Agency in historical institutionalism: Coalitional work in the creation, maintenance, and change of institutions

Institutionalism gives priority to structure over agency. Yet institutions have never developed and operated without the intervention of interested groups. This paper develops a conceptual framework for the role of agency in historical institutionalism. Based on recent contributions following the coalitional turn and drawing on insights from sociological institutionalism, it argues that agency plays a key role in the creation and maintenance of social coalitions that stabilize but also challenge institutions. Without such agency, no coalition can be created, maintained, or changed. Similarly, without a supporting coalition, no contested institution can survive. Yet, due to collective action problems, such coalitional work is challenging. This coalitional perspective offers a robust role for agency in historical institutionalism, but it also explains why institutions remain stable despite agency. In addition, this paper forwards several portable propositions that allow for the identification of who is likely to develop agency and what these actors do.


Introduction
Historical institutionalism sets the temporal unfolding of the stability, creation, and change in institutions center stage. Developed in response to group theories of politics and structural-functionalism prominent in the 1960s and 1970s, historical institutionalism adopted from the former that conflict among rival groups for scarce resources lies at the heart of politics, while it took from the latter the notion that the institutional organization of the polity is the principal factor structuring collective behavior and generating distinctive outcomes (Hall and Taylor 1996: 937). At the conceptual level, Fioretos et al. (2016: 6) argue, historical institutionalism emphasizes structure rather than agency, while giving priority to (material) interests over human cognition.
Yet institutions have never developed, operated, or changed without the intervention of interested groups. Agency must therefore be part of any institutional analysis, not least to be able to explain institutional change (DiMaggio 1988). In empirical research, agency has always been central to historical institutionalist scholarship. However, at the conceptual level, the role of agency in institutionalist analysis has not always been clear. Hence, in the light of the numerous recent advances in historical institutionalism (Mahoney and Thelen 2015), Fioretos et al. (2016: 15) consider attempts to develop "more actor-centered approaches" the next theoretical frontier, whereas Schmidt (2011: 2) calls to endogenize "agency in such a way as to explain the dynamics of institutional change (and continuity)." Clarifying the role of agency in historical institutionalism is challenging, because actor-centered accounts must rise above the particular episode in question. As Thelen (2010: 58) emphasizes, such accounts must generate portable propositions that allow for the identification of broader patterns of political dynamics in order to avoid "the theoretical abyss of agency unleashed." Since institutionalist approaches emerged in response to the predominance of behavioral perspectives, elements of agency must be injected with great care. Yet institutionalist scholarship may have pushed the focus on structure too far. Hence, DiMaggio's (1988: 3) observation still rings true that "the role of interest and agency in institutional theory remains somewhat obscure." This paper does not attempt to square the circle of structure versus agency by identifying the proper blend of the two in explaining social behavior. This would probably be an impossible task. Instead, the goal is to develop a conceptual framework that, starting from historical institutionalism, provides more room for human agency in an otherwise structuralist account. Hence, in this paper, agency is understood to develop under specific conditions that depend on the configuration and structure of institutions. In line with all institutional theory, the proposed conceptual framework maintains that structure holds primacy over agency.
Contributing to the coalitional turn in historical institutionalism, this paper suggests that agency plays a key role in the creation and maintenance of social coalitions that stabilize or challenge institutions. Without such agency, no coalition can be created or maintained. Similarly, without a supporting coalition, no institution can survive. Yet, due to collective action problems, activities related to the creation and maintenance of coalitions supporting or challenging institutionscalled coalitional work in the followingare demanding. This actor-centered coalitional perspective offers a robust role for agencywithin the confines of historical institutionalismbut also explains why institutions remain stable despite agency. In addition, this paper forwards several portable propositions that allow for the identification of who is likely to develop agency and what these actors do.
The discussion is organized in six sections. The next section explores the challenge of injecting agency into historical institutionalism. The goal is to demonstrate that at the conceptual level, the role of agency in historical institutionalism is not clear. Section 3 shows, using the literature on institutional entrepreneurship, that allowing for more agency leads to thorny theoretical issues. Most importantly, an actor-centered historical institutionalism must explain which actors are likely to develop agency and what explains institutional stability in the presence of agency.
The two subsequent sections address these challenges. Reviewing the (often sociological) literature on the constitutive role of institutions, the fourth section asks which actors are likely to develop agency and forwards several portable propositions. The fifth section looks at the source of institutional stability. Combining recent reflections linking historical institutionalism to coalitional analysis as well as the sociological literature on institutional work, this section emphasizes the key role of social coalitions in explaining stability and change in institutions.
The sixth section ties up the loose ends. While the coalitional perspective explains the stability in institutions, it also allows for a more agent-centric and dynamic view of institutions. The reason is the difficulty of collective action. If stability, creation, and change in institutions are a function of the mobilization of social coalitions and if such mobilization is challenging, how can these coalitions be formed and maintained? The sixth section demonstrates that a clear conceptualization of agency is decisive to understand such coalitional work. It is here that there is a role for agents, because without agency, no coalition can be created or maintained. In addition, this section discusses several strategies agents use to do coalitional work.
