Research for this paper has been conducted as part of the research project GreenGrass, which has been financially supported by the German Federal Ministry of Research and Education under the funding program “Agricultural Systems of the Future”, grant no. 031B0734K.
1Policy learning is a challenging concept in theories of the policy process. Often defined as experience-based belief change (Dunlop & Radaelli, 2013), it captures instances of political actors adapting their beliefs in response to evidence about the performance of political measures. Such an understanding resonates with notions of policy-oriented learning as updating beliefs, as developed in the Advocacy Coalition Framework (Moyson, 2017; Sabatier, 1987), but it only partially captures social learning as discussed in research on policy paradigms (Blyth, 2013; Hall, 1993). It also seems squarely at odds with the assumptions of the Multiple Streams Framework (MSF) which characterizes the political process as shaped by contingency and ambiguity (Kingdon, 2011). This implies that policy processes are barely predictable, while social reality can be interpreted in manifold ways. Systematic, evidence-based policy learning seems therefore unattainable under MSF assumptions.
2However, there are strands in the policy learning literature that point in a different direction. Regarding contingency, Kamkhaji and Radaelli (2017) have pointed to crisis situations which do not permit reflections and careful identification of causal mechanisms, but require situational policy responses, building on association and improvisation. Regarding ambiguity, Cino Pagliarello (2022) has shown how ideational change can be driven, rather than distorted, by polysemy and the utilization of boundary objects. This entails acknowledging an ambiguous environment of potential learning processes beyond a mere reduction of uncertainty. Moreover, some research has already combined policy learning and the MSF. Kippin and Cairney (2022) have suggested integrating the varieties of learning approach (Dunlop & Radaelli, 2013) in studies of multiple streams. Kuenzler (2018) has investigated individual learning against the background of multiple streams analysis and highlighted the role of policy entrepreneurs in policy learning. However, reconciling policy learning and the MSF is hindered by epistemological differences, especially with the discourse analytic interpretations of Kingdon (Winkel & Leipold, 2016). Developing a systematic concept of policy learning within the MSF therefore remains a conceptual challenge. We take up this task by addressing the question: How can policy learning be conceptualized and investigated in multiple streams research?
3In section 2, we review the respective literature and argue that the MSF enables scholars to differentiate between learning within and across streams, allowing us to make a conceptual distinction between six different types of learning in the MSF. We then use the fight over the re-establishment of wolf populations in the German state of Lower Saxony as an illustrative case to develop an understanding of policy learning under conditions of contingency and ambiguity. Section 3 introduces the case study. Section 4 presents findings on three within-stream and three cross-stream types of learning and the concluding section 5 discusses the benefits and the limitations of integrating policy learning in multiple streams.
4The integration of policy learning in the MSF requires background on both strands of literature. We start with the MSF. Studying agenda-setting in US politics and building on the Garbage Can Model of organizational choice (Cohen, March, & Olsen, 1972), Kingdon (2011) questioned the rationalist assumptions of a problem-solution sequence in the policy process and portrayed the political process as three independent streams: competing policy ideas; problems vying for attention; and power constellations. Problems, policies, and politics are ambiguous and contingent (Ackrill, Kay, & Zahariadis, 2013). Policy entrepreneurs depict problems and policy ideas to call attention to their issues and their preferred policies and place them on the agenda. This vision was sequentially worked up into the MSF (Herweg, Zahariadis, & Zohlnhöfer, 2023). Successful policy entrepreneurs couple streams by skillfully linking their preferred solution to problems on the agenda, exploiting given constellations of interests and powers (Cairney, 2018). Timing is critical for success. Only if the three streams are aligned – a problem high on the agenda, a persuasive solution available, and a supporting coalition behind it – does a policy window open, i.e. there is an opportunity for policy change (Herweg, Huß, & Zohlnhöfer, 2015).
5In contrast, policy learning studies address the question of who learns what and with what effect (Bennett & Howlett, 1992), focusing on the subject and object of learning and on the effect of the learning process. An important challenge regarding the learning subject is the linkage between individual and collective (Heikkila & Gerlak, 2013), as well as between micro and macro (Dunlop & Radaelli, 2017) learning processes. Regarding the object of learning, it is helpful to differentiate between policy learning, which refers to learning on policies, including their instruments and goals, and political learning on how to utilize power constellations strategically to advance a position (May, 1992). Societal challenges can be the source of a moment of puzzling, i.e., experts wondering what to do to address these problems, as opposed to powering, i.e., pushing policies against an opposition (Heclo, 1974). Policy learning often differentiates between levels at which beliefs can change (instrumental vs. value-related core beliefs), as represented by work on advocacy coalitions (Sabatier, 1987). Work on changes in policy paradigms differentiates between levels of change as an effect of learning (Hall, 1993). The latter is linked to the notion of social learning, i.e. learning that affects policy goals. Finally, another strand of literature focusing on the effect of learning problematizes that policy learning does not necessarily lead to better policy and that the links between the policy learning process and its outcome are not linear (Heikkila & Gerlak, 2013).
6Against this background, investigating policy learning within the MSF may at first seem counter-intuitive. Understanding learning as the updating of beliefs in light of evidence (Dunlop & Radaelli, 2013) can be in line with advocacy coalitions learning about instruments (Moyson, 2017; Nowlin, 2021) and with paradigm change involving goals (Blyth, 2013; Zaki, Pattyn, & Wayenberg, 2022). In the MSF, however, problems, policies, and politics are contingent and not necessarily connected in a logical way. The severity of problems and the merits of policies are ambiguous, so that the emphasis is shifted from evidence to (contested) interpretation. Consequently, Radaelli (1995, p. 174) suggests understanding the MSF as the epistemological opposite of approaches building on falsification, which is in line with more recent lines of work linking the MSF to discourse analytical approaches (Winkel & Leipold, 2016). It is also in line with work combining advocacy coalitions and multiple streams, in which the former provides an access to policy learning, while the latter captures the situational and temporal components in explaining policy change (Bandelow, Vogeler, Hornung, Kuhlmann, & Heidrich, 2019).
