The Programmatic Action Framework (PAF) is a theoretical lens on policy processes developed at the intersection of policy process research, public administration, elite sociology, and social psychology. This chapter is particularly devoted to outlining the foundations of the PAF and putting them in context with other existing theories of the policy process. There are two main bases of PAF assumptions: Firstly, the role of bureaucracy in areas close to the state in formulating policy and the related desire for increased authority gained through advancement in individual careers. Secondly, social psychological perspectives on social identities of groups formed on the basis of shared characteristics are adopted by the PAF to outline the role of shared biographies and resulting policy programs, which are identity-forming, in policy processes and policy change. The particular focus of this study is on the institutional conditions under which such actors form programmatic groups and use their policy programs to shape the policy process over time.

In policy process research, the perspective on shared biographies and the associated promotion of policies has only marginally found its way into recent research. To be sure, there are theoretical perspectives that shed light on the question of why policy actors cooperate in policy processes. Rational choice theorists argue that collective action is a result of strategic consideration, which individual behavior generates—or promises to generate—the maximum benefit for a policy actor. In the original understanding, this presupposes perfect information and knowledge about the interests of all actors involved. The “homo economicus”, who has complete knowledge of his environment, a stable set of preferences, and the ability to weigh among available alternatives to choose the course of action that best suits his preferences, has received much criticism for being too abstract to represent the actual behavior of intendedly rational individuals in particular circumstances. Indeed, this presents a debate as to whether normative or positive claims should be inherent in theories (March, 1978).

Following this critique, the scholarly literature in economics and public policy soon gathered under the common label of bounded rationality (B. D. Jones, 1999; Simon, 1955). Essential in this debate was the work of Herbert A. Simon, who received the Nobel Prize in 1978 for his work on the behavior of actors in organizations (Simon, 1947, 1955, 1978, 1985, 1990). A behavioral model of rationality, they argued, that models real behavior is based on bounded rationality. Rationality is bounded when there are constraints on the completeness of information, the definition of a problem, or the cognitive abilities and/or resources of individual agents to choose among alternative solutions to the problem, distinguishing between different roots of boundedness (Forester, 1984). The model takes into account imperfect information and uncertainty when making rational decisions, while at the same time granting policy actors the ability to learn from mistakes made (i.e., irrational decisions) (March, 1978; Simon, 1955).

In contemporary policy process research, bounded rationality is a frequently received psychological foundation for the behavior of policy actors. Scholars even speak of a behavioral turn in policy sciences (Leong & Howlett, 2020). Challenging the assumption of individual rationality, the Advocacy Coalition Framework (ACF) draws on the psychological concept of belief systems to explain collective action and the formation of advocacy coalitions through shared policy core beliefs (Calanni et al., 2014; Jenkins-Smith et al., 2017; Sabatier & Brasher, 1993). Building on this, the Narrative Policy Framework (NPF), developed only earlier this century, approaches collective action through the use of coherent policy narratives (Gupta et al., 2018; Shanahan et al., 2017). Other policy process frameworks focus on situational occasions exploited by policy entrepreneurs (Multiple Streams Framework, MSF) (Herweg et al., 2017) or on changing attention to policy issues that lead to major policy change (Baumgartner et al., 2017). None of these approaches place an explicit focus on the role of biographies or social identities in explaining collective action, which is what the PAF contributes to explaining policy change and stability. The subsequent chapters will delineate the PAF and embed it in the current state of research according to three main lines: cognitive foundations of policy actors’ motivations, drivers of collective action, and mechanisms leading to policy change.

2.1 Social Groups, Biographies, and Policy Programs

In France, the role of a dominant and powerful public bureaucracy (Rouban, 1989, p. 45; Vernardakis, 2013) and a strong executive due to the semi-presidential system (Keeler, 1993) led to the emergence of a theoretical perspective called the programmatic approach (Genieys & Hassenteufel, 2012). This seeks the roots of policy change in the cooperation of actors close to the state who have similar career trajectories and promote a common policy program over several years. They do this to advance their careers, increase their authority, and realize their policy ideas.

Following these observations, the PAF postulates the argument that individual actors close to the state administration form programmatic groups based on shared biographical experiences. These actors include—among others—bureaucrats and actors from self-governance. Programmatic groups function as social groups and are committed to promoting a joint policy program. A programmatic group encompasses programmatic actors who are hypothesized to coalesce on the basis of shared biographical identities and have direct access to the state’s policy-making apparatus. They are called programmatic actors because they subscribe to a shared policy program that is nameable and on which the programmatic actors agree. This can either be explicit, such as the creation of a document or the name of a group that meets regularly under the same label. Or the program may be implicit in the sense that asking each programmatic actor about their program would lead to highly overlapping results in the formulation. A policy program is defined as a set of policy goals and policy instruments that follow a consensus view of problems and solutions and focus on a particular policy sector. The individual policy reforms represent steps toward implementing the programmatic vision, but they are linked by the common policy goals articulated in the policy program. The immediate first incentive to join a programmatic group is strategic, as programmatic actors associate it with the prospect of advancing their own careers. Following on from this, there is a second, normative, and psychologically based impulse that drives individual policy actors to engage in collective action: a shared view of problems and solutions in a given policy sector, rooted in biographical association with the other programmatic actors and the resulting social identity.

Drawing on elite sociology, it is the career paths and biographies that led a group of scholars around Genieys, Hassenteufel, and Smyrl to formulate the programmatic approach as an alternative perspective on policy processes (Genieys & Hassenteufel, 2001; Genieys & Smyrl, 2008c). The insight that decision-making processes are shaped by a multitude of elites acting in the policy process is not limited to France, but also eminent in German elite research (von Beyme, 2001). Based on the analysis of French policy processes, the “approche programmatique” (French for programmatic approach) finds an endogenous explanation for policy change in the trajectories of policy elites and their policy programs (Genieys & Hassenteufel, 2012, p. 3). Endogenous in that they see policy change as the result of ongoing competition among elite actors and career-seeking individuals within the state (Genieys & Smyrl, 2008c, p. 90). In particular, the homogeneity of trajectories was seen as a key driver for collaboration and the resulting influence on decision-making processes. To specify this, the publication by Genieys and Smyrl (2008b) contains both theoretical assumptions and mechanisms as well as the first empirical case studies that demonstrate the existence of what the authors call programmatic elites. These programmatic elites are described as “a group of actors with direct access to policy-making positions that is self-consciously structured around a common commitment to a concrete and coherent programmatic model for a given policy sector” (Genieys & Smyrl, 2008c, p. 76).

Regarding the underlying assumptions and motives of individuals on which the theoretical mechanisms of programmatic elite competition with resulting policy change are based, Genieys and Smyrl (2008a) discuss these predispositions intensively and place them in the broader realm of bureaucratic theory, micro-level motivations for the pursuit of policy change, and the types of actors that a programmatic elite can encompass. In it, the pursuit of legitimate authority (p. 29) is seen as driving some—but by no means all—policy actors with direct access to the decision-making process, that is, those who directly formulate laws and regulations or have direct access to those who do. Their preferences are described as flexible to some degree (p. 30), and their rationality is assumed only insofar as it involves the pursuit of legitimate authority (p. 44).

