Power-Sharing and Consociational Theory [1st ed.] 978-3-030-14190-5;978-3-030-14191-2

Table of contents :
Front Matter ....Pages i-xv
Introduction (Brighid Brooks Kelly)....Pages 1-17
What Is Consociation? (Brighid Brooks Kelly)....Pages 19-51
Front Matter ....Pages 53-53
The Plural Societies for Which Consociation Is Prescribed (Brighid Brooks Kelly)....Pages 55-72
Empirically Depicting Consociation’s Role (Brighid Brooks Kelly)....Pages 73-113
Quantitative Results (Brighid Brooks Kelly)....Pages 115-140
Comparison of this Project’s Quantitative Findings with Others’ Evaluations of Consociation’s Effects (Brighid Brooks Kelly)....Pages 141-182
Front Matter ....Pages 183-183
Supposed Cases of Historical Success Experiencing Significant Instability: Canada and Belgium (Brighid Brooks Kelly)....Pages 185-237
Some of Consociation’s Most Celebrated Successes: Switzerland, the Netherlands, and Austria (Brighid Brooks Kelly)....Pages 239-268
Newer Political Systems Yielding Valuable Insights Concerning Consociation: South Africa and Northern Ireland (Brighid Brooks Kelly)....Pages 269-339
Conclusions: How Can Stability Be Achieved Most Efficiently in Plural Societies? (Brighid Brooks Kelly)....Pages 341-356
Back Matter ....Pages 357-389

Citation preview

Power-Sharing and Consociational Theory Brighid Brooks Kelly

Power-Sharing and Consociational Theory

Brighid Brooks Kelly

Power-Sharing and Consociational Theory

Brighid Brooks Kelly Andrea Mitchell Center for the Study of Democracy University of Pennsylvania Swarthmore, PA, USA

ISBN 978-3-030-14190-5 ISBN 978-3-030-14191-2  (eBook) https://doi.org/10.1007/978-3-030-14191-2 Library of Congress Control Number: 2019932964 © The Editor(s) (if applicable) and The Author(s), under exclusive licence to Springer Nature Switzerland AG 2019 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Palgrave Macmillan imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

This Work Is Dedicated To My Mother and Grandfather, Brighid Brooks and Patrick Dougherty, Who Inspired My Interest In Helping Divided Societies

Acknowledgements

After learning of Arend Lijphart’s work involving plural societies and his theory of consociation, my first impressions were of the extent to which he has worked to benefit divided societies and stimulated academic interest in the conflicts which afflict them, long before it was fashionable to study “ethnic conflict.” Like so many others throughout the world, I will always feel inspired by his thought-provoking consociational theory. I would like to acknowledge the help and advice given to me concerning this book. More than anyone else, Ken Benoit, Michael Gallagher, and Brendan O’Leary, assisted and encouraged my analysis of consociation. Substantial, valuable feedback was given to me by Paul Mitchell and Robert Thomson. Additional helpful advice was provided by Laura Blankertz, Kathleen Cavanaugh, Michele Commercio, Donald Horowitz, and Rupert Taylor. Allahan Lee, a former legislative researcher at the Canadian Parliament, facilitated my research of consociation in Canada. Dagmar Zuefle and Joe Mariani assisted my final proofreading. My husband, Raymond Kelly, has supported my research in every way he could, for many years. I am also grateful for the encouragement provided by my father, John S. J. Brooks Jr., Sandy Sparrow, my brothers, Patrick and Michael Brooks, and my children, Seamus, Kathleen, and Brighid.

vii

Contents

1 Introduction 1 2

What Is Consociation? 19

Part I  A Quantitative Assessment of Consociation 3

The Plural Societies for Which Consociation Is Prescribed 55

4

Empirically Depicting Consociation’s Role 73

5

Quantitative Results 115

6

Comparison of this Project’s Quantitative Findings with Others’ Evaluations of Consociation’s Effects 141

Part II A Qualitative Examination of Consociation’s Role in Seven Cases 7

Supposed Cases of Historical Success Experiencing Significant Instability: Canada and Belgium 185 ix

x   

Contents

8

Some of Consociation’s Most Celebrated Successes: Switzerland, the Netherlands, and Austria 239

9

Newer Political Systems Yielding Valuable Insights Concerning Consociation: South Africa and Northern Ireland 269

10 Conclusions: How Can Stability Be Achieved Most Efficiently in Plural Societies? 341 Appendix A: T  he Plural Cases—Comprising 88 Countries and 1593 Years of Experience 357 Appendix B: The Variables and Data 361 Appendix C: The Construction of the Dependent, Stability Variable 365 Bibliography 367 Index 385

Abbreviations

ANC African National Congress (South Africa) AV Alternative Vote System DUP Democratic Unionist Party (Northern Ireland) ET External Threats HGC Highly Inclusive Grand Coalition IRA Irish Republican Army (Northern Ireland) MAR Minorities at Risk MMS Moderate Multiparty System MP Member of Parliament MV Minority Veto NCOP National Council of Provinces (South Africa) NP National Party (South Africa) PR Proportional Representation PSNI Police Service of Northern Ireland SA Segmental Autonomy SF Sinn Féin (Northern Ireland) SGC Somewhat Inclusive Grand Coalition STV Single Transferable Vote System

xi

List of Tables

Table 4.1 Table 6.1 Table 9.1

The variables representing consociational components 99 Other quantitative treatments of phenomena similar to consociation 162 Components of consociation found in South Africa’s interim and permanent constitutions 272

xiii

List of Boxes

Box 4.1 Box 5.1 Box 9.1 Box 9.2 Box 9.3

The favorable conditions identified by Lijphart Consociational cases in plural societies between 1975 and 1995 whose existence is confirmed by this dataset Lijphart’s favorable conditions for consociation and South Africa Components of consociation found in Northern Ireland after 1998 Lijphart’s favorable conditions for consociation and Northern Ireland

104 121 273 300 301

xv

CHAPTER 1

Introduction

Violent clashes between the groups in divided societies which appear to be culturally incompatible command substantial media attention because the international community finds it difficult to understand their motivations and how their antagonisms can be resolved. News reports commonly contain very limited information concerning the historical relationships between these antagonistic groups and often give the impression that either the groups have no comprehensible motive for fighting and/or ancient hatreds exist between them which cannot be quelled. These observations seem to intrigue the public even more, since it is difficult to understand why people continue to fight if they are not doing so for any rational reasons and thus seem unlikely to gain anything beneficial from their feud. Not long ago, those who believed that the most antagonistic divided societies could achieve long-term stability were commonly derided as hopelessly naïve. The current surge in academic interest concerning historically intense conflicts reflects a fundamental reversal of expectations, the initiation of which can be attributed very plausibly to the pioneering role of Arend Lijphart and his system of consociation. Acknowledging that his advocacy of this system “challenge[d] the pervasively pessimistic mood,” Lijphart asserted that “it may be difficult, but it is not at all impossible to achieve and maintain stable democratic government” in diverse, antagonistic societies.1

1 Lijphart

(1977, pp. 3, 1).

© The Author(s) 2019 B. B. Kelly, Power-Sharing and Consociational Theory, https://doi.org/10.1007/978-3-030-14191-2_1

1

2  B. B. KELLY

The academic consensus against which Lijphart rebelled in the 1970s is exemplified well by one of the few theoretical treatments of conflict in divided societies existing at that point. Following their exploration of the causes of such instability, Alvin Rabushka and Kenneth A. Shepsle conclude that “[h]istory shows that democratic stability and cultural diversity are often incompatible in the post-independence politics of many plural societies.”2 They believe that certain combinations of “colonial experience, exogenous events, and… ethnic configuration[s]” inevitably yield substantial instability.3 This is thought to occur when the greatest political success is awarded to those politicians who attract support by “outbidding” their rivals through “‘ethniciz[ing]’ politics, to ‘fan the flames’ of ethnic chauvinism.”4 Lijphart has argued for decades that consociation promotes stability in such divided societies. The extensive analysis of consociation presented here confirms that his optimism regarding the potential of institutional engineering is completely justified. The aim of this book is to ascertain whether it is true that consociation promotes stability in plural societies, through quantitative analysis of 88 potentially or actually antagonistic plural societies over 20 years, and qualitative analysis of some societies which have been widely considered to have benefitted from consociation. The cases examined qualitatively were chosen primarily because they have been cited often by scholars as providing evidence that consociation does perform the role ascribed to it by Lijphart. However, the selection of these case studies should not be interpreted as implying that consociation performed this function in all of them or even that consociation operated in all of them. These cases also are focused upon here because a wealth of literature regarding their experiences with consociation is available, with which to compare this study’s other findings. They are analyzed qualitatively to evaluate the role of apparently influential phenomena, such as those involving popular attitudes, which could not be tested quantitatively because of the unavailability of comparable, international data. Some of the cases are also examined qualitatively because they could not be represented in the

2 Rabushka

and Shepsle (1972, p. 207; 2009, p. 207). and Shepsle (1972, p. 92; 2009, p. 92). 4 Rabushka and Shepsle (2009, p. 220). Additional general descriptions of this phenomenon are provided here: Rabushka and Shepsle (1972, pp. 91, 92, 156, 157, 207, 217; 2009, pp. 91, 92, 156, 157, 207, 217). 3 Rabushka

1 INTRODUCTION 

3

dataset. This project emphasizes the greater explanatory value derivable through comparison of statistical findings and appropriate case studies. The wide-ranging exploration presented here confirms that institutional engineering has enhanced stability within even extremely divided societies. Multiple regression tests confirm the positive role of the consociational component corresponding to highly inclusive governance executives. In models including all independent variables found to exert statistically significant effects on stability, one corresponding to a ­consociational component remains significant. This component is minority veto power, but its coefficients in all regression tests indicate that its influence is destabilizing. In regression models finding statistically significant effects by segmental autonomy and proportional representation electoral systems, the signs of these variables’ coefficients also indicated that they were destabilizing. Scatterplot analyses confirm the positive effects of highly inclusive executive coalitions, emphasize the nuanced roles of the other three consociational components, and draw attention to the fact that a sizeable proportion of countries whose constitutions provide minority veto power are also less democratic. Factor analysis indicates that most of the variables representing consociational components, as well as multiple conditions identified by Lijphart as favorable to the system’s success, are influencing factors underlying the dataset. It seems possible that these two underlying factors imply that consociation is not less conducive to stability than other democracies and that somewhat inclusive executives help facilitate stability when countries with large populations face external threats. This body of quantitative ­analysis suggests that specific types of segmental autonomy, minority veto power, and proportionality may exert substantially different effects on stability. The insights discovered through this project’s case studies also suggest that the effects of these consociational components also may be profoundly impacted by conditions Lijphart identified as favorable to the success of consociation but which could not be represented quantitatively here. These qualitative case studies suggest that certain manifestations of segmental autonomy and proportional representation are, in fact, conducive to stability in plural societies. They illustrate the beneficial roles played by two of Lijphart’s favorable conditions, overarching loyalty to a shared state and cross-cutting cleavages. Collectively, the qualitative and quantitative analyses suggest that it is a specific type of proportional representation called the single transferable vote (STV) system which

4  B. B. KELLY

is conducive to stability in plural societies, rather than all proportional representation electoral systems. STV and other phenomena which have optimized stability in conjunction with components of consociation do so by combining guarantees for group autonomy with incentives for intergroup political appeals. This book shows that the ethnic cleavages identified as destructive by Rabushka and Shepsle are not incompatible with democratic stability, if the groups which comprise them are given permanent group autonomy and political rights, as well as reasons to avoid “‘fan[ning] the flames’ of ethnic chauvinism.”5 The extent to which a deeply divided political landscape can be transformed through adoption of carefully tailored institutions is illustrated well by the recent experience of a society specifically identified by Rabushka and Shepsle as incompatible with democratic ­stability. They cited Northern Ireland as providing particularly “[v]ivid,” “instructive,” compelling evidence that “dominant majority societies” are highly conducive to outbidding political dynamics and consequent extremism.6 Some of the most persuasive examples identified by them to support this theory are the “successes of” of Ian Paisley, a Northern Irish political entrepreneur who built his career on the cultivation and representation of intergroup hatred.7 Rabushka and Shepsle’s conclu­ sion published in 1972, that some divided societies are incompatible with real democracy, reflected widespread public opinion at that time. Even as Lijphart was prescribing consociation for other divided societies in the 1970s, he believed that Northern Ireland was an especially intractable case and surmised that the “only” course of action capable of “result[ing] in a durable solution” for Northern Ireland was its dismantlement through partition.8 The pessimistic prediction offered by Rabushka and Shepsle seems obviously incorrect now, precisely because various power-sharing reforms, including elements of consociation, have demonstrated in Northern Ireland and elsewhere that democracy is possible in even the most divided societies. Incentives and protections provided through a power-sharing system motivated Northern Ireland’s population, and even Ian Paisley, to do what had been considered

5 Rabushka

and Shepsle (2009, p. 220). and Shepsle (1972, pp. 82, 83, 143, 6, 151–153). 7 Rabushka and Shepsle (1972, pp. 151–152). 8 Lijphart (1975, p. 105). 6 Rabushka

1 INTRODUCTION 

5

impossible by Rabushka and Shepsle, as well as the worldwide consensus whose assumption they articulated. Consideration of the dramatic political turnaround of Ian Paisley emphasizes the rationale for, and value of, the quantitative and qualitative analysis presented in this book. This is because it demonstrates that institutional reforms are capable of producing profound behavioral change, even in the most extreme political entrepreneurs operating in the most divided societies. Between 2004 and 2007, Paisley transformed from being one of the world’s most notorious cultivators of intergroup antagonism to playing an instrumental role in resolving the very conflict that he had encouraged for fifty years. According to the sectarian belief system he led for most of his life, his decision to join the Northern Irish power-sharing executive in 2007 was impossible. Paisley was so influential in the conflict that he managed to dominate its portrayal in the world media. He characterized it as motivated by religion, apparently because religion is generally recognized as an intrinsically nonnegotiable type of intergroup boundary marker.9 Widespread acceptance of this portrayal is remarkable, considering the true nature of the conflict. Most nationalists, who favored the region’s merger with the Republic of Ireland, are Catholic and those opposing this historically have been predominantly Protestant. However, John McGarry and Brendan O’Leary demonstrate conclusively that this conflict’s consistent motivation was its centuries-old constitutional disagreement, rather than the religious, cultural, or economic perspectives which also separate the two groups.10 Discussions of the region’s constitutional status make for much less intriguing headlines than Paisley’s dramatic rhetoric claiming that his Unionist group hates the Nationalist population because it is Catholic, the religious beliefs of the Free Presbyterian church he led preclude contact with Catholics, and the Pope is the Antichrist.11 This extremism was evident when he initially condemned the European Common Market by claiming that the Catholic Church was involved in its formation, the Pope is the Devil, and so “Satanic Power” is the element which “holds” it “together.”12 At one meeting of the European Parliament in 1988, he sought media 9 Rabushka

and Shepsle (1972, p. 6). evidence demonstrating this statement is provided throughout this publication: McGarry and O’Leary (1995). 11 Moloney (2008, pp. 209, 222, 326). 12 Moloney (2008, p. 221). 10 The

6  B. B. KELLY

attention by repeatedly shouting to the then Pope, “I renounce you as the Antichrist.”13 From the beginning of his career, Paisley encouraged prejudice and violence through initiatives like his 1950s “mock Masses” throughout Northern Ireland, ridiculing Catholic worship practices, beliefs, and priests.14 The official publication of his church depicted Catholics as deranged savages and claimed the Catholic Church is “the mother of harlots and the abominations of the earth.”15 In 1964, the worst riots since the 1930s were provoked when Paisley demanded that a Nationalist flag be removed from public view.16,17 Similarly, when he was sent to jail in 1966 after a demonstration led by him precipitated riots, even more “savage rioting” by his enraged followers ensued.18 While listening to Paisley’s speeches, “many of his hearers simply thought that victims were being pointed out to them.”19 These tactics are familiar to comparativists studying severe intergroup conflicts. However, Paisley’s subsequent transformation is quite unusual, and it constitutes an excellent illustration of the tremendous potential power of carefully tailored political institutions.20,21 Paisley’s willingness to move away from extremism in exchange for political power, combined with his masterful cultivation of support, meant that the emergence of such an opportunity persuaded him and most of his followers to share political power equally with those who he had portrayed as devil worshippers. This development is described in the introduction of this book because it constitutes particularly compelling evidence that institutional engineering can fundamentally alter the 13 Moloney

(2008, p. 326). (2008, p. 83). 15 Moloney (2008, p. 84). 16 Farrell (1980, p. 234). 17 Paisley was identified as the instigator of these riots (Boyd 1969, pp. 181–182). 18 Buckland (1981, p. 120) and Farrell (1980, p. 235). 19 Buckland (1981, p. 120). 20 This short discussion of Paisley’s rhetoric is referred to here only because it emphasizes the extent to which his political decisions were influenced by institutional engineering. 21 & Footnotes beginning with an ampersand indicate that additional relevant information is provided on the website associated with this book. Such a footnote is placed here because the website indicates a valuable resource of evidence demonstrating that there is not, and never has been, a causal relationship between religion and the conflict in Northern Ireland. 14 Moloney

1 INTRODUCTION 

7

political dynamics of even the most divided societies. Paisley’s career trajectory illustrates that appropriate incentives can dramatically change the strategic approach of even those political leaders who historically have contended that they are primarily motivated by nonnegotiable beliefs. His achievements also emphasize that it is incorrect to assume that the vast majority of extremists’ followers will reject their leaders’ movements toward political moderation. Having repeatedly condemned the European Common Market through reference to religion, Paisley successfully ran for office and joined the European Parliament in 1979.22 Ed Moloney explains that this “volte-face was difficult for many in his church to stomach” but “Paisley got his way” and continued on as leader of his church and political party, despite the strongly voiced objections of his three most senior church lieutenants.23 The “central, defining characteristic” of his church is “its belief in the doctrine of ‘separation’ which taught Free Presbyterians to have absolutely no dealings” across the region’s community divide.24 Having firmly reiterated for decades his determination never to contemplate accepting his political party’s joint executive control with a nationalist party, he did just that with the most extreme one, Sinn Féin.25 Moloney’s wide-ranging analysis of Paisley’s life suggests that “he went into government with Sinn Féin because he could.”26 After assuming joint executive office with Martin McGuiness, a former Chief of Staff of a long-dominant paramilitary nationalist group, Paisley revealingly reminisced to the press: “I never thought I would be in a place where I could really influence governments the way I wanted to influence them.”27 Several theories have been offered to explain Paisley’s willingness to share power with nationalists after forbidding mere contact with them since the beginning of his career. Moloney believes that Paisley did so because he desired executive power and his attainment of this goal was made possible by a new set of political circumstances, including many elements of consociation. Paisley’s actions prove that it is incorrect to assume that the nature of any society’s divisions renders it inherently incompatible with stable, true democracy. 22 Moloney

(2008, pp. 221–222). (2008, pp. 222, 221). 24 Moloney (2008, p. 222). 25 Paisley (2006). 26 Moloney (2008, p. 514). 27 Curran (2007). 23 Moloney

8  B. B. KELLY

This book is dedicated to the quantitative and qualitative exploration of consociation’s role because its components arguably are responsible for engendering such democracy in societies in which this achievement had been considered impossible due to the pervasiveness and intensity of historical cleavages. This study attempts to ascertain whether consociation is the optimal system of political power-sharing for divided societies. It constitutes the first comparative, empirical analysis in which all four elements of Lijphart’s consociation are portrayed realistically, and his exact claim for the system is evaluated. If it can demonstrate the extent to which consociation promotes stability, the analysis presented here will assist in the identification of optimal strategies for policymakers, will facilitate international comprehension of severe intergroup conflict, and thus will constitute a substantial contribution to the cultivation of long-term peace in persistently antagonistic societies. As conclusions regarding the most advantageous means of designing power-sharing systems are gradually illuminated, those who design and modify governance institutions presumably will become more able, and more obliged, to do so in accordance with internationally agreed upon guidelines. The theoretical coherence and validity of Lijphart’s consociation already have been extensively analyzed by many academics, ever since its introduction in 1968.28 Although some criticisms of consociation arguably have not been effectively discredited, the system is widely believed to be the only feasible alternative to oppressive forms of majoritarianism and violence in deeply divided, plural societies. Consideration of proposals resembling Lijphart’s consociational model, by constitutional drafters in South Africa, Northern Ireland, and Bosnia, suggests that many policymakers and academics believe the system is capable of resolving many of the most intense intergroup conflicts. Adoption of one or more institutional elements which arguably resemble Lijphart’s consociational components has occurred at some stage in societies as diverse as Austria, Belgium, Canada, Cyprus, Fiji, Lebanon, Malaysia, the Netherlands, Northern Ireland, Nigeria, South Africa, and Switzerland.29 Virtually all of consociation’s critics have failed, and rarely even explicitly attempted, to challenge the assumption that the system promotes conflict resolution. 28 &

Additional references are on the website. (1977), Reilly (2001, pp. 104–105), Evans and O’Leary (2000), and Jung and Shapiro (1995). 29 Lijphart

1 INTRODUCTION 

9

Analyses of individual countries’ experiences with consociation have led to some negative observations concerning its value. However, international constitutional designers’ lack of serious interest in those observations seems justifiable because the apparent shortcomings of the system in individual states can be attributed persuasively to situation-specific factors, such as international intervention in their domestic affairs. This study evaluates Lijphart’s main proposition concerning consociation that it makes it possible “to achieve and maintain stable democratic government in a plural society.”30 Statistical analysis is used to evaluate this proposition because it is conducive to objective ­empirical examination of the effects of political institutions. The precise definition of consociation to be tested in this study necessitates substantial justification because slight variations of it have been published by Lijphart, reflecting his refinement of the concept. For the purposes of this study, consociation will be recognized as the operation of all, or some, of those elements comprising this system that are identified consistently by Lijphart and can be scientifically tested. These include the presence within individual states of an inclusive governing coalition, segmental autonomy, proportional representation electoral systems, and allocation of veto power to politically significant minority groups. Since Lijphart is the recognized definer of the modern concept of consociation, evaluation of the system’s effectiveness requires thorough specification of the functions that he believes that it performs. He has produced the most extensive, persuasive arguments in favor of consociation and his descriptions and revisions of it are used to justify the proposition tested here. This will enable examination of the most widely accepted definition of its goal.31 Lijphart’s general descriptions of consociation’s functions have remained consistent in his writings since 1968. His later discussions have not strayed far from this definition he gave in 1969: “Consociational democracy means government by elite cartel designed to turn a democracy with a fragmented political culture into a stable democracy.”32 While he has become less inclined to use the term, “elite cartel,” he still insists on the pivotal role of elites in establishing and maintaining the system. Chapter 3 explains that, since 1970, he

30 Lijphart

(1977, p. 1). (1998, p. 149). 32 Lijphart (1969, p. 216). 31 Lijphart

10  B. B. KELLY

usually describes the societies it helps as “plural.”33 Since Lijphart’s claim is that consociation promotes “stability,” his use of that term is analyzed in Chapter 4. This “stability” consists of four elements: system maintenance, civil order, legitimacy, and effectiveness. Thorough comprehension of Lijphart’s expectations for consociational systems is facilitated through recognition of his premise that, in deeply divided societies, social and institutional separation of potentially antagonistic population segments most effectively promotes peace in the states they share. Lijphart argued that many of Almond’s “Continental European” democracies were “fragmented but stable.”34 His theory of consociation represents his attempt to understand these states’ seemingly impossible combination of division and stability. His answer to this puzzle is: Such distinct lines of cleavage appear to be conducive to consociational democracy and political stability. The explanation is that subcultures with widely divergent outlooks and interests may coexist without necessarily being in conflict; conflict arises only when they are in contact with each other.35

Lijphart points out that Quincy Wright, David Easton, Sidney Verba, and Walker Connor agree that intergroup contact often heightens conflict and “good social fences may make good political neighbors.”36 He explains that social divisions do not necessarily produce secession or partition and the violence which often accompanies them because, in consociational systems, elites representing the potentially antagonistic segments work together to manage their shared state’s affairs and maintain stability.37 Lijphart also distinguishes between “essentially homogeneous political cultures, where contacts are likely to lead to an increase in mutual understanding and further homogenization, and essentially heterogeneous cultures, where close contacts are likely to lead to strain

33 Examples of Lijphart’s use of these terms are: Lijphart (1972, p. 426; 1977, p. 1; 1979, p. 509; 1981a, p. 355; 1981b, p. 395; 1984, p. 12; 1992, p. 217; 1999, p. 251). 34 Lijphart (1969, p. 211). 35 Lijphart (1969, p. 219). 36 Lijphart (1969, p. 219). 37 Lijphart (1969, pp. 219–220).

1 INTRODUCTION 

11

and hostility.”38 He argues that “clear boundaries between the segments of a plural society have the advantage of limiting mutual contacts and consequently of limiting the chances of ever-present potential antagonisms to erupt into actual hostility.”39 His claims concerning the beneficial effect of such boundaries are consistent throughout his work. Lijphart’s assumption that contact in such societies is conducive to instability is also shown in his discussions of individual countries’ experiences. In response to Brian Barry, who argued that consociation is less appropriate for ethnically divided societies, Lijphart states that the “highly visible differences” often existing between ethnic groups “may also serve the positive function of clearly demarcating segmental boundaries- thus limiting intersegmental contacts and minimizing opportunities for intersegmental conflict.”40 Many specialists focusing on the Northern Ireland conflict believe that the region’s tight congruence of multiple cleavages and the confrontational behavior of its elites are the two main factors that have prevented stability there. In contrast to this common view, Lijphart believes that, in Northern Ireland, the “segmental isolation” that results from coincidence of cleavages is potentially beneficial and “does appear to have the function of the good fences that make good neighbors.”41 Lijphart’s theory of consociation prescribes a system with “segmental autonomy” as one of its four core elements and identifies “segmental isolation” as a condition favorable to establishment of the system. If the effects of consociation are to be tested in a scientifically sound manner, it also must be kept in mind that Lijphart’s view of how conflicts are resolved fundamentally differs from many who believe crossgroup contact promotes long-term stability. One of the four components of consociation is segmental autonomy. However, Lijphart’s thoughtful and nuanced approach to cross-group dynamics is emphasized by his identification of cross-cutting cleavages as a favorable condition for consociation’s success. Like most political scientists including Almond, Lijphart agrees with Seymour Martin Lipset that a society’s stability is enhanced by the presence of cleavages which cross-cut those which

38 Lijphart

(1969, p. 220). (1977, p. 88). 40 Lijphart (1977, p. 233). 41 Lijphart (1977, p. 140). 39 Lijphart

12  B. B. KELLY

are most conducive to antagonism. Cross-cutting cleavages inherently diminish societies’ potential levels of segmental isolation but also help elites to maintain electoral support when they go into government with representatives of those groups who their constituents have traditionally opposed. Lijphart believes that consociation and group separateness are necessary for the more heterogeneous countries but also seems to consider cross-cutting cleavages as beneficial for such countries because they make it easier for elites of different groups to hold their shared countries together. Lijphart’s general approach to conflict resolution and the terms that he uses in his descriptions of consociation suggest that the claim that should be tested in this analysis is as follows: Consociation promotes stability in plural societies. Lijphart’s identification of individual core components of the system in countries which do not possess all of them further emphasizes that his precise claim, and the proposition to be tested here, is actually that there is a directly proportional relationship between consociation and long-term stability in plural societies. Some critics of his theory have assumed that consociation should be considered a necessary, or sufficient, condition for achievement of stability in these situations. Claims positing this relationship are more easily challenged than those asserting only a positive relationship between the extent of a society’s consociation and its stability. However, statistical analysis enables empirically sound calculation of the overall effects of consociational components internationally. Statistical analysis allows objective, identical treatment of all cases and discernment of even slight relationships between variables, so it will enable objective examination of this proposition. It is also important to distinguish between the entire body of Lijphart’s work on consociation, generally recognized to constitute his theory involving the topic, and the core components that make up the consociational governmental structure. A statistical model is constructed which enables valid testing and examination of this theory. To ascertain the effects of consociation on stability in plural societies through statistical analysis, plural s­ocieties’ years of experience must constitute cases, or units of observation. Through reference to Lijphart’s definition of plural societies, it would be possible to impressionistically differentiate between past and present societies that are, or were, plural or not plural. However, this procedure would undoubtedly introduce significant bias to this analysis. For this reason, cases for this analysis will be chosen by categorizing societies

1 INTRODUCTION 

13

according to a threshold, which was derived from a dataset compiled by another scholar, who was not particularly interested in the proposition to be tested here. Although this approach is likely to introduce some bias from this scholar’s work, complete impartiality in choosing societies seems unattainable and this bias is not likely to be directly related to consociation. Ted Robert Gurr’s Minorities at Risk (MAR) dataset contains all of the minorities in the world which he has found to conform to his definition of “minorities at risk.”42 His criteria used for identifying these groups are very similar to Lijphart’s description of which groups are required in a society for it to be considered plural. Gurr’s guidelines are consistent with Lijphart’s most commonly used definition of plural societies, those which are clearly divided into groups whose members’ collective actions reflect the obvious existence of cleavages corresponding to religion, language, culture, ethnicity, and/or race. The similarity between these two scholars’ categories suggests that the most prudent, least biased means of choosing cases for this analysis is inclusion of each country whose population contains a significant proportion of people in Gurr’s MAR. This dataset has been found to contain a number of shortcomings, and these will be discussed in Chapter 3, but their impact on this project is minimal. It is well known that the effects of constitutional systems emerge very gradually so the accuracy of this study’s findings is maximized by inclusion of two decades, from 1975 to 1995. The independent variables and data representing consociation and other phenomena will be described and justified in Chapter 4. These components, rather than consociation in general, are used as independent variables for this quantitative analysis because, as Box 5.1 illustrates, the number of cases indicating use of all components simultaneously is extremely small. Some scholars argue that consociation, defined as the simultaneous operation of all four components, has actually never occurred. Box 5.1 explains that, within this dataset, only 3% of cases corresponding to each year of each country’s experiences exhibit such use of all four components. To arrive at internationally supported conclusions about consociation, data corresponding to components of the system must be used because otherwise conclusions would have to focus on just a few, often quite short, experiences of the overall system.

42 Gurr

(1993).

14  B. B. KELLY

The dependent variables and their data have been chosen to ­represent as accurately as possible Lijphart’s most commonly used definition of stability. This phenomenon’s four elements are system maintenance, ­ civil order, legitimacy, and effectiveness. For the qualitative case studies in this book, stability is defined as comprising these four elements. Analysis of the methodological and theoretical ramifications of including all four elements in an index variable representing stability suggests that only civil order and legitimacy can be thus operationalized, without causing ­serious methodological problems. However, the effects of consociation on system maintenance will be controlled for to some extent in models containing a control variable representing cases’ placement on an ­autocracy-democracy continuum. Some additional, independent variables are included in this study to facilitate confirmation of my findings. In any statistical analysis examining whether factor A is associated with factor B, apparent relationships between these factors may actually be caused by phenomena which are not constitutive of factor A but are systematically related to it. To explain the nature of consociational systems, predict their effects, and judge the likelihood of their successful implementation and maintenance, Lijphart created a theory which attempts to achieve all of these aims. He argues that, if certain conditions do not exist in a society, it is unlikely that consociational systems will be introduced or successfully used there. He particularly emphasizes that the system will definitely not work unless elites are motivated to make this happen. A few of his other conditions include geographical concentration of population segments, cross-cutting cleavages, small population size, external threats, and overarching loyalties. Lijphart’s contention that these conditions facilitate the system has been criticized, in part because some of them, such as overarching loyalties and elite accommodation, are arguably products, rather than stimulants, of conflict resolution. However, this study will include variables corresponding to the conditions whose inclusion will not cause methodological problems, because these conditions arguably account for some stability in plural societies. The quantitative analysis performed for this book examines eightyeight plural societies’ experiences between 1975 and 1995, through employment of variables corresponding to consociation’s components, stability, and several factors which may account for some trends in stability that only appear to result from consociation. The statistical examination shows concrete, international, long-term evidence of the general relationship between consociation and stability. The main fi ­ ­ndings of

1 INTRODUCTION 

15

regression analysis are that variables representing only two consociational components remain statistically significant when all of the independent variables, corresponding to both those components and controlling phenomena, are examined together through multiple regression. The two components which remain statistically significant are highly inclusive grand coalitions, which appear to be conducive to stability, and minority veto power, which this project indicates is actually detrimental. In some models containing different sets of independent variables, PR electoral systems and somewhat inclusive coalitions are found to produce statistically significant, negative effects on stability in plural societies. The consociational component of segmental autonomy is found to exert a statistically significant effect on stability only when it is tested in conjunction with the variables representing other components of the system. In that case, its effect is destabilizing but its influence is severely diminished in the model including these variables with that representing democracy. Scatterplot and factor ­analysis indicate that all four consociational components, as well as three control variables corresponding to favorable conditions, are influential in this dataset. All of the quantitative findings regarding the extent and nature of the relationships between the phenomena tested are quite v­ aluable for several reasons. These include the extent to which these differ from existing assumptions regarding consociation’s operation, their implications involving favorable conditions which were thought to play crucial roles, and their derivation from the first international, large-scale, statistical treatment designed to faithfully represent Lijphart’s power-sharing system and his exact theory concerning its role. Following Chapter 5’s presentation of this project’s quantitative results, Chapter 6 is devoted to comparing them with a wide range of other theoretical, qualitative, and quantitative publications related to consociation and power-sharing. This project is the first using data and variables carefully chosen to maximize their representation of Lijphart’s system of consociation and claims for it. Other quantitative treatments of similar phenomena do not include dependent variables modeled after his concept of stability and many that test political power-sharing represent it using a composite variable rather than individual, independent variables corresponding to consociation’s specific manifestations. In general, the quantitative results derived for this project do not contradict most related findings derived from models which differed considerably from those explored here. This project differs from other related studies in terms of its dataset, definition of power-sharing, and set of key independent variables.

16  B. B. KELLY

Donald Horowitz believes that some policymakers “prefer” consociation to other forms of government “because its requirements are formal, easy to fathom,” and “predictable in their consequences.”43 Statistical evaluations of this system eventually will produce formal, easily understood, more reliable indications of whether consociational components promote resolution of conflicts in plural, divided societies. Part II is devoted to several case studies to maximize our understanding of consociational components through comparison of insights derived through quantitative and qualitative analysis. These cases, in the order in which they are discussed, include Canada, Belgium, Switzerland, the Netherlands, Austria, South Africa, and Northern Ireland. The general insights suggested by this book’s findings are that the group empowerment and protection provided by consociation promote long-term stability best when they are combined with additional motivations. It seems that the potential gridlock and destabilization feared by consociation’s critics can be prevented to a large extent by incentives to avoid conflict and favor intergroup negotiation. The case studies in this book indicate that, if potentially antagonistic groups do feel that they are protected by consociation, they can become increasingly amenable to sharing power and territory when cross-cutting considerations influence their political decision-making.

References Boyd, Andrew. Holy War in Belfast: A History of the Troubles in Northern Ireland. Kerry, Ireland: Anvil Books, 1969. Buckland, Patrick. A History of Northern Ireland. Dublin, Ireland: Gill and Macmillan Ltd, 1981. Curran, E. Paisley. The Ordinary Man in Street Has Come Out and Backed Me: Part Two of the Exclusive Interview with the First Minister. Belfast Telegraph. 11 June 2007. Belfast, Northern Ireland. Evans, Geoffrey and Brendan O’Leary. “Northern Irish Voters and the British-Irish Agreement: Foundations of a Stable Consociational Settlement?” The Political Quarterly. 71:1 (January–March, 2000) 78–101. Farrell, Michael. Northern Ireland: The Orange State. London: Pluto Press, 1980. Gurr, Ted Robert. Minorities at Risk: A Global View of Ethnopolitical Conflicts. Washington, DC: United States Institute of Peace Press, 1993. 43 Horowitz

(1999).