7However, there are at least two reasons to reconsider an integration of policy learning into the MSF. First, recent work on policy learning has shown interest in constellations characterized by contingency and ambiguity. Policy learning approaches often refer to inferential learning, i.e. drawing conclusions from experience. In cases with changing contextual conditions, however, it might be helpful to turn to contingent learning, i.e. learning that takes place simultaneously with change and which operates based on surprise and experimentation (Kamkhaji & Radaelli, 2017; Radaelli, 2022). Recent work has also emphasized the role of ambiguity in ideational change and shown how strategic use of polysemy can help to stabilize coalitions and allow for policy change (Cino Pagliarello, 2022; Radaelli, 2023). Second, pioneering work investigating policy learning in the MSF does exist. Combining the MSF with the varieties of learning approach which distinguishes different modes of learning (Dunlop & Radaelli, 2013), Kippin and Cairney (2022) have highlighted learning processes that exceed mere instrumental considerations. Referring to the MSF’s concept of policy entrepreneur, Dudley (2007, p. 409) has suggested that established approaches toward learning would benefit from paying closer attention to the role of key individuals. Arguing that learning cuts across different stages of the policy cycle, Kuenzler (2018) has pointed out that the coupling of streams by entrepreneurs could entail individual learning. These strands of work suggest that an MSF perspective can help to develop an understanding of policy learning that differentiates different types of learning modes, depending on the situation of the objects of learning in the three streams, and on their interplay across streams.
8The conceptual metaphor of the streams captures the ever-changing circumstances in a policy field or political system regarding the problems, policies, and politics, and the independence of changes in these streams (Kingdon, 2011). Obviously, there are countless empirical examples to argue that the opposite is the case (Mucciaroni, 1992). Yet the assumption of only loosely coupled streams is an integral part of the framework because it functions as a conceptual device (Herweg et al., 2023, p. 51) that emphasizes timing and interpretation. Only if events across the streams are skillfully coupled by policy entrepreneurs during an open policy window, is a policy change considered and potentially adopted. In recent conceptions of the MSF (Dolan & Blum, 2023), coupling is often approached from a discourse analytic perspective, as pioneered by Winkel and Leipold (2016), who argue that the streams represent discursive patterns, while coupling consists of strategic practices establishing new channels between them. From this perspective, coupling can be brought about by making statements (Zittoun, 2013), telling stories (Dudley, 2013), or connecting ambiguous and contingent material from the streams by creating arguments (Blum, 2018). Coupling activities are continuously performed by entrepreneurs, not only during policy windows (Kingdon, 2011, p. 229), and may be limited to two streams as partial coupling (Dolan, 2021).
Table 1: Structured summary of instances of learning within and across the three streams
Learning within streams
|
Within problem stream
|
Within policy stream
|
Within political stream
|
Problem-focused learning
Task: Problem definition; gather knowledge on conditions, turn conditions into problems
Actor: Knowledge brokers, problem brokers
Learning effect: Problem definition/salience, degree of tractability
|
Policy-focused learning
Task: Softening up; design and combine policies, enhance attributes and familiarity
Actor: Policy communities, policy entrepreneurs
Learning effect: Higher feasibility, acceptability, affordability, recognition
|
Politics-focused learning
Task: Powering; acquiring positions, accumulating influence via resources and networks
Actor: Decision makers, interest groups
Learning effect: Balance of power, (a)symmetry of actor certification
|
Learning across streams
|
Problems-policies relations
|
Policies-problems relations
|
Politics-problems relations
|
Policy learning
Task: Puzzling; finding a suitable policy for a given problem
Actor: Policy communities, communities of practice
Learning effect: Policy design, instrument design
|
Coupling-based learning
Task: Consequential coupling; creating discursive links between a given policy and a problem
Actor: Policy entrepreneurs
Learning effect: Causal beliefs justifying given policy, political learning
|
Choice-based learning
Task: De/politicization; manage problems to secure and maintain office
Actor: Decision-makers, interest groups
Learning effect: Problem choice, altered salience
|
Problems-politics relations
|
Policies-politics relations
|
Politics-policies relations
|
Coupling-based learning
Task: Blame management; blame avoidance
Actor: Decision makers, policy entrepreneurs
Learning effect: Improved blame management, political learning
|
Coupling-based learning
Task: Doctrinal coupling; policy suggestion aimed at influential actors
Actor: Policy entrepreneurs
Learning effect: Improved coupling of politics, political learning
|
Choice-based learning
Task: Commissioning; finding and selecting policy to enhance political position
Actor: Decision makers, political entrepreneurs
Learning effect: Policy choice, instrument choice
|
Source: the Authors
9Against this background, two starting points are promising to inquire and conceptualize instances of learning in multiple streams. First, based on the assumption of independent and ever-changing streams, we expect that learning takes place within each stream, resulting in three within-stream types of learning: problem-focused learning, policy-focused learning, and politics-focused learning. Second, in light of the discursive struggles taking place over time in the relations between the three streams (Möck, Vogeler, Bandelow, & Hornung, 2023), we expect that learning can also occur at the intersections of problems and policies, of policies and politics, and of politics and problems, with each stream posing a potential starting point for a link to another. Accounting for the different directions, in which links can be established, each of the three pairs of streams would allow for two learning constellations. Against this background, we derive three cross-stream types of learning: policy learning, choice-based learning, and coupling-based learning. Table 1 summarizes the six types of learning in the MSF by elaborating on the object, subject, and effect of learning (Bennett & Howlett, 1992). In the following, we introduce each of the six types and their main mechanisms.
