Crises and critical junctures in authoritarian regimes: addressing uprisings’ temporalities and discontinuities

Abstract In this article, we aim at sharpening common understandings of the notion of political crisis to better explain the trajectories of authoritarian transformations during popular uprisings. We make three major claims. First, we propose a definition of crisis as brief moments of institutional fluidity and openness in which a process can take different directions. We delineate the crisis concept from the concept of critical junctures and outline how our approach contributes to the methodological debate on ‘near misses’. Second, we indicate how the de-institutionalisation processes leading up to a crisis are to be analytically distinguished from within-crisis moments. We argue in favour of a discontinuity approach that takes into account the different temporalities of gradual lead-up processes and rapid within-crisis dynamics. Finally, we illustrate our theoretical and analytical reasoning with concrete cases from the authoritarian crises of the Arab uprisings, whilst suggesting that our argument can travel to other areas of research in which crisis narratives have gained prominence.


Introduction
'Crises' seem to be among the most unreflectively deployed concepts in the social sciences.
The understandings of what crises actually mean differ considerably. This is a predicament that is hardly new; some 20 years ago already Hay (1999, 317) complained that crisis 'is one of the most underdeveloped concepts in state theory' . He ventured that the term's ubiquity in the literature derived 'precisely from this notorious imprecision' (Hay 1999, 318). Crises can be of structural and long-lasting nature (eg Crozier, Huntington, and Watanuki 1975) or can be brief, tumultuous and fatal moments in time in which actors' decisions become paramount (eg Linz 1978). Crises can also be either acute vs chronic, or manifest vs latent. A crisis can affect the whole functioning of a (political) system or might affect only some parts (Gerschewski 2018). At one end of the spectrum, we find the literature on state crisis and state failure and collapse that addresses the systemic breakdowns of state governance ARTICLE HISTORY (Beissinger and young 2002;eriksen 2011). At the other end, we have the literature dealing with political leadership during crisis episodes (Ansell, Boin and Hart 2014). Whilst the last few years have been dominated by the financial crisis and the (so-called) refugee crisis, there were also repeated suggestions that democracy itself was in crisis, at a time when various autocracies were clearly in crisis as well (eg during the 2011 Arab uprisings).
We do not aim to provide a final answer to the way the terminology of 'crisis' should or should not be used as we recognise that part of its attractiveness resides precisely in that it is a catch-all term, sometimes used as a place holder before a more precise account of the problem is provided. Instead, we build our argument on Hay's (1999, 323-324) point that a more precise crisis terminology can be 'based on the analytical distinction between failure (an accumulation or condensation of contradictions) and crisis (a moment of decisive intervention during which these contradictions are identified)' . In this light, we suggest a specific conceptualisation of political crisis that can be usefully deployed in the literature on democratisation and authoritarianism. In particular, we make two important clarifications, to the concept of crisis and to the empirical analysis of crisis episodes.
First, we propose that we can sharpen the notion of crisis by delineating it from the neighbouring concept of critical juncture. We define a critical juncture as a probability raiser for change that is inherently biased towards 'positive' cases. The critical juncture framework has its major analytical strength in explaining institutional pathways after the juncture happens. In contrast, a crisis episode is a moment of fluidity and openness. The analytical strength of the concept of crisis is that it emphasises how contingency creates new political identities and dynamics at the early stages of what might develop later into a critical juncture. Methodologically, we argue that by sharpening the crisis concept we are better equipped to incorporate cases of 'near misses' or 'negative outcomes' -ie moments in which change was likely, but did not happen (Mahoney and Goertz 2004;dunning 2017).
Second, we propose that a discontinuity approach constitutes the most useful way to organise the empirical study of crisis episodes. The discontinuity approach marks a clear difference between the lead-up to a crisis and the within-crisis dynamics. This approach underscores the analytical value of detaching the within-crisis situational logics from previous developments that led to this moment. Within the critical juncture literature, there is a debate regarding whether a crisis turns into a critical juncture due a preexisting causal dynamic or not (Slater and Simmons 2010;dunning 2017). Here we explicitly argue that due to different temporalities, what leads to a crisis does not necessarily matter in a crisis. The lead-up to a crisis moment and the situational logics within a crisis should be systematically distinguished from each other. They can be integrated into one coherent framework of analysis, yet researchers should always account for potential discontinuities. In particular, we articulate how the production of new within-crisis identities and strategic choices can be distinguished from prior causal dynamics and how they can be included in the overall analysis of a crisis and its outcomes. To illustrate the pertinence of these theoretical and analytical perspectives, we revisit some of the recent political crises in contemporary authoritarian regimes in the Middle east during the 2011 Arab uprisings.