The concluding section discusses the benefits but also the costs of adopting a more actor-centered historical institutionalist approach. It argues that such approaches are rather challenging in terms of data requirements, which is why standard historical institutionalist approaches might be preferable if they offer sufficiently convincing explanations of the phenomena of interest. Yet, the payoffs of the more actorcentered historical institutionalism are considerable, especially in the case of arguments emphasizing endogenous and gradual institutional change (Streeck and Thelen 2005), which typically refer to competing social coalitions, but sometimes struggle to explain where these coalitions come from and why some coalitions prevail over others.

Agency in historical institutionalism
How can we inject agency into the structuralist literature on institutions? Having been created in response to the predominance of behavioral perspectives, institutionalist approaches have been primarily interested in explaining how institutions structure human action rather than how agency shapes institutions. Consequently, Hall and Taylor's (1996) influential review article had little to say about the role of agency in historical institutionalism. Instead, traditional accounts of historical institutionalism emphasize 'sticky' institutional constraints and path dependency (Pierson 2000).
Yet neglecting agency in this way (although without denying its existence), these accounts struggle to explain institutional change (Hall and Taylor 1996). In most accounts, exogenous critical junctures, defined as "relatively short periods of time during which there is a substantially heightened probability that agents' choices will affect the outcome of interest" (Capoccia and Kelemen 2007: 348), are identified as the moments during which agency matters. During these critical junctures, agency is often considered surprisingly unconstrained (Capoccia 2015; but see also Slater and Simmons 2010;Rinscheid et al. 2019). Outside such critical junctures, little room is left to agency.
Newer historical institutionalist accounts understand institutions as being permanently contested, while institutional change is more common than the critical juncture framework suggests (Streeck and Thelen 2005). Yet if institutions can no longer be assumed to be stable per definition, what explains the stability of institutions amid all these changes (Capoccia 2016)? Without some sort of stability (i.e. institutions delaying or shaping how social changes translate into some kind of policy), institutions cease to play a causal role in explanations of social phenomena (Przeworski 2004;Hall 2016).
Hence, while virtually absent in traditional historical institutionalist accounts (outside critical junctures), institutional change and by extension agency seem ubiquitous in more recent historical institutionalist scholarship. Thelen (2010: 55) thus identifies the challenge for historical institutionalism to "inject agency into institutionalist accounts in a way that rises above the particular episode in question, to generate portable propositions that allow us to identify broader patterns of politics and political dynamics." In her own work, she identified different modes of gradual institutional change and developed general propositions about the conditions under which one type is more likely to emerge (Mahoney and Thelen 2010). Yet, at the conceptual level, agency, understood as the motivation and the creativity that drive actors to break away from structural constraints (Emirbayer and Mische 1988), is still conspicuously absent from these accounts. Schmidt (2011: 5) thus claims that "not only do [historical institutionalists] do more to describe than to explain change but they also remain at a macro level of abstraction in which change can only be explained mechanistically, although now they see such change as coming not through structural shifts during a maxi-crisis but rather through structural processes at mini-incremental moments." Instead, she suggests that historical institutionalism should be enriched by a focus on transformational ideas during moments of change, which then become the basis for ideational and discursive reconstructions. This paper agrees with Schmidt (2011: 8) that "agency is the historical institutionalists' Achilles heel" and applaud her attempts to enrich institutionalist accounts by providing more room for agents' discursive activities. Complementing her new (discursive) institutionalism, this paper's ambition is to inject more agency into historical institutionalism and develop portable propositions, while staying within the confines of the historical institutionalist framework. It does so by first reviewing the literature on institutional entrepreneurship.

Institutional entrepreneurship
According to DiMaggio (1988: 14), "new institutions arise when organized actors with sufficient resources (institutional entrepreneurs) see in them an opportunity to realize interests that they value highly." In this process, institutional entrepreneurs engage in agency, which is the planned and strategic persuasion of ends based on a rational assessment of available means and institutional opportunities (Beckert 1999: 778). Institutional entrepreneurs are thus actors who purposefully leverage resources to create new, transform, or defend existing institutions (Battilana et al. 2009: 68). Such action does not imply that accounts emphasizing institutional entrepreneurship are necessarily based on rational choice theory (Whitford 2002). Institutional entrepreneurs are still socially embedded, but they use the room for agency provided by existing structures to pursue certain interests. 1 From the point of view of historical institutionalism, this notion of institutional entrepreneurship seems rather straightforward. Yet a closer look reveals important challenges. At the most fundamental level, institutionalist scholars face the problem that the more they allow agency to play a role, the less causally relevant structure becomeswhat Hall (2016) calls the paradox of plasticity, discussed below. Hence, while institutionalist accounts of entrepreneurship strive to incorporate agency, institutions must simultaneously curtail such agency. Otherwise these accounts cease to be institutionalist (Djelic 2010).
How does the literature on institutional entrepreneurship deal with the challenge of allowing for agency while simultaneously curtailing it? To address this tension, the literature often links institutional entrepreneurship to success, that is, institutional change (Battilana et al. 2009). For instance, Sheingate (2003: 185, emphasis added) defines institutional entrepreneurs as "individuals whose creative acts have transformative effects on politics, policies, or institutions." The advantage of such a conceptualization is that entrepreneurship is rare and exceptional, leading to a high level of institutional stability.
Yet it is problematic to link entrepreneurship to success. As in real life, entrepreneurs may fail. In addition, as recent contributions have argued, action is also necessary to maintain institutions. The literature on institutional work, in particular, emphasizes how institutional incumbents have to work (purposefully) to keep institutions alive (Lawrence and Suddaby 2006). However, if both institutional maintenance and change are the result of action by institutional entrepreneurs and if institutional entrepreneurs are allowed to fail, success or change cannot be used to curtail agency.
