Referendums and Ethnic Conflict [Second Edition] 9780812298581

This updated and revised edition of Referendums and Ethnic Conflict features recent referendums, including Scotland (201

187 93 14MB

English Pages 224 [208] Year 2022

Report DMCA / Copyright


Polecaj historie

Referendums and Ethnic Conflict [Second Edition]

Citation preview

Referendums and Ethnic Conflict


Matt Qvortrup


 Copyright © 2022 University of Pennsylvania Press All rights reserved. Except for brief quotations used for purposes of review or scholarly citation, none of this book may be reproduced in any form by any means without written permission from the publisher. Published by University of Pennsylvania Press Philadelphia, Pennsylvania 19104-4112 www​.­upenn​.­edu​/­pennpress Printed in the United States of Amer­i­ca on acid-­free paper 10 9 8 7 6 5 4 3 2 1 Hardcover ISBN 9780812253993 Paperback ISBN 9780812225266 Ebook ISBN 9780812298581 A catalogue record for this book is available from the Library of Congress.

 Kwa malkia wangu

 T he existence of a nation is a daily plebiscite. —­Ernest Renan, Qu’est-ce qu’une nation? Lex est quod populus iubet atque constituit. [The law is what the ­people have agreed to and established.] — ­Gaius, Institutiones A Citizen in general is someone who is capable of being a ruler and a subject. —­Aristotle, Politics


Preface to the Second Edition


Introduction 1 1. The History and Logic of Ethnonational Referendums, 1791–1945 17 2. Difference-­Managing Referendums


3. Secession and Partition


4. In­de­pen­dence Referendums in International and Constitutional Law


5. Right-­Sizing Referendums


6. Difference-­Eliminating Referendums: E Pluribus Unum?


7. EU Referendums: Nationalism and the Politics of Supranational Integration


8. Regulation of Ethnonational Referendums: A Comparative Overview 134 Conclusion. Identity Politics and Referendums: Patterns and Tendencies


Appendix A. A ­ ctual and Predicted Outcome of Referendums, 1972–2016 163

viii Contents

Appendix B. Ballot Questions


Bibliography 171 Index 187 Acknowl­edgments




his book is about identity politics and referendums. When the first edition of this study was written, the former term was almost unheard of, and referendums on ­these ­matters were in the main only the concern of anoraks such as yours truly. This changed. The rhe­toric of ethnic identity combined with referendums on nationalist issues led to the Brexit vote in the United Kingdom in 2018, as well as votes on in­de­pen­dence in, among other, places Scotland (2014) and Catalonia (2017). Not all of t­hese cases w ­ ere driven by the same ideology. The ethnic nationalism of, respectively, Donald Trump in Amer­i­ca and Boris Johnson in the United Kingdom is miles apart from the civic nationalism espoused by Scotland’s Nicola Sturgeon, though perhaps not that dif­fer­ent from the nationalism of the Junts pel Sí party, which campaigns for Catalonian in­de­pen­dence from Spain. But, in any case, both types of nationalism are indicative of a growth of identity politics, a kind of politics that sees cultural and historical attachment as being more impor­tant than economic issues. In recent years, po­liti­cal economists have coined the term “analytical narratives,” that is an analy­sis that “combines analytic tools that are commonly employed in economics and po­liti­cal science with the narrative form, which is more commonly employed in history” (Bates et al. 1998). Identity politics—of the sort that gave rise to in­de­pen­dence referendums (including the Brexit vote)—­can be analyzed using this approach (see also the early study by McLean 2003). As I have argued elsewhere, the Brexit vote was an example of what economists call an inelastic good (Qvortrup 2016). Viewed from the perspective of economic theory, the Brexit referendum was thus a choice between economic security and sovereignty, and that is true for perhaps all referendums on national issues and self-­determination. Each of t­ hese two options, respectively elastic and inelastic goods, have what economists call dif­fer­ent demand curves. The price of each had implications for the voter’s decision. According to the classic definition as provided


Preface to the Second Edition

by the neoclassical economist Alfred Marshall from his Princi­ples of Economics (1890), We may say generally: the elasticity (or responsiveness) of demand in a market is g­ reat or small according as the amount demanded increases much or ­little for a given fall in price, and diminishes much or l­ ittle for a given rise in price. . . . ​[T]he only universal law as to a person’s desire for a commodity is that it diminishes . . . ​but this diminution may be slow or rapid. If it is slow . . . ​a small fall in price w ­ ill cause a comparatively large increase in his purchases. But if it is rapid, a small fall in price ­will cause only a very small increase in his purchases. In the former case . . . ​t he elasticity of his wants, we may say, is g­ reat. In the latter case . . . ​t he elasticity of his demand is small. (Marshall 1890: 166) Put differently, an elastic good, following Marshall’s definition, is one where a slight change in price leads to a big change in the quantity demanded. Conversely, an inelastic good is one where even a considerable change in price has a negligible effect on the quantity demanded. The contention (or hypothesis) h ­ ere, then, is that the two options—­respectively, “Brexit” and “Bremain”—­ were, respectively, inelastic and elastic goods. And the same, more generally, goes for cases of identity politics versus bread-­and-­butter issues. Applied to the case of Brexit, the main argument proposed by t­ hose who wanted to leave the Eu­ro­pean Union (EU) was sovereignty: “It was clear from the start that we wanted a ­people’s campaign. We wanted to focus on sovereignty. ‘Take back control’, ‘I want my country back’, that [sic] sort of slogans. They are priceless. Who does not want to get his [or her] country back?” This is a quote. The man who uttered ­t hese words was Gerry Gunster, an American po­liti­cal con­sul­tant who ran the campaign for (personal communication with the author 6 July 2016). His argument—­t hough he did not put it in the jargon of economic theory—­was that some voters are willing to accept virtually any increase in economic costs as long as they can “take back control.” In his own words, sovereignty is “priceless.” In economic terms, sovereignty can—­following this logic—be seen as an inelastic good. As in the aforementioned definition, a good is inelastic when a change in the price leads to a negligible change in demand for it. For example, a chain smoker ­will keep buying cigarettes almost no ­matter how expensive a pack of Benson & Hedges or Marlboro becomes. And a commit-

Preface to the Second Edition xi

ted football or baseball fan ­will, no doubt, continue to buy a season ticket even though the price rises exorbitantly. So, too, with Brexit (and in­de­pen­dence): even an increase in the price— as predicted in countless reports by the International Monetary Fund, HM Trea­sury (the British ministry of finance) and o ­ thers—­would only lead to a negligible change in the “demand” for “taking back control”—or so it turned out. Of course, we cannot prove the logic of this analytical narrative in the strictest sense of the word. But the logic was certainly not lost on Nigel Farage. The stockbroker turned politician told the Euroskeptic Daily Express: “The wellbeing of t­hose living and working in our country m ­ atters to me more than GDP figures” (Daily Express 17 June 2016). Farage may not have considered his prediction in the jargon of economic theory, but the logic was clear. An increase in the cost of living—or, as voters saw it, a change in abstract macroeconomic figures—­would be a small price to pay for “freedom” (Farage’s word for leaving the EU). The British government’s document HM Trea­sury Analy­sis: The Immediate Economic Impact of Leaving the EU had claimed that each voter—­according to econometric estimates—­would be £4,000 worse off in 2030 if Britain voted to leave the EU, and “a vote to leave would cause a profound economic shock, creating instability and uncertainty, which would be compounded by the complex and interdependent negotiations that would follow” (HM Trea­sury 2016: 1). However, the Trea­sury’s prediction was, arguably, too abstract for many voters—­and, more to the point—­many voters w ­ ere receptive to and convinced by the “£350 million a week” claim (the suggestion by the “leave” campaign that Britain paid this amount to the EU ­every week). ­These voters believed the claim that this money could be spent directly on public ser­v ices like the National Health Ser­v ice. It was a strong emotional argument, and one that only becomes stronger if analyzed in the light of economic theory. It was clear that organizers of the campaign to stay in Eu­rope wanted to push the economic argument. And they had good reasons for ­doing so. Opinion polls suggested that many voters—­indeed, a majority—­believed that Britain would benefit eco­nom­ically from staying as a member of the Eu­ro­ pean Union (Financial Times 3 May 2016: A3). The prob­lem for the “remain” side was, eco­nom­ically speaking, that the “welfare” argument was a classic example of an elastic good. A small change in the price of Bremain and the demand for staying in Eu­rope would go down.


Preface to the Second Edition

Nevertheless, the economic argument—­and the perceived benefits of staying in the EU—­presented a prob­lem for organizers of the leave campaign. Without any economist of note on their side they faced an uphill strug­gle. They resolved this by resorting to the aforementioned claim that UK taxpayers transferred £350 million a week to the Eu­ro­pean Union. The government and HM Treasury—it should be noted—­dismissed this claim, as one would expect. But the assertion was also dismissed by the UK Statistics Authority, whose chair, Sir Andrew Dilnot, in unequivocal terms denounced the assertion. The Trea­sury Committee in the House of Commons also condemned the claim that Britain was transferring this amount of money to the EU, noting that the “real” figure was closer to £180 million a week. The prob­lem for the remain side was that many voters accepted the “£350 million a week” figure or at least the gist of the argument. “Voters totally internalized the argument when we used it in focus groups. When they ­were told that the figure was inaccurate, they would say, ‘yes, but it is still a substantial amount,’ and we would get nowhere,” said Lord Cooper, Prime Minister David Cameron’s pollster (Lord Cooper, interviewed by the author, 6 July 2016). In economic terms, as soon as the cost of Brexit was believed to be rather minimal—or perhaps even positive—­the demand for leaving the EU went up. The remain campaign had done rather well in the first week a­ fter David Cameron called the referendum; they had been able to control the agenda and had argued that the price of leaving the EU was high. Once this claim had been effectively disputed by leave campaigners like Boris Johnson, Michael Gove, and Nigel Farage, the support for leave increased, and demand for continued membership dropped significantly. In economic terms, Bremain was an elastic good: the perceived change in price had an immediate impact on the demand. When the price of EU membership was claimed to be £350 million a week, demand for EU membership fell. The fact that this claim was acknowledged to be fanciful by Nigel Farage within hours of the referendum makes ­little difference. Esse est percipi—­“what is, is what is perceived,” the phi­los­o­pher George Berkeley (1685–1753) famously noted. Or, in Lord Cooper’s more downbeat assessment: “The other side had the best lines and the best lies” (Lord Cooper, interview with the author, 8 July 2016). Of course, in other referendums, the position was dif­fer­ent. The EU was an emotive issue and a very inelastic good. Scottish in­de­pen­dence was not an inelastic good, except for some hardliners. It was recognized that the “price” was very high, and the famously thrifty Scots w ­ ere more concerned with

Preface to the Second Edition xiii

pennies and pounds than with “putting a kilt on t­ hings,” as the local phrase goes. So much for analytical narratives. And now back to the second edition of the book. Writing a­ fter the very high-­profile referendums (and at a time when identity politics is the dominant theme in everyday debates) this new edition of this book analyzes how national identity and democracy have combined and clashed in the past 230 years. The book is unashamedly a scholarly treatise. It seeks to establish positive and scientific theories of why vari­ous referendums on identity politics are held—­but it also analyzes when they are lost and won. In Chapter 7 of the book I develop a statistical model that can predict the outcome of Eu­ro­ pean referendums. This model, I might add, was developed before the 2016 Brexit referendum, which I predicted in an article published four months prior to the vote (Qvortrup 2016). However, this is not just a po­liti­cal science book. Being a l­ awyer by training, as well as a po­liti­cal scientist by profession, I could not resist delving into the finer points of regulation and litigation. Hence the book contains chapters on when referendums are permissible in domestic and international law. Furthermore, the book has an extended section in Chapter 8 on the regulation of referendums. Unlike the first edition, the book looks at how online campaigning and the use of algorithms have been abused to target voters (e.g., during the Brexit referendum). Some have seen it as almost inevitable that we fall prey to t­ hese tactics. As the chapter on regulation below shows, this is not the case. Many countries regulate and limit the use of online targeting. Democracy needs regulation to remain fair. This book shows how. Referendums and identity politics can be dangerous bedfellows. And referendums are far from unproblematic. They need to be used with care. But so do many other useful devices from cars to chainsaws. Like many other valuable t­hings, referendums require understanding before they are used. This book contributes to this understanding. And can, I hope—­directly or indirectly—­help improve the use of this demo­cratic device. Nationalism and the craving for self-­determination are not universally loved, but they are not likely to go away. ­W hether you condemn or endorse this trend, it is imperative to understand it.



he core prob­lem in po­liti­cal theory is that fundamental and equally axiomatic princi­ples often collide. A paradox can almost be defined as a clash of two equally incontestable maxims of truth. Two such “truths” are (1) that each nation has a right to determine its own affairs and (2) that the majority has a right to govern. Admittedly t­ hese “rights” are tempered by the recognition that no nation and no majority may ­ride roughshod over minorities. But this caveat notwithstanding, national self-­determination and majority rule are princi­ples to which few fundamentally object. Indeed, defending the reverse positions would appear po­liti­cally absurd. But the prob­ lem is that the two princi­ples often are incompatible. To understand why, it might be useful to consider a distinction used in ancient Greek. The Greeks made a distinction between the ­people as a nation (ethnos) and the ­people as a body of citizens (demos). In the classical city-­state—or polis—­the two ­were congruent, and in some present-­day nation-­states, such as Norway and Luxembourg, the same is broadly true. But more often than not, the two concepts are in conflict. To take an example, a small-­town En­glish politician, the Conservative councilor Rob McKella from Corby in Northamptonshire, believes that ­people in ­England should be given a right to vote on Scottish in­de­pen­dence. ­After all, he argues, the voters are citizens in the United Kingdom and collectively constitute the demos. However, most p ­ eople in Scotland, by contrast, believe that only ­people living north of the border should be allowed to vote as ­these p ­ eople—­perhaps alongside Scots living in the diaspora—­constitute the ethnos and hence have a right to self-­determination. As ­will come to be obvious, ­t hese two worldviews, both based on solid arguments, do not combine. That the local politician from the Midlands in the United Kingdom took an interest in this was not accidental. Corby was known as L ­ ittle Scotland, and in the 1930s the former steel town saw a massive influx of Scottish workers.

2 Introduction

This book is about this conflict between the ethnos and the demos, and about the prob­lems that are raised when solutions to ethnic and national issues and conflicts are sought through referendums. Its main focus is on determining when dif­fer­ent kinds of referendums on ethnonational issues occur and also on determining if they lead to exacerbation of conflict or the opposite—if balloting can stop bullets. Referendums have often been perceived to be incompatible with nationalism. William Sumner Maine—­a conservative writer from the end of the Victorian age—­once mused that “democracies are quite para­lyzed by the plea of nationality. ­There is no more effective way of attacking them than by admitting the right of the majority to govern but denying that the majority so entitled is the par­t ic­u ­lar majority which claims the right” (Maine 1897: 88). This book looks at ­t hese conflicts through a comprehensive study of all the referendums held on ethnic and nationalist issues from the French Revolution to the 2020 referendum on in­de­pen­dence for New Caledonia. It is a controversial topic. Some scholars have supported ­t hese referendums on the basis of philosophical conviction and b ­ ecause they believe that they confer legitimacy upon decisions made by elites. Jürgen Habermas, for example, took this view. In light of resentment following new bound­aries ­after the cold war he found that referendums on sovereignty issues—­given certain safeguards—­could be “a way of proceeding which permitted a broader discussion and opinion formation as well as a more extensive—­and, above all, better prepared—­participation,” which would give the voters “the eventual responsibility for the pro­cess” (Habermas 1996: 12). If the ­people w ­ ere given the responsibility through referendums, they would not be able to complain ­later on, as “it would have been the p ­ eople’s own m ­ istake that they would have had to cope with” (Habermas 1996: 12). ­These issues are in­ter­est­ing and impor­tant from a philosophical point of view, but they are ultimately issues for po­liti­cal theory and not the primary concern of the comparative and empirical po­liti­cal scientist. Hence ­t hese normative issues are not the focus of this study. The questions I look at are positive issues of when and why referendums on vari­ous national and ethnic issues occur. ­These polls have played an impor­tant role in attempts to resolve ethnic conflicts for centuries. But it is fair to say that scholars of ethnic and national conflict—as well as po­liti­cal scientists—­have had reservations about ­these plebiscites and referendums. Michael Gallagher, writing about the experi-

Introduction 3

ence in Eu­rope, concluded that “the referendum is least useful if applied to an issue that runs along the lines of a major cleavage in society” (Gallagher 1996: 246), and recently Jonathan Wheatley wrote about the “disruptive potential of direct democracy in deeply divided socie­ties” (Wheatley 2012: 64). While critical, Wheatley was not as dismissive of referendums as Roger Mac Ginty, who noted, “The principal prob­lem with referendums in situations of profound ethnic conflict is that they are zero-­sum, creating winners and losers. ­Simple majoritarian devices do ­little to help manage the complexity of conflict. Instead, they validate the position of one side and reject that of another. Often, they do ­little other than delimit and quantify division” (Mac Ginty 2003: 3). This interpretation may have been correct in the case of the 1973 Border Poll in Northern Ireland (Osborne 1982: 154) and, indeed, in the case of many of the referendums held in former Yugo­slavia in the 1990s (Terret 2000). But Mac Ginty’s conclusion that “the blunt reductionism of a referendum (for example the perception that a conflict is about a line on a map and ­little ­else)” (Mac Ginty 2003: 3) is perhaps a l­ ittle bit cavalier and perhaps ignores the fact that the Schleswigian conflict, which caused two wars between Denmark and Germany, was resolved by a referendum (Laponce 2010: 24). Indeed, Sarah Wambaugh noted in her much-­quoted study Plebiscites Since the World War that “the plebiscite was so fair and excellently administered that the Schleswig question, which caused three wars in the 19th ­Century and rent of councils of Eu­rope for some seventy years, has ceased to exist” (Wambaugh 1933: 98). A referendum is admittedly unlikely to work on its own, and referendums have been followed by vio­lence (the case of East Timor in 1999 comes to mind)—­and, indeed, even the Brexit referendum in Britain in 2016 was marred by the murder of an MP! But this does not prove that the referendum is to blame. Indeed, t­here are examples of the opposite, including the peaceful poll in Eritrea in 1993 (on secession from Ethiopia). Just as we can find contrasting empirical examples, we can also find contrasting academic assessments. Some, such as Ben Reilly, maintain that “despite hollow claims that the ‘­will of the p ­ eople’ must prevail, it is only the most obtuse interpretation that could recommend building peace this way” (B. Reilly 2003: 184), while ­others, notably Arend Lijphart, take the view that “the potential of calling the referendum . . . ​is a strong stimulus for the majority to be heedful of minority views” (Lijphart 1999: 231).

