Springs and their offspring: the international consequences of domestic uprisings

Abstract

A political spring is an abrupt, broad, sustained increase in public dissent in a state that has prohibited it, as in Czechoslovakia in 1968 or Tunisia in early 2011. Some springs produce offspring – clusters of events within neighbouring states (civic unrest, increased state repression, co-option of dissent, revolution) and among those states (intensification of international rivalries, foreign interventions). An English Spring in 1558–9 produced such a cluster in Northwestern Europe. This article addresses the underlying causal mechanism connecting springs and their offspring, rather than the related correlational question (viz. under what conditions a spring is followed by offspring). That mechanism is transnational group polarisation, or the progressive separation of preferences across a population into pro- and anti-government groups. Transnational polarisation along a pro-versus-anti-government axis is an endogenous process triggered by exogenous events, such as violence or public demonstrations that raise the status of, or threat to, one of the groups. It presents powerful actors across states with new threats and opportunities and can help explain how the Tunisian Spring of early 2011 produced throughout the Arab Middle East infectious unrest, serial repressions and reforms, heightened international tensions, and foreign interventions.

Tunisia is unlikely to set off copycat insurrections, for a variety of reasons that could be grouped under the principle that each country’s political situation, despite similarities, has its own unique conditions that facilitate or preclude a successful rebellion.

Tony Karon, Time, 18 January 2011

The Arab Uprising involved more than simply oppressed people trying to free themselves from despotism. Detonated in stepwise fashion by the sudden outbreak and sustenance of dissent in Tunisia – the Tunisian Spring – the uprising comprised a complex set of events across Arab states that included brutal repression, attempts by rulers to co-opt dissent, heightened international rivalries, and foreign interventions.¹ It is barely conceivable that this clustering of events was merely coincidental. In some ways, the self-immolation of a produce vendor in one small Tunisian city in December 2010 is connected not only to similar actions in Egypt and Libya, but to the Muslim Brotherhood’s accession to power in Egypt in 2011, the Saudi-UAE intervention in Bahrain, and the still-unfolding horrors in Syria and Iraq. What spread from country to country, then, was more than just unrest: repression, bribery, and international intervention also appear to have been infectious. We are facing a complex and consequential instance of Galton’s Problem – a contagion of ideas and practices across countries.²

It would be unfair to berate social scientists for failing to predict this tangle of events. Clearly a great deal of contingency was involved.³ Even so, we ought to try to explain their connections. One hindrance to explanation is the division of labour within political science between comparative politics, which traditionally focuses on domestic politics, and international relations, limited to interactions among states. If there are links between a political spring in Tunisia and a Saudi invasion of Bahrain, neither subfield is fully responsible for finding them, and we end up with an intellectual free-rider problem. Related, perhaps, is a robust scepticism among many that rebellions and other practices diffuse across state boundaries. Each country is different, and some scholars argue that fears of ‘falling dominoes’ tell us more about leaders’ paranoia than about how international politics works.⁴

That said, the past few years have produced a growing literature that bridges the comparative-IR gap – in particular, literature on how the international environment affects states’ domestic institutions. Most of this work concerns the alleged diffusion of liberal and democratic institutions across space. The evidence is strong that democracy and liberal ideas and policies do diffuse across national boundaries, although more slowly than some have thought, and via obscure mechanisms.⁵ Scholars have explored more particularly the processes by which democratic revolutions spread across countries. The 1989 revolutions in Eastern Europe – in which dominoes fell toward the West – spurred much of this research, as did subsequent less expansive revolutionary waves.⁶

A number of scholars have advanced explanations for at least some aspects of the extraordinary events in the Arab world in 2011–12. Some of this research has focused on the spread of unrest and rebellion and appealed to demonstration effects, in this case the tendency for protests in one country to inspire protests in other countries.⁷ Kurt Weyland fruitfully compares those events to those in Europe in 1848, finding that in both cases dissenters were captive to cognitive bias and over-estimated the prospects of success; hence, in both cases, despotic governments won in the end.⁸ Clearly demonstration effects and contagion were present in 2011, and clearly protestors did miscalculate. But rounding out the picture requires that we explain why the governments chose to try to defeat the demonstrators; after all, in 1989 the optimistic dissenters in Eastern Europe turned out to be correct, as their governments took the opposite decision. As Stephen M. Saideman writes, in 2011 some Arab rulers evidently were learning from one another’s mistakes in real time and adjusting appropriately.⁹

Furthermore, few scholars have explored the connections between the unrest and repression and the intensification of international rivalries and violence in the Arab world since 2011. Just as wars tend to cluster in space and time, suggesting international contagion,¹⁰ foreign interventions in Arab countries in 2011–12 do not appear to have been coincidental. If dissent, revolt, and repression do spread across national boundaries, then we should expect alert and savvy neighbouring governments to anticipate changes in the distribution of power in the region. A revolution can alter a state’s foreign policies and alignments, and a cascade of revolutions can alter radically patterns of foreign influence in a region. Thus we should expect new patterns of international cooperation and conflict and serial foreign interventions.

The Tunisian Spring had important offspring. But what mechanisms connected cause and effects? A spring can cause the congeries of events we saw in the Arab world in 2011–12 by virtue of its broad and deep polarising qualities. A sustained outburst of public dissent in state A can divide A’s population into pro- and anti-reform groups that harden into pro- and anti-regime groups. It can do the same in states whose populations have extensive social ties to A’s population. In other words, a spring in A can trigger transnational group polarisation, by which I mean the progressive redistribution of preferences among the populations of more than one state into two or more oppositional groups that had been dormant. This transnational polarisation is seen in the nearly simultaneous spread across countries of unrest, on the one hand, and of state repression and bribery, on the other. Transnational polarisation interacts with various factors in countries, leading to varied results, ranging from reform to unsuccessful rebellion to successful revolution. Polarisation also interacts with international structural factors and can alter the distribution of power and influence in a region. Polarised countries can be unstable, which leads to uncertainty about their future regimes and policies; thus they can attract foreign intervention. Each of these types of event, in turn, can exacerbate group polarisation and reproduce the entire cycle.

