Designing stakeholder learning dialogues for effective global governance

ABSTRACT

A growing scholarship on multistakeholder learning dialogues suggests the importance of closely managing learning processes to help stakeholders anticipate which policies are likely to be effective. Much less work has focused on how to manage effective transnational multistakeholder learning dialogues, many of which aim to help address critical global environmental and social problems such as climate change or biodiversity loss. They face three central challenges. First, they rarely shape policies and behaviors directly, but work to ‘nudge’ or ‘tip the scales’ in domestic settings. Second, they run the risk of generating ‘compromise’ approaches incapable of ameliorating the original problem definition for which the dialogue was created. Third, they run the risk of being overly influenced, or captured, by powerful interests whose rationale for participating is to shift problem definitions or narrow instrument choices to those innocuous to their organizational or individual interests. Drawing on policy learning scholarship, we identify a six-stage learning process for anticipating effectiveness designed to minimize these risks while simultaneously fostering innovative approaches for meaningful and longlasting problem solving: Problem definition assessments; Problem framing; Developing coalition membership; Causal framework development; Scoping exercises; Knowledge institutionalization. We also identify six management techniques within each process for engaging transnational dialogues around problem solving. We show that doing so almost always requires anticipating multiple-step causal pathways through which influence of transnational and/or international actors and institutions might occur.

KEYWORDS:

• Multi-stakeholder dialogues
• policy learning
• transnational global governance
• pathways of influence

Introduction

One of the most important challenges facing students and practitioners of public policy in the global era is to identify and nurture policy instruments capable of achieving explicit and implicit policy goals (Capano & Pavan, 2019). Doing so, as the introductory article to this edited volume argues, requires designing policy instruments, including identification of innovative ‘policy mixes’ (Gunningham, Grabosky, & Sinclair, 1998; Howlett & Rayner, 2013) in ways that anticipate what is ‘expected to occur during the process of implementation’ (Capano & Pavan, 2019). Anticipation is important because it turns the attention of those designing policy interventions to identifying obstacles, to implementation and effectiveness before they present themselves. This allows designers to ‘pre-address’ anticipated problems by adjusting instrument design such that the anticipated negative feedbacks are averted before they would have occurred. This ‘[rediscovery] of a planning attitude…’ (Capano & Pavan, 2019) confronts much of the ‘experimentalist’ turn in the social sciences (Gerber & Green, 2000)¹ which draws to a particular type of scientific method in which adjusting instruments follows the collection of empirical data (King, Keohane, & Verba, 1994; Lindblom, 1959). The challenge with the latter approach is that not only is it mathematically impossible to run the thousands of policy experiments that could be unleashed, many critical global policy problems are in such crisis that solving them requires immediate solutions that do not have the luxury of being ameliorated through ‘trial and error’ approaches (Levin, Cashore, Bernstein, & Auld, 2012; Rittel & Webber, 1973). What is clear from the theoretical and empirical analyses in this special issue is that designing, and adjusting, policy instruments, including careful attention to innovative ‘policy mixes’ through forward looking anticipatory exercises, can help minimize the risk of running poorly conceived experiments, and increase the likelihood of achieving policy goals (Capano & Pavan, 2019).

There are many ways to anticipate effects, from scenario building exercises to ‘applied forward reasoning’ approaches (Bernstein, Lebow, Stein, & Weber, 2000). They have in common a requirement to theorize about, and project forward, the multiple causal impacts that a particular instrument might unleash over time (Levin et al., 2012). They also all require integrating policy effectiveness with ‘political effectiveness,’ that is, how the instrument can be designed to foster political legitimation and ‘consensual framing’ (Capano & Pavan, 2019) so that long-term societal support, or acceptance, can also be nurtured. For these reasons, there is increasing recognition that the involvement of stakeholders and other societal actors in the development and implementation of public policy interventions (Howlett, 2000) can help achieve desired results (Kekez, Howlett and Ramesh in press). Arguably no greater scientific attention has been placed on these questions than by those scholars who seek to understand the conditions through which stakeholder learning dialogues influence policy development (Sabatier & Jenkins-Smith, 1993), including their effects on producing marginal versus transformative policy change (Bennett & Howlett, 1992; Hall, 1993). The results have been promising, producing critical insights about how policy learning dialogues might be structured to help domestic public policy makers and practitioners identify strategies for nurturing durable goal-oriented impacts over time (Bryson, 2004; Dunlop & Radaelli, 2011; Heikkila & Gerlak, 2013; Radaelli, 2009; Rietig & Perkins, 2018; Sotirov, Blum, Storch, Selter, & Schraml, 2017).

At the same time, there exists an enduring and largely unresolved tension between advice on how to achieve political effectiveness (i.e. achieving a consensus among a range of stakeholders around what is politically feasible) as opposed to policy effectiveness (i.e. ameliorating the specific problem in question). It is also clear from extant literature that if stakeholder dialogues are to identify, anticipate and nurture policy innovations that contribute to the realization of the desired policy goals (Rietig, 2016), managing these tensions must be explicitly confronted.² Specifically, policy designers need to know how anticipatory multistakeholder policy learning dialogues might be structured such that policy effectiveness is reinforced, rather than compromised, by deliberations over political effectiveness. Similarly, we need to know how multi-stakeholder exercises focused on developing creative and collaborative solutions can be structured to reinforce original policy goals rather than have them undermined by powerful actors whose motivations for participating are to foster compromise in ways that reduces effectiveness.

The purpose of this paper is to reflect on these issues for globally initiated transnational learning dialogues, a form of multistakeholder collaboration that has been given much less attention by comparative and public policy scholars. Transnational learning dialogues have emerged to address almost every major problem facing the planet from species extinctions to biodiversity loss to climate change to the marginalization of traditional peoples and their cultures. Such dialogues have been initiated over the last 25 years among a range of international and non-state actors to address major societal challenges, especially in policy areas where processes and goals are most clearly influenced or affected by economic globalization, external actors and institutions (Bryson, Crosby, & Bloomberg, 2014; Cashore & Lupberger, 2016; Humphreys et al., 2017). The proliferation of these global dialogues and collaborative governance arrangements (Ansell & Gash, 2008) emerged owing to concerns about economic globalization’s obstacles in achieving policy results (Clapp & Fuchs, 2009; Rodrik, 1996) and an awareness of the opportunities that may exist to take advantage of economic linkages or benefit from external actors and resources, or to respond to such pressures in ways conducive to domestic policy goals (Moon, Crane, & Matten, 2005; Ruggie, 2002; Young et al., 2006).

We argue that if advances in such scholarship and practice are to continue, two related conundrums facing students of global affairs and governance must receive careful attention. First, most global or transnational interventions rarely shape domestic behavior on their own (Bernstein & Cashore, 2012), but rather work to ‘nudge’, ‘tip the scales’ or otherwise influence existing domestic policy making processes toward ameliorating particular types of challenges that link to broader global concerns, such as biodiversity loss or the marginalization of indigenous peoples. Second, means-oriented policy learning among stakeholders typically reverts toward compromise (Biermann, Chan, Mert, & Pattberg, 2007; Bryson, Crosby, & Stone, 2006) for two distinct reasons. First, many treat the quest for stakeholder consensus as synonymous with compromise, working to, ironically, undermine the policy effectiveness of the original problem definition. Second, businesses whose activities are often the source of the original problem, and other powerful organizations, often have a vested interest in engaging multistakeholder deliberations in ways that shift problem definitions or narrow solutions toward those that do not threaten their organizational or individual interests (Sotirov & Winkel, 2011).