1 INTRODUCTION 

17

Horowitz, Donald L. Email to Brighid B. Kelly. 9 November 1999. Jung, Courtney and Ian Shapiro. “South Africa’s Negotiated Transition.” Politics and Society. 23:3 (September, 1995) 269–308. Lijphart, Arend. “Consociational Democracy.” World Politics. 21:2 (January, 1969) 207–225. Lijphart, Arend. “Review Article: Toward Empirical Democratic Theory: Research Strategies and Tactics.” Comparative Politics. 4:3 (April, 1972) 417–432. Lijphart, Arend. “Review Article: The Northern Ireland Problem: Cases, Theories, and Solutions.” British Journal of Political Science. 5 (1975) 83–106. Lijphart, Arend. Democracy in Plural Societies: A Comparative Exploration. New Haven: Yale University Press, 1977. Lijphart, Arend. “Consociation and Federation: Conceptual and Empirical Links.” Canadian Journal of Political Science. 12:3 (September‚ 1979) 499–515. Lijphart, Arend. “Comment: Consociational Theory: Problems and Prospects: A Reply.” Comparative Politics. 13:3 (April, 1981a) 355–360. Lijphart, Arend. “Power-Sharing Versus Majority Rule: Patterns of Cabinet Formation in Twenty Democracies.” Government and Opposition. 16:4 (Autumn, 1981b) 395–413. Lijphart, Arend. “The Politics of Accommodation: Reflections—Fifteen Years Later.” Acta Politica. 19:1 (January, 1984) 9–18. Lijphart, Arend. “Democratization and Constitutional Choices in CzechoSlovakia, Hungary, and Poland, 1989–1991.” Journal of Theoretical Politics. 4:2 (1992) 207–223. Lijphart, Arend. “South African Democracy: Majoritarian or Consociational?” Democratization. 5:4 (Winter, 1998) 144–150. Lijphart, Arend. Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries. New Haven, CT: Yale University Press, 1999. McGarry, John and Brendan O’Leary. Explaining Northern Ireland: Broken Images. Oxford: Blackwell, 1995. Moloney, Ed. Paisley: From Demagogue to Democrat? Dublin: Poolbeg, 2008. Paisley, Ian. Making History or History Made? Speech Given at an Independent Orange Demonstration. 12 July 2006. Portrush, County Antrim, Northern Ireland. Rabushka, Alvin and Kenneth A. Shepsle. Politics in Plural Societies: A Theory of Democratic Instability. Columbus, OH: Charles E. Merrill, 1972. Rabushka, Alvin and Kenneth A. Shepsle. Politics in Plural Societies: A Theory of Democratic Instability. Columbus, OH: Charles E. Merrill, 2009. Reilly, Ben. Democracy in Divided Societies: Electoral Engineering for Conflict Management. Cambridge: Cambridge University Press, 2001.

CHAPTER 2

What Is Consociation?

Consociation is now most commonly identified with Lijphart’s description and prescription of the system of government which he believes is most capable of bringing true democracy and intercultural harmony to states and regions plagued by “ethnic” conflict. Consociation is recognized as constituting both this system and the theory concerning its adoption and operation. Lijphart defines the consociational system as a governance structure composed of four core components, which is required to achieve stability in deeply divided societies and is the only form of democracy which can do so. Its components are government by all-inclusive grand coalition, proportional representation, provision of minority veto power, and cultural autonomy for subcultures. The combination of these four elements prevents groups of any size from being consistently deprived of influence over key decision-making. Groups’ continual, political exclusion often provokes conflict and, by guaranteeing groups’ abilities to affect policy, consociation is designed to eliminate this grievance. This book aims to ascertain the validity of Lijphart’s ­theory that his consociational system does promote “stability in plural societies,” by discouraging such conflict. The premise underlying Lijphart’s theory is that potentially antagonistic population segments can peacefully coexist if they are governed by a system which preserves cultural diversity and political influence, through groups’ separate but equal representation. He encourages policymakers instituting consociational systems to “let … groups define themselves” through citizens’ evolving patterns of interaction, instead of © The Author(s) 2019 B. B. Kelly, Power-Sharing and Consociational Theory, https://doi.org/10.1007/978-3-030-14191-2_2

19

20  B. B. KELLY

imposing a set of divisions that may not perfectly correspond to actual cleavages.1 Lijphart’s construction and evolution of his system and theory of consociation were motivated to reflect the qualitative nature of diverse political phenomena and assist the troubled inhabitants of deeply divided societies, with much less concern for how these concepts could be quantitatively operationalized. The task of identifying the definition of consociation which best facilitates quantitative assessment of the system and theory is thus made more difficult because of the complexity of Lijphart’s findings and conjectures concerning the concept. For instance, when prescribing consociational systems, Lijphart encourages adoption of specific political institutions but, when identifying the operation of consociation internationally, he considers any democratic arrangement generating the anticipated effects of these four components as consociational. Arguably because of the complexity and evolution of Lijphart’s theory of consociation, several critics have concluded that political developments and explorations of its theoretical integrity suggest that it is not adequately falsifiable to enable its scientific evaluation. This book demonstrates that a carefully constructed definition of consociation can be used quantitatively to analyze Lijphart’s system and theory, with a replicable and scientifically valid methodology. Lijphart’s definition of consociation is almost unanimously regarded as the definition of the concept. Perhaps for this reason, it seems inadequately appreciated that the complex and arguably contradictory elements in his many years of evolving thought on the subject can be distinguished from his consistent, central claim concerning the system and a solid, lasting concept of consociation, which can be observed in his and others’ works from as early as the seventeenth century. Distillation of a precise definition of consociation appropriate for use in this study can only be accomplished by identifying consistent elements throughout Lijphart’s portrayal of the system and by examining concepts of “consociation” offered by other writers, which seem to have influenced Lijphart’s work. This approach promotes identification and verification of operationalizable versions of the consistent, core components of Lijphart’s consociation, which are necessary for the valid scientific examination of his theory regarding the system.

1 Lijphart

(1998, p. 144).

2  WHAT IS CONSOCIATION? 

21

Apparently because it challenges previous assumptions concerning sources of stability and suggests a means of resolving severe intergroup conflicts, Lijphart’s theory of consociation has inspired extensive commentary. Most discussions of consociation involve the theory’s scientific credibility and the experience of the system in various countries. Some scholars have criticized both the theory and the system it prescribes. O’Leary points out that these critics represent diverse points of view, including conservatism, liberalism, socialism, and feminism.2 Conservative critics “condemn consociational ideas as futile” ones that will likely have no lasting impact on “deeply rooted, zero-sum identity based conflicts.”3 On the other hand, liberal, socialist, and feminist critics argue that group identities are much more malleable than advocates of consociation belief and this system solidifies lines of antagonism that need not be solidified.4 Many critics argue that consociation is undemocratic due to its executive inclusion of potentially antagonistic opposition groups. However, Lijphart and other advocates of the system contend that, “where there is extensive segmental organization and a history of antagonism, the political choice is between consociational democracy and no democracy at all.”5 This book extensively addresses the conservative argument that consociation cannot promote peace in divided societies. Criticisms concerning the changeability of divisions and the relationship between consociation and democracy are discussed in more depth, later in this chapter. Those scholars who have embraced Lijphart’s work on consociation have produced a body of literature recognizing the system in entities as diverse as the European Union and Surinam. Policymakers in at least three deeply divided societies, Northern Ireland, Bosnia, and South Africa, have expressed their respect for Lijphart’s consociation by institutionalizing some of its components. Examination of existing works involving consociation will emphasize the importance of ascertaining whether consociation does encourage stability in plural societies. Recognition of the objections put forth by the academic community will also confirm that the version of Lijphart’s system to be tested in this study is optimal for quantitative analysis.

2 O’Leary

(2005, p. 4). (2005, p. 4). 4 O’Leary (2005, pp. 5–6). 5 O’Leary (2005, p. 7). 3 O’Leary

22  B. B. KELLY

Lijphart has inspired many explorations of the question of where elements of consociation should, and can, be institutionalized to promote stability. Other scholars have extensively discussed the existence and potential for consociation in a wide range of societies, including Austria, Belgium, Burundi, Canada, Chile, Colombia, the European Union, Fiji, the Former Yugoslavia, Gambia, Israel, Italy, Kenya, Lebanon, Malaysia, the Netherlands, Nigeria, Northern Ireland, South Africa, Sri Lanka, Surinam, and Switzerland. A survey of all society-specific commentaries concerning consociation illustrates the influence of the theory and system, as topics of controversy, a seriously considered policy recommendation, and a means of understanding individual countries’ political situations. Pre-Lijphart versions of consociation, given in the Oxford English Dictionary, by Protestant groups, by Johannes Althusius, by David Apter, and by M. G. Smith, agree on a general definition of the term.6 They portray “consociation” as an at least semi-permanent group of individual segments, in which each has autonomy over at least some of its affairs. Earlier descriptions of consociational systems do vary concerning their centralization, decision-making mechanisms, association motives, and predictions of their likely stability. However, they do constitute historical precedents supporting Lijphart’s portrayal of the consociational system as a set of mechanisms which provide diverse forms of political empowerment and autonomy in at least some spheres of life, for potentially antagonistic groups attempting to coexist. These uses of the concept of consociation, including Lijphart’s, all refer to a very similar form of governance. However, it seems possible to distinguish between two schools of thought among them, deriving from their intellectual inspirations. The forms of government observed by Apter and Smith in Africa seem more similar to each other than to the churches’, Althusius’, and Lijphart’s conceptions. The Protestant New England churches, Althusius, who was inspired by the Bible, and Lijphart, whose theory was derived from the significantly Protestant Netherlands, all envisioned a more centralized and permanent version of consociation than those described by Smith and Apter.7 6 & These sources, as well as those mentioned in footnote 50, are discussed at length on the website. 7 Academic contributions by Karl Renner, W. Arthur Lewis, Lorwin, and Lehmbruch provide further evidence that Lijphart’s core concept is one that makes intuitive sense to many people, who came to agree with it from varied international perspectives.

2  WHAT IS CONSOCIATION? 

23

Lijphart’s Concept of Consociation Lijphart’s version of consociation dominates academic treatment of ethnic conflict today. However, Lijphart and analysts of his work have interpreted his system and theory differently and suggested many modifications of it. The system’s consistent core components are more accurately identified when distinguished from their portrayal in other scholars’ many assessments regarding its international applicability. Analysis of Lijphart’s notion of consociation is complicated by his presentation of it as both descriptive and prescriptive. In a 1968 publication, Lijphart identified consociation inductively through analysis of Dutch political patterns.8 However, he only began to prescribe its introduction after recognizing similar systems in other states. His derivation of conditions conducive to the establishment of consociation resulted in an almost unanimous interpretation of the theory’s descriptive and normative aspects as inextricably linked. Lijphart explains that “a distinction must be made between [it] as an empirical theory and as a normative or prescriptive theory.”9 Overall, Lijphart has chosen to focus on the latter aspect of consociation, by examining and recommending the ideal manner in which it should be adopted to help countries to the best of its ability.10 In a 1989 publication, he explains that he: deliberately opted for a definition of consociation in broad and inclusive terms. The obvious disadvantage is that these terms do not lend themselves to manageable operational definitions and precise measurements. Therefore their utility for measuring differences between democracies and changes within democracies is rather limited.11

This attitude obviously frustrates those more concerned with the scientific validity of Lijphart’s theory. They see the “central task for consociation [as] revolv[ing] around trying to actually establish valid causal

It emphasizes that many of consociation’s core components have been consistently associated with one another, by Lijphart and other scholars who appear to have worked independently of him. 8 Bogaards (1998, p. 499). 9 Lijphart (1991, p. 499). 10 Van Schendelen (1984, pp. 30, 44) and Taylor (1994, pp. 167–169). 11 Lijphart (1989, p. 142).

24  B. B. KELLY

relationships.”12 Lijphart argues that “virtually all social scientific knowledge is probabilistic in nature” but some scholars, such as Brian Barry, do argue that Lijphart’s theory could be disproven by country-specific evidence.13 As it has evolved thus far, Lijphart’s consociation aspires to describe a governmental system conducive to ethnic harmony and prescribe this system for use elsewhere. He has not focused on proving empirically the validity of many of its generalizations, apparently because he expected that the small number of cases of consociation makes it very difficult to examine its effects quantitatively. Inconsistencies within Lijphart’s body of descriptions of consociation complicate valid empirical testing of his theory, but distillation of a consistent specification of his system will allow achievement of this goal. His central claim that consociation promotes stability in plural societies is, in principle, falsifiable. Examination of existing analyses of consociation suggests that Lijphart believes that accurate descriptions and maximization of stability in plural societies are much more important than refraining from the revision of his work to facilitate its empirical assessment. He argues that detection of inconsistencies throughout his body of work on consociation is not “fair” because his recent work incorporates “the many refinements and improvements that [he has] been steadily trying to achieve.”14 To distill a valid definition of consociation appropriate for future studies, it seems necessary to identify those elements that have survived unchanged in his work. In addition, comparison between his portrayal of these elements and discussion of them by other academics should illustrate how consociation is now generally interpreted. It should be noted that, although he sees the details of its components as subject to revision, Lijphart “emphasize[s] the ‘inseparability’” of its four core components and maintains that consociational power-sharing “cannot work well- and certainly not optimally- if one or more are missing.”15 Although Lijphart has revised many elements of his definition of consociation over the past thirty years, a general description of the phenomenon undeniably has survived in his work. His general definition of consociational democracy has remained a governance system comprised

12 Taylor

(1994, p. 167). (1985, p. 115) and Barry (1975b, p. 489). 14 Lijphart (1998, p. 149). 15 Lijphart (1985, p. 503). 13 Lijphart

2  WHAT IS CONSOCIATION? 

25

of four components, which promote stability in plural societies. Lijphart has always defined consociation in terms of the maintenance of these four core phenomena: grand coalitions, segmental isolation, proportionality, and mutual veto power.16 He has consistently described this system as one in which “the centrifugal tendencies inherent in a plural society are counteracted by the cooperative attitudes and behavior of the leaders of the different segments of the population.”17 L. J. Boulle explains that academics have approached consociation from three different perspectives. None of these contradict Lijphart’s fundamental portrayal of the concept. The perspectives he identifies are: first, as a pattern of social structure which emphasizes the degree of … segmentation in a society; secondly, as a pattern of élite behavior, emphasizing the process of decision-making and conflict regulation; and thirdly,… as an underlying characteristic of the political culture arising from historical circumstances.18

These perspectives also emphasize the centrality and consistency of the components of Lijphart’s system corresponding to segmental autonomy and grand, intergroup coalitions. Aspects of Lijphart’s general definition of his consociational system must be examined in depth to ensure that the precise version to be used will enable accurate and scientifically valid assessment of his theory. Prior to such extensive analysis of the system’s components, an issue regarding the identification of their manifestations should be addressed. Lijphart has always defined consociation in terms of the maintenance of four core, political phenomena: grand coalitions, segmental isolation, proportionality, and mutual veto power.19 Quantitative representation of these components is made more complicated because Lijphart argues that some of them can be manifested in a wide variety of political systems. Through reference to a number of countries, he has argued that these components have existed in the absence of legislation mandating them because the anticipated effects of such legislation could be observed. Compilation of quantitative data designed to represent consociation 16 Lijphart

(1977, p. 25). (1977, p. 1). 18 Boulle (1984, p. 45). 19 Lijphart (1977, p. 25). 17 Lijphart

26  B. B. KELLY

should be accompanied by the assessment of the appropriateness and possibility of identifying these less straightforward examples of the system’s components. This methodological issue also emphasizes that careful discrimination between consociational components and their effects is required, to maximize the explanatory value of this statistical model designed to assess the system’s ability to bring about the change ascribed to it in Lijphart’s consociational theory. Through reference to Canada, Lebanon, Malaysia, and India, Lijphart identifies consociational effects produced by majoritarian rules, as indicative of the presence of consociational components. For instance, in a 1977 publication, Lijphart observes that the two consociational components corresponding to proportionality and grand coalition exist in Canada even though its “federal elections and cabinet formations are conducted in accordance with majority rule.”20 He argues similarly that Lebanon exhibited proportionality and grand coalition even though the president was elected through majority vote.21 Lijphart believes that, when Malaysia attained independence, the Malay Alliance party’s dominance of the majoritarian system used to elect the executive was equivalent to a consociational grand coalition because of the manner in which groups were represented within that party.22 When Lijphart identifies phenomena as consociational in the absence of formal consociational rules, he does so through reference to the demographic diversity of decision-making bodies. In a 1996 article, he argues that India had been consociational since the 1960s because conditions approximating the components of proportionality and grand coalition had coexisted with the dominance of one political party and a majoritarian plurality election system.23 He believes that sufficient institutional and policy proportionality is engendered by the Indian plurality system, among geographically concentrated “minorities,” and by the predominant Congress Party’s repeated, manufactured, largely inclusive majorities, which have “been deliberately protective of the various religious and linguistic minorities.”24 These observations by Lijphart should be considered when his theory regarding consociation’s impact on stability is evaluated.

20 Lijphart

(1977, p. 125). (1977, p. 148). 22 Lijphart (1977, pp. 151, 152). 23 Lijphart (1996a, pp. 258–261). 24 Lijphart (1996b, p. 261). 21 Lijphart

2  WHAT IS CONSOCIATION? 

27

Recognition of Lijphart’s belief that such phenomena are manifestations of consociational components provides clarification concerning his conception of each component and the optimal manner in which it would be represented quantitatively for this study. He defines each component as a condition or set of conditions, but it is the precise nature of those conditions that are crucial for consociation, rather than the ­mechanisms which produce those conditions. These components, or sets of conditions, make up his consociation system, and it is their impact on stability that is the subject of his theory concerning the system. The quantitative analysis undertaken here tests whether the system performs the role ascribed to it in this theory. Since Lijphart argues that plurality electoral systems operating in very homogeneous constituencies can produce consociational proportionality, his system would be represented more accurately if the presence of such systems were indicated through an independent variable indicating proportionality. Likewise, if parties which dominate executives are sufficiently representative to justify describing such executives as grand coalitions, an argument could be made that quantitative depiction of Lijphart’s consociation should incorporate such situations as independent variable data. However, this study’s data corresponding to consociation is confined to phenomena that even Lijphart’s critics agree are manifestations of its components. Data sources used to represent these phenomena include information regarding political parties holding executive power, legally mandated practices, and the homogeneity and geographic characteristics of constituencies. This is not because diverse single-party executives and proportional election results obtained through majoritarian rules should be categorized as effects, rather than components, of consociation. Lijphart clearly believes that conditions approximating grand coalition, proportionality, minority veto, and segmental autonomy constitute manifestations of these four components. This study’s quantitative representation of consociation is also not motivated by a determination that these data sources are the only ones which provide accurate information regarding the components. For instance, segmental autonomy would certainly be better represented by evidence that it does exist in practice, rather than the limited evidence represented in this dataset that indicates that some manifestations of it are required or are likely to exist due to relative constituency homogeneity. Data incorporating less formalized conditions portrayed as consociational by Lijphart, such as diversity of single-party executives, is not available and cannot be compiled easily at this time for the large set of countries

28  B. B. KELLY

and years covered by this dataset. In addition, phenomena such as diverse single-party executives and proportional results in majoritarian systems would need to be assessed with thresholds in mind, which they would have to surpass to be considered representative of consociational components. Like Lijphart’s identification of such conditions as consociational, the development of these thresholds would be highly debatable among scholars who specialize in the development of power-sharing systems. Perhaps the most compelling reason to leave out diverse single-party executives and proportional results in majoritarian systems from this dataset is that it cannot then be criticized for representing an inappropriate version of consociation. For these methodological and theoretical reasons, such manifestations of consociation will be discussed in this book’s case studies, but not incorporated into its quantitative tests.

Grand Coalition Discussion of the “grand coalition” component of consociation provides further examples of its definition by Lijphart and the extent to which that definition should, and could, be translated into independent variable data for this quantitative study. In one of his earliest relevant publications, Lijphart explains that the “grand coalition of the political leaders of all significant segments of the plural society” is the “prototypal” consociational device.25 However, he explains that the essential characteristic of this element is, not its institutional incarnation but, its inclusion of “all significant segments in governing a plural society.”26 According to a 1996 publication by Lijphart, consociational “grand coalitions” can consist of cabinet coalitions of political parties, coalitions “in sites other than the cabinet,” cabinet coalitions defined in group rather than party terms, proportional allocation of top governmental offices, or even one-party government by very socially inclusive parties.27 Lijphart even considers “informal grand coalition[s] to resolve a sudden crisis” as consociational.28 In a similar fashion, Juan Linz suggested that such informal coalitions are the sole form of consociation compatible with

25 Lijphart

(1977, pp. 25, 31). (1977, p. 31). 27 Lijphart (1996b, pp. 259–260). 28 Lijphart (1989, p. 141). 26 Lijphart

2  WHAT IS CONSOCIATION? 

29

presidentialism.29 Lijphart contends that the “concentration of e­ xecutive power in… one person is inimical to the basic idea of shared power” but agrees with Linz that it is “not completely impossible” for consociational elements to appear in such systems.30 However, Lijphart believes this can also be achieved through “distribut[ion]” of the “presidency and other high offices among the different groups.”31 Ronald Kieve interestingly contends that “presidential councils” can also perform the function of consociational grand coalitions.32 Kieve’s observation and those of Lijphart and other analysts seem to agree that, to be consociational, something organizationally approximating a “grand coalition” must include representatives from all major social groups and must significantly influence political decisions made in their state. To examine the influence of both consociation and its individual components on stability in plural societies, this study includes a quantitative representation of the existence of phenomena constituting “grand coalitions.” This and the other independent, potentially causal variables are described in Chapter 4 and Appendix B. Lijphart advocates allocation of substantial power to each potentially antagonistic segment. For this reason, the variables corresponding to this consociational component were designed to identify those years of states’ histories when the presence of this power could be verified through confirmation that such groups were represented in governing executives.

Segmental Autonomy Lijphart describes the three remaining elements of consociation as “secondary” to “grand coalition.”33 However, Colin Knox understandably interprets Lijphart’s emphasis on the inclusion of culturally distinct groups as indicative of a greater prioritization of one of these three, “segmental autonomy.”34 Soon after Lijphart first introduced his notion

29 Linz

(1992, p. 124). (1992, p. 217). 31 Lijphart (1991, p. 494). 32 Kieve (1981, p. 315). 33 Lijphart (1977, p. 36). 34 Knox (1995, p. 4). 30 Lijphart

30  B. B. KELLY

of consociation, Val Lorwin argued that it was itself a “function of the structure of the social cleavage.”35 Lijphart describes this concept of “segmental autonomy” as consisting of “rule by the minority over itself in the area of the minority’s exclusive concern.”36 “Segmental autonomy” thereby “ramifies subcultural divisions and legitimises subcultural encapsulation.”37 It can manifest itself as “a formal territorial federal system, a non-territorial federation, or a system in which decision-making is delegated to private segmental organisations.”38 Lijphart believes that peace can be maintained in societies suffering from ethnic conflict through institutional segregation of the main subcultures. Matthew Gabel actually interprets Lijphart’s segments as sufficiently segregated to justify their analogy to member states of the European Union.39 Jürg Steiner observes that Lijphart and other academics have not yet conclusively identified the optimal extent of segregation for consociation. However, Steiner has identified minimum indicators of autonomy necessary for a system to be identified as consociational.40 He observes that the presence of “segmental autonomy” requires that cultural groups are clearly distinguishable and have their own identities, that these groups have “political relevance,” and that intergroup mobility and marriage are rare.41 Similarly, Sue Halpern identifies as assumptions of “segmental autonomy” that: The subcultures must be politically organised… they must be unified and responsive to this internal authority;… there are areas of interest which separate the goals of the subcultures from one another and the state; and, of course,… each subculture, its leaders and its goals must be recognised and deemed legitimate by the state.42

35 McRae

(1974, p. 5). (1977, p. 41). 37 Halpern (1986, p. 191). 38 Lijphart (1989, p. 141). 39 Gabel (1998, p. 463). 40 Steiner (1981, p. 1241). 41 Steiner (1981, p. 1242). 42 Halpern (1986, p. 192). 36 Lijphart

2  WHAT IS CONSOCIATION? 

31

It should be pointed out that Halpern assumes that each of these subcultures is unified and homogeneous. While each must elect its own political representatives for consociation to work, Lijphart’s descriptions of the system do not require unity and homogeneity. However, in other respects, Lijphart’s descriptions of subcultural autonomy are consistent with these portrayals of his consociational component of segmental autonomy. Lijphart initially advocated elites’ semi-intuitive identification of those subcultures worthy of sub-state autonomy. However, he has now come to believe that such empowerment structures can more accurately reflect intergroup relations, be designed in a neutral fashion, and be responsive to social change if they naturally evolve through citizens’ interaction with a “proportional representation” electoral system, “multiparty coalition government, and cultural councils.”43 For instance, in reference to South Africa, Lijphart explains that, “instead of positing that ethnic divisions are rigid and immutable, I have argued that their frequent fluidity makes it advisable not to pre-determine any ethnic or other groups in consociational systems and instead to let these groups define themselves.”44 Lijphart specifies that this procedure is best introduced through elections according to a “relatively pure form of PR” which will encourage representation of “even very small parties.”45 He points out that his definition of plural societies “implies” that one indication that a society is “genuinely plural” is that its parties are “organized along segmental lines.”46 These PR elections enhance the extent to which a society is plural according to this definition and facilitate groups’ appropriate political empowerment, by promoting the development of parties representing all politically significant, potentially antagonistic groups. In this way, PR “allow[s] the segments to manifest themselves in the form of political parties” and so “electoral success” can be recognized as one type of “proof of segmental autonomy.”47 Lijphart points out that this situation will also enable distinction between plural and non-plural societies, since the successful parties in plural societies will most closely correspond to ethnic,

43 Lijphart

(1991, p. 499). (1998, p. 144). 45 Lijphart (1995, p. 281). 46 Lijphart (1995, p. 281). 47 Lijphart (1995, p. 281). 44 Lijphart

32  B. B. KELLY

religious, linguistic, cultural, or racial groups, while those in non-plural ones will be “non-segmental policy-oriented parties.”48 He also describes how these electorally self-determined groups should be granted political power. Lijphart recommends that each of the group-oriented parties which will naturally dominate plural societies should be granted representation in the cabinet, if they achieve “a specified minimum size in parliament.”49 Posts and/or opportunities to be given in proportion to group size, as well as veto power and segmental autonomy rights, should also be allocated to each party which achieves this minimum amount of electoral influence.50 Lijphart points out that segregation resulting from segmental autonomy would be voluntary under this system, because groups will only be granted segmental autonomy powers if their members vote in a manner that illustrates their desire for them.51 This refinement of consociation allows groups’ identities and relative strengths to be “naturally and continually self-adjusting,” according to the results of subsequent elections. This avoids conflict inspired by resentment concerning demographic changes that render groups’ population and political strengths incommensurate, like those which seemed to contribute to the collapse of Lebanon’s first consociational system in the 1970s. Lijphart’s revision of his position concerning the optimal means for identifying groups to be granted segmental autonomy illustrates that his modifications of consociational theory were motivated by a desire to assist divided societies. This emphasizes the inadvisability of dismissing it as unfalsifiable because of inconsistencies in his body of work resulting from revision.

Proportionality The remaining two elements required for Lijphart’s consociation involve proportionality and a mutual veto, or “concurrent majority,” rule.52 The proportionality element in these systems is “the principal standard of political representation, civil service appointments, and allocation of public funds.”53 This proportionality is reflected in the makeup of grand coalitions, and it should also determine the constitution of the Assembly 48 Lijphart

(1995, p. 281). (1995, p. 282). 50 Lijphart (1995, p. 282). 51 Lijphart (1995, p. 282). 52 Lijphart (1977, p. 25). 53 (Lijphart 1977, p. 25). 49 Lijphart

2  WHAT IS CONSOCIATION? 

33

cabinet and committees, “broadly in proportion to party strengths” (Lijphart 1996a, p. 271; Lijphart 1994, p. 225). In order for subcultures to be represented in proportion to their sizes, these political parties must coincide with subcultural boundaries and Lijphart seems to recently have improved his prescribed means for achieving this, by recommending subcultures’ self-identification of political categories.54 Lijphart explains that the “normal” and preferable method of reflecting proportionality in elections is the use of a proportional representation system of vote allocation.55 Lijphart discerns that proportionality in election results can also be achieved through other electoral systems, such as “a plurality single-member district system when the segments are geographically concentrated.”56 He believes that this situation in India produced results that were proportional enough to justify identification of its majoritarian electoral system as consociational. With regard to multiple countries, Lijphart implies that the consociational component of proportionality refers to political results, rather than specific electoral formulas.57 In two articles involving Horowitz’s advocacy of the Alternative Vote (AV) electoral system for deeply divided societies, Lijphart focuses on arguing that the relationship between seats and votes produced by AV is very disproportional and that AV does not yield levels of moderation, representativeness, and responsiveness that facilitate stability.58 Lijphart’s overriding concern with the results, rather than the specific configurations, of systems is also emphasized through other academics’ responses to his claims. The extent to which electoral systems allocate seats proportionally can be represented in this analysis through a variable indicating whether a country uses a proportional representation electoral system. Lijphart’s focus on results as indicators of consociation illustrates the justifiability of quantitatively representing the system through data corresponding to both laws mandating the existence of its components, as well as manifestations of their actual use.

54 Halpern (1986, p. 191) pointed out that subcultures would not be represented proportionally, unless political parties perfectly corresponded to subcultures’ boundaries. 55 Lijphart (1996b, p. 261). 56 Lijphart (1989, p. 141). 57 Lijphart (1991, p. 507). 58 Lijphart (1991, p. 93; 1997, p. 9).

34  B. B. KELLY

Veto Power The fourth, and final, element necessary for the achievement of Lijphart’s consociation is mutual veto power that can be exercised by subcultures. Mutual veto power, which he also describes as “negative minority rule” and “concurrent majority” power, is employed in consociations to protect minorities against political defeats judged to jeopardize their “vital interests.”59 To detect consociation in a replicable fashion, some defensible definition of these “vital interests” must be ascertained. Analysts of Lijphart’s work provide slightly different descriptions of them, which may facilitate their delineation. Hennie Kotzé portrays the mutual veto as a guarantee of “sovereignty in matters important to” segments, while Gabel describes it as “the power to block any attempt to eliminate or reduce their autonomy.”60 Perhaps international classification of exactly which policy matters are covered by this throughout the world is unnecessary for a valid definition of consociation, since what is relevant is its interpretation by each population. The only point about mutual vetoes’ potential subject matter relevant to the international study of consociation is whether there is unanimous identification of potential claims and consistent political treatment of them, in individual consociational states. Assuming that all groups have this mutual veto power and consider consequent policy jurisdictions as legitimate, Lijphart argues that mutual veto power is unlikely to cause minority tyranny. He believes this is so because all minorities will have the power and thus each will be aware of the potential for reprisals, deadlock, and general insecurity, should it be used in a manner that seems improper.61 Apparently, Lijphart’s only advice to the individual groups is that this power “works best when it is not used too often and only with regard to issues of fundamental importance.”62 He explains that this mutual veto power can be manifested in either formal, legal form or as an “informal and unwritten understanding.”63 A type of consistently operational mutual veto power can be instituted through a requirement for consensus or some proportion of representatives’ agreement, in the decision-making process.64 59 Lijphart

(1977, pp. 36–37). (2001) and Gabel (1998, p. 465). 61 Lijphart (1977, p. 37). 62 Lijphart (1991, p. 495). 63 Lijphart (1977, p. 38). 64 Lijphart (1994, p. 223; 1996a, p. 271). 60 Kotzé

2  WHAT IS CONSOCIATION? 

35

However, the successful performance of the mutual veto provision depends on the popular legitimacy of its policy jurisdiction, its application, and the identity of those groups given the power to veto. Analysts of Lijphart’s consociation seem to disagree slightly concerning what is required of a group, for it to be given such veto power. Michael MacDonald believes that such vetoes can be “wielded by any group or party,” while Hussein Adam sees this power as only being held by “[e]ach significant subgroup.”65 Michael Hudson emphasizes the necessity for agreement throughout the population about which groups are to be considered deserving of mutual veto power. In his discussion of Lebanese “consociation” before 1975, he points out that, from the perspective of Shi’ites and Druze, this veto power was not legitimate because it was not held by their groups.66 One could argue that, if the identification of empowered groups is seen as illegitimate in a specific state, true consociation has not been attained there. Of course, if no “consociational” state has ever enjoyed this sort of legitimacy, this might suggest that widespread acceptance of designated groups cannot be achieved in consociational systems. Even if groups are allowed to identify themselves as subcultures deserving such empowerment, mutual veto mechanisms may become unsustainable. The more groups that are involved, the greater the likelihood that every legislative proposal will be vetoed. Perhaps we cannot include a specification involving a number of subcultures in a definition that is designed to be applied to the existing states in a replicable manner. The diversity of country-specific perspectives and practices concerning the identification of groups that should be granted this power illustrates the inadvisability of employing one equation to ascertain how many, or which, groups must be empowered in each country, if it is using the minority veto consociational component. To avoid using such an equation, but to simultaneously ensure objective, international identification of this power, a list of “Minorities at Risk” compiled for a separate research program and information taken from each country’s constitution will be analyzed to make the data used in this study to represent minority veto power. Identification of consociational government internationally also requires a definition of these four core components so that they are theoretically compatible with each other, when they exist in their purest forms. Grand coalitions and segmental autonomy can exist in their purest 65 MacDonald 66 Hudson

(1992, p. 713) and Adam (1994, p. 32). (1988, p. 227).

36  B. B. KELLY

manifestations with each other and with proportionality or mutual veto. However, one statement by Lijphart suggests that his notions of proportionality and groups’ veto power may not be completely compatible. He explains that, as far as political representation is concerned, two extensions of proportionality entail even greater deviations from majority rule and give special protection to small minorities: the overrepresentation of small minorities and parity of representation. They are particularly useful alternatives to proportionality….67,68

Overrepresentation is not representation according to the strength of segments, no matter how this strength is measured. It actually constitutes a form of veto power for minorities. Overrepresentation is thus compatible with the veto requirement but not with that of proportionality. Lijphart is the only theorist of consociation who has contributed a list of its core elements, and they are internationally recognized as the components of consociation. Therefore, it seems reasonable to interpret exactly what he means by this prescription of overrepresentation and to attempt to incorporate it into the definition to be quantitatively represented in this study. Lijphart apparently means that each segment in a consociational system should be represented as if it is approximately equally strong, so that segments will treat each others’ demands seriously. After these strengths are determined, he implicitly prescribes that they be fixed at least in the short term, to engender segments’ feelings of security. By proportionality, he means that these groups should be granted policymaking influence in proportion to their designated overall strengths. He does not recommend a precise means of deducing the prudent strength to be allocated to each segment. This implicitly suggested modification of what is normally termed “proportionality” has been adopted by international policymakers. In South Africa and Northern Ireland, it is obvious that they have interpreted consociation as a means of legitimizing the promise of permanent substantial power, regardless of actual numbers of white South Africans and members of both Northern Irish segments. Although they are both designed to enhance the political power of minorities, it must

67 The

italics shown here did not exist in the original text. (1979, p. 501).

68 Lijphart

2  WHAT IS CONSOCIATION? 

37

be kept in mind that Lijphart’s consociational components of proportionality and minority veto power have different specific aims and functions. While proportionality is designed to allocate positions in representative bodies in an egalitarian manner, minority veto is the requirement that certain groups, irrespective of their relative sizes, have an equal ability to actually affect government policy. Since they focus on different things, representation and decision-making capability, they are not actually theoretically antithetical to one another or incompatible, but their coexistence as consociational components emphasizes the theory’s simultaneous dual approach regarding group size. Lijphart has not modified his recommendations for proportionality and minority veto, throughout the body of his work. However, he has extensively revised other elements of his theory and, particularly, the conditions that he believes are favorable to its institutionalization. If his theory is judged only in terms of its descriptive potential, as a means of governance, these conditions are not very significant because they focus on the introduction, rather than the function, of consociation. If, on the other hand, the object is to study his theory’s prescriptive value, some distilled version of these conditions must be derived. Because this study’s primary aim is to evaluate consociation’s effectiveness as a mechanism for resolving severe intergroup conflict, interpretation of these conditions is unnecessary. However, when feasible, they will be included in this study to ensure that they are not themselves responsible for effects which are usually attributed to consociation. For instance, one of these conditions, concerning small population size, may conceivably encourage the introduction of consociation and the effects that it appears to produce. The variables used to represent these conditions in this analysis will be identified in Chapter 4. Although Lijphart has consistently revised his theory and optimal consociational system over the past few decades, his unwavering espousal of the system’s four core elements illustrates that it is methodologically appropriate to define it as comprising these components, to facilitate its quantitative assessment. Other scholars’ work concerning systems similar to consociation that was developed independently of Lijphart also will be examined, to ascertain how consistently these core elements were associated with their systems and the extent to which this book’s conclusions will thus be relevant to their work, as well as Lijphart’s.

38  B. B. KELLY

Analysis of the varied, historical uses of the term, consociation, emphasizes that a theoretically sound, empirically testable definition of Lijphart’s system of governance can be analyzed. Versions of consociation constructed independently of Lijphart are compatible with, and share some element of, his general definition of the concept. Lijphart’s theory reflects the term’s historical meaning, and this enhances its credibility as a representative of a body of thought that is centuries old. For the purposes of this book which is designed to evaluate Lijphart’s claims for consociational systems, they are defined as those which consistently experience all or some of the components, grand coalitions, segmental autonomy, policy influence proportional to designated segmental strength, and provision for mutual veto. Representation of these components in this analysis will enable their repercussions to be individually identified and assessed.