In what follows, I narrate two cases: an English Spring in 1558–9, and the Tunisian Spring of 2010–11. Both are, in methodological terms, ‘typical’, that is, each exemplifies the phenomena in question, inasmuch it produced the offspring described above. As Jason Seawright and John Gerring argue, typical cases are useful in locating and testing causal mechanisms.¹¹ I argue that the mechanism connecting the clusters of events in both cases is transnational group polarisation. Group polarisation is an endogenous process triggered by exogenous events that raise the status of or threat to a pre-existing but latent social group. Transnational group polarisation helps explain how the Tunisian Spring of early 2011 produced the complex cluster of events that followed over the Arab world. I conclude with questions for further research, including why some springs are barren and others spread and succeed.

What is a spring?

‘Spring’, as applied to politics, suffers from a lack of definition. The first political use of spring probably is by the German liberal writer Ludwig Börne, who in 1818 prophesied a Völkerfrühling, or springtime of nations – a coming liberation throughout Europe.¹² Various historical events – in Europe in 1830 and in 1848, in Russia in 1904, in Czechoslovakia in 1968, and in many Arab countries in 2011–12 – have carried the label ‘spring’ and are quite different in some ways. Each of these began in a single country or city and each entailed reformers increasing their demands and meeting with government repression. All but Russia in 1904 were followed by international tensions and foreign intervention.

Similarities notwithstanding, clearly we must impose some discipline on the concept of a political spring. In this article I define a spring as a sudden increase in one authoritarian country in the volume and breadth of open dissent, an increase sustained for at least two weeks. By so defining ‘spring’, I hope to make clear that phenomena that sometimes occur along with or shortly after a spring – regional contagion, state bribery and repression, rebellion, revolution, international tensions, and foreign intervention – ought not to be conflated with the spring itself. In popular usage ‘Arab Spring’ covers a number of these types of event, as if they are logically related. But not all flowerings of dissent in one country have these foreign and international effects. If a spring is related to these other events, that relationship must be contingent, not logical, and our terminology should free us to investigate whether a lifting of repression in one state can cause the phenomena in other states and in an entire region.

I do not limit springs to events that empower liberal democrats alone. Dissenters living under authoritarian regimes adhere to a variety of ideologies. A spring may be a liberalising move, but as the Arab Uprising shows, liberalisation can empower all manner of discontented people, liberal and non-liberal alike.

A brief illustration: an English Spring and Northwestern Europe, 1558–66

The Protestant Reformation that began in 1517 threw much of Europe into more than a century of intermittent political as well as religious turmoil. Under the Peace of Augsburg (1555), Catholics and Lutherans in the Holy Roman Empire (Central Europe) agreed to tolerate one another, but strife was increasing to the west and north with the rapid spread of newer Protestant movement, Calvinism. Excluded from the Peace of Augsburg, Calvinists were more militantly anti-Catholic than Lutherans. By 1559, Calvinists were found in all social classes in France, England, Scotland, and many parts of the Holy Roman Empire. With Calvin’s own Geneva as their nerve centre, Calvinists communicated with each other across political boundaries.¹³ Under threat of persecution, they usually spread the faith quietly.¹⁴

In France, England, and Scotland, Calvinists alternated between suffering bouts of severe suppression and enjoying limited toleration.¹⁵ In late 1558, however, the lifting of persecution in England was more robust. On 17 November, Mary I, arch-Catholic queen, died and her half-sister Elizabeth succeeded her. Raised as a Protestant, Elizabeth I carefully but decisively moved to end the severe persecution of her co-religionists.¹⁶ The Acts of Supremacy and Uniformity, passed in April 1559, made her (rather than the pope) head of the Church in England and mandated that only the Protestant prayer book of 1552 be used in worship.¹⁷ Elizabeth’s changes fell well short of those sought by the most zealous Protestants. Most of her Catholic nobility were sufficiently content with the settlement not to rebel.¹⁸ This, then, was the English Spring of 1558–9: an end to persecution of Calvinists and other Protestants coupled with a relatively moderate policy toward the Catholic faithful.¹⁹

Scotland, then a separate kingdom, was a long-time junior partner of France ruled by Mary of Guise, a French Queen Regent (r. 1554–60). Before 1559 Mary had been relatively tolerant of Calvinist preaching, and Calvinism gained important adherents among the gentry and nobility. An increasing French military presence since 1548 had begun to chafe, and Calvinism provided a language for the discontent.²⁰ In December 1557, the Calvinist Lords of the Congregation formed and pledged to work for the establishment of a Reformed Church.²¹ In March 1559, Mary of Guise held a council to consider church reforms. The evidence suggests that Mary was prepared to grant major concessions to the Congregation. But the English Spring emboldened the Congregation and in that same month some of its leaders began to talk to William Cecil, one of Elizabeth’s ministers in England, about the possibility of English help in their struggle.²² The Congregation ultimately rebuffed Mary’s proposed reforms.

Mary responded by hardening her own position. At month’s end she declared that all Scots who refused to conform to Catholic rites would be tried for heresy. Calvinist clergy ignored the order, and in May Mary summoned them for hearings. The Queen Regent had several motives, but an important one, writes Alec Ryrie, was this: ‘her Protestant subjects, who had once been harmless, were now potentially an English fifth column’.²³ And indeed, in May the Congregation rebelled. By the close of June it had conquered the cities of Perth, Dundee, Stirling, and Edinburgh.

The Congregation being pro-English, Scotland’s foreign alignments hung in the balance. Seizing the opportunity, Elizabeth sent money to the Calvinist rebels and promised more.²⁴ In October, the Congregation declared Mary of Guise overthrown and themselves rulers of Scotland.²⁵ In December, with Scotland slipping from France’s grasp and into that of England, Catherine de Médici, France’s new Queen Regent, sent an expedition to crush the revolution. At least 1,000 Frenchmen were drowned in storms en route, and in January 1560 Elizabeth of England responded to the French expedition with a naval blockade of Scotland’s east coast.²⁶ On 11 June, Mary of Guise died; the next month a treaty was signed and 4,000 French were evacuated.²⁷ The Congregation effectively ruled Scotland, and promptly realigned the country from a French ally to an English one.²⁸

In short: The English Spring of 1558–9 was followed by polarisation in Scotland, as dissenters there gained confidence, pro-government actors increased repression, and civil war erupted. That polarisation, in turn, generated worries in France (and hopes in England) that Scotland would defect from France and become an English ally; intervention by France and England in the Scottish civil war; and with the Protestant victory in Scotland, a diplomatic revolution and shift in the international balance of power. Similar chains of events took place over the next several years in France and the Netherlands. In both countries Protestants, encouraged by the successes of their brethren in England and Scotland, began making more demands on their rulers; were met with new repression; rebelled; and attracted intervention by foreign Protestant and Catholic powers hoping to influence the outcomes.²⁹