We demonstrate, based on a review of relevant scholarship and building on our own collaborative research focused especially on forests and biodiversity policy (Cashore, Visseren-Hamakers et al., 2016; Humphreys et al., 2017), that six processes are key for fostering stakeholder learning in ways that help anticipate, and achieve, policy effectiveness or influence:

1. Problem definition assessments

2. Problem framing

3. Developing coalition membership

4. Causal framework development

5. Scoping exercises

6. Knowledge institutionalization

We argue that implementing specific management techniques within each of these processes increases the likelihood of global interventions ‘tipping the scales’ in domestic settings toward policy goals while avoiding being captured by powerful interests. We proceed in three steps. First, we critically assess the scholarship on international and transnational influences on domestic policy to argue for greater incorporation of domestic and comparative public policy literature. This allows to better inform policy design for improving goal attainment (Section II). Second, we identify management techniques under each of the six processes that, we argue, will improve the likelihood of goal attainment (Section III). We conclude by reflecting on the analytical, political and operational capacities (Howlett & Ramesh, 2017) required for promoting policy learning dialogues capable of enhancing ‘instrumentality’ (Capano & Pavan, 2019), i.e. the substantive effects of problem-focused governance innovations (Section IV).

Anticipating effectiveness of global governance in domestic settings

Over the last 20 years, International Relations scholars interested in governance and cooperation have dramatically transformed their research agendas from a focus primarily on formal international agreements or rule-based regimes to include greater attention to soft law, private standards and the other forms of market-based governance arrangements, goal-based governance and partnerships that increasingly populate the global governance landscape to fill real and perceived governance gaps (Abbott, Genschel, Snidal, & Zangl, 2015; Andonova, 2017; Auld, Bernstein, & Cashore, 2008; Bernstein & Hamish, 2017; Cashore, Auld, & Newsom, 2004; Kanie & Biermann, 2017; Kirton & Trebilcock, 2004; Pauwelyn, Wessel, & Wouters, 2014). Research on these diverse mechanisms and modes of global and transnational governance increasingly recognizes that they rarely directly solve problems, but instead operate through various pathways of influence (Bernstein & Cashore, 2012), via multiple causal processes, to affect or ‘nudge’ behaviors within domestic settings. The work on the effectiveness of these policy mechanisms has built on earlier literatures (Mitchell, 2006), including quantitative efforts that focus on developing ‘collectively optimal’ solutions that seek to enhance social welfare (Hovi, Sprinz, & Underdal, 2003), qualitative techniques that incorporate attention to social learning (Young, 2003, 2011) and historically dynamic efforts for addressing a specified environmental or social problem (Kütting, 2000).

This research has resulted in a much more sophisticated understanding of the ways in which powerful actors frequently work to develop, and constrain, global governance resulting in the prioritization of neoliberal agendas over environmental and social concerns (Cashore et al., 2004; Gale, 1998; Humphreys, 2006; Levy & Newell, 2005). It has also highlighted how non-governmental and international institutions might work to foster environmental and social stewardship (Bernstein & Cashore, 2012).

Less work, however, has focused on generating insights on the interactive effects of these international and transnational actors and processes and the types of causal processes that a range of transnational and domestic stakeholders interested in influencing domestic policy and behavior might trigger. There is thus a need to build on initial insights into such processes, such as the potential of domestic groups to “go global’ to create a ‘boomerang effect’ by lobbying a powerful state or other international actors to put pressure on an initially unresponsive state (Keck & Sikkink, 1998, p. 13). That earlier work also includes longstanding efforts to integrate International Relations, comparative politics and domestic public policy studies to better assess the conditions through which international and transnational actors and processes might influence the domestic public policy process (e.g. Acharya, 2004; Bernstein & Cashore, 2000, 2012; Risse-Kappen, 1995).

We argue that integrating insights from these broad scholarly projects can help design dialogues capable of enabling stakeholders to identify, and nurture, strategies for anticipating effectiveness. There is already strong practitioner interest, and application, in such a project, which has resulted in a range of international and non-governmental organizations and development agencies producing some type of ‘theory of change’ (Anderson, 2014; Elbers, 2014, p. 36) with which to justify a particular global intervention designed to shape domestic policies and behavior. However, we argue that these efforts may suffer from two shortcomings. First, the expected results are often temporary, or fail to materialize at all (Buntaine, Buch, & Parks, 2014; Gibson, Andersson, Ostrom, & Shivakumar, 2005). Second, and partly as a result of this limited durability, efforts to generate ideas by convening multi-stakeholder learning processes or dialogue platforms (Clémençon, 2012) often fail to incorporate insights from comparative public policy scholarship about the promise and pitfalls of stakeholder policy learning initiatives.³

To address these gaps, we posit that management changes can be made to six processes that structure learning dialogues to make them more durable and influential. Taken together, these improved processes are more likely to generate strategies consistent with complex ‘cause and effect’ relationships than a chosen instrument, or policy basket, might unleash. We elaborate the six processes below and the techniques to improve them, which require integrating global affairs research with insights on domestic policy learning scholarship (Rayner, Buck, & Katila, 2010).

Anticipating effective means-oriented global policy learning exercises: management insights

Problem definition assessments: deep, ‘whack-a-mole’

While it may seem obvious that learning protocols intended to address particular challenges must clearly identify the problem at hand, we argue that those promoting global interventions to address domestic practices often fail to do so for two reasons. First, goals can be so abstract that participants have different expectations about the type of ‘on the ground’ problem that the intervention is supposed to resolve. Second, there can be so many goals that cover so many problems that stakeholders frequently only agree that ‘everything matters,’ which undermines the very targeted exercise that problem-definition statements are meant to initiate.

The former problem characterizes the United Nations Forum on Forests in which ‘Sustainable Forest Management’ has come to mean balancing environmental, social and economic goals (Humphreys, 2001). However, just what objectives are meant to be emphasized, and how to ensure that goals are balanced, is hotly contested (Dimitrov, 2005). While some refer to this problem definition as a type of ‘constructive ambiguity’ that explains why diverse stakeholders agree to convene and deliberate (Singer & Giessen, 2017), we argue that these processes are better viewed as a type of ‘destructive ambiguity’ in which two ineffective trajectories are likely. If the process is maintained, participants will be focused on a never-ending debate about what problems to resolve. This debate will continue to play out when discussions turn to concrete attempts to develop and adjust policy instruments along the range of policy settings, and to calibrate potential tools. The result is that stakeholder discussions of causality (of problems and the effects of proposed solutions) will be stunted in favor of constant back referencing to disputed goals and organizational interests (Sotirov & Winkel, 2016). The second possible outcome is that if, and when, it becomes apparent to stakeholders that the process allows or encourages different conceptions of what problems need to be resolved, the process itself is likely to unravel (Skogstad & Wilder, 2018; Sotirov & Winkel, 2011). Failing to anticipate unraveling from destructive ambiguity is not only highly ineffective; it can also lead to the dismantled process being replaced by another ambiguous process constructed around a new policy instrument that simply reproduces the tensions inherent in the original problem definition. Such a pattern has characterized many areas of global environmental governance over the last 30 years, which have displayed a pattern of ‘commitment euphoria and implementation depression’ (Cashore, Auld, Bernstein, & Levin, 2016).

One recent example that risks this pattern is the United Nations’ flagship Sustainable Development Goals (SDGs) (Clark, Kerkhoff, & Gallopin, 2016; Kanie & Biermann, 2017). Collectively they develop a broad approach to problem definitions including poverty alleviation, biodiversity loss, gender and income inequality and economic growth (United Nations, 2015). While laudable on one level for identifying a wide range of public policy problems that confront governments all around the globe and the desire to promote integrative policies, they offer few tools for deliberating over potentially inverse relationships among goals, such as between, say maintaining biodiversity for Orangutan habitat (Voigt et al., 2018), which stands in opposition to other goals, such as rural economic development, especially when doing so relies on extractive sectors (Koh, 2008). Moreover, they also conflate ends-oriented concerns with means-oriented policy interventions.

The result is an implicit bias toward assuming synergistic outcomes and an overly sanguine perspective. This has meant less attention on the more arduous task of carefully disentangling instrumental decisions with their effects on substantive goals. This is problematic because, as Capano and Pavan (2019) point out, effective policies must ultimately solve the problem in question regardless of impacts on secondary, albeit important, instrumental efficiency goals that often bias dialogues toward consideration of market mechanisms, and equity goals that often end up prioritizing balance among competing interests. Effectiveness, therefore, should be defined in terms of ends (in particular, whether there is an improvement in the specified environmental, social or economic challenge at hand) rather than whether there was an (often short lived) agreement over the means of making policy. Thus, the bias that is always present when identifying a problem definition, which inevitably includes debates over values, must be explicit and clearly articulated at the beginning of the process in order to define clearly the ends pursued by the policy so that effectiveness is understood directly in relation to the problem.