Does the System Promote Stability in Plural Societies? For the purposes of this analysis, the most relevant criticisms of consociation are those which challenge the system’s ability to promote stability in plural societies. For instance, it has been argued that stability achieved in consociational systems has actually been caused by other phenomena. One example of these alternative potential explanations is Barry’s contention that elites can achieve stability without consociation.69 Ian Lustick argues that stability can also be maintained in plural societies through systems of “control,” which he describes as “sustained manipulation of subordinate segment(s) by a superordinate segment.”70 Control is probably less capable of maintaining stability in the long term, so Lijphart may have not responded to Lustick’s point because he believes that short-term stability does not constitute true stability. However, Lustick’s and Barry’s contributions do illustrate that Lijphart does not extensively explore other phenomena which may be conducive to stability in plural societies. Another set of criticisms involving the direction of the causal relationship between consociation and stability further illustrates the falsifiability of the central proposition to be tested here. Daalder argues that Dutch 69 Barry

(1975b, p. 129). (1979, p. 330).

70 Lustick

2  WHAT IS CONSOCIATION? 

39

and Swiss experiences suggest that “rather than seeing accommodationist practices as an elite response to threatening political divisions, one might argue that earlier consociational practices facilitate the accommodation of new emerging cleavages.”71 This point is similar to Barry’s observation, regarding the theory of cross-cutting cleavages, that overlapping divisions might be an effect, rather than a cause, of conflict moderation.72 G. R. Boynton and W. H. Kwon’s (1978) quantitative examination of consociation also challenges Lijphart’s claims concerning causal direction. They focus on the performance of Lijphart’s version of consociation in the Netherlands, before 1978. They “perform a formal analytical assessment” of Lijphart’s argument, represented through a set of differential equations “parallel to the structure of [his consociational] relationships.”73 Although their study involves only one country, their overall conclusions suggest that his claims for the system are tautological and that it does not necessarily resolve conflicts in deeply divided societies. They conclude that, “[i]nstead of consociational democracy producing stability[,]… a stable democracy will produce accommodation” and that “more is required to guarantee democratic stability than the structure Lijphart proposed.”74 However, they also find that “democratic stability would be guaranteed if the system were structured as Lijphart suggested and elites were willing to discount their own views of what should be done about political problems.”75 Boynton and Kwon suggest that the direction of the causal relationship between consociation and stability is unclear and that achievement of stability through consociation requires certain types of elite behavior. However, Lijphart also emphasizes the importance of elite behavior, and Boynton and Kwon do not attempt to test the accuracy of their results through the analysis of concrete evidence involving real situations. My study enables further exploration of the issues they examine because statistical analysis can discriminate between causal directions and suggest whether another variable such as elite behavior may be crucial.

71 Daalder

(1974, pp. 615–618). (1975b, p. 487). 73 Boynton and Kwon (1978, pp. 11, 24). 74 Boynton and Kwon (1978, pp. 21, 25). 75 Boynton and Kwon (1978, p. 25). 72 Barry

40  B. B. KELLY

Consociation has been criticized for not promoting desirable effects. Lijphart argues that it enables stability by allowing for substantial intergroup contact only among segments’ elites, who represent the interests of their fellow group members. In his consociational system, Lijphart also explains that elites must make sure that their groups do not successfully challenge their accommodative efforts and that segments must be allocated proportions of power corresponding to their sizes prior to elections. He thus implies that the only function of elections is for the masses to choose who in their groups will represent them, rather than affecting the overall configuration of state power. A number of scholars have examined the degree to which this system is democratic. For instance, Barry argued that “[r]ival politicians may get together to suppress dissent and we may if we wish call the result ‘consociation’, but if it is to be consociational democracy they must carry their supporters with them voluntarily.”76 For slightly different reasons, Van Schendelen comes to a similar conclusion: “In a polyarchy competition between the elites is, more than anything else, essential; in a consociation basically the opposite, namely intense collaboration, is crucial.”77 The literature discussing the extent to which consociation is democratic suggests that, if the system is found to not promote stability in plural societies, one potential explanation for this failure may be because insufficient democracy precludes long-term legitimacy, and hence s­tability. Lijphart’s response to such criticism is that consociation should not be evaluated through reference to the view, almost universally held by academics, that opposition to, and turnover of, governing regimes is necessary for stable democracy.78 He contends that these criteria “are narrowly based on one conception of democracy, the majoritarian conception, which does not exhaust the range of democratic possibilities.”79 He also points out that, while the list proportional representation electoral system he favors gives party leaders great influence over which candidates are elected, party leaders also are “usually able to reserve safe seats” in Britain’s majoritarian system and “[e]lite domination does not

76 Barry

(1975b, p. 500). Schendelen (1984, p. 32). 78 Lijphart (2002, p. 41). 79 Lijphart (2002, p. 41). 77 Van

2  WHAT IS CONSOCIATION? 

41

vary a great deal among democracies.”80 In his discussions of criticisms involving insufficient democracy, Lijphart does not extensively explore the argument that lack of opposition and turnover produces instability and system illegitimacy. He contends that consociation does not require opposition and turnover, which are commonly considered to be indispensable for democracy, but he does not elaborate on why his system does not require them to be successful. To gain some insights into the role of democracy in plural societies and its relationship to consociation, a controlling, independent variable corresponding to democracy will be analyzed here. Barry and Butenschøn, among others, have also discussed the possibility that consociational systems may “exacerbate” conflicts or “create instability” by “freezing” the cleavages which divide groups.81 Lijphart believes that minimizing contact between people of different groups promotes stability in plural societies, but some argue that it is actually conducive to instability. It is said that the system could theoretically result in “frozen,” virtually impermeable lines of cleavage. Lijphart does not prescribe the intentional introduction of incentives to encourage negotiation or cooperation of any kind across these cleavages. Lijphart argues that incentives for intergroup cooperation and negotiation do not need to be intentionally adopted for consociation to work, because he believes that politicians and parties want power and they have to enter into coalitions to acquire and maintain it.82 Particularly because this analysis will yield insights about the specific role of segmental autonomy on stability, it will facilitate a greater understanding of the hypothesis that Lijphart’s consociation inherently leads to frozen cleavages and instability. However, Lijphart’s optimal consociational system also has arguably seemed conducive to instability because he initially recommended that elites and academic advisors identify potentially antagonistic groups and set up systems to grant them autonomy, encourage their isolation, and provide smaller groups with power sufficient to veto decisions. He believes that his recent revisions of the optimal consociational system ensure that populations themselves will determine which cleavages are necessary to

80 Lijphart

(2002, p. 41). (1975a, p. 396) and Butenschøn (1985, p. 99). 82 Lijphart (2002, p. 44). 81 Barry

42  B. B. KELLY

isolate potentially antagonistic groups and determine the makeup of such groups, and that the group boundaries and power granted to groups can evolve according to demographic changes. This analysis tests whether states possessing the consociational components enjoy increased stability so it does not require data indicating the means by which groups were chosen and allocated power in individual states. By showing the influence of consociation on stability, it will provide some hints concerning the validity of the hypothesis that Lijphart’s system destabilizes countries in the long term through excessive “freezing” of cleavages. Some scholars have also argued that consociation contains inadequate incentives for intergroup cooperation and thus long-term stability. Lijphart believes that his system contains sufficient incentives because elites can control their constituents and will form all-inclusive coalitions because they realize that these are necessary for the maintenance of stability and their own power.83 In Lijphart’s words, consociational democracy… assumes that political elites enjoy a high degree of freedom of choice, and that they may resort to consociational methods of decision-making as a result of a rational recognition of the centrifugal tendencies inherent in plural societies and a deliberate effort to counteract these dangers.84

Donald Horowitz disagrees with Lijphart and believes that elites in “severely divided societies” usually do not possess this amount of “freedom.”85 He argues that all-inclusive, grand coalitions are unlikely in democratic systems with such divided populations because the formation of such coalitions substantially heightens “intraethnic competition” and thus jeopardizes the cohesiveness required for the collective articulation of interests.86 Horowitz also believes that, “[u]nless there is an incentive to compromise over ethnic issues, the mere need to form [any type of] coalition will not produce compromise.”87 He thinks that party proliferation sufficient to prevent any one party from monopolizing power must exist

83 Lijphart

(1977, pp. 53, 170, 165; 2002, p. 44). (1977, p. 165). 85 Horowitz (1985, p. 573). 86 Horowitz (1985, p. 575). 87 Horowitz (1991, p. 171). 84 Lijphart

2  WHAT IS CONSOCIATION? 

43

“to require a coalition” and that Lijphart’s prescribed electoral system, List PR, cannot be relied on to produce this level of proliferation.88 Horowitz observes that, whenever possible, politicians “prefer to govern alone.”89 However, Lijphart points out that most coalition theorists believe that, “in multiparty systems, parties will want to enter and remain in coalition cabinets.”90 He explains that, recently, Kaare Strom and other coalition theorists have found that, under “special” conditions, parties may be motivated to not be in “coalitions” but that these situations are “the exceptions, and the more usual inclination of parties is to want to be included in cabinets.”91 Lijphart argues that this is true “[b]ecause” politicians want power and they must be in cabinet to gain power.92 Regardless of whether it is true that parties and politicians almost always want to be in coalitions and cabinets in normal democracies, it needs to be kept in mind that political competition is fundamentally different in divided societies. Ben Reilly explains that most models of political competition assume that societies are not polarized so “both voters and parties can be expected to converge on a middle ground” concerning desired policy.93 When this is true, parties profit most by attracting the “median voter,” who favors moderate policy. Reilly points out that divided societies naturally yield parties that are almost exclusively focused on representing individual groups and that voters in these states are usually “ethnic voters,” who are much more inclined to support parties tailored to their groups, rather than any other type of party.94 Following Giovanni Sartori’s example, Reilly refers to this situation as one of “polarised pluralism,” in which “the logic of elections changes from one of convergence on policy positions to one of extreme divergence.”95 The virtual absence of median voters in this scenario means that political competition occurs “at the extremes” rather than the policy center.96

88 Horowitz

(1991, p. 175). (1991, p. 175). 90 Lijphart (2002, pp. 43–44). 91 Lijphart (2002, pp. 43–44). 92 Lijphart (2002, p. 44). 93 Reilly (2001, p. 9). 94 Reilly (2001, p. 10). 95 Sartori (1976) and Reilly (2001, p. 10). 96 Reilly (2001, p. 10). 89 Horowitz

44  B. B. KELLY

This results in “an increasingly polarised political process, in which strategic incentives for office-seeking politicians often push them in the direction of encouraging ethnic hostilities and perceptions of group insecurity.”97 Under these conditions, politicians can gain more power by gravitating toward policy extremes and parties may prefer not to enter into coalitions because doing so will indicate that they are negotiating concerning their policy preferences. The policies they advocate are also often motivated by the belief that the government system under which they operate is fundamentally unfair and illegitimate. Thus joining coalitions under these circumstances could also be interpreted as “selling out” for power. Lijphart focuses less than Horowitz and Reilly on differentiating between politicians’ motivations in more and less divided societies. The analysis described here facilitates greater recognition of the incentives required for stability in divided societies, by analyzing the effects of highly inclusive coalitions, potential veto power for segmental groups, and moderate multiparty systems. Horowitz suggests means by which politicians and the masses could be motivated to provide greater support for intergroup political systems. He argues that electoral systems which require each voter to provide a list of his/her preferred candidates, chosen from a list containing fewer candidates from his/her group than the number of required preferences, would motivate politicians to appeal for votes from outside their groups and thus encourage their pursuit of moderate policies.98 If politicians do not have to appeal outside of their groups, Horowitz believes that they will be more antagonistic. In that case, he asserts that they will wish to make it appear that they are more dedicated than their competitors to improving conditions for their groups relative to other groups and representing the more extreme factions of their groups. Horowitz also argues that appropriately tailored federalism can be used “to foster interethnic cooperation” by delineating electoral district boundaries in a way that further motivates politicians to appeal to multiple groups.99 Reilly believes that it does not seem reasonable to assume that elites have

97 Reilly

(2001, p. 10). (1991, p. 175). 99 Horowitz (1985, p. 619). 98 Horowitz

2  WHAT IS CONSOCIATION? 

45

good intentions, since there is a “clear pattern in many deeply divided societies of party leaders themselves being the ones who initiate and fuel inter-ethnic conflicts.”100 Since electoral systems exert substantial influence on politicians’ and voters’ incentives, a great deal of debate concerning such incentives in divided societies has focused on electoral systems. Horowitz prescribes the AV majoritarian electoral system as optimal for deeply divided societies. This system is used in single-member districts, and it allocates each district’s single seat to the candidate who wins a majority of votes. Voters can rank-order candidates, and if no candidate receives more than 50% of votes during the first count, voters’ subordinate candidate choices are analyzed to determine the winner of the seat. This sort of rank-ordering system may motivate voters to consider candidates from groups other than their own, but election to the seat still only requires 50% of the votes, and so the system can also result in a lack of representation for almost 50% of voters. The possible alienation of almost 50% of voters who may conceivably belong to a group that is always excluded from power is unavoidable with AV. Jon Fraenkel and Bernard Grofman analyze data involving 1999 and 2001 Fijian elections to search for insights regarding the role of AV, since that system is used for national parliamentary elections there.101 They observe that there was “considerable interethnic movement of preference votes” under AV in both elections but that “far more preference votes were transferred from moderate parties to the more radical parties than vice versa.”102 Fraenkel and Grofman argue that, in these elections, “a proportional representation system” would have promoted the progress of moderate parties more and that the 1999 “outcome exacerbated inter-ethnic tensions.”103 They believe that, for AV to “yield a predictable centripetal impact,” there must be “majority support for ethnic moderation.”104 Fraenkel and Grofman thus imply that AV is only conducive to moderation when moderation exists to an extensive degree in a society already. They come to this conclusion

100 Reilly

(2001, pp. 177–178). and Grofman (2004, p. 489). 102 Fraenkel and Grofman (2006, p. 647). 103 Fraenkel and Grofman (2004, p. 502; 2006, p. 623). 104 Fraenkel and Grofman (2006, p. 648). 101 Fraenkel

46  B. B. KELLY

through the analysis of Fiji, and a primary component of Horowitz’s rebuttal to their argument is that Fraenkel and Grofman incorrectly identify contenders in these elections as moderate or radical.105 Horowitz uses electoral data to support his argument concerning the identification of such contenders in Fiji. In this divided society, “in which there is a struggle for domination between two sizable groups,” Horowitz points out that “more than 80% of second preferences were interethnic” for those seats being filled in 1999 which were not reserved for candidates of “specified ethnic groups.”106 In the 2001 election, the corresponding proportion of interethnic second preference votes declined significantly to 41%.107 Horowitz believes that, “[f] or AV to foster accommodative results, there must at least be heterogeneous constituencies and subethnic divisions reflected in party politics.”108 He argues that the first of these conditions “was only modestly met in both elections” and the second “was no longer really present” in 2001.109 Comparison of these analyses of Fijian political developments by Horowitz, Fraenkel, and Grofman suggests that AV is associated with interethnic secondary vote preferences. Even the lower amount of them in the 2001 election seems very high considering that Fiji is a divided society. These analyses disagree strongly with regard to whether these secondary preferences resulted in support for more moderate or radical candidates, but there also seems to be some common ground between them concerning the political conditions which make AV capable of exerting a positive influence on moderation. According to the widely accepted theory that cross-cutting promotes moderation and consequently stability, Horowitz’s observation that “subethnic divisions reflected in party politics” are necessary for AV to “foster accommodative results” seems to be consistent with the notion that some moderation is needed for AV to perform this role in severely divided societies.110 It seems that Fraenkel and Grofman believe that much more extensive

105 Horowitz

(2006, pp. 656–657). (2006, pp. 660, 656). 107 Horowitz (2006, p. 659). 108 Horowitz (2006, p. 659). 109 Horowitz (2006, p. 659). 110 Horowitz (2006, p. 648). 106 Horowitz

2  WHAT IS CONSOCIATION? 

47

moderation is required for “AV to necessarily yield outcomes that favor moderate parties,” but Horowitz’s prediction concerning when this is likely to work confirms the necessity for the sort of intergroup cross-cutting which is arguably the most obvious manifestation of moderation.111 Since AV requires only 50% support for a candidate to win office and thus makes it possible that almost 50% of voters will be consistently unrepresented, multiple preference-based voting is likely to be deemed more appropriate for divided societies. However, the debate surrounding Horowitz’s prescription of AV for divided societies suggests that multiple preference-based voting is conducive to interethnic vote preferences and that prioritization of encouraging moderation and cross-cutting in such societies is advisable from the time at which power-sharing is first introduced. The single transferable vote (STV) electoral system provides additional incentives for moderation while avoiding the problems which have been associated with AV in societies whose political divisions too closely coincide. Reilly provides additional analysis and evidence concerning the role of multiple preference-based voting. He argues that systems allowing for rank-ordering of candidates, including the non-majoritarian STV procedure, have succeeded in promoting resolution of intergroup conflicts in Australia, Papua New Guinea, Fiji, and Northern Ireland.112 STV operates very differently from AV in severely divided societies because it is used in multimember districts and the rank-ordering it allows for voters thus translates into seats for representatives of multiple groups in districts where multiple groups coexist in substantial numbers. For this reason, STV is much less likely to engender resentment caused by constant unempowerment of significant groups. Lijphart’s preferred system, List PR, allows for greater proportionality in power allocation but does not allow voters to give support for multiple parties, some of which may focus on representing groups other than their own. The approaches of Horowitz and Reilly also differ from that of Lijphart because they believe that cross-group interaction is advisable (in particular between politicians and groups other than their own). This leads them to favor heterogeneous districts in which cross-group political appeals would help candidates

111 Fraenkel 112 Reilly

and Grofman (2006, p. 623). (2001, pp. 167–171).

48  B. B. KELLY

more, while Lijphart favors autonomy and the voluntary separateness of potentially antagonistic groups. Inclusion in this analysis of variables representing PR, segmental autonomy and federalism, and geographical concentration of segments will shed some light on the hypothesis that consociation contains inadequate incentives to promote long-term stability. Criticisms of consociation that can be analyzed to some extent by this set of statistical analyses include the arguments that it does not facilitate stability in plural societies because this aim is achieved by other phenomena, stability actually causes the introduction of consociation, consociation cannot maintain legitimacy in the long-term because it freezes segmental cleavages, it contains inadequate incentives for intergroup cooperation, and it is not sufficiently democratic. These assertions have all been made through reference to case studies, and now they will be analyzed in an objective, international, and quantitative manner with an imperfect but unprecedented set of data specifically designed to represent consociation according to its depiction by Lijphart. The literature concerning consociation, similar concepts not conceived by Lijphart, and criticisms of his work on the subject, demonstrate that the system can be represented quantitatively in a falsifiable form and that an unbiased assessment of its efficacy will help to resolve many controversies. It is clear that consociation is a concept with historical integrity and that Lijphart has always described its system as consisting of four components, which this book will prove can be represented in a scientifically sound statistical analysis. The body of criticisms suggests many considerations relevant to the empirical analysis of the proposition that the system’s elements promote stability in plural societies. Criticisms concerning the scientific credibility of Lijphart’s work emphasize the importance of distinguishing between the components of the consociational system of governance and the entire body of his writings concerning it. Objections involving the imprecision of Lijphart’s terms, means of measurement, and identification of cases similarly illustrate that choosing these carefully is absolutely essential. Discussions of the probable effects of consociation in societies containing different cleavage structures, the system’s arguably undemocratic aspects, and the flaws of existing works attempting to test this proposition all emphasize the importance of this project’s analysis.

2  WHAT IS CONSOCIATION? 

49

References Adam, Hussein M. “Formation and Recognition of New States: Somaliland in Contrast to Eritrea.” Review of African Political Economy. 59 (1994) 21–38. Barry, Brian. “The Consociational Model and Its Dangers.” European Journal of Political Research. 3:4 (December, 1975a) 393–412. Barry, Brian. “Review Article: Political Accommodation and Consociational Democracy.” British Journal of Political Science. 5:4 (1975b) 477–505. Bogaards, Matthijs. “The Favourable Factors for Consociational Democracy: A Review.” European Journal of Political Research. 33 (1998) 475–496. Boulle, L.J. Constitutional Reform and the Apartheid State: Legitimacy, Consociation, and Control in South Africa. New York: St. Martin’s Press, 1984. Boynton, G.R. and W.H. Kwon. “An Analysis of Consociational Democracy.” Legislative Studies Quarterly. 3:1 (February, 1978) 11–26. Butenschøn, Nils A. “Conflict Management in Plural Societies: The Consociational Democracy Formula.” Scandinavian Political Studies. 8:1–2 (June, 1985) 85–103. Daalder, Hans. “The Consociational Democracy Theme.” World Politics. 26:4 (July, 1974) 604–621. Fraenkel, Jon and Bernard Grofman. “A Neo-Downsian Model of the Alternative Vote as a Mechanism for Mitigating Ethnic Conflict in Plural Societies.” Public Choice. 121:3/4 (2004) 487–506. Fraenkel, Jon and Bernard Grofman. “Does the Alternative Vote Foster Moderation in Ethnically Divided Societies?: The Case of Fiji.” Comparative Political Studies. 39:5 (June, 2006) 623–651. Gabel, Matthew J. “The Endurance of Supranational Governance: A Consociational Interpretation of the European Union.” Comparative Politics. (July, 1998) 463–475. Halpern, Sue M. “The Disorderly Universe of Consociational Democracy.” West European Politics. 9:2 (1986) 181–197. Horowitz, Donald L. Ethnic Groups in Conflict. Berkeley: University of California, 1985. Horowitz, Donald L. A Democratic South Africa?: Constitutional Engineering in a Divided Society. Berkeley: University of California, 1991. Horowitz, Donald L. “Response: Strategy Takes a Holiday: Fraenkel and Grofman on the Alternative Vote.” Comparative Political Studies. 39:5 (June, 2006) 652–662. Hudson, Michael C. “The Problem of Authoritative Power in Lebanese Politics: Why Consociation Failed.” Lebanon: A History of Conflict and Consensus. Eds. Nadim Shehadi and Dana Haffar Mills. London: I.B. Tauris, 1988.

50  B. B. KELLY Kieve, Ronald A. “Pillars of Sand: A Marxist Critique of Consociational Government in the Netherlands.” Comparative Politics. (April, 1981) 313–337. Knox, Colin. Emerging Consociation: Prospects for Power-Sharing in Northern Ireland. Northern Ireland: Centre for Research in Public Policy and Management, at the University of Ulster, 1995. Kotzé, Hennie. Federalism: The State of the Debate in South Africa. Stellenbosch: Center for International and Comparative Politics. http://www.unisa.ac.za/ dept/press/politeia/142/hennw.html. Most recently consulted August 2001. No page numbers available on website. Also available in Politeia. 14:2 (1995). Lijphart, Arend. Democracy in Plural Societies: A Comparative Exploration. New Haven: Yale University Press, 1977. Lijphart, Arend. “Consociation and Federation: Conceptual and Empirical Links.” Canadian Journal of Political Science. 12:3 (September, 1979) 499–515. Lijphart, Arend. Power-Sharing in South Africa. Berkeley: Institute of International Affairs, 1985. Lijphart, Arend. “From the Politics of Accommodation to Adversarial Politics in the Netherlands: A Reassessment.” Politics in the Netherlands: How Much Change. Eds. Hans Daalder and Galen A. Irwin. London: Frank Cass, 1989. Lijphart, Arend. “Research Note: The Alternative Vote: A Realistic Alternative for South Africa?” Politikon. 18:2 (June, 1991) 91–101. Lijphart, Arend. “Democratization and Constitutional Choices in CzechoSlovakia, Hungary, and Poland, 1989–91.” Journal of Theoretical Politics 4:2 (1992) 207–223. Lijphart, Arend. “Prospects for Power-Sharing in the New South Africa.” Election ’91: South Africa: The Campaign, Results, and Future Prospects. Ed. Andrew Reynolds. London: James Currey Press, 1994. Lijphart, Arend. “Self-Determination Versus Pre-Determination of Ethnic Minorities in Power-Sharing Systems.” The Rights of Minority Cultures. Ed. Will Kymlicka. Oxford: Oxford University Press, 1995. Lijphart, Arend. “The Framework Document on Northern Ireland and the Theory of Power-Sharing.” Government and Opposition. 31:3 (Summer, 1996a) 267–274. Lijphart, Arend. “The Puzzle of Indian Democracy: A Consociational Interpretation.” American Political Science Review 90:2 (June, 1996b) 258–268. Lijphart, Arend. “Disproportionality Under Alternative Voting: The Crucialand Puzzling- Case of the Australian Senate Elections, 1919–1946.” Acta Politica. 32:1 (Spring, 1997) 9–24.

2  WHAT IS CONSOCIATION? 

51

Lijphart, Arend. “South African Democracy: Majoritarian or Consociational?” Democratization 5:4 (Winter, 1998) 144–50. Lijphart, Arend. “The Wave of Power-Sharing Democracy.” The Architecture of Democracy: Constitutional Design, Conflict Management, and Democracy. Ed. Andrew Reynolds. Oxford: Oxford University Press, 2002. Linz, Juan. “The Perils of Presidentialism.” Parliamentary Versus Presidential Government. Ed. Arend Lijphart. Oxford: Oxford University Press, 1992. Lustick, Ian. “Stability in Deeply Divided Societies: Consociation Versus Control.” World Politics. 31:3 (1979) 325–344. MacDonald, Michael. “The Siren’s Song: The Political Logic of Power-Sharing in South Africa.” Journal of Southern African Studies. 18:4 (December, 1992) 709–725. McRae, Kenneth. “Introduction.” Consociational Democracy: Political Accommodation in Segmented Societies. Ed. McRae, Kenneth. Toronto: McClelland, 1974. O’Leary, Brendan. “Debating Consociational Politics: Normative and Explanatory Arguments.” From Power Sharing to Democracy: Post-conflict Institutions in Ethnically Divided Societies. Ed. Sid Noel. Montreal and Kingston: McGill-Queen’s University, 2005. Reilly, Ben. Democracy in Divided Societies: Electoral Engineering for Conflict Management. Cambridge: Cambridge University Press, 2001. Sartori, Giovanni. Parties and Party Systems. Cambridge: Cambridge University Press, 1976. Steiner, Jürg. “Research Strategies Beyond Consociational Theory.” The Journal of Politics. 43:4 (November, 1981) 1241–1250. Taylor, Rupert. “A Consociational Path to Peace in Northern Ireland and South Africa?” New Perspectives on the Northern Ireland Conflict. Ed. Adrian Guelke. Aldershot: Avebury, 1994. 161–174. Van Schendelen, M.P.C.M. “The Views of Arend Lijphart and Collected Criticisms.” Acta Politica 19:1 (January, 1984) 19–55.

PART I

A Quantitative Assessment of Consociation

CHAPTER 3

The Plural Societies for Which Consociation Is Prescribed

Lijphart’s specific proposition that is tested in this book is that ­“consociation promotes stability in plural societies.” Statistical analysis of this theory requires that cases be identified which represent such plural societies’ experiences. The specific wording of the proposition signifies that the aim is to explore consociation’s role in plural states, rather than in all existing states, all states experiencing violence, or all states that have experienced consociation. Exploration of this proposition thus requires identification of all plural states through a replicable method. Since Lijphart is recognized to be the source of the modern theory of consociation, if the cases are chosen according to his understanding of the term, “plural,” the potential value of this book’s results will be optimized. Lijphart has defined the term, “plural,” a few times so analysis of these definitions and the use of the term by him and other scholars facilitates the identification of the concept which should be used to choose cases. It will become clear that these plural societies are defined by their sharp divisions into groups whose members’ collective actions reflect the obvious existence of cleavages corresponding to religion, language, culture, ethnicity, and/or race.1 Lijphart seems to have first discussed the definition of this term to illustrate how his understanding of it does not completely correspond to those articulated by other academics. In a 1977 publication, Lijphart first 1 & Evidence is provided on the website which demonstrates that this study’s conclusions are applicable to phenomena described as “ethnic conflicts.”

© The Author(s) 2019 B. B. Kelly, Power-Sharing and Consociational Theory, https://doi.org/10.1007/978-3-030-14191-2_3

55

56  B. B. KELLY

directly addressed this issue, when he described Smith’s definition of a “plural society,” Domination of one of the segments is part of his definition of a plural society. But it is not merely a matter of definition. According to Smith, pluralism necessarily entails the maintenance of political order by domination and force: ‘Cultural diversity or pluralism automatically imposes the structural necessity for domination by one of the cultural sections. It… necessitates nondemocratic regulation of group relationships.’2

Lijphart observes that Smith’s definition of “plural society” implied that “homogeneity is a prerequisite for democratic government.”3 In contrast, the rationale behind Lijphart’s work involving democracy is ­ predicated on his belief that a society’s propensity for stable democracy is not necessarily dependent on its level of pluralism. In fact, a ­fundamental assumption of his theory of consociation is that stable democracy can be promoted “by making [societies] more plural” through enhancement of “the organizational strength of the segments.”4 Accurate quantitative portrayal of this theory requires recognition that Lijphart’s definition of “plural societies” differs substantially from those provided by other scholars. Consistent with this view that greater levels of pluralness can actually be conducive to stability, his definition emphasizes that “plural societies” must be clearly and firmly divided but it does not incorporate Smith’s requirement of “domination.”5 Lijphart’s first, explicit definition of “plural societies” was published in 1981: …four criteria may be used to determine whether a society is completely plural or deviates from perfect pluralism to a greater or lesser extent on one or more of the four dimensions:

1. In a completely plural society, it must be possible to identify exactly the segments into which the society is divided.

2 Lijphart

(1977, p. 18). (1977, p. 19). 4 Lijphart (1984, p. 11). 5 Lijphart (1977, p. 18). 3 Lijphart

3  THE PLURAL SOCIETIES FOR WHICH CONSOCIATION IS PRESCRIBED 

57

2. It must also be possible to state exactly what the size of each segment is, that is, how many people belong to each of the segments. 3. In a completely plural society, there must be perfect correspondence between segmental boundaries and the boundaries between the political, social, and economic organizations. 4. Political parties are one type of organization covered by the third criterion. The final test of a completely plural society is that, since party and segmental loyalties should coincide, there should be little or no change in the voting support of the different parties from election to election: in a perfectly plural society, an election is a segmental census. I do not know any empirical example of a society that can be regarded as perfectly plural according to the above four criteria.6

He also reiterated some of these criteria later, “the basic traits of a plural society are that it is divided into clearly identifiable and measurable segments which have their own social and political organizations.”7 Subsequently, he also defined a “fragmented or plural society” as one “that is sharply divided along religious, ideological, linguistic, cultural, ethnic, or racial lines into virtually separate subsocieties with their own political parties, interest groups, and media of communication.”8 He has suggested that plural societies contain “clearly separate and potentially hostile population segments,” that they are “divided along several lines of cleavage,” and that this “multiplicity of cleavages is reflected in the multidimensional character of their party systems.”9 He seems not to equate the number of segments with the intensity of pluralism, since he distinguishes between “the degree of pluralism and the number of groups into which a society is divided.”10 His statistical ­analysis of democracies published in 1999 suggests that, in all but one of the countries he then identified as plural, in “most… cultural-ethnic issues have high salience” and, in two of them, “religious” issues do.11 Particularly in recent decades, Lijphart has discussed degrees of p ­ luralism, suggesting that clear delineation between plural and non-plural societies may not be

6 Lijphart

(1981, p. 356). (1984, p. 11). 8 Lijphart (1989, p. 39). 9 Lijphart (1977, p. 28; 1999, p. 36). 10 Lijphart (1999, p. 168). 11 Lijphart (1999, p. 83). 7 Lijphart

58  B. B. KELLY

possible.12 His means of identifying plural societies are suggested by his implication that the USA is not a plural society and his argument that the European Union is plural because it comprises “‘[d]eep seated and long-standing national differences’” that are unlikely to disappear.13 This observation emphasizes that Lijphart’s concept of “plural” implies much more than simply diversity of ancestral backgrounds or cultural and ideological traditions. Lijphart’s explicit and implicit expositions of his concept of “plural society” collectively suggest that even ideological divisions can render societies “plural” but that, to be fully “plural,” a society must contain extensive coincidence of cleavages, separating virtual subsocieties that have their own political and social organizations. These divisions must be “clear” and they must remain evident, when subjected to replicable, systematic means of measurement. The more clear and evident they are in a society, the more intense is its level of pluralness. Lijphart’s criteria for identifying plural societies are that (1) they are divided into groups clearly enough to enable easy calculation of their groups’ sizes, (2) the cleavages separating the groups are maintained by their collective actions, and (3) these cleavages are obvious but may correspond to a wide variety of phenomena, such as religion, language, culture, ethnicity, race, or even perhaps ideology. The choice of the exact version of Lijphart’s definition of plural society to be analyzed will be confirmed through examination of alternative definitions offered by other scholars. This analysis will ensure that the definition chosen here will seem legitimate to most of the academic community and thus render this book’s findings more relevant to other research. The term, “plural society,” has been defined in a wide range of ways by writers other than Lijphart. John S. Furnivall was generally considered to be the “formulat[or] of the concept” of “plural society.”14 Furnivall defined it as a society “‘comprising two or more elements or social orders which live side by side, yet without mingling, in one political unit.’”15 He also suggested that plural societies lack “common social demand” and Rabushka and Shepsle interpret

12 For

example: Lijphart (1999, p. 252). (1999, pp. 46, 195). 14 Haug (1967, p. 294). 15 Rabushka and Shepsle (1972, p. 10). 13 Lijphart

3  THE PLURAL SOCIETIES FOR WHICH CONSOCIATION IS PRESCRIBED 

59

this as meaning an absence of “consensus.”16 Furnivall’s work seems to agree with Lijphart’s references to the capability of ideological differences to produce plural societies. He explained that, in these societies, “social demands” generate policies that treat communities separately and thus often benefit one more than the others.17 These differential benefits, and the individual communities’ policies that motivate them, could produce ideological divisions which seem to mirror the cleavages which the groups originally organize around. Furnivall believed that these separate demands constitute the characteristic which distinguishes plural societies from homogeneous ones.18 Furnivall’s theory of plural societies was inspired by colonial dynamics. He seemed not to believe that only certain types of groups are capable of engendering plural societies. His theory seems to suggest that it is the behavior of the group, rather than its means of identifying itself, that marks plural societies. Both he and Lijphart seemed to agree that plural societies are those which are clearly divided into groups whose members’ collective actions reflect the obvious existence of cleavages, and that particular types of group identification are not what characterize these societies as plural. Smith also provided a definition of plural societies. He disagreed with Furnivall’s distinction between plural and homogeneous societies, pointing out that heterogeneity can exist without provoking group isolation.19 Smith believed that plural societies are marked by cultural groups who have their own institutions, which may involve “kinship, education, religion, property and economy, recreation and certain sodalities.”20 Smith’s observation that these group differences are institutionalized in plural societies is consistent with Lijphart’s belief that, in plural societies, this diversity is reflected in political structure. According to Marie Haug, Smith believed that “such differences should produce incompatibility, with the society held together only by power concentrated in the hands of one cultural section.”21 Smith’s point that heterogeneous societies are not intrinsically plural is compatible with Lijphart’s belief that pluralness involves collectively created division rather than uninstitutionalized, 16 Rabushka

and Shepsle (1972, p. 11). and Shepsle (1972, p. 11). 18 Rabushka and Shepsle (1972, p. 11). 19 Haug (1967, p. 294). 20 Haug (1967, p. 294). 21 Haug (1967, p. 294). 17 Rabushka

60  B. B. KELLY

unantagonistic diversity. It is consistent with Lijphart’s recommendation that only groups that identify themselves through the electoral process should be granted segmental autonomy. Particularly because it involves both methodological and political issues, identifying the extent to which heterogeneity is antagonistic is considerably more difficult than simply measuring heterogeneity. This may be why Lijphart suggests identification of groups through the electoral process. Lijphart’s beliefs about the definition of “plural society” seem consistent with Smith’s so they do not offer contradictory recommendations concerning the optimal means of collecting data for this analysis. Smith’s view concerning economic interactions between groups contradicted Furnivall’s, but it is unclear whose theory was more compatible with Lijphart’s. Rabushka and Shepsle explain that, according to Smith, economic relationships can be “irrelevant” to group identities and “it is erroneous to equate cultural pluralism with ‘class stratification,’ since one can vary independently of the other.”22 Compared to Furnivall, Smith appeared to believe that ideological divisions are less conducive to the generation of plural societies, than are other types of divisions. Smith’s work emphasized the fact that most definitions of plural society unequivocally agree with all of the types of divisions that Lijphart lists, except “ideological.” However, it should be kept in mind that, if class cleavages related to ideological divisions are pervasive enough throughout a society, resultant lack of social mobility and intergroup interaction can result in separation of groups as extreme as that found in societies afflicted by severe ethnically motivated conflict. Rabushka and Shepsle also provide a definition of plural societies. They do not believe that “the existence of well-defined ethnic groups with generally incompatible values constitutes a necessary condition of the plural society.”23 They think that the distinctive attribute of plural societies, compared to simply heterogeneous ones, is “the practice of politics almost exclusively along ethnic lines.”24 Rabushka and Shepsle observe that these places are characterized by the interpretation of the “preponderance of political conflicts… in ethnic terms” and “cohesive” action by “ethnic” groups “on nearly all political issues.”25 They elaborate that:

22 Rabushka

and Shepsle (1972, pp. 15–16). and Shepsle (1972, p. 20). 24 Rabushka and Shepsle (1972, p. 20). 25 Rabushka and Shepsle (1972, p. 20). 23 Rabushka

3  THE PLURAL SOCIETIES FOR WHICH CONSOCIATION IS PRESCRIBED 

61

Politically organized cultural sections, communally based political parties, the partitioning of major social groups (e.g., labor unions) into culturally homogeneous subgroups, and political appeals emphasizing primordial sentiments serve as unambiguous indicators of a plural society.26

While Lijphart believes that, in a plural society, the groups are reflected in the political structure, Rabushka and Shepsle contend that virtually all political activity is based on the intergroup divisions. They imply that all political attitudes and ideologies are significantly influenced by group identity. Although Rabushka and Shepsle describe these groups as “ethnic,” their contention that they are divided on virtually every issue implies that their antagonism can involve all of the group attributes listed by Lijphart, “religious, ideological, linguistic, cultural, ethnic, [and] racial.”27 However, like Lijphart and Furnivall, Rabushka and Shepsle do not explain whether ideological cleavages, on their own, can make societies permanently deeply divided. Lijphart himself seems to only mention ideological differences as a possible manifestation of pluralness once in his publications, throughout his many decades of discussing consociation.28 Precise determination of which states are plural requires an answer to the question of whether those divided over ideological issues constitute plural societies. Analysis of Lijphart’s discussions of plural societies suggest that these societies should be represented here as being diverse with regard to religion, language, culture, ethnicity, and/or race, but not ideology. The prudence of this course of action is emphasized by the fact that, of these types of identities that he mentions, it is only commonly considered ethically correct to attempt to change someone else’s ideological identity. Generally it is considered unacceptable now to attempt to compel someone to change their religion, language, culture, ethnicity, or race. However, all ideological positions, regardless of how strongly they are held, are considered open to debate. Even though Lijphart mentioned ideology once as a potential “line” of division in plural societies, it seems legitimate to measure levels of pluralism based on all of the attributes he mentioned except ideology. This significant decision is further justified by Lijphart’s description of the optimal procedure for allowing groups 26 Rabushka

and Shepsle (1972, p. 21). (1989, p. 39). 28 Lijphart (1989, p. 39). 27 Lijphart

62  B. B. KELLY

to delineate themselves politically, prior to being granted segmental autonomy. He explains that voters’ formation of political parties through a proportional representation electoral system allows policymakers to ascertain whether their societies are plural or not plural. He explains that, if a society is plural, “the successful parties will be mainly segmental (and presumably ethnic) parties; if it is not… plural…, the parties that will emerge will be non-segmental, policy-oriented parties.”29 Non-segmental, policy-oriented parties are held together solely by belief systems concerning policy, otherwise known as ideologies. Therefore, Lijphart’s identification of them as emblematic of non-plural societies justifies this study’s exclusion of “ideological” as a form of division characterizing plural societies. Definitions of plural societies offered by Furnivall, Smith, and Rabushka and Shepsle collectively support the conclusion that ideological divisions should not be regarded as contributing to the extent to which societies are plural.30 Of the four concepts of plural societies offered by these sources, only Furnivall’s suggested that ideological cleavages can themselves make societies plural. Even he seemed to believe that it is groups’ collective support of separate demands, rather than the nature of these demands, that constitutes pluralism. In addition, if ideological difference were considered a defining characteristic of plural societies, virtually all societies would be plural to some extent. Comparison of the depictions of plural societies offered by Lijphart and these other scholars facilitates confirmation of the optimal definition that should be reflected in quantitative analyses, if his theory concerning consociation is to be tested most accurately. By this definition, plural societies are diverse with respect to religion, language, culture, ethnicity, and/or race and are also sharply divided because of the collective actions of the people who identify themselves through reference to the groups comprising this diversity. A number of scholars have designed means of identifying cases that constitute something akin to the notion of plural societies which has been chosen for use here. A straightforward means of identifying states that are sufficiently plural to be considered in this analysis would be to 29 Lijphart

(1995, p. 281). The Oxford English Dictionary definition of plural societies also supports the definition of such societies used for this project. A discussion of this definition is provided on the website. 30 &

3  THE PLURAL SOCIETIES FOR WHICH CONSOCIATION IS PRESCRIBED 

63

include those which contain a group that Gurr identifies as a “minority at risk.” If his concept of a “minority at risk” is sufficiently similar to the groups which exist in states which are plural according to the definition of plural societies chosen for this study, this procedure for choosing cases would be appropriate. Gurr describes these “minorities at risk” as “politicized communal groups.”31 He considers them to be politically salient if they fulfill either one or both of two criteria: 1. The group collectively suffers, or benefits from, systematic discriminatory treatment vis-à-vis other groups in a state. 2. The group was the focus of political mobilization and action in defense or promotion of its self-defined interests at some time between 1945 and 1989.32 Research conducted since this excerpt was published in 1993 is ­expanding the time period covered by the dataset. Gurr’s approach to studying “minorities at risk” recognizes that “political and economic inequalities” tend to arise between them when they share a state but that these constitute effects of the groups’ interactions, rather than intrinsic elements of their identities.33 All of Gurr’s categories of “minorities at risk” are described as either “culturally,” “ethnically,” or “religiously” “distinct.”34 The fact that these groups are distinct, experience discriminatory treatment, and act collectively signifies that Gurr’s methodology for identifying them is consistent with the definition of plural societies used here. Gurr’s “minorities at risk” also constitute groups that are religiously, linguistically, culturally, ethnically, and racially identifiable. Like the groups which render societies plural according to Lijphart’s definition, Gurr’s “minorities at risk” characterize a divided society but they are not necessarily numerical minorities in proportion to their states’ population sizes. The societies in which they live are also sharply divided by discrimination and group mobilization, and these activities logically facilitate calculation of the groups’ relative sizes. The striking similarities between Gurr’s criteria for “minorities at risk” and

31 Gurr

(1993, pp. 6–7). (1993, pp. 6–7). 33 Gurr (1993, p. 38). 34 Gurr (1993, p. 18). 32 Gurr

64  B. B. KELLY

the definition of plural societies used here, which consists of diversity plus groups’ ­collective, divisive behavior, suggest that its cases should consist of the set of countries in which at least one of Gurr’s groups lives.35 To maximize the validity of this statistical analysis’ findings, a few categories of countries containing Gurr’s MARs have been excluded from the list of cases. Two lists of his minorities have been consulted to identify all potential country cases and to eliminate some that are not appropriate for this analysis. These publications classify the groups by the countries in which they reside and provide estimates of their population sizes in these countries as of 1990 and 1998.36 Optimally, the lists of cases for this book’s analysis would have been compiled through reference to lists reflecting conditions throughout the period to be covered, 1975–1995. However, the 1990 list appears to be the earliest of its kind available from Gurr’s team. The criteria quoted above for inclusion of a group in their study, involving historic, as well as current, collective suffering or benefit, and political mobilization, emphasize that Gurr’s means of categorization take into account developments since 1945. Groups must meet some additional criteria to be identified as MARs. The groups must reside in countries with populations of at least one million and each group in each country must contain at least one hundred thousand people or one percent of the country’s population.37 Gurr’s study also includes advantaged and dominant minorities, if they mobilize collectively or are the object of discrimination, and occasionally aggregations of small groups if they are usually treated by their country’s rump population as if they constitute a single group (e.g. “native peoples” in the USA).38 If only a very small proportion of a country’s population is comprised of members of groups fitting Gurr’s definition of MARs, it seems inappropriate to label their entire country a “plural society.” For instance, the tiny Korean community in Japan is identified as the country’s only MAR presumably because, even though they constitute only .6% of the overall population (as of both 1990 and 1998), they must number more than one hundred thousand people.39 Inclusion of 35 & The socioeconomic and geographic diversity of the cases analyzed for this project is described on the website. 36 Gurr (1993, pp. 326–338; 2000, pp. 321–336). 37 Gurr (1993, p. 8). 38 Gurr (1993, pp. 8–9). 39 Gurr (1993, p. 327; 2000, p. 323).

3  THE PLURAL SOCIETIES FOR WHICH CONSOCIATION IS PRESCRIBED 

65

a country in the statistical analysis undertaken here implies that its entire “society” is “plural” and this cannot be reasonably claimed unless some substantial percentage of its population has the characteristics of Gurr’s MARs. For this reason, a threshold was chosen, without reference to the data, which will enable elimination of insufficiently plural countries. If a country’s MARs do not collectively constitute ten percent of its overall population, according to the 1990 or 1998 statistics provided by Gurr, it is excluded from this analysis. If either set of statistics estimates this percentage to be higher than ten percent for a country, that country is included. This measure eliminates a diverse set of countries from around the globe, some of which are Australia, China, Croatia, Eritrea, Papua New Guinea, and Russia. Examination of this sample of countries containing MARs to be excluded, as well as the assignment of MAR status to certain countries by Gurr, emphasizes that some of the decisions the MAR dataset reflects concerning individual countries would be criticized by comparativist political scientists and/or country specialists. James Fearon and David Laitin point out that “the criteria for inclusion” in this dataset are “subjective and may be contestable for specific cases.”40 They argue that some group boundaries identified for creation of the data are “contested or unclear,” that some groups should have been recognized as MARs, that some groups have been incorrectly lumped together, and that some “politically dominant” groups are labeled MARs.41 They also point out that some groups are labeled as “minorities at risk,” even though, like the Sunnis in Lebanon who comprise approximately 30% of their state’s population, they “form a ‘majority’ in the sense of a plurality.”42 I think that some of these apparent problems are consistently applied, carefully considered coding rules with which some scholars would disagree. Some groups, like the Lebanese Sunnis, have experienced substantial risk even though they could achieve a plurality and the MAR team seems to have a justifiable reason for including them. Some groups have also been both politically dominant and at risk. Fearon and Laitin point out that the Hutus in Burundi are identified as a “Minority at risk,” even though they constitute a majority of their state’s population.43 Considering that there 40 Fearon

and Laitin (2003, p. 3). and Laitin (2003, p. 4). 42 Fearon and Laitin (2003, p. 4). 43 Fearon and Laitin (2003, p. 4). 41 Fearon

66  B. B. KELLY

were mass killings of Hutus by Tutsis in 1972, prior to the mass killings of Tutsis by Hutus in 1994, it seems that including the Hutus as a MAR even though they form a majority makes sense. Fearon and Laitin also believe that, since groups have to be of a certain size to be included, groups which are smaller and thus less likely to engage in active belligerence are less represented in the dataset and so actively belligerent groups are more likely to be included and this “induces an overprediction of rebellion.”44 They also criticize the lumping together of some discernably separate groups. Perhaps this approach should be justified more explicitly by the MAR team, but it seems like an appropriate response to the fact that many, very at risk groups are too small to be included in the dataset otherwise. In a 1993 publication, Gurr acknowledges that his list of MARs “excludes groups that some regional specialists think should be included,” such as the Scots in the UK, the French speakers in Belgium, and Greeks in Albania.45 He then states that “[a]ll these groups should be added the next time we revise the list” and the version provided in the revision published in 2000 includes the Scots, the Greeks in Albania, but not the French speakers in Belgium.46 Fearon and Laitin argue that “ignoring the [French speaking] Flemish in Belgium is surprising.”47 No Belgian groups were inserted in this updated list, and this is significant because Belgium is one of four countries which Lijphart has consistently identified as having had a very successful experience with consociation. These country-specific, arguable drawbacks of the MAR dataset should be kept in mind by those considering the statistical findings derived from this project’s cases, which were derived systematically from this highly respected dataset. Creation of a flawless dataset indicating every group on earth that is “at risk” according to a methodology that is completely beyond reproach has not been achieved yet. However, the fact that a number of the drawbacks identified regarding this data involve methods which were used throughout the data creation process emphasizes that it was performed according to a consistent, carefully considered methodology. If rebellious groups are overrepresented, as suggested by Fearon and Laitin, this may

44 Fearon

and Laitin (2003, p. 5). (1993, p. 9). 46 Gurr (1993, p. 10; 2000, pp. 322–325). 47 Fearon and Laitin (2003, p. 4). 45 Gurr

3  THE PLURAL SOCIETIES FOR WHICH CONSOCIATION IS PRESCRIBED 

67

cause problems for other research projects. However, considering that I am using data regarding the presence of these groups to try to find plural societies, overrepresentation of the potentially or actually antagonistic groups which constitute plural societies does not affect this project in a detrimental manner. Since trends in instability internationally and through time are the phenomenon represented in this project’s dependent variable, consistently lower representation of nonrebellious groups will not hamper our understanding of where instability occurs most. Drawbacks identified involving MAR group selection do not seem to introduce any consistent bias that could be expected to significantly influence my findings.48 In a few cases, Gurr’s choice of specific MARs for individual countries has led to their exclusion here because of his population size threshold, even though they are often portrayed in scholarly works as fitting the definition of plural societies being operationalized. For example, in a chapter-length case study of Papua New Guinea, Reilly describes the country in a way that makes it sound like it should definitely be categorized as a “plural society.”49 He references these passages by other experts, R. Premdas and J. Griffin: … Papua New Guinea is almost certainly the world’s most ethnically fragmented state in terms of the sheer numbers of independent ethno-linguistic groups… Observers of Papua New Guinea politics have long argued that its ethnic fragmentation represents a ‘formidable and intractable’ impediment to nation-building and that Papua New Guinea’s ‘10,000 micro-societies’ and hundreds of language groups, the largest of which numbers only 150,000 people, presents an almost insurmountable barrier to stable democracy.50

Presumably because they are either insufficiently, collectively threatened or mobilized, or they do not meet his threshold requiring a group to constitute one percent of its country’s population or contain one hundred thousand people, most of these groups in Papua New Guinea were not identified as MARs. The only MAR identified in Papua New Guinea 48 & Fearon and Laitin (2003) point out additional drawbacks of the MAR dataset which are less likely to impact this project’s findings. These are discussed on the website. 49 Reilly (2001, pp. 60–61). 50 Premdas (1989, p. 246) and Griffin (1974, pp. 142–143).

68  B. B. KELLY

is the Bougainvilleans, who constituted approximately 4.6% of the population as of 1990 and 3% as of 1998.51 Perhaps it is unfortunate that a country as widely regarded to be plural as Papua New Guinea is not to be included here because its MARs do not total ten percent. The apparent inconsistency between some qualitative analyses of this country and determinations made for the MAR dataset suggest that more research is needed to accurately describe the country’s intergroup dynamics for academic and policymaking purposes. However, to ensure impartiality in case identification for this project, determinations regarding MARs’ sizes and intergroup relationships must be accepted. By not deriving cases according to subjective assessments by its author, this statistical analysis avoids a pitfall more problematic than the exclusion of a few, potentially interesting cases. The choice of cases for this statistical analysis cannot be criticized for containing any bias introduced through considerations involving the subject or outcome of this study of consociation. The cases will be chosen through reference only to choices made by Gurr and his research team, a ten percent threshold that was chosen without consideration of individual countries, and exclusion of the regions inhabited by a few MARs that do not constitute independent countries. Since Gurr’s definition of a MAR, reflecting its collective treatment and activities, closely corresponds to the definition of plural societies which this book aims to analyze accurately, the identification of cases according to his definition does not introduce any undesired effects. However, the significance of his policies of restricting MAR status only to groups and countries of a certain size and identifying advantaged, often dominant, groups as MARs will be acknowledged when the statistical results obtained for this book are examined. It must also be recognized that the arbitrary ten percent threshold for excluding cases does not take into account the strength of problems and collective action experienced, and exhibited, by groups. Biases within the list of cases that are not connected to consociation and not subjectively introduced by this book’s author include a lack of adequate consideration of regions not constituting individual countries, countries with small overall or MAR populations, countries with greater numbers of advantaged MARs, and countries with small numbers of MAR inhabitants whose activities and conditions are more intense.

51 Gurr

(1993, p. 331; 2000, p. 329).

3  THE PLURAL SOCIETIES FOR WHICH CONSOCIATION IS PRESCRIBED 

69

Three categories of countries containing at least one MAR have been excluded from this analysis. They include those whose MARs collectively do not constitute ten percent of their overall populations, those which are not truly independent countries containing the MARs listed for them, and those for which appropriate data was not available for the dependent variable representing stability. These are countries which were included in Gurr’s dataset but have been excluded from the dataset used for this book for methodological reasons. Of the 117 places identified as containing MARs according to either 1990 or 1998 population data, 30 were eliminated because the population figures provided by the MAR research team indicated that the MARs within them did not collectively meet the arbitrary threshold of ten percent. The countries that were eliminated because they are not independent countries containing the MARs identified with them are, in Gurr’s exact terminology: Eurocommunity, Nordic Countries, Taiwan, Cyprus, and Yugoslavia (1992–1995). Taiwan is not included because it is not widely recognized as independent from China. Cyprus is excluded because the MAR identified for it, Turkish Cypriots, mostly inhabits the Turkish portion of Cyprus, which some consider to be a different country. The Turkish Cypriots who live in Greek Cyprus do not constitute 10% of that society’s population. They thereby do not fulfill the threshold for this statistical analysis, so Cyprus must be excluded from it. It is unfortunate that this project’s methodology excludes Cyprus because the breakdown of consociation governance mechanisms there in 1960 is often considered to constitute a failure of consociation. Its absence may bias the results of this analysis because of this experience with the system recommended by Lijphart, but it is not the only case which used consociation and was excluded. Northern Ireland also had to be left out of the statistical analysis so its largely positive experience with the political system seems likely to introduce a bias that may counteract Cyprus’ exclusion. Yugoslavia also is not included for the years, 1992–1995, because it was effectively a different country from 1992 and the only population data I have for it is from 1998, which may drastically differ from the actual makeup of the population from 1992 to 1995. The “Former Yugoslavia,” identified as “Yugoslavia” in the 1993 publication, is included in this analysis for 1975–1991, using the 1990 population data provided in that publication. Burma is listed with its alternative name, Myanmar, in the dataset only because it was identified that way in another dataset used for this analysis. Unfortunately, the Soviet Union had to be eliminated

70  B. B. KELLY

from this analysis because the dependent variable and data used here was constructed through data available in Gurr’s Minorities at Risk dataset, which was not available for the Soviet Union. The one category of countries that is included in this dataset even though they were not identified as containing MARs were three of Lijphart’s cases of apparently very successful consociation, which he specified in addition to Switzerland (which is a case for this analysis). Unfortunately, the data required for making dependent variable data representing stability for these countries, Austria, Belgium, and the Netherlands, were also not available in the Minorities at Risk dataset so data for the independent variables cannot be included here and the influence of consociation on stability in these countries cannot be quantitatively analyzed here. While similar data could conceivably be made to take the place of this data derived from Gurr’s dataset, its separate construction would inevitably provoke criticism regarding subjective, potentially biased decision-making and its comparability with other countries’ values for the dependent variable would be inferior. These countries’ experiences will be analyzed qualitatively in Chapters 6 and 7. Switzerland is included in the statistical analysis because more than ten percent of its population is comprised of MARs. It is also a special case because these MARs, Jurassiens and Foreign Workers, do not correspond to the cleavage identified by Lijphart in connection with consociation. For this reason, the data which were constructed through reference to attributes of MARs for the rest of the countries were made instead according to the cleavage identified by Lijphart for Switzerland. The lack of some other data for additional countries has not required their elimination from this study, but it should be recognized that conclusions concerning certain independent variables will not necessarily reflect the experiences of every country. Some other scholars’ datasets were used to make and collect data for this analysis, and they were occasionally missing data for some years of countries’ experiences. Therefore, these resulting data absences were then also introduced into the dataset to be used here. This problem normally affected only some years for each of these individual countries. However, no data for the minority veto power variable (MV) was created for some countries because it was constructed through analysis of operational, codified, single constitutional documents and some countries either did not use such documents between 1975 and 1995, or such documents could not

3  THE PLURAL SOCIETIES FOR WHICH CONSOCIATION IS PRESCRIBED 

71

be accessed during the data construction process. For example, no MV data exists for Canada, the UK, New Zealand, and Bhutan. However, the absence of MV data for these countries affects their representation in the dataset but does not require their elimination. The representation of one other intriguing place will also necessarily be suboptimal. Since the cases are independent countries and inclusion of other entities would hinder valid comparison of cases, Northern Ireland’s experience of consociational components (which were not used throughout the UK) cannot be adequately assessed in this statistical analysis. The data representing the UK will take Northern Ireland into account, but examination of this region on its own must be restricted to a qualitative treatment in Chapter 9. This quantitative analysis examining the role of consociation in plural societies uses cases that represent each independent country containing one or more of Gurr’s MARs that collectively constitute more than ten percent of its population. These cases comprise eighty-eight countries. When the findings of these quantitative analyses are considered, it should also be remembered that consociation is prescribed for those countries whose populations are not only diverse but also whose segments are at least potentially antagonistic. The theory to be tested here is that consociation produces stability in plural societies so these societies must comprise the cases for the quantitative analyses. However, instability is inherently more likely in these cases and it should be considered that consociational components may actually be more conducive to stability in non-plural societies. Lijphart’s and others’ discussions of plural societies and ethnic and communal conflicts together illustrate which definition of plural societies should be used in this study to ascertain whether the elements of his consociational system do encourage stability in plural societies. Plural societies will be defined as those which are clearly divided into groups whose members’ collective actions reflect the obvious existence of cleavages corresponding to religion, language, culture, ethnicity, and/or race. Since Gurr’s methodology for choosing cases for his dataset, Minorities at Risk, is very similar to this definition, most states including one of the groups constituting his cases, will be adopted as cases here. A few of these states will not be used in this analysis for the reasons described above to maximize the accuracy of the statistical findings. Their exclusion will not introduce any biases connected to the proposition to be tested.

72  B. B. KELLY

References Fearon, James D. and David D. Laitin. “Collaborative Project: ‘Minorities at Risk’ Data Base and Explaining Ethnic Violence: NSF Grant Proposal.” 2003. Griffin, J. “Papua New Guinea.” Modern Asia: Problems and Politics. Eds. R. Brissenden and J. Griffin. Milton: Jacaranda Press, 1974. Gurr, Ted Robert. Minorities at Risk: A Global View of Ethnopolitical Conflicts. Washington, DC: United States Institute of Peace Press, 1993. Gurr, Ted Robert. Peoples Versus States: Minorities at Risk in the New Century. Washington, DC: United States Institute of Peace Press, 2000. Haug, Marie R. “Social and Cultural Pluralism as a Concept in Social System Analysis.” The American Journal of Sociology. 73:3 (November, 1967) 294–304. Lijphart, Arend. Democracy in Plural Societies: A Comparative Exploration. New Haven: Yale University Press, 1977. Lijphart, Arend. “Comment: Consociational Theory: Problems and Prospects: A Reply.” Comparative Politics. 13:3 (April, 1981) 355–360. Lijphart, Arend. “The Politics of Accommodation: Reflections—Fifteen Years Later.” Acta Politica. 19:1 (January, 1984) 9–18. Lijphart, Arend. “Democratic Political Systems: Types, Cases, Causes, and Consequences.” Journal of Theoretical Politics. 1:1 (1989) 33–48. Lijphart, Arend. “Self-Determination Versus Pre-determination of Ethnic Minorities in Power-Sharing Systems.” The Rights of Minority Cultures. Ed. Will Kymlicka. Oxford: Oxford University, 1995. Lijphart, Arend. Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries. New Haven, CT: Yale University Press, 1999. Premdas, R. “Ethnicity and Nation-Building: The Papua New Guinea Case.” Ethnicity and Nation Building in the Pacific. Tokyo: United Nations University, 1989. Rabushka, Alvin and Kenneth A. Shepsle. Politics in Plural Societies: A Theory of Democratic Instability. Columbus, OH: Charles E. Merrill, 1972. Reilly, Ben. Democracy in Divided Societies: Electoral Engineering for Conflict Management. Cambridge: Cambridge University Press, 2001.

CHAPTER 4

Empirically Depicting Consociation’s Role

Statistical assessment of the claim that consociation promotes stability in plural societies requires identification of appropriate variables and data to represent this proposition. Variables must be chosen which reflect consociation and stability as accurately as possible, in a manner which prevents methodological problems which could invalidate this study’s conclusions. For instance, the results would be inaccurate if phenomena other than consociation, instead of this political system itself, actually explained trends in stability which only appeared to be produced by consociation. False conclusions generated in this scenario can be avoided through inclusion of “controlling” variables representing factors which may account for consociation’s apparent successes and failures. Data representing all of the variables was assembled through determination of optimal types of information and subsequent collection and creation of data which resemble these ideal types as closely as possible.

Part 1: Dependent Variable Representing the Goal of Stability Lijphart was initially inspired to identify the consociational system following his observation of long-term stability in societies which had been categorized as intrinsically less likely to achieve and maintain it. Maximization of the accuracy with which his theory of consociation is represented quantitatively is pursued here through meticulous © The Author(s) 2019 B. B. Kelly, Power-Sharing and Consociational Theory, https://doi.org/10.1007/978-3-030-14191-2_4

73

74  B. B. KELLY

comparison of his descriptions of stability and subsequent identification of the available data that most closely corresponds to its components. These components identified through qualitative analysis are system maintenance, effectiveness, civil order, and legitimacy. A thorough search through all of Lijphart’s works involving consociation indicates that his most detailed definition of stability is provided in this passage from a 1977 publication: Political stability is a… difficult and ambiguous term. It will be used in this study as a multidimensional concept, combining ideas that are frequently encountered in the comparative politics literature: system maintenance, civil order, legitimacy, and effectiveness. The foremost characteristics of a stable democratic regime are that it has a high probability of remaining democratic and that it has a low level of actual and potential civil violence. These two dimensions are closely related; the latter can also be viewed as a prerequisite for, and as an indicator of, the former. Similarly, the degree of legitimacy that the regime enjoys and its decisional effectiveness are related both to each other and to the first two factors. Jointly and interdependently, these four dimensions characterize democratic stability.1

This definition indicates that he views the concept as comprised of four elements: “system maintenance,” “civil order,” “legitimacy,” and “effectiveness.”2 Lijphart also believes that stability coexists with low levels of “actual and potential civil violence” and that stable democratic regimes tend to remain democratic.3 Accurate interpretation of this project’s findings requires recognition that Lijphart’s definition of stability differs from those described by other academics. For instance, G. Bingham Powell Jr. differentiates stability from political order and defines it as “[a]verage durability of governments or presidents.”4 In his later application of some of Powell’s data, Matthew Krain analyzes two political order related phenomena, “riots” and “political deaths,” as his dependent variables.5 Examination of other quantitative treatments of stability emphasizes that perfect replication of Lijphart’s notion of stability 1 Lijphart

(1977, p. 4). (1977, p. 4). 3 Lijphart (1977, p. 4). 4 Powell (1982, pp. 19–21). 5 Krain (1998, pp. 139, 148). 2 Lijphart

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

75

for this project requires depiction of stability through a dependent variable incorporating data corresponding to system maintenance, civil order, legitimacy, and effectiveness. The variable should also reflect the existence of actual or potential civil violence as an indicator of instability, and the overall statistical model should be constructed in a manner which enables assessment of the influence of consociation in democratic countries. A number of qualitative issues regarding the optimal means for depicting these phenomena must be considered before the identification of appropriate data for representation of stability. The definition quoted above illustrates Lijphart’s belief that stability and democracy very often coexist. However, it should also be noted that he does not identify democracy as one of the four core elements of stability. Lijphart commonly describes consociation as “consociational democracy,” thus implying that consociation and democracy are closely related. Although the relationship between consociation and democracy has been intensely debated by other academics, Lijphart suggests that consociation is designed to encourage and maintain democracy. Considering both this belief and critiques that consociation is intrinsically, inadequately democratic, it is prudent to include a control variable representing democracy. This will allow comparison of the effects of consociation and democracy on stability. The specific, most justifiable measure of stability for this analysis is identified through qualitative analysis of exactly those situations that Lijphart considers to be stable and unstable. This illustrates his belief that a loss of stability consists of the suspension of the main government institutions, the accession to power of individuals or groups opposed to peaceful coexistence of segments, and/or the eruption of violence and “civil disorder.”6 Lijphart does not consider the end of consociational government to constitute a loss of stability, if it does not result in these problematic developments. He portrays consociation as a form of government that can restore normal political relations and encourage a sense of constitutional loyalty, but that can also eventually be replaced with more competitive democracy without engendering instability.7 He persuasively argues that, in plural societies, more competitive democracy cannot perform these roles as effectively as consociation. Lijphart

6 Lijphart 7 Lijphart

(1977, pp. 152–153, 159). (1977, pp. 50–52).

76  B. B. KELLY

believes that institutional inertia and other factors mean that consociation should not be expected to “vanish” when a society “becomes less plural.”8 However, he has constructed consociation to allow it to “provide for its own abolition,” when a society becomes able to support increased political competitiveness.9 Opposition is more likely in plural societies to result in permanent political exclusion of such groups and consequent instability. Debates concerning the relationship between opposition and stability in democracies thus may be less relevant to this book’s quantitative findings because it focuses on these plural societies.10 Lijphart also identifies measures of political and democratic stability that he judges to be inadequate. He rejects the contention that a system’s stability is related to the number of political parties operating in it.11 Lijphart also disapproves of the use of “executive durability” as a measure of “government stability” because he believes that it is only an indication of the executive’s power vis-á-vis the legislature, rather than of the effectiveness of its policymaking.12 He agrees with Mattei Dogan that cabinet stability is not equivalent to regime stability.13 Lijphart’s conception of democracy as a means of representing groups, more than as a system of alternation in government, is emphasized by his rejection of “turnover test[s],” which imply that turnover of government parties is equivalent to stability.14 Above all, Lijphart equates stability in consociational systems with the peaceful coexistence of subcultures. Lijphart’s focus on subcultures’ representation seems to lead him to assess stability levels through reference to “system maintenance, civil order, legitimacy, and effectiveness,” more than phenomena which have less to do with intergroup stable relations, such as government alternation and specific numbers of political parties.15 The most fundamental element of consociation is “grand coalition” government, which Lijphart defines as constant, joint governance by representatives of all potentially antagonistic population segments. 8 Lijphart

(1989, p. 141). (1975, p. 219). 10 Lijphart (1977, pp. 2, 228–229). 11 Lijphart (1968, p. 35). 12 Lijphart (1999, p. 261). 13 Lijphart (1999, p. 130). 14 Lijphart (1999, p. 7). 15 Lijphart (1977, p. 4). 9 Lijphart

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

77

The central benefit of this component of consociation is precisely that it guarantees political power for minority groups, regardless of their electoral strengths. The intrinsic nature of this system explains Lijphart’s contention that its stability will not be accurately measured according to frequency of party turnover in government, executive durability, and cabinet stability. These variables will not be included in this study because its aim is to determine whether consociation performs the role ascribed to it by Lijphart. Examination of Lijphart’s body of work suggests that the ideal variable for representing stability here would constitute an index incorporating measures of the four components that he identifies as comprising the phenomenon: system maintenance, civil order, legitimacy, and effectiveness. In addition, the data used to represent them should include measures of actual or potential civil violence. Unfortunately, two of these components cannot be incorporated as elements of this dependent variable because their inclusion would jeopardize the methodological soundness of the analysis. The first is system maintenance. In his later writings, Lijphart describes consociation as a system which, in the best case scenario, will evolve into a slightly more competitive system and basically render conflict mitigation techniques of all kinds obsolete. By definition, consociation is a governance system in which available measures of system maintenance are not indicative of stability. Examination of the body of his work involving consociation suggests that Lijphart would consider a system as remaining consociational, even if the rules generally recognized as constituting a political system were drastically changed. For instance, he believes that a very widely representative Congress Party in India is capable of making the country consociational, no matter how it came to power and whether or not it rules alone.16 Lijphart explains that: …the Congress Party has served as the foundation for a consociational grand coalition. Despite never winning a majority of the popular vote in parliamentary elections,… [t]he combination of the Congress Party’s inclusive nature and political dominance has generated grand coalition cabinets with ministers belonging to all the main religious, linguistic, and regional groups.17

16 Lijphart 17 Lijphart

(1996, p. 260). (1996, p. 260).

78  B. B. KELLY

He implies that it is the presence of representativeness, rather than the actual institutionalized mechanisms for making decisions and choosing representatives, which really make a society consociational. If India went from a majoritarian means of making decisions and choosing politicians, to a purely consensual one, but the government remained equally representative, Lijphart would consider the system to have remained consociational. This perspective is in direct contradiction to the general notion of “system maintenance” and the data available to represent it, which focus on the type of institutions rather than their political effects. For this reason, a variable corresponding to “system maintenance” would simply mirror changes in the institution-related independent variables, rather than the sort of political effects implied by Lijphart as constituting stability in plural societies. By definition, Lijphart’s consociational system and institution-related system maintenance as an indicator of stability are fundamentally at odds with one another so the latter will not be employed for this quantitative analysis. The institution-related data available for representation of system maintenance should also not be reflected in the dependent variable because it is too closely related to indispensable independent variables concerning institutional mechanisms. The independent variables corresponding to the consociational components of proportional representation (PR) electoral systems, minority veto power (MV), and “grand coalitions” reflect institutionalized aspects of governance systems. If institution-related data indicating system maintenance were incorporated in the dependent variable, analyses of the effects of consociation would be testing whether the existence of certain institutions led to their existence. However, Lijphart’s identification of system maintenance as a fundamental component of stability emphasizes that his concept of this ­phenomenon should be quantitatively examined here. Lijphart seems to imply that the manifestations of stability in countries with consociational experience are continued representativeness of potentially antagonistic groups and continued democracy. No data covering the cases and dates of this book’s dataset is available to indicate the notion of representation for these country-specific groups. The data corresponding to system maintenance that is available documents the persistence of institutions, and most of it actually focuses on indicating the presence of democracy. Lijphart’s notion of system maintenance as the permanence of a particularly representative form of democracy is indicated better by the four core independent variables corresponding to consociational components.

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

79

If the persistence of these components were reflected in both the independent and dependent variables, the resulting statistical model would have very little explanatory power and it would be fundamentally, methodologically flawed. Lijphart’s four components of stability (system maintenance, civil order, legitimacy, and effectiveness) can be divided into two categories, those involving the operation of the political system, and its effects on society.18 “System maintenance” and “effectiveness” both focus on the operation of the political system. Lijphart’s “effectiveness” element of stability must be excluded from the dependent variable because the data available to represent it seems incapable of representing the sort of effectiveness that Lijphart believes that consociational systems should seek. Examination of Lijphart’s depictions of “effectiveness” indicates that, like many other scholars, he views it to be efficiency of policymaking.19 However, the most decisive policymaking is obviously achieved by those governments which provide the least power to challenge or veto majority decisions. Lijphart’s four components of consociation were designed precisely to guarantee the consideration of minority opinions. Therefore, by definition, consociation is designed to be less “effective,” with less efficient policymaking, than other forms of democracy such as majoritarianism. Lijphart’s claim that consociation promotes “stability” and his identification of effectiveness as a central element of stability together suggest that the sort of effectiveness to be achieved ideally in consociational systems differs from that which is reflected in available data focusing on efficiency. For this reason, existing effectiveness data is not appropriate for incorporation in the dependent variable. It seems reasonable to interpret Lijphart’s works as implying that consociational systems, which he sees as producing democracy, promote their continued operation and that this operation constitutes effectiveness. While the continued operation of democratic governance in undivided, long-term democracies may not seem indicative of an exciting level of effectiveness, such effectiveness is a terrific achievement in the deeply divided, plural societies for which consociation is commonly prescribed. It would appear that Lijphart’s claim regarding effectiveness, and the system he calls consociational democracy, is that it promotes a 18 & Discussion of the relationship between democracy and the dependent variable is on the website. 19 Lijphart (1977, p. 108; 1999, pp. 8, 258).

