnaires were returned,

which represents a percentage of answers of 55 percent. This variable is completed by a

measure of parliamentary importance that is constructed by using the parliament’s office’s

(“bureau du parlement”) decision on the type of debate used for an object in the National

Council.10 The scale varies between 1 (written procedure-low importance) and 5 (free debate-

high importance).11

The juridical type of an object is assessed by a dummy variable comparing the popular

initiatives to all other acts (objects subject to the optional or mandatory referendums).

10 This division into types of debates only applies to the lower chamber. Given the smaller size of the upperchamber, no restriction on debates is in force.11 There are five levels of debate in the National Council, that is written procedure, short debate, reduced debate,organized debate, and free debate. The main differences between the types of debates lie in the members’ right totake the floor, the nature of the individual motions (oral or written) or the floor time. For instance, in a freedebate, the type of debate used for the most important objects, all members have the right to take the floor,motions can be made orally and MPs can talk up to 20 minutes. In the case of a written procedure, in contrast,the right to take the floor is limited to a few members (the committee spokespersons) who can intervene only toanswer a motion. Furthermore, speaking time is limited to five minutes and motions need to be in written form.

11

Regarding the institutional setting, a measure of the degree of pre-parliamentary consultation

is computed by adding the different forms of consultation existing for an object: pre-

consultation, expert committee (0, 1 or 2), consultation procedure (0, 1 or 2), and other forms

of consultation such as hearings.12 Each procedure is worth 1 point. The constructed scale

varies between 0 and 6 and is treated like an interval variable.

The level of internationalization is measured on the basis of Swiss politics experts’

classification of the act on a scale going from 0 to 4 (see appendix for a definition).

To measure the “potential” level of conflict of an act at the onset of the process, we rely on

the judgment (scale from 0 to 4) of the civil servant who followed the most closely the

elaboration of an act. By “potential conflict”, we mean the level of conflict that an act was

likely to generate given its intrinsic features (see appendix for a definition).13

Two dependent variables are analyzed in this study, that is the level of conflict of an act in the

parliamentary phase and in the referendum one. To assess the level of conflict in the

parliamentary phase, we dispose of three different variables that each measures a specific

aspect of the phenomenon. They are the average level of conflict of the intermediary votes,14

the level of conflict of the “ensemble” votes, and the one of the final votes.15 All three

measures are available for both chambers and vary between 0 (unanimity in the chamber) and

50 (50/50 division in the chamber).

Conflict in the referendum phase for its part is estimated on the basis of the objects subject to

the optional referendum (176 objects). A dichotomous variable measuring whether a

referendum was successfully launched against an object or not is used. A referendum is

considered successful when the required number of signatures has been collected, even if the

referendum is rejected in the subsequent popular vote.

12 For the foreign policy objects (10 cases), there are no “pre-consultations”. Instead, we include the forms ofconsultation that took place before or during the international negotiation. 13 We are aware of the fact that the level of “potential conflict” of an act is very difficult to assess. In particular,the measure is likely to be biased by the knowledge of an object’s future fate. Relying on the opinion of civilservants who have been in charge of an act since the very beginning of the decision-making process shouldnevertheless minimize this risk. In a few cases, civil servants did not answer this question. In these cases (19),we relied on experts’ judgment on the level of potential conflict.14 Conflict in the intermediary votes rests on all the votes that have formally taken place, including theunanimous ones. The average level of conflict of the intermediary votes for each object is calculated by dividingthe smallest number of “yes” or “no” by the total number of “yes” and “no”. One obtains thus a number varyingbetween 0 (unanimous adoption or rejection) and 50 (equal number of “yes” and “no”). On this basis, an averagelevel of conflict is computed for each object. 15 Conflict in the “ensemble” votes and in the final ones is calculated by dividing the smallest number (thenumber of “no” in this case given that no object has been rejected in these votes) by the total number of “yes”and “no”.

12

To give an assessment of the level of conflict in the referendum phase on the basis of all

objects, including the ones subject to the mandatory referendum and the popular initiatives, a

second measure is computed in the following way. If no referendum has succeeded and thus

no popular vote has taken place, conflict is considered as null;16 if a referendum has been

successfully launched or a popular vote is compulsory (in case of a popular initiative or a

mandatory referendum), the level of conflict of the vote is computed for each object by

dividing the smallest number (“yes” or “no”) by the total number of “yes” or “no. This

measure is analogous to the one used to measure conflict in the parliamentary phase and

similarly varies between 0 (unanimous rejection or adoption) and 50 (50/50 division in the

popular vote).

4. Empirical tests

Conflict level in the different phases of the decision-making process

We start with a brief overview of our two main dependent variables, that is conflict level in

the parliamentary phase and in the referendum phase. To allow for a comparison with the

initial level of conflict of an act, a measure of the average level of potential conflict is also

included here.17

{TABLE 1}

Given the differences in measurement the scores presented in table 1 are not perfectly

comparable. They nevertheless enable us to examine the dynamics of conflict across the

decision-making system. A few interesting patterns can be highlighted in that respect. First,

results in table 1 tend to confirm the consensus-building capacity of the Swiss system. Indeed,

a trend towards a reduction of conflict through the decision-making process is visible.