Our point here is that if a problem-oriented approach is to achieve some degree of influence, the problem definition must be clearly identified, unpacked from other problems, and understood (Bryson, 2004, p. 23; Bryson, Crosby, & Stone, 2015, p. 653). Hence, the first requirement is to understand, and identify, the state of knowledge regarding the problem itself. For biodiversity loss, for example, this means assessing the degree of land use change, habitat degradation and types of endangered or threatened species at risk. Problem definitions focused on people’s livelihoods require an assessment about how local resource dependent communities engage with the land resources around them as well as an awareness of historical changes, such as extractive industrialization processes. Importantly, and despite well-intended biases among environmental and social activists, it should never be assumed that addressing biodiversity loss, climate change and local livelihoods are synergistic (Larson, Brockhaus, Sunderlin et al. 2013; Stevens, Winterbottom, Springer, & Reytar, 2014). Implicit biases about problem definition synergies, often exacerbated by funding opportunities, will lead to undesirable consequences as the empirics unfold, undermining policy effectiveness in the long run.

For these reasons the problem definition phase requires explicit and honest attention to the ways in which a focus on one problem might lead to ‘whack-a-mole’⁴ inverse impacts, that is, where addressing one problem may generate or exacerbate other problems. These impacts are often diagnosed after an intervention. To take just one illustrative example, Clark et al. (2016) found that while biofuels did, as a means-based solution, seem to lower carbon emissions in sugar cane-sourced fuels in Brazil, they also led to higher corn-based carbon emissions in the United States, and heightened food insecurity through higher food prices overall. Hence, we expect that accurate diagnosis of the problem, and attention to synergies and countervailing effects before a policy experiment is unleashed, make it more likely that smarter experiments can be chosen (Visseren-Hamakers, 2018). Doing so makes it possible to minimize undesirable impacts by considering creative ideas to foster synergistic outcomes or recognize, and anticipate, negative countervailing impacts. Such anticipatory questions, and resultant lessons drawing (Illical & Harrison, 2007), before rather than after running a poorly conceived ‘experiment’, can save significant time and scarce resources. It encourages those involved in the learning dialogue to reason through anticipated whack-a-mole effects to recognize that the dialogue itself, without a clear problem definition, is not likely to have the impacts for which it has been created. We acknowledge that these deliberations can never be a purely objective exercise since knowledge is always constructed (Jasanoff, 1987; Jasanoff & Wynne, 1998) and experts themselves can exercise their own preferences that also have implicit biases (Rietig, 2014). Our point is that learning exercises around problem definitions should incorporate research on how different methodological and epistemological orientations shape social understandings of global challenges (Clapp & Dauvergne, 2005). Thus, our first proposition is as follows:

P1: When learning processes initiate deep-dive problem definition deliberations, including attention to ‘whack-a-mole effects’, global policy interventions are more likely to effectively influence domestic policies and practices.

Problem framing

In addition to deep dives about the nature of the problem in question, stakeholder learning dialogues will need to structure learning deliberations that help participants decide how to anticipate, and adjudicate, multiple synergistic and countervailing effects behind whatever course of action they choose. In this regard it is essential that policy learning dialogues are structured to render implicit biases in problem framing explicit (Bryson, 2004, p. 37; Sotirov, Blum et al., 2017). Doing so will make explicit deliberations about what type of (biased) framing is most appropriate for the problem definition at hand (Bryson et al., 2014, p. 450). In this regard Cashore and Bernstein (2018) confront assumptions of much of the collaborative learning scholarship (Bryson et al., 2015) by arguing that, depending on the nature of the problem and how it is framed or conceived, lessons for designing learning dialogues will differ, often dramatically. Without recognizing these differences, attempts to generalize across all problem types may, inadvertently, prioritize some conceptions of the problem over others. Cashore and Bernstein highlight how four different problem conceptions implicitly shape, and limit, the ways in which policy makers, stakeholders and applied policy scholars think about instrument choice⁵:

• Type 1: win/win, such as those targeting a resource depletion ‘tragedy of the commons’ in which almost everyone is better off through avoiding catastrophic depletion of economically valuable resources and cooperative solutions where all can gain are assumed;

• Type 2: win/lose optimization, often applied in neo-classical cost-benefit analysis that contrasts a problem according to whether addressing it improves social welfare while recognizing trade-offs;

• Type 3: win/lose compromise, in which a problem is seen as requiring compromise with others, such as the need for compromise among environmental, social and economic goals; and

• Type 4: win/lose prioritization, where some problem is deemed so important, or fundamental, that addressing it trumps other problems, because a failure to address it is catastrophic, irreversible and/or violates a fundamental value. Such ‘win/lose’ conceptions could be derived from careful attention to the problem at hand where the scientific body of research is clear that some problems, such as endangered species or climate change, must be given type 4 status if they are to be solved. They could also be derived from norms, such as today’s widespread normative opposition to slavery. Such norms are held so strongly that the idea of compromise (Type 3), or adjudicating their benefits based on their effects on overall utility (Type 2), is deemed to be inappropriate and unethical.

Rendering these implicit biases explicit has practical utility. It can help managers focus learning among government officials and stakeholders about whether, say, a problem such as saving endangered species from extinction owing to biodiversity loss is to be treated as an absolute priority or whether some species loss is acceptable if evidence is clear that protection may hamper livelihoods of local people. Rendering these philosophical choices explicit, rather than keeping them hidden in the guise of Type 2 ‘rational’ optimization strategies, can be expected to expand the range of policy options under consideration. Even more importantly, it can help stakeholders identify the complex causal processes that such interventions might unleash as well as the implications of these processes for ameliorating the problem.

While this point might seem obvious, much of the work designing for effectiveness limits choices based on often highly constructed ‘feasibility’ assessments that narrow instrument choice in ways that implicitly undermine the original problem definition (Kütting, 2000). Similarly, if rendering these implicit biases explicit results in disagreement among stakeholders about whether to apply optimization (Type 2), compromise (Type 3) or prioritization (Type 4), then it will become clear at the start of the process that means-oriented learning processes are unlikely to produce significant positive effects since the issue is not a gap between instrument choice and ends, but differences in problem definition. The logical strategic conclusion under such conditions is simply not to engage in learning dialogues at all since implicit and explicit debates about how to frame the problem will hamper efforts to generate learning about the causal mechanisms around the problem in question.

Recognition that learning dialogues can not only create new opportunities, but also reinforce some interests and biases, can assist with international processes designed to promote a range of problem-oriented solutions. Certainly those championing SDGs will benefit from drawing a dialogue’s attention to the ways in which various ideas might influence, positively and negatively, the range of goals for which these were created (Clémençon, 2012; Kanie & Biermann, 2017). For example, any deliberation over a particular intervention with which to address, say deforestation, will need to be structured in a way that allows for careful assessment of a suite, or combinations, of potential instruments, from market-based eco-labeling to zero net deforestation commitments along the supply chain to traditional protected areas policies. Such efforts would need to assess both the immediate causes of deforestation, which would point analysts to the role of palm oil, soy, beef and wood products in shaping deforestation (TFA, 2018), but also broader patterns of consumption, including diet, and population growth that, though often ignored, are fundamentally important for long term effectiveness. Likewise, such efforts would also need to identify what types of issues are highlighted by a focus on deforestation as opposed to, say, questions of animal rights (Visseren-Hamakers, 2018). And, importantly, such efforts would need to recognize the role of competing scholarly methods, epistemologies and ontologies for illuminating, but also obfuscating and prioritizing, some knowledge over others (Grabs, Auld and Cashore Under Review 2019).

