Introduction
The choice of policy instruments and the formation of corresponding preferences are rarely straightforward in the context of complex sustainability problems. In this context, multiple, often conflicting ideologies, so-called policy core beliefs, about what causes the problem and what needs to be solved exist (Munda Reference Munda2009). Trade-offs between satisfying different policy core beliefs have to be made. Moreover, uncertainty remains about the cause of the problem and the consequences of implementing specific policy instruments (Ingold et al. Reference Ingold, Driessen, Runhaar and Widmer2019; Koppenjan and Klijn Reference Koppenjan and Klijn2004). Actors may have different responses to uncertainty and differ in their willingness to take risks. Therefore, we study the role of key actors’ policy core beliefs and risk attitudes in forming their preferences for different policy instruments. To what extent do risk attitudes matter in policymaking in addition to and in relation to policy core beliefs? We aim to contribute to a better understanding of policy instrument preferences under conflicting beliefs and under uncertainty typical of sustainability problems. Building support for new, ambitious policy instruments to address pressing sustainability problems requires an understanding and consideration of how actor preferences for particular instruments are formed (Dermont et al. Reference Dermont, Ingold, Kammermann and -SteffenStadelmann2017).
In our study, we draw on frameworks and theories from policy studies and decision analysis. Policy studies literature has strongly focused on the relationship between policy core beliefs and actors’ instrument preferences (e.g., Dietz et al. Reference Dietz, Dan and Shwom2007; Glaus et al. Reference Glaus, Wiedemann and Brandenberger2022; Ingold et al. Reference Ingold, Stadelmann-Steffen and Kammermann2019; Lahat Reference Lahat2011). Thereby, scholars have drawn on the advocacy coalition framework (ACF; Sabatier and Jenkins-Smith Reference Sabatier and Jenkins-Smith1993), which suggests that policy core beliefs direct the preferences for concrete instruments to solve a problem (Sabatier and Weible Reference Sabatier, Weible and Sabatier2007; Weible et al. Reference Weible, Ingold, Nohrstedt, Henry and Jenkins-Smith2020). However, the ACF has remained vague on how to anticipate actors’ policy instrument preferences in the case of multiple conflicting policy core beliefs (Kammermann and Angst Reference Kammermann and Angst2021). Although understanding the relative importance of certain policy core beliefs to actors seems relevant in this case (Tetlock Reference Tetlock1986), recent studies have operationalized and studied this only to a limited extent. Bolognesi et al. (Reference Bolognesi, Lieberherr and Fischer2024) used, for example, the top objective of a priority ranking to operationalize the policy core belief about water supply policy, assuming that objectives with lower priority levels were less important to instrument preferences. In contrast, Kammermann and Angst (Reference Kammermann and Angst2021) asked actors about the importance of specific objectives in relation to others for operationalizing policy core beliefs. Literature from decision analysis, especially multi-criteria decision analysis (MCDA; Eisenführ et al. Reference Eisenführ, Langer and Weber2010), can contribute to this discussion about how to study instrument preferences under conflicting beliefs. MCDA offers theories on the response to conflicting beliefs, such as multi-attribute utility theory (MAUT; Keeney and Raiffa Reference Keeney and Raiffa1993). From the perspective of a decision analyst following MAUT, complex policy questions about how to address a sustainability problem are viewed as value-based decisions. These decisions involve evaluating trade-offs and responding to uncertainty (Keeney Reference Keeney2006). Evaluating trade-offs entails translating policy core beliefs into specific objectivesFootnote 1 and comparing the consequences of policy instruments for achieving those objectives. It also entails assessing the importance of achieving one objective relative to another, potentially conflicting objective.
Responding to uncertainty is even less theorized in the ACF than dealing with trade-offs. So far, policy studies have overlooked the relationship between attitudes toward the uncertain consequences of policy instruments and instrument preferences. Based on MAUT, actors’ attitudes toward perceived risks in a specific (objective) domain describe how they respond to uncertainty about the policy consequences for that domain. These risk attitudes are thus relevant to anticipating preferences for particular policy instruments under uncertainty. Several policy studies have investigated the relationship between actors’ risk perceptions and preferences for instruments (e.g., Glaus et al. Reference Glaus, Mosimann, Röthlisberger and Ingold2020; McGuire Reference McGuire2015; Verlynde et al. Reference Verlynde, Voltaire and Chagnon2019; Wiedemann Reference Wiedemann2022). Yet risk perception alone does not allow for unambiguous conclusions about the instrument preferences of actors under uncertainty (Weber and Milliman Reference Weber and Milliman1997). An actor may perceive a high risk, but still be willing to take it. Using MAUT, we expect that domain-specific risk attitudes will better predict actors’ behavioral preferences and thus their instrument preferences under uncertainty. Domain-specific risk attitudes not only describe actors’ willingness to take risks, but also involve domain-specific risk perception considerations to some extent (Weber et al. Reference Weber, Blais and Betz2002). To the best of our knowledge, no policy study has addressed the relationship between domain-specific risk attitudes and instrument preferences so far.