How a spring produces offspring

Social science has produced a great deal of literature on the specific categories of event we see in the sixteenth-century case: rebellions, revolutions, suppressions, spiralling hostility, and foreign interventions. But what interests us is not any one type of event, but the cluster itself – the close proximity in space and time of the events. Decomposing the cluster into its various events, subsuming each type of event into a separate larger dataset (for example, the universe of foreign interventions), and running regressions on various theorised explanatory variables would be helpful in some ways. But it would obscure the coincidence in time and space – coincidence that we have good reason to suspect is not simply coincidental but causal.³⁰

Better, then, to conceive of these clusters of events as causally complex, with numerous interactions among variables, feedback loops, and path dependencies.³¹ Causal complexity may be studied profitably with quantitative methods, namely agent-based modelling (ABM). ABM clarifies causal claims by simplifying highly complex system dynamics, and scholars have built models that test for the effects of certain variables on polarisation and the transnational spread of conflict.³² This article has a different task, however. It seeks to uncover the causal mechanism connecting Tunisian Spring to the rebellions, suppressions, reforms, and foreign interventions that followed over most of the Arab world. Any adequate causal explanation will entail both conditional statements (if X then Y) and specification of a causal mechanism (X produces Y via M).³³ Although the former task is vitally important, this article is confined to the latter. ‘Mechanism’ is variously defined in social science literature; I have in mind, in the words of James Mahoney, ‘something that intervenes between cause and outcome’.³⁴ We want to know, then, the processes through which political loosening in state A produces both protests and repression in B, greater competition between C and D, intervention by D in B, and so on.

Group polarisation as mechanism

If the complex turmoil in the Arab world in 2011 is causally linked, the mechanism must have to do with transnational diffusion and contagion, or the spread of ideas (threats, opportunities, and so on) and practices across state boundaries. Social scientists have defined and explained transnational diffusion in a number of ways, including ‘emulation, persuasion, flows of resources (ideational or material), framing, and power transitions…’³⁵ I return to transnational diffusion in a later section. First, however, it is important to get clear that the thing diffusing is not simply ideas or practices: it is group polarisation.

Polarisation, a concept borrowed from physics, has been treated extensively in social science. Much of that work concerns social polarisation, or segregation into groups that are stable over long periods of time (such as ‘Red America’ and ‘Blue America’).³⁶ By group polarisation, however, I mean more short-term and situational segregation. Group polarisation is the progressive clustering of preferences in a population into two or more oppositional groups. ³⁷ In the case of springs and their transnational progeny, agents polarise into anti- and pro-government groups. Group polarisation is a self-reinforcing identity change that entails changes in utility functions.³⁸ To illustrate: when a population is polarising, at time t an actor in group A an actor in group B prefers a 50-50 allocation of goods; at t+1, each may prefer a 60-40 allocation in its group’s favour; at t+2, a 75-25 allocation; and so on. At the limit of polarisation, each side wishes the other destroyed.³⁹ Group polarisation, then, is one way to explain a progressive worsening of conflict. Inasmuch as it is a change in preferences (mental states) it is not directly observable. Yet its behavioural markers are clear: speech and action that are increasingly biased in favour of one’s group and against an opposing group are evidence of polarisation.

Group polarisation has both exogenous and endogenous features. It is triggered by exogenous events and, as explained below, takes place along axes of identity that are exogenous. But it is endogenous in the sense that it is self-reproducing.⁴⁰ It entails the altering of individuals’ preferences and practices and creates new threats and opportunities for various actors, including actual and aspiring rulers. Polarisation’s endogeneity means that a spring’s offspring may produce still more offspring over time and space. (Of course, a population may depolarise as well, and indeed in a population of overlapping groups polarisation along one axis will entail depolarisation along other axes.)

The process

Individuals and groups

Stated informally, the basic group polarisation model is simple. Assume a population of 100 persons, all belonging to two social groups. Fifty are Muslim, fifty Christian; fifty are socialist, fifty fascist. Half of the Christians (twenty-five) are socialist, half fascist, and the same is true of the Muslims. The population thus has cross-cutting cleavages – one religious, the other ideological – that are perfectly symmetrical.⁴¹ The population is in an unbiased equilibrium, such that individuals’ commitments to each group varies but, in the aggregate, neither religious nor ideological identities predominate. Social interaction does not skew the distribution of resources, including information, to any of the four social groups. Now suppose that, for exogenous reasons, two fascists (one Christian and one Muslim) attack and mortally wound a socialist. There follows the polarisation of the population along the ideological axis, such that individuals care less and less about religious identity (Islam vs Christianity) and more and more about ideology (socialism vs fascism). If not disrupted, the polarisation culminates in inter-group violence.

Different theories may ground that simple model. Rationalist social scientists begin with the assumption that actors have fixed preferences, and social interaction is strategic, that is, actors try to secure what they want. Thus for Kuran open dissent and unrest spread not because people identify more strongly with a group but because they are no longer afraid of persecution for the beliefs they already held. Citizens of an oppressive regime falsify their preferences, so as to avoid persecution. When each is convinced that revealing his or her anti-regime preference is safe, each will reveal that preference. Thus the tendency for anti-regime demonstrations in Eastern Europe in 1989 to cascade.⁴²

Rationalists depict the spread of polarisation and conflict in like fashion, as the revelation rather than the alteration of preferences. David A. Lake and Donald S. Rothchild argue that ethnic conflict spreads within and across states as information flows are distorted.⁴³ More pessimistic rationalists attribute polarisation to accurate flows of information, which reveal genuinely opposed preferences. As members of a population talk, act, observe, and react, each gains information and realises that he has much in common with some and much in opposition to others. No one is changing identities or preferences.⁴⁴

The grounding of polarisation in this article, however, is constructivist, in the tradition of Georg Simmel; Lewis Coser;⁴⁵ Charles Tilly, Sidney Tarrow, and Doug McAdam; Henri Tajfel and John Turner. All of these see individuals not as fundamentally atomised units who choose affiliations to serve interests that they already have formed, but rather as members of groups – many unchosen – that give them their identities and hence structure their interests. As Simmel notes, in modern societies people belong to multiple social groups.⁴⁶ Someone may at once be French, female, married, a resident of Lyon, a European, a state employee, and so on. Each of these properties places her in a corresponding social group (French persons, females, married persons, Lyonnaises, state employees). Depending on conditions, some group affiliations are more politically salient than others. Simmel tells of Australian aborigines whose tribes are cut across by ‘five gentes or totemic associations … If in a fight between two hordes the members of the same totem meet, they avoid each other and seek out another opponent.’⁴⁷ Thad Dunning and Lauren Harrison argue for a similar phenomenon, ‘cousinage’, in explaining weak ethnic identities.⁴⁸