Such an effort would then pave the way for anticipating what types of policy instruments, including the range of policy mixes available to those designing interventions show the most promise in addressing the specific problem definition in question, alongside assessments of both synergistic and countervailing effects of policy choices (Climate Focus 2017). Such an effort would be critical for assessing policy interventions aimed at biodiversity and forest ecosystems, such as SDG 15, with those that call for ending hunger (SDG 2) and expanding export markets (SDG 17). These challenges show up in almost every transnational learning dialogue effort. For example, the NGO The Forests Dialogue, which brings together businesses, environmental groups and other stakeholders, is designed to collaborate on problems as important as indigenous rights and biodiversity loss (The Forests Dialogue, 2018). Its orientation is to discover consensus that prioritizes balancing various interests (The Forests Dialogue, 2008), including those firms and companies whose extractive sectors have often caused the problems they are collaborating to ameliorate (Canby, 2005; Myllynen, 2005; Nussbaum & Simula, 2004). Significant operational and core funding comes from firms either directly or through association of firms such as the World Business Council on Sustainable Development. While these consensus-focused approaches are understandable, their design appears to implicitly reinforce Type 3 compromise solutions, while, at the same time, undermining Type 4 considerations – especially those cases in which scientific research points problem solving in directions that would need to impose costly regulations on participating firms.

Hence, attention to these questions will help dialogues to carefully deliberate over whether they want to give Type 1, 2, 3 or 4 problem conception status to their deliberations. We expect that doing so is important not only to advance transparency in its own right, but also to create the conditions through which problem-focused innovations might emerge in ways that mitigate against subtly and tragically shifting problem conceptions.⁶ Hence our second two-part proposition is as follows:

P2a: The more problem framing biases are made explicit, and deliberated over, the more likely stakeholder learning processes can focus on innovative and appropriate triggers for ameliorating the problem in question rather than constantly debating the empirical manifestations of what success might mean.⁷

P2b: Explicit recognition of Type 4 ‘prioritization’, when appropriate, makes it less likely that powerful interests seeking to shift toward Type 3 compromise or Type 2 optimization conceptions will capture the process.

Developing coalition membership

A third factor, which directly contradicts contemporary expectations of what kind of deliberations produce effective and legitimate global governance, is that the initial (core) learning coalitions ought to be narrow rather than broad. Emphasizing a narrow learning coalition is arguably a heretical point to make, even if supported by extant scholarship, because it confronts existing norms about ‘multi-stakeholder’ participation that underpins the United Nations’ ‘major working groups’ protocol for inclusion and growing norms around ‘multi-stakeholderism’ in global governance more broadly (Pauwelyn et al., 2014; Pülzl & Rametsteiner, 2002; Raymond and DeNardis 2015). The reason for beginning narrowly is to avoid the implicit and institutionally bias of multi-stakeholder dialogues. Implicit biases to compromise almost always occur when a range of actors come to the table with competing problem definitions, and rationales for doing so.⁸ Compromise biases can also be institutionalized formally. An example is the tripartite approach to governance in which environmental, social and economic interests have equal weight, which is manifest in non-state market driven (NSMD) global governance such as the Forest Stewardship Council (Gale & Cadman, 2013). While many scholars and practitioners have identified the ways in which these institutionalized forms of collaborative governance can reinforce some aspects of legitimacy, less conceptual, theoretical and empirical attention has been placed on their role in downplaying Type 4 problems over Type 3. In fact, ‘sustainability science’ (Kates et al., 2001) knowledge generation can reinforce these biases by treating compromise among social, environmental, and economic interests as synonymous with scientific, ‘neutral’ and ‘objective’ (Cashore, 2018) analysis, when the implicit results is to reinforce broad-based stakeholder engagement over Type 3, rather than Type 2, orientations (Cashore, 2018).

While laudable for attempting to generate consensus, the assumption that all problems are amenable to compromises among economic, social and environmental values bypasses a thorny challenge: longstanding research within public policy studies reveals that it is precisely processes that give equal weight to all interests that, in practice, allow powerful interests to shift problem definitions away from the original purpose of the dialogue or, even more broadly, to shift the purpose of the institution itself, often through covert, behind the scenes maneuvering (Bachrach & Baratz, 1962). Such outcomes need not be the direct result of strategic action, but may occur regardless owing to latent power that shapes the perceptions and cognition of actors so they accept an existing order as natural (Lukes, 1974).

To avoid such capture, scholarship on means-oriented multi-stakeholder policy learning processes has demonstrated that attention must be placed on developing coalitions of ‘likeminded’ organizations (Hall, 1993) and individuals with shared values (Sabatier & Jenkins-Smith, 1993) who are united over addressing a particular problem. Such core communities are key for understanding means-oriented learning because, given they agree on the need to ameliorate a specified problem, their attention is focused primarily on better aligning policy goals with instrument choice and design. Although there is no single perfect solution to identifying the membership of a core stakeholder group, efforts must be made that are sensitive to, and foster, ‘shared learning’ among key actors around the causal pathways that a particular policy instrument, and constellation of particular policy component mix, might help unleash. Establishing this initial group is key since its membership will then deliberate over whether, when and how broader coalitions among organizations and individuals might be useful for instrumental reasons, rather than resulting in shifting goals that by definition, undermines effectiveness of the original problem orientation. In the case of biodiversity conservation, for example, this might include groups that share environmental and social concerns as well as likeminded business interests who seek to promote corporate social responsibilities.

The commitment to shared problem-definition within the initial coalition is crucial since the purpose of global governance interventions is to ‘tip the scales’ domestically, where typically an array of governmental agencies and non-state actors are simultaneously advocating a range of often competing problem definitions. While recognizing different interests will be key for functional reasons, including instrument choice and political support, allowing them to join the learning coalition as equal partners runs the risk of shifting problem definitions rather than learning how to address them. Our argument thus adds an important caveat to previous work on advocacy coalitions and cross-coalitional learning. While we agree with Sabatier and Pelkey (1987) and Bryson et al. (2015) that cross-coalition and cross-sector learning can occur when stakeholders see potential in a policy instrument addressing their respective values, we emphasize that the initial learning must occur within a core coalition with shared problem definition since premature searches for ‘win-win’ solutions among those with competing problem definitions too often move policy responses away from the original policy goal. Only once learning on means within that initial coalition occurs should cross-coalitional learning emerge that is aimed at building greater support, and awareness of, the causal processes at play through which a particular instrument design might unleash effective pathways. Similarly, while ‘Bootleggers and Baptist’ coalitions—such as those of businesses who see market advantages over competitors in accepting higher regulation aligning with environmental groups (e.g. Vogel, 1995)—may eventually form, they are ill advised in early stakeholder dialogues for similar reasons (Auld et al., 2008).⁹

Indeed, the startling emphasis on market mechanisms as a way to address environmental destruction by a range of ‘bootlegger and Baptist’ coalitions globally exemplifies this undermining trend, especially when one considers that by almost every measure, global market mechanisms have failed to ameliorate biodiversity loss and species endangerment (van der Ven, Rothacker, & Cashore, 2018).¹⁰ This means that those advocating ‘multi-stake stakeholder dialogues’ must be careful to do so only following a clearly identified problem or set of problems, such biodiversity loss or maintaining forest dependent people’s cultural integrity. Failure to do so runs the risk of powerful interests watering down the problem definition, rather than offering creative solutions for solving the problem. Hence our third proposition is:

P3: When initial coalitions are limited to those whose interests or values are already aligned with the problem definition, the more likely it is that global interventions may positively influence a specified on-the-ground problem.

Causal framework development

The literature on international and transnational influences on domestic policies identifies a variety of causal pathways through which external actors and institutions can help address domestic policy challenges. Problem-focused learning coalitions can apply frameworks that analyze these processes to guide their learning about what might work or not in addressing the policy problems they organize around. While there are many frameworks that could be applied, the extant literature suggests that ideally, they should allow learning processes to engage in three related issues: conceptualizing ‘durable change’ processes; disentangling different types of policy elements through which creative solutions might be identified; and identifying the range of causal pathways of influence through which change is promoted.