These considerations bring us to the next problem. If entrepreneurship is unrelated to success, then who is not an institutional entrepreneur (at least potentially)? Some scholars have therefore conceptualized institutional entrepreneurs as Schumpeterian 'super(wo)men'in contrast to 'managers' who act on the basis of routines (Beckert 1999). Hence, institutional entrepreneurs are simply a different species and therefore rare. While this conceptualization addresses the tension between agency and structure, it is not convincing, either, because institutional entrepreneurs are introduced exogenously. Yet, if they are different, this difference must be explained.
Rather than relying on such 'hero' conceptualizations of institutional entrepreneurs, Djelic (2010) argues that entrepreneurship must be open to everyone. However, some agents might be more likely to have the necessary properties. Generally, it seems reasonable to assume that not all actors are equally positioned to be institutional entrepreneurs. For instance, entrepreneurs need resources (financial and otherwise), must be able to mobilize others and need to be in a position to take advantage of institutional opportunities (DiMaggio 1988;Sheingate 2003;Fligstein and McAdam 2012;Béland and Cox 2016). Clearly, the potential for entrepreneurship is not set in 1 An alternative perspective in sociology starts from the observation that while the stability of institutions is the result of their constant reactivation through behavior, this reactivation's selective nature can lead to change, because structures may be multi-layered or in need of interpretation (Sewell 1992). However, this perspective insists "on the determining power of institutions and suggests that change results from the accumulation over time of acts not necessarily directed to generate it" (Dorado 2005: 385). Hence, this selective reactivation is not purposive. In terms of injecting agency, this perspective does not go beyond the bare minimum. stone but fluctuates over time and is a function of the momentary position in and the situation of the institutional field. 2 Nevertheless, actor-specific attributes such as resources or access to networks should influence who is more likely to become an institutional entrepreneur.
The literature on institutional entrepreneurship thus raises two questions any conceptualization of agency in historical institutionalism must answer. First, what actors are most likely to develop agency? Second, what keeps institutions stable if they are the objects of constant strategic considerations? The following sections address these two questions. Drawing on the sociological literature on the constitutive role of institutions, the next section forwards several portable propositions about which actors are most likely to develop agency.

The paradox of embedded agency
Some researchers criticize the literature on institutional entrepreneurship for downplaying the constitutive role of institutions for social action. The reason is what sociologists call the paradox of embedded agency (Battilana et al. 2009: 72): How can agents, whose beliefs and actions are strongly shaped by existing institutions, break with these very same institutions and innovate? The convenient answer is to say that they cannot and leave no room for entrepreneurship. The sociologist Fligstein (1997: 397) therefore argues that "to the degree that institutional theory […] has a theory of action, it treats shared meanings as constraints on action that limit and determine what is meaningful behavior." Hence, where do institutional entrepreneurs get the "interests that they value highly" (DiMaggio 1988: 14) from? From a sociological point of view, the answer to this question is far from trivial.
Is the paradox of embedded agency simply a function of sociologists' too broad definition of institutions? Although Parsons (2007: 69) claims that most definitions of institutions seem compatible, there is in fact considerable disagreement. According to the sociologist Scott (2014: 56), "institutions comprise regulative, normative, and cultural-cognitive elements that, together with associated activities and resources, provide stability and meaning to social life." The controversial part concerns the cultural-cognitive elements that refer to shared cognitive frameworks and taken-forgranted rules, which inform agents about how the world works and what is in the agents' interest (i.e. the constitutive role of institutions for social action). Due to the sociologists' focus on these cognitive-cultural elements, Parsons (2007: 68, 100) ultimately classifies sociological institutionalism as an ideationaland not institutional approach, because institutional explanations should emphasize "objective rationality." Traditional historical-institutionalist definitions of institutions sidestep the issue of cognition (although without explicitly ruling it out). For instance, Hall (1986: 19) defines institutions as "the formal rules, compliance procedures, and standard operating practices that structure the relationship between individuals in various units of the polity and economy." Yet how do agents know how to comply and what the standard is? Rules are always abstract, while practice is concrete (Jackson 2010). Hence, as constructivist institutionalists have long emphasized, there is always an element of interpretation involved (Blyth 2003). Most historical institutionalists recognize that rules are never self-evident and in need of interpretation (Pierson 2000: 260;Streeck and Thelen 2005: 14). But if interpretation is unavoidable, should this cognitive process how to comply with an abstract rule in a concrete contextnot be part of our definition of institutions?
Responding to this challenge, Hall (2016: 35-36) argues that shared cognitive frameworks defined as "sets of ideas with implications for action" are a constitutive element of "structures." Another constitutive element of social structures are institutions, which he defines as "regularized practices with a rule-like quality in the sense that the actors expect the practices to be observed" (Hall and Thelen 2009: 9). He then adds that although these elements are "conceptually separate, […] their social force often derives from how they operate in tandem" (Hall 2016: 36). Yet this definition does not convince. Institutions are relegated to practices (thereby removing any need for interpretation because institutions are now by definition concrete), while an unproductive distinction is introduced between structure and institutions. 3 If we accept that rules are abstract, while practice is concrete, we must ask how agents know how to follow rules in concrete situations. By pointing to shared cognitive frameworks, Hall (2016) gives the right answer. Yet he does not dare to say what seems to be the logical conclusion: namely that we should not restrict our focus to the regulative aspect of institutional rules at the expense of their constitutive role because you cannot have one without the other. Hence, whether we follow Hall (2016: 36), who keeps institutions and shared cognitive frameworks conceptually separate but argues that "they operate in tandem," or rather Scott (2014), who considers shared cognitive frameworks an element of institutions, is ultimately inconsequential. Rather, the key point is that "once it is agreed that actors' interests are institutionally constructed, and that rational action reflects socially constituted preferences and cognitions, then the processes through which different groups come to have particular beliefs and concepts of the world, their roles in it and their interests become central to any adequate account of socio-economic and institutional change" (Whitley 2007: 549-550).