Three properties of social groups bear mention. First, boundaries are necessary to group definition and cohesion. A group or set is only logically possible if there is a ‘not-group’. Female only has meaning insofar as there also is male.⁴⁹ Second, one’s beliefs and subjective interests derive in large part from his group affiliations. Identifying as a worker entails identifying one’s interests with fellow workers and against those of owners, and hence favouring laws requiring health benefits to employees. Insofar as one identifies as a Frenchman, he identifies his interests with those of fellow French and against those of foreigners.⁵⁰ Thus, in proportion to the strength of his national affiliation, he will tend to desire a distribution of goods biased toward French and against foreigners because he will believe such a distribution to be in his interests. Group identity matters to politics because it entails a particular configuration of interests or preferences.

Third, and most important, group affiliations vary in coherence and salience across time and space. Today, for most French citizens most of the time, being French is less salient than it was in 1916; being European probably is more salient today than a century ago. But even in the short term, the salience of a group affiliation may vary with circumstance. Thus, our hypothetical single female Lyonnaise state employee may, in some circumstances, identify more with the state (and be biased toward fellow state employees and against others), and in other circumstances identify more as a French citizen (with corresponding biases).

Exogenous triggers of group polarisation along a particular axis

What, then, explains variation in the relative salience of any given person’s group affiliations? Put another way, what causes one identity and set of interests to be activated and alternatives deactivated?⁵¹ Scaling up to the population level, what causes a population to polarise along one identity axis rather than another? Psychological literature posits at least two attributes of groups that raise their salience: high status, and threat.

Members of high-status groups are significantly more biased toward fellow members and against non-members than are members of low-status groups.⁵² An exogenous rise (fall) in a group’s status can render it more (less) salient for its members. A rise in status may be triggered in politics by victory in an election or a civil war, or an unexpectedly large public rally. The opposite types of event may trigger a fall in status. Highly committed members of groups whose physical, economic, social, or other standing is threatened also tend to be more biased toward fellow group members. A new threat – such as an attack on a group member qua member – tends to arouse in such persons fears that they may be next and an intuition that other group members have the same fears, and so they tend to see a stronger common interest with fellow group members. Their identification with the group grows, as do their efforts to homogenise it and their emphasis on differences with the out-group.⁵³ Social theorists have made much of this insight that solidarity with some requires conflict with others.⁵⁴ Among persons whose prior commitment to the group was low, by contrast, a threat to the group tends to prod them to exit that group or to try to hide.⁵⁵

Events that raise a group’s status or lower its safety, then, prime people, pressing them to alter their notions of self-interest and to identify more as members of that group and less as members of their other groups. In countless experiments psychologists have found that priming subjects with cues related to a group identity leads them to identify with that group and act accordingly.⁵⁶ Put abstractly, an exogenous event that either raises the prestige of social group A or threatens A may cause people who belong to group A to identify more strongly with A and against not-A and to provoke members of B to identify more against not-B. Large public anti-government demonstrations, common during a political spring, can simultaneously raise the status of being anti-government and threaten those who identify with the government, such as rulers and security forces. As more people act according to this increasingly salient identity axis, still others are effectively primed to follow suit. Group priming begets more group priming, and thus polarisation tends to feed on itself: as members of A observe members of not-A identifying more as not-As, members of A will identify still more strongly with A; and so on.⁵⁷

We must introduce two related complications. The first is based on the empirical finding that real people vary in their prior commitments to a group.⁵⁸ In equilibrium some French factory workers identify more as French and less as factory workers. Following a triggering event, highly committed group members mediate and broker polarisation by spreading and exaggerating information about the threat or increased status and by instigating more events to mobilise or demobilise people.⁵⁹ Public discourse turns in large part to what is to be done; those who hold extreme views tend to have more influence in such times, and moderates either are quiet or move toward the extreme.⁶⁰ Second, the capacity to manage polarisation varies across a population. Some highly committed group members will possess more resources and skill than others. Especially relevant to the case at hand, a government’s superior resources give it particular advantages at manipulation. A government may use propaganda, bribery, and coercion to try to exploit polarisation. Contrary to the recent hopes of many, governments can harness new social media and use them to reinforce their hold on power.⁶¹ A government will be especially determined to keep its military and domestic security forces from defecting to the anti-government pole.⁶²

Prior to a spring, then, some citizens will be fundamentally pro-government and others anti-government, but many in each group will not consider their attitudes toward the government especially important; in each group citizens will vary in level of commitment to their position. As a spring begins, committed anti-government citizens criticise the government more openly than before over a sustained period. A spring thus raises the status of being anti-government and activates the latent anti-government identity of more people. The mounting speech and acts by anti-government people can threaten people who identify as pro-government. Hence a citizenry may polarise into anti- and pro-government groups. The more determined anti-government actors will work to expand the anti-government group. At the same time, any government will try to minimise the size and solidity of the anti-government group via some mixture of coercion, bribery, and propaganda. It may find complete elimination of dissent too costly, and indeed, insofar as the existence of an anti-government group can expand and solidify the pro-regime group, it may wish to keep a small anti-government group. In any case, a spring that increases the size and solidity of an anti-government group will be followed by some mix of government repression (shootings, trials, imprisonment, and threats thereof), favours (cash payments, jobs, amnesty), and propaganda (on the vices of dissidents and the virtues of pro-government citizens).

Transnational polarisation

Empirical studies by David Carment and Patrick James and by Erika Forsberg suggest that ethnic polarisation can travel across state borders as people are inspired by confreres in other countries and ethnic identities come to the fore.⁶³ The same goes for other bases of identity, including anti- and pro-government bases. Preference clustering in one state can produce speech and actions that produce similar preference clustering in other states. An abrupt rise in the status of anti-government actors in state A, for example, from a surprisingly large demonstration, can cause latently anti-government actors in state B to identify more strongly as anti-government and to demonstrate as well. The same event in A can alarm pro-government actors in B and lead them to identify more strongly as pro-government and to take appropriate action, for example, suppress anti-government actors – or, more strategically, bribe them pre-emptively. Polarisation in B can feed back into A and exacerbate polarisation there.