However, the standard deviations are high revealing the great disparity in conflict levels

across objects. Secondly, bearing in mind that 50 represents the highest level of conflict, the

overall level of conflict in the impulsion stage appears as relatively moderate, but again with

16 The absence of a successful referendum does not necessarily mean that an object has not been contested in thereferendum phase. A group of citizens can launch a referendum against a law but fail to collect the requirednumber of signatures. For the 1995-99 legislature, this only occurred for two of the 176 objects submitted to theoptional referendum.17 The score for the initial level of conflict is calculated by modifying the original scale of potential conflict froma scale from 0 to 4 to a scale from 0 to 50.

13

great variations among legislative acts.18 Finally, conflict level in the referendum phase is

very low compared to the other two phases of the decision-making process. However, this

result is mainly explained by the small percentage of referendums launched: Among the 176

federal laws submitted to the optional referendums, only 11 were attacked by a successful

referendum. When one excludes the laws that did not lead to a popular vote from the analysis,

the pattern looks quite different (table 2). In fact, the legislative acts submitted to the optional

referendum and for which a popular vote has taken place appear very conflictual (the average

level of conflict amounts to 37.45), and in fact more conflictual than objects subject to the

mandatory referendum or popular initiatives. This is not surprising since the launching of the

referendum itself already signals a failure of the system at building consensus among the

political elite. The lack of consensus is then reflected in the popular vote.

{TABLE 2}

Results for the parliamentary phase can be analyzed in more details given that we have three

distinct but comparable measures at our disposal. Starting with the evolution of conflict within

both chambers we see that the average level of conflict drops during the parliamentary phase:

In both chambers, conflict decreases almost ten points between the intermediary and final

votes.19 However, as results in table 1 clearly show, conflict is set at a lower level in the upper

chamber. This tendency is corroborated by an examination of the percentage of unanimous

votes in the two councils. In the upper chamber, almost three quarters of the ensemble and

final votes are unanimous, while this is only the case for about three votes in ten in the lower

house.20 These findings can be explained by the greater homogeneity of representation in the

upper house compared to the lower one. The Council of State is traditionally dominated by

18 A more detailed look at the distribution of potential conflict reveals that 13% of the objects were considered aspotentially highly conflictual and almost half of them were classified as presenting none or a small risk ofpotential conflict.19 The low level of conflict of the ensemble votes should be interpreted with some caution for two main reasons.First, given that we do not take into account the timing of the different votes, it is hard to place them in achronological order. They can intervene before or after some of the intermediary votes. However, most of thetime, the majority of ensemble votes take place after the intermediary ones. This problem does not exist for thefinal votes, which always take place at the very end of the parliamentary process, that is after all the intermediaryvotes. Secondly, according to some authors (see for instance Trivelli 1975), these votes are mainly proceduraland they have little substantial meaning. 20 This pattern is further confirmed by the distribution of “tacit” intermediary votes in the two chambers, that isvotes for which no MP asked for a formal vote. Unfortunately, complete information on the percentage ofcollective votes is only available for the first year of the 1995-99 legislature. In this period, 84.9% of theintermediary votes in the upper house were tacit compared to 64.9% in the lower chamber.

14

the parties of the right which tend to support similar positions,21 while leftist and far-right

parties are underrepresented: During the legislature under study, the Social Democrats

occupied more than a quarter of the seats in the lower house, while they only had five elected

officials in the upper house (about ten percent of the seats); a similar asymmetry, albeit of a

lower magnitude, also holds for the far-right Swiss People’s Party.

The frequency of the resort to a “navette” or a conciliation conference informs us about the

patterns of convergence and divergence between the two chambers. As results in table 3 show,

exactly half of the objects were adopted by both houses of parliament without divergence, that

is without resorting to shuttling. Furthermore, 80% of the time the differences were solved

after one round of the “navette”, meaning that the object is discussed a second time in both

houses before being accepted. This pattern highlights the consensus-building capacity of

perfect bicameralism. The need for an agreement between the two chambers in order for

legislation to be adopted pushes members of parliament (MP’s) to cooperate. As Sciarini and

Hug emphasize, “the mutual neutralisation of the veto power that both councils have at their

disposal makes consensus necessary” (1995, p.60).

{TABLE 3}

However, a comparison of these findings with earlier studies on the topic shows that although

the level of convergence between houses remains high, it tends to decline relative to previous

legislatures. Ricklin and Ochsner (1984, p.90) for instance observe that 90% of the projects

are adopted after a maximum of two deliberations by chamber. The weakening of the

consensus is confirmed by a comparison with Huber-Hotz’s results (1991, p.174s) which

highlight a drop in the resort to the “navette” between 1875 and 1942. She shows in particular

that this decline was the sharpest between 1972 and 1989 where 77% of the constitutional

acts, and 55% of the laws and federal acts were adopted without resorting to shuttling. The

pattern for the 1995-99 legislature points rather towards an increase in the level of conflict

between the two chambers, since only 33% of the constitutional amendments and 44% of the

laws and federal amendments were adopted without shuttling.22

The upwards trend in conflict between the two houses also transpires from an analysis of the

resort to conciliation conferences: In the four years of the legislature under study there are as

21 In his study of decisions on fiscal and economic policy, Lehner shows that the correlation of parliamentarydecisions between the PRD, UDC, and PDC, the three main parties of the right, reaches .95 (1984, p.32).22 Owing to their different nature, popular initiatives are not included in the computation.

15

many conferences called (fifteen) as between 1902 and 1989 (Huber-Hotz 1991).23 However,

this sharp increase in the frequency of conferences may also be explained by the fact that

since 1992 the shuttling procedure is no longer unlimited.