To fill the theoretical gaps in policy studies on instrument preferences in complex contexts such as sustainability problems, we combine the ACF with MAUT from MCDA. MCDA is a collection of methods for supporting complex decision problems, including those at the policy level (French and Argyris Reference French and Argyris2018; Gregory and Wellman Reference Gregory and Wellman2001; Keeney Reference Keeney2006; Tsoukias et al. Reference Tsoukias, Montibeller, Lucertini and Belton2013). The established methods evaluate the decision alternatives such as policy instruments in terms of satisfying actors’ policy core beliefs. The evaluation is based on how well alternatives achieve specific decision objectives and how actors value these consequences. MCDA methods allow for identifying conflicting policy core beliefs and integrating the beliefs of different actors when evaluating policies (Munda Reference Munda2009). To date, MCDA has been applied in policy studies mainly as method to identify the policy beliefs of different actors (Ingold Reference Ingold2011). The theories behind the MCDA methods, such as MAUT, were less of a focus. However, MAUT finds implicit reference in policy studies, in the discourse on utility as a driver of individual or organizational behavior and in the maximization of social welfare through public policy (Andrews Reference Andrews, Fischer and Miller2007; Weimer and Vining Reference Weimer, Vining and Boca Raton2017). Hereby, utility functions assign a value to uncertain policy consequences and can be used to evaluate the effectiveness of policies and maximize social welfare in view of actors’ policy core beliefs and uncertainty response. Against this background, we use MAUT in our exploratory policy study. We test whether actors evaluate policy instruments in terms of their utility under uncertainty (response to uncertainty) and in accordance with the relative importance of policy core beliefs (evaluation of trade-offs) and whether this is reflected in their instrument preferences.
We study the relationship between instrument preferences and the relative importance of specific policy core beliefs and domain-specific risk attitudes in Swiss pesticide risk reduction policy. This case is ideal for gaining new insights into this complex relationship: Agricultural pesticide use has become a serious sustainability problem owing to growing concerns about its adverse effects on human health and the environment (e.g., Chagnon et al. Reference Chagnon, Kreutzweiser, Mitchell, Morrissey, Noome and Van der Sluijs2015; Schaub et al. Reference Schaub, Huber and Finger2020; Stehle and Schulz Reference Stehle and Schulz2015; Tang et al. Reference Tang, Lenzen, McBratney and Maggi2021; Zubrod et al. Reference Zubrod, Bundschuh, Arts, Brühl, Imfeld, Knäbel, Payraudeau, Rasmussen, Rohr and Scharmüller2019). Policy choices to reduce these effects shape pesticide regulations, farming systems, and cultivated crops, which are three key factors that explain the agricultural pesticide lock-in, resulting pollution, and adverse health and environmental effects (Hüesker and Lepenies Reference Hüesker and Lepenies2022; Wuepper et al. Reference Wuepper, Tang and Finger2023). These choices require making trade-offs between conflicting policy core beliefs related to human health, environmental protection, agro-economic productivity, and socio-political costs. These choices also require dealing with great uncertainty, both about the adverse effects of pesticides (Möhring et al., Reference Möhring, Bozzola, Hirsch and Finger2020a; Spycher et al. Reference Spycher, Mangold, Doppler, Junghans, Wittmer, Stamm and Singer2018) and about the consequences of different instruments (Kaiser and Burger Reference Kaiser and Burger2022; Pedersen et al. Reference Pedersen, Nielsen and Daugbjerg2020; Schaub et al. Reference Schaub, Ghazoul, Huber, Zhang, Sander, Rees, Banerjee and Finger2023). The Swiss case is interesting because policy choices for pesticide regulation must also take into account a particularly large number of different actors and their instrument preferences. Switzerland is a consensus-oriented direct democracy (Varone and Ingold Reference Varone, Ingold, van Gerven, Rothmayr Allison and Schubert2023). Popular initiatives or, retroactively, referenda and popular votes allow not only policymakers but also other affected or targeted actors to participate in policy processes. To ensure the support of the majority of actors who could potentially become politically active, they are proactively consulted for policymaking. For policymakers, understanding the instrument preferences of these actors is crucial to designing future policies to reduce the adverse effects of agricultural pesticide use —policies that are urgently needed (Candel et al. Reference Candel, Pe’er and Finger2023; Möhring et al., Reference Möhring, Ingold, Kudsk, Martin-Laurent, Niggli, Siegrist, Studer, Walter and Finger2020b).
Theoretically, this study contributes to policy studies by using MAUT in addition to the ACF to better understand the role of conflicting beliefs and uncertainty in forming policy instruments. Methodologically, we contribute by adapting MCDA tools to policy studies. MCDA provides tested and well-established tools to elicit the relative importance of policy core beliefs and the risk attitudes of actors. Empirically, we contribute by discussing the potential of different instruments to regulate agricultural pesticide use and reduce its adverse effects given conflicting policy core beliefs and risk attitudes of actors in Switzerland. The discussion could shed light on why certain policy instruments are supported and what adjustments could lead to broader support from various actors. Insights into how particular policy instruments and their characteristics might be related to actors’ evaluation of trade-offs between conflicting policy core beliefs and their response to uncertainty in the form of domain-specific risk attitudes could be valuable beyond our case study.
Theory and hypotheses
Policy instrument preferences
Often, political negotiations involving diverse actors turn around the concrete design of public policies, including policy objectives and instruments. The formation of actors’ preferences among possible policy instruments is complex and influenced by several criteria, such as the efficiency, effectiveness, equity, manageability, and legitimacy of the instruments (Howlett Reference Howlett2004; Salamon Reference Salamon2002). These criteria are in turn influenced by different instrument dimensions, namely the instruments’ automaticity (re-use of existing polity structures), directness (distance between adopters and implementers), coerciveness (degree of constraint), or visibility (appearance in budgeting or reporting processes) (Salamon Reference Salamon2002). The relevance of each criterion and dimension to instrument preferences is highly contextual (Howlett Reference Howlett2004; Rigby Reference Rigby2007). For example, the legitimacy of subsidies may depend on budget constraints in the respective policy subsystem (Howlett Reference Howlett2004). In our study, the dimensions of directness, automaticity, and visibility appear to be less relevant since the institutional and economic context is similar for all policy instruments discussed for pesticide risk reduction at the national level in Switzerland (Rigby Reference Rigby2007; Salamon Reference Salamon2002). The coerciveness, in turn, is more relevant to instrument preferences as it is strongly associated with ideological debates about policy instrument choice, instruments’ consequences, and associated uncertainties (Salamon Reference Salamon2002; van der Doelen Reference van der Doelen, Bemelmans-Videc, Rist and Vedung1998). Conflicts among actors in subsystems can arise over policy instruments with different degrees of coercion exercised on the target groups of the policy.