Durable change

Drawing on decades of research on policy change, Cashore and Howlett (2007) and Howlett and Cashore (2009) have identified four different change processes that, initially, might appear the same or be misdiagnosed (Table 1). Two are in ‘equilibrium’ and hence can be seen largely as reinforcing the status quo. They are classic incrementalism, in which only marginal change is possible, and faux paradigmatic, where initial signs of large change instead revert to the previous equilibrium. In contrast, classic paradigmatic change represents those (rare) cases in which a single large step creates a new equilibrium. While useful, Cashore and Howlett (2007) argue that whereas most learning processes focus on promoting classic paradigmatic change, a more promising and realistic focus for creative solutions and mechanisms is progressive incrementalism in which a new equilibrium emerges through several smaller, evolutionary steps. Given the rarity of single shot classic paradigmatic shifts, learning processes that focus attention on multiple steps for influencing policy goals are more likely to prove fruitful in anticipating results over time, and identifying forward looking strategies for nurturing them.

Table 1. Policy change and durability: four models of policy change.

Source: Cashore and Howlett (2007), adapted from Durrant and Diehl (1989).

Disentangling policy elements

A second critical component to developing learning frameworks about causation is to encourage stakeholders to identify, and disentangle ends-based discussions concerning overall goals, such as environmental degradation or economic development, that target stakeholders to focus on a precise concern and the specific content of policies; and means-based discussions that render explicit the normative framework through which different policy tools are identified, and choices about the literally thousands of ways to calibrate particular tools in the hope of achieving desired results. Doing so requires distinguishing means from ends. Such an orientation is important for stakeholders to render explicit, and benefit from, previous processes around problem definition and problem framing discussed above. While the ultimate goal is to influence ends, once a mechanism is created, such as, say, ‘eco-labeling’ or ‘no-deforestation’ commitments by purchasers of commodities who have historically caused deforestation, stakeholders often inadvertently end up promoting the instrument regardless of the results ‘on the ground.’ This shift from ends to means can be so significant that stakeholders maintain support even when the empirical evidence is overwhelming that it is having either little, or an inverse, effect on the original problem definition (van der Ven et al., 2018).

Second, policy research has long recognized (e.g. Hall, 1993; Sabatier & Jenkins-Smith, 1993) that different types of policy elements, from instruments to settings to mechanisms, change for different reasons, with some being easier to change than others. Hence, distinguishing types of policy elements allows stakeholder learning on identifying ‘easy to pull’, but ‘hard to reverse’, levers to address specified problems (Levin et al., 2012). For these reasons, Cashore and Howlett (2007) and Howlett and Cashore (2009) expanded the analyses of Hall and Sabatier from identifying three to six elements of policy (Table 2). Doing so challenged Hall’s expectation that only ‘societal learning’ can lead to transformative change. In some cases, ‘easy to pull’ policy settings or instrument ‘calibrations’ can also lead to transformative change through progressive incremental change. While often overlooked in much of the policy literature, these tinkering efforts can, evidence suggests, help address some of the most thorny questions of our times, including Type 4 challenges such as climate change (Levin et al., 2012). Hence, concerted efforts must be made to understand how the six elements of policy levels interact, which are easiest to change, and which might cause, through a series of steps, meaningful influence.

Table 2. Elements of a policy.

Source: Cashore and Howlett (2007, p. 536) and adaptations since then.¹¹

¹¹

Elements of policy

Pathways of influence

Finally, some type of framework with which to orient stakeholder policy learning around complex cause and effect relationships is necessary for success, especially when the impact is likely to occur through multiple, progressive incremental steps that draw on and distinguish all six policy elements. This is important because, as in the case of global interventions in general, and market mechanisms in particular, forward looking trajectories may not unfold in linear paths, but may take ‘U curve’ trajectories and may influence behavior indirectly (Cashore, 2016; Cashore, Matas and others 2012). For example, Cashore and Stone (2014) show that legality verification along global supply chains requires relatively modest policy settings so that efforts can be focused on building supply chain tracking tools without imposing undue costs on producers. However, they argue that once legality verification becomes routinized as standard operating procedures and shirking is no longer possible, the substantive requirements (i.e. the policy settings) can be revised upwards without risking support of legal producers because at this point, consumers would pay through higher prices rather than have costs borne by individual firms (Cashore & Stone, 2012). The point here is that to realize effectiveness, stakeholder learning must be focused on generating consensuses, and understanding, about ‘instrument choice to effectiveness’ pathway on which they are embarked, which will almost always entail traveling a series of unique steps, each of which have their own temporal strategic implications, but which are connected through historical trajectories that must also be unleashed.

Bernstein and Cashore’s (2000, 2012) ‘pathways of influence’ framework offers one such way to guide these learning processes around the multiple causal pathways that global interventions might unleash. Their approach focuses stakeholder attention not only toward identifying the types of policy mixes, or configurations, that Cashore and Howlett’s six elements help identify,¹² but also target stakeholder deliberations around anticipating the multiple simultaneous influences that any particular tool, including calibrations, might have in championing, or undermining, broad goals, concrete objectives, and precise substantive requirements. In other words, their approach offers one way for stakeholders to deliberate over, and anticipate, what might be expected to occur before they experiment with a particular instrument that might be the ‘flavour du jour’, but which might be less than optimal for long-term effectiveness. Hence, their pathways framework allows for deliberations that could very well lead to changes in settings and calibrations because, following careful causal projections forward they are more likely to enhance effectiveness as well as and to anticipate and avoid negative feedbacks (Humphreys et al., 2017).

Albeit, Bernstein and Cashore’s framework does not tell those involved with means-oriented learning processes what particular intervention is most useful at a particular time. Rather, it provides a framework for identifying questions policy makers will want to ask, and the overall trends to consider, when deliberating over specific interventions and strategies to nurture them. Our point here is not that learning dialogues must draw on this particular framework, but rather they must guide stakeholders around some type of learning about complex ‘cause and effect’ relationships that exist independently of individual or organizational material interests.

Markets pathway

Bernstein and Cashore identify a markets pathway useful for focusing stakeholder deliberations around the causal processes that might be unleashed through some type of market incentive or disincentive (Bernstein & Cashore, 2000, p. 76–78, 2012, p. 9). Several strategic insights have emerged from assessing specific interventions under the markets pathway. For example, research has found that while boycotts as a policy are often useful to generate support for the problem definition in question (Sasser, Prakash, Cashore, & Auld, 2006), they tend to be short lived unless they are matched by other market instruments, such as certification, that contain wide-ranging policy settings and that can foster significant learning across a range of stakeholders across countries (Elliott, 2005). But even here, Bernstein and Cashore (2007) acknowledge that this market instrument can only result in an effective pathway if stakeholders work to foster calibrations that reinforce institutionalization of support across global value chains. Doing so often requires adjusting substantive standards as support improves. Standards are initially set rather modestly to avoid negative feedbacks in which firms would vacate support, but can in progressive incremental steps have substantive requirements (i.e. policy settings) ‘ratcheted up’ in ways that reward, rather than penalize, participating firms as broader-uptake occurs along the supply chain (Cashore, Auld, Bernstein, & McDermott, 2007).

The point here is that careful attention to the markets pathways provides one way for stakeholders to think about, and project forward, not only the strength of the market tool at any one time, but also the historical trajectories it could be expected to unleash. In the case of certification systems, this approach allows stakeholders to think about the multiple steps through which institutionalization occurs and to identify strategies, in advance, for navigating the progressive incremental increases in standards (i.e. policy settings) in ways that might be expected to plausibly foster durable impacts and long-term effectiveness (Cashore, 2016). Likewise, identification of three other pathways elaborated below points stakeholders to deliberate not only on mixes among Cashore and Howlett’s six elements, but also around what constellation of pathways might be traveled that are potentially synergistic (Cashore, Elliott, Pohnan, Stone, & Jodoin, 2015), while avoiding those that could result in negative feedback processes.