While the policy program and the programmatic elite are mutually reinforcing, it is unclear whether the program precedes the elite group that gathers around it (Genieys & Michel, 2005, p. 187) or whether the group exists before its program is developed. One can bluntly argue that this is a question of empirics rather than theory, as theoretical considerations are primarily concerned with what drives successful programmatic action and under what conditions it occurs. Thus, it may be both that a programmatic group forms around an existing policy idea or that it emerges before it formulates its policy program.

Indeed, looking at career trajectories and “life cycles” (Michel, 2008, p. 165) in the study of programmatic elites might lead us to consider related concepts of generational change not only in policy processes (Obinger, 2012), but also in terms of the values and attitudes that different generations exhibit (Fisher, 2020; Shaykhutdinov, 2019; Wu & Lin, 2019). Generational change, however, implies a compelling exchange of generations as a function of elapsed time, regardless of how a particular group’s positions change, how its policy program changes, or how networks are built and dissolved. Consequently, what is captured by a consideration of generations is not the focus of the programmatic approach.

Already in the early versions of the programmatic approach, there is a reference to programmatic elites as social groups, their sectoral meaning, and corresponding identity (Genieys, 2010, p. 14). In particular, the emphasis on programmatic actors coalescing into social groups and social identities driving the group dynamics of these actors to explain their collective action in the policy process has been the subject of ongoing reinterpretation and development of the PAF. This represents the only recent starting point for incorporating social identities in the policy process (Hornung, Bandelow & Vogeler, 2019).

In an attempt to integrate psychological and social psychological insights into the study of public administration and public policy, new strands of research have emerged under the labels of Behavioral Public Administration (Bhanot & Linos, 2019; Grimmelikhuijsen et al., 2017) and Behavioral Public Policy (Ewert, 2019; John & Stoker, 2019; Lodge, 2019; Strassheim, 2019). The concepts of bounded rationality have long been part of policy process research and, in particular, actor-centered theories of the policy process assume a model of the individual from which psychological assumptions allow the derivation of behavioral hypotheses. In contrast, social identity theories have found little entry into policy process research, but their disregard is considered an “enormous blind spot” (Béland, 2019, p. 29). Far better known in political science are the big five personality traits of extraversion, agreeableness, conscientiousness, emotional stability, and intellect/imagination (Goldberg, 1990; McCrae & Costa, 1987), which are increasingly referenced in research on political behavior (Ackermann, 2016; Duckitt & Sibley, 2016; Weinschenk, 2017). However, these are difficult to operationalize for policy actors relevant to policy processes. Questionnaires tend to be extensive, and even the brief measures of personality traits rely on 10-item measures (Gosling et al., 2003), which are unlikely to be fully answered by a sufficient number of policy actors. It should not be neglected here that the big five personality traits may be able to shed light on important behavioral observations of policy actors, for example, how high levels of extraversion and agreeableness make policy actors more likely to engage in policy entrepreneurship (Timmermans et al., 2014, pp. 5-6). However, addressing how policy actors work together and how collective action leads to policy change suggests a more social psychological integration of psychological insights in the form of social identity research.

Compared to existing psychological foundations for policy process theories, the Social Identity Approach (SIA) offers a complementary theoretical perspective on individual preferences and behavior. The SIA circumscribes the combination of the Social Identity Theory (SIT) and the Self-Categorization Theory (SCT). The former was developed largely by Tajfel (1974) and the latter was refined by his student Turner (1982) (Hornsey, 2008, p. 208). Their approaches are interrelated and both relate to the concept of social groups, albeit with slightly different emphases. Originally developed as a theory of intergroup relations, the SIT seeks to examine intergroup behavior to explain cooperation and conflict between groups, focusing on in-group and out-group behavior, and the behavior that individuals exhibit in intergroup relations. Here, Tajfel defines social identity as “that part of the individual’s self-concept which derives from their knowledge of their membership of a s social group (or groups) together with the value and emotional significance attached to that membership” (Tajfel, 1982, p. 3). The SCT is more concerned with the intragroup processes and cognitive aspects of identification and aims to uncover the underlying cognitive processes of individuals as they assign themselves and others to social categories and behave accordingly. The cognitive processes that lead individuals to identify with a social group are the parallel desires for sameness and distinction, summarized in a model of optimal distinctiveness (Brewer, 1991). A social identity consists of a triad of cognitive and affective components articulated as belongingness, positive evaluation, and emotional attachment (Tajfel, 1978, p. 372).

Although political science has adopted identity concepts from psychological research, not all of these identity concepts are related to the SIA (Weiner & Tatum, 2020), or the identity theory as part of the self-concept besides and apart from social identity (Hogg et al., 1995). Those that are related to the SIA use social identities as a refinement of already familiar concepts, such as partisanship (Greene, 1999, 2004). Conceptualizing partisanship as a social identity allows for examining the dynamics of partisan polarization (Iyengar & Westwood, 2015). At a micro level, Goren (2005) outlines how party group membership shapes core values and preferences, rather than party group membership being the result of preexisting core values and preferences. Party group members have been shown to be more attached to the party as a group than to what the party stands for, and to remain loyal to the party even when issues, platforms, and leaders change, which has been described as expressive rather than instrumental partisanship (Huddy et al., 2015, p. 15). This holds true in multiparty systems (Bankert et al., 2016). Nevertheless, this relation is still the subject of controversial findings and ongoing research (Egan, 2019)

Others relate the identity concept to the influence of peer networks and friends on preferences and behavior, for example when the salience of partisan identities depends on the party affiliation of friends (Parsons, 2015, p. 681). In these cases, social group membership is also associated with social pressures that lead individuals to follow group norms (Gerber et al., 2008). Other examples include when the online political behavior of individuals with high belongingness needs is influenced by whether their friends engage in online political activity (Bäck et al., 2020), or when personal friendships of actors in city governments determine the extent to which information is shared across those governments (Ki et al., 2020, p. 23).

In view of the need in policy process research to develop alternative models of the individual beyond rational choice (Millar et al., 2019, p. 114), Hornung et al. (2019) transfer the perspective of social identities to explain the behavior of policy actors in the policy process. A novelty here lies in the assumption that preferences and behavior are not antecedent but descendent of group memberships and that the thinking and behavior of group members converge over time and gain cohesion. Hornung et al. (2019) distinguish five macro-level identities that can be salient in policy processes and are attributed to the following different social groups in the political arena:

  • any type of policy-related organization, including political party groups (Bartle & Bellucci, 2009);

  • any type of group related to the region or locality to which the individuals belong. This refers to the conflict between subnational and national levels of policy-making (Hildebrandt & Trüdinger, 2020) but also to geographical affiliations of place (Devine-Wright & Howes, 2010) or supranational, more abstract identities such as European identities (Kuhn & Nicoli, 2020);

  • sectoral groups that emerge from sectoral professionalization and ongoing sectoral collaboration on a particular topical issue (Eriksson, 2017), and resulting policy styles (Padgett, 1990);

  • demographic and biographical identities that involve an individual’s early socialization through education and work experience, but also less formalized life experiences and biographical events;

  • informal groups in the policy process, consisting of an informal collaboration familiar from the US Congress (Stevens et al., 1981) and other parliaments (Osei & Malang, 2016; Steinert & Yordanova, 2016).