4 Introduction

We have a welter of empirical evidence, but so far ­there have been no systematic study, no developing of hypotheses, and no general theories. One of the aims of this book is to alter this state of affairs. It is surprising that this task has not been undertaken before. The ethnonational referendum has been around for a long time, and it is still popu­lar. From the Swiss referendum on in­de­pen­dence from Bo­na­parte’s France in 1802 (and before that Avignon’s incorporation into France in 1791) to more recent referendums on power sharing in Northern Ireland in 1998, autonomy for East Timor in 1999, and statehood for Puerto Rico in 2020, voters have been called upon to decide ethnic issues. Th ­ ere have been more than two hundred of t­ hese referendums, and their outcomes have de­cided the fates of p ­ eople and p ­ eoples from Malta to Micronesia and from Mongolia and Montenegro to Montreal. Some have been held in despotic—­and indeed totalitarian—­regimes (like the Anschluss Referendum in 1938  in Austria), o ­ thers have been held in countries with impeccable demo­cratic rec­ords such as Switzerland (Jura), Denmark (Greenland), and Canada (Nunavut). This geo­graph­i­cal spread, and the fact that the same institution has been used across dif­fer­ent countries and by very dif­fer­ent types of regimes, make t­ hese referendums both puzzling and challenging. Although they are never far from the media news stream, ­t hese referendums (I reserve the word plebiscite for votes held in autocratic states) have rarely received systematic treatment. Mentioned in passing in one of my previous books (Qvortrup 2003) and in Larry LeDuc’s The Politics of Direct Democracy (2003), as an aside in David Altman’s Direct Democracy Worldwide (2011), as an afterthought in Markku Suksi’s Bringing in the P ­ eople (1994), and in sections of John T. Rouke, Richard P. Hiskes, and Cyrus Ernesto Zirakzadeh’s Direct Democracy and International Politics (1992), referendums on national issues and ethnic conflict have been relatively marginalized in scholarly research. The only existing monographs on the subject are Lawrence T. Farley’s fairly empirical and theory-­free Plebiscites and Sovereignty (1986) and, much ­earlier, Sarah Wambaugh’s Plebiscites Since the World War (1933). Before that Johannes Mattern’s doctoral thesis, The Employment of the Plebiscite in the Determination of Sovereignty (1921), dealt with some l­egal and historical aspects of referendums on national self-­determination and related issues. At the time when the first edition of this book was all but finished, Jean Laponce’s Le Référendum de souveraineté: Comparaisons was

Introduction 5

published in French (2010). While the late Professor Laponce’s book dealt with some of the same issues, my book differed in taking a more positivist approach that is more devoted to testing hypotheses and finding general and statistical patterns, whereas Le Référendum de souveraineté is a more descriptive study in the very best sense of this word. The two books therefore are complements rather than competitors. I hope that my book ­will spur other scholars on to consider t­ hese issues and that it w ­ ill be superseded by works that further elaborate on a theme that has received too l­ ittle treatment before.

Methodology and Testing of Hypotheses Many social scientists have been strangely fascinated by the physical sciences—­t hough they have not always practiced what they have preached. Ever since at least Thomas Hobbes’s Leviathan and Baruch Spinoza’s Ethics it has frequently been the ambition of po­l iti­c al and philosophical writers to analyze public affairs with the stringency of Euclidian geometry or, to paraphrase Spinoza’s Ethics, to conduct po­l iti­c al science in the manner of polıtica more geometrico demonstrate, “demonstrated by the geometrical method.”

The Competition Proximity Model of Ethnonational Referendums

In recent years in par­tic­u­lar it has become de rigueur to develop formal models to pre­sent theoretical arguments, which subsequently can be tested empirically (Aldrich, Alt, and Lupia 2010), and this tendency has begun to influence scholars specializing in direct democracy (e.g., Setälä 1999). However, neither Hobbes nor Spinoza slavishly followed the razor-­sharp logic of strict deductive reasoning. The geometric method that they both espoused provided “mathematical illustrations” rather than proofs in the Euclidian sense (Jesseph 1993). This book does the same. Formal models are proposed and even subjected to quantitative tests. But the statistical arguments do not stand alone. Throughout the book numerical data are contrasted with and complemented by case studies and historical narratives. Together they provide a mosaic that point

6 Introduction

t­oward general tendencies and trends. The conclusions drawn in the study are inspired by and seen through the prism of formal models; however, the truthfulness of t­ hese is determined not solely by statistical data but also by circumstantial and even anecdotal evidence, something that make so-­called analytical narratives particularly useful. This approach may appear defeatist to the positivist purist, but to the realist living in the practical world of politics the departure from pure mathematical models is justified by the greater ability to draw general conclusions. ­These caveats notwithstanding, this book loosely follows the positivist approach by proposing that the decision to hold referendums on ethnic and national issues can be explained by a formal model that sees the likelihood (probability) of a referendum being held as the result of an inverse relationship between po­liti­c al competition and support for the proposed policy among the constituents. The formal model looks as follows: C 2 i=1 ( I − P ) m i n

Pref = ∑

According to this competition proximity model, the probability, Pref, that an actor, i, ­will submit a national or ethnic issue to a referendum depends on the relationship between the competition, C, the actor is facing, and the squared distance between the actor’s preference point Pi, and the preference point of the median voter Im. If an actor is facing considerable competition (large value for C), and if the actor’s preferred policy is a popu­lar one (the value of [Im—­Pi]2 is small), then holding a referendum—­especially if this is opposed by his competitors—is likely to give the initiator a boost and strengthen his or her legitimacy. The bold hypothesis in this book is that most referendums on ethnic and national issues are held as a result of the logic under­lying this model. In other words, if the value for C is large (something that, admittedly, is difficult to mea­sure), and if the preferred policy is a popu­lar one (the value of (Im—­Pi)2 is small), then the probability of a referendum w ­ ill be larger than one. Formally speaking, n

Pref = ∑



(Im − Pi )2


Of course, this model does not—­and cannot—­stand alone and throughout w ­ ill be contrasted with other hypotheses. I return to t­ hese shortly, but

Introduction 7

before even starting the analy­sis it is useful to outline the methodological considerations under­lying this study.

The Scientific Method for the Social Sciences

At the most basic level, academic or scholarly research differs from other endeavors ­because of its method and as a result of the more stringent criteria applied. Scholarly research requires at the most basic that we have: 1. clear definitions of the concepts we are studying; 2. criteria for when we can make firm conclusions; and 3. considerations of when we can compare dif­fer­ent instances and facts. In this section I set out to answer t­ hese questions in an operational way with a view to addressing the prob­lems in this book. This introduction is not intended to provide an analy­sis of the more philosophical issues pertaining to social science analy­sis. It is merely intended to offer an overview and a statement of the methodology that I accept a priori. “The ultimate goal of a positive science is the development of a ‘theory’ or, ‘hypothesis’ that yields valid and meaningful (i.e., not truistic) predictions about phenomena not yet observed,” noted Milton Friedman in “The Methodology of Positive Economics” (1966: 7). Developing t­hese theories means, in this case, to foresee conditions ­under which the dif­fer­ent kinds of ethnonationalist referendums are likely to occur. For the pre­sent purposes I take Karl Popper’s falsificationist theory as the point of departure. Karl Popper’s argument was based on s­ imple propositional logic. In his early study Logik der Forschung (­later published in En­glish as Logic of Scientific Discovery in 1959), Popper challenged the view espoused by the Vienna Circle, namely that scientific statements must obey two conditions: (1) have semantic reference and (2) be of a so-­called modus ponens form. That is—­ formally speaking—­t hey must be of the form “If p then q, p, therefore q.” However, as Popper pointed out—­with reference to David Hume—­while statements should have semantic reference, they should be of modus tollens form, that is, be of the form “If p then q, not q, then not p.” Popper further stressed that the antecedent should be a universal statement (i.e., “all x are y”) and the consequent should be a singular statement (e.g., of the form “this x is y”).

8 Introduction

I can explain this by a practical example from po­liti­cal science, namely Duverger’s Law (Duverger 1972: 23). The argument, developed by the French po­liti­cal scientist Maurice Duverger, can be summed up as follows: All countries that have first-­past-­t he-­post (FPTP) systems have two-­party systems. Canada (which uses FPTP) does not have a two-­party system. Not all countries that use FPTP have two-­party systems. This hypothesis is true for many countries, such as the United States, Jamaica, Botswana, and so on. But ­t here are cases that break the “law.” Box I.1. Formal Example of Hy­po­thet­i­cal Syllogism and Auxiliary Hypothesis ­ nder normal circumstances the formula for testing hypotheses is a modus U tollendo tollens. It can formally be stated as follows: P → Q −­Q −P With the addition of an auxiliary hypothesis (where R denotes the auxiliary hypothesis), we get P (&R) → Q −­Q −­P (&R) Source: Based on Popper (1959).

Now the prob­lem with this statement is that it strictu sensu should be considered as falsified, as a first-­past-­the-­post country like the aforementioned Canada—to take but one example (India is another one)—­has a multiparty system. Clearly t­ here is something to Duverger’s Law, but a c­ ouple of exceptions exist. What do we do? To proceed one might invoke auxiliary hypotheses. One might thus restate the theory thus: countries with FPTP electoral systems and without strong regionally based po­l iti­c al movements have two-­party

Introduction 9

systems. However, to use such auxiliary hypotheses we must require that such caveats are universal and not ad hoc. Thus, the existence of a par­tic­u­lar (nonuniversal) f­actor—­such as, for example, Quebecois nationalism—­will not be regarded as acceptable as we are striving for universal laws. In other words, and to quote a scholar from another field, “An auxiliary hypothesis ­ought to be testable in­de­pen­dently of the par­tic­u­lar prob­lem it is introduced to solve, in­de­pen­dently of the theory it is designed to save (e.g. the evidence for the existence of Neptune is in­de­pen­dent of the anomalies in Uranus’s orbit)” (Kitcher 1982: 46). Moreover, we need to add that we should not simply reject a theory once one single case falsifies it and such an approach is—­moreover—­not consistent with Popper’s original formulation of the theory of falsification. It is beyond the scope of this book to delve deeply into the finer points of the theories of science. Suffice it to say that Popper, in his Logic of Scientific Discovery, actually stresses that a theory should be superseded by a rival theory only if the new theory can explain the ­t hings the former theory could explain as well as the t­ hings it could not explain (Popper 1959: 57). But d ­ oing this alone does not solve the prob­lem. Often in po­liti­cal science we are faced with a rather simplistic approach to causal inference. We assume a generality, and then proceed to prove it. Popper notwithstanding, we are typically dealing with statements of the form “If p then q, p therefore q” (or, formally speaking, what logicians call a modus ponens). This form of statement w ­ ill be transformed into hypotheses based on the formular above, that is, n

Pref = ∑



(Im − Pi )2

and tested statistically. However, t­ here is no suggestion that merely a high r2 and high levels of statistical significance are sufficient to render the model plausible. While some analyses of the policy effects of electoral systems may have suggested this (Lijphart 1999), it is rarely the case that ­t here is a direct, quantifiable, and causal relationship between variables. A high r 2 may indicate an under­lying causal relation, but the explanation is key. For example, it is pos­si­ble that t­here is a correlation between the number of referendums and a par­tic­u­lar system’s degree of democ­ratization (e.g., as mea­sured by the Polity IV scores). Polity scores, as the reader may know, represent a mea­ sure of levels of democracy, with −10 being the worst and + 10 being the best.

10 Introduction

The worst are hereditary monarchies and the best are fully competitive democracies. The mea­sure is based on six component mea­sures that rec­ ord key qualities of executive recruitment, the existence (or not) of constraints on executive authority, and the level of po­ liti­ cal competition. Further, Polity IV mea­sures changes in the institutionalized qualities of governing authority. As is well known, a correlation does not establish a fact. A high level of statistical association does not in itself imply or indicate that ­t here is a causal relationship between referendums and dictatorship. Rather this seeming correlation might be a result of intervening and far more impor­tant, and indeed in­ter­est­ing, variables. For example, it has been argued that ­t here is a correlation between economic growth and first-­past-­t he-­post electoral systems. This is not, however, b ­ ecause this electoral system in itself leads to growth, but more likely a result of an intervening variable, namely that single-­ party governments (which are more common in countries with first-­past-­t he-­ post systems) are less likely to negotiate costly deals with minority parties (Marsiliani and Renström 2007), which in turn would have reduced the funds available for investment. Finding ­these regularities, however, is not just about conducting large comparative studies across dozens of cases. To fully understand the logic of the regularities (if they exist) also requires us to probe into par­tic­u­lar examples. For, to quote a prominent American comparativist, “ “If [on the one hand] statistics address questions of propensities, [then] narratives address questions of pro­cess” (Laitin 2002: 636). Or put differently, an understanding of the “laws” is aided by case studies that explain the mechanics of the causal relationships. ­Needless to say, ­there are other methodologies, and ­t here is always a danger of underestimating the complexities in the social sciences. Charles Ragin has drawn attention to what he seminally called “multiple conjunctural causation,” namely that a phenomenon can be the result of dif­fer­ent ­causes (Ragin 2014: 42). That is especially true when analyzing the etiology of referendums, which are never just the result of one f­actor—­say strategic considerations—­but can be the consequence of dif­f er­ent ­causes—­and frequently a combination of ­t hese. To delve into their origin, we need qualitative data but we also need to put the quantitative finding into contextual perspective, and this requires an understanding of the past. As Charles Tilly has written, “­every significant po­liti­cal phenomenon lives in history, and requires historically grounded analy­sis for its explanation. Po­liti­cal scientists ignore historical context at their peril” (Tilly 2009: 536). To acquire this

Introduction 11

knowledge means that we need to look at macrohistorical trends, but we must also complement the apparent general patterns with illustrative (or contrasting) case studies, as t­ hese “may allow one to peer into the box of causality to locate the immediate f­actors lying between some structural cause and its purported effect” (Gerring 2009: 1146). Hence, in addition to the theoretically based analy­sis, the book contains descriptive, yet analytical, case studies. ­These case studies w ­ ill also help us to find special circumstances that explain anomalies. This is sometimes a complex business and suffers from the fact that we sometimes have to “feel” our way into a prob­lem. This more qualitative addendum to the more formal analy­sis has been summed up as the pro­cess of “so­cio­log­i­cal imagination,” namely “the capacity to shift from one perspective to another—­from the po­ liti­cal to the psychological; from examination of a single ­family to comparative assessment of the national bud­gets of the world; from the theological school to the military establishment; from considerations of an oil industry to studies of con­temporary poetry” (Mills 1951: 11). Getting this right is a tricky business, and ­t here is no guarantee that we get it right even though we follow the logic.