In fact, some historical institutionalists have pointed to institutions' constitutive role for social action from the beginning. For instance, in their classic contribution, Thelen and Steinmo (1992: 9) observe that "one, perhaps the, core difference between rational choice institutionalism and historical institutionalism lies in the question of preference formation." Historical institutionalism has thus always emphasized "the endogenous (institutional) origins of preferences by offering a more structural rendering of the world than rational choice, one in which institutions and organizations, not individual-level traits were the primary building blocks in accounting for political preferences and outcomes" (Fioretos et al. 2016: 7).
At the same time, historical institutionalism leaves more room to agency than (traditional) sociological institutionalism. Thelen and Steinmo (1992: 3) note that "what is implicit but crucial in this and most other conceptions of historical institutionalism is that institutions constrain and refract politics, but they are never the sole 'cause' of outcomes." Historical institutionalists have thus recognized that interest and agency must have a role in institutional analysis. However, historical institutionalism has never offered a coherent logic of action. Hence, while (traditional conceptions of) rational choice and sociological institutionalism suffer from undersocialized and oversocialized views of action respectively (Wrong 1961;Granovetter 1985), historical institutionalism might be in the position to strike a balance between the two views. Yet, being torn between two extremes, historical institutionalism never developed its own microfoundation.
Historical institutionalism has the potential to allow agency without necessarily resorting to behavioralism. Yet if historical institutionalism accepts the constitutive role of institutions for social action, it too faces the paradox of embedded agency, because institutions shape what agents believe to be in their interest. We must therefore ask which actors are likely to take a calculating position with regard to taken-forgranted rules. Based on a review of the (often sociological) literature, this paper identifies four factors that increase the likelihood of agency.
First, the literature emphasizes the situation of the institutional field (Fligstein and McAdam 2012;Capoccia 2016). Less institutionalized fields are more open to change because conventions and routines have not yet acquired the status as taken-for-granted rules. This point is related to the key role of uncertainty and institutions' role in reducing it. In emergent and not fully institutionalized fields, the level of uncertainty is higher, which leaves more room for agency (Capoccia 2015). In contrast, with the incremental institutionalization of cultural categories, or path dependence in the case of Pierson (2000: 260), the degree of embeddedness increases (Blyth 2001;Rixen and Viola 2015), thus reducing agents' ability to adopt calculating positions vis-à-vis takenfor-granted rules.
Second, the potential for agency is likely to vary with the position in the institutional field. A peripheral position in the field or a position at the intersection of two fields allows agents to develop different perspectives on what is taken for granted or on what is in their interest (Sewell 1992;Battilana et al. 2009). This is because such actors are less exposed to the field's cultural categories or exposed to possibly conflicting ones. However, it should be noted that such actors might be less likely to be successful because their peripheral position in the field is often reflective of their overall lack of influence. As they move into the center of the institutional field (and become influential), these actors are likely to adopt the field's routines and conventionsotherwise they would not be able to progress into the field's center.
Third, agency may result from the interaction with other fields. Changes in adjacent fields can diffuse into settled institutional fields, thereby breaking up cognitive locks and normative certainties (Blyth 2001). Settled institutional fields thus become temporarily unsettled, which creates room for agency. In addition, diffusion processes imply a translation effort, which forces actors to engage in some creative combination of new externally given elements received through diffusion and old locally given ones inherited from the past (Campbell 2004: 80). Clearly, this driver of institutional change parallels the historical institutionalist literature on (exogenous) critical junctures (Capoccia and Kelemen 2007). However, while Capoccia and Kelemen (2007) mainly emphasize the increased causal role of agency during critical junctures, the argument can be developed one step further. Critical junctures not only increase the causal role of agency, they may also reduce the degree of social embeddedness, which allows agents to adopt a calculating position with regard to taken-for-granted rules.
Finally, historical institutional accounts argue that institutions are the political legacies of historical battles and thus reflect foundational conflicts (Mahoney and Thelen 2010). Once created (or changed), institutions turn into "weapons of coercion and redistribution" (Moe 1990: 213). This foundational assumption has far-reaching implications, which are also increasingly recognized in sociological institutionalism (Fligstein and McAdam 2012). If institutions are the result of political conflict, then there are winners and losersand the losers do not suddenly disappear (Moe 1990). In addition, the resulting institutions reflect the interests (however defined) of the winners (incumbents), while the losers (challengers) will oppose the institutions, although they might not be in the position to ignore them (Hacker et al. 2015). Put differently, in case of such contested institutions (which, historical institutionalists would argue, are quite common), there are always (more or less organized) actors with some resources that wait for an opportunity to realize interests that they value highly. However, as the peripheral actors mentioned above, the problem of these institutional challengers is to find a way to outmaneuver the institutional incumbents (who tend to be structurally privileged).