For group polarisation to cross state borders, two conditions must obtain. First, parallel latent social groups must already exist on all sides of those borders, for example, anti-government actors must exist in at least two countries at once. Some members may be highly committed activists who form transnational networks, such as the international proletariat or the Calvinist International. Group members in each country may or may not already be communicating. Second, information must travel across state borders – meaning that rulers must be unable or unwilling fully to control the flow of information into and out of their countries. Governments of states near a political spring may have interests in preventing their citizens from learning of it. But most governments need to tolerate some outside news. (The only exception in today’s world is North Korea.) Under these two conditions, a sudden increase in one group’s prestige or in threats to that group in state A can trigger polarisation between members of parallel groups in neighbouring states B, C, and D, leading to the emergence of transnational groups.

In the offspring countries B, C, and D, as in the spring country A, various actors have incentives to manipulate and modify polarisation via propaganda, bribery, and coercion. The chief interest of B’s government being the retention of power, it will need to minimise the potency of the anti-government group and maximise the potency of the pro-government group. At its disposal will be the same tools that A’s government has (surveillance, propaganda, repression, co-option). Committed anti-government actors in B will try to counter B’s government’s efforts. Insofar as the anti-government actors succeed, B will undergo civil unrest and polarisation may worsen.⁶⁴

International Relations

Transnational group polarisation can affect relations among states in a region by increasing uncertainty about the future of the distribution of influence and power across that region. That uncertainty, in turn, can provoke foreign realignments and interventions. The more polarised a state – the more intense are anti-government and pro-government sentiments – the higher the probability of government or regime change. Government or regime changes may cause changes in foreign alignments.⁶⁵ Precisely because anti-government actors in a highly polarised country have intense preferences, if they prevail they are likely to follow radically new policies, including a quest for new foreign allies; that, in turn, would degrade the power of the ex-allies and increase that of the new allies. Possible or actual changes in the balance of international power in a region can cause tensions among the major powers and among any extra-regional powers that have a stake in the region. It also can give rulers of other states an incentive to intervene in highly polarised states so as to steer the outcome. Foreign interventions can provoke counter-interventions by rival states.

Rulers of other states have a second incentive to intervene in polarised states: uncertainty about their own domestic security. Precisely because polarisation is transnational, rulers of polarised state who find their own hold on power loosening have reason not only to manipulate polarisation in their own state but also to do so in other states. If the government of B can defeat the anti-government group in A, it also might weaken the anti-government group in B by lowering the prestige of anti-government groups and raising the status of pro-government groups across the region.⁶⁶ Here, too, interventions will tend further to exacerbate international tensions and may provoke counter-interventions. Indeed, foreign interventions may intensify polarisation within countries.⁶⁷Figure 1

Figure 1 The causal mechanism.

presents the argument in graphic form.

The Tunisian Spring and its offspring: Hypotheses

The argument that group polarisation diffused across states, and that this transnational polarisation is the causal mechanism by which the Tunisian Spring produced such wide and complex offspring, implies a number of observations. The usual test of a causal mechanism in a particular historical case is process tracing, and that is the method I follow in this article. Process tracing entails examining a case closely for signs that the hypothesised mechanism is present. Of particular value are ‘hoop tests’, or tests for the presence of events that must have occurred if the mechanism were present; and ‘smoking gun tests’, or tests for events that, if present, definitely confirm the mechanism.⁶⁸ Inasmuch as group polarisation is a change in distribution of preferences (mental states) in a population, it is not directly observable, and so no smoking gun test is available. Hence I shall rely on a set of hoop tests: the close temporal proximity among similar events across Arab states; similar forms and tropes of protest and reaction; anecdotal accounts of contagion; and evidence from social media.

The scope conditions of the argument apply to the Arab world in 2011. First, notwithstanding the predominance of authoritarian regimes, communication across Arab-majority states was robust. The Arabic language comes in many local dialects, but a generic form, Standard Arabic, is used across the region.⁶⁹ Indeed, some scholars argue for a new Arab public sphere, a kind of bottom-up pan-Arabism driven by the plummeting cost and high velocity of long-distance communication.⁷⁰ The Al-Jazeera satellite channel, out of Doha, Qatar, is most famous. Other such channels include Saudi-based al-Arabiy and the US-based al-Hurra. Al-Manar, run by Lebanon’s Hezbollah, has been on the air since 1991 and has sought to compete with these other channels.⁷¹ Internet-based social media also contributed to the formation of a dissenting Arab consciousness. A survey of 16 Arab countries in late 2009 found between 40 and 45 million Internet users. The Arab world had an estimated 35,000 blogs. Social media sites were new vehicles by which Arabs voiced dissent against authorities.⁷²

Second, dissenters were active in most countries prior to 2011 and indeed already formed transnational networks. Millions of devout Sunnis belonged to the most famous transnational Islamist group, the Ikhwan or Muslim Brotherhood. Founded in Egypt in 1928, and still centred there, the Brotherhood (according to Wikipedia; caveat lector) had chapters in Bahrain, Jordan, Syria, Iran, Iraq, Israel & Palestine (Hamas, which rules Gaza, began as the Palestinian chapter of the Ikhwan), Saudi Arabia, Kuwait, Yemen, Oman, Algeria, Sudan, Somalia, Tunisia, and Libya.⁷³ The Egyptian Ikhwan website links to material about various other national chapters.⁷⁴ It is by no means a centralised group; disagreements among and within national affiliates are well-known. The Egyptian website, however, makes clear that Brotherhood national chapters share ideas and ‘have a special and amicable bond’.⁷⁵

Smaller but also important were liberal networks that rejected both authoritarianism and Islamism and sought democracy, human rights, transparency, free markets, and globalisation. Arab liberals tend to be highly educated, many with degrees from Western countries. Arab liberal networks were active after the fall of communism in Eastern Europe in 1989, particularly the fall of Nicolae Ceauşescu; the US-led defeat of Saddam Hussein’s Iraq in 1991 had a similar effect.⁷⁶

Tunisia: Increases in threat and status were followed by signs of polarisation

On 17 December 2010, Mohamed Bouazizi, a street vendor in the Tunisian provincial city of Sidi Bouzid, set fire to himself to protest a police seizure of the produce he was selling. Late the following day, hundreds of young people gathered to protest the death. Police dispersed them with tear gas. A Facebook page showed footage of the protests.⁷⁷ Unrest spread to other towns, reaching the capital Tunis on 19 December where 1,000 gathered outside union headquarters. Two protesters died. Some began to call for President Zine al Abadine Ben Ali not to run for re-election in 2014. The following day, Ben Ali visited the families of the dead, and the hospitalised Bouazizi; via televised speech he threatened to crack down on protests.⁷⁸ Yet protests continued, and hence events in Tunisia qualify as a political spring.