The definition of social groups thereby follows Turner’s as “two or more individuals who share a common social identification of themselves or, which is nearly the same thing, perceive themselves to be members of the same social category” (Turner, 1982, p. 15) and the important observation that “members of a social group seem often to share no more than a collective perception of their own social unity and yet this seems to be sufficient for them to act as a group” (Turner, 1982, p. 15). In considering these types of identities, the role of demographic identities in particular emerged as not being adequately addressed in current policy process research.

What need to be considered when applying the theoretical perspective of social identities to policy process research are the professional identities of policy actors. Aschhoff and Vogel (2019, p. 715) note that working in collaborative projects can establish a shift from professional identities to an identity created through collaboration. Similarly, agreements can create a shared identity by institutionalizing values that provide an anchor for identification (Duina, 2019). Given the multiplicity of existing social identities, their use as a valuable adjunct to explain behavior in the political sphere raises several key challenges, among the most important of which is defining the boundaries of political groups (Huddy, 2001, p. 145). In order for groups to be clearly delineated and for individuals to clearly identify the boundaries of the in-group versus the out-group, it is imperative that the group has identifying features such as a group name (Ren et al., 2012, p. 847). A second challenge relates to determining which identities actually matter in the policy process, that is, which are salient and strong, and what happens when the social identification with a group no longer leads to positive ingroup perceptions of distinction. The issue of identity salience is particularly relevant under different context-specific conditions (Vogeler, Hornung & Bandelow, 2020). Different ways of dealing with negative feelings of identity include not only leaving the group, but also changing the reference base for comparison with other groups and actively trying to change status hierarchies and intragroup dynamics (Hornsey, 2008, p. 207).

In short, the core argument of social identity in policy process research is that social identification with a group arises from the cooperation of actors who initially form a group and that individual preferences and behavior are guided by social identities. In the case of multiple identities, the salience of an identity depends on institutional circumstances as well as the strength of the social identity.

According to the flow diagram shown in Fig. 2.1, institutional opportunities and constraints influence the entire process of programmatic action. Programmatic action is defined as an instance in which programmatic groups successfully form and pursue their policy program to the ultimate outcome of policy change. This process is characterized by several steps that correspond to hypotheses about group formation and the success of programmatic groups and policy programs.

Fig. 2.1
A model diagram for the influence of institutional opportunities and constraints on the programmatic group. Policy actors via biographical intersections lead to a programmatic group that commits to policy programs that result in policy change.

The programmatic action framework. Source: Slightly modified on the basis of Bandelow, Hornung, and Smyrl (2021, p. 11)

Full size image

Briefly summarized, the model begins at the bottom with the entire field of policy actors in their “natural state”, that is, they are policy actors with different social identities (since each individual is member in multiple social groups) but do not (yet) have a programmatic identity, since there are no programmatic groups. These actors follow a duality of incentives in their behavior: a rational motive to advance their careers and authority (what to achieve), and an ideational motive to normatively shape policies (how to achieve it). Some of these policy actors already hold positions and resources of moderate importance because they began careers in one of the sectoral associations or were recruited through party-political channels. At some point, they may (but do not necessarily) coalesce into programmatic groups to further their goals. This is hypothesized to happen if there are homogeneous career paths, biographical linkages, and institutionalized or informal venues of exchange. Once there is a programmatic group, this social group forms a social identity of programmatic actors and at the same time commits itself to a policy program, which binds the group together and further strengthens social identification. Thus, a previous biographical identity then becomes a programmatic identity (no programmatic group exists without a program, and no program exists without programmatic actors). The program is thus an expression of the ideational positions of the programmatic actors and an instrument for achieving their goals. The success factors for the programmatic group and its policy program that ultimately help them achieve policy change lie in the programmatic group’s ability to spread its ideas, institutionalize them, form alliances, or use the breaking of existing alliances to its advantage.

In the step-by-step description of the mechanisms inherent in this theoretical model, it becomes clear that one element in the PAF is still under-theorized and its influence on programmatic action is not yet sufficiently differentiated. This is the case with institutional opportunities. As a result, although empirical applications of the PAF already exist, they rarely address the institutions that enable or hinder programmatic action. Instead, they are merely points of contact or indications of potentially relevant institutions, such as the particular importance of self-governance in German health policy (Hornung & Bandelow, 2020). Hassenteufel et al. (2010) provide a starting point by arguing that variations in institutional settings influence the types of programmatic actors one is likely to find. However, they do not focus on the similarity of institutional settings that enable programmatic action in the first place. Bringing these institutions together with the study of policy process research is an attempt that has been successfully made by other theoretical frameworks (Fischer, 2015; Lubell, 2003; Wenzelburger, 2015; Zohlnhöfer et al., 2016). This study intends to do the same for the PAF, asking about the very institutions in different states that enable or hinder programmatic action.

Policy actors refer to those actors who are directly involved in policy-making. Assuming that policy actors seek authority in competition with other policy actors, cooperate with other policy actors based on shared biographies, and normatively use a policy program to achieve their strategic goal, the social group resulting from this mechanism is referred to as a programmatic group. Programmatic groups are thus defined as social groups that form around a policy program and thus form a social identity. Their members join together on the basis of biographical ties, pursue a strategic and an ideational goal, the latter being a means to achieve the former (Saurugger, 2013), and commit to the group’s policy program to this end. When policy actors form programmatic groups, they become programmatic actors.

The concept of policy programs differs from existing conceptualizations of policy content in policy process research. A policy program is defined as not only arguments, measures, and instruments of policy change to achieve specific policy goals, but also a shared strategy in doing, so as well as a previously shared view of existing policy failures and problems (Hassenteufel & Genieys, 2021). Due to its visionary nature, programmatic change in policy sectors is often only visible after several years, when first steps of the program have been implemented and a coherent vision can be observed in the adopted reforms. In this respect, it is comparable to a so-called institutionalized reform that is implemented gradually and planned as a learning, adaptive system (Pannowitsch, 2009, p. 142). Nevertheless, observing innovative impulses in a policy sector that have the potential to shape it over several years and that are tied to a group of actors who publicly relate to the idea of the policy program and are biographically bound always allows for speculation about possible future policy programs.

Policy programs differ from policy paradigms, regimes, and alternatives (Hall, 1993; Howlett et al., 2006; Wilson, 2000) in two major respects. Firstly, a policy program is associated with a programmatic group. While policy paradigms, regimes, and alternatives may be promoted by collective actors such as collective policy entrepreneurs (Stephenson, 2010, p. 730), and policy entrepreneurs may also make a policy alternative to their pet policy, these connections are not hypothesized to be driven by social identities, nor are they based on biographical intersections between the actors. The PAF assumes that a programmatic group is a social group that creates its own social identity and uses as a defining element of the group the policy program to which it commits. Through social identification, individual members are guided in their behavior by the norms and values of the programmatic group, which, unlike in the ACF’s understanding, do not come from beliefs developed through socialization but are handed down by the social group.