There are several different policy instruments available to address the problem that arrived on the political agenda. The degree of coercion of these instruments heavily influenced the most traditional classification into regulatory, market-based, and persuasive instruments (Ingold et al. Reference Ingold, Lieberherr, Schläpfer, Steinmann and Zimmermann2016; van der Doelen Reference van der Doelen, Bemelmans-Videc, Rist and Vedung1998). From the first to the last, the coercion exercised on the target actors, and thus the degree of freedom in these actors’ individual decision-making decreases. The degree of coercion is further refined by subdividing stimulative and repressive instruments in each class (van der Doelen Reference van der Doelen, Bemelmans-Videc, Rist and Vedung1998; see overview in Table 1). In the end, it is the approach (persuasive, market-based, or regulatory) and the strategy (stimulative or repressive) of an instrument that determine the degree of coercion: Regulatory instruments approach a change in the behavior of the target actors causing the problem by using a legal basis. More repressive regulatory instruments for pesticide risk reduction are, for example, the approval or banning of pesticides. More stimulative (less repressive) regulatory instruments are industry agreements on the certification of pesticide-free products. Market-based instruments give more room of maneuver to the target actors and use monetary remunerations or sanctions to incite behavior change. Remunerations such as state subsidies aim to encourage a behavior (stimulative strategy). In contrast, sanctions through a levy tend to be more discouraging (repressive strategy). Examples of policy instruments to reduce pesticide risks include direct payments to farmers for pesticide-free production and incentive taxes on pesticide use based on the predicted adverse effects of the substance. Finally, persuasive instruments nudge target actors to change their behavior. Here, information campaigns, public appeals, education, or, in the extreme, propaganda aim to raise awareness of a problem and to promote behavior change. Nudging can be stimulative, or it can also be repressive if certain information is considered to be standard (e.g., training standard). We know from pesticide risk reduction policy, for instance, that there are information campaigns on risk prevention and a requirement for training in agricultural pesticide use. Among other dimensions, less relevant in our case, the degree of coercion is used to make assumptions about the effectiveness of these policy instruments in satisfying certain policy core beliefs, with preference given to more effective instruments (van der Doelen Reference van der Doelen, Bemelmans-Videc, Rist and Vedung1998). However, the definition of effectiveness, and thus policy instrument preferences, depend on individual core beliefs about what needs to be achieved and individual attitudes toward uncertainty in satisfying those beliefs.
Table 1. Examples and characteristics of policy instruments (adapted table from van der Doelen Reference van der Doelen, Bemelmans-Videc, Rist and Vedung1998). The approach (persuasive, market-based, or regulatory) and the strategy (stimulative or repressive) of the instruments determine their degree of coercion. The instrument examples stem from Swiss pesticide risk reduction policy
Evaluation of trade-offs given conflicting policy core beliefs
The ACF illuminates policy processes and policy change, with a focus on advocacy coalitions, their belief system, and policy-oriented learning therein. The ACF speaks of a three-tiered belief system, which includes preferences for policy instruments (Sabatier and Weible Reference Sabatier, Weible and Sabatier2007; Weible et al. Reference Weible, Ingold, Nohrstedt, Henry and Jenkins-Smith2020): Deep core beliefs are fundamental values rooted in the actors’ religious and cultural ideology. Policy core beliefs, in turn, are the translation of the deep core beliefs into concrete, issue-related beliefs in policy subsystems. They refer to beliefs about what needs to be solved and guide secondary beliefs, actors’ preferences for technicalities, and instrumental aspects of policy design, typically including preferences for policy instruments. The hierarchy of the belief system would thus tell us that (deep and policy) core beliefs are an important driver for understanding and explaining actors’ preferences for certain policy instruments over the others. According to the ACF, actors with similar beliefs form coalitions in which they coordinate and share resources to strengthen their advocacy in negotiations (Weible et al. Reference Weible, Ingold, Nohrstedt, Henry and Jenkins-Smith2020). Negotiations are a fundamental part of policy processes, where actors agree or disagree on the choice of instrument(s) to address a problem or find compromises. In these negotiations, actors strive for preferred instruments that most effectively satisfy their policy core beliefs (rational choice paradigm; Geels Reference Geels2010). Actors therefore make assumptions about the probability that a particular instrument will satisfy their core beliefs, given the actions of others and the institutional or economic context (Andrews Reference Andrews, Fischer and Miller2007; Griggs Reference Griggs, Fischer and Miller2007). To reach an agreement, actors may have to compromise on preferred policy instruments to accommodate the core beliefs of their negotiation partners while maximizing the utility of newly proposed instruments in light of their own core beliefs (Metz et al. Reference Metz, Lieberherr, Schmucki and Huber2021). However, the (deep or policy) core beliefs of actors remain rather stable (Sabatier and Weible Reference Sabatier, Weible and Sabatier2007), and different core beliefs need to be considered to satisfy the various actors involved (e.g., Rosenow et al. Reference Rosenow, Kern and Rogge2017). Yet the ACF remains vague about how actors account for multiple conflicting beliefs when forming instrument preferences.