While all four factors increase the likelihood of agency, they differ in important ways. Some emphasize foundational conflicts that introduce agency. In these cases, challengers of a given social order are rather easy to identify. Others highlight how in less institutionalized fields (including during critical junctures), when rules are up for grabs, there is a heightened role for agency. Both of these perspectives are well established in the historical institutionalist literature. Yet, there are also more subtle ways of injecting agency, which historical institutionalism has tended to overlook. Most notably, actors in peripheral positions in the field or at the intersection of different fields are prime candidates to take a calculating position with regard to taken-forgranted rules even in highly institutionalized fields. Here, historical institutionalism could learn from the literature on social movements (McAdam 1996), which has examined how taken-for-granted rules are being challengedinitially by small and peripheral groups, which however grow in influence over time and force the incumbents to justify and defend the institutional status quo (e.g. the feminist and civil rights movements).
These considerations have a further implication that has been so far largely overlooked in the literature. While there is a (natural) tendency in the literature to focus attention on incumbents and challengers, there are in fact further actors populating institutional fields. DiMaggio (1988: 15) has described these actors as 'external' constituencies (as opposed to 'core' constituencies, which include both incumbents and challengers). These external constituencies have no skin in the game. They are primarily interested in the certainty that institutions are supposed to provide (Beckert 1999). However, these actors' inability to develop agency on their own does not rule out the possibility that they can be mobilizedin particular by institutional challengers. This point is further developed below.
In sum, torn between rational choice and sociological institutionalism, historical institutionalism has never developed its own micro-foundation. However, this section has developed several portable propositions that allow for the identification of situations in which agency is more likely. Yet, what ultimately forms the agents' purpose is left to empirical research.

The paradox of plasticity
While historical institutionalist scholars have paid little attention to the paradox of embedded agency, they have recently begun to highlight another problem, which is more or less the flipside of the paradox of embedded agency: the so-called paradox of plasticity (Hall 2016: 39). In simple terms, the paradox raises the following conundrum. The more attention scholars devote to the factors shaping institutions, in particular agency, the more they call into the question the power of institutions to shape politics. Hence, while sociological institutionalists are concerned that all-powerful institutions leave no room to agency (paradox of embedded agency), historical institutionalists have begun to worry that too much agency may make institutions seem too fluid (paradox of plasticity).
At first sight, it seems odd that institutionalist scholarship could be concerned with this problem. Although the exact mechanisms differ (e.g. Nash equilibrium, path dependency, isomorphism), all three classic institutionalisms emphasize the stability of institutions. Hence, Hall and Taylor (1996) spend much of their influential review article discussing the problem that institutionalist approaches are insufficiently able to explain institutional change. In general, they argue, institutional change is only likely during critical junctures, typically caused by exogenous shocks. Otherwise, the institutional constraints are too strong to allow for change.
In recent years, however, there has been a turn in institutionalist scholarship towards more incremental, informal, and endogenous change. While this turn can be observed in all institutionalisms (Greif and Laitin 2004;Lawrence and Suddaby 2006), it has been most pronounced in historical institutionalism (Streeck and Thelen 2005;Mahoney and Thelen 2010;Hacker et al. 2015). Quite generally, this development is a function of two empirical observations. First, there is considerably more institutional change than the critical juncture framework suggests. The literature has convincingly demonstrated that institutions are often permanently contested and change gradually over time. Second, although these gradual changes often appear to be of minor importance, they can lead to substantial change in the end. Hence, an exclusive focus on critical junctures risks missing a large part of institutional change (Streeck and Thelen 2005).
Yet if we accept that institutions are permanently contested and constantly change, we are possibly facing the paradox of plasticity, because institutions, at the most extreme, may not have causal efficacy on their own (Riker 1980;Przeworski 2004;Hall 2016). Rather, as conditions shape institutions, the institutions' role might be restricted to (automatically) translating the causal effects of these antecedent conditions. Therefore, any institutionalist argument, according to Capoccia (2016Capoccia ( : 1115, is "ultimately premised on the gap between social change and institutional change" because although "actors and context change continuously, the constant stream of social change does not necessarily, automatically, and fully translate into institutional change." The task of institutional analysis is to explain why this is the case. What explains the stability of institutions, which prevents the automatic translation of social change into institutional change? Consistent with their ontological assumptions about groups, conflict, and power, historical institutionalist scholars have increasingly converged on the key role of social coalitions in stabilizing institutions (Hall and Thelen 2009). Interestingly, this focus on social coalitions supporting institutions parallels recent developments in sociology, which increasingly emphasize the role of agency in institutional work (Lawrence and Suddaby 2006). The remainder of this paper argues that a focus on the micro-foundations of coalition building is a particularly productive way to inject agency into historical institutionalism.
The coalitional perspective stresses that institutions are both supported and challenged by social coalitions. 4 While a coalition of incumbents actively maintains and defends the institution, other social coalitions might oppose and challenge it. In this perspective, institutional stability is a function of the activities of these social coalitions. 5 Regarding the incumbents, the literature discusses several strategies incumbents can use to prevent institutional change (Fligstein 1997;Capoccia 2016). For instance, incumbents can use their agenda setting power to delay and defer the attacks of challengers to (from their point of view) more convenient points in time. Numerous other strategies are conceivable. Regarding the challengers, starting from a minority position in the field, their main task is to build a coalition strong enough to displace the incumbent one. However, as long as the challengers are not able to build such a coalition and as long as the incumbent coalition remains in place, institutions persist despite social change.