Crises vs critical junctures
A critical juncture can be described as a moment in which change is of heightened probability. The juncture is critical because it is more likely that a specific long-term institutional outcome (legacy) is produced (Collier and Collier 1991;Collier and Munck 2017). In one elegant definition, Capoccia and Kelemen (2007, 348) introduce critical junctures as 'relatively short periods of time during which there is a substantially heightened probability that agents' choice will affect the outcome of interest' . They argue that the criticalness of a critical juncture can be measured by two components: its temporal leverage and its probability jump (Capoccia and Kelemen 2007, 360-363). Temporal leverage means that the episode is all the more critical the briefer it is compared to the resulting institutional outcome. The probability jump, in turn, refers to the idea that the more critical this episode is, the more it affects the probability that an institutional legacy is produced. In other words, a very critical 'critical juncture' would be one in which a short sequence of events produces a long-lasting institutional outcome that had a very low probability of happening at the start of the sequence. A critical juncture is thus understood as an incubator of change.
Soifer (2012) proposed a notion of critical juncture that went beyond the idea of Capoccia and Kelemen. He introduced a 'causal logic of critical junctures' (Soifer 2012(Soifer , 1572. This causal logic lies in the separation between permissive conditions that allow for a loosening of structural constraints, and productive conditions that act within the boundaries of the opened possibility space and produce the outcome. He argues that a critical juncture is characterised by the presence of permissive conditions and the presence of productive conditions (Soifer 2012(Soifer , 1580. It is a moment in which change is very likely due to the presence of both the opportunity and the push for change. Here, it becomes even more evident that the study of critical junctures has an inherent bias towards 'positive cases' (dunning 2017).
Importantly, however, the literature on critical junctures retains a degree of ambiguity when it comes to qualifying 'near misses' . Collier and Munck (2017, 6) insist that a critical juncture needs to produce a legacy, which is 'an enduring, self-perpetuating institutional inheritance of the critical juncture that persists and is stable for a substantial period. If a legacy in this sense does not emerge, then the prior episode is not considered a critical juncture' . In a similar vein, Slater and Simmons (2010, 888) define critical junctures as 'periods in history when the presence or absence of a specified causal force pushes multiple cases onto divergent long-term pathways, or pushes a single case onto a new political trajectory that diverges significantly from the old' . They further demarcate critical junctures from other types of transformations by stressing the importance of what they call 'critical antecedents' , which are 'factors or conditions preceding a critical juncture that combine with causal forces during a critical juncture to produce long-term divergence in outcomes' (Slater and Simmons 2010, 889). other scholars, by contrast, seek to include different types of 'near misses' . As Capoccia and Kelemen (2007, 352) explicitly note: 'contingency implies that wide-ranging change is possible and even likely but also that re-equilibration is not excluded' . In their view, 'change is not a necessary element of a critical juncture' . Bernhard (2015) adds that a critical juncture might not always produce a stable legacy, but that it can result in 'chronic instability' and the rapid succession of regimes. From this perspective, instability, despite its lack of specific institutional content, constitutes a type of legacy that makes the preceding episode qualify as a possible critical juncture.
These differences underscore a significant conceptual disagreement regarding when it is appropriate to analyse specific episodes of change as critical junctures. In effect, despite important works on the gradual and endogenous developments that lead to a critical juncture (Streeck and Thelen 2005;Thelen and Mahoney 2010), the bulk of the literature analyses cases that produce specific legacies. The majority of recent scholarly works are more interested in the legacy effect of critical junctures and the creation of path-dependent developments than in 'near misses' . As a result, the literature tends to overemphasise positive cases and does not problematise identifying 'near misses' .