In MCDA, decision analysts support actors in complex decision problems that require making trade-offs between multiple conflicting decision objectives, such as policy core beliefs (Eisenführ et al. Reference Eisenführ, Langer and Weber2010). One theory to decide about the “best” decision alternative among policy instruments when there are multiple policy core beliefs is MAUT (French and Argyris Reference French and Argyris2018; Reichert et al. Reference Reichert, Langhans, Lienert and Schuwirth2015). Following MAUT, the utility of a policy instrument (decision alternative) for an actor is a function of all relevant consequences of the instrument and the value that the actor associates with these (uncertain) consequences (Keeney Reference Keeney2006; Keeney and Raiffa Reference Keeney and Raiffa1993). Values or the utility function are not a natural human expression, but a theoretical model and approximation of how policy core beliefs may translate into behavioral preferences (Beinat Reference Beinat and Beinat1997; Eisenführ et al. Reference Eisenführ, Langer and Weber2010; Weimer and Vining Reference Weimer, Vining and Boca Raton2017). In public policy, where decisions must ideally be transparent and plausible to be legitimate, the few axioms of the theoretical model (MAUT) are considered justified (French and Argyris Reference French and Argyris2018; Reichert et al. Reference Reichert, Langhans, Lienert and Schuwirth2015). According to the model, actors first translate their policy core beliefs into concrete policy objectives. Second, the consequences of a policy instrument for these objectives are assessed. Third, the actors assign values to the consequences according to the satisfaction of their policy core beliefs, i.e., how well an instrument achieves a particular objective. When there are multiple policy core beliefs, one policy instruments can usually not satisfy all of them at the same time and to the same degree (Glaus Reference Glaus2021). For example, a ban on agricultural pesticides may be effective in protecting the environment but is not designed to secure farm incomes. In this situation, actors evaluate the policy instrument and resulting trade-offs based on which of their policy core beliefs is more important to satisfy. In the end, the utility of a policy instrument is an aggregation of these considerations.Footnote 2 The consequences of an instrument for each policy core belief are evaluated, weighed, and aggregated into an overall utility estimate. MAUT expects actors to prefer the policy instruments with the highest utility to them (Keeney and Raiffa Reference Keeney and Raiffa1993). Similar to the ACF, MAUT suggests that the policy core beliefs are ideally the main drivers of instrument preferences and decision-making. The values that actors assign to policy instruments as measures of utility are derived from their policy core beliefs and, in the case of conflicting policy core beliefs, the relative importance of those beliefs (“value-focused thinking”; Keeney Reference Keeney1982). This leads to our first hypothesis (H1):
H1: The degree to which actors are likely to prefer a particular policy instrument depends on the relative importance they associate with specific policy core beliefs.
Domain-specific risk attitudes as response to uncertainty
Although uncertainty is inherent in almost all public policy problems or decisions (Koppenjan and Klijn Reference Koppenjan and Klijn2004), the ACF remains vague in how actors respond to it when forming their preferences for policy instruments. When policy analysts consider uncertainty, they assume that actors who perceive high risk in effectively achieving their policy core beliefs are more likely to prefer policies that reduce that risk (e.g., Glaus et al. Reference Glaus, Mosimann, Röthlisberger and Ingold2020; McGuire Reference McGuire2015; Verlynde et al. Reference Verlynde, Voltaire and Chagnon2019; Wiedemann Reference Wiedemann2022). This assumption is rooted in psychology: Because policy choice can have serious consequences, actors use risk assessments to hypothetically or empirically evaluate uncertainty, gain insight into the probable consequences (risks or benefits), and make decisions (Slovic Reference Slovic1987). Slovic (Reference Slovic1987) describes risk perception as an intuitive risk assessment of an actor, i.e., a subjective risk judgment based on experience, evidence, or heuristics. Policy analysts use risk perception to anticipate actors’ instrument preferences under uncertainty. However, this theoretical basis lacks in terms of behavioral conclusions as actors’ behavioral preferences in situations of uncertainty do not necessarily correlate with their risk perception: Actors can perceive a risk as high but still want to take it because of anticipated benefits and their willingness to take the risk (Weber and Milliman Reference Weber and Milliman1997).
In MAUT, the actors’ responses to uncertainty through their attitudes toward associated risks play an important role in deciding among alternatives with uncertain consequences (Eisenführ et al. Reference Eisenführ, Langer and Weber2010; Keeney and Raiffa Reference Keeney and Raiffa1993; Reichert et al. Reference Reichert, Langhans, Lienert and Schuwirth2015). MAUT postulates that the preference for a decision alternative such as a policy instrument is based on the subjective expected utility associated with the alternative—that is, how effective the instrument satisfies policy core beliefs and achieves specific objectives given uncertainty (Keeney and Raiffa Reference Keeney and Raiffa1993). MAUT expects actors to prefer the instrument that has the highest expected utility associated with its probabilistic consequences. It is thus necessary to specify utility as a function of the consequences of a policy instrument, the probability of those consequences occurring, and the values assigned by actors to the uncertain consequences. The utility function represents how actors respond to uncertainty, their risk attitudes, and how these attitudes may translate into behavioral preferences (Keeney Reference Keeney2006). The function models the values that actors assign to possible lotteries between the probabilistic consequences of policy instruments for a specific policy core belief. In other words, actors’ “risk attitudes” capture how willing actors are to take a risk and accept the consequences of an instrument when those consequences are uncertain (Keeney and Raiffa Reference Keeney and Raiffa1993). The actors’ risk attitudes can be domain-specific (Weber et al. Reference Weber, Blais and Betz2002). Thus, the willingness to take risks or the acceptance of uncertainty depends on the policy core belief at stake. For example, an actor could be relatively open to accepting uncertain consequences for agro-economic productivity, but reluctant to do so for environmental protection. Domain-specific risk attitudes are expected to better approximate behavioral preferences than risk perceptions alone. Psychological literature suggests that both risk perception and risk attitudes must be considered to understand decisions and preferences under uncertainty (Weber and Milliman Reference Weber and Milliman1997). Domain-specific risk attitudes result from considering risk benefits, the degree to which beliefs are expected to be satisfied (expected utility), and perceived risks, the degree of potential loss (Weber et al. Reference Weber, Blais and Betz2002). Thus, a risk-tolerant attitude does not automatically imply acceptance of negative consequences, but rather a belief that there is a low risk that the consequences will be worse than the status quo. Taking the above example, a ban on pesticides might have a positive environmental impact with certainty, but the consequences for national food security are uncertain. In contrast, a less repressive and coercive instrument, such as direct payments for pesticide-free farming, might be more uncertain in protecting the environment because of uncertainty about the effectiveness of the instrument, which includes uncertainty about how many farmers will make use of such payments. Whether actors prefer the ban or the direct payments depends on their risk attitudes in the domain of agro-economic productivity and environmental protection. This leads to our second hypothesis (H2):
H2: The degree to which actors are likely to prefer a particular policy instrument depends on their domain-specific risk attitudes.