10The MSF assumes that problems are not externally given. Rather, policy entrepreneurs argue that given conditions are not acceptable and that something can and should be done about them (Kingdon, 2011). Indicators such as unemployment figures, focusing events such as hurricanes, and feedback about the effects of existing policies are particularly effective sources of problem definition, which can be exploited by ‘problem brokers’ (Knaggård, 2015). Different from policy entrepreneurs, problem brokers define and advocate problems without linking them to policies, with knowledge brokers specifically focusing on the promotion of scientific knowledge (Knaggård, 2016). By advancing knowledge on indicators, events, and policy effects, problem and knowledge brokers contribute to learning within the problem stream. Their updating of beliefs is based on understanding the characteristics, causes, and consequences of specific problems, but does not yet involve policies. Regarding, for example, extreme weather events (Dolan, 2021), knowledge brokers could provide insights that improve forecasting, allow assessments of who is affected in which way, or indicate climate change as a main driver of their frequency and severity. While such a problem-focused learning process can be analytically isolated, it enhances problem tractability, i.e. the degree to which problems are understood and controllable (Dunlop & Radaelli, 2022), which is a main factor influencing learning mode choice.
11Kingdon (2011) has described the policy stream as ‘policy primeval soup’ that develops in an evolutionary way with policy ideas constantly emerging, and being adjusted, recombined, or split up, and eventually being forgotten or adopted. Three criteria determine evolutionary success and are the likely focus of policy-focused learning (Goyal, Howlett, & Taeihagh, 2021): technical feasibility, value acceptability, and financial viability. To increase the chances that their (pet) policies survive, policy entrepreneurs must make them appear more doable, agreeable, or affordable. They conduct test-runs to assess the potential of a policy, refer to resistance and skepticism against a proposal, identify possible adaptations, and make the policy generally more familiar. Such ‘softening up’ is a learning activity within the policy stream (Kagan, 2019) if limited to improving feasibility, acceptability, and affordability. If a draft policy faces concerns about technical barriers, ethical and legal problems, or high costs, policy-related beliefs may be updated by adjusting features of the policy proposal itself. Policy entrepreneurs might adjust policies in response to incoming information to reduce costs, align with changing norms and values, or facilitate challenging implementation. Cost-effectiveness provisions, legal interpretations, ethical justifications, and technical propositions can be regarded as instances of policy-focused learning as long as the softening up activities do not entail an argumentative coupling of streams (Blum, 2018).
12The political stream represents the balance of power changing over time in favor – or to the detriment – of certain policies. While the overall development, based on national mood, interest-group campaigning, and electoral turnover, is beyond the influence of individual policy entrepreneurs (Kingdon, 2011), they do aim at influential positions and seek to increase their influence on decision-making processes relevant to their pet policies, e.g. by securing access to key decision-makers, building and maintaining coalitions, and establishing reliable relations with supporters in networks (Reardon, 2018). Unsuccessful coalition building can result in a political stream not being ready for coupling (Hornung, Bandelow, & Iskandar, 2024). Activities in the political stream resonate with the notion of powering, traditionally associated with non-learning, where beliefs and positions are regarded as stable and political decisions are explained by varying power constellations (Heclo, 1974). Therefore, while powering in the political stream does not necessarily entail learning, interest groups and decision makers can evaluate the success and failure of their activities in coalition building and resource acquisition, update their beliefs, and adjust their strategies. Such politics-focused learning would be confined to the political stream and more limited than political learning, as defined by May (1992, p. 339) as ‘learning about strategies for advocating policy ideas or drawing attention to policy problems’, which represents cross-stream learning. Yet political learning is not limited to advocacy strategies which focus on framing and narratives; it also involves socialization and governance strategies such as trust building and capacity development, as recent work on scientific experts in EU climate policy-making has shown (Zaki & Dupont, 2024). More generally, learning on how to influence the balance of power effectively can also result in an alteration of the certification of actors, i.e. the degree to which some actors are recognized as more influential than others (Dunlop & Radaelli, 2022). As with problem-focused learning, this would influence the choice of learning modes and therefore undermine a strict distinction between puzzling and powering, as the latter would pose a context for the former (Dunlop, Radaelli, Wayenberg, & Zaki, 2024).
13The first type of cross-stream learning is policy learning and occurs at the interface of the problem and the policy stream. If a problem is defined for which no solution is available at the respective point in time, this could create a moment of collective puzzling (Heclo, 1974), in which actors have not yet decided what they want to do – as they would in powering – but try to learn about the problem so as to improve policy. In the MSF, this constellation is represented by cross-stream policy learning which originates from the problem stream and seeks to establish links to a policy stream that is not ready for coupling, i.e. a problem is looking for a solution. In this context, the MSF accounts for policy learning under ambiguity, in which beliefs about policies are updated in light of problem-policy relations. Potential sources of policy learning are new information on the extent or severity of a problem, but are also feedback from existing policies. Responding to evolving insights, policy communities may revise their beliefs about policy goals or instruments in social or instrumental learning modes (Hall, 1993; May, 1992). The relationship between problem stream and policy stream therefore allows for policy learning, with the caveat that policy change does not necessarily indicate previous learning (Heikkila & Gerlak, 2013; Sandström, Morf, & Fjellborg, 2020). If, for example, focusing events, such as weather-related disasters, are linked to policy failure, this could create a context for instrumental learning about early-warning systems, for social learning about adequate building codes, or for denial and finger-pointing which lead to non-learning or learning failure (O'Donovan, 2017). The loci of learning from problem-policy relations are generally policy communities, but in situations characterized by a low degree of actor certification (Dunlop & Radaelli, 2022) are also communities of practice, for example living labs (Möck & Feindt, 2024).