Michael Biggs argues that acts such as self-immolation are examples of ‘communicative suffering’, designed to mobilise collective action by inducing feelings of guilt or shame upon fellow group members or by showing the uncommitted the depth of the group’s convictions.⁷⁹ Bouazizi’s self-immolation seems to have effectively raised the status of Tunisians with deep complaints against their government, and so street protests multiplied across the country. In reaction, rulers and security forces responded with force.

Polarisation in Tunisia was soon followed by polarisation in other countries, marked by similar acts and tropes

Over the next few months, what spread to other Arab countries was not only civil unrest but also government efforts to suppress it via varying combinations of repression and bribery. A number of Arab countries were shaken by self-immolations and protests in imitation of those in Tunisia. Governments in these states reacted to such unrest with repression, and even anticipated unrest with attempts to buy off their citizens with money or promises of reform. Security forces could have defected to the protesters, but generally obeyed orders and acted so as to maintain regimes in power. In other words, polarisation, not just simple rebellion, spread: civil unrest was infectious across national boundaries, but so were efforts to suppress that unrest.

In early January 2011, as unrest continued in Tunisia, young protesters in neighbouring Algeria fought with police in several cities, looting shops in the capital Algiers. On 7 January, protests erupted around mosques and football matches.⁸⁰ On 16 January, Mohsen Boutefif set himself on fire, evidently imitating Bouazizi in Tunisia.⁸¹ Various Algerian cities had had regular protests since 1992; but this time these protests grew and spread even to the capital city of Algiers. The government responded by using its oil revenues to keep food prices from rising too steeply.⁸²

In Cairo, Egypt’s capital, a man set himself on fire on 17 January near the parliament building.⁸³ Anti-government groups declared 25 January a ‘day of revolt’ in Cairo and other Egyptian cities. The police responded with water cannon and rubber bullets. Suppression only aggravated unrest. Interaction between the government and protesters (and outside diplomatic intervention, particularly by the United States) escalated into a revolution the following month, in which Mubarak resigned and a new regime began to take shape. Polarisation according to attitude toward regime meant depolarisation along other axes, including religious ones. For example, during the Egyptian revolution Muslims – including Islamists – and Coptic Christians cooperated.⁸⁴

In Sanaa, Yemen’s capital, thousands protested in mid-January. On the 23 January, more protests led to the arrest of Tawakul Karman, a prominent female dissident and Islamist. Riot police used water cannon on protesters at Sanaa University.⁸⁵ On the same day President Saleh attempted to pre-empt unrest by declaring a 50 per cent cut in the income tax and price controls on basic commodities. Saleh also increased the salaries of the civil service and armed forces.⁸⁶ Four days later, tens of thousands protested on the streets of the capital, calling for Saleh’s resignation.

On 28 January, thousands of protesters took to the streets of Amman, Jordan. The group included labour union members and Islamists.⁸⁷ Three days later, King Abdullah appointed a new prime minister, Marouf Bakhit, and announced new political reforms.⁸⁸

On 31 January, Syria’s President Bashar al-Assad told the Wall Street Journal that unrest would not spread to his country, proclaiming that Syria’s unifying ideology would maintain stability. He also announced new political reforms, including more freedom for the press and for non-governmental organisations.⁸⁹ Six weeks later, police cordoned off the southern city of Daraa following demonstrations and the killing of five protesters by police.⁹⁰ So began the ongoing Syrian rebellion, the bloodiest and longest-running offspring of the Tunisian Spring.

On 16 February, in Libya’s eastern city of Benghazi, protesters numbered in the hundreds fought with police. The protesters were calling for the release of a human rights activist. The government of Muammar Qaddafi staged counter-protests, broadcast on state television. So began a spiral of violence in which the regime cracked down brutally on protesters and civil war ensued; the eventual Libyan revolution toppled Qaddafi, then the longest-ruling autocrat in the Arab world.

In late February, members of Saudi Arabia’s Shia minority protested, asking for an elected parliament; 22 were arrested.⁹¹ On 6 March, the government banned public protests.⁹² Five days later, heavy police deployments thwarted a planned ‘day of rage’.⁹³ On 18 March, King Abdullah announced on live television a package of benefits to citizens worth billions of dollars. The package included pay increases, loans, apartments, and – in a dual-purpose initiative – the hiring of 60,000 new security personnel, addressing problems in employment and in domestic security.⁹⁴

In other oil-rich Gulf monarchies the stories were similar. In February the Kuwaiti government increased pay to civil servants by 115 per cent and handed 1,000 dinar (approximately $3,500) to each citizen. The Gulf Cooperation Council collectively pledged $20 billion for public works projects in Bahrain and Omar.⁹⁵

On 14 February, Shia Muslims in the tiny Gulf state of Bahrain staged a ‘day of rage’ against discrimination by the ruling Sunni al-Khalifa family. They were met with rubber bullets and teargas fired by the police; 14 people were injured.⁹⁶ Demonstrators converged on Pearl Square, where they set up camp for many days to demand a new constitution and an end to anti-Shia discrimination. Bahrain’s largest Shia party, al-Wifaq, withdrew its 18 representatives from parliament, leaving only 22 in that body. Police fired upon a funeral procession of 10,000. King Hamad visited the families of two killed in demonstrations.⁹⁷ Crown Prince Sheikh Salman bin Hamad al-Khalifa pledged a nationwide dialogue, parliamentary and electoral reform, and a national referendum for constitutional change.⁹⁸

Sceptics of international contagion argue that similar conditions within countries explain similar outcomes.⁹⁹ The evident transnational monitoring and imitation – self-immolations and ‘days of rage’, for example – cast serious doubt on that scepticism. It is important to note, too, what was passing from country to country was not only unrest, but also efforts to contain or suppress that unrest by governments and their agents. Some governments tried to pre-empt unrest with reform. These patterns of interaction are consistent with the diffusion of social polarisation along a pro- versus anti-government axis.