Methodology
Case
Swiss pesticide risk reduction policy is embedded in a consensus-oriented direct democracy and thus involves various actors and their instrument preferences (cf. Varone and Ingold Reference Varone, Ingold, van Gerven, Rothmayr Allison and Schubert2023). Involved are the political parties represented in the federal parliament, the federal government, and associated executive agencies covering different jurisdictions (e.g., environmental protection, agriculture, food safety). But national trade associations of farmers, beekeepers, fishermen, water suppliers, the pesticide industry, or retailers, consumer and environmental organizations, scientific institutes, and agricultural experts are also involved (see Data collection section). Recent policy processes to reduce (agricultural) pesticide risks have largely been influenced by such non-governmental actors (Huber et al. Reference Huber, Benni El and Finger2024; Huber and Finger Reference Huber and Finger2019). Different popular and parliamentary initiatives have put the issue of agricultural pesticide use at the top of the political agenda in Switzerland. In 2017, the national action plan “Risk Reduction and Sustainable Use of Plant Protection Products” was launched. The plan proposed about 50 action points to reduce pesticide risks by 50% until 2027 (Schweizer Bundesrat 2017). These action points mainly extend existing policy instruments, such as supporting pesticide-free production systems through higher direct payments or regulating the use of pesticides through a training requirement. In response, two popular initiatives requested more ambitious action, suggesting stricter cross-compliance standards and a ban on pesticides, respectively (Finger Reference Finger2021). As a compromise, the Swiss federal assembly adopted the parliamentary initiative “Reduce the Risk Associated with the Use of Pesticides” (Pa.Iv. 19.475) in 2021 (Schweizer Bundesversammlung 2019). This initiative induced inter alia an obligation to report agricultural pesticide use, established a national risk evaluation system and the monitoring of set targets, and linked pesticide approval to observed effects. Despite this initiative, the search for policy instruments to reduce the adverse effects of agricultural pesticide use remains relevant, as trade-offs between agro-economic productivity and environmental protection persist (Candel et al. Reference Candel, Pe’er and Finger2023; Möhring et al., Reference Möhring, Ingold, Kudsk, Martin-Laurent, Niggli, Siegrist, Studer, Walter and Finger2020b).
Swiss pesticide risk reduction policy is an ideal policy subsystem for studying the role of trade-offs between conflicting policy core beliefs and of domain-specific risk attitudes in forming preferences for policy instruments. Trade-offs exist, for instance, between environmental protection and agro-economic productivity (Schmidt et al. Reference Schmidt, Mack, Möhring, Mann and BenniEl2019). While farmers use pesticides to protect their crops and meet market requirements (Bakker et al. Reference Bakker, Werf, Tittonell, Wyckhuys and Bianchi2020), the pesticide use exposes them and other non-target organisms to risks (Graczyk et al. Reference Graczyk, Hopf, Mediouni, Canu, Sanvido, Schmid and Berthet2018; Humann-Guilleminot et al. Reference Humann-Guilleminot, Binkowski, Jenni, Hilke, Glauser and Helfenstein2019). With respect to these trade-offs, different actors have different priorities. Two coalitions and an intermediate group exist among the actors in Swiss pesticide risk reduction policy (Wiget Reference Wiget2024). There are those actors who are less concerned about agro-economic productivity, fully supporting repressive regulatory instruments to reduce pesticide risks. Others are concerned about environmental protection, food security, and the economic viability of farming. They do not unconditionally support additional regulations and clearly oppose tax incentives to govern pesticide use and risks. The actors have to from their instrument preferences under substantive and strategic uncertainty.Footnote 3 Numerous ways of exposure, various environmental factors, missing analytics, misleading indicators, and the delayed occurrence of effects make risk assessments of agricultural pesticide use difficult (Kiefer et al. Reference Kiefer, Bader, Minas, Salhi, Janssen, Guntenvon and Hollender2020; Möhring, Bozzola, et al. Reference Möhring, Ingold, Kudsk, Martin-Laurent, Niggli, Siegrist, Studer, Walter and Finger2020; Möhring et al. Reference Möhring, Gaba and Finger2019; Riedo et al. Reference Riedo, Wettstein, Rösch, Herzog, Banerjee, Büchi, Charles, Wächter, Martin-Laurent and Bucheli2021; Schulz et al. Reference Schulz, Bub, Petschick, Stehle and Wolfram2021; Spycher et al. Reference Spycher, Mangold, Doppler, Junghans, Wittmer, Stamm and Singer2018). In addition, reactions to policy instruments are heterogeneous, as many actors and factors along the food value chain influence agricultural pesticide use (Bakker et al. Reference Bakker, Werf, Tittonell, Wyckhuys and Bianchi2020; Kaiser and Burger Reference Kaiser and Burger2022; Pedersen et al. Reference Pedersen, Nielsen and Daugbjerg2020; Schaub et al. Reference Schaub, Ghazoul, Huber, Zhang, Sander, Rees, Banerjee and Finger2023).