So far, our analysis of the search for consensus in the decision-making process has focused on

a first description of the level of conflict in each phase taken separately, with a particular

emphasis on the parliamentary arena. In the next section, we turn to an examination of the

determinants of conflict (reduction) across the different phases of the decision-making

process.

Analysis of the determinants of conflict and consensus: the parliamentary phase

The hypotheses presented in section two are first assessed with regard to the determinants of

conflict and consensus at the end of the parliamentary stage (final votes). They are tested

using structural equation models (SEMs). These types of models allow to simultaneously test

a set of regression models (Arbuckle and Wothke 1999; Byrne 2001) and thus to measure the

influence of a number of independent variables on a set of dependent variables. All estimates

are computed using AMOS 4.0 (Arbuckle and Wothke 1999).24

As shown in figures 1 and 2, an identical model is applied to both chambers in order to allow

for a comparison between the two houses. This model has been developed by including all the

relationships between variables posited by our set of research hypotheses. In addition, in order

to optimize the model and account for the indirect influences exerted by variables, a number

of relationships identified as significant have been added to the model while the ones that

were non-significant have been removed.

Table 4 presents the detailed results of the structural equation models for both houses. All fit

measures suggest that one can be confident about the overall quality of the model for both

chambers.25

{TABLE 4}

23 Tsebelis and Money confirm these figures. According to their study, there were only 13 conferences for the4’803 pieces of legislation debated between 1902 and 1972 (1997, p.189).24 The distribution of the final votes does not follow a normal curve given the large number of unanimous votesin both chambers. To correct for this bias, we use a logarithmic transformation of the dependent variable. 25 Different fit measures are commonly used to measure the overall quality of a model. One admits usually that amodel should be rejected if the adjusted GFI is inferior to .9, the comparative fit index inferior to .95, and theRMSEA superior to .05. Furthermore, the quality of the model is considered as unsatisfying if the value of theChi-square is superior to the degrees of freedom multiplied by two.

16

The model tested on the lower chamber explains 38% of the variance in the level of conflict

of intermediary votes and 44% of the variance in the level of conflict of final votes. These

figures are slightly lower for the upper chamber (30% and 31% respectively), which is not

surprising given the lower level of variation in the scores of the intermediary and final votes

in the upper chamber. As mentioned in the previous section, in contrast to the National

Council, a majority of votes were unanimous in the upper chamber.

{FIGURE 1}

The validity of each of the three hypotheses will be discussed in turn on the basis of the

coefficients for both chambers. One notices that a number of coefficients are identical

between the two figures. This is explained by the fact that only the values for the intermediary

and final votes are specific to each house while the other variables remain the same in both

models.

Our results bring strong supporting evidence to hypothesis 1a regarding the influence of an

act’s importance. As expected, the more important an object, the higher its level of conflict in

the parliamentary phase. The importance of an act exerts an effect both directly and indirectly.

On the one hand, the two measures of importance (experts and parliament) have a positive

direct effect on conflict in the intermediary votes. This direct effect is reinforced by the

impact exerted indirectly by the experts’ importance through other variables (importance in

the parliament, potential conflict) resulting in a total effect of .50 for the National Council

(respectively .48 for the Council of States) on conflict in the intermediary votes.26

A similar pattern is visible for conflict in the final votes. In this case, the importance assessed

by the experts no longer exerts a direct effect on the final votes but a significant indirect effect

remains (.40 for the National Council; .30 for the Council of States).27 In addition, our

measure of parliamentary importance also affects the level of conflict in the final votes.

However, the total effect is stronger in the lower chamber (.31) than in the upper one (.11).

This weaker impact can again be explained by the generally lower level of conflict in the

26 The total effect of a variable is computed by adding its direct effect on a dependent variable to its indirecteffects via other variables. For instance, for the National Council, the effects of the experts’ importance onconflict in the intermediary votes are the following:

- direct effect: .35- indirect effect via importance (parliament): .43 * .29 = .12- indirect effect via potential conflict: .57 * .05 = .03- total effect: .35 + .12 + .03 = .50

17

latter: Over 70% of the final votes were unanimous in the Council of States, while it was only

the case for 28% in the lower chamber.

{FIGURE 2}

The hypothesis that popular initiatives are less prone to conflict than the other types of

legislative acts (hypothesis 1b) is confirmed in the Council of States but not in the National

Council: In the latter there is no difference in level of conflict between popular initiatives and

other acts; consensus is much larger regarding initiatives compared to other types of acts only

in the upper house. This pattern can be explained by the differences in representation between

the two chambers. Political actors that resort most frequently to popular initiatives or usually

support them (trade unions, leftist parties, new social movements, radical right parties…) are

fairly well represented in the National Council, but not in the Council of States. As a result,

the upper chamber rejects popular initiatives with larger majorities than the lower house.

The evidence with regard to the impact of the degree of pre-parliamentary consultations on

consensus-building (hypothesis 2b) is mixed for both houses. On the one hand, the level of

conflict in the final votes diminishes as a consequence of pre-parliamentary consultations.

This is congruent with our hypothesis and the function traditionally attributed to the pre-

parliamentary phase, that is to favor the emergence of compromises. However, a more

thorough examination of the results shows that the effect of consultation is not unidirectional.