Secondly, a policy program is an instrument in the power struggle between competing policy actors. It is not necessarily associated with a particular set of policy beliefs. However, actors tend to formulate programs that correspond to some of their values and attitudes. More important is the alignment of the policy program with the strategic goal of the social group, which is to gain authority within the sector in question. Policy programs are thus used by programmatic groups to gain power positions in the sector. The development, maintenance, or—in the case of new members joining the programmatic group—takeover of a policy program follows strategic individual interests. Competing policy programs may appeal to similar beliefs and values. Once a programmatic group establishes dominance of its policy program, it defends it not out of conviction or ideological beliefs, but because of inherent strategic interest, since the programmatic group’s authority rests on the program’s dominance. Individual members of the programmatic group have an interest in the stability of the program because it legitimizes the institutions that have emerged in the wake of the program and guarantees their authority.

The persistence of a policy program is explained by precisely this long-term social identification. Even if advocating a different policy is more beneficial to an individual in the short or long term than adhering to the policy program, social identity prevents rejection of the jointly developed program. Although a policy program is rarely a paradigm, since it need not provide a holistic view and need not state causal relations or assumptions, a paradigm can be used as a policy program if it is adopted for a defined subsystem. Policy programs are also much more explicitly linked to policy change than the current discussion of the relation between policy paradigms and policy change would suggest (Daigneault, 2014, p. 463).

2.2 Ambiguity and Multiple Streams in the Policy Process

Ambiguity and time are the key constraints on rational behavior in the MSF (Zohlnhöfer & Rüb, 2016). To cope with the complex environment in which policy actors find themselves, they not only engage in routines (Lindblom, 1959; Wildavsky, 1964) (Lindblom, 1959; Wildavsky, 1964) that would suggest incremental change, but their cognitive capacities force them to selective attention. As a result, policy actors may pay only serial attention to issues (Simon, 1985), with only one issue receiving the most attention. Analogously, the political system may consider only one issue at a time at the highest level of the political agenda. Originally developed to explain agenda setting, the MSF draws on this understanding of prioritization of attention to formulate hypotheses about how situational factors in an ambiguous environment promote a policy proposal to the top of the agenda. The core of this idea is found in the title “Agendas, Alternatives and Public Policies” by Kingdon (2003).

Rationality is bounded not only by the limited time available to actors to process information, but also by the phenomenon of ambiguity. The MSF assumes that the definition of problems and the solutions available to solve them are subject to ambiguity in the sense that they cannot be interpreted in a definite way but comprise different aspects that can be considered. A high degree of ambiguity among policy actors then leads to a situation in which many policy actors have no preconceived opinion on many policy issues and are therefore easily persuaded that a proposed policy measure is necessary or desirable at a particular time. Such a predisposition leads to the idea that situational aspects are crucial in the adoption of a policy. In developing the MSF, situational relevance was portrayed as a parallel flow of problem stream, policy stream, and politics or political stream, which has to be coupled by a policy entrepreneur when a window of opportunity opens. The problem stream includes problem definitions and matters of public attention; the policy stream entails possible policy solutions; the political stream consists of features that are elemental to everyday politics, such as the national mood, elections or changes in government, and interest groups (Béland & Howlett, 2016, p. 222; Zahariadis, 2003, 2016). By national mood, scholars often mean public opinion (Durr, 1993; Enns & Kellstedt, 2008), operationalized, for example, by the salience of certain issues to the public (Bromley-Trujillo & Poe, 2020). The particular pertinence of situations and entrepreneurs inherent in the MSF is informed by insights from organizational theory, particularly the garbage can model of organizational choice, which views solutions as floating independently of designated problems and emphasizes the chaotic, anarchistic nature of decision-making in organizations (Cohen et al., 1972). Criticisms of translating the garbage can model into an analytical frameworks call for more explicit consideration of the institutional context (Sætren, 2016, p. 27).

The application of the MSF has successfully answered this claim, translating the perspective from US-American politics to European policy process research and the EU (Ackrill et al., 2013; Herweg, 2016). In the course of assessing the traveling capacity of the MSF, several adaptations accompanied this endeavor. One is the adaptation of the MSF to parliamentary party systems with a particular focus on institutions (Herweg et al., 2015). Another concerns not only the application of the MSF to agenda-setting processes, but its extension to the decision-making process with a separate coupling process (Zohlnhöfer et al., 2016) or implementation (Fowler, 2018, 2020; Howlett, 2018). This includes the observation that in the agenda-setting stage, it is sufficient to generate interest among policy actors, while in the later stage of decision-making, it is necessary to organize the required majority for the adoption of the policy proposal (Zohlnhöfer, 2016). A consideration of the institutional conditions under which policy actors operate is not only important for the MSF studies on policy entrepreneurs, but indispensable for any theoretical perspective dealing with agency and actor behavior in the policy process. It is the essence of the connection between ideas and institutions.

In addition to the policy entrepreneur, other types of actors populate the current landscape of MSF research. Although the ontological underpinnings are different, advocacy coalitions can be integrated into the political stream, and epistemic communities and instrument constituencies have been theorized to be relevant in the problem and policy stream, respectively (Mukherjee & Howlett, 2015). It is in the political stream, at the decision-making stage, that a political entrepreneur can excel. By holding an elected position with formal decision-making power, the political entrepreneur is able to adopt an idea prepared by the policy entrepreneur in the policy stream and agenda-setting stage, increasing the likelihood of adoption depending on the veto players standing in their way (Herweg et al., 2015; Zohlnhöfer et al., 2016). Several case studies empirically confirm the notion of the political entrepreneur, such as research on labor market policy (Zohlnhöfer, 2016), infrastructure policy (Bandelow & Vogeler, 2019), local energy policy (Kagan, 2019), and administrative reforms (Lichtmannegger & Tobias, 2020). Other types of actors include the problem broker, who defines and frames problems in the problem stream and provides tying knots for the policy entrepreneur (Knaggård, 2015). In this context, what is ultimately considered a problem also depends on the institutional conditions of the dominant coalitions and networks (Reardon, 2018). Another is the “bricoleur”, who takes responsibility for finding the best possible solution to a given problem, choosing from a variety of existing solutions and ideas (Carstensen, 2011, p. 154; Deruelle, 2016, p. 49).