14Choice-based learning bridges the political stream, and either the problem or the policy stream, and involves a strategic choice among competing problems or policy solutions. It is therefore similar to powering but involves one of the other streams.
15i) At the interface of the political and the problem streams, actors face the challenge that problem selection and management have effects on the acquisition and preservation of power. Here, choice-based learning occurs if power considerations guide problem selection. It involves the development of de/politicization strategies (Feindt, Schwindenhammer, & Tosun, 2021): Actors will seek to politicize issues that support their positions, coalitions, and resources and aim to depoliticize issues that could hurt them. Choice-based learning occurs if decision-makers or interest groups update their beliefs about the effects of (relative) problem salience on their standing, eventually leading to altered problem salience. Recent MSF-based work has shown that problem constitution and salience are subject to de/politicization effects and that problems which potentially hurt organized interest groups are exposed to depoliticization efforts and agenda denial (DeLeo & Duarte, 2022), thereby affecting problem salience.
16ii) At the interface between the politics and the policy stream, the MSF concedes decision-makers a proactive role (Eckersley & Lakoma, 2021) and the ability to make a conscious selection among competing policy options in a process of ‘commissioning’ (Ackrill & Kay, 2011). If influential actors advocate for a policy in order to gather sufficient support for it, they link the policy and the politics stream and act as political entrepreneurs (Herweg et al., 2015). The selection of policies and respective instruments entails searching for a suitable solution. It is similar to puzzling but follows different, more strategic criteria of evaluation. Choice-based learning occurs if decision-makers or political entrepreneurs update their beliefs about the usefulness of policies to secure and advance their influence and position, based on their experience in selecting policies. The ensuing effects on the policy stream are more directly observable than the mental learning processes.
17Policy entrepreneurs are not necessarily successful (Cairney, 2018). Learning can enable them to adapt their course of action based on experience. Policy entrepreneurs then become learning subjects (Kuenzler, 2018) with a focus on political learning (May, 1992). Policy entrepreneurs evaluate previous experience with agenda-setting strategies and update their beliefs. The notion of coupling-based learning builds on the concept of partial coupling (Dolan, 2021) between two of the three streams. Coupling-based learning can occur at each dyad of streams.
18i) At the interface of the policy and the problem streams, ‘consequential’ coupling (Zahariadis, 2003) refers to the typical garbage can situation where ‘solutions look for problems’ with enhanced chances of success during problem windows. At this interface, policy entrepreneurs do not necessarily learn something substantially about policies or their interaction with problems or politics, but adapt their beliefs on how problems can be discursively linked to policy solutions. Based on experience with such problem-policy arguments, policy entrepreneurs update their approach of utilizing ambiguity to their advantage and of making sense of the world strategically.
19ii) At the interface between the policy and the political streams, a similar learning situation can be observed if policy entrepreneurs couple the policy and the political streams and draw on this experience for updating strategies. The ‘doctrinal’ style of coupling (Zahariadis, 2003) refers to entrepreneurs who suggest a specific policy to influential actors as being in line with their overall ideological agenda. This is particularly promising during open political windows, for example with new governments eager to implement their agendas (Herweg et al., 2023).
20iii) The interface between the problem and the political streams has received less attention despite its potential for agenda-setting. By attributing responsibility for problems to decision-makers, policy entrepreneurs can engage in coupling beyond the policy stream and generate demand for solutions and a sense of urgency. Such blame games are particularly observed in connection with focusing events (Boin, 't Hart, McConnell, & Preston, 2010) and entail strategies of blame avoidance (Weaver, 1986). Changes in blame management could indicate coupling-based learning, while instances of learning, processes of reflection, and conscious adaptation are more difficult to observe.
21Finally, the success of coupling styles depends on the type of policy window, i.e. its origin in the streams. Therefore, coupling-based learning entails learning about the selecting and timing of different partial couplings.
22To illustrate and test the typology of learning within and across multiple streams, we looked for a case of policy-making with high levels of ambiguity in the problem, policy, and political streams, settling on wolf policy in the German state of Lower Saxony as a highly ambiguous environment for policy-related learning. The wolf, Canis lupus, has been repopulating Germany since around 2000, after having disappeared at around 1900 as a result of hunting and loss of habitat (Trouwborst, 2010). The comeback was facilitated by an international and European wildlife protection regime implemented in unified Germany in the 1990s. This is based on the Bern Convention (1979) and the European Union’s Habitats Directive (1992) and grants the wolf the status of a strictly protected species (Trouwborst & Fleurke, 2019). In Germany, wildlife protection policy falls under concurrent legislation, which allows German states to pass laws on issues not covered by federal legislation. This makes German wolf policy a multi-level governance issue: While states must comply with German and European legislation, they have ample leeway in problem description, agenda-setting, and the implementation of solutions.
23Wildlife-livestock conflicts emerge where livestock farming coexists in shared spaces with wild carnivores. The reintroduction of wolves in livestock areas revives ancient conflicts because wolves depredate livestock, particularly grazing animals on pasture, creating financial losses for farmers, which may threaten some farmers’ livelihoods (Enserink & Vogel, 2006). Lethal measures to reduce depredation, however, can come at the cost of carnivores being dispelled or becoming extinct (Kuijper et al., 2019). Therefore, carnivore-livestock conflicts are characterized by an inherent ambiguity in understanding the carnivore as threat vs. threatened. Livestock depredation by carnivores may be defined as a problem to be solved by suitable policies, while also stirring political conflict. Furthermore, the predators’ role in ecology and society can be strategically interpreted, while policies undergo continuous evolutionary change.