While it decreases conflict in the final votes, it increases conflict in the intermediary votes;

the more one consults on an object, the greater the degree of polarization of the elite in the

article-by-article debate. This result suggests that the solutions found in the pre-parliamentary

phase are hotly debated in the initial stage of the parliamentary process (intermediary votes),

but nevertheless result in a fairly high level of consensus at the end of that process (final

votes).28

To get a clearer view of the overall impact of the pre-parliamentary phase on conflict, the total

effect on conflict level in the final votes can be computed by accounting both for the direct

effect and indirect effect (via the intermediary votes) of consultation procedures. This effect,

although less strong, remains negative (-.16 for the lower chamber; -.22 for the upper

27 The experts’ importance exerts an indirect influence via the importance (parliament), potential conflict, andthe intermediary votes. 28 On the basis of the available data we can not determine whether MPs go back to the old compromise reachedin the pre-parliamentary consultations or come up with a new (or partially new) project that minimizesopposition.

18

chamber),29 highlighting the contribution of the pre-parliamentary phase to the building of

consensus in the parliamentary arena.

This finding sheds new light on the relationship between pre-parliamentary procedures and

conflict in the parliamentary phase. Previous studies focusing on conflict level in the

intermediary votes or in the final votes taken separately had cast some doubts on the

usefulness of the pre-parliamentary phase for conflict reduction (see for instance Sciarini,

Nicolet, and Fischer 2002). Taking into account the relationships between the sub-phases of

the parliamentary process, our analysis tends to demonstrate that the effect of consultations is

overall congruent with Neidhart’s implicit assumption about the consensus-building effect of

the pre-parliamentary arena. This finding also underscores the usefulness of structural

equation modeling. By allowing for a more comprehensive approach of causal links, it

enables us to bring to light a subtle relationship between pre-parliamentary consultations and

conflict in the parliamentary phase.

While pre-parliamentary consultation reduces conflict, it does not modify the “hierarchy of

conflict” among the different acts: In both chambers, the impact of potential conflict on

conflict in intermediary votes and in final votes is positive. In other words, acts that were the

most conflict-prone at the beginning of the process remain so at the end.30

Finally, our results show that internationalization exerts almost no impact on conflict and

consensus in the parliamentary arena. Internationalization has a very weak indirect influence

(-.07 for the National Council; -.06 for the Council of States) and no significant direct effect

on conflict in the final votes. In addition, its impact on conflict in the intermediary votes is

also limited. The direct effect of internationalization favoring consensus (-.19 for the lower

chamber; -.16 for the upper chamber) shrinks once one accounts for its indirect effect exerted

via the importance of an act, resulting in a total impact of -.05 in both chambers.31 This

29 Based on figure 1 (National Council), the effects of degree of consultation on conflict in the final votes are thefollowing:

- direct effect: -.25- indirect effect: .25 * .36 = .09- total effect: -.25 + .09 = -.16

30 Note that this result does not necessarily contradict the one regarding the impact of the pre-parliamentaryconsultations. It simply suggests that the ordering of acts with respect to the level of conflict hardly variesbetween the initial phase and the end of the legislative process. But pre-parliamentary consultations proceduresmay nevertheless lead to an overall reduction of the level of conflict.31 Based on figure 1 (National Council), the effects of internationalization on the intermediary votes are thefollowing:

- direct effect: -.19- indirect effect: .40 * .35 = .14- total effect: -.19 + .14 = -.05

19

finding contradicts our two research hypotheses that posited either a positive or a negative

effect of internationalization.

With respect to the comparison between the two chambers, our results point towards the

existence of a similar dynamic. With the exception of the higher degree of consensus

regarding popular initiatives in the upper chamber, there appears to be no significant

difference in the determinants of conflict and consensus in the two houses. In other words,

while the level of conflict is overall higher in the National Council than in the Council of

States (see table 1), both the dynamics of conflict reduction and the determinants thereof seem

to be quite similar in the two chambers.

Analysis of the determinants of conflict and consensus: the referendum phase

We now turn to the analysis of the determinants of conflict and consensus in the referendum

phase. Among the 210 objects under study, only 40 were submitted to a popular vote. This

narrow set of objects raises statistical problems. Therefore, rather than limiting our analysis to

the popular votes we choose instead to study conflict in the referendum phase by focusing on

the objects subject to the optional referendum. We distinguish legislative acts for which a

referendum was successfully launched from objects for which the referendum failed or was

not asked. Given that the dependent variable is dichotomous (successful launching of a

referendum or not), we use a binary logistic regression.

The model tested here rests on the same variables as the ones included in the analysis of the

determinants of conflict and consensus in the parliamentary phase.32 In addition, in order to

capture the effect of the dynamic of conflict between the chambers on the referendum phase,

two variables measuring the number of “navettes” and the existence of a conciliation

conference are added to the model. Furthermore, owing to a high level of collinearity between

the two measures of importance used previously, only the measure of perceived importance

by the experts is kept.

{TABLE 5}

On the basis of the results presented in table 5, it appears that only two variables significantly

affect the likelihood of successful launching of a referendum. First, and in agreement with our

32 The dummy variable “popular initiative” is excluded from the model. Given that the analysis is restricted tothe objects submitted to the facultative referendum, this distinction no longer makes sense.

20

first research hypothesis (1a), acts that have a high degree of importance are more likely to be

conflictual and thus to be attacked by a referendum.33

Secondly, the results of the binary regression provide some support for the hypothesis that the

parliamentary phase influences the level of conflict in the referendum phase. As expected in

hypothesis 2c, the more divided the representatives are in the final vote, the higher the risk of

a referendum. However, our results show that only the level of consensus in the lower

chamber has a significant impact on the launching of the referendum. This finding

corroborates in a multivariate setting the relationship observed by Sciarini and Trechsel

(1996) between votes in the National Council and optional referendum on a bivariate basis.