In this light, we argue that our approach to crisis is well suited to address these 'near misses' . We follow in the footsteps of Mahoney and Goertz (2004) regarding the 'possibility principle' . We agree with these scholars that the identification of negative cases is crucial and consequential for theory building and testing, but is often not as straightforward as one might assume. Negative cases are cases for which an event was possible, but the expected outcome did not materialise. The critical junctures literature, with its emphasis on path-dependent processes producing legacies, commonly adopts a forward-looking explanatory approach that starts from a well-defined episode. Certainly, all investigations of critical junctures involve some reflection on preconditions. Illuminating studies like the one by Collier and Collier (1991) on labour movements in Latin America, or the study of Mahoney (2001) on the legacies of liberalism, include antecedent conditions that define the range of historical options available to the actors. Nevertheless, the main point of the critical juncture approach (at least from the perspective of the empirical scholars) is to understand what follows from such watershed moments and how legacies crystallise.
In contrast, crisis is a concept that is less directional and more attuned to the uncertainty and fluidity of a particular episode. Crisis does not include an in-built quest for significant legacies and is therefore more attuned to explaining why and how these episodes occur in the first place. In our understanding, a crisis is a precondition for the possibility of change. In other words, there are crises that do not produce critical junctures, but there are no critical junctures without prior crises. A crisis is necessary but not sufficient for a critical juncture. This has important consequences for capturing the re-equilibration processes taking places in 'near-miss' cases. When the focus is primarily on events that have a lasting legacy, then we over-value the mechanisms found in 'successful' critical junctures and under-value those found in episodes where change was likely, but the situation reverted to the situation ex ante. By and large, 'near misses' are characterised by what they lack to become actual critical junctures with a lasting legacy.
The analytical focus on crisis enables us to highlight the specific temporality of these episodes, and particularly the acceleration of the tempo of change, as described by Grzymala-Busse (2011). This acceleration of tempo and the contingencies associated with it increase the relevance of situational logics for the strategies of the actors. In this context, as outlined by Kurzman (2004) andermakoff (2015) in different revolutionary situations, confusion and mutual uncertainty become important factors of change. Whilst these dynamics, and the actors and strategies creating them, may be transient, accounting for these factors helps refine the analysis of the trajectories of crises and their eventual institutional legacies (from state collapse to democratic consolidation).

Crisis in authoritarian regimes: lead-up vs within-crisis dynamics
As an empirical illustration, we use crises in authoritarian regimes. We follow the broad definition of authoritarian regimes by Geddes, Wright, and Frantz (2014). They identify three ways by which an authoritarian regime starts to exist: undemocratic elections; democratic elections but a subsequent change in formal and informal rules that inhibits competition; and competitive elections, but the military preventing parties from competing. our argument does not directly engage with the debates concerning the different types of authoritarian rule, though we realise that our approach could be refined by studies analysing specific sub-categories of authoritarianism.
We deliberately use authoritarian systems in order to bring into sharp focus the issue of political crises in regimes with less reliable institutionalised mechanisms of de-escalation. We agree with the recent institutionalist literature that highlights how autocratic regimes can also rely on party structures, parliaments, judiciaries and elections, but also informal rules, to maintain their rule (Magaloni 2006;Brownlee 2007;Ginsburg and Moustafa 2008;Schedler 2013;Morgenbesser 2016). However, we stress that whilst authoritarian regimes use court rulings, snap elections, 'independent' investigations and executive concessions in order to de-escalate tensions, they do not have dependable political mechanisms to address the challenge of mass protests directed at the incumbent elites. Whereas in democratic systems elections provide an institutionalised mechanism to diffuse popular discontent by rotating the elites, holding elections in authoritarian regimes (including those with non-competitive elections) in response to a crisis usually heightens the extraordinary character of the crisis more than it de-escalates tensions. At the time of writing, a vivid illustration of this situation is the wave of mobilisation sweeping Algeria against the holding of elections organised by the regime (Grewal, Kilavuz, and Kubinec 2019).
Phenomenologically, we focus here on one type of authoritarian crisis that displays three main characteristics: (1) an explicit public challenge to the regime (eg in the form of street protests), (2) a sustained level of mobilisation that disrupts routine authoritarian governance (eg police or army mobilisation, partial shutdown of institutions), and (3) the emergence of explicit intra-elite challenges (eg a split between soft-and hardliners).