Anecdotes about infectious polarisation

Anti-government actors during and after the Arab Uprising noted to journalists and scholars that they were inspired by counterparts in other Arab countries. ‘The Tunisian tsunami gave everyone hope’, said one Egyptian activist.¹⁰⁰ Another Egyptian who used social media is quoted as saying:

After the Tunisian revolution … Egyptian people were saying, ‘Egyptians, why don’t you revolt? You can see the Tunisians, what they did.’ I didn’t like these comments. Then I found an event calling people to revolt on the 25^th of January, which for me was … a provocative joke, because a revolution should not be scheduled … I just expected protests, maybe a mass protest, but I never expected a revolution.¹⁰¹

A female Saudi protestor, Alia al-Faqih, was quoted as saying, ‘The protesters in Egypt and Tunisia did something that was almost impossible. If they could bring down two tough presidents, why can’t we demand our rights?’¹⁰² Demonstrators in Yemen in late January chanted, ‘Look at Tunisia with pride … Yemen has strong people too.’¹⁰³ Demonstrators in Qatif (Eastern Saudi Arabia) told interviewers they were inspired by fellow Shia in Bahrain to act.¹⁰⁴

At the same time, resistance to rebellion spread as government actors anticipated transnational revolutionary contagion, observed their counterparts in other countries suppressing and co-opting it, and imitated them. As mentioned earlier, Assad of Syria explicitly stated that revolt would not spread to Syria and that he was implementing reforms to that end. On 9 March, King Mohamed VI declared to his people that substantial constitutional reforms were on the way.¹⁰⁵ Queen Rania of Jordan told a prominent US journalist that the Arab Uprising helped Jordanian elites see their way to supporting reforms that King Abdullah had long intended; Abdullah himself said that the transnational wave of unrest prodded him to initiate dialogue with Jordan’s Muslim Brotherhood.¹⁰⁶ Adel al-Toraifi, a Saudi columnist for London-based Asharq al-Awsat (and now the Saudi Minister of Culture), wrote on 2 February that ‘governments [should] take the initiative to absorb these lessons’ lest their own illegal opposition parties ‘monopolise the current opportunities, and ride the wave of change’.¹⁰⁷

Contagion on social media

The social medium of Twitter provides evidence of transnational contagion (if not of polarisation). Claims that the revolutions were ‘caused’ by social media are overblown. Many researchers, however, have exploited the availability of social media data to track the spread of sentiment concerning unrest. Published studies make clear that Arabs in one country were activated by events in other Arab countries. Philip Howard and his colleagues find that ‘social media helped spread democratic ideas across international borders’. Figure 2

Figure 2 Number of tweets about a given Arab country in other Arab countries, early 2011.

(reprinted from Howard et al., ‘Opening Closed Regimes’, 2011) shows changes over time in the volume of tweets about politics in one Arab country that were posted in other Arab countries (that is, not the country that is the subject of the tweets).¹⁰⁸ Note that #sidibouzid refers to the city where the Tunisian Spring began, and that #feb14 refers to the Day of Rage in Bahrain.

Here again, the thesis that the Arab Uprising was caused entirely by conditions within each country, with no transnational contagion, is difficult to maintain in the face of the evidence.

Note, too, that the resignations of Ben Ali in Tunisia and Mubarak in Egypt follow sharp rises in tweets about those respective countries. Howard et al. report that ‘a spike in online revolutionary conversations often preceded major events on the ground’. They acknowledge that cause-and-effect is difficult to tease out, but the close temporal connection between tweets and dictator resignations is consistent with endogenous polarisation.¹⁰⁹

Anecdotes about contagion abound. On 1 February, Time reported Egyptian social media saying, ‘Tell your friends. Look what is happening in Tunisia. This is how people change their country.’ A Facebook page from Eastern Saudi Arabia reading ‘Qatif and Bahrain are one people not two peoples’ garnered 19,000 ‘likes’. (Qatif is the eastern Saudi province with a majority Shia population.) In Syria a Facebook page, ‘We Are All Hamza Alkhateeb’, was remarkably similar to the ‘We Are All Khaled Said’ Facebook page in Egypt.¹¹⁰ The New York Times reported on 14 February that Egyptian protestors conferred with their Tunisian counterparts on tactics.¹¹¹

International tensions and interventions

That regional powers believed that the Arab Uprising had the potential to alter the balance of power is consistent with the claim that Arab countries were polarised. Just as the revolution in Scotland in 1560 removed that country from France’s column and placed it in England’s, the Arab rebellions of 2011 threatened to change who ruled which countries and hence foreign alignments patterns and the distribution of power in the region. Particularly in play was the rivalry between Iran and Saudi Arabia, states that glower at one another across the Persian Gulf; one Arab, the other Persian; one Sunni, the other Shia; both claiming to be the exemplary pious regime and competing to be the leading power in the Middle East. Transnational polarisation interacted with that regional rivalry.

Iranian and Saudi rulers were aware that Egypt, Jordan, Yemen, and the Gulf Arab states were all internally polarised and that if the anti-government ‘pole’ triumphed in any, that state could move from the Saudi into the Iranian column. All knew that the reverse was true with polarised Syria: regime change there would likely move it out of Iran’s column and into Saudi Arabia’s.¹¹² An early flashpoint was Bahrain, a tiny Gulf state with a Sunni king ruling a majority-Shia population. Unrest in Bahrain threatened to topple King Hamad al-Khalifa, a Saudi client, and bring to power a Shia-dominated regime that would be friendlier to Tehran and more hostile to Riyadh. Riyadh faced a second problem as protests began to spread from Bahrain into Saudi Arabia’s oil-rich Eastern Province, the majority of whose population is Shia.¹¹³

Thus did group polarisation within Bahrain draw the Saudi-led Gulf coalition to invade on 14 March and suppress the unrest. The Saudi and Bahraini rulers denounced Iran for stirring up the unrest.¹¹⁴ Tehran did rhetorically support the protesters and condemn the invasion.¹¹⁵ The head of Bahrain’s main Shia party, Sheikh Isa Qassem, has close ideological affinities with Iran’s regime.¹¹⁶ A Western journalist interviewed several protesters who said they were members of the Bahraini Hezbollah (Party of God, the Iran-supported Shia party in Lebanon).¹¹⁷ Hezbollah’s leader offered to train Bahraini protesters, but denied actually doing so.¹¹⁸ All told, evidence of direct Iranian interference is difficult to come by. But the Saudi belief that Iran was behind the rebellion shows just how seriously Riyadh took Bahrain’s internal polarisation.