Based on how agency is conceptualized in the MSF, an important aspect of interest here is what entrepreneurial strategies ensure the success of a policy proposal. A policy entrepreneur is a corporate or individual actor, for example, a rapporteur in the European Parliament (Thierse, 2019), who promotes a proposal in the policy stream to make it a notable alternative in the policy community (Herweg et al., 2017, p. 28). Together with resources and access to decision-makers, entrepreneurial strategies represent the key to entrepreneurial achievement (M. D. Jones et al., 2016, p. 16). Entrepreneurial strategies include bargaining, the use of narrative stories (Tullia Galanti & Sacchi, 2019) or other discursive strategies such as framing and issue linking (Zahariadis & Exadaktylos, 2016, p. 62f), the continuous formation of supportive coalitions (Mintrom & Norman, 2009), venue shopping (Brunner, 2008, p. 505), promoting their own expertise (Witting & Dudley, 2020), and “salami tactics”, that is, dividing the effort into several phases and promoting a policy proposal stepwise (Zahariadis, 2003, p. 14f). By organizing themselves into smaller groups, policy entrepreneurs can also strategically prevent opposition actors from gaining access to decision-making processes in order to ensure their own success (Johannesson & Qvist, 2020, p. 4). Thus, in their actions, entrepreneurs and policy actors are always enabled and constrained by institutions (Zahariadis & Exadaktylos, 2016, p. 62). Recent research has also identified a type of institutional entrepreneurship that describes cases in which policy actors span multiple institutional contexts and draw inspiration for policy innovation and institutional change from them (Bakir & Jarvis, 2017, p. 466). There is evidence of the existence of Institutional entrepreneurs in health policy, for example, when actors with professional experience in a policy sector enter the policy-making process to bring their normative ideas to policy-makers and public officials (V. Smith & Cumming, 2017, p. 532). Drawing on the literature on knowledge utilization, Blum (2017) develops the concept of argumentative coupling, which describes the strategic use of scientific arguments and evidence to couple streams. The active role of the policy entrepreneur may even extend to influencing the opening of a policy window, especially at smaller levels of government (Cairney, 2018, p. 202). Such strategies for success may also be fruitful, at least in part, for other types of actors in the policy process, including programmatic groups.

With respect to the PAF, there are some parallels that can be drawn when comparing it to the MSF, and some blind spots that the PAF may fill. Kingdon’s original book remains intentionally non-specific about the source of ideas, finding them “anywhere” (Béland, 2016, p. 231). By focusing on the emergence of policy programs, the PAF echoes this view, but sheds more light on how a policy program is composed from the multitude of existing ideas and how it is purposefully designed by intellectual circles. It thereby implies a much stronger linkage to the policy design literature (Capano & Lippi, 2017; Howlett et al., 2018; Peters et al., 2018). In this regard, the PAF’s understanding of a programmatic group that deploys a policy program for strategic and normative purposes is compatible with the MSF’s understanding of the motivational drivers of policy entrepreneurs. Kingdon finds these drivers in “self-interest, the promotion of values, passion for shaping policy, and the pleasure that advocacy and public exposure give them” (Béland, 2016, p. 233; Kingdon, 2003, p. 123). Moreover, programmatic groups resemble policy entrepreneurs in that they also link their policy solutions (policy programs) to a defined problem, whether framed by themselves or by problem brokers. MSF research could thus benefit from integrating the concept of programmatic groups as a specific type of entrepreneur with different resources and a more narrowly defined policy program. However, the important distinction of the PAF remains in the social psychological view of social groups and identities in the policy process. Programmatic groups are bound by the policy program, and their shared experience and resulting mutual trust ensures long-term cooperation among programmatic actors beyond purely rational interests. Thus, the longevity of programmatic action and reliance on long-term strategies and careers rather than situational opportunities for major policy change is the key distinguishing feature of the MSF and the PAF.

Given the often complex and historical nature of health care institutions, combining entrepreneurship and agency-based explanations for major reforms with institutional predispositions is critical to explaining policy outcomes (V. Smith & Cumming, 2017). The combination of the Multiple Streams Framework and institutional veto players (Zohlnhöfer, 2016, p. 89), as well as the MSF and historical institutionalism, in which policy entrepreneurs that use policy windows present critical junctures (Spohr, 2016) has proven fruitful. This is close to the PAF’s idea of linking individual actors with the institutional positions they hold and integrating both aspects into a coherent framework of policy change. Following on from the roots of collective action, the next two subchapters turn attention to different types of groups, networks, and coalitions, both to highlight the limitations and hypotheses of programmatic groups and to work out the essence of programmatic groups compared to other types of networks.

2.3 Beliefs and Advocacy Coalitions in the Policy Process

One of the most prominent and widely applied actor-centered frameworks in policy process research focused on networks is the ACF. Consistent with the line of thought that traces the psychological roots of actor-centered approaches in public policy, the ACF continues the understanding of a boundedly rational individual, but seeks the limits to rational processing of information not in ambiguity and time constraints, but in the existence of belief systems. Belief systems are cognitively existing sets of elements that collectively represent coherent individual attitudes, values, and preferences (Converse, 2006, p. 3). Explicitly following Converse, Sabatier adapted this understanding to elite belief systems, proposing that they encompass perceptions of causal mechanisms that relate to policy areas (Sabatier & Hunter, 1989). Accordingly, in the ACF, belief systems are defined as “sets of value priorities and causal assumptions about how to realize them” (Sabatier, 1988, p. 131). The notion that they are hierarchically structured into three levels, with (1) deep normative core beliefs, (2) policy-related core beliefs, and (3) secondary aspects, is borrowed from Lakatos’ conceptualization of scientific research programs (Jenkins-Smith et al., 2017, p. 136; Lakatos & Musgrave, 1970).

Over the past three decades, the conceptualization of belief systems has been both challenged and advanced (Brandt et al., 2019; Hann, 1995; Ripberger et al., 2014). In operationalizing deep normative core beliefs, Sabatier and Jenkins-Smith (1993) originally referred to basic convictions regarding the relation between nature and humans that were reinforced by socialization (Searing et al., 1973). However, this operationalization has been specified using cultural theory, which approaches them through the concepts of individualism, fatalism, egalitarianism, and hierarchy (Jenkins-Smith et al., 2014; M. D. Jones, 2011; Sotirov & Winkel, 2016). Underlying this research is the central research interest of what drives individual political preferences, with current explanations for political attitudes ranging from biological to partisan, cultural, and belief-driven (Bachner & Hill, 2014). Bachner and Hill (2014, p. S65) emphasize that further research is needed to specify under which conditions which factor is more relevant. Context and individually different ways of interpreting the world, caused either by belief systems or other types of cognitive filters, thus considerably shape the formation of policy preferences. Because belief systems influence how actors interpret and process information, actors are more susceptible to cognitive bias and differences in how they perceive the world and the information presented in it, leading to segregated policy networks (Henry, 2011a) and biased assimilation of policy preferences and policy beliefs (Moyson, 2017). Therefore, psychological and social psychological approaches to explaining individual preferences have proven insightful when integrated into the study of policy processes.

The contemporary understanding of the ACF largely runs along three lines—advocacy coalitions, policy change, and policy-oriented learning (Weible & Nohrstedt, 2012). Shared policy core beliefs present the unifying force of advocacy coalitions that strive to translate their policy preferences into adopted policies (Zafonte & Sabatier, 1998). An important element of cooperation among coalitions is the trust that follows from shared policy core beliefs, as well as from small and stable groups (Leach & Sabatier, 2005). Advocacy coalitions compete for dominance in a policy subsystem and use their collective resources to achieve policy change in the direction of their own policy preferences. Depending on the degree of cooperation between competing advocacy coalitions, policy-oriented learning becomes more or less likely (Koebele, 2018, p. 50). The defining characteristics of advocacy coalitions are that they consist of policy actors who share policy core beliefs, act in a coordinated manner, pool resources, and are stable over time (Weible et al., 2019).