How does the coalitional perspective resolve the paradox of plasticity? The reason is that it is difficult to build a coalition sufficiently strong to enforce institutional change. As Hall (2016: 40) observes, "discontent with existing institutions has to reach certain levels" because "actors have to be convinced they should abandon procedures with which they are familiar to enter uncertain territory." As a result, "they have to develop new interpretations of their circumstances, agree that specific types of reforms are likely to address their problems, find ways to proceeding collectively, and assemble the relevant power resources. Sometimes, they have to be persuaded to enter costly contests for power" (Hall 2016: 40).
Hence, while the coalitional perspective allows for permanent contestation and adopts a considerably more dynamic and political view of institutional stability, creation, and change, it emphasizes stability over change because the creation of sufficiently strong challenger coalitions is difficult. The reason for this is that new coalitions must first disrupt existing cognitive frameworks and overcome collective action problems. Given these challenges, the coalitional perspective avoids "the theoretical abyss of agency unleashed" (Thelen 2010: 58). Yet, at the same time, the coalitional perspective also stays clear of the theoretical abyss of all-powerful structures because it leaves room to agencyalbeit within institutional constraints. While the building of challenger coalitions is difficult, it is possible. Moreover, while incumbent coalitions are privileged, they can still weaken over time.
Yet, there is one striking omission in this exciting literature on the coalitional foundations of institutional stability and change. While it emphasizes the role of coalitions, it never explicitly addresses the question of who is actually building or maintaining them. Put differently, if new interpretations have to be provided, actors have to be convinced, and ways of proceeding collectively have to be found, who is doing all this work? 6 The answer to this question seems clear: actors who develop agency. While the section on the paradox of embedded agency has discussed the conditions under which actors are more likely to develop agency, the next section turns to the question of how actors can build social coalitions.

The problem of collective action
Collective action problems denote situations in which agents would benefit from a certain action, but associated costs make it implausible that any agent will undertake it. Three sets of factors complicate collective action such as coalition building (Wijen andAnsari 2007: 1080). Agents may want to free ride on other agents' contributions, action may be held up because agents are waiting for others to take the lead, and, finally, agents may abstain from engaging in collective action because they feel their contribution to the problem is insignificant. If we add to this list the fact that there are incumbents who have an interest in actively undermining the challengers' coalitional work (Capoccia 2016), it becomes clear that the obstacles to collective action are formidable.
Nevertheless, the problem of collective action is the key to reconciling the paradox of embedded agency with the paradox of plasticity because this coalitional perspective allows for a robust role for agency in institutionalist approaches. As discussed, agency is more likely to develop if (1) institutional fields are not fully settled, (2) actors occupy a peripheral position in the field or a position at the intersection of two fields (exposing them to alternative perspectives), (3) changes in adjacent fields diffuse into the settled institutional field (exogenous shocks), or (4) institutions reflect a foundational conflict that still reverberates.
However, the existence of such agency does not automatically translate into institutional change. Rather, institutional change is still the improbable outcome because other coalitions stabilize institutions (Hall 2016). Coalitional work, understood as activities to create and maintain social coalitions, stabilizes institutions through two different mechanisms. First, because of collective action problems, the creation of a sufficiently powerful challenger coalition is difficult, which makes institutional change unlikely (Wijen and Ansari 2007). Second, incumbent coalitions may actively defend institutions against attacks by emerging challenger coalitions (Capoccia 2016). Given the incumbents privileged structural position, they are likely to succeed. Yet, institutional change is possible. What is more, the coalitional perspective in combination with a focus on agency makes it possible to generate portable propositions that allow for the identification of broader patterns of political dynamics.
Some actors work to create and maintain social coalitions. What factors increase the likelihood that they succeed? This paper distinguishes between three sets of factors: resources, skills, and institutional opportunities.
First, the availability of material and social resources facilitates coalitional work (DiMaggio 1988). Financial and organizational resources allow actors to sustain mobilization efforts, organize an information campaign, or buy professional support. For instance, Sheingate (2003) discusses how U.S. Presidents, due to their position of direct authority, have the potential to be powerful actors. However, resources need not be material. The position in actor networks is of course also an important resource. Culpepper (2005) demonstrates how network centrality provides actors with the possibility to induce a collective belief shift among the other network members because the central actor's decision to change its beliefs suggests to the other network members that their old cognitive maps may be wrong (see also Rinscheid et al. 2019). However, exclusively resource-based entrepreneurship is rare because it is unlikely that institutional challengers achieve such positions of authority. Given their structural privileges, why would the institutional incumbents tolerate a U.S. President from the challenger coalition? Hence, resource-based explanations must typically be combined with other factors such as skills and opportunities.