Typically, these crises are the results of two main processes of de-institutionalisation: gradual and rapid. While the first one is stretched over a longer period of time and accounts for incremental change that leads to a crisis situation, the latter takes into account fast and unexpected shifts that disrupt governance routines in a short period of time. These disruptions are identifiable in terms of both formal de-institutionalisation (eg imposition of martial law, suspension of the constitution, of habeas corpus, etc.) and de facto de-institutionalisation (eg branches of the state administration not having the means or the personnel required to function beyond minimal capacity or outside specific locations).

Gradual de-institutionalisation and lead-up to crisis
Consider the following citation by Brownlee: The collapse of a dictatorship seems a sign of the supremacy of the human will. Jubilant masses flood the streets, statues fell, and parliaments reawaken with new faces. At that moment, the 'inhuman' forces of social structure and political organization appear peripheral the action at best. yet, in another sense, the victory of dissidents and activists is but the final act of a long drama. (Brownlee 2007, 17-18) To analyse the 'long drama' is to address the underlying, slowly moving factors that shape the lead-up to the crisis moment. In this perspective, the analysis must consider cumulative causes in which a slow but steady change in one condition is observed over time and leads gradually to a crisis. The incremental change can either be observed from the very beginning or take place for a significant period of time under the radar, ie without observers noting it at first until it leads ultimately to a big 'blow' (Pierson 2003). In addition, these cumulative causes can present themselves also as a chain of interrelated phenomena in which change in one condition triggers a sequence of (prospectively) unexpected events. While the first process corresponds phenomenologically to the evolution of one set of conditions A→A→A→y, the other entails a sequence of distinct phenomena A→B→C→y. What both have in common is a prolonged time frame in which these sequences unfold.
A prominent macro-structural example of the structural cascade pathway (A→B→C→y) is the path-dependent explanation of regime outcomes in Central America by Mahoney (2001). It explains the differing regime outcomes of military authoritarianism, traditional-dictatorial regimes, and democratic polities. Mahoney's explanatory model of antecedent conditions, critical junctures, and reactive sequences is of course much more elaborated than the structural cascade pathway that we propose here. But what is important for our argument is that Mahoney builds his influential account on distinct and sequential phases of a development towards a regime outcome. each stage programmatically and explicitly sets the parameters for the next one.
This structural cascade contrasts with the mounting up of one set of (observed or unseen) conditions (A→A→A→y) leading to a crisis. A well-known illustration is that of mounting economic difficulties leading to a major political crisis. In the 1980s, there was a wave of crises in developing countries induced by the rolling out of structural adjustment plans devised by the International Monetary Fund (Walton and Seddon 1994;Sadiki 2000). In those cases, a progressive worsening of socio-economic conditions for a large part of the population caused by the rolling back of welfare provisions led to widespread social unrest during so-called 'food riots' .
These gradual developments are usually time-consuming and extend over a longer period of time. They form the underlying, slow-moving causes and represent the lead-up to a crisis moment. Their temporalities differ in most cases from within-crisis dynamics in three fundamental ways: their duration is long, their tempo slow and their acceleration low.

Rapid de-institutionalisation and within-crisis dynamics
The second process is a rapid process of deinstitutionalisation triggered by the actions of professional or amateur political entrepreneurs. As such, it is more an actor-driven process that usually takes place in a short time span and that leads to a disruption of governance routines. In generic terms, these dynamics have their roots often in a cognitive disequilibrium between the objectives of the actors and their achievements (Aoki 2001). In authoritarian systems, more specifically, they are also often an outcome of the dynamics of expressing preferences publicly -ie when preference falsification (Kuran 1995) is overcome. From this perspective, the positionality of the actors deciding when 'enough is enough' or when there is an opening for them to safely express dissent in large numbers is key to suddenly upsetting institutionalised mechanisms of authoritarian governance. empirically, rapid de-institutionalisation processes can be illustrated by the wave of 'colour revolutions' of the 2000s. elections that were planned by regimes as a means to prolong the authoritarian status quo backfired. They produced a situation of heightened delegitimisation and strengthened the political opposition. The electoral debacle in turn created the condition for an open post-electoral challenge by a reorganised opposition movement (Bunce and Wolchik 2010). extra-constitutional post-electoral mobilisation produced in many cases a rapid change of regime. In this context, electoral mismanagement highlighted how actors' miscalculation, overreaction and/or improvisation constituted short-term causes for a rapid de-institutionalisation process. The dynamics of the Arab uprisings that we illustrate in the next section provide more insights into how within-crisis dynamics produce specific processes of rapid de-institutionalisation of authoritarian regimes. Before we do so, we highlight the main tenets of our methodological suggestion of a discontinuity approach. It emphasises the distinct nature of long-term lead-up to crises on the one hand and situational logics within crisis dynamics on the other hand.