For policy change to occur, the original ACF places a strong emphasis on the role of policy-oriented learning. Policy-oriented learning occurs when experiences and knowledge lead to the modification of beliefs or secondary aspects (Sabatier, 1987). However, there is an important distinction between minor change (change in secondary aspects) and major change (change in the policy core) in policies that has received less attention in ACF studies (Pierce et al., 2017, p. S29). Hypotheses expecting policy learning under conditions of forums, quantitative performance measures, and policy subsystems of natural sciences were added and modified, while the core assumptions defined above were left largely unchanged (Wellstead, 2017, p. 558). However, learning processes can also be the result of adaptive reactions to criticism in the absence of external shocks (Lertzman et al., 1996).

Meanwhile, research on learning has expanded beyond the boundaries of the ACF and is currently discussed as a stand-alone framework or theory (Dunlop & Radaelli, 2018). While learning and belief change are not of central interest to the theoretical development of the PAF, not least because of the fundamentally different understanding of how preferences are formed, one aspect in the learning literature appears to be transferable to the concept of social and programmatic groups, that of socialization (Dunlop & Radaelli, 2017, p. 313). Whenever policy actors enter new positions with new constraints on their daily actions or find themselves in new situations, including new memberships in groups to which they begin to belong, their values change and so do their preferences.

It is precisely this influence of social groups on individual preferences and behavior that allows the PAF to propose a group-theoretic explanation of policy change. This goes beyond the cooperation among individuals to translate their shared beliefs into adopted policies, which defines collaborative action as an ideologically driven act to achieve subjective goals. Instead, the PAF sees the group itself as the main generator and driver of ideas. Not only is it a resource in promoting policy, but the success of the group itself becomes the central goal of its members’ engagement in policy-making.

In addition to learning, the ACF describes three other pathways to policy change. First, negotiated agreements can serve as a mediating element between advocacy coalitions and lead to major policy change (Metz et al., 2020). For negotiated agreements to be concluded, policy brokerage can be a critical determinant of success (Ingold, 2011). Second, internal shocks to the subsystem and external perturbations can trigger change in the subsystem. These latter two types of shocks are similar to what the MSF calls policy windows that open from the political (internal shock) or problem (external shock) stream. Empirical studies of the ACF often combine several of these pathways to identify policy learning and, in particular, the possibility of major policy change as more likely under the impact of external or internal shocks (Albright, 2011; Ardıç et al., 2015; Crow et al., 2018; Markard et al., 2016; Sandström et al., 2020). Negotiated agreements receive relatively little attention compared to learning and external events, but more than internal events and superior jurisdictions (Pierce et al., 2020, p. 71).

Regarding the roots of collective action, the ACF complements research by focusing on the role of shared beliefs as binding elements of coalitions or networks. However, the initial desire to engage in collective action is a rational one, as actors expect benefits from joining coalitions and the opportunity to realize their policy preferences (Schlager, 1995). Since Sabatier’s original thinking, this has been grounded in the Theory of Reasoned Action (Ajzen & Fishbein, 1980), which was later developed into the Theory of Planned Behavior (Ajzen, 1991) Criticisms focus on the ACF’s inability to explain collective action problems, to which it responded by integrating access to information (Yagci, 2019), repeated interaction, or interdependencies (Fenger & Klok, 2001) as factors to overcome these problems. In addition to shared policy beliefs, empirical evidence supports the claim that perceived actor influence contributes to building actor relationships (Weible, 2005, p. 471) and that the preexistence of organizational structures fosters the formation of advocacy coalitions (Kübler, 2001, p. 628). Schmid et al. (2019) combine the ACF with PFT to outline how policies influence the structures of and available resources of advocacy coalitions. Such a view corresponds well with the added value of PFT, seen in this book particularly in the way policies account for the resources and dominance of a dominant programmatic group.

While the ACF assumes shared policy core beliefs as a predisposition for coalition formation, this unilateral causal direction is controversial in the current state of the art on networks and attitude formation. Numerous empirical studies have confirmed that actors adopt political values, attitudes, and preferences from those with whom they regularly interact in social relationships, not that they form social relationships based on their preexisting views (Goren, 2005; Lazer et al., 2010; McPherson et al., 2001). The PAF presents an attempt to integrate social psychological research on the SIA into theories of the policy process to fill this gap (Hogg et al., 2017; Hornung et al., 2019). The proposition that preferences are shaped later than group membership and that social identities are a crucial explanation for policy actors’ behavior is a major innovation that the PAF contributes to existing policy process research.

In terms of the role of institutions, the ACF incorporates into its flow diagram and theoretical model political opportunity structures, by which it understands the stable parameters of institutional settings (Nohrstedt et al., 2020, p. 71). Specifically, these are the ways in which institutions influence actors’ belief systems and their resulting behavior (Lubell, 2003). In the research literature, these have been operationalized as consensus or majoritarian democracies and openness of the political system, and more concretely as federalism and corporatism (Kübler, 2001; Weible et al., 2009, p. 124). For example, the ACF certainly recognizes that corporatist structures shape advocacy coalitions differently than non-corporatist settings with regard to the actors involved (Gronow et al., 2019). Institutional settings can also facilitate policy change through advocacy coalitions if they provide forums or institutionalized venues where negotiated agreements are reached (Metz et al., 2020, p. 20). Despite these advances, scholars still call for more research to identify the political institutions that actually matter for coalition-building (Heikkila et al., 2019, p. 163). Importantly, institutions can also change within subsystems and thereby influence coalition behavior, which Gupta (2014, p. 360) attributes at least in part to the role of political parties. The institutions included in ACF research therefore are similar to those analyzed in this study.

2.4 Network Theory and Agency in the Policy Process

Within the broader field of policy process research, there exist several theories about the formation of networks and their influence on policy-making. Since the notion of a programmatic group is comparable to a closely collaborating network, it is worthwhile to place PAF in the realm of network theories and to demarcate programmatic groups as a very specific type of network. Several studies have provided an overview of what can be considered a network and the different types of networks (Carlsson, 2000; G. Jordan & Schubert, 1992). Essential to any network conceptualization is the assumption about what drives individuals to engage in collaboration and joint action. A review of the MSF and ACF literature has highlighted the strategic use of alliances (as in the MSF) and the formation of coalitions around shared policy core beliefs to translate them into policy (as in the ACF) as potential motivations for individuals to collaborate with other policy actors. This chapter will consider a selection of the remaining types of networks, extrapolate their boundaries, and relate them to the new type of network, the programmatic group.

In order to use the concept of actor networks in policy process research, it is essential to adequately define their scope. This is most easily done by answering the following four questions: What actors does the network consist of, or can it theoretically consist of? Why do these actors work together, what is their motivation for joining the network? What holds the network together (the question is related to the decision of individuals to join a network, but the answer is not necessarily the same)? What is the goal of the network and what strategies are available? In reviewing the existing notions of networks that are prominently used in public policy research, these four answers are kept in mind to answer. While the literature on networks differs on whether individual actors or corporate actors, usually organizations, are the nodes within a network, when applying psychological explanations of collective action and group (or network) formation, the unit of analysis must logically be an individual. As a consequence, network theories that focus on corporate actors such as organizations are excluded from consideration, while acknowledging the extensive research on corporate actors such as the organizational state (Laumann & Knoke, 1987).