24Recent research on Germany has investigated ways of mitigating wolf-livestock conflicts by assessing the effectiveness of livestock protection measures (Bruns, Waltert, & Khorozyan, 2020) or by predicting local depredation risks (König et al., 2023). Earlier work on German wolf governance highlighted a high degree of political opposition in Saxony (Stöhr & Coimbra, 2013) and pointed to the high mobility of wolves as an important driver of governance arrangements (Thiel, Schleyer, & Plieninger, 2012). Niedziałkowski (2023) has documented long-term German wolf policy and explained its institutional pattern as being safeguarded by European norms.
25Our case study focuses on the state of Lower Saxony and investigates learning within and across streams over six years of wolf policy-making from 2017 to 2022. Lower Saxony was selected because it has been more affected by the issue than most other German states. Although wolves returned to Lower Saxony as early as 2011/2012, the investigation period was selected due to developments in the three streams that make the spatio-temporal context highly interesting. Increasing numbers of wolves settling in Lower Saxonian regions, along with livestock-depredation events, constituted continuous shifts in the problem stream. In the political stream, the investigation period covers two state elections with subsequent government turnovers. Most importantly, in the policy stream, two significant policies were discussed and decided. In 2020, the Lower Saxonian Wolf Ordinance (NWolfVO) was issued, providing regulations on the scaring, startling, and killing of wolves under specific circumstances. In 2022, these regulations became a part of the amended Lower Saxonian Hunting Law (NJagdG), changing the legal context of wolf governance. From a multiple streams perspective, these instances of agenda setting and policy change indicate intensive cross-stream activity. Although extensive material from within streams, and the high potential for interactions of streams in the investigation period, do not yet indicate learning – which is always less frequent than its potential sources – they do pose a promising starting point in looking for instances of learning within and across streams.
26We follow a case study approach, situated in the MSF methodology (Zohlnhöfer, Herweg, & Zahariadis, 2022). Desk research identified and gathered documents up to the required level of saturation, i.e. until the three streams could be mapped over the case’s time period. The empirical material includes legislation, management plans, parliamentary debates, electoral programs, coalition agreements, interest group publications, articles in business magazines and newspapers, and scientific reports, especially data from the Federal Documentation and Advisory Centre Concerning the Wolf (DBBW). For contextualization in broader policy developments, we draw on Niedziałkowski (2023). Based on document analysis, we reconstructed the development of each stream. An integrated time line, specifying occurrences and streams as well as a list of sources, is available in the appendix. Subsequently, this material was analyzed to identify deductively those instances of the six types of learning within and across streams, based on the key tasks and main subjects associated with each type. In order to assess whether these activities entail evidence of learning, policies or political strategies were examined for adaptations, alterations, or improvements over time.
27MSF case studies are organized according to occurrences in the problem, policy, and political streams (Zohlnhöfer et al., 2022), which provide the material that can be interpreted by entrepreneurs under conditions of ambiguity (Winkel & Leipold, 2016). Below, we trace the development of each independent stream and present evidence of instances of problem-focused, policy-focused, and politics-focused learning within streams.
28As in other wildlife management conflicts, the interaction between wolf, livestock, and humans is an ambiguous condition with potential for different problematizations. After wolves had been widely perceived as a threat to humans for centuries, the late 20th century saw the rise of a nature-restoration paradigm, emphasizing the ecological benefits of wolves as a capstone species threatened by humans. While this paradigm was dominant at the beginning of the investigation period, the reverse interpretation of the wolf as a threat to livestock and human livelihoods constitutes the area of problems that change-oriented proposals sought to address. There are two primary sources of problem perception: wolf sightings and wolf attacks. Boundaries blur between indicators and focusing events. The first wolf entering a jurisdiction, or the first pair of wolves reproducing, are noticeable events. With established populations, wolf monitoring produces numbers that can indicate conservation success or rising threats. Similarly, each instance of livestock depredation can serve as a focusing event, while corresponding statistics can be used to indicate an accumulating problem.
29The policy community soon learned that wolves live in packs with pairs raising cubs and with young wolves dispersing, sometimes across long distances, sometimes returning to their families, and eventually reproducing and founding new packs in new territories (Packard, 2019). Problem-focused learning included emerging evidence of established wolf territories. The first wolf territory crossing the Lower Saxonian border was established in 2008/2009, while the first within the state occurred in 2011/2012 as reported by the DBBW (2023). The official DBBW counts show a continuous increase in the number of territories of packs and pairs from 14 to 44 between 2016 and 2022 [see Figure 1 in the supplementary material].
30Wolf attacks on livestock became a particularly important area of problem-focused learning with a focus on the number of incidents of wolves killing animals. The policy community also learned about a phenomenon known as surplus killing: As grazing livestock must be fenced in Germany, in particular small grazing animals, such as sheep, cannot escape once a wolf has overcome the barrier, and the mere availability of prey can trigger wolves killing more livestock than they require for consumption (Kiffner, Uthes, Ostermann-Miyashita, Harms, & König, 2022). As focusing events, surplus killings allow the portrayal of wolves as dangerous and bloodthirsty. In May 2022, a wolf attack near Osnabrück, with 65 sheep killed, received specific attention, despite being one of 296 wolf attacks on livestock that year. The focusing event had more effect on the agenda than did the decreasing number of incidents in 2021 [see Figure 2 in the supplementary material]. Problem brokers had learned to use surplus killings as focusing events.