A Typology of Ethnonational Referendums

­ ese methodological rules are the bare bones of the analy­sis. But ethnonTh ational referendums are not a uniform category. The referendum on devolution in Wales in March 2011 was vastly dif­fer­ent from the referendum held on in­de­pen­dence in South Sudan a few months before. Similarly, the referendum held in the Soviet Union in March 1991 was vastly dif­fer­ent from the vote held in Saarland between Germany and France in 1955. In other words, the category “ethnonational referendums” is so broad that it might be meaningless. To paraphrase Aaron Wildavsky, if ethnonational referendums are every­t hing then maybe they are nothing (Wildavsky 1973). What we need is a set of more precise definitions, which can subsequently form part of testable hypotheses. Social science is—or o ­ ught to be—­a cumulative endeavor. The research developed by scholars provides the basis for the research undertaken by a subsequent generation. Furthermore, research in a subarea is often based on a larger framework developed for more general prob­lems. This model is generally inspired by and based upon the taxonomy developed by Brendan

12 Introduction

O’Leary and John McGarry, who distinguish between, respectively, “difference-­ managing policies” and “difference-­eliminating policies” (see, e.g., McGarry and O’Leary 1993; updated in O’Leary 2001). Using O’Leary and McGarry’s definition we can thus have referendums on: 1. Difference elimination, that is, referendums that aim at legitimizing a policy of homogenization, such as the Anschluss referendum in Austria in 1938; 2. Difference managing, that is, referendums aimed at managing ethnic or national differences, such as the referendums on devolution in Scotland, Wales, and Northern Ireland in 1997 and 1998. Yet, in order to be more specific, this model is expanded by two categories. In addition to O’Leary and McGarry’s taxonomy, we thus expand our model to include: 3. Secession referendums, that is, votes to endorse (or other­w ise) a territory’s secession from a larger entity (e.g., the referendum in Jamaica in 1963 or the referendum in Eritrea in 1991); 4. Right-­sizing referendums, that is, votes dealing with the drawing of disputed borders between countries, such as the border between Croatia and Slovenia, which was the subject of a referendum in 2010. This model can also be stated in a more logical way, namely by developing a typology of dif­fer­ent types of ethnonational referendums. Broadly speaking, we can distinguish between referendums that are initiated by politicians who take diversity as an accepted fact and want to manage ­these differences and, on the other hand, referendums initiated by politicians who do not accept diversity. The former may be categorized as “heterogenizing” referendums. The latter may be categorized as “homogenizing.” Heterogenizing referendums can be divided into “international” and “national,” and the same is true for “homogenizing” referendums. D ­ oing this we get a two-­by-­two model of four logically pos­si­ble types of ethnonational referendums (­Table I.1). For example, heterogenic referendums can be e­ ither right-­sizing referendums (international and heterogenizing), for example the Saar referendum in 1955, or they can be internal, that is, held within a single state, for example the referendum on the ­f uture of Greenland in 2009 and

Introduction 13

­Table I.1. Typology of Ethnonational Referendums International/Homogenizing

International Heterogenizing

Secession referendums E.g., Scotland (2014) or Bougainville (2019)

Right-­sizing referendums E.g., Belizean territorial dispute referendum (2019) National heterogenizing Difference-­managing referendums E.g., Puerto Rico (2020)

National homogenizing Difference-­eliminating referendums E.g., Egypt and Syria (1958)

the referendum in Wales in 2011. Homogenizing referendums can similarly be divided into internally held plebiscites (such as the poll in the Soviet Union in 1990 on maintaining Moscow control) (Sheehy 1991) and international homo­genizing votes that create a new homogeneous entity, for example, a new monoethnic state, as was the case in Norway in 1905. Of course, ­there may be referendums that do not fit into this neat categorization. We should always be mindful of social anthropologist Gregory Bateson’s observation in Steps to an Ecol­ogy of Mind that t­ here is a difference between the “maps” and the “landscape”; that is, that the theories are the maps and not the landscape (Bateson 1962: 459). Categorizations are preludes to theories but are only social constructs that serve as heuristic tools, that enable us to get a better overview of the empirical world. A group of referendums that do not fit the categories are the votes on EU integration and membership. ­These could be seen as marginal examples of difference-­managing referendums, but it is not an easy fit, and hence they ­will be treated separately; again, the theories are the maps not the landscape, and I am more interested in understanding the latter than in drawing up the former. The aim of this book is to determine when ­t hese dif­fer­ent kinds of referendums occur. Overall, I propose that referendums on ethnic and national issues are most likely to occur when the conditions of the competition proximity model outlined above are met. But this model does not and cannot stand alone. In addition to testing the competition proximity model, I develop other testable propositions. The hypotheses are the following: 1. Secession/partition referendums tend to occur following the lifting of a long-­standing imperial hegemony—­but only if t­ here is a broad-­based elite commitment to polyarchic government in the

14 Introduction

country in question (as was the case in the former Soviet republics of Estonia, Latvia, and Lithuania). 2. Right-­sizing referendums tend to occur in the wake of a major conflict or in the wake of a regime change, such as in the case of the referendums on border demarcation in the wake of World War I. 3. Difference-­eliminating referendums tend to occur in authoritarian regimes, to lend legitimacy to a policy of ethnic and national homogenization, such as in the case of the referendums or­ga­nized by Mikhail Gorbachev in 1990 and in the case of the 1958 referendum in Egypt-­Syria on an amalgamated state. 4. Difference-­managing referendums tend to occur in countries with high Polity IV scores, especially following long-­standing ethnic disagreement (as in the cases of the Canadian Charlottetown referendum in 1992, the 1998 Good Friday referendum in Northern Ireland, and the referendums on Greenland in 1980 and 2010). In addition to t­hese hypotheses, I surmise that ethnonational referendums tend to result in peaceful resolutions of conflicts only if two conditions are met: (1) elite consensus for the proposed solution and (2) international backing of (or not outright opposition to) the referendum.

Chapter Overview Chapter 1 (“The History and Logic of Ethnonational Referendums, 1791–1945”) looks at the history of ethnonational referendums up to World War II in light of the competition proximity model. The birth of demo­cratic politics in the modern sense of the word coincided with the emergence of the ideology of nationalism. It is not surprising, therefore, that referendums aimed at resolving ethnic tensions have a long history. Some of the first exercises of demo­cratic involvement ­were referendums on self-­determination held in the aftermath of the French Revolution. The first chapter pre­sents a history from the French-­ organized referendums at the beginning of the nineteenth c­ entury through the plebiscites that legitimized the Italian Risorgimento in the 1860s and the referendums on national self-­determination in the wake of World War I. Chapter  2 (“Difference-­Managing Referendums”) deals with referendums on difference management. Often referendums are used to approve policies that manage differences. In this chapter, I analyze referendums on

Introduction 15

the relations between center and periphery in federal states (e.g., the 1992 referendum in Canada). I also examine polls in quasi-­federal states (e.g., the referendums on Scottish and Welsh devolution in 1997 and the Welsh referendum in 2011). But difference-­managing referendums do not just pertain to territorial differences. The referendums on power sharing in places as dif­fer­ ent as Northern Ireland (1998) and Burundi (2005) are examples of votes held to add legitimacy to policies of (ethnic) difference management. Why ­were ­t hese polls held? When do governments feel ­t here is a need to get the voters’ seal of approval for policies of difference management? Using a large­N analy­sis and selected case studies, the chapter seeks to establish a general pattern of difference-­managing referendum occurrence and relate ­t hese to the competition proximity model. In addition to analyzing the genesis of ­t hese referendums, the chapter also outlines a model that explains what determines the outcome of ­t hese referendums in demo­cratic countries. Chapter 3 (“Secession and Partition”) looks at “po­liti­cal divorce settlements.” ­These w ­ ere the first referendums to be called by this name and w ­ ere concerned with secession and partitions. Since then scores of referendums have been held on ­whether territories should secede. Sometimes ­t hese have resulted in peaceful resolutions of long-­standing conflicts (e.g., in the case of Norway in 1905, where a referendum was held on secession following negotiations between Norway and Sweden). But more often than not, t­ hese polls have resulted in exacerbated conflicts and deepening of tensions (as was the case in Bosnia-­Herzegovina and East Timor). Drawing on the lit­er­a­ture on secessions, this chapter uses case studies from Schleswig in 1920, the Faroe Islands in 1946, and South Sudan in 2011 to analyze when referendums on secessions and partitions can resolve issues. The chapter also looks at the conditions for using referendums to resolve conflicts peacefully. Chapter 4 (“In­de­pen­dence Referendums in International and Constitutional Law”) deals with the legality of referendums in constitutional law. Often it is believed that a local or state government is simply entitled to hold a referendum on in­de­pen­dence without further ado, but—as the chapter shows—­t his is, legally speaking, rarely the case. Right-­sizing referendums, typically referendums on the drawing of borders between countries that have irredentist populations in neighboring countries, such as Poles in Germany in the aftermath of World War I, are dealt with in Chapter 5 (“Right-­Sizing Referendums”). Chapter 6 (“Difference-­Eliminating Referendums”) considers referendums to promote homogenization (difference-­eliminating referendums). ­Whether

16 Introduction

they are conducted in demo­cratic or nondemo­cratic regimes, ­t hese votes are held to secure the legitimacy of controversial policies. In divided socie­ties, referendums are often held to approve policies aimed at assimilating and integrating groups. Using selected case studies, the chapter seeks to establish a general pattern of difference-­eliminating referendum occurrence. The chapter also includes a theoretical section on the justification for difference-­ eliminating referendums as outlined by the controversial German scholar Carl Schmitt. Chapter 7 (“EU Referendums”) analyzes referendums on Eu­ro­pean integration. I analyze when such referendums are held, and develop a novel theory of why and when referendums on this ­matter are won or lost. The chapter pre­sents a statistical model that correctly predicted the outcome of the UK’s Brexit referendum, a result that was largely unforeseen by pollsters. Chapter 8 (“Regulation of Ethnonational Referendums”) considers the regulation of ethnonational referendums, through a comparative overview of m ­ atters such as the registration of voters, the role of the electoral commission, campaign spending, and broadcasting of campaign material. This chapter also addresses issues such as: Who is eligible to vote? Who counts the votes? Who formulates the question? and Should ­t here be a special majority quorum? The book concludes with a summary of the law-­like regularities found in the chapters and sums up the arguments and makes recommendations. Last, a note on the data. Th ­ ere have been many referendums on ethnic and national issues. This book is—­unless other­wise stated—­based on the data set developed by Laponce (2010), though it has been expanded to include some of the cases in Mattern’s doctoral thesis (1921). Some have proposed to include more referendums. For example, Germann and Mendez include referendums that this study does not consider warranting inclusion (Mendez and Germann 2018). ­These authors pre­sent strong theoretical reasons for a broader criterion of inclusion. However, while their arguments are persuasive and insightful, the se­lection criterion for this book remains that of the previous edition, to wit, votes that explic­itly dealt with ethnic issues (what many would call “identity politics”).


The History and Logic of Ethnonational Referendums, 1791–1945


elf-­determination and democracy,” E.  H. Carr has argued, “went hand in hand. Self-­determination might indeed be regarded as implicit in the idea of democracy; [if] ­every man’s right is recognized to be consulted about the affairs of the po­liti­cal unit to which he belongs, he may be assumed to have an equal right to be consulted about the form and extent of the unit” (Carr 1942: 39). This was a revolutionary development. Before the French and American Revolutions, it was accepted that the ruler had a mandate from God. In fact, in the libretto of Mozart’s The Magic Flute, first performed in 1791 but originally penned in 1774, the protagonist Sarastro eulogized the divine rights of kings: “The Earth a heavenly kingdom, and mortals like the gods.” But time was ­r unning out for this idea. When Avignon (then one of the papal territories) voted to join Republican France in a referendum in the same year, the Pope complained—­through his emissary—­that the consequences of the vote would be that “henceforth every­body [would be able] to choose a new master in accordance with one’s plea­sure” (Cardinal Resonico, quoted in Freudenthal 1891: 3–4); this view the pontiff, with his a priori ac­cep­tance of the doctrine of rex dei gratia, found plainly “absurd” (Freudenthal 1891: 44). By the time the pope expressed this opinion his views already had an obsolete ring to them. A ­ fter 1776, and still more a­ fter 1789, it became the unquestioned axiom in politics—­whether in autocracies or polyarchic states—­ that the support of the p ­ eople, or the “consent of the governed,” as it was called in the U.S. Declaration of In­de­pen­dence, was the gold standard of po­liti­cal legitimacy. As Johannes Mattern put it in a classic study,


Chapter 1

The French Revolution proclaimed the dogma that we now term self-­determination. . . . ​The m ­ ental and logical pro­cess was s­ imple. The ­people are the state and the nation; the p ­ eople are sovereign. As such they have the right to decide, as the ultimate ratio, by popu­lar vote and ­simple majority, all m ­ atters affecting the state and the nation. A p ­ eople held by force and against their own w ­ ill within the bound­aries and u ­ nder the sovereignty of any state are not in real­ity part of that state. They have, consequently, the right to declare their separation from the dominant state and proclaim their in­de­pen­ dence. (Mattern 1921: 77) But the in­ter­est­ing question is how this affected the rulers. The answer is, perhaps, that the “dogma” of self-­determination became an “informal institution” in the sense that it was—to use Douglas North’s definition—­ ­­a “humanly devised constraint that [structures] po­l iti­c al . . . ​i nteraction” (North 1991: 97). Due to the ac­cep­tance of this “informal institution” of the “­will of the p ­ eople,” it became necessary for rulers to show that they heeded the w ­ ills of the majority of the p ­ eople; they w ­ ere institutionally constrained by what James March and Johan P. Olsen call “the logic of appropriateness” (March and Olsen 1984: 1). This meant that even regimes that ­were far from ­free or demo­cratic (in the present-­day sense of the word) had to find a way of showing that they ­were supported by the ­people and that they legitimately could claim to act in the name of the nation. Of course, the decision to heed the w ­ ill of the p ­ eople also had ele­ments of the “logic of consequentiality” to it; ignore the preferences of the demos and the ethnos and you ­will suffer the consequences. Politics is rarely an altruistic endeavor. As already noted in the Introduction, this overall logic can be summed up formally. Taken as a ­whole an actor, i, is likely to benefit from a referendum if ­t here is considerable competition, C, and if the squared distance between the actor’s preference point and the preference point of the median voter is small. Hence the probability of holding a referendum can be illustrated using the following example from France. In 1992 the French president, François Mitterrand, was facing stiff po­liti­cal competition from the far right Front National as well as from the moderate right Rassemblement pour la République. He perceived—­wrongly it turned out—­t hat the distance between his preference point and the preference point of the median voter was minimal; (Pi −Im)2 was small, and the electoral competition, h ­ ere denoted as C, was large. The utility of calling a

The History and Logic of Ethnonational Referendums, 1791–1945 19

referendum was perceived to be considerable; he could show that he was willing to heed the demands of the ­people, that he was in touch. Hence, he de­cided to call a vote on the Maastricht Treaty. But—as he found to his near peril—he almost lost the referendum as (Pi − Im)2 was relatively large; t­ here was ­little congruence between the president’s preferences for more transfer of sovereignty and t­ hose of other members of the elite, and the Euroskepticism of large sections of the electorate was considerable. The same logic, or so it ­will be argued, can be discerned in the referendums held between 1791 and 1944. To be sure, statistical evidence is hard to come by, but the case studies of historians point in this direction. The first main user of referendums was Napoleon Bonaparte. Though very far from being a believer in democracy, Bonaparte readily submitted policy issues to votes. When Bonaparte returned from Elba for the famous Hundred Days in 1815, he had the phi­los­o­pher Benjamin Constant draw up a constitution—­ known as La Benjamine—­which was duly submitted to a plebiscite. Yet his defeat at Waterloo put paid to the French experiments with direct democracy. At least for the time being. The Bourbons did not seek to justify their return by references to popu­ lar sovereignty, though they ­were prudent enough not to quote Jean Bodin’s dictum that “the sovereign Prince is only accountable to God” (quoted in Maritain 1961: 34). The referendum was only revived in France when Louis Napoleon—­ Bo­na­parte’s less glamorous nephew—­became emperor in the early 1850s. A perceptive young journalist named Karl Marx reflected on this and found that the referendum was particularly suitable to autocrats as it enabled them to speak on behalf of the nation: “While each representative represents only this or that party, this or that city, this or that dunghill, the President, on the contrary, is the elect of the nation, and the act of his election is his trump card” (Marx 2016: 26). Why did the French hold referendums at the end of the nineteenth ­century? One reason was that the newly established republican regime faced considerable “competition” (i.e., military threats) from both erstwhile royalists in France and the foreign powers that ­were anything but enthused by the prospect of the new administration in Paris. The challenges to the revolutionaries’ rule ­after 1789 ­were considerable. But by holding a referendum on the sovereignty over Avignon in 1792, the Assembly in Paris could show that their actions ­were close to the views of


Chapter 1

the ­people in the areas in question, which gave them legitimacy. Using po­ liti­cal science language, the decision to hold a referendum provided an opportunity for Paris to show that the distance between the preference point of the Assemblée Nationale and that of the median voter in Avignon was small if not identical to that of the preferences of the legislature. This thinking was evident when the Assembly passed a resolution a­ fter the vote in the erstwhile papal enclave: “The majority of the communes and citizens have expressed freely and solemnly their wish for a ­union with France . . . ​the National Assembly declares [itself] . . . ​in conformity with the freely expressed wish of the majority” (von Martens 1801: 400).

The Risorgimento Referendums The 1820s and the 1830s w ­ ere periods of drought in terms of submitting issues to a vote among the ­people. But the nationalist sentiment regained momentum in the 1830s and did so with a vengeance. In the wake of the rekindled patriotism, a number of irredentist groups began movements that led to a rediscovery of the idea of the referendum as a mechanism to resolving ethnonational conflict. This was especially true in Italy. The fragmented Italian states had long wanted unity and unification. Niccoló Machiavelli had dreamed about “seizing Italy and f­ree[ing] her from the barbarians” (Machiavelli 2002: 95). While the Italians had been briefly unified u ­ nder Napoleon, the French emperor had treated Italy as a vassal state (he made his ­sister Elisa Baciochi ruler of Naples, and his stepson Eugène Rose de Beauharnais was crowned king of Rome) (Davies 1997: 181). The Italian nationalists who came of age in the 1830s had a dif­fer­ent idea. In his influential treatise L’idea dell’Italia Giuseppe Mazzini had plainly said that the “princi­ple of nationality is sacred for me. I regard it as the guiding princi­ple of the ­future. I am prepared to greet—­without fear—­any change of the map of Eu­rope, which is the result of a spontaneous manifestation of the w ­ ill of a ­people” (quoted in Harder 2006: 29). And as early as 1833, Mazzini had proposed that referendums rather than terrorism (as advocated by Filippo Buonarotti) be used to win legitimacy for the proj­ect of unification (Mack Smith 1959: 16). Part of the reason for this was, according to Stephen Tierney, a change in ideology: “The demand for increased popu­lar participation grew, [and] the plebiscite became a tool in the in­de­pen­dence strug­gles of the latter part of the ­century” (Tierney 2012: 62).