In policy process research, network theories are typically concerned with the study of elites. While there are several elite theories in other disciplines such as sociology (Rahman Khan, 2012; Scott, 2008), the study of policy networks dates back to C. Wright Mills’ work on national power elites in the US to outline the importance of occupying key institutional positions in influencing the policy process in general and decision-making in particular (Knoke & Kostiuchenko, 2017, pp. 92-93; Mills, 2000, p. 9). Accordingly, most theories of the policy process explicitly designate the types of actors that are potentially of interest or subject to their theoretical propositions. In this regard, most policy process theories use the policy subsystem as the level of analysis (Nowlin, 2011, p. 54). The ACF has a fairly broad understanding of policy actors as those who populate a subsystem. Policy subsystems are distinguished as systems of specialized policy actors who engage with and seek to influence the policy issue (Sabatier & Weible, 2007, p. 192). Other ACF studies refer to it as “a partition of a broader governing system that focuses on a policy topic within a geographical area and involves specialized policy actors attempting to influence policy subsystem affairs” (Olofsson et al., 2018, pp. 645-646). The concept of policy subsystems overlaps with that of policy domains in that it encompasses public and private actors that revolve around a commonly perceived (policy) problem and are characterized by a shared desire to solve it (Laumann et al., 1985, p. 2).

In addition to the policy subsystem, other conceptualizations of networks and aggregates of actors have emerged. Rhodes refers to them all as policy networks (Roderick A. W. Rhodes, 2008, p. 425f). What they all have in common is a policy-oriented clustering of actors (corporate and/or individual) and the description and analysis of their interconnections, especially those of state versus non-state (above all industrial) actors, with respect to their influence on policy-making processes. In this context, they denote a type of sectoral policy-making and are often understood as running along a continuum (Miller & Demir, 2007, p. 140). The terminological confusion and overlap of concepts also stem in part from the fact that some terminologies emerged from US contexts and empirical research on US politics and others are explicitly communicated from a British perspective, for example, policy communities (Roderick A. W. Rhodes, 1990, p. 302).

Howlett and Ramesh (1998, p. 469) distinguish between policy communities and policy networks, with the former encompassing a large set of actors who have knowledge of a policy topic and the latter comprising only those who collaborate out of an interest in influencing policy-making; both are seen as subsets of policy subsystems. Exactly how the concepts relate to each other, however, is a matter of debate in the research literature. While some agree with Howlett and Ramesh and see policy communities as broader than policy networks, with networks explicitly referring to the connections between actors, others see policy communities as a specific type of network (M. M. Atkinson & Coleman, 1992, p. 158). Policy communities are generally referred to as being limited—and therefore tantamount to—policy subsystems and sectors. Comparable to sub-governments, they are characterized by shared attitudes and temporally stable relationships between government groups in bureaucracy and politics and interest groups (G. Jordan, 1990, p. 331f). Because they are so exclusive, they promote exchange between members (Grant & MacNamara, 1995, p. 509). In doing so, they can also be territorially bound (Keating et al., 2009, p. 53), transnationally active (Stone, 2008), or competing in policy arenas for policy change (Dudley & Richardson, 1996). The terminological derivation of policy communities is essentially rooted in the critique of iron triangles and sub-governments as overly simplistic and exclusive representations of policy-making processes (G. Jordan & Maloney, 1997, pp. 559-561). This discussion dates back to the seminal work of Freeman and Stevens (1987) and concerns the early use of the term “policy subsystem”, which was later opened up to include actors with specialized interests, making it issue-driven rather than actor-driven (McGee & Jones, 2019, pp. S140-142). Nonetheless, a conceptualization of policy communities allows for the emergence of subgroups such as advocacy coalitions (Birkland, 1998, p. 57).

In a similarly issue-driven perspective, the concept of issue networks, originally introduced by Heclo (1978), refers to the linkages between actors in a fragmented and highly diversified web of relationships among actors that influence policy-making. Freeman and Stevens (1987, p. 14) place them on a par with subsystems because of their vagueness in designating members. Unlike Rhodes’ network perspective, Heclo sees members of issue networks as less bound and fixed, but floating, and the main motivation of actors is emotional or intellectual advancement rather than interests (Schneider, 2015, p. 375). Consequently, issue networks are much more open to actors who are not institutionally linked to the state, allowing them to be conceived as shapers of public policy, such as non-profit policy activists (Nyland, 1995). They are specifically tied to an issue, although they may differ in terms of their agreement on the concreteness of legislation, advocacy arguments, and conceptualization of the issue, as well as membership, member-to-member exchanges, and leadership (Kirst et al., 1984). In contrast to iron triangles and the neo-corporatist picture of policy-making, issue networks are less likely to resolve conflicts among their members than the latter, but more open than the former, with iron triangles limited to the presence of interest groups, and corresponding administrative and legislative bodies (A. G. Jordan, 1981, pp. 96-98). Within issue networks, evidence suggests that past experiences with network interactions also influence whether alliances will form in the future and that those who have had contact in the past are more likely to contact again compared to other actors in the same larger network (Heaney, 2004). As the network changes, so do the terms used to describe it. For example, M. J. Smith (1991) noted that as scandals and media attention on the issue of salmonella in eggs increased, the policy community morphed into an issue network, taking in actors who did not originally deal with this policy issue on a regular basis.

Similarly revolving around an expertise on a policy issue, but much narrower in scope than issue networks, epistemic communities include scientific experts on a policy issue. Originally introduced by Haas, the concept of “networks of knowledge-based experts” (Haas, 1992, p. 2) builds on the argument that state actors in contexts of uncertainty and complex policy problems form preferences based on the advice and knowledge integration of these epistemic communities. Therein lies their influence on the policy process. As they disseminate their knowledge, epistemic communities can cross national boundaries and shape international views that are translated into policies and treaties (Adler, 1992). Going beyond this initial narrow definition is the observation that policy actors with a biographical trajectory relevant to the issue at hand can be recruited by epistemic communities in the process of forming a strategic alliance to advance a policy proposal. In the case analyzed by Grødem and Hippe (2019), civil servants from the Norwegian Ministry of Finance formed an epistemic community and strategically cooperated with the prime minister, whose expertise was similar to theirs due to his educational background in economics. Alternatively, epistemic communities may also support advocacy coalitions, thereby impeding policy-oriented learning and a change in preferences based on incoming scientific information (Meijerink, 2005, p. 1070). In this context, which knowledge is provided by which epistemic communities is highly dependent on the respective policy actors. For example, the way international organizations such as the OECD and the World Bank select epistemic communities and draw on their knowledge depends on the mission pursued and the internal structures of the organization (Francesco & Guaschino, 2020, p. 115).

Policy communities, policy subsystems, sub-governments, iron triangles, issue networks—they all have a special focus on relationships between groups of actors and thus concern the meso level of action. While they often start from the policy issue before deriving the actors involved, they consider not only the structural setting of actors but also their interactions and influence on policy processes. In addition, they usually allow for the inclusion of interest groups and other actors who do not have formal institutional power to be relevant to the policy process. In considering individual action, many network approaches thereby look for the reasons why individual actors cooperate in interest-driven behavior—or do not specify the psychological basis for micro-level decisions to join networks at all. Moreover, interest-based action is usually an expression of rational behavior. However, there are other roots for collective action that explain cooperation in the absence of rational cost-benefit analyses. The PAF complements these perspectives by looking at biographies and career ambitions as drivers of collaboration and by integrating social identities.