This link can be explained once more by the greater representativeness of the National

Council in contrast to the more conservative Council of States. A solution approved by the

different interest groups represented in the lower house is likely to hold through the

referendum phase, highlighting the consensus-building capacity of the decision-making

process.

In addition, the analysis shows that it is only the level of consensus reached at the end of the

deliberations in the lower chamber (final votes) that affects the course of the referendum

phase. Conflict level in the intermediary votes for its part does not have a significant effect on

the launching of the referendum. The different nature of the two votes is likely to explain this

result. As evidenced in the structural equation models, the intermediary votes give an

opportunity to MPs to once more discuss the details of a legislative act. In contrast, in a final

vote MPs have to approve or reject a legislative act in its entirety. Final votes provide a more

reliable signal regarding elite's attitude regarding a project. It is thus not surprising that it is

these votes rather than the intermediary ones that directly affect the success or failure of a

referendum.

The effects of the importance of an act and of the final votes are summarized visually in graph

1: The predicted probability of a successful referendum is calculated according to different

degrees of importance (horizontal axis: minimal value = 1, maximal value = 5) and for two

levels of consensus in both chambers’ final votes (maximal consensus = 0, minimal consensus

= 50). All other variables are held constant at their mean.

{GRAPH 1}

33 Note, however, that the measure of importance used here could be problematic for acts that were actuallychallenged by a referendum: Indeed, the experts who were consulted are probably more likely to give a highdegree of importance to an object knowing that it has been attacked by a referendum.

21

As visible in graph 1, for acts with no or little importance the level of conflict and consensus

in the final votes has no influence on the launching of the referendum. However, for more

important acts, the degree of cohesion in the National Council has a marked effect. To take an

example, for an important object (value: 4) the likelihood of a successful launching goes from

2% when the National Council is unanimous to 70% when it is maximally divided.

The other independent variables included in the model have no significant effect on our

dependent variable (see table 5). However, given the small number of cases included in the

analysis (176 acts subject to the optional referendum among which only 11 cases of

successful launchings), it is problematic to base our interpretation only on the significance

tests. If one takes into account the direction and the values of the coefficients, a number of

patterns are visible that are summarized in graph 2.

{GRAPH 2}

First, one notices an impact of the degree of internationalization on the level of consensus in

the referendum phase. The more internationalized an act is, the less it is likely to be attacked

by a referendum, lending support to hypothesis 3a of a reduction in conflict for more

internationalized acts. However, this effect only affects more important objects. For the most

important acts, the passage from a minimal degree of internationalization to a maximal one

creates a difference of almost 50% in the probability of a referendum being launched.

Obviously, this effect should not be over-interpreted since the predicted probabilities are

calculated on a limited number of cases. Nevertheless, the curves in graph 2 suggest that the

objects that are both important and internationalized lead to less conflict than the others.

Secondly, the graph confirms the limited impact of pre-parliamentary consultations on

conflict. Their effect is not only non significant (table 5), but their magnitude is also weak

(graph 2). Even for the most important acts, the predicted probability of successfully

launching a referendum differs less than 10% between acts submitted to no pre-parliamentary

consultation and acts submitted to a maximum of five different kinds of consultation

procedures. This finding tends to question the existence of a relationship between pre-

parliamentary consultation and acceptability of an act in the referendum stage, a relationship

assumed in Neidhart’s famous hypothesis (1970). The pre-parliamentary phase seems to have

22

a sharper effect on consensus-building in the parliamentary arena than in the referendum

stage.34

5. Conclusion

Although the consensus-building capacity of the Swiss decision-making system has often

been praised, only few studies embrace a comprehensive view of the decision-making process

and examine which factors favor or hinder the emergence of compromise. This contribution

aims at overcoming this weakness by offering a quantitative analysis of the determinants of

conflict and consensus in the different stages of the decision-making process. Relying on a

simple but comprehensive model, we examine how conflict-conducive factors and

compromise-facilitating institutional mechanisms affect consensus-building. Moreover, we

analyze which effect the changing context of internationalization has on conflict level. The

use of statistical techniques that are for the first time applied to the Swiss decision-making

process, as well as the richness of the collected data, permit to bring to light a number of

interesting findings. Some of them confirm and refine conclusions drawn in previous studies,

but others lead us to qualify these conclusions.

First, regarding the role played by the intrinsic characteristics of a legislative act, our study

confirms the overriding impact of the importance of an act on its level of conflict. In line with

previous studies, we find that important legislative acts lead to higher levels of conflict in all

phases of the decision-making process.

While intrinsic characteristics such as the importance of an act favor conflict, pre-

parliamentary procedures are supposed to facilitate the search for consensus. Our analysis

provides new evidence regarding the influence of the pre-parliamentary stage. Overall, our

findings tend to support the implicit assumption of Neidhart’s hypothesis about the

consensus-building effect of pre-parliamentary consultations. However, the relationship is

subtler than what is commonly assumed. While consultation mechanisms favor the emergence

of compromise among MPs in the final stage of the parliamentary process, the effect is

opposite in the article-by-article discussion, where objects submitted to consultation are again

intensely debated. In addition, our findings show that, although Neidhart developed his

hypothesis with the referendum phase in mind, the consensus-building effect of pre-

34 Since, as noticed before, parliamentary consensus itself leads to a strong reduction of the likelihood of areferendum, one could think that part of the consensus-building effect of pre-parliamentary consultations isexerted through its influence on the parliamentary arena. However, additional regression models show that evenwhen one excludes the influence of the parliamentary stage from the model, pre-parliamentary consultations donot exert any significant effect on the likelihood of a referendum.