Second, actors must be skilled. Fligstein and McAdam (2012) define such social skills as the ability to induce cooperation in others. More precisely, social skills concern "the way in which individuals or collective actors possess a highly developed cognitive capacity for reading people and environments, framing lines of action, and mobilizing people in the service of broader conceptions of the world and of themselves" (Fligstein and McAdam 2012: 17). Actors use these social skills to build challenger coalitions (Béland and Cox 2016). In this process, they are looking at other already existing institutional challengers (to join forces), but also at the aforementioned external constituencies. While these external constituencies have an interest in certainty and predictability, they are neither part of the incumbent nor the challenger coalitions (DiMaggio 1988: 15). Hence, these actors are not strongly wedded to concrete institutions. However, the absence of agency on the part of the external constituencies (as opposed to incumbents and challengers) does not rule out the possibility that challengers can mobilize them. Put differently, like in elections, those who win the independents, typically win the vote.
Actors must therefore engage in skilled action. What tactics are available to them? Broadly speaking, the literature offers two sets of tactics. On the one hand, researchers have emphasized tactics that strengthen the incentives of calculating actors to engage in collective action (Ostrom 1998;Wijen and Ansari, 2007). Such tactics include attempts to gain control over agenda setting, link different policy dimensions to build coalitions, or improve cooperation by means of repeated interaction. On the other hand, researchers have highlighted "the distinctive human capacity and need to fashion shared meanings and identities to ensure a viable existential ground for existence" (Fligstein and McAdam 2012: 18; see also Lawrence and Suddaby 2006;Béland and Cox 2016). These tactics primarily concern attempts to frame situations as constituting a significant new threat to or opportunity for the realization of intersubjectively desirable goals. Hence, skilled actors are able to fashion meanings and identities that are relevant to others. This is not the place to discuss all available tactics. Fligstein (1997) alone provides a non-exhaustive list of fifteen tactics (see also Vis and van Kersbergen 2007). Rather, it suffices to note that actors, once they have developed agency, have a large number of tactics at their disposal, which they can use to create and maintain coalitions. Which tactics actors' use and which ones are considered most appropriate or effective is of course largely influenced by these actors' concrete situation. Yet, the relative importance of skilled action certainly increases, as differences in resources become smaller.
There is clearly an affinity between historical institutionalism and interest-based tactics, on the one hand, and sociological institutionalism and identity-based tactics, on the other. Yet as emphasized by Smith (2004: 304), even though the relationship between identities and interests is complex, there is reason to think that it is reciprocal, as identities are likely to shape interests and vice versa. Moreover, coalitions are probably most successful if they have coalitional aspects to them, which are based on interests but also provide collective identities to mobilize action. Institutionalized social movements such as feminism and civil rights are certainly cases in point. 7 Finally, coalition building is also a function of institutional opportunities, which may be defined as the likelihood that an institution permits actors to mobilize, based on available resources and by means of skilled action, a sufficiently strong coalition to create, change, or stabilize institutions (McAdam 1996;Meyer 2004). The literature suggests five factors that influence institutional opportunities. As will become clear, these factors partly overlap with the factors that increase the likelihood of developing agency. This should not come as a surprise, however, because the factors that enable actors to develop a sense of purpose may also influence others in a given institutional field. While the latter may not necessarily develop agency on their own, they might become more open to other actors' mobilization efforts.
Hence, institutional opportunities are conducive to coalition building when the degree of institutionalization of a field is low or there is a multiplicity of interpenetrating institutional structures. In both cases, actors' ability to adopt a calculating position vis-à-vis taken-for-granted rules is increased because the degree of embeddedness is relatively low (Blyth 2001;Capoccia 2016) or because possible tensions between interpenetrating institutional structures lead to ambiguities or even contradictions that actors can try to expose (Sewell 1992;Beckert 1999). Third, exogenous shocks, often triggered by changes in adjacent fields, can temporarily unsettle an institutional field, thereby creating a critical juncture during which the actors' capacity to mobilize (challenger) coalitions is substantially strengthened (Capoccia and Kelemen 2007). Put differently, these institutional opportunities increase the relative importance of resources and skilled action.
Two further factors can be derived from the literature. The endogenous weakening of incumbent coalitions can also provide institutional opportunities. This possibility is a logical implication of previous arguments, which have suggested that the creation, change, and stability in institutions are the result of skilled institutional work. If stability (also) requires skilled action by incumbents, there is always the possibility that these actors show little skill. For instance, the literature on institutional exhaustion shows how negative (rather than positive) feedback effects may lead to situations in which the normal functioning of institutions undermines the institution's external preconditions (Greif and Laitin 2004;Streeck and Thelen 2005). In these situations, institutional incumbents have shown little skill in maintaining an institution.
Finally, the social movement literature, in particular, emphasizes how the institutions governing political decision-making may create opportunities for both challenger groups and incumbent groups (McAdam 1996). These considerations imply thatdue to the multilayered structure of institutional fieldssome institutions continue to structure a field even in times of institutional change. Hence, even in times of great political change, the rules of democratic politics may remain in place. However, not all polities feature the same rules, which may make some polities more accessible to challenger groups. For instance, Kitschelt (1986) demonstrates how the pre-established channels and opportunities that political regimes provide shape the possibilities offered to institutional challengers to disseminate their message and disrupt established policies. Of course, these political 'rules of the game' may also benefit institutional incumbents. For instance, Capoccia (2016) shows how institutional incumbents can use agenda setting power to undermine challengers' efforts to induce institutional change. 8

Conclusion
This paper's aim has been to conceptualize the role of agency in historical institutionalism. Based on recent contributions following the coalitional turn in historical institutionalism and by drawing on recent scholarship in sociological institutionalism, it has argued that a focus on the micro-foundations of coalition building offers a productive way to generate portable propositions about agency. In addition, this focus can resolve the paradox of the two paradoxes, i.e. the fact that while sociological institutionalists are concerned about the paradox of embedded agency, historical institutionalists are increasingly concerned about its flipside, the paradox of plasticity. This paper has suggested that actors are more likely to develop agency when institutional fields are not fully institutionalized, actors occupy a peripheral position in the field or a position at the intersection of two fields, changes in adjacent fields diffuse into the settled institutional field (i.e. exogenous critical junctures), or institutions reflect a foundational conflict that still reverberates. However, solving the paradox of embedded agency in this way necessarily leads to the next problem: If institutions are constantly attacked and regularly change, what explains their stability? Following Hall and Thelen (2009), this paper argues that social coalitions stabilize institutions. They do so in two ways. On the one hand, the incumbent coalition may actively defend the institution against attacks from challenger coalitions. On the other hand, because of collective action problems, the creation of sufficiently strong challenger coalitions is difficult. The coalitional perspective thus explains why social change does not automatically translate into institutional change.