80  B. B. KELLY

situation in which democratic governance continues in the long-term. This interpretation is consistent with his definition of stability quoted above, which explains that one of the two “foremost characteristics of a stable democratic regime [is] that it has a high probability of remaining democratic.”20 Such continuation of democracy intrinsically requires that a country’s government can make policies and, therefore, exhibit some level of the type of effectiveness that is reflected in available data. This situation inspires consideration of a means of data manipulation to create appropriate effectiveness information for incorporation in the dependent, stability variable. Data could conceivably be created which would indicate continued policymaking abilities but would not grant higher values of effectiveness to extremely efficient and efficient policymaking countries. This data would perhaps allocate the same value to all systems achieving some level of effectiveness indicating continued, rather than very efficient, policymaking activity. However, a theoretical difficulty would be incurred by this course of action that would not arise if the stability element of effectiveness is incorporated in a component of this statistical model that is not part of the dependent variable. This difficulty is caused by the fact that Lijphart’s concept of effectiveness stability in consociational countries is of the continued operation of democratic policymaking mechanisms, rather than policymaking mechanisms in any sort of country. Policymaking is obviously performed very decisively by dictators and other leaders of undemocratic countries. Data focusing only on policymaking does not represent accurately Lijphart’s concept of the effectiveness component of stability in consociational societies. Lijphart’s concept of effectiveness can be included in these statistical analyses without relying on such problematic, policymaking-focused data. Data combining information regarding the presence of democracy and policymaking could be used to represent the continued democratic policymaking that Lijphart seems to consider to be the manifestation of effectiveness in consociational states. If democracy were included in the dependent variable while a number of the core independent variables indicate the presence of institutions and practices closely related to democracy, the relationships between these variables’ components would contain inadequate variation and would thus be methodologically

20 Lijphart

(1977, p. 4).

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

81

invalid. The inadvisability of including democracy in the dependent variable becomes more obvious when one considers that Lijphart views consociation as a form of democracy. It is illogical and unproductive to study the effects of four components of a type of democracy, on democracy. Like Lijphart’s system maintenance element of stability, the effectiveness element also cannot be represented in the stability dependent variable. The other manifestations of stability ideally represented in the dependent variable correspond to Lijphart’s remaining two elements of stability, civil order, and legitimacy. Civil order and legitimacy are criteria for the measurement of governance mechanisms’ effects on society, rather than indications of whether these governance mechanisms are operational. They are reflected in the dependent variable. A point must be made about the interpretation of results for those analyses conducted with the larger set of cases. The absence or low incidence of the violent and nonviolent forms of opposition, which indicate levels of civil order and legitimacy, cannot be relied upon to indicate true legitimacy and consequent prospects for long-term stability when undemocratic regimes exert substantial repression of opposition. Lustick has emphasized that the absence of violence cannot be assumed to signify legitimate stability because “control” systems repress grievances, unlike democratic systems, such as consociation, which aim to promote consideration of them.21 In undemocratic countries, apparent stability could just mask severe oppression. Thankfully, the cases will be analyzed with a democracy control variable that will suggest the extent to which levels of democracy are yielding levels of stability, in the forms of violent and nonviolent opposition. In these models, if this democracy variable explains stability better than the other independent variables, it will become obvious that stability levels in undemocratic countries are being influenced much more by levels of democracy than the phenomena represented by the other independent variables. A plausible means of creating a dependent variable that is more independent of democracy levels would be to represent the legitimacy component of the variable through data indicating the attitudes of, rather than the actions of, the inhabitants of each country. Although their results may also be warped by respondents’ fears of oppression, survey data indicating levels of legitimacy could probably be relied upon to be more indicative of legitimacy levels than

21 Lustick

(1979).

82  B. B. KELLY

active protest, in undemocratic societies. However, the unavailability of this sort of survey data for most of the countries and years included in this study prevents its use as a component of the dependent variable. It should be recognized that research by some scholars suggests that the relationship between repression and opposition may not be as strong as is sometimes assumed. In quantitative studies focusing on resource mobilization, Edward Muller and Muller and Mitchell Seligson have shown that violence is not simply positively correlated with repression, since it is most common in situations of moderate levels of repression and resources.22 Their findings illustrate that it is inappropriate to assume that a straightforward, consistently negative relationship exists between democracy and opposition expressed violently and nonviolently. The occurrence in democratic societies of large-scale, nonviolent forms of opposition, such as protests, is also arguably an indication of well-developed, stable democracy. Some scholars may argue that nonviolent forms of opposition should not be incorporated into the dependent variable, as manifestations of instability. However, this statistical model is designed to test Lijphart’s argument that, in plural societies, consociation promotes stability and this stability is composed of four elements, one of which is legitimacy. The term, opposition, suggests activity that is not provided for in states’ formal or informal constitutions and ­formal governance structures. Regardless of whether it is violent or nonviolent, active opposition by a state’s inhabitants against their state or its ­policies indicates that its participants consider the channels of participation outlined in their formal or informal constitutions to be inadequately effective means of influencing their state, and thus lacking in legitimacy. While active opposition to state policies may be regarded to be an almost official channel for participation in a relatively small number of countries, such opposition to one’s state inherently constitutes an assertion that that state policies are illegitimate. Even violent opposition could theoretically be condoned by the leaders of a state as a normal means of political participation. Ideally, those situations in which opposition is not as motivated by legitimacy related grievances could be distinguished from those in which participation is inspired by such grievances. However, such a distinction would require extensive survey data indicating protestors’ relevant opinions. Sufficient amounts of such data are unavailable

22 Muller

(1985) and Muller and Seligson (1987).

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

83

for this analysis so legitimacy feelings associated with opposition cannot be quantitatively distinguished. Creation of a dependent variable which approximates Lijphart’s concept of stability as closely as possible requires data reflecting legitimacy. Since data for nonviolent forms of opposition is the best information available for representation of this component of stability, the dependent variable corresponding to stability will combine this data with that corresponding to the civil order element of violent opposition. Frank Cohen describes “nonviolent protest” as “the most frequent but least severe form of conflict.”23 Horowitz explains ethnic conflict’s “ubiquitous character,” which includes developments as diverse as “business rivalries” and nonviolent secessions.24 Gurr, whose dataset is arguably the most respected one focusing on divided societies, similarly defines conflict as comprising “nonviolent protest, violent protest, and rebellion.”25,26 The Minorities at Risk dataset provides data that can be used to represent the legitimacy and civil order elements of the dependent variable corresponding to stability. A dependent index variable has been constructed to combine data concerning these elements. It includes information on group protest and intercommunal rebellion in these plural societies prone to group organized antagonism. The Minorities at Risk codebook confirms that the rebellion and protest data together represent violent and nonviolent activity.27 This data covers a wide range of activities, including verbal protest, petitions, massive nonviolent demonstrations, riots, strikes, rebellions, and civil war.28 For each of the protest and rebellion variables in the Minorities at Risk dataset, higher values indicate greater amounts and intensity of conflict. There are four such variables: a protest index with values corresponding to five-year periods in each country’s history from 1945 to 1999, another protest index for which annual values are available for the period 1985–2000, and two indexes covering rebellion with values available for the same periods and years as the two protest indexes. Data for the dependent index variable, which incorporates both protest and rebellion, was made by adding 23 Cohen

(1997, p. 618). (1985, pp. 5, 113, 229). 25 Bonneuil and Auriat (2000, p. 565). 26 & Nonviolent manifestations of instability are discussed in more detail on the website. 27 Gurr et al. (2002, pp. 152–181). 28 Gurr et al. (2002, pp. 152–181). 24 Horowitz

84  B. B. KELLY

each country’s quinquennial data for protest and rebellion for each year before 1985 and adding the corresponding annual data to construct values for each year of countries’ experiences from 1985 until 1995. To conclude: in reference to plural societies and consociation, Lijphart provides his most detailed definition of the elements of stability, system maintenance, effectiveness, legitimacy, and civil order.29 Analysis of his portrayal of these elements suggests that he understands them to correspond to the persistence of a consociational-like, inclusively representative form of democracy, continued democratic policymaking, and nonviolent and violent forms of opposition. The former two factors relating to government structure cannot be represented in the dependent variable for statistical analysis. Fortunately, it is possible and methodologically sound to create an index, dependent variable that combines data corresponding to the legitimacy and civil order elements of stability. The importance of all four of Lijphart’s elements of stability is emphasized by his statement that “[j]ointly and independently, these four … characterize democratic stability.”30

Part 2: Independent Variables Representing Consociation Statistical analysis of the effects of consociation requires independent variables representing this political system, as well as independent, controlling variables which are also thought to significantly influence stability in plural societies. The consociational variables correspond to the four components of this system, grand coalition, segmental autonomy (SA), PR, and MV. The controlling, independent variables chosen for inclusion in this study involve external threats, moderate multiparty systems, geographical concentration of segments, population size, socioeconomic equality, and democracy. Data for the controlling variables and PR is borrowed from other datasets, while newly created data represents MV and new data incorporating information from some other datasets represents grand coalition and SA. Grand coalitions including representatives from all potentially antagonistic population segments are presented by Lijphart as the most crucial 29 Lijphart 30 Lijphart

(1977, p. 4). (1977, p. 4).

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

85

institutional element of consociation. The variables designed to measure the presence of grand coalitions could either indicate the proportion of voters represented by politicians in coalitions, or simply whether or not all-inclusive coalitions exist. The latter option is the optimal means of constructing these variables because it must correspond to an element of the consociational system, instead of an effect of consociation. It has been claimed that Lijphart does not distinguish adequately between components and ramifications of consociation. Unintentional testing of the connection between different ramifications of the system would suggest a positive relationship that is not necessarily connected to consociation itself. For instance, intergroup elite negotiation is obviously conducive to decreased group violence but identification of this relationship implies nothing about consociation. Government by all-inclusive coalition, especially when it is permanent, might very possibly become illegitimate in the eyes of voters because it would virtually preclude cabinet turnover following elections and effective opposition to cabinet decisions. Therefore, all-inclusive coalitions are not unarguable manifestations of long-term stability. In contrast, coalitions which do not include all politically significant groups in society do not preclude government turnover or opposition. For these reasons, the presence of all-inclusive, but not semi-inclusive, coalitions ideally should be represented in this independent variable corresponding to grand coalition. However, examination of the actual experiences of the countries to be represented here, and indeed all countries in the world, indicates that such all-inclusive coalitions are extremely rare. Many coalitions identified by Lijphart as being all-inclusive actually do not contain all political parties in their societies. Representation of this consociational component through data with very little variation, resulting from the extreme rarity of grand, all-inclusive coalition in its ideal form, would render this variable methodologically problematic. To avoid methodological problems that would result from identifying only the ideal form of grand, all-inclusive coalition, variables indicating the much more common variations of this phenomenon were used in this project. Coalitions including one or more, but not all, Minorities at Risk (MARs) will be coded positively for the first “grand coalition” variable used here, “somewhat inclusive grand coalition,” indicating that they at least somewhat resemble Lijphart’s grand coalitions. It is accurate to identify executives that do not include all MARs as grand coalitions because, as O’Leary explains, “[w]hat matters [for consociation]

86  B. B. KELLY

is meaningful, cross-community, joint decision making within the executive.”31 Coalitions including members of all of a country’s MARs will be allocated positive scores for the second “grand coalition” variable, “highly inclusive grand coalition.” When countries only contain one MAR and a member of it is included in executives, they are allocated positive scores for this second variable. Particularly, when the MARs are included in coalitions specifically to promote the mitigation of antagonism, the spirit of true, grand, all-inclusive coalitions is present and they thus constitute some degree of consociation. Identification of a coalition as inclusive because of its representation of only one group is more appropriate than it may first appear because this group has been identified by Gurr’s team as a “minority” and “at risk,” both of which are attributes which make a group much less likely to possess supreme political power. Its extreme rarity means that Lijphart’s consociational component of grand coalitions cannot be represented statistically in its purest form. Its absence should be kept in mind. Completely inclusive executive coalitions, with their inherently very limited, accompanying potential for opposition, also have a much less obvious relationship with stability than do executives which simply contain one or more MARs. For this reason, relationships found in this dataset between the inclusive coalition variables and stability will be less valuable and interesting than those involving the other consociational components, whose variables are less clearly associated with stability. It should also be kept in mind that the methodology of identifying cases as consociational based on reference to the inclusion of its MARs in executives unfortunately decreases the accuracy with which local and regional consociations are represented in the analysis. The case which is impacted negatively in this way here is the UK, since only one of the historically antagonistic groups in Northern Ireland is included as a MAR, so the data representing grand coalition underrepresents the extent to which MARs are included in executives there. This shortcoming should be considered when the quantitative results are assessed and it emphasizes the importance of considering Northern Ireland’s experience qualitatively. The variables corresponding to the presence of grand coalitions for this quantitative analysis are represented with data made through reference to the World Bank Database of Political Institutions,

31 O’Leary

(2005, p. 12).

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

87

which is compiled by Philip Keefer, and a wide range of internet sites found through extensive Google searches.32 For each country, a substantial attempt is made to ascertain whether its MARs have been in executive coalitions during the years from 1975 until 1995. Separate pieces of data are made for each year of each country’s experience during this period. The list of MARs to be investigated for each country corresponds to the group identified in the Appendix of the 1993 Minorities at Risk publication, unless a new country was not listed in that publication.33 In that case, the MARs residing in it are found in Appendix C of the 2000 version of that publication.34 For each year between 1975 and 1995, Keefer’s data for individual countries and a number of variables are consulted to ascertain whether any MARs were in coalition governments. This data indicates whether the chief executive was from a political party (rather than a military, royalty, etc.), how many coalition parties existed and the identities of the three biggest ones, if a “primary component” of those parties’ platforms was “nationalist,” “regional,” or “religious,” and the number of opposition and government seats existing each year.35 If examined collectively, this information often indicates whether parties focusing on representing MARs were in the executive. When it is not adequate to enable determination of this question, terms relevant to the situation in each country are used in internet searches that usually facilitate accurate discovery of the required data. If complete information regarding a country’s coalition partners cannot be found through these two avenues, it is concluded that no evidence of coalitions including MARs can be found. A case corresponding to a year of a country’s experience is allocated a “0” for this variable in any one of three scenarios: (1) if no evidence can be found of such a coalition, (2) if it is confirmed that no such coalition existed, or (3) if one of the MARs identified as “advantaged” by the Minorities at Risk team dominates power. It should be pointed out that these variables do not indicate whether or not a country is democratic. Grand coalitions, or any executive

32 Keefer

(2002). (1993, pp. 326–338). 34 Gurr (2000, pp. 321–336). 35 Keefer (2002, p. 9). 33 Gurr

88  B. B. KELLY

power holders for that matter, do not necessarily hold any democratic mandate to rule.

Segmental Autonomy The second consociational component is Segmental Autonomy (SA), which Lijphart describes as “rule by the minority [or other group] over itself in the area of the minority’s exclusive concern.”36 It is prescribed to reassure groups that their cultures and other collective behaviors are respected and will not be impinged upon. This component thereby encourages extensive separation of groups and arguably diminishes intergroup contact to an extent that may facilitate prejudice, scapegoating, and secession. It cannot be assumed that it is a manifestation of stability, so it can be represented without modification in this analysis and without causing methodological problems. Lijphart’s thesis is that there is a strong, positive relationship between stability and the separation of groups at all but the most elite leadership levels.37 He explains that intergroup “[e]lite cooperation is the primary distinguishing feature of consociational democracy” and that “clear boundaries between the [nonelite members of] segments of a plural society have the advantage of limiting mutual contacts and consequently of limiting the chances of ever-present potential antagonisms to erupt into actual hostility.”38 He argues that, in “essentially heterogeneous cultures,… close contacts are likely to lead to strain and hostility” so “it may be desirable to keep transactions among antagonistic subcultures in a divided society… to a minimum.”39 In my view, Lijphart’s other three core consociational components are subsidiary to this more fundamental element requiring group autonomy and separation. Grand coalition is the necessary cooperation among elites, MV is one manifestation of groups’ autonomy, and PR is another means of maximizing groups’ power over their own affairs. These three governance-related components constitute a fair way of representing groups in joint decisions and

36 Lijphart

(1977, p. 41). (1969, pp. 216–221; 1977, pp. 1, 88–89). 38 Lijphart, (1977, pp. 1, 88). 39 Lijphart (1969, pp. 220–221). 37 Lijphart

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

89

allocating shares of joint resources. The performance of the SA variable in this analysis will give some indication of whether Lijphart’s entire notion of consociation promotes intergroup harmony. The accuracy of any statistical representation of Lijphart’s notion of SA can be maximized through analysis of his descriptions of this component. In a 1996 article, he provides a very clear description of it that seems to lend itself to quantifiability: Cultural autonomy for religious and linguistic groups has taken three main forms in power-sharing democracies:

(1) federal arrangements in which state and linguistic boundaries largely coincide, thus providing a high degree of linguistic autonomy, as in Switzerland, Belgium, and Czecho-Slovakia; (2) the right of religious and linguistic minorities to establish and administer their own autonomous schools, fully supported by public funds, as in Belgium and the Netherlands; and (3) separate ‘personal laws’—concerning marriage, divorce, custody, and adoption of children, and inheritance—for religious minorities, as in Lebanon and Cyprus.40

This excerpt suggests that segmental (“cultural”) autonomy should be represented through data indicating the presence of federalism, the extent to which federal boundaries coincide with group boundaries, the right of groups to establish their own publicly funded schools, and the existence of separate “personal laws” for groups. This list emphasizes the fact that SA does not require the existence of federalism, the option of separate schools, or separate personal laws. These are indicators of, rather than requirements for, groups’ autonomy. Lijphart’s consistent use of the term, “autonomy,” and his choice of the word, “right,” in reference to schools, highlights that SA is about groups’ power to do things separately, rather than their actual separateness. His early differentiation between voluntary separation and apartheid, with regard to the Netherlands, also aids in clarification of Lijphart’s beliefs regarding the role of this separation:

40 Lijphart

(1996, p. 260).

90  B. B. KELLY …segregation in Holland means self-imposed social isolation, whereas apartheid implies the imposition of segregation by a dominant group upon a less favored group. The Dutch case does show, however, that a ‘separate but equal’ doctrine can work. The religious-ideological blocs have fought for and have largely achieved equal status in Dutch society while maintaining their separate identities and their mutual isolation. In other words, social segregation does not necessarily lead to instability nor is it inherently unequal or unjust.41

Lijphart seems not to have provided a more detailed description of this distinction, but analysis of his body of work on consociation indicates that it is a valid interpretation of his overall point of view. Some scholars understand Lijphart as advocating separation, and/or imposed “segregation,” of groups. Separation can be evidence of forced segregation or the exercise of an option to be separate. It is obvious that Lijphart is advocating optional, rather than forced, segregation. The data for the SA variable is thus made to represent groups’ empowerment, rather than forced, segregation. Although groups’ separate schools are often identified as a manifestation of SA, no data indicating the option for such schools could be found to incorporate into this dataset. There is more data indicating the existence, as distinguished from the option, of separate education. However, including it in this analysis would be problematic for two reasons. Firstly, such data could not distinguish mandatory from optional school separateness, mandatory separateness does not conform to Lijphart’s definition of SA, and the mandatory and optional types are antithetical to one another with regard to equality and democracy. The second reason why the available school data should not be included in this analysis is that, although Lijphart implies that separate schools are evidence of SA, this model is designed to represent the existence, rather than the effects, of empowerment. Lijphart differentiates two categories of SA which are fundamental types of empowerment that can be translated into data representing, territorially based and nonterritorially based federalism and autonomy. This SA variable constructed here constitutes the first quantitative representation of the concept that incorporates both territorial and nonterritorial forms. Lijphart explains that consociation can manifest itself as 41 Lijphart

(1975, p. 186).

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

91

“a formal territorial federal system, a nonterritorial federation, or a system in which decision-making is delegated to private segmental organisations.”42 The variable will indicate whether or not any of these situations exist. Lijphart also believes that, to bring stability to plural societies, the substate divisions separating the units which comprise territorially based federations and other autonomy structures ideally should coincide with the cleavages separating potentially antagonistic groups. For this reason, countries with territorially delineated federalism and/or autonomy will only be recognized as containing SA if their substate governance units contain populations that are relatively culturally homogeneous (in terms of MAR groups). Lijphart’s advocacy of the unit homogeneity promoted by such systems is in fundamental disagreement with many other scholars’ recommendations for antagonistic, plural societies. For instance, Horowitz and Reilly argue that achievement of stability in such societies is facilitated when governance and group boundaries do not coincide.43 Unlike any previous quantitative treatment of consociation, this analysis recognizes that Lijphart’s SA is manifested through separation of the potentially antagonistic groups which make up a plural society. If federal and territorial autonomy systems are made with units that integrate these groups more than separating them, their presence is actually inimical to the theory and spirit of consociation. Nonterritorially based systems for separating groups and maintaining their autonomy will also be reflected in this variable. Groups’ control over their own cultural affairs can be manifested through a huge variety of practices, so it is a difficult phenomenon to quantify accurately. Countries will be identified as containing SA if their MARs’ cultural organizations’ activities are not substantially restricted, and if they have one of these three provisions: (1) territorial autonomy with relatively homogeneous governance units, (2) at least one MAR’s language as a legally official language, or (3) MARs are permitted to govern themselves to some extent through their own laws or justice systems. The only requirement of these four which must be in place for SA to be coded as existent is that involving cultural organizations. While it can be assumed that MARs do not want their cultural organizations to be restricted in their activities, they may not all want territorial (as opposed to nonterritorial) autonomy, official language status, or their own laws. 42 Lijphart

(1989, p. 141). (1985, pp. 619–620) and Reilly (2001, pp. 185–192).

43 Horowitz

92  B. B. KELLY

The existence of one of these latter three conditions seems to indicate the presence of empowerment in a society that is segmented. Although official language status does not itself provide any power to groups, if a MAR’s language is made official, it is implied that the MAR is as powerful as any other group inhabiting the country. While these conditions obviously do not comprise all of the seemingly unlimited, potential types of SA, they are unarguably manifestations of this phenomenon and they collectively cover a diversity of experiences and feelings of empowerment. In the absence of data covering all existing forms of group empowerment in segmented societies, some countries containing SA may accidentally not be coded as such for this analysis. The variable corresponding to this consociational component will indicate whether SA can be identified in a country, following a search for these manifestations of it. Optimally, the presence of these manifestations would also have been separately tested for each year during the 1975–1995 period, for every country. Since most of these factors are unlikely to change often in each society, and due to the unfeasibly large undertaking that would be required to calculate separate, annual data, only one piece of information was made for each country, corresponding to each manifestation and to SA as a whole. The data for SA was made through reference to information involving these manifestations, found through the Minorities at Risk dataset, Ann L. Griffith’s Handbook of Federal Countries: 2002, Keefer’s World Bank Database of Political Institutions, the 1988, 1989, and 2004 editions of the CIA World Factbook, the United Nations High Commissioner for Human Rights, the Islamic Family Law project, Herbert Kritzer’s encyclopedia of Legal Systems of the World, and the results of numerous internet searches. First, data was collected for each of the forms of SA to be analyzed: restrictions on cultural organizations, territorial autonomy with relatively homogeneous units, state language status, and groups’ control of their own legal affairs. The first of these regarding organizations corresponds perfectly to a variable in the Minorities at Risk dataset that is entitled “Restrictions on Cultural Organizations.” Unfortunately, this variable is not available for many years in the 1975–1995 period, but it is calculated for 1990. The 1990 mean values for each country are used as the data for this element of SA. In the Minorities at Risk dataset, there are four potential values for it: 0: “no restrictions,” 1: “activity informally restricted,”

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

93

2: “activity somewhat restricted,” and 3: “activity sharply restricted.”44 These data were used to decide whether each country had such restrictions or did not. At first, only countries with a “0” were coded as not having restrictions. However, this method eliminated many countries which were commonly believed to have SA, such as Bosnia. For this reason, countries with a “0” or “1” are recorded as unrestricted. The second phenomenon to be identified through data is territorial autonomy with relatively homogeneous units. This information was made through two steps: identification of all countries with federalism or other forms of territorial autonomy, and determination of which countries seemed to have substantial geographical concentration of MARs in the resulting substate units. The list of countries containing some form of territorial autonomy includes all those identified as federal in Griffith’s Handbook of Federal Countries: 2002, and those which were allocated positive values for two variables in Keefer’s database corresponding to the existence of autonomous regions and devolution of powers often delegated to regions in federal systems. Separate values were calculated for these two variables for each year of countries’ experiences between 1975 and 1995. If a country received positive values for either of them at any time during that period, it was coded for analysis here as if it has contained territorial autonomy. Like the information compiled for restrictions on cultural associations, one piece of data was made for each country indicating whether it had territorial autonomy at any time during the 1975–1995 period. Next, the data provided for six variables in the Minorities at Risk dataset was analyzed to ascertain the likelihood that the countries with territorial autonomy had subnational units which were relatively homogeneous with regard to MARs. Unfortunately, no data exists that specifically indicates the percentages of groups akin to MARs in subnational governing units. However, this data from the Minorities at Risk dataset does show to what extent MARs were geographically concentrated and what proportions of them were living inside and outside of their respective “regional base[s].” This is the best data available that provides an indication of the extent to which such groups are concentrated in subnational governance units, so it was examined for each country with territorial autonomy. A determination was made that, 44 Gurr

et al. (2002, pp. 88–89).

94  B. B. KELLY

during the 1975–1995 period, either territorial autonomy with relatively homogeneous units existed or it did not. The third manifestation of SA analyzed to make data for this core component of consociation is the provision for MARs’ distinctive languages to be granted official status. Such designation implies that MARs should have a right (but not any obligation) to have the separate treatment in educational and other social spheres that would be required for constant use of their language. For a country to receive a positive score for this factor, a MAR that constitutes less than half of the overall population must have a distinctive language which is formally recognized as having status equal to that of the more dominant, majority language. Such a MAR can be historically dominant, such as the whites in South Africa, if it comprises less than half of the country’s population. This language need not actually be called an “official” language. If a country uses a different term for this status, such as “national language,” that is acceptable. However, designation as a “national” or “official” language is not sufficient if another language has a more prestigious designation. For instance, Sri Lanka is not given a positive score for this factor because, although Tamil is a “national” language, Sinhala is given supreme status as both a “national” and the “official” language. If a language is given the supreme designation for the MAR rather than its entire country, that country also receives a positive score for this factor because that seems to imply that the group would be empowered to create its own publicly funded institutions. Since only one value is provided for SA for each country’s experience during the entire 1975–1995 period, if a MAR gained such status for its language sometime during this period, the entire period was given a positive value. For instance, Maori was declared an official language in New Zealand in August 1987 so that country is given a positive value for this factor covering the entire period. The entries listing each countries’ languages in the 1988, 1989, and 2004 editions of the CIA World Factbook were compared to the list of MARs, to determine values for this manifestation of SA. In the rare cases for which the CIA World Factbook did not contain information regarding official language status, Internet searches yielded the information required for making this data. The list of countries was simply divided into those containing such a provision and those which do not. Some drawbacks of the information compiled for this factor illustrate the importance of making the SA data through reference to four of its manifestations. Countries whose MARs share the languages of the

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

95

dominant group in their societies would automatically receive a negative score for this factor, even though they may possess very significant SA. In addition, sometimes official language status does not seem to correlate very strongly with positive treatment of MARs. For instance, during Saddam Hussein’s dictatorship, Kurdish remained an official language in the Kurdistan region of Iraq but this certainly was not evidence of the group’s fair or inclusive treatment. The fourth, and final, element of SA that was examined to make data for this consociational component is MARs’ ability to control their own culturally related legal affairs. The most common cases of this provision are for Muslim groups within largely non-Muslim countries, which are often recognized as preferring to operate according to distinctive customary laws. Like the SA factor involving official language status, this phenomenon is also not something that would be desired by a lot of MARs that are not interested in ruling themselves according to a distinctive set of rules. For instance, African Americans in the USA and “Catholics” in Northern Ireland desire equal treatment according to laws shared with their co-populations, rather than a separate set of laws for themselves. However, the presence of such legal autonomy is a definite indication of the presence of SA. Provisions allowing for measures of such legal autonomy for MARs were searched for, through reference to three sources: “core documents” for individual countries available through the website for the Office of the United Nations High Commissioner for Human Rights, the website for the Islamic Family Law Project at Emory University, and a set of encyclopedias edited by Herbert Kritzer, entitled Legal Systems of the World. These sources were identified as the most useful for collection of this information, following consultation with a number of international legal experts. Countries were coded as either possessing, or not possessing, separate legal treatment for one or more of their MARs, manifested through separate laws or courts. If a territorially autonomous substate possesses a separate legal system that is applied to everyone in that substate, it is not recognized as evidence of this type of legal autonomy. While this determination does not take into account the fact that some of these substates may be culturally homogeneous, it is not problematic because homogeneity of such units will produce positive SA scores for their countries, through reference to data made for the factor described above that indicates territorial autonomy. However, if a country possesses multiple MARs but not all of these are granted this legal autonomy, and if this right is allocated to them as

96  B. B. KELLY

a group rather than as a territorial unit, the country is positively coded for this factor. The group in question may be the only one who would desire, or benefit from, such a provision. The data to be used in the statistical analysis, corresponding to the presence of SA, was made through reference to the information collected for its four manifestations described above. For a country to be coded as possessing SA during the 1975–1995 period, it must have received a positive score indicating little repression of MARs’ cultural organizations, as well as a positive value indicating possession of either territorial autonomy with likely coincidence of substate and MAR boundaries, formal official status for at least one MAR’s distinctive language, or MARs’ ability to control their own culturally related legal affairs. When evidence of these elements was found in legal and constitutional sources, the extent to which these sources reflect the reality of life could usually not be verified. Fortunately, considering that a country theoretically could be inaccurately coded as possessing some manifestations of SA, the data indicating restrictions on cultural organizations was derived from observation of reality rather than legal documents. Any country found to have significant restrictions of this type was coded negatively for SA, regardless of its values for other manifestations of SA. A survey of resulting data illustrates that a positive value for SA is not necessarily indicative of democracy. For example, Iraq gets positive scores for four of the manifestations of SA and a positive overall score for the variable during this period, when Saddam Hussein brutally mistreated the group he officially gave the most autonomy to, the Kurds.

Proportionality Proportionality of electoral systems and other institutions comprises the third component of Lijphart’s consociational system. Evidence of proportional allocation of opportunities might involve employment and education positions, or public expenditure allocated to each group. These three phenomena are difficult to represent statistically due to lack of comparable data corresponding to a large set of countries. The availability of such opportunities in approximate proportion to groups’ sizes should eventually, naturally evolve when people receive equal treatment and groups peacefully coexist. For this reason, proportional allocation of opportunities is arguably an effect, rather than a cause, of stability. Therefore, representation of it in an independent variable

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

97

within this analysis would introduce methodological problems. The variable assigned to proportionality for this quantitative analysis must be restricted to proportionality of electoral results. Electoral rules which engender a close correspondence between votes and awarded political seats naturally encourage groups’ fair representation in government. PR is not necessarily associated with, or an effect of, stability, so its presence can be indicated through an independent variable. The variable regarding proportionality in this analysis will be represented by data indicating whether PR has operated in each state for each year during the 1975–1995 period. The data for this variable will be taken directly from Keefer’s World Bank Database of Political Institutions. Cases which Keefer’s dataset indicated as using PR were coded for this analysis as positive for PR. The consequent translation of Keefer’s missing values as negative for PR constitutes a deficiency of this book’s dataset.

Minority Veto Power The final component of consociation is each politically significant group’s power to veto legislation. Since extensively granted veto powers can, at least under some conditions, cause deadlock and consequent instability, this component cannot be considered a manifestation of stability so it can be directly represented in this statistical analysis. The independent variable corresponding to veto power will indicate whether all politically significant groups, only some, or none, possess this instrument. This variable’s identification of situations in which some, but not all, groups can veto legislation will significantly increase the ability of this analysis to explain stability in plural societies. If states allocating veto power to some or no groups were identified as belonging to the same category within this variable, the analysis would be biased toward the conclusion that consociation promotes stability. For instance, situations in which only some politically salient groups are given this right would obviously be conducive to tremendous resentment and instability, whereas allocation of veto power to no groups may promote stability through deterrence of stalemate. This is the first data indicating the presence of minority groups’ veto power (MV) in a large set of countries. With the exception of two sets of cases which are described on the website, this data was made for all of the country cases, including those four which Lijphart has identified most consistently as having experienced long-term success with

98  B. B. KELLY

consociation. These data values were calculated through reference to demographic statistics and the contents of all constitutions and constitutional amendments that could be found which governed each country’s entire populations. The unfeasibility of finding out whether groups veto policies in practice meant that the MV variable is confined to indicating whether, if groups were represented in legislative bodies in proportion to their sizes, they could exert a veto resulting in a changed policy decision. This criterion contains a highly significant and unfortunately necessary assumption. However, if the MV and PR variables are used in conjunction with each other in the statistical analysis, inability of groups to gain such representation at least will be partly controlled for. With the exception of “schedules” used to provide details, and attached to some constitutions, every word of each constitutional document used between 1975 and 1995 and found for use in this analysis was read in search of provisions that could result in MV. Individual pieces of data were made for each year of a country’s experience. Notes made for each document were compared to the percentages of each country’s population made up of individual MARs and all of its MARs collectively. If MARs and countries were listed in Gurr’s 1993 publication, the 1990 population figures used in it were provided for this purpose. If they were included in his 2000 publication but not the 1993 one, the 1998 figures provided in the 2000 publication were referred to. Austria is the only one of Lijphart’s core country examples of consociation for which MV data could not be made, and it was not included in Gurr’s study.45 In a couple cases for which I knew that information contained in constitutions did not reflect the reality in a specific country, I did a little more research through internet searches to confirm my impression and changed the data to reflect reality. For example, one of the MARs for South Africa, from as early as 1993, is “Europeans.” Although Europeans had lots of veto power, South Africa is coded “0” until the very end of the 1975–1995 period because this “minority at risk” was dominant in terms of political power. The data is adversely affected by the fact that information showing the exact proportions of legislative seats held by members of different cultural groups is unavailable. Another deficiency is that the data inadequately represents the domination of some majority groups by minority ones. 45 & More extensive details concerning the compilation of MV data are provided on the website.

SA

(continued)

Is there evidence that any non-advan- • Gurr’s MAR dataset • Data indicates presence of highly taged MARs are represented in the • Keefer’s World Database of inclusive, rather than all-inclusive governing executive? Political Institutions coalitions* (Actually indicates the presence of • Wide ranging internet searches • Data may not represent some highly inclusive coalitions, for which no highly inclusive executive coalitions) evidence could be found** Is there evidence that MAR groups • Gurr’s MAR dataset • Data may not represent some SA, had the power to govern some • Griffith’s Handbook of Federal for which no evidence could be aspects of their lives autonomously? Countries: 2002 found** • (SA identified when low levels of • Keefer’s World Database of (Including data representing separate repression of cultural organizations Political Institutions schooling, since data indicating the are combined with either (1) territo- • 1988, 1989, and 2004 editions option for such education is not rial autonomy for fairly homogeneof the CIA World Factbook available)** ous regions, (2) formal official status • “Core documents” for individof one MAR’s distinctive language, ual countries available on webor (3) a MAR’s ability to control its site of the Office of the United own cultural legal affairs) Nations High Commissioner for Human Rights • Website for the Islamic Family Law Project at Emory University • Kritzer’s Legal Systems of the World • Wide ranging internet searches

Limitations

Grand coalition

Sources of data

What do their variables indicate?

Consociational components

Table 4.1  The variables representing consociational components

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

99

**Limitations

that could not be eliminated, if the methodological validity of the quantitative analysis was to be preserved that could have been eliminated, if more information had been available More details about the construction of these variables are available in Appendix B

*Limitations

MV

Limitations

• Almost no missing values for Keefer’s PR variable could be investigated. Needed to assume that missing values indicate absence of PR** If it is assumed that MAR groups • Constitutions • Needed to rely on constitutions for can gain representation in legislative • Gurr’s MAR dataset evidence of potential MV, so data bodies in proportion to their sizes, • 2004 edition of the CIA World may not always reflect reality** does their country’s constitution Factbook • Countries whose populations conindicate that they could possibly exert tain large proportions of MARs are a veto resulting in a changed policy more likely to be coded as if they decision? have MV**

If there is electoral competitiveness, is • Keefer’s World Database of a PR system used? Political Institutions

Proportionality

Sources of data

What do their variables indicate?