The discontinuity approach
We propose a discontinuity approach that explicitly distinguishes between the lead-up to the crisis that is usually connected to gradual deinstitutionalisation processes and within-crisis dynamics that are usually connected to rapid deinstitutionalisation processes. Grzymala-Busse's (2011) perceptive analysis of the role of temporality draws special attention to the importance of disentangling them from each other. due to their different temporalities with respect to duration, tempo and acceleration, we argue that what is important for the emergence of a crisis does not necessarily matter for what happens in the crisis itself. dobry (2009) explains the articulation of situational logics onto prior causal mechanisms by analogy with Clausewitz's argument about war being the continuation of politics. He notes that although war 'emerges out of the very diverse political ends or aims pursued by states or their governments' , once it breaks out something new becomes noticeable: 'because of the situational logic that engulfs its protagonists, war tends to develop logics of its own, independent from its original political end, cause or determinant' (dobry 2009, 79).
In our argument, we make a similar point regarding political crises. While the emergence of these crises may be understood as an outcome of specific combinations of local, national and international factors and strategies, once there is a crisis situation, new factors and strategies generated by the crisis itself gain prominence. Importantly, we emphasise that we cannot assume that 'within-crisis' dynamics only matter when we have deviant outcomes, and that they are irrelevant when the crisis outcomes are in line with structural trends. Tracking how institutions fail (and are then rebuilt) enables the identification of false positive cases -ie cases in which post-crisis institutional outcomes align themselves with the predicted evolution of a particular regime but not due to the reasons posited in structural models. As Kurzman (2004b) illustrated in the case of the 1979 Iranian revolution, analysing rapid sequences of de-institutionalisation not only provides insights into how unexpected outcomes come about, but also highlights the causal pertinence or irrelevance of structural factors for strategic choices.
Methodologically, the discontinuity approach asks us to first place ourselves at the beginning of the lead-up to a crisis and to observe empirically the factors behind its unfolding. Secondly, we should place ourselves anew at the beginning of the crisis itself in order to evaluate specifically within-crisis dynamics. This second step is crucial to identify the views, actions and strategies of old and newly emerging protagonists as they are expressed at the time (and not as they might be second-guessed by an external observer (Kurzman 2004b;ermakoff 2015)). This discontinuity approach does not mean that causes prior to and in crises necessarily diverge. However, this approach ensures that explanations of what happened in a crisis do not fall into either the extrapolation trap or the post hoc rationalisation trap. Below, we illustrate our reasoning with empirical case studies. We focus on the 2011 Arab uprisings in which specific within-crisis dynamics reinforced or disrupted the trajectories of de-institutionalisation of the authoritarian regimes of the Middle east and North Africa.

Structural trends and strategic choices during crisis: the case of the Arab uprisings
during the 2011 Arab uprisings, the sequence of mobilisation that occurred in Tunisia between 17 december 2010, when Bouazizi set himself on fire, and 14 January 2011, when President Ben Ali fled the country, illustrated well in-crisis dynamics. While the factors that led to the crisis -economic failure, suppression of dissenting voices, etc. -provided grounds for people to revolt, it remains controversial how far they directly shaped the way the protest unfolded and the resulting institutional reconfigurations (Gana 2013). While the established pro-and anti-regime actors were actively implementing their preferred strategies, they were repeatedly unable to control the protests or cope with the (intended and unintended) consequences of protest interactions. Instead, a leaderless protest movement began to set the tempo and sequence of political change in the polity (Volpi 2017).
It was a leaderless (or leader-full) movement in the sense that important events and processes were repeatedly triggered in an ad hoc way by groups and individuals who were not previously significant actors in the Tunisian political landscape. Whereas the routine governance of discontent in the country and the previous management of crises set the parameters for managing known protest actors (Camau and Geisser 2003;Chouikha and Geisser 2010), the 2011 uprisings gained additional momentum by involving new categories of actors and practices. As we indicated in our theoretical outline, the lead-up to the crisis was important for the emergence of the protest movement, but it did not determine the type of actors who would direct the movement or the modalities of contestation, which were the results of within-crisis dynamics.