Data collection
The study data stem from an online survey and a focus-group workshop. In summer 2022, we conducted an online survey among 54 key actors in Swiss pesticide risk reduction policy with a response rate of 85% (N = 46; see also Wiget Reference Wiget2024). We considered decision-, position-, and reputation-relevant actors (Hoffmann-Lange Reference Hoffmann-Lange, Best and Higley2018) and identified the actors using policy documents of the national action plan and parliamentary initiative, previous stakeholder analyses (Metz et al. Reference Metz, Leifeld and Ingold2019; Metz et al. Reference Metz, Lieberherr, Schmucki and Huber2021), and interviews (with four experts from administration and non-governmental organizations). We surveyed representatives in leading positions about the policy instrument preferences of their respective organizations or institutions (see actors list in appendix, Table A-1). In summer 2023, we conducted a follow-up workshop with a representative sub-sample (N = 24). The actors were selected based on a social network analysis considering the type, relevance, and coalitions of the actors (Wiget Reference Wiget2024). During the workshop, we elicited their policy core beliefs and risk attitudes. Finally, we studied the relationship of instrument preferences with policy core beliefs and risk attitudes based on the data from those actors who participated in both the survey and the workshop. Additional survey data were used for control variables.
Data operationalization
We operationalized the dependent variable, the policy instrument preferences, by measuring the actors’ level of support for 15 examples of different policy instruments (Table 2). We considered regulatory, market-based, and persuasive instruments aiming at pesticide risk reduction in a stimulatory or repressive manner (Table 1). All the examples we used were discussed in the context of the national action plan and the Pa.Iv. 19.475 (see also Wiget Reference Wiget2024). We validated the instrument examples used with experts and based on a comprehensive, context-specific collection of instruments by Lee et al. (Reference Lee, Uylden and Runhaar2019). In the survey, we asked: “How supportive is your organization of each of these policy instruments?” The actors had to rate their support on a Likert-scale ranging from 1 (not at all) to 5 (fully).
Table 2. Policy instruments to operationalize the actors’ instrument preferences (adapted table from Wiget Reference Wiget2024)
We elicited the actors’ policy core beliefs (1st explanatory variable) in the workshop with tools from MCDA. The policy core beliefs and their relative importance were operationalized as relative weights assigned to the achievement of 16 objectives covering four domains (Table 3). The objectives were compiled based on the same policy documents used to identify the actors. In addition, we used our system knowledge and the actors’ feedback on objectives in the survey (see also Wiget Reference Wiget2024). In an individual pen-and-paper questionnaire (see SI-1.2), we asked the workshop participants to rank the 16 objectives according to the importance of achieving them. The ranking was used to calculate the weights (Kuller et al. Reference Kuller, Beutler and Lienert2023). Although weight approximation can induce biases (Riabacke et al. Reference Riabacke, Danielson and Ekenberg2012), the used rank sum calculation approach is a good substitute for more cognitively demanding weight elicitation methods (Roberts and Goodwin Reference Roberts and Goodwin2002). Objectives’ weights are range sensitive. For example, actors might weight the objective “cost fairness” differently depending on the magnitude of external costs. Hence, weights are specific to the context, i.e., to information about the best and worst possible consequences across all considered policy instruments. To avoid a range insensitivity bias when eliciting weights, the ranges must be provided for each objective (Montibeller and von Winterfeldt Reference Montibeller, von Winterfeldt, Dias, Morton and Quigley2018). For each objective, we thus identified suitable indicators, the status quo, and the best/worst possible consequences together with experts (see details in the supplementary information section SI-1). In our models, we included the most variant policy core beliefs of each domain (see appendix, Table A-1), as the variance in the relative importance indicates divisive beliefs between actors (Karimo et al. Reference Karimo, Wagner, Delicado, Goodman, Gronow, Lahsen, Lin, Schneider, Satoh and Schmidt2023).
Table 3. Policy objectives to operationalize the actors’ policy core beliefs (adapted table from Wiget Reference Wiget2024). For information on the indicators for each objective and the best and worst possible consequences for them, see Table SI-1.1
We also elicited the actors’ risk attitudes (2nd explanatory variable) in the workshop in groups of five to seven actors. Following common practice in MCDA, risk attitudes were elicited with a basic reference lottery using the bisection version of the variable certainty equivalent method (Eisenführ et al. Reference Eisenführ, Langer and Weber2010). In an iterative questioning process, the actors compared the uncertain consequences of a hypothetical policy instrument A with the certain consequences of a hypothetical instrument B (for the detailed process, see SI-2). They had to indicate the certainty equivalent (CE) at which they were indifferent between preferring A or B. The CE is the level to which the objective must be achieved with certainty by instrument B so that it is equivalent to instrument A and its uncertain consequences. It indicates the utility of the policy instrument A, where there is an equal chance that the objective will be either fully achieved or not achieved at all (i.e., 50/50 chance lottery). We elicited the risk attitudes in each of the four domains based on the objective that the respective group had considered as most important in that domain. Next, we normalized the resulting CEs on a scale between 1 (i.e., instrument B fully achieves the objective) and 0 (i.e., instrument B does not achieve the objective at all). Depending on the elicited CE, the actors were identified as risk-averse (CE < 0.5), risk-neutral (CE = 0.5), or risk-tolerant (CE > 0.5) of uncertain consequences for achieving the objectives in the corresponding domain.
