Introduction

Recently, while sorting through a pile of assorted papers that had been collecting dust in the corner of my office and were destined for the recycling bin, I came across a yellowing seven-page single-spaced letter dated January 17, 1997, from the late Paul Sabatier. In great detail, he provided a chapter-by-chapter review of my M.Sc. thesis “The role of the advocacy coalition framework in understanding forest policy change: Alberta and Ontario.” Like a growing number graduate students during the mid-1990s, the advocacy coalition framework (ACF) was starting to come into vogue. At this time, controversies surrounding old-growth logging and the long-term horizon of policy change in British Columbia’s forest sector made for fertile ground that also advanced policy process scholarship (Wilson 1987; Cashore et al. 2001; Rayner et al. 2001) including debating the applicability of the ACF (Lertzman et al. 1996a, b; Hoberg 1996). My comparative study examined the shift from sustained yield timber forestry to sustainable forest ecosystem-based management in two provinces with large public forests, but had received far less academic attention (Wellstead 1996). Sabatier’s review began by highlighting its positive contributions to ACF scholarship, but more importantly most of the comments were dedicated to critically assessing the application of concepts, areas of disagreement, the need to consider quantitative methods and revised causal path diagram. Many of these suggestions were explored in a subsequent ACF application (Wellstead and Stedman 2007).

This personal vignette is important because it illustrates, in a small way, how the ACF came to be and continues to be a dominant approach in the policy process theory literature. It is safe to assume that if Sabatier painstakingly reviewed a modest Master’s thesis, he also learned from many other applications. Sabatier and other ACF scholars are committed to a “Lakatosian” research program whereby the core attributes of the ACF outlined in his 1988 Policy Sciences paper, “An advocacy coalition framework of policy change and the role of policy-oriented learning therein” (herein referred to as the “1988 Paper”), remain intact today (Cairney 2012). Despite constant revisions of the ACF, including a number by Sabatier himself, the 1988 Paper is the most cited paper in Policy Sciences (approaching 3000 Google Scholar citations) and shows no signs of decreasing its yearly citation output.

This 50th Anniversary review highlights a number of key reviews of the ACF that has been generated over that past 25 years. First, however, the ACF as originally presented in 1988 is highlighted. A quantitative analysis of papers—with journal article abstract data obtained from the Scopus—citing Sabatier’s paper is undertaken. The VOSviewer software visualization program reveals networks of key terms from article abstracts citing the 1988 Paper (van Eck and Waltman 2007, 2010). Two types “reviews of ACF reviews” are highlighted: those examining contributions, modifications, and shortcomings of the ACF, and comparisons of the ACF with other policy process theories and frameworks.

A framework from a framework: continuous critical self-assessment

The earliest formulation of the ACF can be found in Sabatier’s 1986 Journal of Public Policy paper “Top-Down and Bottom-Up Approaches to Implementation Research: A Critical Analysis and Suggested Synthesis.”Footnote 1 The ACF was seen as a product of “critical self-assessment” in his earlier implementation research (the Sabatier/Mazmanian framework) as well as the growing complexity of US policy-making (Heclo 1978). Specifically, he advocated a general model of the policy process which combines the best features of the bottom-up and top-down approaches, while also applying them to a longer time frame than what was the case in most implementation research (Sabatier 1988, 36). Sabatier was primarily concerned with understanding and explaining causality within complex multi-level policy-making processes. The outline of the proposed ACF is sketched out in last five pages of the article, including the well-known and universally copied heuristic “flow diagram” (Fig. 1) containing a “common vocabulary including major conceptual categories and general relations between them” (Weible and Nohrstedt 2012, 127).