Rules pathway

The rules pathway focuses stakeholder attention on the current, and potential future, role of binding agreements in shaping domestic policies over some preconceived problem definition such as reducing biodiversity loss (Bernstein & Cashore, 2000, p. 78–80, 2012, p. 6–7). While binding international agreements are currently out of fashion, deliberations over this pathway can be important for not only identifying innovative ideas that could lead to new binding agreements but also to finding existing rules that might be drawn on to link synergistically with other pathways. For example, identification of this pathway-focused learning not only allows directing policy goals and the objectives of formalized multilateral agreements (MAs) around the problem definition in question, but also to other treaties that work to influence the problem definition indirectly. This broader effort to identify existing or new policy mixes is important and empirically justified. Efforts to promote environmental goals and related objectives such as species preservation and rural livelihoods in a (failed) global forest convention in 1992 still found traction in related treaties such as the Convention on Biological Diversity (CBD) or the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES). Likewise, EU and US policy mechanisms aimed at curbing imports of illegal timber by promoting domestic policy settings that foster legality compliance in producer countries have sent multiple ripple effects across global supply chains (Leipold, Sotirov, Frei, & Winkel, 2016; Sotirov, Stelter, & Winkel, 2017). Similarly, trade agreements are a particular type of policy mechanism that can contain specific policy requirements around forest practices (i.e. policy settings) such as were included in the US-Peru Free Trade Agreement (Jinnah & Lindsay, 2016).

Norms pathway

The norms pathway focuses attention on the role of deeply engrained values and ideas about appropriate behaviors that often precede any self-interested calculations (Bernstein & Cashore, 2000, p. 80–83, 2012, p. 7–8) and can shape identification of policy goals, objectives, and intervention logics identified in Table 1. It also focuses attention to how these norms might be changed, or generated. For example, deliberations over ‘high conservation value forests’ (HCVF) that occurred first through Forest Stewardship Council (FSC) multi-stakeholder tripartite deliberations about how to foster appropriate forest management in sensitive ecosystems, ended up generating a ‘high conservation value’ discourse generally that now permeates almost all stakeholder deliberations over resources management (Hamzah, Nik, Efransjah et al. 2007).

This pathway is important for stakeholder deliberations as it points their collaborative learning to the role of norms as generating instrument logics about the prevailing cognitive framing of the appropriateness of a range of policy instruments. For instance, Bernstein (2001) has found that norms of liberal environmentalism have given primacy to market tools over other policy interventions, regardless of their merits in addressing global environmental challenges. These, in turn, often trumped the development of stringent policy requirements (i.e. the policy settings in Table 2) that might have been more effective in addressing a range of environmental problems. Learning about the role of norms in creating instrument logics allows stakeholders to deliberate over tools that might fit these norms but still have a chance to develop meaningful policy settings over time, or, conversely, to recognize that these norms must be changed or challenged if the problem definition in question is to be effectively addressed. This norms pathway also focuses deliberations of stakeholders around whether the current narrowing of policy tools that reinforce market or neoliberal norms, is not owing to their problem solving superiority, but rather to the dominance of a particular intervention logic that serves to direct, and constrain, policy instrument options (Bernstein, 2001; Jacques & Lobo, 2001). Doing so, might lead them to not simply take these norms as granted, but to reflect on how their very own deliberations might shift norms in ways that open up the choice of tools available, and the types of calibrations that are employed.

Likewise, stakeholders can learn about powerful norms such as forest livelihoods, indigenous rights, and ‘subsidiarity’ as key objectives that explain why governments promoting payments for REDD+ must include, rather than bypass, non-carbon benefits such as the livelihoods of local peoples. Hence, these seem particularly useful for creating normative ‘pulls’ that might help ‘tip the scales’ toward identified problems such as promoting biodiversity conservation, indigenous rights and sustainable community forestry.¹³ Bernstein and Cashore’s articulation of the norms pathway also points stakeholder learning to deliberate over distinguishing norms that have little explanatory power (i.e. they rarely trump pragmatic or material interests) from those that are durable in the face of countervailing incentives.

Direct access pathway

Finally, the ‘direct access’ pathway focuses stakeholder learning deliberations on the role of non-domestic influences in shaping domestic level policy through capacity building, technology transfer or other resources that might be able to alter domestic power dynamics among differing interests and sectoral level policy networks (Bernstein & Cashore, 2000, p. 83–85, 2012, p. 9–10) in ways that prioritize the problem definition in question. Traveling this pathway requires that relevant stakeholders learn about, are aware of, and work to avoid, the risk that efforts are viewed as (inappropriate) foreign or international intrusions. Learning deliberations must also focus on understanding the structure of, and navigating, existing domestic public policy subsystems through which government agencies engage domestic stakeholders in general (Bernstein & Cashore, 2000, p. 83–85, 2012, p. 9–10), the impact of these networks on domestic choices about relevant tools, calibrations and settings, and their capacity to advance policy decisions once made (Risse, 2011). Hence our fourth proposition is:

P4: A prerequisite for effective means-oriented learning requires the identification of analytical frameworks through which stakeholders and policy makers can identify the causal processes behind specific global interventions, as well as the associated strategic implications for unleashing these effects. For success, such frameworks must include attention to deliberations of policy durability, multiple-step change processes and multiple pathways.

Scoping exercises

A fifth key management technique that seems to correlate with global learning efforts aimed at identifying and designing anticipatory policy interventions, tools and calibrations best suited to achieve identified objectives, requires undertaking some type ‘scoping’ exercise with which to assess whether existing instruments already exist that are amenable to changes, or whether new tools need to be identified.

To foster this thinking, Table 3 identifies four overall scoping strategies: survey existing global interventions; create new and innovative options; or develop modestly through progressive incremental logics, or rapidly through comprehensive paradigmatic interventions, ‘grafting’ into existing tools that do not currently or explicitly recognize the identified policy goal and objectives in question. These are important distinctions for fostering insights based on both past experience and for generating creative new ideas for moving forward. For example, relatively modest changes might be identified that stakeholders aim to expand on in progressive yet incremental steps. Indeed, we argue that such a scoping exercise, though rarely applied today, offers considerable promise for triggering durable change processes (Hacker, 1998) compared to attempts to create large scale ‘single shot’ approaches (Levin et al., 2012). Historical efforts within political science and policy studies show that what originally appeared to be rather minor changes in policy calibrations or policy settings can ultimately yield very significant impacts on a range of social and environmental challenges, positive or negative (Levin et al., 2012). Literally thousands of creative ideas regarding instrument design and effects can be conceptualized from this orientation.

Table 3. Approaches to intervening for the problem at hand.

		Existing (Graft) An existing instrument to problem definition in question	New (Create) Draw on new instrument designed to address problem definition in question
Starting point	Moderate	Scope existing interventions that already have support	Champion the problem directly, but modestly
	Comprehensive	Target an existing effort that has salience globally	Champion comprehensive global approach to address the problem at hand

Source: Adapted from Cashore et al. (2016).

For example Humphreys et al. (2017), Cashore et al. (2016)’s application of this step during a multi-stakeholder learning dialogue with indigenous and environmental stakeholder groups in Peru¹⁴ projected forward that by moderate grafting onto the existing global policy tool of legality verification by creating, and training, auditors about how to undertake legal auditing for community forestry, strategists could be expected to, progressive incrementally, do more to enhance local indigenous rights, than abstract formal agreements on indigenous rights that may never be enforced. This exercise also allowed them to identify ‘comprehensive grafting’ as another possible strategy drawing on existing attempts to embed indigenous rights within transnational REDD+ collaborations. However, they also raised flags about the ‘on the ground’ effects of such strategies, since many efforts have not moved beyond written recognition in formal agreements (Visseren-Hamakers, Gupta, Herold, Pena-Claros, & Vijge, 2012; Visseren-Hamakers, McDermott, Vijge, & Cashore, 2012; Visseren-Hamakers & Verkooijen, 2013).

Based on these insights our fifth proposition is:

P5: Mean-oriented policy learning processes initiated at the global level are more likely to influence problems on the ground when stakeholders engage in a systematic scoping exercise of existing interventions and new creative ideas for intervening to ameliorate one or more ‘on the ground’ problems.