Consociational components

Table 4.1  (continued)

100  B. B. KELLY

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

101

“Minorities at Risk” that are actually numerical majorities are very rare within Gurr’s dataset, but they do exist in a few countries. The method for constructing this data treats all MARs the same so, if one of a country’s MARs had a numerical majority of the population and was judged to have potential veto power, the country was given a positive “1” score. However, examination of Gurr’s cases indicates that the countries with a majority MAR either had potential veto power for other MARs as well, or the majority MAR was labeled a MAR because it was dominated by a smaller group. The first category resulted in no introduction of inaccuracy. For countries in the second category, constitutional provisions indicating potential veto power for the majority MAR did constitute promises of empowerment for groups that traditionally were politically dominated. However, this second category of countries, consisting of South Africa, Burundi, and Iraq, often did not operate in practice in a manner reflecting this power portrayed in their constitutional documents. That is a problem related to the drawbacks of using constitutions to make this data, rather than treating all MARs the same way. Yet another regrettable aspect of this data is that, since countries are given a positive “0.5” score if all of their MARs can apparently exert a veto if they act together, countries whose populations contain large proportions of MARs are actually more likely to be coded as if they have more MV. Introduction of an additional independent variable corresponding to size of population comprised of all MARs helps to control for this effect, but such a variable would have a systemic, directly proportional, and unfalsifiable relationship with the dependent variable, stability. More people at risk in a society obviously make it more likely to be unstable.46 Values of “0” were not allocated to years for which minority veto data was unavailable because that would introduce a problematic bias, since countries with constitutions that were difficult to find or unformalized constitutions would be penalized. Although it has some drawbacks, this variable’s data is the first designed to represent MV. Consociational institutions and practices will be represented in this analysis through four variables corresponding to Lijphart’s core components of consociation: the presence of inclusive coalitions, SA, PR, and MV. Particularly, since many states have adopted elements of 46 Since population size is also a factor Lijphart believes to be favorable to successful consociation, it is analyzed through an independent variable but found to have no statistically significant effect in regressions on stability.

102  B. B. KELLY

consociation, rather than the complete system, separate representation of the elements through these variables is optimal for this analysis. Individual representation of these elements will enable identification of their independent effects on stability in plural societies. Table 4.1 facilitates comparison of the consociational components, the variables representing them, and the limitations of these data.

Part 3: Control Variables Accurate identification of any causal relationship requires confirmation that apparent effects are in fact resulting from the variables which are assumed to be decisive. Although components of Lijphart’s consociational system are designed to engender stability in plural societies, other factors may also be performing this function. If Lijphart is correct in saying that certain conditions facilitate the introduction and successful function of consociation, variables representing them should influence the system’s performance. It is not possible to include all of these conditions for methodological reasons, but six can be examined, at least to some degree. These are: external threats, segmental isolation and federalism, moderate multiparty system, geographical concentration of segments, small population size, and socioeconomic equality. The relationships between these conditions and other elements of consociational theory were considered prior to identification of the best data available for their representation. Although Lijphart consistently identifies external threats as a favorable condition for consociation, some instances in which this system arguably created instability have instead been attributed to external threats and interventions. The relationship between at least one category of external threats and stability in plural societies will be analyzed with a controlling variable indicating the extent to which ethnopolitical rebellion spreads in each country’s surrounding region. Another of Lijphart’s favorable conditions, segmental isolation and federalism, will be examined through one of the primary independent variables corresponding to SA. The effects of moderate multiparty systems, small population size, and socioeconomic equality will be tested with data borrowed from other databases. An additional variable representing these favorable conditions records geographical concentration of segments. Its data indicates the presence of federal or

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

103

autonomous regions in which MARs are concentrated. This variable is also important because it enables examination of Lijphart’s claim that stability is best achieved in plural societies when the boundaries separating countries’ regions correspond as closely as possible with the boundaries separating potentially antagonistic groups.47 Federal and other forms of autonomy for relatively homogeneous units were also considered during the process of constructing data to represent the SA variable. However, the other conditions that influence a country’s value for SA all reflected the ability of groups to have autonomy in various spheres of life, rather than the actual presence of that autonomy. Since the actual presence of autonomy might be expected to play a different role than this ability, the geographical concentration of groups in autonomous and semi-autonomous regions will also be analyzed separately through a controlling, independent variable. The only control variable included which is not related to Lijphart’s favorable conditions will represent democracy. Lijphart’s work on consociation aspires to be empirical, normative, and predictive. He presents it as capable of explaining stability in plural societies, constituting a set of policy recommendations, and predicting where it will be adopted. Lijphart claims that certain “favorable conditions” are conducive to the introduction and successful operation of consociation. He does emphasize that these conditions are “helpful but neither indispensable nor sufficient… to account for the success of consociational democracy.”48 The different lists of conditions presented by Lijphart in various publications over the past three decades have inspired considerable analysis by other academics. Matthijs Bogaards provides a comprehensive list of these conditions, indicating the years of their introduction by Lijphart. This list is provided here in Box 4.1. Lijphart claims that discrepancies between these lists result from his continual refinement of consociational theory.49

47 Lijphart

(1977, p. 42). (1977, p. 54). 49 & Description of Lijphart’s classifications of conditions with reference to South Africa is on the website. 48 Lijphart

104  B. B. KELLY

Box 4.1  The favorable conditions identified by Lijphart*

Introduced in 1968 • Distinct lines of cleavage between subcultures • A multiple balance of power among the subcultures • External threats • A relatively low load on the system • Moderate nationalism • Popular attitudes favorable to government by grand coalition Introduced in 1969 • Widespread approval of the principle of government by elite cartel • The length of time a consociational democracy has been in operation • Internal political cohesion of the subcultures • Adequate articulation of the subcultural interests Introduced in 1977 • Segmental isolation and federalism • Small country size • Overarching loyalties • Moderate multiparty system • Representative party system • Crosscutting cleavages (in some instances) • Tradition of elite accommodation Introduction in 1985 • Geographical concentration of segments • No majority segment plus segments of equal size • Small population size • Small number of segments • Socioeconomic equality * This compilation was contributed by Matthijs Bogaards (1998, p. 478).

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

105

Six favorable conditions can be examined in this analysis because data is available with which to represent them and their inclusion will not cause any significant methodological problems. These are external threats, segmental isolation and federalism, moderate multiparty systems, geographical concentration of population segments, population size, and socioeconomic equality. The other conditions in Lijphart’s most recently revised list cannot be included because no data is available for them, and, even if it was, their representation in the analysis would weaken its explanatory integrity. If assigned to variables, three of these conditions would create endogeneity problems, which occur when “the values our explanatory [independent] variables take on are… a consequence, rather than a cause, of our dependent variable.”50 For example, if one wanted to test whether eating more made people gain more weight, one would not want to use gaining weight as an independent variable and eating as a dependent one because an analysis of them would indicate that gaining weight makes one eat more. Lijphart argues that “overarching loyalties” and “traditions of elite accommodation” promote consociation’s ability to bring stability to plural societies. However, these phenomena are possibly causes, but also unarguably manifestations (and effects) of stability. He argues that the absence of a majority group segment and the rough equality of such segments are conducive to consociation’s success. Observations for these conditions would not contain enough variation to avoid systematic biases. Only a tiny proportion of the countries being studied contain either a majority MAR or roughly equal size MARs. The very small sample of countries with more than a few MARs means that inclusion of a variable corresponding to this factor would also be likely to introduce systematic bias. Lijphart does provide numerous, compelling examples of how threats with origins external to countries promote both patriotism and conflict. Lijphart identifies international threats as a condition favorable to consociation but also argues that international intervention heavily contributed to the breakdown of consociational systems in Lebanon and Cyprus.51 For instance, Lijphart claims that “it is primarily to Lebanon’s increasingly unfavorable international environment- combined with the

50 King

et al. (1994, p. 185). More information regarding Lijphart’s discussions of Cyprus and Lebanon is provided on the website. 51 &

106  B. B. KELLY

internal flaw of consociational rigidity- that the 1975 breakdown of the ­democratic regime must be attributed.”52 External threats must be represented in the analysis because he believes that they can substantially influence both the adoption and the success of consociational systems and can sometimes transform a peaceful consociational society into a disaster area. Lijphart has consistently, convincingly argued that at least some external threats are felt by all inhabitants of a country and thus motivate them to feel more loyalty to their country and temporarily forget their inter-segmental disagreements to some extent. For example, if a country was made up of only As and Bs, an attack by Cs would constitute this sort of external threat but a secessionist movement by Bs in a bordering country would not. Unfortunately, Lijphart has not specifically described which categories of external threats result in greater intergroup feelings of comradeship and which actually precipitate instability. These phenomena are obviously very situation-specific, and it would seem almost impossible to create data to represent them, especially for a large sample of countries and time periods. Identification of the optimal means for controlling for international intervention in this study can be facilitated through reference to discussions of its various manifestations.53 Most intervention types identified in academic contributions seem to be positive ones, initiated by other countries and designed to increase stability. However, most examinations of the causes of group conflict, such as the well-respected and fairly comprehensive ones contributed by Powell and Douglas Hibbs, do not explore international effects.54 Externally applied belligerence seems to have affected the performance of consociation, and data representing it to some degree is available in the Minorities at Risk dataset.55 Two types of data related to destabilizing international intervention are available in it, corresponding to the “spread of ethnopolitical protest and rebellion through [the] region” surrounding each country.56 The two variables are indexes of “international contagion” corresponding to protest and rebellion. This data is the best available for representation of destabilizing international intervention in a large number of countries, 52 Lijphart

(1977, pp. 154, 161; 1985, p. 92). Vine (1997, p. 75). 54 Powell (1982) and Hibbs (1973). 55 Khosla (1999). 56 Gurr et al. (2002, p. 217). 53 Le

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

107

but it should be recognized that it does include both the intentional and unintentional influence implied in the term, “contagion.” Another drawback of the data is that it does not reflect the rare cases when an outside entity that does not border a country exerts such a destabilizing influence. The variable for rebellion will be used here because it is the strongest, and so likely the most influential, form of this contagion represented in the Minorities at Risk dataset. However, it must be kept in mind that such rebellion is, by no means, the only form of external threat which can influence a country’s level of stability. Political, nonviolent intervention cannot be analyzed here because of the unavailability of relevant data. Separate pieces of rebellion-related data are provided for each decade of a country’s experience in the Minorities at Risk dataset, so some of the years of each country’s experience covered here are allocated different values. The higher the value for this variable, the more contagion and intervention was being exerted. Inclusion of this variable is also methodologically unproblematic. Intervention intended to cause instability and rebellion in bordering countries can occasionally motivate a greater sense of patriotism in response to such external threats. This possibility ensures that this variable’s effects on stability can be expected to contain variation adequate to produce valid statistical results. Another control variable to be included in this analysis will facilitate assessment of Lijphart’s belief that “moderate multiparty system[s]” are more conducive to the success of consociation.57 Unfortunately, data indicating the polarization represented by all of the parties operating in each country is unavailable. Philip Keefer’s World Bank Database of Political Institutions does provide information for all of the countries and years covered here, indicating the maximum polarization between the executive party and the four principal parties of the legislature. Data from this dataset will be used to represent the variable corresponding to “moderate multiparty systems” here. This variable is methodologically unproblematic because a lack of polarization may facilitate the progress of the system, but it might also be responsible for stability that initially appears to be caused by consociation or some other prescribed remedy for deeply divided societies.58

57 Bogaards

(1998, p. 478). More discussion of the compilation of data representing moderate multiparty systems is on the website. 58 &

108  B. B. KELLY

Inclusion of a variable controlling for the influence of geographical concentration of population segments enables assessment of another of Lijphart’s favorable conditions. To some extent, the coincidence of governance and group boundaries is reflected in data made for the SA variable. However, unlike the other factors incorporated in that data, this boundary coincidence is a form of autonomy, rather than simply a provision allowing groups to get autonomy. Since the phenomenon reflected in this boundary coincidence variable is also reflected in the SA variable, the two variables cannot be used together if methodological soundness is to be preserved. If they were analyzed in conjunction with one another, this would cause multicollinearity, which is a problem that occurs when “we can perfectly predict one explanatory [independent] variable from one or more of the remaining explanatory variables.”59 The relationship between the variables, corresponding to SA and geographical concentration of segments, is close enough that, if information was known about cultural restrictions, values for the boundary coincidence variable could be used to predict a substantial number of values for SA. If multicollinearity exists between independent variables, their causal effects simply cannot be ­distinguished accurately from one another. For this reason, these two variables will not be used in the same analysis. By alternately trying both of them, some valuable insights will be reached concerning the possibly different roles of autonomy. Inclusion of the boundary coincidence variable on its own creates no methodological problems involving variable effects on the dependent variable because two, highly respected theories contend that it produces completely opposite effects on countries’ stability. Like the boundary coincidence information that was considered during the process of constructing data for SA, the data for this variable indicates whether countries with federal systems and/or autonomous regions have high or low concentrations of MAR groups in those subcountry territories. This data was made through comparison of a list of these countries and data from the Minorities at Risk dataset, indicating the extent to which MARs are geographically concentrated. The list of countries with both autonomy provisions and this geographical concentration was also examined according to Horowitz’s and Reilly’s observations about which countries contained this coincidence of autonomy and group boundaries. When the data for this variable differed from their descriptions, it was modified to agree with them. For this reason, 59 King

et al. (1994, p. 122).

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

109

the data for three countries was changed: Fiji, Malaysia, and Sri Lanka.60 The analysis required for evaluation of these countries with regard to this phenomenon emphasizes that every piece of data for this variable necessitated a somewhat impressionistic judgment concerning whether or not governance systems separated potentially antagonistic groups and restricted their members to voting only within their groups. The data corresponding to “small population size” in this analysis is taken directly from the Minorities at Risk dataset. It is the estimated population size for each country, and it is provided for two years, 1990 and 1995. This data is particularly appropriate because it was used by Minorities at Risk staff to calculate the percentages constituted by each MAR in individual countries, and these figures were used to identify cases here. Since 1990 is within the period to be covered, each case representing one year of a country’s experience is allocated the 1990 population estimate for that country. Only one figure is thus given for each country but, since this analysis contains a large number of countries, that information will be adequate to identify statistically significant relationships between population size and stability. For countries that did not exist as of 1990, the 1995 data was used for the analysis. The influence of this variable is unpredictable enough to ensure that its effects on the dependent variable, stability, will contain some variation. Although Lijphart believes that smaller populations are more conducive to stability, the greater diversity likely in larger populations could plausibly result in less corruption, competition, and other forms of antagonism among groups and individuals. The final favorable condition to be examined, “socioeconomic equality,” also will be represented through inclusion of a variable from the Minorities at Risk dataset. This variable is an “economic differentials index” that reflects the situation of MARs regarding income, land and property, higher education, and presence in commerce, professions, and official positions, compared to the rest of their countries’ inhabitants. The higher a value given for this index, the greater the extent to which a country’s MARs are disadvantaged. The codebook explains that only one piece of data for this variable was made for each country that corresponds to the period immediately preceding the year that it was included in the database. The Minorities at Risk data collection

60 The manner in which this data was developed for Fiji, Malaysia, and Sri Lanka is explained and justified on the website.

110  B. B. KELLY

undertaking was initiated in 1986, so the period analyzed to make the data for most countries was the early 1980s. In contrast to the description of this variable in the Minorities at Risk codebook, the data for this variable provided for each country in the MAR dataset is not the same for all years within the 1975–1995 period covered here. Very few countries were represented in this way. For those, the value which most often occurred, which was usually, roughly an average of the figures provided, was used for all of the years of its experience to be covered. Like the data corresponding to population size, the data used for this variable was also the same for every year of a country’s experience but the large number of countries included precludes methodological problems. For both of these variables, international differences are much more significant than any intra-country differences that may have occurred during the 1975–1995 period. The influence of socioeconomic equality on stability will also vary enough to ensure that this variable will not jeopardize the methodological integrity of this analysis. While socioeconomic equality is often believed to result in greater stability, occasionally countries containing tremendous disparities of wealth, such as Singapore, have maintained impressive levels of stability. The influence of democracy should be controlled for in this analysis for two reasons. Firstly, it is widely believed to exert a strong independent effect on stability which could be confused with the role of consociational components, if a variable corresponding to it was not included. Secondly, although Lijphart implies that only democratic arrangements can be considered consociational, his description of the grand coalition component places great stress on the superiority of governments including all potentially antagonistic groups and elite predominance, which arguably could constitute an elite cartel government. His notion of grand coalition has been interpreted by some critics as implying a system with no effective opposition and very little turnover of government members, in which the general population is provided with very little information concerning the government and its decisions. Lijphart has identified several countries as possessing grand coalitions which do not fit this description. For instance, he claims that the Indian Congress Party was adequately representative of its country to justify its identification as a grand coalition, even when it was alone in government. Comparison of his theoretical treatments of consociation and descriptions of real governments that he believes constitute grand coalitions implies that

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

111

consociational governments could be either democratic or relatively undemocratic. Inclusion of a controlling variable corresponding to democracy will enable discrimination between the effects on stability of more, and less, democratic consociational systems. A variable contained in the Polity IV dataset, which was also included in the Minorities at Risk dataset, can be used to represent democracy in this variable. The higher its value for this variable, the more democratic a country is, and negative values indicate various levels of autocracy. Separate values are provided for each year of a country’s experience, and they represent democracy in the first quantitative analysis to examine its influence on stability and its operation when combined with consociation, in a large number of cases throughout the world. This variable is methodologically unproblematic because, while democracy is commonly associated with stability, it also allows exertion of more dissent which could produce substantial disruption and regime change. Therefore, the effects of democracy on stability can be expected to contain substantial variation.

Conclusion This chapter presents, justifies, and discusses the variables and data necessary for the following statistical investigation of the proposition that consociation promoted stability in plural societies, between 1975 and 1995. The roles of consociational components will be verified and further explored through their analysis in conjunction with six additional variables which are also specified in this chapter. These control for phenomena that analysis of consociation has indicated are likely to affect the system’s impact on stability. Lijphart is, by far, the foremost authority on such phenomena, which he calls “favorable conditions.” Beginning in the 1960s, he has derived them through extensive examination of a wide array of case studies and theoretical considerations. Unfortunately, not all of these factors could be represented and/or analyzed quantitatively for this study, while using existing data sources and maintaining scientific validity of statistical results. However, five of them could be quantitatively assessed here, as well as democracy, which is widely assumed to also exert a significant influence on consociation’s performance. The case studies comprising Part II of this book were chosen in part to maximize our understanding of those of Lijphart’s favorable conditions which could not be tested here quantitatively.

112  B. B. KELLY

References Bogaards, Matthijs. “The Favourable Factors for Consociational Democracy: A Review.” European Journal of Political Research. 33 (1998) 475–496. Bonneuil, Noël and Nadia Auriat. “Fifty Years of Ethnic Conflict and Cohesion: 1945–94.” Journal of Peace Research. 37:5 (2000) 563–581. Cohen, Frank S. “Proportional Versus Majoritarian Ethnic Conflict Management in Democracies.” Comparative Political Studies. 30:5 (October, 1997) 607–630. Gurr, Ted Robert. Minorities at Risk: A Global View of Ethnopolitical Conflicts. Washington, DC: United States Institute of Peace Press, 1993. Gurr, Ted Robert. Peoples Versus States: Minorities at Risk in the New Century. Washington, DC: United States Institute of Peace Press, 2000. Gurr, Ted Robert, Monty G. Marshall, and Christian Davenport. Minorities at Risk: Dataset Users Manual.1002. Center for International Development and Conflict Management, University of Maryland, 2002. Hibbs, Douglas A. Mass Political Violence: A Cross-National Causal Analysis. New York: Wiley, 1973. Horowitz, Donald L. Ethnic Groups in Conflict. Berkeley: University of California, 1985. Keefer, Philip. DPI2000: Database of Political Institutions: Changes and Variable Definitions. The World Bank, March 2002. Khosla, Deepa. “Third World States as Intervenors in Ethnic Conflicts: Implications for Regional and International Security.” Third World Quarterly. 20:6 (1999) 1143–1156. King, Gary, Robert O. Keohane, and Sidney Verba. Designing Social Inquiry: Scientific Inference in Qualitative Research. Princeton: Princeton University Press, 1994. Krain, Matthew. “Contemporary Democracies Revisited: Democracy, Political Violence, and Event Count Models.” Comparative Political Studies. 31:2 (April, 1998) 139–164. Le Vine, Victor T. “Conceptualizing ‘Ethnicity’ and ‘Ethnic Conflict’: A Controversy Revisited.” Studies in Comparative International Development. 32:2 (Summer, 1997) 45–75. Lijphart, Arend. “Typologies of Democratic Systems.” Comparative Political Studies. 1:1 (April, 1968) 3–44. Lijphart, Arend. “Consociational Democracy.” World Politics. 21:2 (January, 1969) 207–225. Lijphart, Arend. The Politics of Accommodation: Pluralism and Democracy in the Netherlands. Second Edition. Berkeley: University of California, 1975. Lijphart, Arend. Democracy in Plural Societies: A Comparative Exploration. New Haven: Yale University Press, 1977.

4  EMPIRICALLY DEPICTING CONSOCIATION’S ROLE 

113

Lijphart, Arend. Power-Sharing in South Africa. Berkeley: Institute of International Affairs, 1985. Lijphart, Arend. “From the Politics of Accommodation to Adversarial Politics in the Netherlands: A Reassessment.” Politics in the Netherlands: How Much Change. Eds. Hans Daalder and Galen A. Irwin. London: Frank Cass, 1989. Lijphart, Arend. “The Puzzle of Indian Democracy: A Consociational Interpretation.” American Political Science Review. 90:2 (June, 1996) 258–268. Lijphart, Arend. Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries. New Haven, CT: Yale University Press, 1999. Lustick, Ian. “Stability in Deeply Divided Societies: Consociation Versus Control.” World Politics. 31:3 (1979) 325–344. Muller, Edward N. “Income Inequality, Regime Repressiveness, and Political Violence.” American Sociological Review. 50 (1985) 47–61. Muller, Edward N. and Mitchell A. Seligson. “Inequality and Insurgency.” American Political Science Review. 81 (1987) 425–51. O’Leary, Brendan. “Debating Consociational Politics: Normative and Explanatory Arguments.” From Power Sharing to Democracy: Post-conflict Institutions in Ethnically Divided Societies. Ed. Sid Noel. Montreal and Kingston: McGill-Queen’s University, 2005. Powell, G. Bingham. Contemporary Democracies: Participation, Stability, and Violence. Cambridge, MA: Harvard University Press, 1982. Reilly, Ben. Democracy in Divided Societies: Electoral Engineering for Conflict Management. Cambridge: Cambridge University Press, 2001.

CHAPTER 5

Quantitative Results

While there are a number of limitations to the data being analyzed here, they do represent this theory with an unprecedented level of accuracy. This chapter is devoted to describing this project’s quantitatively derived observations. Highly inclusive coalitions were found to deter violent and nonviolent instability. However, in at least one regression analysis, somewhat inclusive coalitions, potential minority veto (MV) power, segmental autonomy (SA), and PR electoral systems were all found to exert statistically significant destabilizing influences. Scatterplots with lines fitted to facilitate their interpretation corroborate almost all of the regression-derived findings. They also confirm that the effect of the SA variable on instability is slight and explained better by other variables. One scatterplot shows that a large concentration of cases found to provide constitutionally protected MV power is actually relatively undemocratic. Factor analysis provides evidence of two factors, both of which are found to exert statistically significant effects on instability, but only one of which explains much variance in stability. The first appears to possibly be discriminating between the larger body of democracies and those cases exhibiting more elements of consociation and consociational theory. The tiny extent to which the variable for this factor explains instability may provide the first quantitative evidence that Lijphart was correct in 1968 when he asserted

© The Author(s) 2019 B. B. Kelly, Power-Sharing and Consociational Theory, https://doi.org/10.1007/978-3-030-14191-2_5

115

116  B. B. KELLY

that consociational democracy is no less conducive to stability than other democracies. The variable representing the second factor identified here explains 17% of the variance in stability and appears to elucidate a phenomenon by which somewhat inclusive executives facilitate stability in countries with large populations that face external threats. Consideration of this body of quantitative results also emphasizes the importance of unavailable data.

Describing the Data The relationships between data representing consociation and stability are the most interesting element of this study because of their implications concerning the nature of optimal governance systems for divided societies. The extent to which countries have ever contained all four consociational elements has never been quantified as objectively and extensively as in this dataset. A survey of the dataset suggests that some countries have been less consociational than was previously assumed and draws attention to others, which have never been examined in reference to consociation but have contained substantial levels of consociational components. The data illustrates the system’s influence throughout the large percentage of the world’s population covered by this study, during the period of 1975–1995. Grand coalition is often presented as the most critical element of consociation, but definitions of it provided by Lijphart and others vary greatly regarding their required extent of groups’ and parties’ inclusion. So few examples exist of completely inclusive coalitions that a variable indicating their existence would contain too little variation and would not be useful in any statistical analysis. The only means of creating data for this phenomenon, which contains enough variation to maintain methodological integrity, is the definition of this type of coalition as a government executive that includes representatives of one or all of each country’s Minorities at Risk (MAR) groups. This data is useful because it indicates the extent to which a government resembles a “grand coalition.” However, since the criterion for a top score of highest inclusiveness does not indicate perfectly inclusive coalitions, it cannot be assumed to correspond exactly to this consociational component. Although high values indicating more inclusive coalitions are allocated for executives including all MAR groups listed in Gurr’s study as existing in each country, this list

5  QUANTITATIVE RESULTS 

117

cannot be relied upon to include all potentially antagonistic population segments. In addition, the inclusion of one representative politician from each such segment does not guarantee his/her equal power within the executive, which is another facet of Lijphart’s stricter descriptions of grand coalition. The extent to which countries have executives similar to grand coalitions is represented through two variables. The first of these is labeled, “somewhat inclusive grand coalitions,” and it indicates whether one or more MAR groups, which do not comprise all of their country’s MAR groups, and are not identified as “advantaged” in Gurr’s dataset, are represented in the governing executive. Both of these variables are dummy ones, indicating either the presence or absence of the coalition type. The second variable shows the presence of “highly inclusive grand coalitions” and its data indicate whether all of a country’s MAR groups, which may be only one, and which are not identified as “advantaged” in Gurr’s dataset, are represented in the governing executive. A figure provided on the website shows the distribution of somewhat and highly inclusive executives in the continents where the cases are located.1 Of all of the years of countries’ experiences examined here, 87% had no inclusion of MAR groups in executives. 6% contained somewhat inclusive coalitions and 7% had highly inclusive executives.2,3 Countries which experienced executive representation for all MAR groups between 1975 and 1995 included: Afghanistan (1989–1995) Belgium (1978–1995) Bosnia and Herzegovina (1995) Czechoslovakia (1989–1992) Ethiopia (1992–1995) Lebanon (1991–1995)

Madagascar (1975–1995) Malaysia (1975–1995) Netherlands (1975–1977) Nigeria (1980–1983) South Africa (1995) Switzerland (1975–1995)

1 & Four figures are provided on the website, indicating the percentages of each continent’s cases found to have evidence of the four consociational components. 2 All of the proportions described in this section, “Surveying the Body of Data,” are rounded to the nearest whole number. 3 & More details concerning the construction of the data representing such executives are on the website.

118  B. B. KELLY

The extent to which this data accurately represents the existence of such coalitions is emphasized by the fact that a number of the countries identified as consociational by numerous scholars are included in this list. Madagascar’s experience with this consociational component, as well as for SA and MV, illustrates that analysis of a long period of its experience would likely produce some interesting insights. Identification of Malaysia and Nigeria as possessing executives similar to grand coalitions illustrates the fact that the data for this variable does not represent the component of consociation in an ideal fashion. The distinction between highly and completely inclusive coalitions is very significant because, while the former indicates a willingness to allocate power to all potentially antagonistic groups, the latter renders government opposition at least powerless and potentially, just completely nonexistent. The data used to represent grand coalition here shows the existence of highly inclusive, rather than totally inclusive, executives and even they have not ever occurred in most countries. The second consociational component, SA, was represented through a variable whose data was compiled through reference to information regarding restrictions on cultural organizations, provisions for official language status and legal autonomy for MARs, and the existence of homogeneous, geographically defined governance units.4 This data indicated that 60% of the years of countries’ experiences included SA. Of the countries listed above as possessing highly inclusive coalitions, only Czechoslovakia and the Netherlands were identified as not employing SA. This finding regarding the Netherlands may reflect the dramatically decreased pervasiveness and “sharpness” of “the social cleavages between the blocs” there, which Lijphart observed in one of his first publications regarding consociation.5 Three countries which have been identified by Lijphart as using consociation at some stage but did not use these coalitions, Austria, Canada, and India, have experienced SA as it is defined for this study. Due to the complications involved in weighing certain types of autonomy as more significant, only positive and negative values were calculated for this variable.

4 The only exception to this methodology involving SA concerned Fiji. It is described on the website. 5 Lijphart (1975, pp. 197, 210).

5  QUANTITATIVE RESULTS 

119

However, the manifestations of autonomy that were examined seem to represent the concept of, and spirit behind, Lijphart’s consociational component of SA.6 The variable for MV power represents whether countries’ constitutions indicate that MAR groups could possibly exert a veto resulting in a changed policy decision, if it is assumed that they can gain representation in legislative bodies in proportion to their relative population sizes. A separate piece of data was constructed for each year of countries’ experiences, indicating that either one MAR could theoretically exert such vetoes alone, all of a country’s MAR groups could do it collectively, or none could do it independently or together. Data for this variable could not be created for certain countries and time periods. During the 72% of countries’ years of experience for which this data was made, 43% had the theoretical possibility that a MAR group could exert this veto power itself. The countries that used highly inclusive executive coalitions (HGC) at some stage between 1975 and 1995 and for which MV data could be compiled, also exhibited this potential veto power for an independent MAR for at least one year of this period. According to Philip Keefer’s World Bank Database of Political Institutions, proportional representation electoral systems were used during 29% of the years of countries’ experiences examined here. Some countries which have experienced other consociational components, as they are represented in these variables, but not proportional representation electoral systems, include Canada, Afghanistan, Bosnia and Herzegovina, Ethiopia, Lebanon, and Malaysia. Although Lijphart identifies proportional allocation of some career opportunities as an element of this consociational component, its role in the promotion of stability cannot be tested here. Such allocation of posts is arguably an effect, rather than a cause, of stability so its inclusion in this statistical analysis would have caused methodological problems. The extent to which a year of a country’s experience is consociational can be approximated by adding the values allocated to it for the variables corresponding to the components of the system. For instance, if a country exhibits the maximum amounts of grand coalition, SA, MV, and proportional representation shown through the variables used here, it could be allocated 5 points. These points include 2 that a case may achieve 6 & More details concerning the construction of the data representing SA are on the website.

120  B. B. KELLY

through inclusiveness of MARs in executives, 1 for SA, 1 for potential MV power for independent MAR groups, and 1 for the use of a proportional representation electoral system. If a case is found to have a “highly” inclusive executive coalition, it is assigned two points and, if it is found to have a “somewhat” inclusive coalition that includes some but not all non-advantaged MARs, it is given one point. Box 5.1 lists the consociational situations whose existence is confirmed by this dataset. Only 3% of countries’ years of experience can be allocated all five points. These situations include: Belgium (1978–1995), Madagascar (1993–1995), Nigeria (1980–1983), South Africa (1995), and Switzerland (1975–1995). Therefore, this analysis indicates that, even if consociation is defined in the very broad terms represented by these variables, the combination of all four components judged to collectively constitute this system has operated extremely infrequently. Of course, it should be kept in mind that a combination of these elements which does not include all four is very often described by scholars as constituting some level of consociation. Even the countries’ years of experience warranting an overall consociational score of four constitutes only 4%. These additional cases are: Afghanistan (1989– 1991), Belgium (1975–1977), Ethiopia (1992–1995), India (1975), Lebanon (1991–1995), Madagascar (1976–1992), Malaysia (1975– 1995), and the Netherlands (1975–1977). If cases achieving 4 or 5 points are considered to be at least somewhat consociational, this category makes up 7% of all of these cases. It should be remembered that this is not 7% of the world’s experience between 1975 and 1995, but only of those eightyeight countries which contained large enough populations of MAR groups to be considered plural and did not have any attributes that rendered them methodologically inappropriate for this analysis. The extent to which the countries comprising this 7% of cases have multiple manifestations of consociation suggests the accuracy of Lijphart’s belief that the components of this system often exist and complement one another. Unfortunately, the data from MAR that was used to make data representing stability for this analysis is not available for Austria, Belgium, the Netherlands, and Czechoslovakia. The relationship between stability and their experiences with consociational components cannot be statistically tested here. Since they are generally considered to constitute some of the most clear examples of consociation, Austria, Belgium, and the Netherlands will be qualitatively analyzed in depth, in Chapters 6 and 7.

5  QUANTITATIVE RESULTS 

121

Box 5.1 Consociational cases in plural societies between 1975 and 1995 whose existence is confirmed by this dataset

Identification of the presence of consociation has been performed through allocation of a maximum of five points for each case. Each case’s point allocation corresponds to the sum of its values for the variables representing consociation’s four components. Cases deserving all five points comprised only 3% of the countries’ years of experience represented in this dataset and they were confined to two continents, Africa and Europe. Africa Madagascar (1993–1995) Nigeria (1980–1983) South Africa (1995)

Europe Belgium (1978–1995) Switzerland (1975–1995)

Some additional cases deserved four points and thus exhibit several elements of consociation. They are listed below and include countries from all of the continents except the Americas and Oceania. The situations allocated four and five points collectively constitute 7% of the years of experience examined here.

Africa Ethiopia (1992–1995) Madagascar (1976–1992) Malaysia (1975–1995)

Europe Belgium (1975–1977) Netherlands (1975–1977)

Asia Afghanistan (1989–1991) India (1975)

Middle East Lebanon (1991–1995)

122  B. B. KELLY

Some caveats should be kept in mind regarding the representation of consociation by the data used here. For instance, a country could theoretically have a proportional representation electoral system that never translates into executive representation for a permanently excluded group and MV power that only appears to exist in constitutions but cannot actually be exerted. If a permanently excluded group was somehow overlooked by the MAR project or if only an unrepresentative element of it is included in a coalition, its country’s achievement of a positive score indicating inclusive coalitions might not constitute a genuine grand coalition. Even the inclusion of some element of such a group in a coalition does not render this body representative of all potentially antagonistic elements. Therefore, this situation would not fulfill Lijphart’s strictest definition of grand coalition, as including all potentially antagonistic groups. Fortunately, for the purposes of this study, it was possible to construct data for SA which avoid conflation with repressive segregation. The ability to voluntarily restrict one’s contact to only members of one’s group is thus counted as SA, while involuntary restrictions of this sort are not. This was accomplished by coding each year of a country’s experience with moderate to high restrictions on cultural organizations, as indicative of an absence of SA. If a case receives a cumulative score of “4” or “5” for the variables corresponding to consociational components, it seems reasonable to assume that it does constitute an example of something approximating consociation. However, it should not be assumed that even a score of “5” perfectly corresponds to a strictly defined consociational situation. It is unfortunate that consociation cannot be represented more accurately in this quantitative analysis. Nevertheless, the data available for use here provides a completely unprecedented amount of knowledge concerning the use of consociation around the globe and its capacity for facilitating stability in plural societies. Examination of these data before the statistical analyses illustrates that 7% of the years of plural countries’ experiences between 1975 and 1995 can be described as somewhat reflective of consociation.

Statistical Results Multiple Regression Analyses Exploration of the relationship between consociation and stability, and empirical examination of several claims that have been made by various scholars about the system, is performed with the Beck-Katz crosssectional time-series model, using fixed-effect country dummies.

5  QUANTITATIVE RESULTS 

123

The data collected for representation of all of the variables here are known as cross-sectional time-series data. This means that they are data where multiple cases like countries or people are observed at more than two time periods. This dataset comprises 88 cross-sectional cases that each correspond to one country. The best type of cross-sectional time-series analysis to use on one’s data can be ascertained through a Hausman Test, examining which means of analysis will produce the most efficient, explanatory model with consistent, optimally reliable results.7,8 The Hausman Test for this database indicates that the random effects cross-sectional time-series variant of multiple regression analysis should be used. It should be remembered that, although the dependent variable is being analyzed in order to find out which factors contribute to stability, higher values for this variable actually correspond to higher levels of instability, as it is manifested by protests and rebellions. Appendix B lists all variables and the means by which their data was collected. Appendix C describes the manner by which the dependent variable and data were designated. The role that each consociational component plays independently is explored. The somewhat inclusive grand coalition variable (SGC) is not significant when the dependent variable is regressed on it, in models with it as the only independent variable and in models incorporating all independent variables. However, in a test using only the independent variables representing consociational components, this SGC variable is found to be exerting a statistically significant impact. While this set of analyses suggests that some other variables explain the variance in the dependent variable better than SGC, they also indicate that this variable exerts an influence in plural societies in the same direction as instability. Therefore, this project finds that somewhat inclusive grand coalitions exert a detrimental effect in these places, ostensibly because they exclude potentially antagonistic groups from executive power. Examination of this variable actually suggests that coalitions which do not contain representatives from a country’s MARs are more conducive to stability than those including members of some, rather than all, such groups. In contrast, the variable corresponding to highly inclusive grand coalition executives (HGC) is found to have a positive influence on stability, of a 7 “Panel.” 8 & More information regarding the manner by which this form of regression analysis was chosen is provided on the website.