The History and Logic of Ethnonational Referendums, 1791–1945 21

­These sentiments led to a referendum in Lombardy in May 1848, in which a majority voted to join the Kingdom of Sardinia. This vote was thwarted by the Austrian armies, but a de­cade ­later—­and with the support of Napoleon III—­Austria was forced to concede the territory to the Kingdom of Sardinia. In return for his support, the Italians agreed that two territories, Savoy and Nice (which had been transferred to the Kingdom of Sardinia in 1815), would be allowed to hold plebiscites on w ­ hether to join France. Both t­ hese votes resulted in large majorities in ­favor of unification with France. While ­these referendums ­were undoubtedly rigged, the same cannot be said of the referendums that—at the insistence of the Italian statesman Conti di Cavour—­were held in Tuscany, Reggio, Sicily, Naples, and Umbria. However, the referendums ­were not massive manifestations of unquestioned support and high participation rates. As a British envoy in Naples noted, “The proportion of the estimated population actually voting in the plebiscites was a mere 21.17 ­percent in Tuscany, 20.09 ­percent in Emilia and 19.17 ­percent in Naples” (Goodhart 1971: 106). But why did the Italian Risorgimento rest on referendums? ­Under the theoretical framework introduced above, the plebiscites proved useful to show that the distance between the ­people and their rulers—­formally (Pi − Im)2—­was small, and at a time when Guiseppe Garibaldi and his allies w ­ ere threatened by other forces in both Italy and abroad, not least the Austrians and the French (who initially had supported them), ­t here was a need to show that the unification had popu­lar support (Hibbert 1987: 171). While it is not a neat fit, it is difficult to contest the view that the Risorgimento referendums follow some of the logic of the overall model: Garibaldi faced po­liti­cal competition and needed to show that he was close to the ­people. The referendums ­were an excellent way of achieving this. In this way, perhaps, the referendum also provided a way of establishing the new state. Massimo d’Azeglio had famously noted that “we have made Italy, it remains to make the Italians” (quoted in Hobsbawm and Kertzer 1992: 3). It is pos­si­ble that the referendum, as an act of public participation, contributed to the establishment of a new demos.

Nationalist and Separatist Referendums in Amer­i­ca The situation in the United States was less glamorous. The referendum has been deep-­seated in American po­liti­cal culture since the War of In­de­pen­dence,


Chapter 1

and it was used early on in the life of the American republic to resolve issues pertaining to sovereignty. The first example occurred in 1788 in Mas­sa­chu­ setts. By the mid-1850s it had become commonplace to consult the citizens in major issues of constitutional importance (Lee 1981: 46). Most of t­hese referendums had a deliberative character (Fisch 2006: 485). However, when the Confederate states turned to plebiscites, the mechanism of direct democracy was used in a hitherto unknown way in North Amer­i­ca. So why did some of the southern states resort to referendums? Like the Italian Risorgimento politicians, the southerners ­were ­under severe po­liti­cal pressure and faced considerable competition from the North (Wooster 1961: 117). To maintain legitimacy, it was vital for them to show that they understood the concerns of their constituents, and that the voters consented to their actions. A referendum was a con­ve­nient way of showing this. Referendums assenting to secession would send a strong signal to President Abraham Lincoln and the abolitionists in the northern states as well as provide a rallying point for the voters in the South. It is not surprising, therefore, that Texas, V ­ irginia, and Tennessee submitted the decision to secede from the Union to the voters. What is perhaps in­ter­est­ing—­and what might show that the decision to hold a referendum was not based on solid facts—is the fact that the support for secession was far from unan­i­mous. In Tennessee, for example, 104,913 voted for secession while 47,238 voted against (Anderson 1994: 123) and in Texas the figures ­were 44,317 for and 13,020 against. In ­Virginia 125,950 voted for and 16,646 voted against (Anderson 1994: 120). ­These ­were not endorsements of epic proportions, though ­t here ­were significant majorities, nearly 69  ­percent in Tennessee, 72  ­percent in Texas and 86 ­percent in V ­ irginia (Mattern 1921: 119). (Arkansas also voted on secession from the United States, but not in a referendum. The decision was not made directly by the ­people themselves but by a so-­called session convention of men elected to a special assembly for that purpose.) We s­ hall return to t­ hese cases in Chapter 3.

Schleswig: The Dog That ­Didn’t Bark ­ fter the polls in Italy and the United States, in the m A ­ iddle of the nineteenth ­century the use of the referendum died down once again. But referendums ­were proposed. Most notable, perhaps, is the case of the Second Danish-­ German War of 1864. Following the Danish government’s decision to incor-

The History and Logic of Ethnonational Referendums, 1791–1945 23

porate the two duchies ­under a Unified Danish Constitution, Prus­sia and Austria declared war in 1864. ­After heavy Danish losses, a cease-­fire was negotiated in ­England. During ­these negotiations, “[the] use of a [referendum] had been urged by the British government” (Goodhart 1971: 106). The British proposed that a referendum be held in each of the towns and parishes in the now occupied Schleswig (Ward and Gooch 1923: 579). The German negotiator, Count Albrecht von Bernstorff, declared, “I have nothing against sending this proposal ad referendum.” However, the chief Danish negotiator, Foreign Secretary Georg Quaade, rejected the proposal and declared that it was against his “instructions” to “submit the ­matter ad referendum.” As a result, hostilities erupted again. The Prus­sians and the Austrians occupied the w ­ hole of Schleswig and—­predictably—­became less keen on a plebiscite. But why did the Danes reject the proposal of a referendum? In view of the competition proximity model ­t here is a ­simple explanation. Danish public opinion was inflexible, and ­after the military victory in the Three-­Year War against Prus­sia (1848–1851), a large segment of the Danes was in ­favor of a military solution, which they unrealistically believed would be pos­si­ble to win. To consent to a referendum would signal that t­ here was a gulf between the position of the government of D. G. Monrad and the ­people. Formally (Pi − Im)2 was large. The p ­ eople did not ­favor a referendum. ­There was no strategic mileage in forcing a vote. In addition, Monrad and the National-­Liberals had a solid majority and ­were not challenged electorally. ­There was l­ittle utility in a vote; consequently, the issue was not sent “ad referendum” (Bak 1975: 274). ­After the war, the Prus­sian chancellor, Otto von Bismarck, in 1866, told the Prus­sian Parliament that a referendum was necessary. A month ­later, realizing the significant opposition to German occupation, he annexed the duchies before negotiations over the promised referendums could take place. Again, Bismarck’s logic was consistent with the competition proximity model; formally (Pi −Im)2 was large, hence ­there was a gulf between Bismarck’s position and that of the voters—­and Bismarck was u ­ nder threat neither in the German Reichstag nor internationally. The British, while not enthusiastic about German expansion that could alter the balance of power on the continent, ­were not in a position to assert pressure on Berlin. Consequently, Bismarck’s incentives for holding a referendum ­were small; nothing would be gained strategically from sending the proposal to the voters, and indeed he might even lose. At this stage, of course, ­matters had changed considerably


Chapter 1

in Denmark. Having lost on the battlefield, the Danish government was in ­favor of a vote. The defeat had not been well received domestically (Monrad had to resign soon a­ fter and emigrated to New Zealand), and a referendum was a popu­lar option—­indeed the only option—­for regaining the lost territories (Goodhart 1971: 107). This decision not to hold a referendum is in sharp contrast to the Danes’ insistence on a referendum on the fate of the West Indian Islands Saint Croix, Saint Thomas, and Saint John, which the Americans had offered to buy. Contrary to their opposition to plebiscites in 1864, Copenhagen now argued that a vote was necessary as “a new custom in Eu­rope of holding plebiscites was now so entrenched that any failure to stage a referendum would cause comments” (Goodhart 1971: 109). A referendum was duly held in 1868, in which “all male residents, black as white, had a vote. The vote was overwhelmingly in favour of transfer to the United States” (Goodhart 1971: 109); 98  ­percent voted “yes”—­a nd yet, the treaty was not ratified by the U.S. Senate. When the Americans made another offer in 1902, the sale was blocked by the Danish second chamber (Landstinget). The American made yet another offer in 1915, which led to the collapse of the Conservative-­ Liberal government. To survive, the Social Liberals invited the Socialists to join the government, but it was clear now that the sale of the Virgin Islands was controversial, though largely a popu­lar option. The government was u ­ nder threat electorally, having incorporated the socialists—­who only recently had renounced revolution. With an electoral threat and a popu­lar issue, Prime Minister Thomas Zahle submitted the issue to a referendum. However, the inhabitants of the islands w ­ ere not asked w ­ hether they wanted to be sold to the United States. (For this reason, the Virgin Islands are not included in ­Table 1.1). The vote was opportunistic—­u nprincipled if you like—­but entirely consistent with the rationality of the competition proximity model; 64 ­percent of the Danish voters voted “ja” (yes) to the sale.

Nationalist Referendums in the English-­Speaking World The En­glish w ­ ere never in the vanguard of the movement t­ oward using referendums and plebiscites to resolve national issues. To be sure, the British ­were not adverse to using the referendum as a tactical means of international politics (e.g., in the case of the referendum in Moldova in 1857, where the referendum was a con­ve­nient excuse to curb the influence of the Rus­sian

The History and Logic of Ethnonational Referendums, 1791–1945 25

Empire ­after the Crimean War) and occasionally suggested referendums (as they did in the case of Schleswig). ­After the Crimean War (which Britain, France, and the Ottoman Empire fought against the Rus­sians), a poll was held to unify the two territories of Moldova and Walachia (previously an area that had been u ­ nder Turkish suzerainty, though it was often dominated by Rus­sia u ­ nder the name Romania) (Laponce 2010: 71). However, it should be noted that the referendum was anything but f­ree and fair; “intimidations and arrests ­were not infrequent,” and up to “nine-­tenths of the population w ­ ere denied the right to vote” (Mattern 1921: 104). Furthermore, the vote was held only ­after some “bizarres manoevres diplomatiques” (Laponce 2010: 71). While the British ­were—on occasion—­willing to suggest referendums outside their own territories, they ­were rarely keen to conduct referendums within the empire and its territories. The British accepted a referendum regarding the transfer of the Ionian Islands to Greece (Goodhart 1971: 104), but this was the exception. The government in London—­unlike that in Paris—­was not smitten by the idea of referendums to resolve issues of sovereignty. Whereas the French willingly submitted the issue of the transfer of sovereignty of Saint Bartholomew from Sweden to France to a plebiscite in 1877 (Mattern 1921: 116), the British ­were generally opposed to this course of action. In the case of the transfer of sovereignty of Heligoland (an island close to Germany), Lord Salisbury, the foreign secretary, rejected a proposal by Lord Rosebery (­later the prime minister) for a plebiscite, stating, “My answer must be negative. The plebiscite is not among the traditions of this country. We have not taken a plebiscite; and I can see no necessity of ­doing so” (House of Lords Debates, vol. 345, June  1890, cols. 1311–1312). The same view was taken by Prime Minister William Gladstone, who told Parliament that he was similarly opposed to a referendum. The proposal for a referendum was rejected in the House of Commons by a vote of 172–76 (Mattern 1921: 112). But this does not mean that the referendum was not used at all by the British, or rather by ­people of British extraction. Referendums ­were held in Canada (Nova Scotia), South Africa (Natal), and Australia. The British ­were keen to give a mea­sure of autonomy to areas with a large white population—­ such as present-­day Canada, Australia, and South Africa. As far as the latter was concerned, a conference was held among the South African colonies with a view to establishing an autonomous u ­ nion within the British Empire shortly ­after the Boer War. Based on the recommendations of the conference,


Chapter 1

the UK Parliament passed the South Africa Act 1909, which was subsequently ratified by the South African colonies. However, in Natal, the smallest of the previously existing states, ­t here was considerable concern that a centralized state would be detrimental to the interests of the area. While the South Africa Act was passed in its entirety in the Transvaal and Orange River parliaments, opposition in Natal was so strong that the local administration de­cided to call a referendum. That settled the issue. Support for ­union was strong—­perhaps ­because t­here was l­ittle alternative. The South Africa Act was passed 11,121 votes to 3,701 (Kahn 1960: 1). The establishment of Canada in 1867 did not involve any official referendums. But the poll held in Nova Scotia in 1867—on leaving the newly established federation—­was an unofficial one and was ignored by the authorities, despite 65 ­percent voting for separation (Laponce 2010: 50). The situation was dif­fer­ent in Australia, not as a result of British pressure, but ­because the po­liti­cal class in Australia felt compelled to win support from constituents before g­ oing ahead with the pro­cess of federation. That the Australians ratified the unification of their country through a series of plebiscites was not, it seems, due to their British ­legal and constitutional heritage, but rather a result of the more progressive ideas they had received from another settler society, namely the United States (Williams and Hume 2010: 7). The 1891 Australian Constitutional Convention agreed that before proceeding with federation, the constitution for governing the new nation should be approved by the p ­ eople. The intention was affirmed at the Corowa P ­ eople’s Convention in 1893. This enabling legislation was passed in each colony. In 1898 referendums on the Commonwealth Constitution Bill w ­ ere held in New South Wales, South Australia, Tasmania, and Victoria. A majority of yes votes was recorded in each colony, but in New South Wales the enabling legislation required a quota of eighty thousand. This was not achieved. In 1899, as a result of amendments to the constitution recommended by New South Wales, the colonies or­ga­nized a second round of referendums. This time New South Wales required only a s­ imple majority of yes votes. Queensland also joined the pro­cess. Majorities ­were achieved in all colonies, albeit with dif­fer­ent levels of enthusiasm. ­There was not widespread support for the federation everywhere. Mineral-­ rich Western Australia was especially hesitant. By 1900 the colony had still not taken steps to hold a referendum. In protest, residents of the Eastern Goldfields took steps to form a separate colony. This set the ball rolling. Fi­nally, on July 31, 1900, when the Commonwealth Constitution Bill had already been

The History and Logic of Ethnonational Referendums, 1791–1945 27

enacted by the British Parliament, a referendum was held in which a large majority voted in ­favor of federation. But why w ­ ere the referendums held in the first place? Was this another case of the logic outlined at the beginning of this chapter? Without wanting to force the m ­ atter too far, t­ here are indications that some of the under­lying logic was at play. In a famous analy­sis of the pro­cess of federalization, Willian Riker Jr. has pointed out that “external” and “internal” threats are significant (Riker 1975: 116). This was perceived to be the case in Australia. Moreover, ­t here was a need to show that the policy of federation was supported by the voters. Holding a referendum was an ideal way of demonstrating this and was, moreover, consistent with the po­liti­cal fashion at the time, especially as expressed by prominent En­glish l­awyers and po­liti­cal theorists (MacIntyre 2004: 138). Yet, as one observer noted, “unlike the Italians, it [Australia] experienced no Risorgimento. The turnout in federal referenda was lower than for parliamentary elections” (MacIntyre 2004: 138). The most celebrated referendum to be held before World War I was perhaps the 1905 poll in Norway. When Norway’s parliament, Stortinget, in 1905 sent notification to Sweden that Norway had seceded from the ­union established in 1814, the response was initially negative. The Swedish Riksdag responded that the ­union was two-­sided, and that in strict ­legal terms, the ­union could not be dissolved without the consent of King Oscar II and the Riksdag (Vedung 2007: 5). Yet the Swedes conceded that the request would be accepted if it was preceded by “a fairly conduced [sic] plebicite” (Nordlund 1905: 365). The statement went on to say that if the “conditions [of a fair referendum] ­were complied with negotiations would be entered into” (Eden 1905: 23). The Swedes had not expected that the Norwegian prime minister, Christian Michelsen, would take up the challenge and or­ga­nize the referendum. Michelsen, according to a recent study, “had ‘teft’—­t his strange and almost animalistic ability to sense, feel, and gauge t­ hings as opposed to the ability to analyse, calculate and rationally . . . ​even while in the ­middle of the maelstrom” (Hegge 2010: 197). Michelsen’s ­gamble paid off. More than 99 ­percent voted to sever the ties between the two countries; Sweden almost immediately entered practical negotiations in the border town of Karlstad and divided the spoils in an amicable way. That this was pos­si­ble had, perhaps, just as much to do with the fact that the Swedes w ­ ere not an aspiring power and that the relationship with Norway was not eco­nom­ically or po­liti­cally beneficial to Stockholm.