31Compared with livestock depredation, attacks by wolves on humans can easily be defined as problems of exceptional severity. However, wolves generally avoid humans and in Germany no attack has been documented since 1980, while wolf attacks on humans are common and sometimes fatal in countries such as India and Iran (Dobson et al., 2023). Nevertheless, reported encounters with wolves can become a source of updated problem definitions. Media reported a woman killed by wolves in Greece in 2017 and a man bitten by a wolf in Lower Saxony in 2018. In both cases, later investigations could not confirm the wolf attacks, but the 2017 incident, in particular, was utilized during the ongoing Lower Saxonian election campaign. In this case, a problem-focused learning would entail a change in the potential victims from mere livestock to including humans, thereby reinforcing the priming effect of the wolf as highly dangerous.
32In wolf policy, feedback occurs in both the successful and the unsuccessful execution of lethal measures against specific wolves and in the costs involved in payments for depredation compensation and livestock protection subsidies, discussed in detail in the policy stream. During our investigation period, Lower Saxonian authorities repeatedly failed to remove individual wolves because experts were not able to localize, trace, and shoot them. During this period, costs of prevention/compensation measures increased from €487,502/€39,745 in 2017 to €3,515,438/€117,004 in 2021. Already, by 2017, costs had become a popular reference for problem definition, for example in a parliamentary inquiry by the Christian Democrats. Problem brokers learned to present targeted wolf management as a problem in itself.
33There are three subgroups of policies that approach the carnivore-livestock conflict from different perspectives: wolf conservation, conflict mitigation, and wolf reduction (see table 2). First, wolf conservation policies represent the dominant policy paradigm during the investigation period. They are rooted in the perception of the wolf as being threatened and were established in the late 20th century in overarching policy architectures including the Bern Convention, the Habitats Directive, and the German Federal Nature Conservation Act. As a consequence, the wolf is strictly protected in Germany. As these policies were already enacted, they are not part of the material in the policy stream, but pose the main background.
Table 2: Summary of the main strands of policy-making in the German wolf-livestock conflict
Policy approach
|
Wolf conservation
|
Conflict mitigation
|
Wolf reduction
|
Main policy goal
|
Reestablish/conserve wolf populations
|
Stable coexistence of livestock and wolves
|
Removal (i.e. killing) of wolves to protect grazing livestock
|
Governance mode
|
Protection provisions
|
Targeted payments
|
Exemptions from protection provisions
|
Functionality
|
Specify provisions which ensure that wolves are safe and that their natural habitats are preserved
|
Provide farmers with information and resources to prevent and/or deal with depredation events
|
Allow for, and specify, conditions of applying lethal measures to remove wolves based on reasons, e.g. attacks
|
Policy relations
|
Long established, but jeopardized by wolf-reduction approaches
|
Problems caused by wolves are recognized, yet more tolerance supports conservation
|
Wolf conservation is challenged, but legal norms limit scope
|
Examples
|
1998 Federal Nature Conservation Act
|
2014 Lower Saxonian Wolf Directive
|
2022 Lower Saxonian Hunting Law
|
Source: the Authors
34Second, conflict mitigation policies mostly address farmers affected by rising numbers of wolves and by livestock depredation. Examples are information campaigns and guidelines on how to adapt farm management to wolves. In Lower Saxony, the first wolf management plan was published in 2010 with updated editions in 2017 and 2022. Conflict mitigation is also pursued through payments to farmers. Public funding of herd protection measures is intended to enable farmers to establish practices and infrastructures preventing wolf depredation. In Lower Saxony, these subsidies are limited to investments in wolf-deterring fences and guard dogs, but do not include running costs. Additionally, farmers can apply for compensation payments based on livestock losses caused by wolf attacks. Compensation requires proof of established basic protection measures, in particular wolf-deterring fences which comply with predefined criteria, and documentation of the depredation event, including official veterinary confirmation that the damage was caused by wolves.
35Both investment and compensation payments were introduced before the beginning of the investigation period in the 2014 Wolf Directive. At first, subsidies and protection requirements for compensation applied only in selected counties. Successively, the geographical scope was extended and in late 2017 all jurisdictions within Lower Saxony were included. These adjustments point to potential policy-focused learning. Financial viability concerns, for example, were reflected in the limited eligibility for subsidies based on the location of the farm and the exclusion of running costs, but play a subordinate role in the extension of scope in 2017, which entailed increasing costs. Learning about technical feasibility issues, in terms of implementing livestock protection measures on farms, is indicated by the nuanced calibration of specifications on types of fences and their height or electrical current. Further learning is embodied in updates of the Lower Saxonian Wolf Management Plan, which provide guidance on how to react to unusually behaving or injured wolves.
36Third, policies aiming at wolf reduction seek ways to provide legal means of controlling wolf populations utilizing lethal measures. Although the killing of wolves is generally prohibited, Article 16 of the EU Habitats Directive defines goals that may justify the provision for a member state to allow wolf ‘removal’ if there is no alternative and if the conservation of the respective wolf population is not threatened (Trouwborst & Fleurke, 2019). The second amendment to the Federal Nature Conservation Act in 2020 introduced a wolf-specific clause (§45a) aimed at the prevention of serious economic damage and allowing lethal measures against non-specific wolves in a narrow spatio-temporal context of such damage. The Lower Saxonian Wolf Ordinance of 2020 explicitly adopts this clause. If an individual wolf causing serious economic damage cannot be identified, other wolves in a narrow spatio-temporal context can be killed as well. These changes can be interpreted as a process of policy-focused learning with reference to value acceptability. As European law defined strict limits to wolf-reduction policies in Germany (Niedziałkowski, 2023; Thiel et al., 2012), identifying ways to implement them in Lower Saxony also constituted a legal puzzle to be solved within the policy stream. Finally, an amendment to the Lower Saxonian Hunting Law in 2022 replaced the Wolf Ordinance, changing the legal status of wolf management and reassigning responsibilities for lethal measures to hunters. However, a year-round grace period ensures that wolves are not generally hunted despite the shift in legal context.