Situational logics are crucial here because new actors of change are guided more by achieving (often confusedly conceived) local, short-term objectives than nationwide or programmatic institutional transformations. Allal (2011) provided a vivid description of how young Tunisia 'revolutionaries' seized the opportunities created by mass unrest to reorganise local politics and economic activities to their advantage in their suburbs and to become the new 'security' providers. Tellingly, whereas organised activists and institutional actors may have had grand projects of reform, in the public sphere talks of a 'revolution' and of 'regime change' only became noticeable after several weeks of popular protests (Volpi 2017). Initially, it is a sum of uncoordinated collective endeavours -each one a reflection of particular local circumstances and interactions -that directly shaped the options for governance at the national level at that moment (Mekouar 2017). The flight of President Ben Ali is illustrative of an institutional transformation triggered by the situational logics of unrest in Tunis. When protests came too close to the seat of power and Ben Ali began to have doubts about the willingness and/or ability of the security forces to guarantee his safety, he decided to leave the country and in so doing deepened and accelerated the process of deinstitutionalisation of the regime (Jebnoun 2014;Volpi 2017).
The strategic choices of the president during the Tunisian uprising underscore the tension between situational logics and longer term strategies in times of crisis. Situational logics engulf elites as much as other actors. For newly mobilised grassroots protesters or previously depoliticised actors, the situational logics of crisis provide grounds for discarding earlier strategies of quiescence to routine authoritarian governance (el Chazli 2018; Pearlman 2018). For elites, in-crisis situational logics could also supersede established strategies of governance and repression (Goodwin 2011;Jebnoun 2014). Some of their choices reflected the longer term evolution of the power balance between different factions of the ruling elites -typically soft-vs hardliners -in which case the crisis merely provided an opportunity to implement established strategies of institutional takeover. other choices, by contrast, did not appear to fit well within pre-existing trends and were instead primarily an ad hoc product of the crisis itself (eg the flight of Ben Ali to Tunisia or, in a different register, the decision of the Algerian regime not to use lethal force to repress the uprising in the country; Volpi 2013). As we discussed in our theoretical introduction, those strategic choices that are shaped more by the tempo and acceleration of a crisis than by an evaluation of longer term costs and benefits are both an outcome and a cause of within-crisis dynamics.
The role of the military during crises is illustrative of the tension between these logics and temporalities of change. empirically, the 2011 Arab uprisings presented the established scholarship with a number of difficulties as a series of supposedly strong security apparatuses and regimes were forced to revise their strategic and tactical aims in the face of mass protests (Bellin 2012). In this context, to note the congruence of the military elite's preferences before and after the crisis in order to highlight a posteriori the coherence of their strategies is not entirely satisfactory. It certainly helps to make sense of some of the changes that occurred but it does not demonstrate that these choices were made and these strategies were implemented. In this situation, Goodwin (2011, 454) thus summarised the institutionalist rational-choice perspective: 'the more professional and institutionalized armies in Tunisia and egypt calculated that they could best safeguard their interests by abandoning dictators' . But querying how far this estimate could have been given at the time of the uprisings, he concluded that 'the structural characteristics and dispositions of armies often become apparent only after they begin to fight for their survival' .
during the uprising, protesters actively sought to entice the security forces deployed on the street to join them, with various levels of success (Ketchley 2014;Mekouar 2017), and in so doing they introduced a specific in-crisis factor that weighed upon the decisions of the military leadership. While the issue of the institutional cohesion of the military is a long-term concern of any armed forces, specific within-crisis circumstances requiring a shift from targeted covert coercion to open mass repression not only changes the strategic calculations of existing 'security' forces (as in egypt or Tunisia), but can also create new armed actors and strategies (as in Libya and Syria). In our argument, we draw attention to the variations in the institutional trajectories of regimes that are introduced by the situational logics of the crisis. In particular, we stress that we cannot mechanically extrapolate the factors that mattered prior to crisis to the crisis itself.