124  B. B. KELLY

high magnitude, that is statistically significant in every regression test in which it is included. The results obtained using these models corroborate Lijphart’s belief that high levels of potentially antagonistic groups’ inclusion in executives are necessary for promotion of stabilization in plural societies. The roles played by variables representing the other consociational components are not as consistent with Lijphart’s observations. In the tests including the MV variable, its positive coefficients indicate that the potential MV power represented by this variable’s data is associated with higher levels of instability. When the variable is regressed with all of the independent variables, just those representing consociation, and those found to exert a statistically significant effect in any test, it is found to influence instability to a degree that is statistically significant. Interestingly, its influence on the dependent variable when it is the only independent one included in the regression is also positive but it is not statistically significant. In the four tests analyzing the SA variable, it is found to exert a statistically significant effect on instability, but in only that model which includes only the independent variables representing consociation. SA’s coefficient in this test is positive, suggesting a destabilizing effect. All of the coefficients assigned to the variable corresponding to PR electoral systems are positive, indicating an overall destabilizing influence. Those coefficients were statistically significant when this was the only independent variable tested and when it was examined in collaboration with those representing components of consociation. When it was evaluated in tests including all independent variables and those which were found to exert statistically significant effects in any tests, the PR variable lost its statistical significance. Perhaps this differential performance in two sets of tests is explained by the fact that it was not statistically significant in those tests including the democracy variable and so democracy, as it was operationalized here, may be responsible for effects seemingly exerted by PR. Interestingly, this democracy variable exerted very slight but statistically significant, detrimental effects on stability in every test in which it was examined. This result seems to suggest that the widely observed, long-term stabilizing effect of democracy throughout the world can be slightly less capable of deterring protest and rebellion than the control exerted by some nondemocratic governments determined to eliminate such phenomena. This result reminds one of Lustick’s emphasis that stability can be achieved by control, in the form of “sustained manipulation

5  QUANTITATIVE RESULTS 

125

of subordinate segment(s) by a superordinate segment.”9 The fact that control is widely assumed to be less capable of maintaining stability in the long term suggests that it is important to realize that the data representing this project’s dependent variable indicates levels of stability associated with each year, rather than figures adjusted to indicate the length of previous years over which such stability has been maintained. This set of regression tests collectively indicates that the three independent variables exerting a statistically significant influence on instability, when combined with the control variables and each other, are HGC, potential MV power, and democracy. The first of these exerts a stabilizing influence and the latter two destabilizing effects. Other variables which sometimes exert a statistically significant influence are PR, SA, somewhat inclusive ­executive coalitions (SGC), and external threats. The coefficients for these variables when they are statistically significant are all positive, indicating detrimental effects of different intensities. The test results are provided on the website. The relationships between the independent variables and stability indicated by this book’s statistical analysis can be understood better through R-squared statistics. They show the percentage of change in the dependent variable that is caused by the combination of independent variables used in each model. Although the tests provide substantial evidence confirming that elements of consociation influence stability levels, their R-squared statistics suggest that the manifestations of consociation which could be represented through independent variables accounted for only 6.7% of change in levels of instability indicating the occurrence of protest and rebellion. When all independent variables found to exert statistically significant influences on the dependent variable were examined, they collectively accounted for 9.9% of change in levels of instability. The value of the controlling variables is demonstrated in the regression test indicating that 14% of change in levels of instability is explained by the model incorporating all of the independent variables. These control variables were chosen to maximize our understanding of consociation, rather than to maximize the model’s ability to explain all change in levels of instability, as it is operationalized here. There are undoubtedly many other factors influencing the extent of protest and rebellion in all societies but 14% of these manifestations of instability are explained here by variables representing phenomena related to consociation. 9 Lustick

(1979, p. 330).

126  B. B. KELLY

Regression analyses indicate that a number of factors influence the role of consociation in plural societies. They include potential MV power, proportional representation electoral systems, somewhat and HGC, and democracy. It is certainly possible that other phenomena could better explain stability in plural societies but this dataset does include additional variables representing more factors which are capable of being examined statistically and are commonly believed to strongly influence the operation of consociation. The two prescriptions for lasting peace in divided societies which have so far commanded the most respect from both academics and policymakers are Lijphart’s system of consociation, based on the belief that intergroup cooperation is necessary only at the elite level, and Horowitz’s suggested mechanisms for promoting incentives for intergroup cooperation among both leaders and their constituents. This project’s regression analysis provides insights into the benefits and drawbacks of both academics’ preferred sets of institutions and practices. At first glance, it may seem that only the control variable of geographical concentration of segments involves Horowitz’s suggestions. These tests did not detect a statistically significant influence of this variable on stability in plural societies. Horowitz believes that electoral constituencies containing substantial amounts of people from different antagonistic groups, and electoral systems requiring voters to choose more than one candidate for each office, will promote moderation of campaign promises and policies, as well as consequent declines in polarization and conflict. This idea that incentives for moderation should be promoted through mass-level intergroup contact is completely at odds with Lijphart’s belief that, by keeping the non-elite memberships of groups separate from one another, permanent peace can be achieved in states that they will still share. Since Lijphart’s consociational components and theory derive from this general premise, if it was proven that they do facilitate stability, Horowitz’s conception of human nature in plural societies and the prescriptions designed according to it would be wrong. The variables used here do not perfectly represent consociation and the factors thought to influence its operation. Although findings ­regarding consociation suggest hypotheses concerning Horowitz’s suggestions for quelling intergroup conflicts, it must also be recognized that this analysis was not designed to evaluate Horowitz’s work. The arguable drawbacks of consociation would have been more effectively assessed if the variable or variables corresponding to the grand coalition component

5  QUANTITATIVE RESULTS 

127

of consociation had represented the presence of executives commanding the allegiance of all potentially antagonistic groups in each country. The analyses of these variables indicate that coalitions containing representatives of all MARs are conducive to stability, while those containing some but not all such groups seem to have a detrimental effect. Horowitz claims that no real grand coalitions have existed and defines these bodies as they would ideally have been represented with such variables, but could not be for methodological reasons. The contrary influences of highly, and somewhat, inclusive coalitions suggested by these analyses do not prove or disprove Lijphart’s contentions regarding the absence of opposition in a governance system. However, they are consistent with Lijphart’s belief that, the more a coalition represents all potentially antagonistic groups, the more stable their plural society will become. Proportional representation is required for consociational systems. Lijphart believes that stability is maximized in plural societies when power is allocated to potentially antagonistic groups as proportionally as possible and in a manner which helps them achieve cultural autonomy. To this end, he prescribes list proportional representation for these countries because it results in very proportional legislative seat allocation and includes no element designed to motivate mass-level cross-group negotiation. The single transferable vote electoral system is the only proportional representation system which does inspire such negotiation because it allows voters to list several preferences for individual candidates competing for the same position. However, it accounts for only a tiny percentage of proportional representation systems. This data does not differentiate between types of PR. Thankfully, they all allocate seats much more proportionally than majoritarian and undemocratic systems. Only one rare type, STV, promotes intergroup negotiation so this variable can be relied upon to reflect Lijphart’s general recommendation regarding PR. The variable corresponding to the consociational component of SA is only found to exert a statistically significant effect on stability when it is tested with the set of variables representing consociational elements. It does not influence stability in a statistically significant manner when it is the only independent variable examined, or when it is analyzed in conjunction with control variables. When its effect is statistically significant, the variable’s coefficient indicates that it promotes protest and rebellion. Consideration of those manifestations of SA which could be incorporated into its variable facilitates the derivation of knowledge from the

128  B. B. KELLY

regression results. These manifestations involve cultural organizations’ repression, MAR groups’ legal cultural autonomy, official language status, and relative geographic homogeneity of group populations. The tests presented here indicate that these manifestations can have a statistically significant effect on stability but also suggest that the autonomy they provide can be conducive to protest and rebellion. SA is seen to exert this influence in collaboration with other consociational components. However, comparison of two regression tests suggests that the democracy variable explains change in protest and rebellion levels better than that representing SA. This seems plausible, since it would appear that the overall detrimental impact of SA may be due to the extent to which it provides political and social freedom of expression and organization. These regression analyses suggest that it must also be kept in mind that all aspects of cultural autonomy cannot be relied upon to serve as universal deterrents to those aspects of Lijphart’s concept of stability corresponding to protest and rebellion. The consociational component of MV power is analyzed with a variable corresponding to its potential availability to MAR groups in situations where they can gain legislative representation in rough proportion to their numbers. Data indicating the actual exertion of MV power does not exist and would be extremely difficult to produce. It provides the first indication of the role of potential, if not actual, MV power in plural societies. Its examination in conjunction with the other variables indicates that it exerts a detrimental effect on stability, as manifested by nonviolent and violent opposition, in countries with democratic and other types of political systems. All of the regression analyses find clear, statistically significant influences of the data representing MV on the data corresponding to stability. However, the existing literature regarding the relationship between MV and stability in plural societies does not seem to provide sufficiently clear hints regarding the source of this relationship. Perhaps the most valuable insight to be derived concerning it, prior to future research and additional data collection, is that power granted to exert such veto power may need to be accompanied by incentives for intergroup cooperation, if it is to maximize stability. The results of this analysis suggest that the international community of constitutional drafters and academics needs to recognize a crucial distinction involving the effects of majoritarianism in different situations. Particularly in countries with certain constellations of potentially antagonistic groups, simple majority decision-making in the electoral or

5  QUANTITATIVE RESULTS 

129

legislative arenas will permanently disempower them and result in terrible injustice and likely eventual instability. However, not all majoritarian decision-making is of this type. If a country is divided into regions or constituencies with sufficient heterogeneity of such groups, requirements for assent by relatively large proportions of electors or representatives in elections and for other decisions may actually motivate them to work together, without forcing them to assimilate or surrender their distinctive identities. For example, if majoritarianism unavoidably compelled the assimilation of potentially antagonistic groups, the USA would either have soon disintegrated following a war even more bloody than the civil war that did take place or it would be the most ethnically homogeneous immigrant society imaginable, after its two centuries of existence. The analyses performed suggest that elements of Lijphart’s and Horowitz’s premises and prescriptions are both correct and perhaps even compatible. The effects exhibited by the variables corresponding to consociational components suggest that stability may be maximized when MAR feel that they can gain representation in their countries’ executives but also feel motivated to work with members of other groups to achieve their common and separate cultural goals. Scatterplots These findings are further explored through nine scatterplots showing the relationships between the variables corresponding to consociational components, instability, and democracy. A scatterplot confirms that the data points representing those cases whose executives were in one of the “grand coalition” categories fell in a manner consistent with the results obtained through regression analysis, indicating a beneficial impact of the more inclusive category of executives. Regression tests indicate that SA has a statistically significant detrimental influence but loses statistical significance when tested in combination with democracy. The relevant scatterplot shows a substantial concentration of cases which exhibit lower levels of instability and were recognized as exhibiting “segmental autonomy.” However, the line fitting this scatterplot is so close to being flat that it is difficult to interpret its meaning.10

10 &

More details concerning the individual scatterplots are on the website.

130  B. B. KELLY

Regression analyses indicate that MV power and PR both exert a statistically significant, detrimental effect. The scatterplots for both of these relationships indicate the largest concentration of cases lies neither in the lowest nor the higher levels of instability. Lines fitted to the scatterplots’ cases do very slightly slope down, corroborating the regression analyses’ findings that each of these variables exerts a detrimental effect. However, the scatterplots also emphasize and confirm that these detrimental effects are less pronounced than the positive influence exerted by the variable corresponding to HGC. These scatterplots illustrate that a large proportion of cases indicating the presence of these two components were observed to enjoy relatively low amounts of instability. Perhaps more detailed data, indicating factors such as exertion, rather than availability, of MV and particular PR systems, will yield more valuable quantitative insights. The variables corresponding to the system’s components were not regressed on democracy. The impact of democracy was controlled for through an additional independent variable. However, scatterplots including democracy were conducted here because of their potential insights concerning other scholars’ commentaries involving the extent to which consociation is democratic. A scatterplot focusing on grand coalition and democracy confirms that most cases are not governed by executives falling into the categories represented by this project’s grand coalition variable. However, the scatterplot and fitted line emphasize that those which are governed by highly inclusive executives are more democratic. Another plotting SA and democracy suggests that places using the elements of SA which could be represented quantitatively here are fairly evenly spread along the continuum indicating levels of democracy. This is consistent with most of the regression analyses’ inability to detect a statistically significant impact of SA on stability. One scatterplot and the line fitted on it suggest an intriguing hypothesis concerning the data used for representation of the consociational component of MV power. It would appear that more cases with MV power are undemocratic, than democratic. The data used to represent this consociational component was the first ever compiled to do so but, due to a lack of data necessary for indication of the exertion of such veto power, the variable is represented through information indicating whether countries’ constitutions seemed to provide an opportunity for such exertion. This scatterplot suggests that a concentration of ­country cases exists in which a lack of democracy is present alongside constitutional provisions that appear to enable MV. Lijphart envisions

5  QUANTITATIVE RESULTS 

131

this power as an element of democratic systems so it would appear that additional data indicating its exertion is necessary for comprehension of its role. Some critics of consociation argue that the system is insufficiently democratic so the implication of this scatterplot is not incompatible with all scholarly analysis. However, the contrast between the scatterplot and Lijphart’s consociational theory inspires one to speculate that perhaps the association between MV power and democracy is in some way related to the destabilizing effect of the variable corresponding to this consociational component in regression analyses. If data can be compiled documenting the exertion of this power, it may yield much greater comprehension of the nature of this component’s role in plural societies. Two additional scatterplots corroborate the regression analyses conducted with the variables on which they focus. One confirms that PR electoral systems are much more associated with democracy, but also reminds us that PR systems are sometimes used in less democratic countries. Another shows that democracies are very slightly more associated with instability, thus reminding one again of the ability of undemocratic regimes to deter protests and rebellion. In general, the only finding derived through scatterplots which is surprising considering this project’s regression analyses is the revelation that more cases coded positively for MV power are coded as autocratic, rather than democratic. Factor Analyses Another method of statistical analysis which sheds light on the relationships between all of the variables explored for this project is factor analysis. This tool enables discernment of underlying, latent factors with which multiple variables are associated. Each such factor explains some of the variance in a set of variables. The results of three factor analyses are provided on the website. Some information regarding the nature of factor analysis output facilitates interpretation of these tables. For each factor analysis, the number of variables analyzed and the number of factors identified are the same. The “eigenvalues” provided for each factor indicate the total variance accounted for by each factor. The Kaiser criterion “suggests to retain those factors with eigenvalues equal or higher than

132  B. B. KELLY

1.”11 Looking across factor analysis tables to the right, the “proportions” indicate the “relative weight of each factor in the total variance.”12 Thus a proportion of 0.9096 indicates that the factor to which it is assigned “explains” 90.96% of the variance among the data.13 The second part of these tables involves the factor loadings and their uniqueness. The factor loadings, the values corresponding to each factor for each variable, are highest when their variables are “more relevant in defining the factor’s dimensionality.”14 Negative factor loadings indicate “an inverse impact on the factor.”15 The uniqueness assigned to each variable in this portion of the factor analysis output indicates the percentage of their variance that is not shared with other variables tested in the analysis. For instance, a uniqueness value of 0.4973 indicates that 49.73% of its variable’s variance is not shared with the analysis’ other variables. Factor analysis identifies axes within the set of data points contained in one’s dataset and rotation of the factor axes “simplifies the factor structure and therefore makes its interpretation easier and more reliable.”16 Varimax orthogonal rotation of the factor axes, which was used for this project, facilitates the derivation of a set of results in which “each original variable tends to be associated with one (or a small number) of factors, and each factor represents only a small number of variables.”17 Such rotation clarifies the role being played by the variables within the dataset. Screeplots of the eigenvalues associated with factors provide further evidence of which factors explain most of the variability in one’s dataset. Screeplots for this project’s data indicated which factors were appropriate for conversion into additional independent variables and then those new variables’ influences on the dependent variable were examined through additional regression analysis, to maximize the breadth and accuracy of quantitative results. This general information regarding the mechanics of factor analysis facilitates consideration of the findings obtained through its use regarding this dataset focused on assessing the role played by consociation. 11 Torres-Reyna, “Getting Started in Factor Analysis (Using Stata 10) (ver.1.5).” http:// dss.princeton.edu/training/Factor.pdf. 12 Torres-Reyna. 13 Torres-Reyna. 14 Torres-Reyna. 15 Torres-Reyna. 16 Abdi (2003, p. 2). 17 Abdi (2003, p. 3).

5  QUANTITATIVE RESULTS 

133

The two sets of factor analysis conducted for this project are designed to derive factors from the data representing the independent variables corresponding to consociational components and factors existing in the larger dataset comprising all of the independent variables. The set of factor analysis examining only consociational component data found no factors with eigenvalues reaching the threshold of 1, customarily used to decide whether factors should be analyzed further. Factor 1’s eigenvalue was .0885 so, out of curiosity, I carried out the further stages of factor analysis. The regression analyses conducted with variables comprising the two factors indicated by the screeplot indicate that their effects on the dependent variable are statistically significant but very small. The main aspect of interest regarding the first set of factor analyses is that, although its eigenvalue is below the threshold, the signs corresponding to independent variables for the first factor’s loadings are in the same pattern ascertained for the first factor in the second set of factor analysis. For this more fruitful derivation of factors from the data corresponding to all independent variables, two factors are found to have eigenvalues above the threshold of 1. The insights suggested by these factors remain largely unchanged as a varimax rotation and screeplot are conducted. Regression analyses using variables corresponding to the first two factors suggest that the first has a statistically significant but extremely small effect on stability, while the second’s influence on stability is both statistically significant and quite useful as a means of explaining variation in the dependent variable. The factor loadings for Factor 1 indicate that the variables most relevant to it are, in decreasing order, PR, democracy, external threats, political polarization, MV, SA, socioeconomic equality, homogeneous governance units, and somewhat inclusive executives. The first three of these are significantly higher than the others. The signs allocated to the variables indicate that the factor is differentiating external threats, SA, MV, and somewhat inclusive executives from the other variables whose factor loadings indicate relatively substantial relevancy. The first factor identified through varimax rotation also assigns much higher factor loading values to the same three variables, PR, democracy, and external threats. The signs on this factor’s variables are the same as those ascertained for the first factor from the analysis prior to varimax. However, this varimax produced first factor indicates a substantially greater amount of influence by the PR variable. Consideration of the first factors found through initial and varimax rotation factor analysis suggest that some phenomenon distinguishing two

134  B. B. KELLY

groups of this dataset’s variables exists but has a small overall impact on stability, as it is represented by the dependent variable. Lijphart observes that external threats can be a condition favorable to the success of consociation and SA, MV, and somewhat inclusive executives attempt to operationalize elements of consociation. It should be noted that highly inclusive executives much more closely resemble the consociational component of “grand coalitions.” However, this set of variables which is allocated the same sign for both of these tests’ first factors seems to reflect some substantial elements of Lijphart’s consociational system and theory. While PR is also a fundamental component of consociation, all of the quantitative analysis conducted for this project collectively seems to indicate the necessity for future operationalization of PR through data specifying individual systems of PR. The list PR electoral system is identified by Lijphart as optimal for consociation systems because different types of PR can introduce dramatically different incentives for voters and politicians. The case studies examined in later chapters suggest that PR systems engendering incentives for greater intergroup consideration could have profound effects on stability. The strong association between the PR and democracy variables in this factor analysis’ first factor ostensibly derives from the fact that PR systems are widespread and most are not operating in conjunction with substantial external threats and practices resembling consociational components. Perhaps this first factor suggested by this analysis and its rotation is related to the dominant academic consensus regarding stability in plural societies, which inspired Lijphart’s initial identification of consociation. This consensus is discussed in Chapter 1. Some of the most accomplished political scientists working at that time, including Almond, Bluhm, Downs, Duverger, Lipset, and Neumann agreed that division of societies by sharp cleavages inherently discouraged stability. For example, Rabushka and Shepsle contend that “[h]istory shows that democratic stability and cultural diversity are often incompatible in the postindependence politics of many plural societies.”18 Identification of this factor, even though its impact on the dependent variable is not sizeable, provides quantitative evidence that multiple elements of consociation and consociational theory are related to a single, underlying factor, which is different from other democracies. Factor analysis suggests

18 Rabushka

and Shepsle (1972, p. 207; 2009, p. 207).

5  QUANTITATIVE RESULTS 

135

that the effects of the consociational components are more complicated than are indicated through regression analysis. In the regression analysis in which all variables representing the consociation system were statistically significant, highly inclusive executive coalition was the only variable found to be exerting a stabilizing influence in plural societies. In contrast, factor analysis indicates that this factor is not influenced much by such coalitions and consistently differentiates PR and democracy from some consociational components and favorable conditions. This contrast is probably partially explained by the regression analyses’ focus on ascertaining variables’ influence on stability, and factor analyses’ identification of factors regardless of variables’ relationship to the dependent variable. Regression analyses’ distinction between highly inclusive executives and the other variables representing consociation also seems consistent with factor analysis’ distinction between the two variables representing highly and somewhat inclusive executives. The influence of this factor on stability, as it is operationalized in the dependent variable, suggests that this influence is extremely small. However, the low r-square statistic provided there could be interpreted as tremendously valuable evidence that Lijphart was correct to contend that plural societies requiring consociation are not necessarily less stable than those democracies preferred by the academic consensus which first inspired Lijphart’s identification of consociation. The second factor identified as influential in this dataset exerts a statistically significant and substantial effect on stability. Its eigenvalue establishes the advisability of its consideration and, when the variable constructed to represent this variable is regressed on the dependent stability variable, it attains an overall R-squared value of 17.03%. This indicates that this model containing only one independent and one dependent variable explains 17% of the variance in the dependent variable. Comparison of findings provided by factor analysis and varimax rotation for this factor indicate that the three variables influencing it the most are population size, SGC, and external threats. When the rotation is performed to help clarify which variables strongly contribute to it, all of the factor loadings for other variables are below .20. All three of these are allocated the same sign, indicating that they are exerting an influence in the same direction on this factor. There is something they have in common that has a big impact on stability, represented by data indicating protest and rebellion levels. This suggests that these phenomena are strongly associated with each other. Intriguingly, Factor 1 is not related

136  B. B. KELLY

closely to stability but also indicates a close association between SGC and external threats. Prior to varimax rotation, factor analysis also found that the variables indicating homogeneity of governance units, SA, and MV played a significant role in the constitution of this factor. This homogeneity and the SA variable were found to be exerting an influence in the same direction as the three phenomena which remained important in the varimax rotation, while MV’s sign was in the opposite direction. The MV variable had its factor loading drop from −.268 in the factor analysis, to a tiny −.061 in the varimax rotation results. Factor two suggests that the combination of a large population, external threats to their country, and somewhat inclusive executives, is clearly conducive to stability as it is operationalized here. Lijphart expects smaller populations and substantial external threats as contributing to the success of consociation. He also observes that more inclusive executive coalitions are constitutive of consociation so this factor confirms that these elements of consociational theory are related but this is not consistent with Lijphart’s expectations. However, it does make sense that, in countries with large populations and faced with external threats, executive coalitions including multiple potentially antagonistic groups will increase the likelihood of domestic stability. The factor analyses and varimax rotations performed for this project collectively suggest a number of insights, some of which are not shown through regression analysis and have never before been quantitatively identified. The SA variable is statistically significant in only one of the regression analyses, but its influence is also indicated through factor analysis. The factor analyses corroborate Lijphart’s observation that SA plays a significant role in relation to the other consociational components and factors likely to influence consociation’s performance. Another intriguing aspect of the factor analyses is that they all yield identical signs for the MV and HGC variables. While these variables’ factor loadings are sometimes low, these variables were found to be exerting opposite effects in all of the regression analyses so these factor analysis signs are intriguing. The signs allocated to the highly inclusive executive variable in regression analyses are all in the opposite direction to those allocated to the other variables representing consociational components. In contrast, Factor 1 indicates an influence in the same direction by all consociational component variables except PR. PR’s opposite sign in Factor 1 perhaps reflects the fact that the category of PR electoral systems is so varied and, in turn, suggests the value that could be derived

5  QUANTITATIVE RESULTS 

137

from more detailed data indicating the use of the specific types of such systems. Considering the likelihood of similar motivations for adoption of joint executives, MV power, SA, and PR, evidence of their associations in regressions and factor analyses is not surprising. Although lack of more detailed and extensive data representing these consociational components has likely frustrated these statistical tests’ ability to elucidate the nature of their relationships to each other and to stability, this project’s demonstration that they are connected and are related to stability is unprecedented. Collectively, these factor analysis results suggest that the data examined here reflect the validity of many scholars’ generalizations concerning the likelihood of governance mechanisms being used in conjunction with one another. For example, Factor 1 can be interpreted as exhibiting the existence of the set of democratic, less divided cases whose identification inspired Lijphart to identify consociation decades ago. It also provides the first suggestive evidence that neither this set of democracies, nor consociational ones, is clearly more conducive to stability as it is operationalized here. The factors identified in these tests also corroborate the existence of a tendency for consociational components and favorable conditions to be used in conjunction with one another.

Assessing the Value of These Quantitative Results Multiple regression analyses are often evaluated based on the extent to which their independent variable data explains variation in their dependent variable data. According to that criterion, the multiple regression results presented here are not particularly exciting. Multiple regression analysis searches for regression lines which indicate trends that best explain their datasets. The R-squared statistic indicates the percentage of variance in a dataset which is explained by a given regression line. R-squared statistics derived for this project indicate that 14% of the variance in the dataset is explained by a regression line. Maximization of such R-squared statistics obtained for my regression analyses would be of much greater importance if this project’s goal was to develop an overarching, predictive model of how stability is achieved in plural societies. This is not the goal for which this project was designed. Regression analysis was conducted with this dataset in order to test hypotheses about the effects of consociational components.

138  B. B. KELLY

The multiple regression results presented here constitute the first quantitative, international, long-term evidence that some elements of Lijphart’s system of consociation do exert a statistically significant effect on stability. The data represented by this project’s variables may not account for a large percentage of the variation in stability but the international influence of some on it has now been demonstrated statistically. In Chapter 6, the comparison of this study’s statistical results with other scholars’ related quantitative findings emphasizes that no other dataset has represented consociation and Lijphart’s claims concerning it, to anywhere near the same extent as the dataset analyzed here. The unprecedented degree to which this project’s dataset corresponds to Lijphart’s claims concerning consociation means that the confirmation of statistically significant relationships for this project is valuable to those who construct governance mechanisms throughout the world. Extensive qualitative interpretation of developments in individual countries has provided a wealth of insights into the operation of consociational components but objective, international, long-term examination of their results is just getting started. The analysis of this dataset through scatterplots and factor analysis further emphasize its explanatory value and the need for more data collection regarding some of the phenomena under examination. The scatterplots provide more evidence that the more inclusive category of executives is conducive to stability. In addition, they illustrate that the influence of the variables designed to represent consociational components is complex and that more detailed data corresponding to these phenomena would likely result in much greater understanding of consociation’s effects.

Conclusion The body of quantitative results presented here indicates that highly inclusive coalitions deter violent and nonviolent manifestations of instability, while somewhat inclusive coalitions, potential MV power, and PR electoral systems actually exacerbate it. However, potential MV power is the only variable corresponding to a consociational component which is still found to exert a statistically significant detrimental effect when analyzed through regression in conjunction with control variables. It would appear that a slightly detrimental effect of democracy on instability explains variance in this dependent variable better than somewhat inclusive coalitions, SA, and PR electoral systems. Scatterplot analysis

5  QUANTITATIVE RESULTS 

139

corroborated the regression results but yielded the observation that a large concentration of cases whose constitutions provide MV power are, in fact, relatively undemocratic. However, regression shows that MV power’s influence on instability is not explained better by the democracy variable. The provision of such power in constitutions and the existence of HGC are found to exert statistically significant effects on instability in the plural societies represented in this dataset. The positive role of executive inclusivity corroborates Lijphart’s observation that it is crucial for members of MAR to feel as if their representatives could attain some measure of executive power. The negative influence associated with somewhat inclusive coalitions suggests that exclusion from executive power of only some of a country’s MARs may be sufficient to destabilize a plural society. Factor analysis conducted to further explore this dataset confirms the explanatory power of the variables representing several consociational components and conditions identified as Lijphart to be conducive to their promotion of stability. The first factor found to exert a statistically significant influence on the dependent variable appears to discriminate between the larger body of democracies and those exhibiting more elements of consociation and consociational theory. Assuming that this interpretation is correct, the very low R-square statistic calculated in the regression of this factor’s variable on instability is actually of tremendous value. It would suggest that neither set of countries is more conducive to instability as it is operationalized here. This would vindicate Lijphart’s belief that consociations are not inherently less stable than more majoritarian democracies operating in less plural societies. While the second factor again emphasizes the explanatory potential of the variables representing consociation and Lijphart’s theory concerning it, the results of its varimax rotation seem to suggest that it has much less to do with consociation than the first factor. The variable representing a consociational component shown to heavily influence the second factor is somewhat inclusive coalitions, which are associated with consociation to some extent. In general, this project’s quantitative analysis demonstrates that elements of consociation do affect stability in plural societies, highly inclusive executives are beneficial, and the influence of MV power is nuanced. It also emphasizes that additional data will greatly expand our understanding of consociation, and Lijphart was correct to assert that consociation in plural societies is not an inherently less stable system of government. These findings are explored further through the qualitative case studies which comprise Part II of this book.

140  B. B. KELLY

References Abdi, Hervé. “Factor Rotations in Factor Analyses.” Encyclopedia for Research Methods for the Social Sciences. Eds. Michael S. Lewis-Beck, Alan E. Bryman, and Tim Futing Liao. Thousand Oaks, CA: Sage, 2003. Lijphart, Arend. The Politics of Accommodation: Pluralism and Democracy in the Netherlands. 2nd ed. Berkeley: University of California, 1975. Lustick, Ian. “Stability in Deeply Divided Societies: Consociation Versus Control.” World Politics. 31:3 (1979) 325–344. Rabushka, Alvin and Kenneth A. Shepsle. Politics in Plural Societies: A Theory of Democratic Instability. Columbus, OH: Charles E. Merrill, 1972. Rabushka, Alvin and Kenneth A. Shepsle. Politics in Plural Societies: A Theory of Democratic Instability. Columbus, OH: Charles E. Merrill, 2009. Torres-Reyna, Oscar. “Getting Started in Factor Analysis (Using Stata 10) (ver.1.5).” Princeton University. http://dss.princeton.edu/training/Factor.pdf.

CHAPTER 6

Comparison of this Project’s Quantitative Findings with Others’ Evaluations of Consociation’s Effects

This project fills a gap within a large body of scholarly analysis of consociation and power-sharing, including studies conducted using logic, case studies, and quantitative data. To an unprecedented degree, the international dataset presented here was designed to represent consociation and Lijphart’s theory concerning its role. This chapter explores the implications of this project’s findings, in comparison with other scholars’ qualitatively and quantitatively derived contributions concerning consociation and power-sharing. However, those which challenge whether consociation promotes stability are addressed in Chapter 2 because they suggest several hypotheses which ideally would be explored statistically. Those works influenced virtually every methodological decision which contributed to this project’s design. Their hypotheses include the suggestion that other phenomena may be responsible for stability commonly attributed to consociation. They also explore the theoretical possibilities that stability causes consociation, this governance system cannot maintain long-term legitimacy because it freezes cleavages, it contains inadequate incentives for intergroup cooperation, and/or it is insufficiently democratic. This project’s quantitative analyses presented in Chapter 5 suggest that some elements of consociation do promote stability, defined as including evidence of legitimacy. Therefore, with regard to those elements, those speculations appear to be unsubstantiated. However, the quantitative results suggest that other elements of the system may not contribute to stability. The case studies presented in Part II suggest that group autonomy and empowerment, combined with intergroup loyalty © The Author(s) 2019 B. B. Kelly, Power-Sharing and Consociational Theory, https://doi.org/10.1007/978-3-030-14191-2_6

141

142  B. B. KELLY

and certain incentives for moderation, may constitute the optimal basis for achievement of stability in plural societies. This chapter explores the implications of the quantitative results, discerned through their comparison with multiple categories of literature examining consociation and power-sharing. The categories to be explored are those which are most pertinent to the testing of consociation’s effects. These are scholarly contributions involving the extent to which consociational theory is falsifiable, the conditions which have been identified as conducive to the system’s success, and other quantitative studies involving phenomena similar to those explored here. The construction of this project is relevant to the discussion of consociational theory’s falsifiability because it demonstrates that consociation can be tested in a methodologically sound manner. While Lijphart’s evolution of thought concerning the system contains small contradictions, these were caused by scientifically sound theory progression. A falsifiable version of his concept of consociation, and claims for it, is easily identifiable. Comparison of this project’s findings, with others’ observations and theoretical explorations of Lijphart’s favorable conditions, facilitates greater understanding of those conditions’ role in plural societies. This chapter concludes with a comparison of this project’s conclusions with those arrived at by other scholars using quantitative analysis to explore related phenomena. These other quantitative analyses also indicate that some elements of power-sharing are conducive to long-term peace. Their findings involving PR electoral systems and territorial autonomy suggest interesting subjects for further research.

Responses to Lijphart’s Work in This Area: Is It Scientifically Falsifiable? It has been argued that Lijphart’s body of work concerning consociation includes numerous invalid observations, assumptions, and conclusions. Ian Lustick believes that most of the theory’s scientific defects crept into Lijphart’s work as he revised his approach in response to criticisms and a “changing evidentiary base.”1 Lijphart responds to this claim by arguing that his constant improvement of consociational theory means that it is not “fair” of Lustick to “detect all kinds of ‘inconsistencies’” within 1 Lustick

(1997, p. 90).

6  COMPARISON OF THIS PROJECT’S QUANTITATIVE FINDINGS … 

143

his body of work on the subject.2 In an article published fifteen years before, Lijphart also points out that revisions he had introduced were “most[ly] … small,” and “do not affect the fundamental propositions of consociational theory.”3 The fact that a core definition of consociation, comprising the consistent aim and four core components of the system, has remained intact throughout his work, indicates that its basic elements have not been substantially changed by his revisions of the theory. The persistence of these elements is illustrated through the literature review conducted to identify them for operationalization here, as well as the description of their quantitative analysis. Lustick argues that Lijphart’s elaboration of the theory, to remedy problems identified by other scholars, renders his body of work unfalsifiable. Lustick contends that this evolution has become problematic because such a broad range of phenomena can be portrayed as consistent with some version of Lijphart’s theory. However, even if his entire body of work on consociation was, as a whole, collectively unfalsifiable, this would still not preclude distillation of his most recent version, specified in a scientifically falsifiable form. As a policy recommendation, consociation remains pretty clear because the fundamental elements of it have remained unchanged through his work. Lijphart’s requirement of the same four core components to justify identification of consociation suggests that, regardless of how he has changed other details of his theory, these components can be considered to collectively define consociational systems. His theory’s falsifiability has also been questioned through criticisms involving his terminology, methodology for detecting consociation in practice, the relationship between cleavages and consociation, and the conditions which he believes are favorable to consociation. Lustick observes that the “imprecision” of Lijphart’s terms has been repeatedly criticized and claims that, without well-specified concepts, the validity of consociational theory cannot be empirically tested in a replicable manner because data would necessarily be chosen subjectively.4 Lustick points out that, a few years after Lijphart first articulated this theory, Eric Nordlinger criticized “the imprecision of his terms, the awkwardness of

2 Lijphart

(1998, p. 149; 2000, p. 425). (1984, p. 15). 4 Lustick (1997, p. 99). 3 Lijphart

144  B. B. KELLY

his typology, and his mischaracterization of key cases.”5 Lustick writes that Lijphart only defined key terms like “plural” and “stable” in publications following Nordlinger’s criticisms and that these characterizations were “explicit” but “far from precise.”6 Reference to Nordlinger’s work confirms that he argued that Lijphart’s consociational theory was “semitautological,” he questioned Lijphart’s interpretation of some historical experiences, and he observed that the meanings of the term “nationalism” in Lijphart’s works involving consociation are not always the same.7 Additional, and often similar, criticisms involving his terms and observations have been made by Nils Butenschøn, M.P.C.M. Van Schendelen, Sue Halpern, and Jűrg Steiner.8 Lustick asserts that Lijphart’s “imprecision in the definition of key concepts … precludes empirical testing.”9 However, the objective, internationally replicable means used here to define “plural,” and thus choose quantitative cases, exemplifies how imprecisions in Lijphart’s descriptions of such concepts do not render this theory untestable or unfalsifiable. The quantitative exploration presented here confirms that descriptions of the theory’s core concepts, provided throughout his body of work, are sufficiently consistent and clear to enable their quantitative testing in a methodologically sound manner. Lustick’s 1997 condemnation of Lijphart’s consociational theory is entitled “Lijphart, Lakatos, and Consociation” because Imre Lakatos provided widely respected criteria for the scientific credibility of theories and Lustick argues that consociational theory does not meet these criteria. Lakatos discriminated between naïve and sophisticated definitions of falsification, as well as progressive and degenerative research programs. Lakatos believes that naïve falsification consists of labeling an entire theory as falsified based on the existence of one observation “which conflicts with it.”10 He argued that naïve falsification “is not a sufficient condition for eliminating a specific theory.”11 Lakatos defined sophisticated falsification as occurring when three phenomena have occurred: a new theory 5 Lustick

(1997, pp. 99–100). (1997, p. 105). 7 Nordlinger (1972, pp. 15, 69, 90–92). 8 Butenschøn (1985, p. 99), Van Schendelen (1984, pp. 31, 32), Halpern (1986, p. 194), and Steiner (1981, p. 341). 9 Lustick (1997, p. 105). 10 Lakatos (1978, p. 32). 11 Lakatos (1978, p. 36). 6 Lustick

6  COMPARISON OF THIS PROJECT’S QUANTITATIVE FINDINGS … 

145

explains more than the one to be falsified, the new theory explains “all of the unrefuted content” of the original, and some of the new content of the new theory is corroborated by evidence.12 This logic led to Lakatos’ differentiation between progressive and degenerative research programs. He argued that modifications of their auxiliary hypotheses render theories progressive, rather than degenerative, when these modifications are always theoretically, and at least sometimes empirically, progressive. In other words, if they are progressive, these modifications will enable the updated theory to predict more than its predecessor and empirical evidence should confirm this new prediction. Lakatos summed up many of these points regarding theoretical integrity when he explained that “[w] e may rationally decide not to allow ‘refutations’ to transmit falsity to the hard core [of a theory] as long as the corroborated empirical content of the protecting belt of auxiliary hypotheses increases.”13 Lustick portrays Lijphart’s consociational research program as degenerative and several scholars argue that the theory has been falsified by changes evident in Lijphart’s work involving his assessment of cases, descriptions of terms, and conditions favorable to the system’s promotion of stability. As Lijphart has revised the auxiliary hypotheses of his consociational theory, it has explained more than the original and all of the original’s unrefuted content.  In addition, some of the revised theory has been corroborated with evidence. Therefore, according to Lakatos, the theory has not been falsified. Its research program is also progressive because it allows it to predict more than its predecessor and its revisions are confirmed by empirical evidence. This book finds that the hard core of the theory is evident throughout Lijphart’s writings and it is that consociation promotes stability in plural societies. Inconsistencies in the definition of terms such as nationalism in these writings do not affect this hard core. Lijphart has always defined consociation as consisting of the same four components. The fact that he later argued that they could exist in the absence of institutions mandating them does not change his central theory that these phenomena are conducive to stability, as he precisely and consistently defines it, in plural societies, as he precisely and consistently defines them. His evolving impressions of the extent to which societies have used consociation, and to what extent conditions in those societies

12 Lakatos 13 Lakatos

(1978, p. 32). (1978, p. 49).