37The political stream is typically delineated by drawing on the changes of governments in office, the constellations of organized interest groups, and the national mood (Kingdon, 2011). In particular, the governing party (or coalition) has been recognized as a major factor determining the condition of the political stream in parliamentary systems such as Lower Saxony (Herweg et al., 2015). During the investigation period, two elections led to a change in government and allowed for closer analysis of the party positions at state level. In summer 2017, wolf policy became a salient issue in Lower Saxonian electoral campaigns with Christian Democrats and Liberals calling for lethal measures and a transfer of wolf-related regulations to hunting law. Social Democrats did not support the latter proposal, but acknowledged the need for lethal measures. The Green Party, which since 2013 had held the environmental department, highlighted the positive aspects of reestablishing wolves. Table 3 provides an overview of all party positions.
Table 3: Positions of Lower Saxonian parties on wolf policies in electoral programs of 2017
Position
|
The Left
|
SPD Social Democrats
|
Green Party
|
FDP Liberals
|
CDU Christian Democrats
|
AfD Extreme right
|
Strengthen the return of wolves
|
Yes
|
|
|
|
|
|
Return of wolves has pros and cons
|
|
|
Yes
|
|
|
|
More monitoring and information
|
|
|
Yes
|
|
|
|
Support livestock owners financially
|
|
Yes
|
Yes
|
Yes
|
Yes
|
|
Keep wolves away from settlements
|
|
Yes
|
|
Yes
|
Yes
|
|
Kill wolves displaying unusual behavior
|
|
Yes
|
|
Yes
|
Yes
|
|
Transfer wolves to hunting law
|
|
|
|
Yes
|
Yes
|
Yes
|
Source: the Authors
38Two interest groups were particularly influential on wolf policy at federal level (Niedziałkowski, 2023). The Nature and Biodiversity Conservation Union (NABU) advocated coexistence, calling for herd-protection measures, better understanding of – and communication about – the predator’s behavior, and enhanced wolf protection (Naturschutzbund Deutschland, 2022). A competing coalition of interest groups, including the German Hunting Association and the German Farmers’ Association, advocated the removal of wolves and formed the Action Alliance Forum Nature (AFN) in 2004 (Niedziałkowski, 2023). Independent of both problem and policy streams, coalition formation was a core aspect of powering in the political stream, which can be a source of politics-focused learning. As a well-connected policy entrepreneur, the AFN is highly active in lobbying and advocacy, engaging in debates in policy communities and developing proposals on wolf management and ‘protective hunting’ (Aktionsbündnis Forum Natur, 2019).
39Identifying changes in public mood is challenging regarding operationalization and data availability. Opinion polls may provide insights (Zahariadis, 2015), but need to be considered with care, especially if commissioned by interest groups. Exit polls in the 2017 parliamentary election showed a majority rejecting stronger measures against wolves and credited the Green Party for the best handling of matters regarding protected animals. In 2021, a survey commissioned by the Lower Saxonian farmers’ association and others found a majority supporting wolf management, including lethal measures, but also a similar majority welcoming the return of wolves.
40Against the background of developments and learnings within streams, we turn to learning across streams.
41The successive repopulation of certain regions by wolves was turned into a problem by highlighting the issue of livestock depredation and the urgent need to act. This triggered extensive episodes of puzzling for the policy community. Although the policy stream was ready for coupling, with proposals such as the funding of prevention or financial compensation, prioritizing measures and defining admissible basic protection can pose open questions in need of answers. Implementation deficits in herd protection schemes and emerging evidence on their effectiveness (Frank & Eklund, 2017), generated in epistemic communities, informed policy-making regarding, for example, the minimum height or depth of wolf-deterring fences to prevent wolves from digging under or jumping over them. Similarly, adjustments of the spatial scope of eligibility to the dispersal of wolves indicates cross-stream policy learning based on changing conditions. This was particularly visible in the successive amendments made to the Wolf Directive until it covered all of Lower Saxony in 2017. The situation on the North Sea coast was identified as a specific problem. Here, sheep graze on dikes where herd protection measures, such as wolf-deterring fences, conflict with coastal protection. Looking for a solution to this problem, a specific exemption (§6) was added to the Lower Saxonian Wolf Ordinance, allowing lethal measures against wolves which repeatedly overcome protection on dikes and which hurt or kill sheep.
42Investigating the links between the political stream and the other two streams, the inclusion of wolves in the Lower Saxonian Hunting Law in 2022 is of particular interest as this policy change entailed only minor practical consequences and was sometimes regarded as merely symbolic. However, a nuanced shift in problem definition was a win for wolf reduction proponents and indicates choice-based learning. Following the logic of hunting law, the year-round grace period for wolves implies that not shooting wolves poses an exception from hunting, while under conservation law each lethal measure had been framed as an exception from strict protection. The result is a move towards a new guiding policy principle which tilts more toward deliberate wolf reduction than did the previous legislative context. Advocates of wolf conservation, including the NABU, recognized the implications and shifted the issue to juridical venues, referring to the legal framework and particularly to the European level.
43These developments go back to autumn 2017, when Social Democrats and Christian Democrats formed a coalition government and agreed on regulations to enable lethal measures against ‘unusually behaving’ wolves and on transferring wolves to the hunting law, subject to the assignment of a favorable conservation status at European level and a transfer of wolves from annex IV to annex V of the Habitat Directive, entailing a lower protection status. The coalition agreement is the result of deliberate coupling between the political and the policy streams through commissioning, i.e. strategic choice between policy options guided by political considerations. Similarly, when the Lower Saxonian Minister for the Environment, Olaf Lies, a Social Democrat, introduced lethal measures for wolves in the 2020 Wolf Ordinance, the policy option was selected in response to intense pressure by interest groups and the conservative coalition partner.