Knowledge institutionalization

The sixth key process, which also may seem obvious but is rarely applied, is that there must be some way to institutionalize insights and knowledge that emerged from the core stakeholder policy learning deliberations. Because individuals come and go and memories are faulty, participants will need a ‘playbook’ with which to guide specific decisions once the initial step is undertaken. Failure to do so may explain why some well-intended learning processes failed to produce enduring outcomes. For example, extant research has found that generating progressive incremental support for global certification systems requires nurturing a ‘chicken and egg’ back and forth between supply and demand such that substantive requirements of policy are in line with existing market incentives (Auld & Cashore, 2012; Cashore et al., 2007). The result is a multiple step process in which ultimate impacts on policy objectives will be realized following, rather than preceding, widespread global uptake. Accordingly, multistakeholder learning dialogues must, at Time 1, identify modest policy settings and calibrations in Table such that firms evaluate them creating more benefits than costs. However, at Time 2, once, and if, the market has institutionalized support, a plan needs to be in place allowing for increased standards in line with enhanced market benefits. Such an approach requires carefully documenting just what type of impact is anticipated at each step, and to then identifying strategies consistent with them.

Failure to recognize these temporal logics may explain why environmental groups raise the Forest Stewardship Council’s policy settings in British Columbia (Cashore et al., 2004) and the Canadian Maritimes (Cashore & Lawson, 2003) which caused many participating firms, who were already operating under relatively high government forest practice regulations (Cashore & Auld, 2003; McDermott, Cashore, & Kanowski, 2010), to vacate support for the FSC. A longer term focused strategic playbook might have allowed NGOs to maintain instrumentalist alliances with firms aimed at collaboratively enhancing market uptake for the FSC. This in turn might have permitted incrementally progressive adjustments in standards (policy settings) consistent with maintaining firm level support. Similarly, a playbook seems also critical for anticipating, and avoiding, ‘Time 1’ modest standards (i.e. policy settings) that, owing to powerful actors or lack of market uptake, get ‘stuck’ in classic incremental ways that are not in line with addressing the original problem.

Although there is evidence stakeholders dialogues can indeed learn about these historical dynamics informally (Auld, 2006), we argue that absent some type of formalized playbook, this learning will be limited or short lived. Whatever the specific format, such a playbook could especially focus on identifying, and travelling multiple steps across two or more synergistic pathways while also helping uncover and avoid countervailing strategies. Hence our sixth proposition is:

P6. The development, and application, of a playbook that identifies multiple steps through which an instrument is nurtured, increases the likelihood that policy learning processes will influence domestic policies and behaviors.

Conclusions: lessons for analytical, political and operational expertise

Policy learning scholarship emerged and remains in the realm of domestic and comparative public policy studies. Yet, insights from this literature are highly useful for addressing what many at the global level feel is frustration with the slow pace and small scale of positive change. Our review reveals two broad conclusions about the analytical, political and operational capacities (Howlett & Ramesh, 2017) needed to foster means-oriented policy learning around global interventions (Kekez et al., in press).

First, concerted and proactive efforts are needed to overcome existing stasis. This is crucial due to the paucity of structured learning processes at the global level.¹⁵ This means that some type of policy entrepreneur (Faling, Biesbroek, & Karlsson-Vinkhuyzen, 2018; Faling, Biesbroek, Karlsson-Vinkhuyzen, & Termeer, 2018; Kingdon, 1995) capable of knowledge brokering (Bernstein, Cashore, & Rayner, 2015; Bull et al., 2018; Cashore, Bernstein, & Rayner, 2015; Cashore & Lupberger, 2015, 2016; Cashore, Visseren-Hamakers et al., 2016; Humphreys et al., 2017) is needed to foster such change.

Second, and relatedly, our review raises key themes for the types of skills such an entrepreneur would be required to have. At the analytical level, it is quite clear that the policy entrepreneur would require a background broadly based in the social sciences and have a strong awareness of both the descriptive aspects of global and domestic affairs, as well as the ability to develop propositions to explain the past. The policy entrepreneur should also be able to propose creative ideas for enabling anticipatory policy design that incorporates projecting forward multiple causal steps. Such analytical skills require a deep appreciation of historical evidence and data, but not to be locked in by them. This is particularly important as the policy world can only run a handful of the almost infinite number of possible ‘experiments’. Uncovering potentially more influential policy instruments therefore means having the ability to think conceptually and theoretically in a manner that fosters ‘real world’ impacts. It also requires having the political skills necessary to build a tightly focused learning dialogue that can take on board this framework first before thinking about building across coalitions, and to build strategies for doing so out of the process—but not as a replacement for this process.

Likewise the entrepreneur must have the skills to engage transnational networks, but also to examine the domestic historical context and policy setting through which global efforts might improve local problems (Bull et al., 2018). Finally, the entrepreneur must possess the necessary operational skills for managing processes that include diverse organizations. This requires not only adequate financial resources to convene meetings and engage in teaching and discussions about causal frameworks, but also the human resources techniques that can recognize the potential and values of participants, without changing the particular problem-focused dimension in question. These are not easy skills to find in one place, and we realize that implementing the six processes is a challenging task. What we do know is that if means-oriented policy learning is to advance at the global level, it must incorporate considerations of multiple steps and multiple causal pathways through which creative policy mixes might be uncoverd. This can ultimately result in advancing problem-focused pathways capable of making a positive difference on the ground.

Acknowledgments

We are grateful to two anonymous reviewers for their thoughtful comments on a previous version. We thank Metodi Sotirov and John Bryson for helpful comments on a previous draft. We thank, for collaborations on related theory building projects Iben Nathan, Wil de Jong, Kathleen McGinley, Constance McDermott, Graeme Auld, Jeremy Rayner, Gary Bull, A.K. Boedhihartono, Gabriela Bueno, Chris Elliot, J.D. Langston, R.A. Riggs and Jeffrey Sayer; and for collaborations on related substantive applications, Sarah Lupberger, Audrey Denvir, Michael Stone, Sebastien Jodoin, Reem Hajar, Pablo Pacheco, Chelsea Judy, Daniela Göhler, Michael Howlett, Paloma Caro Torres, Sara Sax and Daphne Yin. We are also grateful for the generous feedback we received on a different, but related draft paper, from the participants of the IWPP Pittsburgh workshop on ‘The causes and effects of policy learning’, especially to Claudio Radaelli, Claire Dunlop, Susanna Borras and Stephane Moyson.

Disclosure statement

No potential conflict of interest was reported by the authors. Cashore was briefly engaged in the early 2000s with an NGO-business stakeholder process on global certification systems that was supported by the Forests Dialogue.

Additional information

Notes on contributors

Benjamin Cashore

Benjamin Cashore is Professor of Environmental Governance & Political Science at Yale University’s School of Forestry and Environmental Studies. He is courtesy joint appointed in Yale’s Department of Political Science and is a Senior Research Fellow at the Yale MacMillan Center for International and Area Studies. He is the Director of the Governance, Environment and Markets (GEM) Initiative at Yale and is the Joseph C. Fox Faculty Director of the Yale International Fox Fellows Program.

Steven Bernstein

Steven Bernstein is a Professor of International Relations in the Department of Political Science and Co-Director of the Environmental Governance Lab at the Munk School of Global Affairs, University of Toronto. His academic publications span the areas of global governance, global environmental and sustainability politics, international political economy, and international institutions. He also co-edits the journal Global Environmental Politics. Current major research projects include: “Coherence and Incoherence in Global Sustainable Development Governance” and “Transformative Policy Pathways Towards Decarbonization.

David Humphreys

David Humphreys is Professor of Environmental Policy in the Faculty of Arts and Social Sciences, The Open University, UK. His research focuses on global environmental policy with specialisms in forestry and climate change. He holds a PhD in international relations from City University London and is an Honorary Fellow of the Institute of Chartered Foresters.

Ingrid Visseren-Hamakers

Dr. Ingrid Visseren-Hamakers is Associate Professor and Graduate Programs Director at the Department of Environmental Science and Policy at George Mason University, and affiliated with Radboud University. Her research program on transformative global environmental governance can be organized into four main focal areas: Integrative Governance; transforming international biodiversity governance; learning from current governance instruments; and the relationships between animal and environmental governance. She holds an MBA, and MS and PhD degrees in Environmental Studies.