As control variables, we included the collaboration partner support index (CPSI) and the actor type. To calculate the CPSI, we collected data on the actors’ collaboration partners in the survey (see also Wiget Reference Wiget2024). The CPSI operationalizes the instrument preferences of an actor’s collaboration partners. The coordination and exchange with other actors in policy processes can influence individual instrument preferences (e.g., Metz et al. Reference Metz, Leifeld and Ingold2019). Actors might adapt their policy instrument preferences to coordinate with allies with similar policy core beliefs but different instrument preferences or to find a compromise in negotiations with opponents (Metz et al. Reference Metz, Lieberherr, Schmucki and Huber2021; Sabatier and Weible Reference Sabatier, Weible and Sabatier2007). We calculated the CPSI for each actor and policy instrument: The support ratings of an actor’s collaboration partners for a given policy instrument were averaged and divided by the maximum possible support of the respective group of partners. The resulting index can take a value between 0 (none of the partners support the instrument) and 1 (all partners fully support the instrument). Furthermore, we differentiated between actors with an administrative role, particular interests, or a scientific background. Actor types differ in their experiences and role. These differences can influence their policy instrument preferences (e.g., Metz and Leifeld Reference Metz, Leifeld, Bréthaut and Schweizer2018). Our sample included six administrative actors, thirteen interest groups, and five science actors. As administrative actors can be expected to have less partisan policy instrument preferences than other actors (Jenkins-Smith et al. Reference Jenkins-Smith, Nohrstedt, Weible and Ingold2017), they served as the reference actor type in our analysis.
Data analysis
Our study aimed to understand actors’ preferences for policy instruments in the face of multiple conflicting policy core beliefs and uncertainty. We used Bayesian ordinal logistic regression (OLR) models (Bürkner and Vuorre Reference Bürkner and Vuorre2019) to assess the relationship between policy instrument preferences and the explanatory variables (Figure 1). Five separate models were created for regulatory, market-based, and persuasive policy instruments and their stimulative and repressive strategies (Table 1). In the case of the persuasive policy instruments, we did not distinguish between stimulative and repressive strategies. There was no clear example of a repressive persuasive instrument in our case. In each model, we controlled for the differences among individual examples for a particular instrument (Table 2) by comparing a reference example to all others. As references we used additional pesticide approval requirements (repressive regulatory instrument), voluntary industry agreements (stimulative regulatory instrument), risk-based tax incentives on pesticides (repressive market-based instrument), direct payments for low-emission application techniques (stimulative market-based instrument), and improved information access (persuasive instrument), respectively. OLR (Agresti Reference Agresti2010) was employed, as the dependent variable was polytomous (i.e., the level of support for an instrument was ranked in order: not at all < mostly not < neutral < mostly < fully).Footnote 4 For modeling, we centered all continuous variables and applied the brms() function of the Stan package (Bürkner Reference Bürkner2017; Reference Bürkner2018) in R. Owing to limited prior knowledge, we used flat uniform priors for all explanatory variables. For model diagnostics, we conducted graphical posterior predictive checks using the ppc_bars() function of the bayesplot package (Gabry et al. Reference Gabry, Simpson, Vehtari, Betancourt and Gelman2019) and Rhat metrics (Bürkner and Vuorre Reference Bürkner and Vuorre2019). We also conducted a descriptive analysis to check the model and understand how differences in the preferences for policy instruments relate to specific policy core beliefs, risk attitudes, the CPSI, and the type of actor (see details and results in SI-3).
Figure 1. Causal model for policy instrument preferences. Blue: the policy instrument preference of an actor as the dependent variable; green: specific policy core beliefs (CB) and risk attitudes (RA) in the domains of human health (Health), environmental protection (Env), agro-economic productivity (Agro), and socio-political costs (Socio) as explanatory variables; white: the collaboration partner support index (CPSI), actor type, and policy instrument specificity as control variables that need to be adjusted for in the Bayesian ordinal logistic regression models.
Results
The Bayesian OLR models converged well, with Rhat convergence measures consistently being 1. The effective sample size was above 2,500 for all model parameters except for the stimulative market-based instrument model. Here, the effective sample size was above 1,300. All effective sample sizes were satisfactory. The posterior predictive checks revealed that the simulated data (i.e., posterior predictive distribution) fitted the observed data (distribution) quite well (Figure A-2). The results of the models were also consistent with those of the descriptive analysis (see SI-3).
The results of our model for repressive regulatory policy instruments showed a relationship between the actors’ specific policy core beliefs about environmental protection and their instrument preferences. Actors who weighted the protection of non-target organisms higher were more likely to prefer additional pesticide approval requirements (Figure 2: violet CB (Env) estimate with 95% credibility interval). In other words, holding all other variables constant, the model estimated higher odds of fully supporting repressive regulatory instruments with an increasing weighting of environmental protection (Figure 3). The model results also revealed relationships between the preferences for repressive regulatory instruments and the actors’ risk attitudes in the domains of environmental protection, agro-economic productivity, and socio-political costs. For example, the actors who were more risk-tolerant of uncertain consequences for the protection of non-target organisms and water bodies or for food security and the economic viability of farming were more likely to prefer additional pesticide approval requirements. Moreover, science actors were less likely to prefer additional pesticide approval requirements than administrative actors (Figure 4: difference in full instrument support between the actor types).