Chapter 1

But—as area specialists and historians have pointed out—­there w ­ ere also more sinister or strategic reasons ­behind the referendums. True, Norway was not eco­nom­ically beneficial to the Swedes, but the right-­wing parties in the Swedish Riksdag sought to capitalize on the national romanticism among their voters. Seeking to win support in the upcoming Swedish general election in the autumn of 1905, the parties on the right proposed that the issue of the u ­ nion could be resolved through a referendum. Well aware that the policies of Prime Minister Erik Gustaf Boström had antagonized public opinion in Norway, the parties on the right—­especially the Protectionist Party of the majority—­k new “even at the time before mass opinion polls that the Norwegians would vote for in­de­pen­dence.” However, “calling a referendum could delay ­matters” and would provide the parties on the Swedish right the opportunity to score party po­liti­cal points (Bjørklund 2003). The strategy, as suggested by Thomas Wyller, was internal, and the Swedish right was focused on the opportunity to win votes in the upcoming Swedish elections rather than preventing Norwegian in­de­pen­dence (Wyller 1992: 40). Seen in this light, it can be argued that the referendum—at least to a degree—­conforms to the competition proximity model; the Swedish po­liti­ cal parties w ­ ere competing while locked into the “competitive strug­gle for votes,” to use Joseph Schumpeter’s expression (Schumpeter 1942: 242), and the parties on the right knew that public opinion in Sweden favored a delay for nationalistic and romantic reasons. Arguably, the logic of the competition proximity model can be detected. What the parties in Sweden could not have foreseen was Michelsen’s “teft” and the Norwegians’ ability to or­ ga­nize a referendum in less than two weeks. Furthermore, Michelsen’s decision was itself consistent with the model. U ­ nder threat from Stockholm, the Norwegian leader faced stiff po­liti­cal competition; and derogatory remarks by Swedish prime minister Erik Gustaf Boström meant that a referendum was a popu­lar option in Norway. Both sides had—at dif­fer­ent times—an incentive to call a referendum (Vedung 2007). But as can be seen from ­Table.1.1, this type of referendum was a rarity at the time.

Referendums A ­ fter World War I It is commonly assumed that Woodrow Wilson was responsible for the referendums in the wake of World War I (Qvortrup 2012). To be sure, before he became president, Wilson (­running on a progressive ticket) championed the

The History and Logic of Ethnonational Referendums, 1791–1945 29

­ able 1.1. Ethnonational Referendums from the French Revolution to the First T World War Country



France France France Belgium France France France France France Italy Italy Turkey Italy Italy Italy Italy Italy Italy France United States United States United States United Kingdom Italy Canada Denmark Italy Sweden Australia Australia

Avignon Savoy Nice Wallonia Moselle Mul­house Geneva Switzerland France Lombardy Reggio Romania Parma Sicily Tuscany Naples Marches Umbria Savoy Texas

1791 1792 1792 1793 1793 1798 1798 1802 1816 1848 1848 1857 1860 1860 1860 1860 1860 1860 1860 1861







Ionian Islands


Venice Nova Scotia Virgin Islands Rome St. Bartholemew Tasmania New South Wales Victoria South Australia Western Australia

1866 1867 1868 1870 1877 1898 1898

1 1

1898 1898 1898

1 1 1

Australia Australia Australia

Difference-­ Difference-­ Secession Right-­Sizing Eliminating Managing Referendums Referendums




1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

1 1

1 1 1 1



Chapter 1

­Table 1.1. (continued) Country



Australia Australia Australia Australia Australia

Queensland South Australia Tasmania Victoria New South Wales Western Australia Norway Natal

1899 1899 1899 1899 1899

1 1 1 1 1




Sweden United Kingdom Rus­sia Finland Denmark Iceland Finland Aaland

Difference-­ Difference-­ Secession Right-­Sizing Eliminating Managing Referendums Referendums

1905 1909 1918 1918 1918



1 1

use of referendums as the ­people’s means of “responding to subsidized machines” (Wilson, quoted in Munro 1912: 87). But while Wilson was still the “enthusiastic apostle of self-­determination . . . ​he had no enthusiasm for plebiscites as such,” and, moreover, “this view was shared by the most impor­tant members of the American del­e­ga­tion” (Goodhart 1971: 115). That the eight referendums ­were held to establish the borders between the previously warring states was more a result of internal pressures and strategic considerations (Wambaugh 1933). The British del­e­ga­tion of Lord Balfour, who had advocated the use of referendums to s­ ettle the issue of tariff reform in Britain (Bogdanor 1981: 51), was at least consistent and was responsible for the referendums that ­were held. Indeed, it was “the British del­e­ga­tion [that] suggested the plebiscites in Allenstein on the German-­Polish border and in Upper Silisia, an area which American experts had originally assigned to Poland” (Goodhart 1971: 115). It can always be discussed w ­ hether t­hese referendums resolved the issues of irredentism that so preoccupied the Versailles Conference. Furthermore, the fact that referendums w ­ ere denied in territories that w ­ ere claimed by Germany—or in which t­ here was a German majority (e.g., Tyrol and Alsace-­ Lorraine)—­suggests that the referendums w ­ ere not as neutral and idealistic

The History and Logic of Ethnonational Referendums, 1791–1945 31

as some enthusiasts would have us believe. For example, the referendum or­ga­nized in Tyrol was not endorsed by the Americans, although 90 ­percent voted for joining Germany. That not all areas w ­ ere allowed to let the p ­ eople decide might—­with the benefit of hindsight—­have paved the way for irredentist claims l­ ater. Indeed, it is in­ter­est­ing that—as Vernon Bogdanor has observed—”it was precisely in ­t hose areas where plebiscites ­were refused (with the exception of Alsace-­Lorraine)—­Danzig, the Polish corridor and the Sudetenland—­ that w ­ ere the subject of revisionist claims by the Nazis in the 1930s” (Bogdanor 1981: 145). And it is equally noteworthy that ­t here ­were no claims on lands that had de­cided their allegiance through plebiscites such as Schleswig-­ Holstein. Even for a dictatorial regime such as Nazi Germany “frontiers that w ­ ere fixed by plebiscite could not easily be undermined” (Bogdanor 1981: 145). The main supporter of referendums ­after World War I was Germany. The country, which had opposed referendums and plebiscites only a few years before, was now advocating their use whenever given the chance. It did this not only vis-­à-­vis the Western powers at Versailles but also in its negotiations with the new Soviet regime in the east. At the negotiations with the Soviet Union at Brest-­Litovsk, Austrian foreign minister Count Czernin proposed to Leon Trotsky, the Soviet foreign minister, that referendums be held in areas currently ­under the control of German, Austrian, or Turkish troops. The Axis powers affirmed that the ­f uture status of Poland and the Baltic States would be de­cided in “consultation with their populations” (Goodhart 1971: 112). However, only one plebiscite was held ­under the auspices of the Brest-­Litovsk Treaty; in 1918 a referendum was held in Kars while it was occupied by Turkish troops (previously a part of the Rus­sian Empire). That the Soviet Union had signed up to ­t hese conditions did not mean, however, that it was committed to the princi­ple of self-­determination. On the 25th of June, 1918, at the Kiev Peace Conference, the Soviet regime signed an agreement with the in­de­pen­dent government of Ukraine for a referendum to be held in certain territories claimed by both lands. Two days l­ater, the Red Army—­u nder Trotsky—­ invaded Ukraine and annexed the ­whole country without a referendum (Rein 2000: 33). ­After World War I, the number of ethnonational referendums diminished. The votes held between the two world wars outside Germany ­were largely inconsequential, and in some cases almost po­liti­cal curiosities, such


Chapter 1

as the referendum in 1933 in Western Australia. On April 8, 1933, the government of premier and Nationalist Sir James Mitchell or­ga­nized a plebiscite on secession alongside the state parliamentary election. Mitchell campaigned in f­avor of secession, while the L ­ abor Party campaigned against breaking from the federation. Of the 237,198 voters, 68 ­percent voted in ­favor of secession, but at the same time the Nationalists ­were voted out of office. Only the ­Table 1.2. Ethnic and National Referendums, 1918–1945 Country



Denmark Finland Turkey Austria Germany Germany Germany Belgium Germany

Iceland Aaland Kars Vorarlberg Schleswig 1 Schleswig 2 Allenstein Eupen Marienweder Klagenfurt Upper Silesia Tyrol Salzburg Sophron Rhodesia

1918 1918 1918 1918 1920 1920 1920 1920 1920

Austria Germany Austria Austria Austria United Kingdom Spain Spain Australia Germany Germany Germany/ Spain Germany/ France United States Austria

Difference-­ Eliminating

Difference-­ Managing



Right-­ Sizing 1 1 1 1 1 1 1 1

1920 1921

1 1

1921 1921 1921 1922


1 1 1

Catalonia Basque Country Western Australia Germany Germany Saar Galicia Germany

1931 1933

1 1

Philippines Austria

1935 1938

1933 1933 1934 1935 1936 1936

1 1

1 1 1



1 1

The History and Logic of Ethnonational Referendums, 1791–1945 33

mining areas, populated by keen federalists, voted against the move. The state sent a half-­hearted petition to the British Parliament requesting in­de­pen­ dence. It got nowhere ­a fter the petition was ruled out of order ­because the convention dictated that it be made by the Commonwealth (of Australia) and not by the individual state (Williams and Hume 2010: 8). The fact that Mitchell had lost the election effectively killed the proposal. The proposal was stillborn and died. The same cannot be said of the votes held in Germany, where Hitler (ab) used the referendum to eliminate differences and to create unity in the reich. What is perhaps in­ter­est­ing (and disturbing) is that most of t­ hese votes—at least according to con­temporary observers—­were relatively fair. Writing about the referendum regarding withdrawal from the League of Nations, an American observer concluded, “Even a­ fter discounting intangible official pressure, of which t­ here undoubtedly was a g­ reat deal, and downright coercion and intimidation at the poll of which ­t here was prob­ably very ­little, the electoral rec­ord remains an amazing one” (Zurcher 1935: 95, emphasis added). This book mainly examines referendums ­after World War II (see ­Table 1.2). ­Until this time referendums ­were to a large degree held for strategic reasons, often in a way that was consistent with the competition proximity model—­ though the evidence analyzed so far is, admittedly, based on impressionistic evidence drawn from historical accounts by area specialists. W ­ hether the same conclusion can be drawn if we look more closely at the evidence and data for the period a­ fter 1945 is a m ­ atter I seek to answer in the remaining chapters.


Difference-­Managing Referendums


he referendum is an emphatic assertion of the princi­ple that nation stands above parties,” according to a prominent ­legal scholar (Dicey quoted in Cosgrove 1981: 67) The difference-­managing referendum is exactly this. It is a mechanism that limits the scope for action and offers the possibility of what we, paraphrasing the Declaration of In­de­pen­dence, may call the dissent of the governed. Typically, ­t hese referendums are held on ­matters regarding so-­called devolution. One of the prob­lems with ­t hese referendums is that, to quote Markku Suksi, “­There does not exist any solid theory about autonomy or devolution, perhaps ­because autonomy arrangements are often very pragmatic ad hoc solutions that escape generalizations.” Having said this, however, Suksi is happy to accept that devolution normally consists of a situation in which the “inhabitants of the autonomous territory have the right to elect their own self-­governing bodies” (Suksi 2003: 221). In this book this definition w ­ ill be followed. Generally speaking, difference-­managing referendums come in two forms: “

• Devolution referendums, which deal with the del­e­ga­tion of power (home rule) to a geo­graph­i­cally defined territory, like when the Australian Capital Territory was offered self-­government in 1978 and when Nuuk (the capital of Greenland) and Copenhagen submitted their renegotiated relationship to a referendum in 2009; • Power-­sharing referendums, that is, votes on agreements on consociational power sharing, like the referendum in Burundi in 2005, the plebiscite in Northern Ireland in 1998, and the referendum on power sharing in Suriname in 1987.

Difference-­Managing Referendums 35

But why and when are such referendums held? On one level this is a normative question: difference-­managing referendums are held when legitimacy is required for a decision of considerable magnitude and constitutional irreversibility. That, at least, is the theory favored by ­legal theorists such as A. V. Dicey. Yet t­ here is also a less idealistic and empirical view, which differs from the normative ideal. The overall framework of this book proposes that the decision to hold a referendum on an ethnic or national issue, ceteris paribus, follows the competition proximity model:

C >1 i=1 (I − P )2 m i n

Pref = ∑

The probability, Pref, that a government, i, ­w ill submit a proposal for difference-­managing policy (such as, for example, the system of home rule in Scotland) to referendum depends on the relationship between the competition, C, the actor is facing and the squared distance between the government’s policy point and the position of the median voter. In the context of difference-­managing referendums, we would expect that ­t hese polls are held when the initiating party (for example, the British ­Labour Party in 1997) is ­under electoral pressure and has reason to believe that its preferred option is close to that of a majority of the voters. For example, in the more empirical lit­er­a­ture on British devolution referendums, it is commonly argued that the polls w ­ ere held for electoral rather than for idealistic or philosophical reasons. In the words of Richard Wyn Jones and Roger Scully, “Mr. Blair’s decision was a ­matter of electoral strategy; neutralizing the potentially contentious part of ­Labour’s constitutional reform agenda as election issues by allowing the electorate the power of decisions in ­f uture referendums” (Wyn Jones and Scully 2012: 15). The ­Labour Party was locked in competition with the Conservatives. ­Under the circumstances ­Labour needed to show that its constitutional reform program coincided with that of the median voter. By promising a referendum on Scottish and Welsh devolution, the government effectively removed the issue—­which played well with conservative voters—­from the competitive strug­gle between ­Labour and the Conservatives. This model seems convincing—at least superficially—­but it is not the only one. I cannot and ­will not ignore other models that are closer to Dicey’s idealistic justification for holding difference-­managing referendums. I start with the other contenders.


Chapter 2

Dicey: The Referendum as a Constraining Instrument In a series of articles and papers, including in the sixth edition of his influential treatise An Introduction to the Study of the Law of the Constitution and in the article “­Ought the Referendum to Be Introduced into E ­ ngland?” (Dicey 1890, 1981), Albert Venn Dicey made a case for the referendum as an “alternative second chamber” (quoted in Qvortrup 1999: 531). Dicey’s concern was the policy of home rule for Ireland proposed by William Gladstone’s Liberal government. In 1886, the First Home Rule Bill was proposed, but it was dropped following intense opposition, which split the Liberal Party. ­After the Bill’s defeat in the House of Commons, the Second Home Rule Bill was introduced in 1893. This Bill was rejected in the House of Lords. But it was gradually becoming clear that the unwritten British Constitution did not have the checks and balances to ensure that radical changes such as this could be prevented. It was with this in mind that A. V. Dicey (then a professor of constitutional law at All Souls College, Oxford, and a staunch conservative) came up with the idea that the referendum could perform the function of an alternative second chamber. Unlike the House of Lords, the referendum was—in Dicey’s opinion—­ “the one available check on party leaders” and the only institution that could “give formal acknowledgement of the doctrine which lies at the basis of En­glish democracy—­t hat a law depends at bottom for its enactment on the consent of the nation as represented by its electors” (Dicey 1911: 189). Steeped as he was in the tradition—if not the practice—of liberal constitutionalism, Dicey was at pains to show that the referendum was consistent with the ideal of l­imited government, but his prob­lem was that the prevailing system, that is, a balance of power between, in his case, the House of Lords and the House of Commons, was out of sync with the realities of popu­lar politics at the time of writing. The po­liti­cal fault line was no longer between the aristocracy and the bourgeoisie, but between the elected representatives and the emerging ­middle class, which had been granted the vote through dif­fer­ent reforms. In Dicey’s view, the elected politicians w ­ ere not sensitive to the national issue and, through the policy of home rule for Ireland, w ­ ere in danger of jeopardizing the unity of the nation (as he saw it). The referendum was, in Dicey’s view, a mechanism that could put a stop to this, and do so in a way that was consistent with the new doctrine of popu­lar sovereignty. As he wrote, “I value the referendum first ­because it is ­doing away with the strictly

Difference-­Managing Referendums 37

speaking absurd system which at pre­sent exists, of acting on the presumption that electors can best answer the question raised, e.g. by Home Rule, when it is put together with such totally dif­fer­ent questions of prohibition” (Dicey quoted in Cosgrove 1981: 67). For Dicey, therefore, the referendum was a strictly negative po­liti­cal instrument. It was an instrument that ­limited the scope of government. It was “nothing more nor less than a national veto” (Dicey quoted in Cosgrove 1981: 106). Of course, Dicey’s theory was predominately normative. He was not in the business of predicting when a referendum would be used. He was concerned only with the question of when the p ­ eople ­ought to be allowed a say on an impor­tant issue. And his answer was that the referendum predominately should be held for ­matters concerning the devolution of power. At the time of writing—in the late Victorian and early Edwardian age—­ such referendums ­were unheard of. As we have seen, no referendums—­with the exception of the votes in Australia at the turn of the c­ entury—­were difference-­ managing referendums. Rather like Karl Marx, who expected that his theories would be most relevant in developed cap­i­tal­ist socie­ties, and not in countries with shorter traditions, Dicey assumed that his normative theory would be most applicable in advanced liberal polyarchies. However, taking stock of three-­quarters of a ­century of devolution referendums, Rouke, Hiskes, and Zirakzadeh asserted that “referendums on regional self-­rule seldom occur in advanced industrialized socie­ties” (Rouke, Hiskes, and Zirakzadeh 1992: 112). Admittedly my focus in this chapter is wider, as I consider not only territorial power sharing (e.g., the referendum on the Charlottetown Agreement in Canada in 1992 and the referendum in Catalonia in Spain in 2006) but also nonterritorial consociational agreements such as the Good Friday Agreement in Northern Ireland in 1998. But nevertheless, the statement does not seem to hold true. Looking at the list of difference-­managing referendums, most appear to have been held in advanced industrial democracies. Moreover, the assertion that t­ hese types of referendums are rare or “seldom occur” is questionable as a total of eighty-­ four difference-­eliminating referendums have been held. Rouke, Hiskes, and Zirakzadeh had a point when they w ­ ere writing. ­There ­were fewer difference-­managing referendums in the 1980s, and the ones held ­were, arguably, of ­limited consequence. The conclusion Rouke et al. drew in the early 1990s was based on what appeared to be a trend at the time. The prob­lem is, however, that t­ hings change. In the following de­cade, the number grew to twenty-­one cases, and ­t here ­were twenty-­two referendums