Yet while this coalitional perspective allows for a considerably more dynamic and political view of institutional stability, creation, and change, it lacks a micro-foundation and does (too) little to address the unclear role of agency in historical institutionalism. This paper has suggested that a robust role for agency can be found in the activities of creating and maintaining these social coalitions (coalitional work). The literature on collective action problems leaves no doubt that coalition building is difficult. Hence, it requires resources and skill. In addition, actors engaged in coalitional work need to recognize institutional opportunities. Nevertheless, without agency, no coalition can be created or maintained. Coalitions do not come out of nowhere. Their origins and persistence must be part of the explanation too, and for this, we need actors.
The actor-centered account introduced in this article explains why institutional change often occurs in the form of an exogenous critical juncture. During critical junctures, weakened institutional constraints heighten the probability that agents' choices will affect the outcome of interest (Capoccia and Kelemen 2007). However, during critical junctures, there is also a reduced degree of social embeddedness, which allows actors to develop agency, understood as the motivation and the creativity that drive actors to break away from structural constraints. More importantly, however, this actor-centered historical institutionalism also allows for endogenous and gradual institutional change (Streeck and Thelen 2005), because it emphasizes the microfoundations of coalition building and maintenance. Arguments emphasizing endogenous and gradual institutional change typically refer to competing social coalitions, but they often struggle to explain why some coalition prevails over the other. Historical institutionalist scholars therefore often face the criticism of describing rather than analyzing institutional change (Schmidt 2011). The more actor-centered historical institutionalist approach presented in these pages allows researchers to address this criticism.
Admittedly, there is a rather clear downside to more actor-centered historical institutionalist approaches. Such approaches are rather challenging in terms of data requirements, which is why standard historical institutionalist approaches might be preferable if they offer sufficiently convincing explanations of the phenomena of interest. In the best of all worlds, actor-centered empirical analyses should be able to demonstrate coalition-building activities on the ground (Seitzl and Emmenegger 2019). However, the empirical evidence needed to demonstrate the actors' coalitional work is not always readily available. In addition, these data requirements make comparative-historical analyses considerably harder. There is thus a clear trade-off between adding more cases and injecting agency.
To develop propositions about agency in historical institutionalism, this paper has also drawn on recent contributions in sociological institutionalism. Conceptual innovations in historical institutionalism have often been developed in dialogue with rational choice institutionalism (Hall and Thelen 2009;Hall 2010;Mahoney and Thelen 2010). This paper has followed this tradition but turned its attention to sociological institutionalism, which, paradoxically, has advanced the furthest with regard to conceptualizing agency. Originally, sociological institutionalism left little room for agency (Jepperson 1991). However, in recent years, it has increasingly turned to understanding how agents wittingly change the institutions in which they are embedded. In particular, my efforts parallel the ones by Fligstein and McAdam (2012), who have offered a powerful new account of the role of agency in sociological institutionalism. In this process, they have also drawn on insights from historical institutionalism, such as the focus on the conflict and distributional consequences, which demonstrates the benefits of a dialogue between the different institutionalisms.
However, in spite of this cross-fertilization, important differences between the institutionalisms remain. Most notably, sociological institutionalism is more cultural constructionist, whereas historical institutionalism is more realist, which implies that the former sees rules as more or less institutionalized, whereas the latter considers rules to be more solid. 9 Although this is more a matter of emphasis than definitional fiat, it has consequences for the comparative advantages of the two institutionalisms. Historical institutionalism lends itself to more parsimonious accounts, which allows researchers to cover longer periods and facilitates comparison. In contrast, sociological institutionalism is more attentive to subtle forms of change, as it emphasizes how the meanings of institutional fields are being adapted. Moreover, while historical institutionalism tends to assume the existence of conflicts and pays more attention to distributional consequences (which might also be the result of its strong roots in political science), sociological institutionalism seems more attuned to explain where challenger coalitions come from in the first place.
Several theoretical challenges remain. For instance, future research may be able to relate different modes of gradual institutional change identified in recent historical institutionalist scholarship to the general propositions about agency offered here. For instance, it may be that gradual institutional change by means of conversion is particularly likely in case of weakly institutionalized fields, while institutional change by means of layering presupposes considerable resources by institutional challengers. In addition, it remains an important task for future research to develop expectations about when gradual institutional change is more likely than transformative change and how these different forms of institutional change might relate to the general propositions about agency discussed above.