23

parliamentary consultations is strongest in the parliamentary phase, the compromise reached

in this phase mitigating in turn the risk of a subsequent successful referendum.

By contrast, our results regarding the impact of internationalization on decision-making

processes are not clear-cut. On the one hand, the level of internationalization of a legislative

act has only minimal effects on the level of conflict in the parliamentary phase. On the other

hand, highly internationalized acts display a greater level of consensus in the referendum

phase. However, this result is statistically hardly significant. The clarification of the effects of

internationalization will require further research.

A comparison between the dynamics of conflict and consensus in the two houses of

parliament reveals few differences in their functioning. While the level of conflict is higher in

the lower house owing to the greater diversity of its representation, similar influences explain

conflict and consensus in both houses. The two houses differ only with regard to their

treatment of popular initiatives, which raise more consensuses against them in the more

conservative Council of States. While our contribution highlights the similarities of the

mechanisms at work in both houses, the available empirical evidence does not allow for a

study of the dynamics of conflict and consensus between the two houses, that is of the

influence of one chamber on the other. This is an issue that also deserves further inquiry.

Finally, our contribution shows the importance of developing more comprehensive studies of

decision-making processes. In particular, it underscores the need for models that account for

the relationships between the different phases of the process. Our model developed on the

basis of the relatively complex Swiss institutional design could be profitably adapted and

applied to other decision-making systems, at the national or at the supra-national (European)

level.

24

Appendix: Definitions

The importance, internationalization, and potential conflict of an act were all assessed on a

scale varying between 0 “very low” and 4 “very high”.

The importance of an act was defined in the following way in the questionnaire sent to

experts of Swiss politics.

• “The importance of an act is defined as the importance of the content of this act and of

the specific measures proposed, and not as the importance of the more general issues

surrounding the act.

• We are interested in the importance of an act for the country as a whole and not for you

personally nor for the organization that you represent.

• We ask you to evaluate an act’s importance at the end of the parliamentary debates,

that is after the possible modifications made by the Federal Assembly to the project

presented by the Federal Council.”

The internationalization of an act was defined in the following way in the questionnaire sent

to experts of Swiss politics.

• “The degree of “internationalization” refers to the position of an act on an axis whose

extremes represent on the one hand an act for which Switzerland disposes of a very wide

room for internal maneuver ( very low degree of internationalization) and on the other

hand an act strongly influenced by the international environment (world, European) (

very high degree of internationalization). There exist varying degrees of

internationalization between these two extremes.

• Next to the objects that come formally within the foreign policy domain (convention or

international treaty), there exist domestic policy objects that are characterized by a high

degree of internationalization, that is objects for which Switzerland’s room for maneuver

is largely influenced by the international context. This influence can be direct (for

instance: legislative adaptations linked to an international agreement) or indirect (for

instance: anticipation of the expected effects of an act.”

The potential level of conflict of an act was defined in the following way in the questionnaire

sent to civil servants.

• “Potential conflict is defined as the level of conflict that an act was likely to produce

given its intrinsic features. We ask you thus not to evaluate the level of conflict actually

25

generated by an act for instance in the parliamentary phase but to assess the potential

level of conflict of an act at the outset of the process, that is when the first proposal for

the act is drafted.

• The level of conflict of an act is “high” when this act is likely to produce a strong

polarization among the elite. Conflict is “low” when this act is likely to produce a broad

consensus for (or against) it.”

26

References

Arbuckle, James L., and Werner Wothke. 1999. Amos 4.0 user's guide. Chicago: SmallWatersCorporation.

Armingeon, Klaus. 1998. The impact of globalization on Swiss policy making: A comment.Swiss Political Science Review 4 (2):104-111.

Bendor, Jonathan B. 1985. Parallel systems: Redundancy in government. Berkeley, CA:University of California Press.

Busch, Andreas. 1998. Unpacking the globalization debate: Approaches, evidence and data.In Demystifying globalization, edited by D. Marsh and C. Hay. London: Macmillan.

Byrne, Barbara M. 2001. Structural equation modelling with AMOS: Basic concepts,applications, and programming. Mahwah, N.J.: L. Erlbaum.

Cerny, Philip. 1997. Globalization and politics. Swiss Political Science Review 3 (4):122-128.Deutsch, Karl W. 1976. Die Schweiz als ein paradigmatischer Fall politischer Integration.

Bern: Haupt.Germann, Raimund E. 1994. Staatsreform. Der Übergang zur Konkurrenzdemokratie. Bern:

Haupt.Grote, Jürgen, and Philippe Schmitter. 1999. The renaissance of National Corporatism:

Unintended side - effect of European Economic and Monetary Union or calculatedresponse to the absence of European Social Policy? transfer 1-2:34-63.

Häusermann, Silja, André Mach, and Yannis Papadopoulos. 2002. Reconfiguration ofnational decision-making structures under external and financial pressure: The Swisscase. Paper read at ECPR Joint Sessions of Workshops, at Turin, Italy.

Hefeker, Carsten. 1997. Globalization: New or déjà vu? Swiss Political Science Review 3(4):116-121.

Huber-Hotz, Annemarie. 1991. Das Zweikammersystem - Anspruch und Wirklichkeit. In LeParlement - Autorité suprême de la Confédération?, edited by M. Bovey Lechner, M.Graf and A. Huber-Hotz. Bern: Haupt.