146  B. B. KELLY

promoted its successful use, have also been identified as evidence that his theory is false. However, they do not involve the hard core of the theory and his thoughts about them are auxiliary hypotheses that fit Lakatos’ definition of those which characterize progressive research programs. Lijphart’s identification of consociational effects as evidence of the system’s operation in India is often identified as the most persuasive example of the theory’s lack of scientific credibility. Lustick argues that this portrayal is factually incorrect and points out that the proportional representation decision-making rules which comprise one core component of consociation are, by definition, antithetical to majoritarianism.14 However, regardless of the extent to which one agrees with Lijphart’s interpretation of those political conditions, the auxiliary hypothesis Lijphart describes in his discussions of India is evidence of a progressive research program. This hypothesis is that political phenomena corresponding to the four components of consociation do not have to be institutionalized or legally enforced to be recognized validly as constituting evidence of his power-sharing system. Even if one believes that he identifies some phenomena as consociational incorrectly, this auxiliary hypothesis fits Lakatos’ definition of a progressive research ­ program by predicting more than the theory as it existed before and being confirmed by some empirical evidence. Lustick observes that Lakatos himself changed his position on what constitutes a progressive research program but this shift similarly did not result in an implausible, theoretically inconsistent theoretical change.15 This book proves that Lijphart’s core theory regarding consociation is falsifiable by demonstrating that he has consistently and clearly defined his claim for the system and that its influence internationally can be tested in an objective and scientifically valid manner. Lijphart’s motives for conceiving many of the auxiliary hypotheses can be debated as extensively as those hypotheses’ credibility and the assessments of individual countries’ experiences with power-sharing. However, they fit Lakatos’ definition of a progressive research program and this analysis shows how rigorously consociation can be tested.

14 Lustick 15 Lustick

(1997, pp. 115–116). (1997, p. 89).

6  COMPARISON OF THIS PROJECT’S QUANTITATIVE FINDINGS … 

147

Lijphart carefully considers criticisms made of his consociational theory. For instance, in regard to arguments that he has neglected some definitional issues, he explains that his consideration and production of increasingly precise definitions of “plural” societies have not yet yielded criteria that are “exactly measurable” and do not require “impressionistic” estimation.16 He agrees that it is unfortunate that no means of precise measurement has been invented yet.17 So far, most academic attention has focused on specifying and measuring the “plural” element of his theory to be tested here, that consociation promotes stability in plural societies. In response to critiques focusing on whether or not individual countries should be categorized as “plural,” Lijphart agrees with Steiner’s proposal that it is more accurate to “think in terms of varying degrees of pluralism” rather than “a sharp dichotomous distinction” between plural and “homogeneous” (non-plural) societies.18 Steiner’s discussion of potential means for creating an index through which to identify these degrees illustrates that Lijphart has not been able to come up with a means of quantitatively measuring pluralism in a replicable fashion.19 Steiner explains that factors hindering the formulation of such an index include the relative importance of cleavages in different spheres of life (such as “interest groups, media of communication, schools, and voluntary associations”) and of different types of cleavages, such as those which are linguistically, ethnically, racially, and ideologically based.20 He also points out that manifestations of cleavages within all organizations, and particularly those which apparently contain one, particular segment, would also have to be measured and weighed in a fashion generally agreed upon by the academic community, in order to create an acceptable index.21 Steiner persuasively argues that the possibility of developing such an index designed to identify segmentation levels is remote and emphasizes that any index incorporating variables related to segmentation cannot be expected to provide the precise, appropriate cutoff value to distinguish plural from non-plural societies.22 Like Steiner, 16 Lijphart

(1985, p. 87). (1985, p. 87). 18 Lijphart (1981, p. 355). 19 Steiner (1981, pp. 343–344). 20 Steiner (1981, pp. 343–344). 21 Steiner (1981, p. 344). 22 Steiner (1981, p. 343). 17 Lijphart

148  B. B. KELLY

Lijphart recognizes that the index which he uses, designed by Douglas W. Rae and Michael Taylor, measures diversity, rather than potential or actual antagonism. However, he convincingly argues that it should be used to aid identification of “pluralism” because a more appropriate index does not exist and, while not all diverse societies are plural, pluralism does require diversity.23 Lijphart observes that “it has not been possible so far to devise quantitative indices” for pluralism but argues that “it is better to make well-informed impressionistic judgments based largely on qualitative data than to make no judgment at all.”24 Fortunately, Gurr’s Minorities at Risk team uses criteria unrelated to consociation to identify which countries are divided into culturally distinct, and potentially or actually antagonistic, groups that seem very akin to those groups whose presence Lijphart portrays as the hallmark of his “plural” societies. The Minorities at Risk dataset provides an independently identified, empirically replicable means of distinguishing plural societies which should be examined in an evaluation of Lijphart’s consociational theory. The possibility of thus quantifying the concept in this theory that has been most criticized as imprecise emphasizes that ambiguities in Lijphart’s descriptions of such phenomena do not necessarily preclude their scientific evaluation. The two least precise concepts that must be specified to enable analysis of this theory are “plural societies” and “stability.” Lijphart’s descriptions of them have been sufficiently clear to enable the quantitative representation of both with data collected by other scholars for studies unrelated to consociation. Lijphart supports his theory through reference to his impressionistic judgments concerning countries which he identifies as having experienced consociation. He persuades policymakers in other countries to adopt the system by referring to its apparent successes. Several critics argue that he cannot prove that the cases he refers to as consociational actually have employed the system. Particularly since Lijphart concedes that he cannot demonstrate which societies are plural in a replicable, empirically unassailable fashion, this makes it possible to challenge his identification of cases that he offers to exemplify the validity of his theory that consociation promotes stability in plural societies. In the absence of any agreed-upon indexes through which levels of pluralism could be

23 Lijphart 24 Lijphart

(1981, p. 357). (1981, p. 357).

6  COMPARISON OF THIS PROJECT’S QUANTITATIVE FINDINGS … 

149

empirically measured, Lijphart and his critics focused on analysis of individual cases to debate consociation’s performance. For instance, in his most comprehensive discussion of consociation, Lijphart identified Switzerland as providing one of the four best ­“concrete examples” of how consociation can “be a stable and effective system of government in plural societies.”25 He originally developed his consociational theory after discerning the system’s presence in the Netherlands. However, Lijphart revealed the importance of the Swiss case when he explained that “the concrete Belgian and Dutch cases are only imperfect instances of consociational democracy and approximate the ideal type of consociational decision-making less closely than Austria and Switzerland.”26 Lijphart’s portrayal of Switzerland has inspired substantial analysis by numerous specialists but Barry provided the most plausible argument that the country’s experience does not support Lijphart’s theory.27,28 Lustick succinctly explains that: Citing the mildness of ethnic conflict in Switzerland, the use of majoritarian techniques such as binding referenda, and successful challenges to elite decisions, Barry observe[d] that Switzerland is neither deeply divided, nor conflictual, nor consociational in design, nor stable because of undesigned consociational practices.29

Barry’s assessment that Switzerland was neither conflictual nor consociational also is not completely persuasive because its 1847 civil war was motivated by a cantonal autonomy dispute and referendums proposing constitutional change there require double majorities in both the federation and a majority of the cantons. Lijphart responded to Barry’s charges by explaining why he disagrees with him, pointing out that others who specialize in Swiss politics have also identified the existence of consociation and its positive effects in the country. He maintains that it is

25 Lijphart

(1977, p. 2). (1977, p. 185). 27 Barry (1975, pp. 481–490). 28 Some examples of works which have examined the claim that Switzerland is consociational are: Bohn (1980, 1981), Daalder (1971), Dunn (1972), Hughes (1978), Lehner (1984), Kriesi (1990), Linder (1994), and Steiner (1983, 1991). 29 Lustick (1997, p. 101). 26 Lijphart

150  B. B. KELLY

“clearly both consociational and spectacularly successful in maintaining peace and democracy.”30 Barry, Lijphart, and other scholars have also debated the extent to which consociation promoted stability in Lijphart’s other most “ideal” case, Austria. Barry accepted that Austrian consociation did coincide with an introduction of greater political stability in the decades after World War II.31 However, he argues that consociation was not a “necessary condition” for this stabilization, since economic and international factors also facilitated this development and there was little inter-segmental hostility among the Austrian masses.32 Lijphart addresses only the economic component of Barry’s argument, contending that high increases in prosperity do not necessarily promote political stability and demonstrating this through reference to trends in Belgium, Canada, and generally within Western Europe.33 Van Schendelen also “raises serious questions about Lijphart’s characterization of the Netherlands as a culturally fragmented society saved from democratic instability only by consociation.”34 Van Schendelen points out that specialists in Dutch politics do not all agree with Lijphart that there was a period in which there was serious intergroup antagonism and polarization along significant “cleavages” of political debate.35 He argues that Dutch elites actually sought to “maintain their positions” by exerting “[c]ontrol through enforced pillarization” and encouragement of a religious cleavage which weakened the potential threat to their power by socioeconomic “class antagonisms.”36 In addition, Van Schendelen argues that, while the Netherlands did experience stability as indicated by levels of violence and protest, data involving cabinet stability between 1917 and 1967 actually indicate a lack of stability.37 In contrast, Lijphart (1984, p. 11) believes that elites’ management of cleavages and levels of stability as he defines it, along with other evidence he provides, confirm that Dutch society was originally very divided and consociation ameliorated a potentially 30 Lijphart

(1985, pp. 90–91). (1975, pp. 490–491, 501). 32 Barry (1975, pp. 495, 493, 499). 33 Lijphart (1985, p. 94). 34 Lustick (1997, p. 103). 35 Van Schendelen (1984, p. 36). 36 Van Schendelen (1983, pp. 23, 25; 1984, p. 48). 37 Van Schendelen (1983, p. 14). 31 Barry

6  COMPARISON OF THIS PROJECT’S QUANTITATIVE FINDINGS … 

151

explosive situation. The complexities of such case studies are explored further in Chapters 7, 8, and 9, but contrasting Lijphart’s and others’ observations and interpretations of them emphasize the value of international, quantitative analysis of trends related to consociation. Although Lijphart’s core cases of the Netherlands, Switzerland, Austria, and Belgium seem to have attracted the most scrutiny by other scholars, his portrayals of other countries as consociational have also inspired substantial commentary.38 Halpern observes that the body of relevant literature contains extensive “[c]ontroversy” concerning the identification of which cases are accurately described as consociational.39 She believes that “virtually every case that can be located within the universe can also be argued out of it” and argues that this predicament persists because Lijphart’s theory is based on “faulty and imprecise concepts.”40 However, the four core components of consociation, and his general theory concerning their role, do remain consistent throughout his work and this project’s quantitative representation and exploration of them emphasize that the theory is not unfalsifiable.

Critical Analyses of the Conditions Identified by Lijphart as Conducive to Consociation’s Success The conditions identified by Lijphart as favorable to consociation have inspired extensive qualitative evaluation by scholars attempting to assess the theoretical accuracy of his claims concerning them. He explains that,“[g]enerally a factor that is favorable for the establishment of a consociation will also be a positive condition for its maintenance.”41 Since this maintenance constitutes one aspect of Lijphart’s definition of stability, it is reasonable to surmise that these conditions are also expected to promote stability in conjunction with consociation. This project’s findings are also relevant to critiques of the conditions focusing on their appropriateness as aspects of a theoretically sound, coherent theory of consociation. These discussions involve Lijphart’s depiction of the 38 Some works which have discussed Belgium in relation to consociational theory are: Covell (1981), Deshouwer (1994), Dierickx (1978), Dunn (1972), Hooghe (1993), and Pijnenburg (1984). 39 Halpern (1986, p. 181). 40 Halpern (1986, p. 181). 41 Lijphart (1985, p. 119).

152  B. B. KELLY

condition known as “cross-cutting cleavages,” as well as the extent to which his conditions are identified consistently, are uniformly favorable, and can be measured in a replicable fashion. Criticisms involving these subjects, Lijphart’s reactions to such commentary, and their relevance to this project, will be discussed here. One set of criticisms concerning when, and if, consociation promotes stability in plural societies examines Lijphart’s treatment of cleavages, and levels of pluralism. While Lijphart identified cross-cutting cleavages as a favorable condition, his theory of consociation also was inspired by his observation that an alternative theory explaining stability, prioritizing cross-cutting cleavages, did not explain stability in some more divided societies where groups lived more separately. Lijphart’s sharp departure from the initially widespread view that “cross-cutting” intergroup divisions are necessary for peace understandably has stimulated comparison of these two alternative theories for explaining stability. In contrast to consociational theory, Barry argued that the notion that “cross-cutting between political parties and socio-cultural cleavage lines moderates conflict seems… much more plausible and better supported.”42 Evidence exists to support both the consociational and cross-cutting theories. Lijphart consistently argues that consociational, autonomous freedom for segments, rather than cross-cutting allegiances, produces stability. However, he also believes that certain types of cross-cutting cleavages help antagonistic groups to decide to retain their shared state through consociation.43 Unfortunately, many cross-cutting political phenomena could not be included in this project’s statistical analysis. One factor related to socioeconomically motivated cleavages, socioeconomic equality, was examined quantitatively. It was not found to exert a statistically significant effect on stability when tested along with variables representing consociation. Conceivably socioeconomic equality could be more involved in the system’s establishment than its maintenance. Part II’s case studies suggest a positive role is played by cross-cutting cleavages, even in societies which provide extensive segmental autonomy. This socioeconomic cleavage which it was possible to analyze quantitatively is also the cleavage most widely portrayed as capable of mitigating divisions which render societies plural. Other scholars have argued

42 Barry

(1975, p. 487). (1977, p. 81).

43 Lijphart

6  COMPARISON OF THIS PROJECT’S QUANTITATIVE FINDINGS … 

153

that these cleavages’ observed and expected roles should be clarified more completely by Lijphart. Lijphart mentions some potential types of cleavages, such as those related to religion, social class, and language.44 However, Luigi Graziano correctly points out that Lijphart’s discussion of consociation “does not provide any systematic classification of the cleavages.”45 Graziano observes that Lijphart does not mention which cleavages consociation can more “effectively mediate.”46 Barry’s discussion concerning the applicability of consociation in ethnically divided societies also highlights the possibility that the system is more conducive to promoting stability in situations with certain cleavages, rather than others. The following quote, concerning Dutch experience, illustrates the extent to which Kieve believes that social cleavages influence developments involving consociation: the accommodationist arrangements were themselves determined by shifts in the balance of class forces between the bourgeoisie and the proletariat. For that very reason it is not wholly surprising that these arrangements should start breaking down with the intensification of class conflict, that is, precisely during those periods of social and political unrest when they are supposed to be most effective.47

Lijphart responds to allegations that he has neglected class issues by pointing out his introduction of socioeconomic equality as a favorable condition for promoting consociation.48 He also argues that the most divisive cleavages in plural societies “usually do not coincide with class divisions” and their dominance in these situations means that “class interests will have little chance to be articulated and promoted.”49 Lijphart does not discuss Kieve’s argument concerning the potency, as distinct from the suppression, of “class forces.” In general, Lijphart’s discussion of cleavages emphasizes the validity of Hans Daalder’s distinction between “mechanistic” and “qualitative” approaches to this subject.50 44 Lijphart

(1977, pp. 73, 75). (1980, p. 349). 46 Graziano (1980, p. 349). 47 Kieve (1981, p. 332). 48 Lijphart (1985, p. 98). 49 Lijphart (1985, p. 98). 50 Daalder (1974, pp. 612–613). 45 Graziano

154  B. B. KELLY

Daalder observes that most analysis concerning cleavages focuses on counting and identifying them, and measuring their relative strengths.51 He advocates more attention to the “qualitative differences of specific cleavages” and their relationships to levels of accommodation.52 Lijphart has not prioritized analysis of the specific roles played by cleavages of different types in plural societies. Greater comprehension of these roles promotes understanding of their interaction and prediction of each unique society’s potential for achieving stability with consociation. While the variable used here to represent socioeconomic equality is not found to have a statistically significant effect on stability when tested in combination with consociational elements, that in itself is notable considering existing descriptions of the influence of socioeconomic factors in divided societies. The case studies provided in Part II also point to several trends involving the positive effects of cross-cutting cleavages. Publications related to consociation’s performance have also considered extensively the conditions other than these cleavages, which Lijphart suggests are favorable to the system’s success. Box 4.1 provides a comprehensive list of them compiled by Bogaards, which was used to develop this project’s control variables. One set of criticisms argues that all of the conditions have been identified in an inconsistent manner, by Lijphart and other scholars. Lijphart first described conditions conducive to consociation in a 1968 publication. Due to his dedicated evolution of his thoughts concerning consociation, only a few of the conditions he portrayed then were presented in the same terms in a 1985 publication. Bogaards provides a systematic examination of this evolving list.53 He detects “four different sets of favourable factors,” exemplified by works published in 1968, 1969, 1977, and 1985.54 Bogaards observes that most academics who discuss Lijphart’s conditions either refer to the conditions endorsed in Lijphart’s most recent publication or they pick and choose which ones they wish to refer to.55 Lijphart’s list was highly unlikely to remain unchanged, following his tireless revision of this theory which aims to enhance the ability of consociation to benefit divided societies. 51 Daalder

(1974, pp. 612–613). (1974, p. 613). 53 Bogaards (1998, p. 478). 54 Bogaards (1998, p. 480). 55 Bogaards (1998, p. 480). 52 Daalder

6  COMPARISON OF THIS PROJECT’S QUANTITATIVE FINDINGS … 

155

Another set of criticisms is best represented by Adriano Pappalardo’s analysis of whether each condition is “unambiguously favourable to consociational cooperation.”56 He finds that only two fit this description: “inter-subcultural stability and elite predominance over a politically deferential and organizationally encapsulated following.”57 Pappalardo defines an empirically valid favorable condition as one which has been proven to be always conducive to consociation in all situations studied, in no way contradicted by any other condition identified as performing this function, not itself explained by variables representing consociation, and present in degrees that match the degrees to which consociation exists. Similarly, having qualitatively evaluated case studies, Van Schendelen concludes that the presence of Lijphart’s conditions is also “neither indispensable nor sufficient” for the achievement of stability in plural societies. For instance, having qualitatively analyzed the Dutch case, Van Schendelen argues that “[i]f the population is deferent and passive, if the antagonistic segments are isolated from each other and produce only scarce hot issues, and if the elites tend to converge their policy-views, then there is already stability in the country…”58 Such observations contradict Lijphart’s portrayal of a number of favorable conditions, including “popular attitudes favorable to government by grand coalition,” “widespread approval of the principle of government by elite cartel,” “adequate articulation of the subcultural interests,” “segmental isolation,” and “tradition of elite accommodation.”59 Lijphart’s contrasting views of the role performed by these conditions will be described in more detail, following a survey of Van Schendelen’s arguments that they are unmeasurable and “empty.” While this is one approach with which to assess the value of the conditions, these criteria are not the ones used by Lijphart to identify them. Although it is regrettable that data does not yet exist to quantitatively test these conditions more thoroughly, the quantitative and qualitative explorations presented in this book illustrate that it is not difficult to define these conditions in a manner which enables large-scale, methodologically sound analysis. Statistical confirmation of their role internationally would maximize our

56 Pappalardo

(1981, p. 363). (1981, p. 364). 58 Van Schendelen (1984, p. 43). 59 Bogaards (1998, p. 478). 57 Pappalardo

156  B. B. KELLY

comprehension of the issues raised by Pappalardo, which question the conditions’ conducivity to consociation. In addition to arguing that historical trends show that it is incorrect to assume that these conditions promote consociation, Van Schendelen believes that they are insufficiently defined to enable measurement of their existence in any valid, cross-national, and systematic manner. He describes the conditions as “empty,” by which he means that they are “neither indispensable nor sufficient,” and in other words, they “may be present and absent, necessary and unnecessary, in short conditions or no conditions at all.”60 Van Schendelen contends that, if these observations are untrue, it “should be possible to predict, on the basis of the presence or absence of the conditions, the chances of elite co-operation occurring in a plural society.”61 He argues that many “indicators” identified in consociational theory are not suited to cross-national empirical analysis because he believes they are “loosely formulated, theoretically biased, and not indicative of the complexities of consociational politics.”62 Lustick summarized Van Schendelen’s comments as suggesting that none of the conditions “can be tested in any practical way.”63 Such criticism suggests that large-scale quantitative analysis of such conditions is not worth attempting because they cannot be operationalized in a methodologically sound manner. A number of these conditions are not included in this project’s quantitative analysis but it should be pointed out that this is not because it would be impossible to represent them quantitatively. Perusal of the list of conditions compiled by Bogaards, which is provided in Box 4.1, illustrates that, even if Lijphart’s depictions of them are “loosely formulated” or “theoretically biased,” methodologically sound quantitative representation of all of them may require huge data collection efforts but the conditions are all identifiable, and thus quantifiable, phenomena. Consideration of Lijphart’s response to criticisms regarding his conditions suggests that his approach actually does not preclude valid, largescale, comparative empirical analysis.64 Lijphart explains that he intends

60 Van

Schendelen (1983, pp. 10–11). Schendelen (1983, p. 11). 62 Van Schendelen (1983, p. 11). 63 Lustick (1997, p. 107). 64 Lijphart (1985, pp. 114–116). 61 Van

6  COMPARISON OF THIS PROJECT’S QUANTITATIVE FINDINGS … 

157

these conditions to be “merely probabilistic instead of decisive” and that, “even if most or all of the favorable factors are missing, it is still possible to have a successful consociation.”65 Even before the contribution of the above criticisms, in a 1977 publication, Lijphart describes these conditions as “helpful but neither indispensable nor sufficient” for conflict resolution.66 Van Schendelen acknowledges this statement by Lijphart but interprets it as precluding valid measurement of the conditions’ effects.67 However, particularly through mechanisms such as statistical analysis, even slight causal relationships that may not exist in every case can be detected in a scientifically sound manner. Lijphart’s clear specification that these conditions do not necessarily always promote consociation also indicates that it is invalid to argue that Pappalardo’s evaluation of whether conditions are “unambiguously favorable” may possibly falsify Lijphart’s theory. Perhaps the most persuasive criticism involving these conditions is that they may only promote consociation when the most crucial one is present, namely elites’ intentions to promote accommodation. In reference to the role of Austrian elites, Barry went so far as to say that “what mattered was the attitude manifested in the offer and acceptance of a ‘grand coalition’ and that the same attitude without any consociational institutions could also have produced a stable democratic system.”68 Even if this is true, Lijphart’s conditions should still have explanatory value because this argument does not challenge the proposition that factors like elites’ positive attitudes exert some potential influence over consociational systems’ likelihood of success. Data for six of the conditions specified by Lijphart were easily accessible when this dataset was compiled so their influences on the stability, or maintenance, of consociation are quantitatively examined. It should be noted that Lijphart also expects that these conditions should promote the establishment of the system. That aspect of consociational theory is not tested for this project. The conditions which could be treated quantitatively are: external threats, segmental isolation, federalism, moderate multiparty system, geographical concentration of segments, small population size, and socioeconomic equality.

65 Lijphart

(1985, p. 116). (1977, p. 54). 67 Van Schendelen (1984, p. 34). 68 Barry (1975, p. 500). 66 Lijphart

158  B. B. KELLY

This project’s quantitative results emphasize that Lijphart’s claims concerning the conditions’ influences are falsifiable. In multiple regression tests of their effects on stability in the large number of countries represented in this dataset, the only variable corresponding to these conditions which is found to be statistically significant represents external threats. Perhaps not surprisingly, the variables representing elements of the consociation system were found in these tests to explain more variance in levels of stability than these conditions. On the other hand, factor analysis indicates that three of these do exert an influence on underlying factors, which eigenvalues suggest do account for a significant amount of variance in this dataset. These three conditions, operationalized through control variables, correspond to external threats, geographical concentration of segments, and population size. Particularly considering the findings of other scholars who have studied consociation and the qualitative insights involving individual cases presented in Part II of this book, it is not surprising that factor analysis identifies these conditions as influential. Perhaps the most intriguing aspect of this project’s case studies is their illustration that the favorable conditions for which data is least available, appear to exert a profound impact on consociation. Those conditions involve nationalism and overarching loyalties, approval of grand coalitions and elite cartels, internal cohesion of subcultures, segmental isolation, cross-cutting cleavages, and traditions of elite accommodation. Lijphart’s presentation of these conditions has been criticized for not describing them precisely and consistently enough to allow for their empirical measurement and evaluation. Pappalardo and others also argue that the conditions are not “unambiguously favorable.” However, the qualitative analysis presented in Part II facilitates understanding of Lijphart’s approach and criteria for identifying such conditions, which differ somewhat from the standards according to which some critics have judged these aspects of his theory. While it is very difficult to identify phenomena which are “unambiguously favorable” through case study and theoretical analysis, large-scale quantitative tests such as those performed for this project are capable of contributing to the demonstration of the most widespread effects of such phenomena. While it is unfortunate that more of Lijphart’s favorable conditions could not be incorporated into the tests presented here, inclusion of those for which data was available illustrates that their roles can be tested in a methodologically sound manner. The qualitative and quantitative aspects of this project show that there is merit to Lijphart’s approach to

6  COMPARISON OF THIS PROJECT’S QUANTITATIVE FINDINGS … 

159

identifying conditions and that their effects can be evaluated through objective, international, and data analysis. Careful analysis of theories similar to consociation, Lijphart’s many discussions of this system, and other academics’ commentaries concerning his work in this area suggest that consociation can be quantitatively represented in a falsifiable form and emphasize the value of an unbiased, quantitative study of its efficacy. It is important to recognize that Lijphart has contributed both a theory, containing all of his claims concerning consociation, and a description of the structure of this form of governance, which can be considered separately from the theory. While some criticisms attempt to falsify the entire theory by discrediting elements that are not crucial to its central proposition, others focus on attacking the proposition by examining whether the core elements of the government structure do promote stability in plural societies. Gary King, Robert Keohane, and Sidney Verba remind us that positivist “scientific research” is based on inferences involving empirical information, is conducted according to “public” procedures, is falsifiable, and “adheres to a set of rules of inference on which its validity depends.”69 Criticisms involving the scientific credibility of consociation can be evaluated through reference to these requirements. If they are found to have merit, it can be ascertained whether they are damaging to Lijphart’s theory in a way that does, or does not, challenge his fundamental claim regarding this government structure. Objections to the imprecision of definitions offered by Lijphart for some terms involve the requirements of falsifiability and for research according to public procedures. The specific problems with these definitions have already been discussed, but it should be recognized that, even if it is true that all of Lijphart’s descriptions of concepts considered as a unit are not falsifiable, this project demonstrates that versions of them can be adopted in a way which enables scientifically sound analysis of the central proposition involving consociation’s role. A more serious potential problem for this analysis would exist if the phenomena central to Lijphart’s theory and government structure were not measurable, in any replicable, empirical way. For instance, “segmental autonomy” is one of the four core elements of the consociation governance system but, if means of measuring its existence could not be decided upon and effectively implemented, this would render the entire

69 King

et al. (1994, pp. 7–9).

160  B. B. KELLY

theory unfalsifiable. Lustick argues that Lijphart’s alternative impressionistic and precise judgments make the theory unfalsifiable, Employing precise use of evidence and analysis to defeat alternatives to his position, he delegitimizes this technique as a basis for criticizing consociation and makes every possible allowance for the ‘impressionistic’ nature of social science where evidence or logic is inconvenient.70

Lijphart believes that, occasionally, searching for “exact” measurements is a “futile” distracting exercise.71 However, regardless of Lijphart’s specificity of definitions and measurements, the versions of them represented in the analyses described in Chapter 5 are precise enough to be tested and they collectively comprise a version of his theory which is falsifiable. It is possible that Lijphart’s presentation of consociational hypotheses over the years has been scientifically problematic and, as a result, this theory has not been properly evaluated. It is also possible, as I believe, to maintain that Lijphart’s successive presentations have been wholly honest developments and adjustments of arguments, and truly constructive responses to criticism, but that these have been misread by critics who often have been mission-committed to rejecting the political prescriptions suggested by consociational theory. Whatever one’s judgment on Lijphart’s own successive texts and how to read them, this book shows that it is possible to represent his arguments clearly and to test them in a proper social scientific manner.

Previous Quantitative Evaluations of Concepts Related to Consociation The ambition to quantitatively test consociation’s progress in plural societies has existed for decades. Previous attempts to do so have not focused on evaluating Lijphart’s consociation system and his theory that it promotes stability in plural societies. No data has been available to represent some of the four core components of consociation, such as grand coalitions, minority veto power, and segmental autonomy. To accurately assess the effects of this system, which Lijphart has carefully detailed, revised, and recommended for many years, data for these elusive variables was compiled 70 Lustick

(1997, p. 112). (1985, p. 88).

71 Lijphart

6  COMPARISON OF THIS PROJECT’S QUANTITATIVE FINDINGS … 

161

specifically for this book through analysis and comparison of other ­datasets and hundreds of constitutions. Publications by Caroline Hartzell, Matthew Hoddie, Barbara Walter, Michaela Mattes, Burcu Savun, Pippa Norris, Powell, Joel Selway, and Kharis Templeman have quantitatively analyzed the effects of systems similar to consociation. Table 6.1 facilitates their comparison, categorizing them according to their cases, hypotheses, and results. With the exceptions of these contributions by Norris and Powell, these analyses share very similar hypotheses. Those by Hartzell and Hoddie, Walter, Mattes and Savun, as well as Selway and Templeman, all focus on testing the impact of power-sharing on long-term peace. Norris’ 2002 analysis examines the influence of proportional representation electoral systems on the difference between majority and minority support for their political systems. Her 2008 contribution evaluates the impact of power-sharing on democracy. The findings by Powell published in 1982 include quantitative measurement of the impacts of PR on executive stability and deaths by violence.72 None of these analyses use the same cases as this book’s quantitative tests but many involve divided societies. Hartzell, Hoddie, Walter, Mattes, and Savun all use cases that involve divided societies because they represent the conditions present at the conclusions of civil wars. However, their cases constitute agreed upon plans, rather than countries’ fulfillment and use of power-sharing institutions and mechanisms. The cases used by Selway and Templeman are diverse, but not necessarily comprised of potentially antagonistic groups, and the governance mechanisms they test are operational.73 For Norris’ 2008 publication, she performs additional quantitative analysis of a subset of cases found to be “multiethnic (plural) societies, [as] measured by the Alesina index of ethnic fractionalization.”74 While these cases are plural according to the definition incorporated into the Alesina index, this definition of plural is different from that used by Lijphart regarding consociation.75 The most significant difference between these analyses and mine involves their definitions of power-sharing. Hartzell and Hoddie’s analyses published in 2003 and 2007 use a composite power-sharing 72 Powell

(1982, pp. 63, 64). and Templeman (2012, pp. 1551, 1554). 74 Norris (2008, p. 32). 75 & The cases that are quantitatively analyzed were compiled through reference to the MAR dataset, rather than datasets constructed through reference to fractionalization data. Evidence supporting this decision is provided on the website. 73 Selway

2003 and 2007

2005

Caroline Hartzell and Matthew Hoddie

Caroline Hartzell and Matthew Hoddie

The agreements that settled 38 civil wars between 1945 and 1998 (Timeseries cross-sectional data)

Agreements to end civil wars 1945–1999 (2 sets of cases: (1) 38 full negotiated settlements; (2) 49 negotiated agreements, including those settlements and 11 negotiated truces) (Timeseries cross-sectional data)

Publication Cases year

Authors

Impact of power-sharing on lasting peace

Impact of power-sharing on lasting peace

Issue explored

Comparison with my findings

(continued)

They define power-sharing differently from Lijphart’s definition which is operationalized here and also differently from that used for quantitative analysis by Pippa Norris. Their definition includes four components: political, territorial, military, and economic. The political and territorial elements include “electoral,… administrative,… [and] executive proportional representation,” as well as division of political power through federalism, confederalism, and/or regional autonomy. These elements are related to the components of consociation as they are quantitatively analyzed here but Hartzell and Hoddie tested their elements of power-sharing by combining them in a composite measure. Consociational components are tested here separately. In addition, their indicator of “executive proportional representation” is quite different from Lijphart’s “grand coalition” They find that politi- They define power-sharing differently from Lijphart’s definical power-sharing “at tion, which is operationalized here, and also differently from least in isolation, does that used for quantitative analysis by Pippa Norris. Their not have demondefinition includes four components: political, territorial, milstrated ability to assist itary, and economic. To arrive at the findings presented in this in the transition from 2005 publication, Hartzell and Hoddie tested these four sets anarchic civil war to of phenomena separately. Elements of power-sharing which a peaceful and funcare tested separately by myself are combined in their variable tioning democratic entitled: “Central power-sharing institutions.” In addition, polity” the definitions they use for crucial components of this variable differ significantly from their counterparts used here.

They find that “the prospects for peace remaining intact are enhanced with each additional power-sharing aspect included as part of the settlement to end the war”

Main findings

Table 6.1  Other quantitative treatments of phenomena similar to consociation

162  B. B. KELLY

2009

2002

Michaela Mattes and Burcu Savun

Pippa Norris

12 states varying with regard to democracy, socioeconomic development, and electoral systems (Cross-sectional national election survey data from the 1996–1998 Comparative Study of Electoral Systems)

Agreements for 46 civil war cases, from 1945 to 2005 (Time-series cross-sectional data)

Publication Cases year

Authors

Table 6.1  (continued)

Impact of proportional representation on relative majority– minority differences in support for these states’ political systems

Impact of “fear-reducing and cost-increasing provisions” on lasting peace following civil wars

Issue explored

Comparison with my findings

(continued)

Mattes and Savun’s results can be compared easily to those found by Hartzell and Hoddie for their 2005 chapter, since both studies share many methodological conditions and they each analyze these four elements of power-sharing as individual independent variables. In contrast to Mattes and Savun’s results, Hartzell and Hoddie found that there was no statistically significant relationship between this political element and long-term peace but there was a relationship between such peace and the territorial and military independent variables. However, like Mattes and Savun, Walter finds a statistically significant relationship between political pacts and implementation of civil war peace pacts (although the p-values for both findings were p