Chapter 2

in the first de­cade of the new millennium, before the number dropped slightly to sixteen between 2010 and 2019 (see ­Table 2.1). What accounts for this tendency to have more referendums on difference-­ managing policies? In the first edition of this book, I tested the hypothesis that governments hold referendums when they face po­liti­cal competition. I mea­sured this by a dummy variable with the value of one if the granting of self-­government to the territory was an election issue. I also mea­sured support for the policy by using the yes vote in the referendum as a proxy for the support of the devolution of power. In that analy­sis I found statistical support for the proposition, though I also found that countries with higher levels of democ­ratization ­were statistically more likely to hold difference-­managing referendums. In the research for this book, I was able to identify a number of previously unknown cases. As some of t­ hese cases ­were from the period before 1973 (when Freedom House began to grade democracies), I used the Polity IV mea­sure of democ­ratization instead, which goes all the way back to 1800. Using Polity IV over Freedom House in this case is not a general preference. Indeed, I use Freedom House mea­sures in some of the other chapters. Both mea­sures are respected and widely used, and often correlate. The reason for using Polity IV is merely for inclusion purposes. ­Table 2.2. provides a statistical overview of the ­factors in the form of a regression model. This tendency is further illustrated by the error bars in Figure 2.1. As the figure shows, the statistical margin of error gets smaller the higher the country in question is on the Polity IV Scale. This indicates that the probability that a difference-­managing referendum ­w ill be held is higher the more demo­cratic the country in question becomes. This tendency can be illustrated by a number of case studies. Given the strength of the correlation between Polity IV scores and difference-­managing referendums, I concentrate on the finding that higher levels of po­liti­cal freedom lead to higher chances of this type of referendum being held. Throughout the period from 1987 when the Philippines became a democracy again, ­there have been a number of referendums on regional government, for example in 1998  in the Cordillera Autonomous Region (it was rejected), in Mindenao in 2001 (it was approved), and in Shariff Kabunsuan, Dinegat Islands, Quezon del Sur, and in Daveo, in respectively, 2006 (two referendums), 2008, and 2013. All of ­these ­were approved. The votes took place at a time when the Philippines scored eight on the Polity IV Scale.

Difference-­Managing Referendums 39

­Table 2.1. Difference-­Managing Referendums, 1931–2019 Country



Yes (%)

Polity IV


Spain Spain

Catalonia Basque Country Galicia Northern Ireland Jura Mayotte St. Pierre Aruba Australian Capital Territory Virgin Islands Basque Country Catalonia Northwest Territory Greenland Scotland Wales Andalusia Galicia Guam Palau Palau Falklands Suriname New Caledonia Cordillera Canada Puerto Rico Curaçao Virgin Islands Curaçao Palau Puerto Rico Crimea Bonaire

1931 1933

99 99

8 8

1 1

1936 1973

96 99

8 10

1 1

1974 1976 1976 1977 1978

51 99 92 82 78

10 9 10 10

0 0 0 0 0

1979 1979

44 94

10 9

1 1

1979 1979

88 40

9 10

1 1

1979 1979 1979 1980 1980 1982 1983 1983 1986 1987 1987

73 52 21 93 77 49 56 56 96 93 98

10 10 10 9 9 10 10 10 10 3 10

1 1 1 1 0 0 0 0 0 1 1

1990 1992 1993 1993 1993 1993 1993 1993 1994 1994

26 45 48 18 81 73 68 48 78 89

10 10 10 10 10 10 10 7 10

Spain United Kingdom Switzerland France France Netherlands Australia United States Spain Spain Canada Denmark United Kingdom United Kingdom Spain Spain United States United States United States United Kingdom Suriname France Philippines Canada United States Netherlands United States Netherlands United States United States Ukraine Netherlands

0 0 1 0 0 0 1 0 0 (continued)


Chapter 2

­Table 2.1. (continued) Country



Yes (%)

Polity IV


Netherlands Netherlands

Saba Saint Eustatius Saint Marten Scotland Wales Northern Ireland Puerto Rico Nunavut New Caledonia Northern Territory Puerto Rico Cordillera Mayotte Saint Maarten Zamboanga Mindanao Gibraltar Guadeloupe Cyprus Taiwan Saba Albanians Burundi Curaçao Saint Eustatius Catalonia Shariff Kabunsuan Dinegat Algeria Andalusia Greenland Quezon del Sur Santa Cruz Mayotte Martinique

1994 1994

86 90

10 10

0 0

1994 1997 1997 1998

59 74 50 73

10 10 10 10

0 1 0 0

1998 1998 1998

50 54 71

10 10 10

0 1 1





1998 1998 2000 2000 2001 2001 2002 2003 2004 2004 2004 2004 2005 2005 2005

46 39 72 69 70 81 1 72 24 91 86 95 79 68 76

10 8 10 10 8 8 10 10 10 10 10 9 7 10 10

0 1 0 0 1 0 0 1 1 0 1 0 1 0 1

2006 2006

78 98

10 8

0 0

2006 2007 2007 2008 2008

52 98 87 76 43

8 6 9 10 8

0 1 0 0 0

2008 2009 2010

80 61 20

8 10 10

0 0 0

Netherlands United Kingdom United Kingdom United Kingdom United States Canada France Australia United States Philippines France Netherlands Philippines Philippines United Kingdom France Cyprus Taiwan Netherlands Macedonia Burundi Netherlands Netherlands Spain Philippines Philippines Algeria Spain Denmark Philippines Bolivia France France

Difference-­Managing Referendums 41

­Table 2.1. (continued) Country



Yes (%)

Polity IV


Netherlands United Kingdom Kosovo United States Philippines Netherlands

Bonaire Wales North Kosovo Puerto Rico Davao Saint Eustatius La Manga Bonaire District of Columbia Veneto Lombardy Macedonia Bangsamaro Sidama

2010 2011 2012 2012 2013 2014

87 63 1 61 76 65

10 10 8 10 8 10

0 1 0 0 1 0

2015 2015 2016

64 34 86

6 10 10

0 0 0

2017 2017 2018 2019 2019

98 95 94 88 98

10 10 9 8 3

1 0 0 1 1

Ec­ua­dor Netherlands United States Italy Italy Macedonia Philippines Ethiopia

­ able 2.2. Regression Models of Difference-­ T Managing Referendums Variables in the Equation Support Competition Polity IV Year Constant N: 103. Nagelkerte R 2: .65 *Significant at .05.

B .022 (.056) −1.06 (2.33) .85* (.409) .041 (.039) −84.3 (79.3)


Chapter 2

Mean Referendum

Simple Error Bar Mean of Referendum by PolityIV 2.00 1.00 0.00 –1.00 –10.00





PolityIV Error Bars: 95% CI

Figure 2.1. Error Bars for Difference Managing Referendums and Polity IV Level of Democ­ratization.

However, ­there have been no difference-­managing referendums since Rodrigo Duterte took office in 2016, a­ fter which the country arguably became more authoritarian. Yet, despite the demo­cratic tendency, it remains the case that the referendums in the Philippines are “focused on government control, and that the use of direct democracy is used to create a mandate for the government” (Reilley 2018: 112) As we have seen, some of the most prominent difference-­managing referendums have been held in demo­cratic countries. Thus, in Canada the Charlottetown Agreement was submitted to a vote in 1992. And in Spain between 1979 and 1980—­and ­later in 2006 (Catalonia) and 2007 (Andalusía)—­ schemes of difference-­managing policies ­were submitted to voters, as w ­ ere, as we have seen, vari­ous policies regulating the relationship between the center and the periphery in the United Kingdom. Apart from a large number of polls in the Philippines, schemes of difference-­ managing policies in countries with lower levels of democ­ratization ­were not generally submitted to referendums. The devolution granted to Tobago in 1980 was not submitted to a referendum, nor have the inhabitants of Aceh in Indonesia been granted a right to vote on a scheme of home rule (Ross 2005: 35). This is consistent with the statistical tendency and the hypothesis from the beginning of this book, to wit: difference-­managing referendums tend to be held in countries with high levels of democ­ratization. So what, then, explains the outliers? What accounts for the demo­cratic countries that did not hold difference-­managing referendums? Given that demo­cratic countries tend to hold devolution referendums, one might ask

Difference-­Managing Referendums 43

why the self-­government for Südtyrol (Alto Adige) in Italy was not submitted to a referendum when this was established by the Austro-­Italian Treaty in 1971? The answer is perhaps that the doctrine that such ­matters have to be ratified by referendum had not yet taken root. It is certainly in­ter­est­ing that the proposed federalization of Italy in 2006, which 61 ­percent of voters rejected, was submitted to a referendum (Pinelli 2006)—­though, strictly speaking, it is questionable if this was a difference-­managing referendum as defined in this book. But it is also pos­si­ble that no referendum was held ­because t­ here was no strategic reason for holding one. Th ­ ere was no electoral threat to the governing parties at the time, save from the—in Giovanni Sartori’s sense—­“irrelevant parties” (J. Evans 2002: 157). Hence, the incentives for holding a referendum ­were—­consistent with the competition proximity model—­limited. We have already looked at the example of the United Kingdom, where it has become a convention of the constitution that policies of difference management are submitted to referendums. But is ­t here a reason why this norm seems to have been established in countries that have a high level of demo­ cratic and civil freedoms such as Spain and Canada? ­There is some anecdotal evidence that the decision to hold referendums on devolution in Spain in the post-­Franco era was a deliberate attempt to break with the authoritarian past (Gunther, Sani, and Shabad 1988: 249). Similarly, in Canada the decision to hold a referendum on the Charlottetown Agreement was a deliberate attempt to show that the perceived elitism of the failed Meech Lake Agreement—­which failed ­after a filibuster in the Manitoba Assembly—­could be remedied only by the ­people’s direct involvement in a referendum. In the words of Ronald Watts, “The Constitution does not require a referendum for ratification of constitutional amendments, but during 1992 the conviction developed among the negotiating governments that popu­lar support indicated by a favorable consultative referendum result would facilitate ratification by the required legislatures by giving the proposals po­liti­cal legitimacy” (Watts 1999: 5). ­There are several examples that stand out and call for an explanation, though some more than ­others. The fact that Burundi held a referendum on a power-­sharing agreement in 2005 would appear to be an anomaly. Yet with a relatively demo­cratic system (Polity IV score of seven), the African country is not a clear anomaly. Having emerged from a period of authoritarian rule (the country had a Polity IV score below zero in the 1990s), Burundi is not much dif­fer­ent from Spain in 1979, when the latter country held a series


Chapter 2

of difference-­managing referendums and was rated with a Polity IV score of nine. The fact of the m ­ atter, as far as Burundi is concerned, is that its political system was in the pro­cess of becoming more demo­cratic as a result of the agreement that had been reached between the Tutsi and Hutu groups u ­ nder the transitional government of Domitien Ndayizeye (Basedeau 2011: 205). The same argument holds true for the referendums in Suriname in 1987 and in the Philippines in 1989 and 1990 (though t­ hese w ­ ere not mainly ethnonational referendums and ­were more focused on major constitutional changes a­ fter the fall of dictatorships). In the latter case, the fall of the Marcos regime opened up a demand for a resolution of the territorial and regional disputes that had been kept ­u nder wraps during the years of authoritarian rule and u ­ nder the policies of difference elimination that had been initiated with the referendum in 1977 (Machado 1978: 202). The prob­lem with the 1990 referendum in Cordillera in the Philippines was that Filipinos ­were split along religious lines: “While the four Muslim-­dominated provinces ­were in favour all the nine Christian-­dominated provinces ­were against” (He 2002: 75). This split arguably exacerbated the differences between the two communities. The fact that the Philippines was relatively in a pro­cess of democ­ratization (Polity IV score of eight in 1990) and the other examples given ­here support the auxiliary hypothesis, that difference-­eliminating referendums may occur in countries that do not have the highest democ­ratization scores if the country is in the pro­cess of becoming a democracy. The same is arguably true for Suriname (MacDonald 1988: 105). Suriname—­like the Philippines, Burundi, and, indeed, Spain in 1979—­was a country in transition, although its democracy score was rather low (Polity IV score was three). In the 1980s the country had scored a dismal minus six on the index. This was improving at the time. To be sure, the pro­cess of democ­ratization in Suriname was uneven and marred by setbacks, yet the country was in the pro­cess of becoming a polyarchy, which culminated in 1996, when the country scored six on the Polity IV index. I thus have to modify the original hypothesis to this: “difference-­managing referendums occur in countries with regionally based ethnic differences that are demo­cratic or demo­cratizing.” This model, as w ­ ill be seen, is consistent with the methodological criterion that the auxiliary hypothesis (in this case the pro­cess of democ­ratization) is universal and in­de­pen­dently testable. But the dynamic of the referendums can sometimes go deeper than statistical analyses can reveal. My ambition in this book was—­and is—to write

Difference-­Managing Referendums 45

a po­liti­cal science treatise based on a positivist methodology, in which patterns of h ­ uman be­hav­ior are laid down in precise mathematical formulas, which can be tested empirically in the manner known from Newtonian physics. The prob­lem is that the world—­while it is not completely random—is a bit more complicated than that. A case study of Catalonia in Spain ­will show some of t­ hese dynamics.

Difference-(Mis)managing Referendums: A Case Study of Catalonia Spain has a long tradition for holding referendums. And not always a very distinguished one at that. In the 1920s, the dictator Miguel Primo de Rivera—­a man described as having “encyclopedic incompetence” and who suffered from “a hunger for crowds” (Preston 2019: 185)—­wished to be reaffirmed in his self-­perceived greatness and consequently held a rigged plebiscite on a new constitution in 1926. This abuse of direct democracy did not mean that the referendum was somehow tainted when Spain became a democracy in 1931. During the eight years of democracy, t­ here w ­ ere referendums on devolution for Catalonia (1931), the Basque Country (1933), and fi­nally Galicia (1936). While Francisco Franco held dubious plebiscites during his dictatorship (1939–1975), the referendums w ­ ere not tainted by association when Spain returned to normalcy. ­There ­were votes on democ­ratization in 1976 and on a new constitution in 1978. And, in an echo from the 1930s, the voters in the three “historic nations” mentioned in the previous paragraph (i.e., Catalonia, the Basque Country, and Galicia) ­were invited to participate in referendums on new statutes of autonomy in the late 1970s and in the early 1980s. Once again, Catalonia was in the forefront of the developments. The reasons for this w ­ ere manifold. Unlike the Basque Country, Catalonia did not have a violent separatist movement seeking outright in­de­pen­dence through terror. Moreover, support for secession was bordering on being non­ex­is­tent in the 1970s. A second reason for the referendums was the person of Adolfo Suárez. The new prime minister was a careerist from Franco’s movement, but also a man who realized that Spain had no choice but to become a democracy. True to his pragmatist philosophy, Suárez met with Josep Tarradellas, the


Chapter 2

leader of the exiled Generalitat de Catalunya, and agreed on a re-­establishment of the 1931 statute subject to its endorsement by a vote by the p ­ eople (Preston 2019: 486). At the time the referendum was held, the vast majority had already endorsed the new constitution, which confirmed the unity of the country—­ albeit in somewhat ambiguous terms. In the 1979 referendum, no less than 88 ­percent of the voters in Catalonia supported the new estatuto. The only concern was the low turnout. Only 59 ­percent both­ered to participate in the referendum. The following years ­were relatively calm. Indeed, Spain was seen as a model for how to deal with nationalist movements. And the case was explic­itly cited by the L ­ abour government led by Tony Blair in 1997 as an example of how to subdue separatist movements. That is not what happened. As we ­shall see elsewhere in this book, Scottish nationalism grew a­ fter the nation was granted its own parliament a­ fter a referendum in 1997 (Keating 1998). In Spain it was much the same. But it was not inevitable that ­things turned out the way they did. Indeed, unlike in Britain, t­ here was a kind of truce—­ even a marriage of convenience—­between the center-­right Partido Popu­lar (PP) and the parties that sought more autonomy for Catalonia. Thus in 1996, the new PP government relied on the votes of the center-­right Catalan party Convergència Democràtica de Catalunya (CDC). History can be odd in hindsight. The compact between the two parties—­k nown as the Pact of ­Hotel Majestic (Pacto del Majestic) was negotiated by Jordi Pujol the head of CDC and Mariano Rajoy—­who ­later became an uncompromising hardliner when he became prime minister a de­cade and a half ­later (discussed ­later). The Catalans ­were able to double the tax revenue from 15 to 30 ­percent—­a nd every­one seemed happy (Preston 2019: 524). But when the Socialist Pasqual Maragall and co­a li­tion partners won a majority in the Catalan assembly in 2003, and when the Spanish ­Labor Party won the election in 2004, the new prime minister too had to rely on support from the Catalans. The situation was not that dif­fer­ent from the one that had faced José María Aznar, the PP prime minister in 1996. But the new leader of the PP, Mariano Rajoy, immediately began to obstruct any rapprochement between Madrid and Barcelona. Much of what ­Labor Party prime minister José Luis Rodríguez Zapatero enacted was progressive and went against the conservative preferences of the PP base. For example, the requirement that 40 ­percent of a com­pany’s