Fig. 1
figure 1

(Reproduced with permission from Weible et al. 2009)

The ACF flow diagram

Full size image

Sabatier’s 1988 Paper fleshes out his earlier proposal. The framework’s main assumptions include that the unit of analysis required to understand policy changes occurs within what he referred to as a “political subsystems.” These subsystems will often be populated by more than 30 governmental and non-governmental organizations. Events external to the subsystem are considered primary inducements for major policy shifts and constrain the actions of the policy actor. A second path of policy change can occur within subsystems providing that they are on minor or less controversial aspect of policy. Measuring the impacts of policy change and policy learning requires a time perspective of a decade or more. Sabatier found that there is typically one dominant coalition along with two to four key contending or emerging coalitions within a subsystem (Lindquist 1992). What differentiates one advocacy coalition from another is a three-leveled hierarchical belief system. From these belief systems, coalitions develop their overall policy direction and devise specific programs. A policy-oriented belief system is arranged according to three distinctive categories: a deep normative core, a policy core, and the secondary aspects. The deep core, which is equated with the personality of an individual, is nearly impossible to change. The policy core belief, which is the basic strategy and overall policy position of a coalition, is possible to change but very difficult. Secondary aspects are the instrumental decisions associated with the policy core. Changing a coalition’s core policy belief would eventually alter the basic perception of policy problems as well as the general policy prescription of an issue (Sabatier 1988). But as long as the dominant advocacy coalition remains in power within the subsystem, belief systems are unlikely to change, and the core attributes of a government program are unlikely to be significantly revised. Most routine policy changes occur at the secondary aspect because it does not threaten the dominant coalition’s core policy belief because subsystem actors are willing to give up secondary aspects more readily. This type of policy change occurs via “policy learning” which Sabatier defined as a relatively enduring alteration of thought or behavioral intentions that are concerned with the attainment (or revision) of the secondary aspects of the policy belief system (Sabatier 1988). The advantage of this definition is that it transcends an information-based view of learning and considers alterations in frames, values, and meanings (Kemp and Weehuizen 2005).

Sabatier maintained that policy learning’s influence upon policy change requires a number of common factors. The most significant contribution made in the 1988 Paper was the introduction of nine core hypotheses (Table 1) which represented the first attempt to permit empirical testing of the framework and Sabatier’s desire to take causality seriously in policy research. A great deal of the ACF scholarship over past 30 years has been dedicated to developing new hypotheses or modifying the central assumptions in core hypotheses and has been the subject of reviews highlighted later in the paper.

Table 1 The original core ACF hypotheses in Sabatier (1988)
Full size table

The durability of Sabatier’s 1988 Paper: more than advocacy coalitions

Before reviewing the ACF reviews, it is important, first, to consider the wider impact of the 1988 Paper. A Scopus database search of abstracts fields of documents published between 1988 and 2017 citing “An advocacy coalition framework of policy change and the role of policy-oriented learning therein” yielded 1033 results, of which 737 were peer-reviewed articles. In comparison, using the same criteria, the term “advocacy coalition framework” yielded 227 peer-reviewed articles. Since 2007, there has been a sharp rise in the number of citations of the 1988 Paper (Fig. 2). This trend roughly corresponds to articles citing the ACF. The USA remains the most popular location for the 1988 article and the ACF in general followed by the UK and Canada (Fig. 3). Outside of Anglo-American countries, papers originating primarily from northern European countries dominate. The geographic trend also holds true for ACF-related articles. Henry et al. (2014) found that 75% of ACF empirical applications were found in North America and Europe. The 1988 Paper has been cited most frequently in Policy Studies Journal (40 times) followed by Policy Sciences (33 times), Journal of European Public Policy (27 times), Review of Policy Research (19 times), and Forest Policy and Economics (18 times). In all but one case, the 1988 Paper was, as expected, cited far more frequently than ACF-based article. The exception was Policy Studies Journal with 35 papers examining the ACF (Table 2).

Fig. 2
figure 2

Yearly citations of the 1988 Paper

Full size image
Fig. 3
figure 3

Country of origin of the 1988 Paper

Full size image
Table 2 Number of times the 1988 Paper and ACF have been cited in peer-reviewed journals
Full size table

A freely available bibliometric mapping computer software, VOSviewer, was employed to analyze the abstracts from 737 peer-reviewed articles citing the 1988 Policy Paper. The program produces easy-to-interpret two-dimensional maps in which the distance between nodes (key terms in the abstract) is a measure of similarity. The dataset contained relevant 223 terms that occur in the abstracts. The most frequently occurring terms were: “advocacy coalition framework” or “ACF” (139), “policy subsystem” (35 abstracts), “hypothesis” (30 abstracts), “advocate” (29 abstracts), and “failure” (28 abstracts). Network visualization of sources referenced 10 or more instances of a particular theme from article abstracts referencing the 1988 Paper. A node in this network depicts them. The size of the node indicates the number of instances on which a theme was mentioned in the abstracts. The nodes are clustered based on similarity—nodes in the same color are more similar than nodes in different colors; nodes closer to each other are more similar than nodes farther from each other. A link between two nodes indicates a co-citation relationship.