This last observation acts as a counterweight to the structural analyses of contemporary political crises that tend to offset contingency-generated strategic choices by invoking longterm calculations and institutional reequilibration. For Brownlee, Masoud, and Reynolds (2015, 62), protesters during the 2011 Arab uprisings 'may have perceived a structural opportunity where one did not exist' , and their mobilisation succeeded 'partly because of a misperception that regimes were vulnerable' . In their explanation, the dynamics of within-crisis agency are stripped of most of their rationale, and the relevant causal dynamics are named a posteriori in view of measurable institutional outcomes. As a result, the only possible outcomes of these crises were the ones we obtained, and within-crisis agency had a causal impact only insofar as it purposely or inadvertently facilitated pre-existing trends.
In relation to earlier revolutionary episodes in Southeast Asia, Slater (2009, 206) constructed a similar argument when he noted that 'unless we do the historical work to uncover where oppositional political cultures come from, it is difficult, if not impossible, to explain why so many societies have produced neither democratic revolutions nor large-scale authoritarian crackdowns but persistent quiescence' . In this context, Slater limits the relevant within-crisis strategies to the ones that articulate themselves onto longer term nationalist and religious repertoires of established communal elites. While we do not discount these factors out of hand and acknowledge their particular importance for comparative work (Møller 2013), the study of within-crisis dynamics is about creating a space for analysing the sui generi causal impact of new actors and strategies. Capoccia and Ziblatt (2010) noted that during a critical juncture there are multiple possible scenarios which are equally plausible, and only process tracing can show how one set of outcomes came into being. Methodically, we stress the relevance of their argument about episode analysis. We strongly concur with their point that we should 'go back and see' . However, we also stress that to do so effectively, the institutionalist analysis has to problematise more systematically the creation of new actors, and not have as its default position that the actor's identity is somewhat known. We focus here on those identity markers that are shaped by institutions (which facilitate or constrain specific views and practices) and that are therefore tied to institutional capacity and effectiveness (North 1990;Clemens and Cook 1999). This known identity of the actors is implicit in Capoccia and Kelemen's (2007, 354-355) argument when they state that episode analysis should reconstruct each step of the decision-making process, discuss the availability and viability of options in the eyes of the actors, and uncover the relation to and effect on other decisions. unlike in the social movement literature (Jasper and Volpi 2018), from an institutionalist perspective, less attention is paid to how new actors become relevant actors -a tendency that is compensated by a greater focus on the novel decisions taken by known actors in new circumstances. Ahmed and Capoccia (2014) illustrate this point in their analysis of the 2011 Arab uprisings in egypt. The focus of their explanation is on the crucial role played by an established Islamist organisation (Muslim Brotherhood) and workers' unions. yet other cases during the Arab uprisings provide clear examples of how the causal impact of newly created actors shaped specific trajectories of crisis. As we outlined in our theoretical presentation, within-crisis dynamics are producing not only new factors and strategies that were not entirely identifiable in the lead-up to the crisis, but also new actors' identities.
A powerful illustration of these crisis-generated identities is the role played by some of the revolutionary brigades during the Libyan uprising. Whereas Libya may now be turning into a model of chronic instability, many of the actors responsible for this situation were not present (as political actors) before the uprising. Not only were they not present as structured organisations, but pre-existing cleavages within Libyan society did not indicate that such entities were about to emerge at this juncture. The formation of armed movements was easily conceivable along tribal, ethnic or religious lines (Anderson 1986); but the within-crisis contingency of the uprising opened up new opportunities for city-framed identities to emerge without clear antecedents. The rise to prominence of the armed brigades of the town of Misrata during 2011 shows how new collective actors emerged and became key players of the Libyan crisis as a result of the conflict itself (Lacher 2011;McQuinn 2015). A strong local identity that challenged and remodeled previous affiliations was created in a previously unremarkable part of the country directly as a result of the siege of the city by Gaddafi's forces. At least in part due to the NATo military intervention at that time, the strategies of these actors that asserted their newly created identity and newly defined interests became a crucial factor in locking in the institutional trajectories of instability that became entrenched in the country. one of the major merits of the discontinuity approach is to facilitate the analysis of such new players who are the product of situational logics and within-crisis dynamics.