Difference-­Managing Referendums 47

board members must be female (if the com­pany has more than 250 employees), laws against domestic vio­lence, enactment of a law on marriage equality—­ and, for the pre­sent purposes, the most impor­tant reform, a new statute governing Catalonia (Preston 2019: 547)—­was anathema to many who voted for PP. The PP sought to use this opportunistically to solidify its electoral base. But t­ here ­were also other sinister reasons. The new statute was approved by all parties in the Cortes—­apart from PP. It was also endorsed by 74 ­percent of the voters in Catalonia in a referendum on 18 June 2006. At this stage only 14 ­percent of Catalans supported in­de­pen­dence (Centre d’Estudis d’Opinió 2006). It is often reported in the media that Catalonia’s demand for more autonomy is based on its higher level of prosperity. In the run-up to a referendum in 2017, for example, the other­wise well informed Financial Times put the demand for separatism down to the “nation’s prosperity”: “On its own, Catalonia’s output corresponds to about 19 per cent of Spain’s gross domestic product” (Financial Times 28 September 2017), the paper reported ­under the headline “Catalonia’s Economic Strength Fuels In­de­pen­dence Push.” To be sure Catalonia is a rich part of Spain. And it might seem unfair that a region with 16 ­percent of the population contributes 21 ­percent of the tax revenues and only receives 8 ­percent of the public sector investments. Yet the same is true for other parts of the country. As Paul Preston has noted, “in fact, a comparable in­equality prevailed in Spain’s richest region, Madrid” (Preston 2019: 555). What originally contributed to the anger was po­liti­cally motivated. The PP played hardball opportunistically and challenged the new statute in the courts, which declared parts of it unconstitutional in an 881-­page ruling in 2010 (Tribunal Constitucional 16 June  2010). Without ­going into the jurisprudence of the judgment, the use of litigation as a po­liti­cal tool fueled the flames of discontent. In 2011 support for in­de­pen­dence r­ ose to 34  ­percent (Centre d’Estudis d’Opinió 2012). (I return to the jurisprudential issues in Chapter 4.) It is tempting to see the hardening of opinion as a harbinger of ­t hings to come in other places if separatist sentiments are blocked—­Scotland immediately comes to mind for a Brit! Yet, it should not be overlooked that the Catalonian case was part of a proverbial perfect storm. The austerity policies enacted in the wake of the 2008 financial crisis caused resentment in Barcelona that added to an already tense situation. Add to this that Artur Mas,


Chapter 2

the CDC politician who became president of the Generalitat de Catalunya in 2010, like Rajoy in Madrid, had to b ­ attle corruption charges, and it is clear that the two protagonists had personal reasons for diverting attention away from themselves and onto other ­matters. The chicken-­game-­like situation led to a further increase in support for in­de­pen­dence. By the end of 2012, ­t hose favoring secession could muster the support of 44 ­percent of the electors (Preston 2019: 555). It needs to be added, though, that it never r­ ose significantly above this. Mas and the Catalans wanted negotiations. They ­were greeted by a collective shrug from Madrid. In response, the Catalan president proposed a referendum in 2014. When this proposal was rejected by a vote in the Cortes in April 2014, Mas proposed a consultation. However, this was declared ultra vires by the Constitutional Tribunal. When the poll went ahead regardless, it was managed by volunteers. The recorded turnout was a respectable 37 ­percent in the “vote” on 9 November 2014. The vast majority—­just over 80 ­percent—­opted for “in­de­pen­dence.” In response, Madrid used its favored option, the weaponization of the courts. Mas was charged with “misuse of public funds and ­later found guilty” (Preston 2019: 556). Certainly, the Catalan president was legally on thin ice. But his position was hardly criminal. Further, the Madrid government did not shy away from using dirty tricks. It was ­later revealed that “the PP government, with the knowledge and encouragement of Rajoy, had been ­r unning covert and illegal operations to investigate and smear Catalan politicians” (Preston 2019: 557). ­W hether this worked is an open question. In any case, the regional elections in Catalonia w ­ ere far from impressive for the separatists. Junts pel Sí—­a newly minted alliance of secessionist parties—­won 39 ­percent of the vote, but that was still enough to win a majority of the seats. They claimed, somewhat unconvincingly, that this was a mandate for a vote on in­de­pen­ dence. But their proposal was immediately struck down by the courts—­and on solid l­egal grounds that could not be regarded as merely po­liti­cal (Buck 2019: 49) From an external point of view—­and perhaps even from Barcelona—it seemed that the discussion in Spain revolved entirely around the Catalan crisis. But that was far from being the case. In fact, it could be argued that the issue of Catalonian in­de­pen­dence was a welcome distraction from the scandals that consumed the Madrid elite—­including that of the prime min-

Difference-­Managing Referendums 49

ister and even the king. In politics every­thing is context, and no developments can be divorced entirely from the milieu in which they take place. Titillating stories about King Juan Carlos’s liaison with the Danish businesswoman Corinna Larsen, the fraud investigation of the king’s son, and Rajoy’s involvement in a corruption case (that ­later led to his downfall) meant that “the festering crisis in Catalonia did not create the alarm that it might have done b ­ ecause media attention was more concerned with the links between the Head of State and corruption” (Preston 2019: 558). While all this was happening, support for in­ de­ pen­ dence ­ rose to 47 ­percent (Centre d’Estudis d’Opinió 2016). Following a conviction for illegal use of public funds, Artur Mas was replaced by Carles Puigdemont who was—if anything—­even more uncompromising. The new president called yet another referendum on 9 June 2017. He sought external support; he appealed to the Eu­ro­pean Parliament. He told the parliamentarians in Strasbourg that the Catalans aimed at “achieving in­de­pen­dence peacefully, in a civic way, and armed exclusively with democracy.” He further stressed that the Catalans w ­ ere willing to compromise but that if the Spanish government would not give way, he would “hold it anyway” (Puigdemont quoted in Guidi and Casula 2019: 185) However, while claiming to be committed to a compromise, he did not open the possibility of anything short of holding the referendum; for example, reinstating the original 2006 Statute, or the possibility of a constitutional amendment that could legalize an a fortiori vote on in­de­pen­dence ­under a new constitutional framework. Not every­one was happy with this strategy. The escalation of tensions within the Catalan government led to the sacking of Jordi Baidget on 3 July. The minister of enterprise, a moderate politician, had suggested Puigdemont’s confrontational tactic. And three days ­later, a further three ministers w ­ ere shown the proverbial door for questioning the wisdom of demanding a referendum on secession. The referendum was declared illegal by the courts. But the Madrid government also used other techniques, not all of which ­were welcomed by international opinion. Ten days before the referendum, an estimated 10 million ballot papers ­were seized by the Guardia Civil. And on the same day, 20 September, fourteen public officials, including the finance secretary, w ­ ere arrested. In a show of force that appeared heavy-­handed, the Madrid government deployed sixteen thousand police officers to interrupt the vote.


Chapter 2

This notwithstanding, the referendum went ahead on 1 November 2017. The vast majority voted “yes,” albeit in a very low turnout of 43  ­percent (almost precisely the percentage that supported secession). A few weeks ­later the Catalan Parliament declared in­de­pen­dence. In response, the Spanish government imposed direct rule. Puigdemont fled to Belgium and the Spanish government refused negotiations. The Madrid government received support from other capitals, while several ministers ­were charged with sedition. This notwithstanding, the separatist parties won a slim majority of parliamentary seats (70 out of 135 seats) in the Catalan election on 21 December 2017. But they fell short of a majority in the popu­lar vote by securing only 47.6 ­percent. Both sides claimed a mandate. The situation was a classic example of Murphy’s Second Law, all that cannot go wrong goes wrong. Or, in the words of Paul Preston, the doyen of studies of modern Spain, The culpable mishandling of the Catalan situation by both Rajoy and Puigdemont could lead only to disaster. Puigdemont’s recklessness would allow Rajoy to bolster his own position by taking a strong Anti-­Catalan position instead of seeking to reach out to the majority of Catalans who w ­ ere not committed to outright separatism. Hoping to exploit Anti-­Catalanism that had been nurtured over the previous de­cade, Rajoy gambled that a hard line would boost the PP’s popularity in the rest of Spain. In d ­ oing so he was ignoring the pattern in the last hundred years of Spanish history where Catalan separatism has fed of Madrid’s centralist intransigeance. (Peston 2019: 564) The Spanish—­and Catalan—­experiences with difference-(mis)managing referendums failed to resolve the prob­lem b ­ ecause Madrid was more interested in playing hardball than in resolving a legitimate feeling of cultural distinctiveness. The Catalans, on their side, took the bait, and sadly, the ­people of Spain and Catalonia ­were caught in the ­middle. To use the referendum to manage differences requires that you accept and recognize ­these as legitimate, and that a solution is sought which is based on reconciliation and compromise. The latter two words ­were in short supply in Spain in the first two de­cades of the twenty-­first ­century. It is tempting to quote the last lines of George Orwell’s Homage to Catalonia and conclude that “in such circum-

Difference-­Managing Referendums 51

stances ­there can be no argument; the necessary minimum of agreement cannot be reached” (Orwell 2000: 247). Box 2.1. New Caledonia: The Referendum as a Mechanism for Defusing Tensions? In the 1970s, the Kanak in­de­pen­dence movement was launched in New Caledonia. It drew steady support from other countries in the region and grew stronger in the 1980s. In 1984 the Front de Liberation Nationale Kanak et Socialiste (FLNKS) was founded as an umbrella organ­ization for the pro-­independence parties, and ­later that year FLNKS established a provisional in­de­pen­dent government. However, tensions continued to grow, and between 1984 and 1988 around eighty p ­ eople died in clashes and terrorist attacks (R. Aldrich 1993). In response to this the French government initiated negotiations with a view to a settlement. ­These ­were concluded with the Matignon Accords in 1988 among FLNKS, the loyalist Rassemblement pour une Calédonie dans la République (RPCR), and the French government. The Matignon Accords provided for greater local autonomy (including provincial governments) and substantial aid designed to redress deep inequalities between the French and Kanak communities, while committing the territory to a self-­determination referendum ten years ­later. The Matignon Accords—­negotiated by the French prime minister M. Michel Rocard—­were approved in a referendum in 1988. They ­were supported by 88 ­percent of the voters. In 1998 the three Matignon Accords partners agreed on a new statute defining the territory’s institutions and its relations with France. The agreement, termed the Nouméa Accord, which steered a m ­ iddle course between the respective po­liti­cal aspirations of the RPCR and FLNKS, and avoided the need for a divisive yes-no referendum on in­de­pen­dence. It was signed on 5 May 1998, during a visit to New Caledonia by the French prime minister Lionel Jospin and approved by 72 ­percent of New Caledonians in a referendum on 8 November 1998. The accord was subsequently ratified by the French National Assembly and Senate. As a result, New Caledonia is no longer a French overseas territory but has its own special status as a collectivite within the French Constitution (Bensa 2003). Yet the situation in New Caledonia is far from resolved. As we ­shall see in the next chapter, the devolution of powers did not quell the demands for in­de­pen­ dence among the Kanaks. In 2018, as stipulated by the Nouméa Accord, ­t here was a vote on in­de­pen­dence, which narrowly lost: 56 ­percent voted against the proposal, but in the areas where ­t here was a Kanak majority, the voters opted for severing the ties with France (Chauchat 2019). The inconclusive result in 2018 led to another vote in 2020, when 53 ­percent voted against. In the case of New Caledonia, the granting of devolution has not quelled the quest for in­de­pen­dence.


Chapter 2

Conclusion Since the 1980s difference-­managing referendums have tended to occur in industrial countries with high Polity IV scores. But why is this the case? Some would argue that referendums of this kind have been held not out of self-­interest (or the logic of consequentiality) but as a result of an understanding prevalent in demo­cratic socie­ties that issues that concern the sharing of power are so fundamental that they must be submitted to the ­people for their approval. In the words of He, “democ­ratization has facilitated conditions u ­ nder which referenda can be a­ dopted and accepted as a means of settling [questions]. The ideas of popu­lar sovereignty and the referendum princi­ple emerge when countries embark upon democ­ratization” (He 2002: 71). And, yet, not all countries—­not even demo­cratic ones—­have shown this willingness to compromise. The situation in Spain—­especially in Catalonia—­ shows how referendums can lead to conflict and exacerbate differences if they are used as a party-­political tool. In this par­tic­u­lar case, the Partido Popu­lar’s refusal to accept the outcome of the 2006 devolution referendum on a new statute for Catalonia led to confrontation and brought Spain to the abyss. Box 2.2. The Establishment of the Swiss Canton of Jura in the 1970s Difference-­managing referendums can be carried on almost ad absurdum. This was perhaps the case in Switzerland in the 1970s. Between 1974 and 1978 a series of referendums took place to determine if the French-­speaking minority in the canton of Bern should be allowed to form their own new canton (Laponce 2001). The prob­lem had started many years before at the Congress of Vienna—­where precious ­little thought was given to national and ethnic groups and where the powers ­were keen on restoring the pre-­ Napoleonic order (Kissinger 1957). The conservative, mainly Catholic, and French-­speaking Jura—­once an “autonomous po­liti­cal entity”—­was joined with the liberal, German-­speaking, and Protestant canton of Bern (Buechi 2012: 187). ­After several unsuccessful attempts to establish a new canton—­ including a failed citizen initiative in 1959—­t he cantonal government in Bern de­cided to put the ­matter to a referendum (Buechi 2012: 189). Initially the entire population of the canton was asked in 1973 to endorse—or other­w ise—­t hat a referendum could be held in the enclave of Jura (the area of the proposed new canton). So, basically, a referendum on a referendum. The Bern administration was ­under pressure, but the “expectation was that, as in 1959, a majority would vote against the creation of the

Difference-­Managing Referendums 53

Box 2.2. (continued) new canton of Jura” (Buechi 2012: 189). The result of this poll was a narrow majority in support of the establishment of a new canton—51 ­percent voted yes, with 88 ­percent turnout (turnout in Swiss referendums is normally below 50 ­percent) (Lutz and Marsh 2007). However, the referendum showed a split between the French-­speaking Protestants in the southern part of Jura, who wanted to remain part of Bern. But support was strong among the French-­speaking Catholics, who ­were in ­favor of establishing a new canton, and thus in f­ avor of splitting from their German-­speaking co-­religionists. To resolve the impasse, “citizen-­initiated referendums in favour of remaining in the canton of Berne w ­ ere now organised [in the three southern districts]” (Buechi 2012: 191). This second referendum rejected the proposition to create a new canton. The net result was that the three northern districts of Jura formed a new canton, whereas the three southern districts remained part of Bern. ­After the vote, a further thirteen citizen-­initiated local votes ­were held along the new cantonal border; “majority Protestant districts voted to remain with Berne [sic] and 8 majority Catholic districts opted for the Jura” (Buechi 2012: 191). As if this w ­ ere not enough, the establishment of the new canton of Jura was ratified in a federal referendum, and ­t hose in the new canton voted for a new cantonal constitution in a referendum. So why did the authorities hold a referendum? ­There is some suggestion that the initial vote in 1973 was held for strategic reasons that ­were not inconsistent with the competition proximity model (Mayer 1968). However, it is not a neat fit, and in any case the citizen-­initiated referendums that followed ­were held for other reasons, namely in order to get recognition for dif­fer­ent minorities’ rights. The model does not account for citizen-­initiated votes, nor does it account for constitutionally mandatory referendums. But this limitation in the theoretical model notwithstanding, the referendum worked; the Jura issue that was created by elitist bargaining in 1815 was resolved by citizen-­initiated referendums in the 1970s.