Every student of the ACF should appreciate the VOSviewer network’s coalition-like output (Fig. 4). The dominant network is the tightly clustered one (yellow nodes) with the largest nodes and the most number of intra-network links, meaning that these terms were frequently used together. Here, the abstracts address advocacy coalitions, beliefs, policy change. This is consistent with previous discussion regarding the large ACF-oriented literature. A smaller but still very densely clustered network (green nodes) presents a methodological/empirical theme (e.g., methods, policy, analysis, comparison, researcher, science, policy network). There are also strong linkages between the ACF network and the methodological/empirical network nodes. A third more scattered network represented by the red nodes presents a less cohesive theme based on ideas, technology, innovation, and decision-making, policy entrepreneurs. While there are many intra-network linkages, very few inter-network linkages exist. Finally, the fourth network represented by the blue nodes has virtually no inter- or intra-network linkages. Papers about stakeholders, science, research uncertainty, and individuals provide a looser and less coherent theme. The “heat” map in Fig. 5 illustrates the overall density of terms with red areas representing the highest density. Policy, advocacy coalition framework, method, stakeholder were the most important concepts from the VOSviewer analysis.

Fig. 4
figure 4

Co-citation networks of sources referenced on 10 or more terms. (Color figure online)

Full size image
Fig. 5
figure 5

Density view of sources referenced on 10 or more terms. (Color figure online)

Full size image

Review of the ACF reviews

In 1993, Sabatier and Jenkins-Smith undertook the first significant review and revision of the ACF. Their critical reassessment for the framework was in part motivated by six empirical case studies. Table 3 lists the subsequent major reviews of the framework and their contribution to ACF research. These reviews provide an accounting of ACF studies, namely their location, or the types of analysis undertaken. With the inclusion of new case studies, there has been the development of new hypotheses. Most of the revisions have focused on the composition and stability of coalitions, the structure of beliefs, interaction in subsystems, coalition behavior, strategies for influencing behavior, policy learning, and policy change.

Table 3 ACF reviews
Full size table

Since 1999, there have been four editions of Theories of the Policy Process edited or co-edited by Paul Sabatier. In each there, there has been a chapter reviewing the ACF along with discussions of other key policy process theories or frameworks. However, it was not until the third edition that a stand-alone chapter was dedicated to comparing the approaches (Cairney and Heikkila 2014) and was subsequently updated in the 2018 edition (Heikkila and Cairney 2018). In Table 4, two lines of inquiry are followed: comparing the ACF with other theories and the role that the ACF plays in synthesizing efforts. Heikkila and Cairney (2018) provide the most extensive comparative overview of the ACF (across 15 indicators) along with the multiple streams approach, punctuated-equilibrium theory, policy feedback theory, the narrative policy framework, the IAD framework, and the innovation and diffusion model across three broad criteria: theoretical elements, research and methodological approaches, and their ability to explain the policy process.

Table 4 Assessing the ACF in the larger policy theory field
Full size table

Conclusion

Thirty years after the 1988 publication of “An advocacy coalition framework of policy change and the role of policy-oriented learning therein,” the fourth edition of Theories of the Policy Process was published. Destined to be a staple textbook for another generation of policy students in seminar classes across the world, the ACF chapter takes it place along other approaches. Its popularity is in part due to Sabatier and others developing a general theory of the policy process with attention to causality and hypothesis testing. Despite the modifications, the flow diagram and key hypotheses, the ACF’s core assumptions, remain largely unchanged. The goal of the next generation of ACF scholarship will involve more compelling and measurable causal explanations. Rather than emphasizing “why” things happen, which been the case for many ACF studies, future research will should measurable causal claim and to describe “how” things happen.