Most of the crises of the Arab uprisings did not generate sufficient de-institutionalisation to produce new identities that could supersede pre-established identities and protest practices. Commonly, as could be seen in Jordan (Schwedler 2018), within-crisis dynamics gained momentum as different opposition actors jumped on the protest bandwagon and mobilised their supporters. yet this momentum quickly fell as the regime negotiated with these different groups to ensure that their mobilisation remained within existing repertoires of contention. As these protest dynamics did not underpin novel political subjectivities, the regime was able to tolerate these expressions of dissent. Nonetheless, even in these contexts, the notion of crisis is useful to analyse how contingency works to create new political dynamics in the early stages of a potential critical juncture. Before having a situation of increasing returns which makes specific identities and strategies the recognisable drivers of institutional reconfiguration (Pierson 2003), we have a period of indeterminacy when new contingent identities and practices that are only loosely connected to pre-existing cleavages in society are able to emerge.
The focus on within-crisis processes enables us to bring to light those causal factors and actors having a transient but decisive impact on the institutional trajectory of regimes. Identifying the role of those protagonists coming together as politically relevant collective actors during a crisis is an important first step to explaining changes in the strategy of established actors. As dobry (2009) stresses, another part of this inquiry is to investigate how some identities are less likely to reconfigure themselves even as social and political fluidity increase, thereby empowering entrenched actors and practices. Just as the institutional constraints on identity are never so strict that the actors cannot evade them (Thelen and Mahoney 2010), within-crisis fluidity is never so absolute that institutional factors are totally irrelevant. Thus, alongside identifying how novel actors and strategies are the product of crises, recognising how institutional influence can remain effective in situations of deinstitutionalisation is part of the equation. These two dynamics help to better explain the direction of change at specific within-crisis junctures. They may also have a longer term impact on the trajectories of institutional change and the successes and failures of institutional legacies.

Conclusion
In this article, our analysis highlighted the difference between critical junctures and crises. Substantively, whereas crises are moments of heightened political fluidity and uncertainty, critical junctures are biased towards 'positive cases' and long-lasting legacies. In turn, the concept of crises represents a more open-ended set of political events. Analytically, the study of crisis is more focused on explaining the causal dynamics present at this specific moment; it is therefore better able to capture near misses and negative outcomes.
We argue that changes in causal interactions within a crisis can be mapped in the same way as we map the dynamics leading up to crisis. The crisis dynamics are linked to the longer term evolution of the factors stabilising (and destabilising) an authoritarian regime, but are distinct from them. In a structural logic, crises result from an accumulation of small changes either in one causal factor (mounting pressure) or in a sequence of interlocking factors (cascade). Whereas crises are articulated on those predictable factors (and actors), they are also shaped by rapid processes of de-institutionalisation set in motion by the more ad hoc strategic choices of the protagonists who purposely or unintendedly disrupt routine authoritarian governance.
For within-crisis dynamics, we present the contingencies created by the crisis itself as a distinct causal factor. Here, far from being the outcome, the crisis becomes the cause of change. As the pace and tempo of change increase (Grzymala-Busse 2011), the crisis generates confusion (Kurzman 2004) and mutual uncertainty (ermakoff 2015) which shape decisively the strategic choices of the protagonists. Within-crisis dynamics can therefore be quite distinct from the interactions that led to the crisis (and, by extension, from those following the crisis). We stress the need to avoid extrapolating the relevance of causal factors that were important before a crisis to the crisis itself. only an in-depth examination of crisis episodes can establish how far earlier factors continue to be relevant and how far they are superseded by within-crisis dynamics. Importantly, this demands a more systematic inclusion of new identities and the strategies generated by the crisis itself.
The discontinuity approach that we outline summarises these methodological considerations that ought to be part of a comprehensive explanation of crises. It emphasises that during crises, causal interactions need to be re-examined and re-weighted to reflect situational logics, and the outcomes need to be also explained in relation to these dynamics. A discontinuity explanation grounded in process tracing thus includes both the strategic choices corresponding to the longer term considerations of the protagonists, and the choices generated by the situational logics of the crisis. While longer term strategic choices and causal interactions can be approximated more readily by structural accounts focusing on the outcomes of a crisis, a discontinuity approach constitutes a crucial step in integrating within-crisis causal processes into a comprehensive explanation of institutional change.

Disclosure statement
No potential conflict of interest was reported by the authors.