44Instead of learning from problems to improve policies, policy entrepreneurs often present their preferred policy as a solution to a problem coming along. A common point of reference for such consequential coupling is provided by numbers. For example, the AFN (2019, pp. 6-7) argued that wolf population management, including lethal measures, was needed to respond to a doubling in livestock depredation events in Lower Saxony in 2017, while the NABU (2022, p. 19) highlighted that livestock made up only 1.6 percent of wolves’ diet. After the 2022 election, the new Minister for the Environment from the Green Party, Christian Meyer, interpreted decreasing livestock depredation as the success of prevention policies, while the Lower Saxonian Hunting Association pointed to unrecorded cases and a decline in pasture grazing.
45The change in government in 2022 was more generally a source of concern for advocates of wolf reduction policies as the Ministry of Agriculture was taken over by the Green Party whose actors seek to increase transparency in assigning shooting permits and to establish a dialogue with livestock farmers regarding the wolf. During the formation of the government, and after inauguration, policy entrepreneurs – most prominently Jörn Ehlers representing the Lower Saxonian Farmers’ Association – expressed concerns that, under the new government, there would be talking instead of action and called for the establishment of ‘wolf-free’ regions and enhanced remuneration for farmers. These instances of coupling are not doctrinal in a conventional sense, as suggestions are not in line with the Green Party’s original political agenda, and nor do they entail discursive shifts that allow for the diagnosis of learning. However, they do illustrate the relevance of timing with actors recognizing, and seeking to utilize, shifts in the political stream.
46Finally, the problem stream was also coupled to the political stream by discursively connecting wolf attacks on livestock to authorities and legislators, thereby constructing responsibility and generating pressure to act. In the Lower Saxonian legislative term of 2017-2022, the Liberals, in particular, used emerging information to blame the government. They first linked the government to rising problem indicators and increasing costs from prevention and compensation schemes. Policy feedback about implementation challenges involved in wolf-reduction policies provided further opportunities for coupling-based learning, in particular unsuccessful attempts to kill wolves for which shooting permits had been issued, but which experts, such as hunters and trappers, were unable to track or catch. A focusing event for coupling-based learning was the unsuccessful hunt of male wolf GW717m, from the ‘Rodewald pack’, at a high cost, for about a year in 2019/2020. In spring 2020, this episode became a major focus of blame attribution and blame avoidance, e.g. in a plenary debate in the Lower Saxonian parliament, initiated by the Liberals.
47Combining concepts of policy, political, or social learning with the MSF is an attempt to integrate two different visions of the policy process: Learning is often associated with methodical proceeding, evidence, evaluation, and improvement, whilst the MSF presents a contrasting vision of the policy process as rather chaotic, unpredictable, and driven by opportunistic actors who are often fixated on specific pet problems or policies (Kingdon, 2011). Building on earlier attempts to understand policy learning within the MSF (Kippin & Cairney, 2022; Kuenzler, 2018), this paper aimed to make a conceptual distinction between different types of learning associated with the MSF’s problem, policy, and politics streams and their interfaces. Based on the assumption of loosely coupled streams, we suggest that problem-focused learning contributes to the problem stream in the context of problem definition, policy-focused learning to the policy stream in the context of softening up, and politics-focused learning to the political stream in the context of powering. In addition, we distinguished three types of learning across streams: policy learning at the intersection of problem and policy streams (in contexts of puzzling), choice-based learning at the interface of the political and either the policy or the problem streams (in the context of strategic de/politicization of problems or strategic commissioning of policies), and coupling-based learning, which enables policy entrepreneurs to update strategies of partial coupling between a preferred policy and politics, a preferred policy and the problem stream, or a priority problem and the politics stream (e.g. blame management by attributing or avoiding responsibility for problems).
48Assuming that situations with great ambiguity are particularly amenable for different types of learning, we applied the combined MSF-learning typology framework to the case of wolf-management policy in the German state of Lower Saxony. The combination of multi-level governance and regulation, competing policy paradigms, and lack of experience of wolf management made this an illustrative case for social learning in policy-making under high ambiguity and uncertainty. The MSF-based learning typology allowed identification not only of instances of learning within and across streams, but also of possible effects on both the separate streams and stream coupling. The case analysis found learning among all key actors in the policy arena and also it could often describe how actors benefited from updating their beliefs.
49The findings suggest that actors in an MSF world have strong incentives to learn in order to improve their chances of successfully influencing the three streams or stream coupling. Learning should therefore be understood as an important mechanism and even as a driver of stream development. However, whether learning among actors leads to improved policies is a different question. The answer depends not just on the overall constellation of problems, policies, and actors, but also on the standards of evaluation. In situations with high ambiguity and uncertainty, problem agendas and policy solutions are likely to be contested. Political polarization might lead actors to adopt a focus on power struggles that dominates any attempt at policy-making. Actors might learn – and often want to learn – how to downplay inconvenient problems, undermine unwanted solutions, or obstruct decision-making processes that would be in the public interest but which affect their interests and positions.
50We suggest four lines for further discussion: First, we should aim to understand whether the different types of learning within streams and across streams are linked to different types of political processes, in the way that Hall’s (1993) distinction between routine and crisis modes of policy-making is linked to different types of learning (instrumental, strategic, goal revision) and to different degrees of policy change. Second, we need a better understanding of the relationship between the different types of learning and their effects on the three streams and on stream coupling. Third, it is necessary to include the strategic dimension of learning in the policy process and to analyze how actors respond to, anticipate, or even manipulate the learning processes of other actors in a policy arena. Finally, it is important to gain a better understanding of the relationship between learning and other internal and external mechanisms that affect the development of the three streams and their de/coupling.