A second strategic country deeply polarised in 2011 was Syria, whose demographic situation is the reverse of Bahrain’s: the majority of Syrians are Sunni, while Bashar al-Assad, the country’s dictator, and most of his regime elites are Alawites, who form a variant of Shia Islam. Polarisation put Syria’s foreign alignments in question along with its domestic regime. Syria was allied with Iran; the Tehran regime had found Assad useful in keeping Hezbollah in Lebanon loyal.¹¹⁹ The Saudis, on the other hand, saw in Syria’s internal polarisation an opportunity for a regime change that would move Syria into the Saudi column.¹²⁰

Iran has helped the Assad regime from the start, beginning with the suppression of demonstrations in March 2011 using mass door-to-door raids.¹²¹ By January 2012, Iran’s leaders were committed to keeping Assad in power at a very high cost. Iran’s Quds (Jerusalem) Force and Republican Guard have trained and provided arms to the Syrian army, and the Republican Guard has fought alongside them.¹²² In February 2013, Brigadier General Hassan Shateri of the Quds Force was assassinated near Damascus.¹²³ Iran’s unstinting support for Assad is widely known and has caused its reputation in the Arab world to plummet.¹²⁴

A second Saudi rivalry – with Qatar, a small but extremely wealthy Gulf state with outsized ambitions – hampered Riyadh’s early efforts to help the Syrian opposition. The Saudis, along with Jordan and the United Arab Emirates, funnelled arms to an array of Syrian rebels, from Salafists (strict Islamists) to liberals; Qatar sent weapons to the Syrian Muslim Brotherhood, long-time adversaries of the Saudis within Sunni Islam. Neither Riyadh nor Doha trained troops or sent combat units.¹²⁵ In the spring of 2013 the Saudis met with the Free Syrian Army and the Qatari role shrank. Complicating Saudi aims has been the emergence of the so-called Islamic State (a.k.a. ISIS, ISIL, and Daesh) as at once the most potent anti-Assad force and a possible future threat to the Saudi monarchy itself.¹²⁶ Notwithstanding, Saudi and Iranian actions have been consistent with the presence of polarisation within Syria.

Conclusion

In the Middle East 2011–12, as in northwestern Europe in 1558–66, a spring in one country was followed by a tangle of regional events, including repressions and reforms, unrest, revolutions, civil wars, and foreign interventions. I have argued that transnational group polarisation is the causal mechanism connecting the initial spring and the other events. Transnational group polarisation is the spread not only of unrest and rebellion, but of resistance to it. The segregation of preferences into anti- and pro-government took place along pre-existing but latent cleavages within and across the states in question. The Tunisian Spring, an exogenous event, activated these cleavages across Arab states in early 2011, presenting rulers with new threats and opportunities and interacting with other conditions to produce the peculiar outcomes of the Arab Uprising. A great deal more theorising and empirical testing is needed, but it is clear that what James N. Rosenau called linkage politics¹²⁷ is real and highly consequential for international security. In regions where people have social ties across state boundaries, a spring can produce all manner of offspring across and among countries, interacting with structures and disrupting normal threats and opportunities.

This article has addressed only half of the larger causal question. If transnational group polarisation is indeed the causal mechanism, we are left with questions of correlation. Why do some sustained liberalisations in one country not set off such chains of events across regions? Under what conditions are springs barren? A recent example comes from May 2009, when the Iranian Green Movement tried to nullify Iran’s allegedly fraudulent national election. For some days it appeared that the Green Movement might prevail, and yet there was little spill-over into other authoritarian Muslim countries. In particular, the Arab societies that were to see revolts in early 2011 did not register much change in May–June 2009. It may be that one of the scope conditions of transnational polarisation did not obtain – namely, extensive social ties across national boundaries. Although Persians and Arabs have links, particularly through Shia networks, linguistic and ethnic differences may mean that these networks are not as robust as those among Arab societies. Marc Lynch refers to an Arab public square, not a Muslim one.¹²⁸ Yuri M. Zhukov and Brandon M. Stewart advise scholars to theorise about what constitutes a region,¹²⁹ and the case at hand confirms the wisdom of that advice.

Another question arises when we consider still a different, more hopeful, kind of outcome. In 1989 the peoples of Eastern Europe revolted and were met not with repression and international intervention but a cascade of mostly peaceful revolutions unopposed by a great power (the Soviet Union) with vast leverage and a great deal at stake. Doubtless there was some transnational polarisation – the communist regimes retained some supporters who would have strongly preferred to crush the rebellions – but most internal security forces and high communist officials themselves defected or acquiesced to the anti-government side. Under what conditions, then, will polarisation be so imbalanced that the anti-government group wins decisively, in country after country? To what extent do conditions within states, as opposed to with international-structural conditions, affect outcomes? The answer may tell us much about why some waves of revolution succeed and others do not.¹³⁰

For now, we have good reason to believe not only that unrest sometimes spreads from state to state, but also that a determination among governments to pre-empt and suppress it can spread along with it. When anti-versus-pro-government polarisation is transnational, it can create new threats and opportunities that lead to complex tangles of events. It is important that we see not only the events but also the tangle, and that explaining the former entails explaining the latter.

Biographical information

John M. Owen (PhD, Harvard) is Ambassador Henry J. and Mrs. Marion R. Taylor Professor of Politics, and a Faculty Fellow at the Institute for Advanced Studies in Culture, at the University of Virginia. He is author of Confronting Political Islam: Six Lessons from the West’s Past (Princeton, 2015); The Clash of Ideas in World Politics: Transnational Networks, States, and Regime Change, 1510–2010 (Princeton, 2010); and of Liberal Peace, Liberal War: American Politics and International Security (Cornell, 1997), and co-editor of Religion, the Enlightenment, and the New Global Order (Columbia, 2011). He has published in the European Journal of International Relations; Foreign Affairs; International Organization; Internationale Politik und Gesellschaft; International Relations; International Security; International Studies Quarterly; National Interest; The New York Times; Perspectives on Politics; and a number of edited volumes. From 2011 through 2014 he was Editor-in-Chief of Security Studies. During 2015–16 he is a visiting scholar at the Free University of Berlin and the Social Science Research Center Berlin (WZB), supported by a Humboldt Research Award.