A rather newer strand of research focuses on the concept of instrument constituencies. As the term implies, they are particularly concerned with one policy instrument and how it has been innovated and developed over the course of the policy process. This also means that they are not tied to one subsystem but span multiple policy cycles (Weible, 2018, p. 63). Despite obvious overlaps with the concept of epistemic communities, due to the focus on shared and exchanged knowledge and ideas, as well as a potential transnationality, Zito (2018) finds considerable differences between instrument constituencies, which are centered around and exist through the policy instrument, and the more knowledge-oriented epistemic community. However, both concepts include the possibility that their defining feature may become a source of shared group identity (Zito, 2018, p. 44). Following the MSF perspective of policy-making structured around the problem stream, the policy stream, and the politics stream, Béland, Howlett, and Mukherjee (2018, pp. 7-8) situate instrument constituencies in the policy stream because they develop and promote the solutions to which problems can be assigned and then because the promotion of those very solutions is their binding element. Consequently, the bond between the instrument constituency and the policy instrument it promotes may be so strong that one does not exist without the other, thus providing an important source of identification (Voß & Simons, 2014, p. 738). Interestingly, the concept of instrument constituencies also holds the structural promise that the implementation of a policy instrument may entail an increased demand for expertise and the creation of instrument-specific institutions, from which promoters of that instrument may benefit career-wise by occupying new positions (Simons & Voß, 2018, p. 22). This aspect is quite close to the career-related motivation of programmatic actors, but programmatic groups, in contrast to instrument constituencies, initially turn out to be rooted differently and characterized by a higher degree of cooperation and social identity.

Often, the alternative theoretical explanations outlined above share the argument that cooperation is a result of (or influences in a reciprocal way) some sort of sameness and cohesion among individuals. This sameness and cohesion can stem from similar cultural worldviews (Weare et al., 2014), ideology and policy core beliefs as in the ACF (Henry, 2011b), expert knowledge as in epistemic communities (Adler & Haas, 1992), issue-specific knowledge in issue networks (Kalafatis et al., 2015), coming together around specific policy design elements (Haelg et al., 2020, p. 331) or other characteristics shared by the actors. Narratives can also emerge as a unifying element of networks, referred to as narrative networks or discourse coalitions (R. Atkinson et al., 2011; Bulkeley, 2000; Ingram et al., 2019; Lejano et al., 2017). While sameness contributes to network integration, efficiency (in the sense of mobilizing resources to influence policy-making) and innovation in network activities grow from heterogeneity of actors in what regards their positions and resources (Sandström & Carlsson, 2008).

Based on a summary of several network approaches, the theory of collaborative policy networks proceeds starts the main defining features of networks to propose hypotheses about structural signatures of policy networks. For example, it states that collaborative policy networks are more likely to be characterized by diversity, reciprocity, horizontal power structures, long relationship histories, trust, transparent decision-making processes, and structurally equivalent leadership (DeLeon & Varda, 2009). The role of trust and reciprocity in particular is emphasized by theories of social capital, most prominently by Putnam’s work “Bowling Alone”, which proposes to define social capital as “features of social organization such as networks, norms, and social trust that facilitate coordination and cooperation for mutual benefit” (Putnam, 1995, p. 67). In public policy, social capital and trust have been studied in relation to how to overcome problems of collective action and management of common pool resources. Factors seen as contributing to mutual trust include: frequent interactions (Lubell, 2007, p. 248), cohesion among group members in terms of similar beliefs (Leach & Sabatier, 2005, p. 493; Lubell, 2007, p. 239; Parsons, 2020, p. 39), opportunities for and experiences with cooperation and collaboration (Fischer & Sciarini, 2015, p. 65), professional competence and expertise (Calanni et al., 2014, p. 917), resources, and power derived from network centrality (Berardo, 2008, pp. 186-187). From a psychological perspective, biased perceptions of opponents as more evil and powerful reduce trust (Fischer et al., 2016, p. 330) as do personal and individual predispositions that make it easier for certain individuals to trust others (Dinesen et al., 2014; Freitag & Bauer, 2016, p. 141). Research findings suggest that trust is a mandatory prerequisite for long-term close collaboration and that frequency of contact and commonalities foster trust. This comes closest to the idea of the PAF.

The development of the study of policy networks also contributes to research on institutions and public administration. Firstly, it also reveals the replacement of the long-held dichotomous view of politics and administration, in which bureaucrats are seen as purely functionally driven individuals directed by their respective political superiors (Skok, 1995, pp. 329-330). Instead, in the policy networks studied here, bureaucrats and administrative actors themselves can actively participate in the meaningful construction of agendas and the formulation and adoption of policies. Second, with respect to institutions, the argument that the institutional positions of policy actors and their structural positions in a network determine the perceived power of actors is underscored (Ingold & Leifeld, 2014). Overall, this confirms the added value of the PAF perspective, which also addresses these aspects.

Summarizing perspectives on policy networks and typologies of aggregates of policy actors, it is clear that an important prerequisite for comparing and advancing research is to be clear about the terms one uses to label one’s work on policy actors. Depending on the scientific community in which research is conducted, different terms have different meanings and, accordingly, different assumptions about actors are predefined. Rather than getting caught up in trying to define structural limitations of subsystems, “it may be more fruitful to focus instead on the functional characteristics of subsystems, what is their behavior under specific conditions” (McCool, 1998, p. 558). This means that in choosing an appropriate term to describe the object of research, attention is drawn to how and what kinds of conflicts arise in policy-making processes and how they are constrained by existing conditions. Similarly, it is important to be clear about the goals of using network approaches, whether they serve a purely descriptive goal or are explicitly understood as dependent and independent variables embedded in effect mechanism (Thatcher, 1998, pp. 410-411).

The scope and terms in the PAF have already been presented and distinguished from other similar network definitions in this subchapter. Essentially, the literature on policy networks emphasizes that policy actors in policy networks always share one or more characteristics that serve as defining features of that network. Sometimes these characteristics are rather loose, such as repeated engagement with and reflection on a particular policy issue. Sometimes the defining characteristics of a policy network are narrower and include only those who share similar beliefs (such as advocacy coalitions). In the view of other perspectives, networks are mere products of institutional settings and follow the rules and scope that are institutionally imposed on them for rational cost-reducing interests, as is the perspective of the Actor-Centered Institutionalism (Bevir & Richards, 2009, p. 6; Scharpf, 1997).

Complementing these perspectives, the PAF brings a new approach to research on network formation in policy process research. The first aspect is the role of shared biographies in creating a sense of belonging and thus the roots for cooperation. Second, social psychological insights into the dynamics of group processes have only partially found their way into these studies. Among the rare contributions are those dealing with learning processes across advocacy coalitions (Koebele, 2019; Wagner & Ylä-Anttila, 2020, p. 196) or biased perception in the context of devil and angel shifts (Nilsson et al., 2020). The main argument made in the following chapters is that a focus on these group-based drivers of identification with a policy program can explain major policy change beyond existing policy network approaches. The main goal of the empirical analysis is to systematically test hypotheses about which institutional influences promote programmatic action and the success of programmatic groups.