Hug, Simon, and Pascal Sciarini. 1995. Switzerland - still a paradigmatic case? In Towards anew Europe: Stops and starts in regional integration, edited by G. Schneider, P.Weitsman and T. Bernauer. New York: Praeger/Greenwood.

Katzenstein, Peter J. 1984. Corporatism and change. Austria, Switzerland, and the politics ofindustry. Ithaca and London: Cornell University Press.

Katzenstein, Peter J. 1985. Small states in world market. Industrial policy in Europe. Ithacaand London: Cornell University Press.

Keohane, Robert O., and Helen V. Milner, eds. 1996. Internationalization and domesticpolitics. Cambridge: Cambridge University Press.

Kriesi, Hanspeter. 1980. Entscheidungsstrukturen und Entscheidungsprozesse in derSchweizer Politik. Frankfurt: Campus Verlag.

Kriesi, Hanspeter. 1995. Le système politique suisse. Paris: Economica.Kriesi, Hanspeter, Claude Longchamp, Florence Passy, and Pascal Sciarini. 1993. Analyse de

la votation fédérale du 6 décembre 1992 (EEE). Genève/Berne: Département descience politique/GfS-Forschungsinstitut.

Landau, Martin. 1969. Redundancy, rationality, and the problem of duplication and overlap.Public Administration Review 29 (4):346-358.

Lehner, Franz. 1984. Consociational democracy in Switzerland: A political economicexplanation and some empirical evidence. European Journal of Political Research12:25-42.

27

Lijphart, Arend. 1974. Democracy in plural societies. New Haven: Yale University Press.

Lijphart, Arend. 1984. Democracies. Patterns of majoritarian and consensus governments intwenty-one countries. New Haven: Yale University Press.

Lijphart, Arend. 1999. Patterns of democracy. Government forms and performance in thirty-six countries. New Haven: Yale University Press.

Linder, Wolf. 1987. La décision politique suisse. Genèse et mise en oeuvre de la législation.Lausanne: Réalités sociales.

Linder, Wolf. 1998. Schweizerische Demokratie. Bern: Haupt.Linder, Wolf, Stefan Schwager, and Fabrizio Comandini. 1985. Inflation législative? Une

recherche sur l'évolution quantitative du droit suisse 1948-82. Lausanne: IDHEAP.Morand, Charles-Albert. 1987. La formation et la mise en oeuvre du droit. Pouvoir 43:73-86.Neidhart, Leonard. 1970. Plebiszit und pluralitäre Demokratie, Eine Analyse der Funktionen

des schweizerischen Gesetzesreferendum. Bern: Francke.Ossipow, William. 1994. Le système politique suisse ou l'art de la compensation. In Elites

politiques et peuple en Suisse. Analyse des votations fédérales: 1970-1987, edited byY. Papadopoulos. Lausanne: Réalités sociales.

Papadopoulos, Yannis. 1997. Les processus de décisions fédéraux en Suisse. Paris:L'Harmattan.

Poitry, Alain-Valéry. 1989. La fonction d'ordre de l'Etat. Analyse des mécanismes et desdéterminants sélectifs dans le processus législatif suisse. Berne: Lang.

Ricklin, Alois, and Alois Ochsner. 1984. Parlament. In Manuel système politique de la Suisse,edited by U. Klöti. Bern: Haupt.

Scharpf, Fritz W. 1996. Föderalismus und Demokratie in der transnationalen Ökonomie. InPolitische Theorien in der Ära der Transformation, edited by K. v. Beyme and C.Offe. Opladen: Westdeutscher Verlag.

Sciarini, Pascal. 1994. La Suisse face à la Communauté européenne et au GATT. Le cas-testde la politique agricole. Genève: Georg.

Sciarini, Pascal. 1999. La formulation de la décision. In Handbuch der Schweizer Politik,edited by U. Klöti, P. Knoepfel, H. Kriesi, W. Linder and Y. Papadopoulos. Zürich:NZZ Verlag.

Sciarini, Pascal, Sarah Nicolet, and Alex Fischer. 2002. L'impact de l'internationalisation surles processus de décision en Suisse. Revue Suisse de Science Politique 8 (3/4):1-34.

Sciarini, Pascal, and Alexander H. Trechsel. 1996. Démocratie directe en Suisse: L'élitepolitique victime des droits populaires. Revue Suisse de Science Politique 2 (2):201-232.

Steiner, Jürg. 1974. Amicable agreement versus majority rule: Conflict resolution inSwitzerland. Chapel Hill: University of North Carolina Press.

Trechsel, Alexander H., and Pascal Sciarini. 1998. Direct democracy in Switzerland: Do elitesmatter? European Journal of Political Research 33:99-124.

Trivelli, Laurent. 1975. Le bicamérisme. Institutions comparées. Etude historique, statistiqueet critique des rapports entre le Conseil national et le Conseil des Etats. Lausanne:Diffusion Payot.

Tsebelis, Georges, and Jeannette Money. 1995. Bicameral negotiations: The navette system inFrance. British Journal of Political Science 25 (1):101-129.

Tsebelis, Georges, and Jeannette Money. 1997. Bicameralism. Cambridge: CambridgeUniversity Press.