Secession and Partition

When in the Course of ­human events, it becomes necessary for one p ­ eople to dissolve the po­liti­cal bands which have connected them with another, and to assume among the powers of the earth, the separate and equal station to which the Laws of Nature and of Nature’s God entitle them, a decent re­spect to the opinions of mankind requires that they should declare the c­ auses which impel them to the separation. We hold ­these truths to be self-­evident, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights, that among t­ hese are Life, Liberty and the pursuit of Happiness.—­T hat to secure t­ hese rights, Governments are instituted among Men, deriving their just powers from the consent of the governed. —­U.S. Declaration of In­de­pen­dence, 4 July 1776


­ ere is a reason that the evocative words penned by Thomas h Jefferson—­imperfect man though he was—­have resonated through the ages. The words seem to sum up the aspirations of e­ very ­people seeking statehood. This chapter ­will—­after some historical introductions—­ consider when referendums are held, and when they are successful—by which I mean, when they result in the establishment of a new state. I am making no value judgment as to w ­ hether establishing a new state is a desired outcome. First a few definitions. Using Weberian ideal types, one can distinguish, in the main, three types of referendums,

Secession and Partition 55

• Ad hoc referendums (questions to solve a perceived po­liti­cal issue—­ such as David Cameron’s decision to hold a referendum on UK membership in the Eu­ro­pean Union 2016); • Initiatives (votes initiated by a specified number of electors on (1) already enacted legislation, as in Switzerland, or (2) on new laws, as in Hungary), and; • Constitutional referendums (see next paragraph). The constitutional doctrine normally distinguishes between three types of constitutional referendums: on the approval of the constitution, on its revision, and on sovereignty issues (like the foundation of a new state). This chapter focuses on the latter. It is impor­tant to stress that ­t hese are ideal types and that ­t here might be borderline cases. Thus, the types are not what phi­los­o­phers call “universals.” Hence, the concepts are approximations to real existing examples. In philosophical terms I thus follow a nominalist or nonessentialist approach (Goodman and Quine 1947). Overall, the number of ­t hese referendums is comparatively small. Thus, out of the twelve hundred nationwide only sixty-­two (or 5 ­percent) have pertained to in­de­pen­dence, of which only four (or 6 ­percent) have returned a “no” vote (Quebec, in respectively, 1980 and 1995, Scotland in 2014, and New Caledonia in, respectively, 2018 and 2020)—­t hough other referendums have failed b ­ ecause they did not satisfy supermajority requirements (e.g., in Nevis in 1998 and in several referendums in Palau in the 1980s). Historically the idea and the doctrine of in­de­pen­dence referendums can be traced back to the beginning of the seventeenth c­ entury. No less a figure than Hugo Grotius (1583–1645) observed in The Law of War and Peace, “In the alienation of a part of sovereignty, it is required that the part which is alienated consent to the act” (Grotius 1853: 108). Samuel Pufendorf (1632–1694), the other g­ reat theorist of international law, was even more explicit when he stressed that “in the alienation of a part of the kingdom, t­ here is required not only the consent of the p ­ eople which continues to be with the old king, but the consent of that part too, especially, whose alienation is at stake” (Pufendorf quoted in Phillipson 1912: 234). And the third of the g­ reat l­egal theorists of sovereignty, Emer Vattel (1714–1767), roughly one hundred years ­later, held that such fundamental changes had to be supported by the p ­ eople “by a majority of votes” (Vattel 2009: 118).


Chapter 3

At the time of the French Revolution, it was acknowledged, at least at the theoretical level, that the ­people had a right of veto over territorial changes. Hence, it was not surprising that the Constituent Assembly in Paris passed a decree declaring that “the French nation renounces any war of conquest” (Duvergier 1824: 191, my translation). In practice, though, the French w ­ ere reluctant to put their subsequent conquests to a vote. The reasons for this ­were understandable. In 1802, the French or­ga­nized a constitutional referendum in Switzerland after the country had been conquered by Napoleon. A majority of the voters voted “no.” This unexpected outcome, perhaps, explains why the French henceforth ­were reluctant to or­ga­nize sovereignty referendums in conquered territories. None of ­t hese ­earlier referendums ­were ­free or fair, and the votes ­were not secret. One can, therefore, question if they are relevant for the purposes of this chapter. As mentioned in Chapter 1, the first referendums on in­de­pen­dence, in the form we know ­today, ­were held in the 1860s, when the U.S. states of Arkansas, Tennessee, Texas, and V ­ irginia held referendums on in­de­pen­dence following the election of Abraham Lincoln to the U.S. presidency. All the referendums w ­ ere won but no country recognized the results (Mattern 1921: 118–123). The result of the referendum in V ­ irginia is particularly noteworthy. Before the vote, representatives from counties in West ­Virginia declared that they, in the event of a “yes” vote for in­de­pen­dence, would establish a new state and that the constitution of this new state would be approved by the voters in a referendum (Mattern 1921: 118). ­Virginia as, we have seen in Chapter 1, on the ­whole voted for secession. However, in the western counties 8,375 out of the 9,758 votes cast ­were against secession (Mattern 1921:120). Subsequently, ­these counties sent delegates to a specially convened convention, which declared that the referendum in ­Virginia was “illegal, unoperative, null, void and without force and effect” (quoted in Mattern 1921: 123). They then “passed an ordinance providing for the ‘formation of a new state out of the portion of the territory of this state [­Virginia].’ This ordinance was to be and was submitted to a plebiscite” (Mattern 1920: 123). Eigh­teen thousand voted for a new state, 781 voted against it (cited in Mattern 1921: 123). A ­ fter the American Civil War the U.S. Supreme Court established in Texas v. White that unilateral declarations of in­de­pen­dence are unconstitutional, a case that was most recently used by the Alaskan Supreme Court in 2006 in Kohlhaas v. Alaska to ban a constitutional initiative for in­de­pen­dence for this state. Figure 3.1

Secession and Partition 57

30 25 20 15 10 5

18 60 18 70 18 80 18 90 19 00 19 10 19 20 19 30 19 40 19 50 19 60 19 70 19 80 19 90 20 00 20 10


Number of Referendums

Figure 3.1. Referendums on In­de­pen­dence, 1860–2018. Note: This figure does not include the four multioption referendums in Puerto Rico (1967, 1993, 1998, and 2012) that formally included “in­de­pen­dence” as one of the options. However, the figure includes the two-­round multioption referendum in Newfoundland in 1948 as in­de­pen­dence was one of the choices in the runoff. The in­de­pen­dence options lost to “statehood” and the former British territory became a Canadian province.

shows the growth of in­de­pen­dence referendums up to the high point in the 1990s. ­After the American secession votes t­ here was gap of a few de­cades before the aforementioned Norwegian referendum, then a hiatus again u ­ ntil the mid-1930s, when the number of in­de­pen­dence referendums began to pick up with the unrecognized, but successful, in­de­pen­dence referendum in Western Australia in 1933—as we saw in Chapter 1—68 ­percent voted in ­favor, but the vote was ignored as the secessionist party lost the state election on the same day—­and the vote for in­de­pen­dence for the Philippines in 1935. Especially a­ fter the Second World War, the referendum was increasingly used to show popu­lar approval for decolonization, though not all countries held plebiscites before they broke with their erstwhile colonial overlords ­After a drop in the 1970s, t­ here was an explosion of in­de­pen­dence votes in the years immediately following the fall of the Berlin Wall and the collapse of Soviet Communism. ­Table 3.1 lists the referendums on in­de­pen­dence held a­ fter 1990.


Chapter 3

­Table 3.1. Secession Referendums, 1990–2020 Parent Country

Seceding Country


Turnout (%)

Yes Vote (%)

Soviet Union Soviet Union Soviet Union Soviet Union Soviet Union Georgia Georgia Yugo­slavia Croatia Yugo­slavia Soviet Union Bosnia Serbia Serbia Soviet Union Soviet Union Soviet Union Macedonia Moldova Rus­sia Yugo­slavia Yugo­slavia Georgia Bosnia Ethiopia Bosnia United States United States Georgia Quebec Canada St. Kitts and Nevis United States Indonesia Somalia New Zealand Yugo­slavia Sudan United Kingdom Iraq Spain France Papua New Guinea France

Lithuania Estonia Latvia Georgia Ukraine South Ossetia Abkhasia Croatia Serbs Macedonia Armenia Serbs Sandjak Kosovo Turkmenistan Karabagh Uzbekistan Albanians Transnistie Tartarstan Bosnia Montenegro South Ossetia Krajina Eritrea Serbs Puerto Rico Palau Abkhasia Cree Quebec Nevis Puerto Rico East Timor Somaliland Tokelau Montenegro South Sudan Scotland Kurdistan Catalonia New Caledonia Bougainville New Caledonia

1991 1991 1991 1991 1991 1991 1991 1991 1991 1991 1991 1991 1991 1991 1991 1991 1991 1991 1991 1992 1992 1992 1992 1992 1993 1993 1993 1993 1995 1995 1995 1998 1998 1999 2001 2006 2006 2011 2014 2017 2017 2018 2019 2020

91 77 74 98 70 98 99 98 98 70 95 90 96 99 94 NA 98 99 100 82 99 96 NA 99 99 96 48 64 96 95 94 57 50 94 NA NA 86 97 83 72 43 81 87 81

84 83 88 90 85 90 58 83 83 75 90 NA 67 87 97 NA 94 93 NA 62 64 44 NA 64 98 92 73 68 52 75 49 61 71 78 97 95 56 98 44 92 92 43 98 47

Secession and Partition 59

Not all in­de­pen­dence referendums are comparable. Dif­fer­ent historical and ­legal circumstances play a role. Further subdividing the in­de­pen­dence referendums, we can distinguish between three forms (see Şen 2015: 213), namely, 1. Postcolonial (e.g., Philippines 1935); 2. By agreement (e.g., Montenegro and New Caledonia, Bougainville); 3. Unilateral (e.g., Catalonia, Quebec, and Estonia). So, when are ­t hose seeking to establish a new state successful?

Empirical Analy­sis Good research, according to a much-­quoted text, should “make a specific contribution to an identifiable scholarly lit­er­a­ture by increasing our collective ability to construct verified scientific explanations of some aspect of the world” (King, Keohane, and Verba 1994: 15). This chapter makes a “specific contribution” to the study of in­de­pen­dence referendums by analyzing the Scottish in­de­pen­dence vote in 2014 in a comparative perspective. The chapter does not provide a single perspective on the referendum, rather the vote is seen through dif­fer­ent conceptual lenses: a comparative statistical one, an elite perspective, and a grassroots a­ ngle. ­There is a considerable lit­er­a­ture on the determinants of in­de­pen­dence referendums. Most of this is country specific and focuses on idiosyncratic ­factors b ­ ehind a unique event (Conley 1997). While t­ here are some studies that contrast in­de­pen­dence referendums, for example Quebec in 1995 and Montenegro in 2006 (Oklopcic 2012), ­t here are relatively few studies devoted to the comparative study of in­de­pen­dence referendums (though see Dion 1996). A subset of this lit­er­a­ture deals with the determinants of the vote (e.g., Clarke and Kornberg 1996 and Leduc 2002). The aim of this chapter is first to contribute to the comparative psephology of in­de­pen­dence referendums and to determine if ­t hese votes follow a recurrent pattern. And then to look at how t­ hese f­ actors square with the experiences of the Scottish referendum on in­de­pen­dence 2014.


Chapter 3

Methodological Considerations It is the goal of “scientific research” to make “causal inferences on the basis of empirical information” (King, Keohane, and Verba1994: 7). The question is how? It is often objected that quantitative analyses fail to uncover the essential issues. However, statistical data can give us an understanding of general and recurrent patterns, which can direct us t­ oward more qualitative data. Scholars in the broadly qualitative tradition maintain that “we want social science theories to provide causal explanations of events . . . ​and to give an account of the reasons for or meanings of social action” (Ferejohn 1993: 228). While a statistical analy­sis cannot provide us with the latter it is a useful starting point for a subsequent analy­sis of meaning. Consequently, if we find a statistical pattern, this w ­ ill be subjected to a more qualitative analy­sis to determine if t­ here is a congruence between the qualitative and the quantitative analy­sis. Epistemologically, the two perspectives thus constitute two dif­fer­ ent aspects of a social phenomenon. It is recognized that t­ here is no superior epistemological vantage point (della Porta and Keating 2008). The quantitative analy­sis only provides one perspective. It w ­ ill be complemented by a qualitative study. The approach used ­here is thus akin to that of Graham T. Allison (1969), in which dif­fer­ent conceptual models are used to explain the same phenomenon or event. In Allison’s words (1969: 690): “Although the standard frame of reference has proved useful for many purposes, it is now clear that this model must be supplemented, if not supplanted, by frames of references which focus upon the detailed functioning and malfunctioning of organ­izations and individuals in the policy pro­cess.” This alternative approach ­will be pursued following the statistical analy­sis.

Hypotheses Some of the previous studies make generalized claims, for example, that secession is difficult in developed democracies (Dion 1996). Another possibility is that in­de­pen­dence is correlated with a high level of elite consensus (as in Norway in 1905). Altogether ­t here are three distinct hypotheses: • Support for in­de­pen­dence is correlated with elite consensus. • Support for in­de­pen­dence ­will be correlated with higher turnouts. • The yes vote is correlated with the level of democ­ratization.

Secession and Partition 61

Data The data covers the period a­ fter 1990. This cutoff was based on the assumption that in this period ­t here ­were no ideological overlays but also on the fact that this period had a sufficient number of cases to make statistical analy­sis legitimate. As in the previous chapter the level of democ­ratization is based on Polity IV. Data is for the individual referendums on specialist assessment by country experts and is from C2D, Zentrum für Demokratie, Aarau.

Statistical Findings ­ ere is support for the hypothesis that referendums on in­de­pen­dence are Th less successful in countries that are undemo­cratic. For e­ very point down the Polity IV scale the percentage voting for in­de­pen­dence decreases by 1.5 ­percent. A cursory look at the cases ­will bear this out. Countries like South Sudan and Eritrea are very far from being model democracies, and the same goes for Kurdistan. In all t­ hese cases, the rulers w ­ ere—to put it delicately—­able to forge the results to suit their own purposes. This is not to say that, for example, the p ­ eople in South Sudan w ­ ere opposed to in­de­pen­dence, or that the poll in that country in 2011 was rigged. In all likelihood, the vote was a fair reflection of the preferences of the voters, but given the less than pluralistic culture in the country and a press dominated by the government, the voters w ­ ere hardly able to make up their minds in a fair way. It is not surprising, therefore, that they voted for in­de­pen­dence. Conversely, in Catalonia and in Scotland, t­ here was an open debate, the media w ­ ere vocal on both sides of the argument, and the countries ­were fully fledged democracies in which could-be secessionist governments ­were not able to adversely influence the result. Likewise, a low turnout is a strong predictor for a yes vote. It seems that governments are able to troop the core supporters to take part, while the opponents of in­de­pen­dence boycott the referendum. This is what happened in Catalonia. In this region only the supporters of in­de­pen­dence turned out. Interestingly, ­t here is nothing to suggest that in­de­pen­dence referendums are more likely to yield a yes if ­there has been a war. The animosities following an armed conflict do not, statistically speaking, act as recruiting sergeant for support for in­de­pen­dence.


Chapter 3 Simple Error Bar Mean of Yes by EliteConsens 90.00

Mean Yes

80.00 70.00 60.00 50.00







EliteConsens Error Bars: 95% CI

Figure 3.2. Error Bars for Elite Consensus and Yes Vote in Referendums.

The strongest ­factor, statistically speaking, is elite consensus. If the ruling elite is in agreement that they want in­de­pen­dence t­ here is a strong indication that the voters ­w ill vote for secession. This is also shown from the error bars in Figure 3.2; the margin of error, as mea­sured by the width of the bar, is short in cases where the yes vote is high. Conversely, when t­here is no consensus, ­t here is ­little statistical chance that a vote for in­de­pen­dence w ­ ill yield a high yes vote. That is, normatively speaking, as it should be. Democracy means the rule by the p ­ eople. It does not mean factionalism, or government by a small majority—or even the largest minority. Countries vote for in­de­pen­dence if this is a view that is shared by all the po­liti­cal parties and across cleavages, as it was in East Timor in 1999 and in Bougainville in 2019—­but not in New Caledonia in 2019, or in Scotland in 2014. Box 3.1. Velvet Divorce Without a Referendum in Czecho­slo­va­kia ­After the 1992 elections (Václav Klaus on the Czech side and Vladimir Mečiar on the Slovak side), Czecho­slo­va­k ia broke up. The leaders did not have a mandate for this decision, and the election had not been fought on this theme (Harris 2002: 31). Indeed, opinion polls showed that t­ here was support for continuation of the unified state. However, the two leaders ­were intent on dividing the state, and both knew that a referendum on the issue

Secession and Partition 63

Box 3.1. (continued) would be lost. More than 80 ­percent of the Czechs and roughly 70 ­percent of the Slovaks supported federal president Václav Havel’s campaign for a referendum (Wolchik 1995: 233). This law was blocked by Slovakian separatists in the Federal Assembly (Innes 2001: 135). And although the law on break-up of the federation kept being rejected by the Federal Assembly, eventually a new end-­of-­t he-­federation law passed by parliamentary parties. The National Councils ­adopted recommending resolutions, which w ­ ere passed by the Federal Assembly on the second reading on 25 November 1992. ­After this, each of the two state legislatures voted for dissolution of the federation (Harris 2002: 90). How can we reconcile this case? Does it force us to reject the overall conclusion regarding when separatist referendums take place? Not ­really. The separation took place at the end of a period of international upheaval, that is, the breakdown of communism, and the only reason that a referendum did not take place was due to the fact that opinion polls suggested that a majority was skeptical if not outright hostile t­ oward separation (Wolchik 1995: 233). That the referendum did not take place was not due to a flaw in the theory that such a poll was an anomaly or considered unacceptable. Indeed, as noted, Havel, the federal president, called for a referendum and even managed to get signatures supporting a vote from two million voters (Pavkovic and Radan 2007: 73). But the po­liti­cal elites in each of the two would-be states, and especially Mečiar (a socialist) and Klaus (a Thatcherite conservative), knew that their ideological plans could be carried out only if the federation split up. However, it is impor­tant to note that separation was not equally supported by both sides. The Czechs—­including Klaus’s Civic Demo­cratic Party—­­were opposed to separation and accepted it only a­ fter they realized that no other option was available. Conversely, Mečiar’s Movement for a Demo­cratic Slovakia demanded international recognition, while the more hardline Slovak National Party demanded immediate in­de­pen­dence. In other words, the “velvet divorce” was above all a demand supported by the Slovakian elite, but not something demanded by the demos or even the ethnos. So why was no referendum held? Perhaps ­because the two leaders knew that t­ here was very l­ imited support for their policies and ­because neither of them would face competition if no referendum ­were held. Using the logic of the competition proximity model n

Pref = ∑