Katharine Rietig

Dr. Katharine Rietig is Lecturer (Assistant Professor) in International Politics at the School of Geography, Politics and Sociology of Newcastle University. Her work sits at the intersection between Global Environmental Governance and Public Policy. She is interested in factors that facilitate transitions to sustainable economies such as learning, non-state actors and multilevel governance processes. She holds a PhD and MSc in Environmental Policy from the London School of Economics and a M.A. in Political Science, Law and Economics from Ludwig Maximilan University Munich.

Notes

¹ Some scholars use the term ‘experiments’ to advocate an approach consistent with anticipatory approaches, such as Hoffmann (2011) and Bernstein and Hoffmann (2018).

² Bryson argues that attention to both the interest of an organization, and how much power it has to influence the issue, helps identify the key players as well as advance interests of the powerless (Bryson, 2004).

³ For example, in a recent review initiated by The Forests Dialogue (2018) housed at Yale University’s School of Forestry & Environmental Studies aimed at ‘learning’ from 17 years of fostering ‘community engagement’, there were no references to scholarship on means-oriented policy learning, dispute resolution and/or dialogues. At the same time, several prescriptions were developed based on implicit causal claims, none of which were subject to theoretical or scientific assessments.

⁴ Whack-a-mole is a popular North American arcade game in which players attempt to hammer moles that pop out of one of nine holes, only to find that doing so causes another mole to pop up in another hole. The game is won when no more moles pop up.

⁵ The importance of rendering implicit biases explicit is a longstanding theme in stakeholder engagement scholarship. For example, Bryson (2014) notes that Meynhardt (2009) turns to ‘moral-ethical, political-social, utilitarian-instrumental, and hedonistic-Aesthetical’ cognitive frames with which to generate stakeholder engagement regarding problem definitions. Where we differ slightly from Meynhardt is that we argue cognitive frames, especially for global problem focused efforts, ought to be tied back to the nature of the problem in question, rather than using cognitive frames to determine the problem definition.

⁶ Cashore and Bernstein (2018) argue that much of current climate-focused learning dialogues tend to identify type 3 interventions despite implicit recognition that climate is a type 4 problem.

⁷ The rationale being that failure to do so will open up debates later in the process, threatening to unravel the coalition from means-oriented learning to interest-based processes.

⁸ In three wide ranging reviews, inspired by the concept of ‘public value’ Mark Moore and Barry Bozeman (Bozeman, 2002; Moore, 1995), Bryson (2004), Bryson et al. (2006, 2015)) make three key points about the number, nature, and types, of stakeholders to be engaged in collaborative governance. First, Bryson posits that the ‘one way to avoid outcomes that do not create public value is to begin with an inclusive definition of stakeholders, so that the net of considerations about who and what counts is cast widely to begin with’ (Bryson, 2004, p. 47). Second, Bryson, Crosby and Stone (2015, p. 51) argue that ‘collaborations are most likely to create public value when they build on individuals’ and organizations’ self interests.’ Third, and somewhat in contrast, Bryson et al. (2006), acknowledge that ‘Stakeholder analysis never should be seen as a substitute for virtuous and ethical practice’ (Bryson, 2004, p. 47) and that ‘what is being said does not imply that all possible stakeholders should be satisfied, or involved, or otherwise wholly taken into account, only that the key stakeholders must be, and that the choice of which stakeholders are key is inherently political, has ethical consequences, and involves judgment’ (Bryson, 2004, p. 26).

Our motivation to foster learning about ‘political effectiveness’ in ways that reinforce, rather than undermine, ‘policy effectiveness’, leads us to both advance and challenge these design principles in three ways. First, Cashore and Bernstein’s framework offers a way to adjudicate the ‘inclusionary/ cast a wide net’ design principle in ways that do not undermine an ‘ethical or virtuous practice’ by targeting stakeholder learning processes to first decide which of the four problem conceptions will be applied to generate learning. Second, and related, this approach requires that stakeholder learning deliberations constantly refer back to the original problem definition that was the rationale for creating the dialogue in the first place. In cases where, owing to efforts on the part of powerful organizations, or instrumentalist discussions of political effectiveness, stakeholder learning dialogues agree to change the original problem, then it follows that such deliberations, while effective in generating a consensus, are ineffective in addressing specified problem. Put another way, to refer to such a (highly common) process as ‘effective’ is essentially tautological, since the measure of substantive effectiveness is whatever the dialogue agrees to after the deliberations. Third, and related, Cashore and Bernstein’s conceptions render explicit attention to what is almost always implicit in the stakeholder learning design literature: that generalizing design principles for collaborative governance across all problem domains works to implicitly bias some problem conceptions over others. In particular, design principles focusing on legitimacy and trust across stakeholders tend to prioritize Type 3 conceptions, while design principles on organizational self-interest tend to prioritize types 1, 2 or 3. In almost all cases, even when paying attention to ‘ethical virtue’ the application of most design principles undermine type 4 conceptions, despite these constituting most of the thorny questions facing global governance in general, and the environment in particular. In fact, it may be that universalist design principles behind stakeholder dialogues explain why global environmental governance efforts have been so ineffective in ameliorating key problems, such as global climate change and species extinctions.

⁹ See also Bryson (2004, p. 39) for discussion of the promise but also pitfalls of ‘winning’ coalitions.

¹⁰ While Lambin and Thorlakson (2018) argue that ‘Contrary to widely held views, interactions between governments, NGOs, and private companies surrounding the adoption of sustainable practices are not generally antagonistic, their conclusions concern support for Type 3 compromise-oriented standards development, rather than specified problems such as deforestation’.

¹¹ An earlier version of this table appeared in Cashore and Howlett (2006, p. 150). Since the AJPS article, this table has been reproduced a number of times including in Howlett and Rayner (2008, p. 388), Howlett and Cashore (2007) and Howlett (2009). Howlett and Cashore’s (2009, p. 35) reproduction changed the term ‘policy measures’ in the title to ‘policy components’ as well as modifying some of the writing within each cell. In 2013, Rayner and Howlett (2013, p. 174) modified the title from ‘policy components’ to ‘components of a policy mix’. In 2014 Howlett, Mukherjee, and Rayner (2014) modified further the title to ‘Components of a Policy Mix and the Position of Policy Programs Therein’. In 2019 Howlett (2019, p. 15) presents the title as ‘Components of public policies involved in policy design’. The reproduction of Cashore and Howlett (2007) table here includes modifications by B. Guy et al. (2018, p. 28) which includes replacing the term ‘levels’ ‘content’, while ‘content’ is replaced by ‘focus’. Following recommendations from Howlett and Mukherjee in 2019, this table uses the term ‘tools’ in the bottom middle cell, since the original term ‘mechanisms’ has taken on, since the original article, a broader meaning about causality that the table is meant to help uncover, rather than identify.

3.5. This approach is consistent with political science work on assessing the range of ways in which learning might trigger multiple types of policy diffusion (Dobbin, Simmons, & Garrett, 2007), from one level to another (Jordan, Wurzel, & Zito, 2003).

¹³ Jodoin (2017) has found that even when focused on universalist issues as ‘human rights’, norms may still work to empower some groups over others, reinforcing the need for strategists to think carefully about, and nurture, particular pathways of influence.

¹⁴ This project focused on community forestry as an end in itself: i.e. the goal in this case is local involvement in forest management. We recognize that for many, community forestry is actually a means to achieve other ends such as livelihood improvements and maintaining traditional culture.

¹⁵ The Collaborative Partnership on Forests (CPF) has initiated a policy learning effort that has engaged several authors of this article, and its core scientific body, the International Union of Forest Research Organizations (IUFRO) has supported our work in developing this framework, and its application in Peru. Our point is that these efforts are both rare, and nascent. We argue that whether these efforts will end up reinforcing compromise, rather than problem solving, depends, in part, on whether they adopt the very techniques offered in this article and our other work.