28

Tables

Table 1. Average level of conflict in the different phases of the decision-making process(standard deviation in parentheses)Average level of conflict Lower chamber Upper chamber

Initial level of conflict (Potential conflict) 22.51(16.71)

Parliamentary phase

Intermediary votes 20.66(16.90)

12.80(15.61)

Ensemble votes 12.73(15.34)

3.41(7.44)

Final votes 11.75(13.73)

3.20(7.44)

Referendum phase 6.09(13.29)

N 20735

Table 2. Average level of conflict in the referendum phase for objects submitted to a popularvote (standard deviations in parentheses)

Optional ref. Mandatory ref. Popular initiative

Referendum phase 37.45(7.11)

26.39(12.85)

31.35(8.95)

N 11 12 17

Table 3. Percentage of objects for which a “navette” or a conciliation conference have takenplaceNumber of “navettes” (Shuttling) %

0 50

0.5 19

1 14

1.5 6

2 4

Conciliation conference 7TotalN

100210

35 The number of cases included in the analysis is reduced to 207 here. This is explained by the fact that in thecases of popular initiatives against which a counter-project is prepared, the Federal Assembly pronounces itselfboth on the initiative and on the counter-project in a single final vote, a “yes” vote meaning the rejection of theinitiative and the acceptation of the counter-project. In this situation (3 cases), we counted the final vote as voteon the counter-project and excluded the popular initiative from the analysis.

29

Table 4. Structural equation modeling for level of conflict in the parliamentary phase;National Council and Council of State (standardized coefficients)Relationship Lower chamber Upper chamberImportance (experts) Pop. initiative .25*** .25***

Importance (experts) Internationalization .40*** .40***

Potential conflict Importance (experts) .57*** .57***

Potential conflict Pop. initiative .16*** .16***

Consultation degree Potential conflict .41*** .41***

Consultation degree Pop. initiative -.35*** -.35***

Importance (parliament) Pop. initiative .20*** .20***

Importance (parliament) Importance (experts) .43*** .43***

Intermediary votes Consultation degree .25*** .14**

Intermediary votes Importance (parliament) .29*** .15**

Intermediary votes Importance (experts) .35*** .27***

Intermediary votes Internationalization -.19*** -.16**

Intermediary votes Potential conflict .05 .25***

Intermediary votes Pop. initiative -.01 -.30***

Final votes Intermediary votes .36*** .37***

Final votes Consultation degree -.25*** -.28***

Final votes Importance (parliament) .21*** .05

Final votes Potential conflict .33*** .31***

N 207 207

Chi-square 12.288 15.338

Degrees of freedom 10 10

Adjusted GFI .947 .934

Comparative fit index .995 .987

RMSEA .033 .051

Note: *** p < .01; ** p < .05

30

Table 5. Binary logistic regression of the determinants of the success of the launching of theoptional referendum (unstandardized coefficients; standard errors in parentheses)

Determinants

Importance (experts) 2.94**(1.30)

Internationalization -.53(.51)

Potential conflict .33(.48)

Consultation degree -.12(.41)

Parliamentary conflict within the chambers

Intermediary votes (lower house) .04(.06)

Final votes (lower house) .10*(.05)

Intermediary votes (upper house) -.03(.04)

Final votes (upper house) -.00(.05)

Parliamentary conflict between the chambers

Number of navettes -.38(1.02)

Conciliation conference -8.52(42.34)

Constant -14.84(4.39)***

N 176

Nagelkerke R-squared .54

Percentage predicted correctly 94.9

Note: *** p < .01; ** p < .05; p < .1

31

Figures

Figure 1. Structural equation modelling for level of conflict in the parliamentary phase;

National Council (standardized coefficients)

p o p u la r in i tia tiv e

.2 2im p o r ta n c e (e x p e rts )

.3 9p o te n t ia l c o n f lic t

in te rn a tio n a liz a tio n

.2 0c o n su lta t io n d e g re e

.3 8in te rm e d ia ry v o te s

.4 4fin a l v o te s

.2 6im p o r ta n c e (P a r lia m e n t)

e 3

e 5

e 6

e 7

e 8

.2 5

.5 7

.1 6

.4 1

.3 6

.2 0

-.2 5.2 1

.3 3.2 5

.2 9

.3 5.4 3

.0 5

-.0 1

e 1

.4 0

-.3 5

- .1 9

Figure 2. Structural equation modelling for level of conflict in the parliamentary phase;

Council of States (standardized coefficients)

p o p u la r in i tia t iv e

.2 2im p o r ta n c e (e x p e rts )

.3 9p o te n t ia l c o n f lic t

in te rn a tio n a liz a tio n

.2 0c o n s u lta t io n d e g re e

.3 1in te rm e d ia ry v o te s

.3 0fin a l v o te s

.2 6im p o r ta n c e (p a rlia m e n t)

e 1

e 2

e 3

e 6

e 4

e 5

.2 5

.5 7

.1 6

.4 1

- .3 5

.3 7

.2 0

- .2 8

.3 1

.2 5

.1 4

- .3 0

.1 5

.2 7.4 3

.0 5

.4 0

- .1 6

32

Graphs

Graph 1. Probability of a successful referendum launching given the importance of an act and

the level of parliamentary consensus

0.00

0.20

0.40

0.60

0.80

1.00

1.00 2.00 3.00 4.00 5.00

Importance

NC_consensus NC_conflict CS_consensus CS_conflict

Graph 2. Probability of a successful referendum launching given the degree of

internationalization of an act and the level of pre-parliamentary consultations

0.00

0.20

0.40

0.60

0.80

1.00

1.00 2.00 3.00 4.00 5.00

Importance

Pred

icte

d Pr

obab

ility

Internat_min Internat_max

Consult_min Consult_max