Figure 2. Result of the Bayesian ordinal logistic regression models. The upper point plot shows the estimates (x-axis) for the model intercepts, followed by a plot showing the estimates of the model parameters for the explanatory variables (CB: policy core beliefs; RA: risk attitudes). The last three plots show the parameter estimates for the control variables, the collaboration partner support index (CPSI), actor type (SC: science actors; IG: interest groups), and policy instrument specificity. Model estimates that differ from zero, including their 95% credibility interval, indicate a systematic and directional relationship between the corresponding variable and the actors’ policy instrument preference for regulatory (violet), market-based (dark blue: tax incentives; green: subsidies), cooperative (orange), or persuasive (yellow) policy instruments. Reading example (black CB (Env) estimate): Actors who weighted the protection of the environment higher were more likely to prefer regulatory instruments with a 95% credibility interval (for details on the parameter estimates and credibility intervals, see Table A-2).
Figure 3. Marginal effect of policy core beliefs about the protection of the environment, and specifically non-target organisms (NTO), on the degree to which actors are likely to prefer a regulatory instrument such as additional pesticide approval requirements. The line plot illustrates the predicted probability of different levels of support (y-axis) depending on the weight the actors assign to the protection of non-target organisms (x-axis). The weights are shown as deviations from the average weight across all actors. The shaded areas indicate the 95% credibility intervals of the model predictions (for details on the marginal effects of other explanatory variables, see SI-4).
Figure 4. Marginal effect of the actor type (AD: administrative actors; SC: science actors; IG: actors with particular interests, so-called interest groups) on the degree to which an actor is likely to prefer a regulatory instrument such as additional pesticide approval requirements. The point plot illustrates the predicted probability of different levels of support (y-axis) depending on the type of actor (x-axis) with a 95% credibility interval (for details on the marginal effects of other control variables, see SI-4).
For stimulative regulatory instruments (cooperative instruments), our model showed a relationship between the policy core beliefs about the relative importance of low agricultural workload and the actors’ policy instrument preferences. If low workload was relatively important, the actors were more likely to prefer voluntary industry agreements between bulk buyers and farmers (Figure 2: orange CB (Health) estimate). We also noted a relationship between the actors’ risk attitudes toward uncertain consequences for the health of applicants or consumers and their preferences for cooperative instruments (Figure 2: orange RA (Health) estimate). In addition, interest groups were more likely to prefer voluntary industry agreements than administrative actors if they did not differ in other matters (Figure 2: orange Actor type (IG) estimate).
In the case of repressive market-based instruments, we found a relationship between the actors’ policy core beliefs about agro-economic productivity and their preferences for tax incentives, considering a slightly smaller credibility interval than 95% (Figure 2: dark blue CB (Agro) estimate). Actors who placed higher weight on the economic viability of farming were less likely to prefer tax incentives on pesticides based on health and environmental risks. We also found a relationship between the preferences for stimulative market-based instruments and the actors’ risk attitudes in the domain of agro-economic productivity (Figure 2: green RA (Agro) estimate). Actors who were less risk-tolerant concerning the economic viability of farming were more likely to prefer direct payments for low-emission application technology. Controlling for the specific instrument examples, we found that all the actors more likely preferred subsidies for research on pesticide effects and crop protection practices than direct payments (Figure 2: green Policy (SubResearch) estimate).
Persuasive instruments received unanimous support from all actors (Figure SI-3.1). However, the degree to which the actors preferred the instruments depended on the relative importance they assigned to policy core beliefs about environmental protection. Actors who put more weight on the protection of non-target organisms were more likely to prefer improved access to information on pesticide effects and crop protection practices (Figure 2: yellow CB (Env) estimate). Table 4 summarizes our results.
Table 4. Summary of the Bayesian ordinal logistic regression model results for each of the five instrument categories. The table summarizes the relationships (R) of domain-specific policy core beliefs (CB), risk attitudes (RA), and control variables (CO) with the preferences for policy instruments. The (+) and (–) signs indicate the direction of the systematic relationship. Reading example: Actors who weighted the protection of the environment higher were more (+) likely to prefer repressive regulatory instruments
Discussion
The results supported our first hypothesis that the actors’ preferences for particular policy instruments depend on the relative importance they assigned to specific policy core beliefs. We found systematic relationships between the relative importance assigned to environmental protection, agro-economic productivity, or agricultural workload and the preferences for all instruments except stimulative market-based subsidies. The following three examples illustrate this result: First, the higher the actors weighted the protection of non-target organisms, the more likely they were to prefer additional pesticide approval requirements. This result is consistent with the findings of Hunka et al. (Reference Hunka, Meli, Palmqvist, Thorbek and Forbes2015) that environmental concerns are one of the most important drivers of changes in risk assessment regulations for pesticides in the European Union. The key role of environmental concerns as drivers of pesticide approval regulations also appears to be true for Switzerland. Similar to the findings in Swiss agriculture policy (Metz et al. Reference Metz, Lieberherr, Schmucki and Huber2021), the actors in our study about Swiss pesticide risk reduction policy all assigned relatively high importance to the protection of human health and the environment. This shared policy core belief seems to be a key explanatory factor for the compromise reached in 2021 with the parliamentary initiative on linking pesticide approval to observed effects of pesticides in Switzerland (Schweizer Bundesversammlung 2019). Second, the more important the economic viability of farming was to actors, the less likely they preferred tax incentives on pesticides based on health and environmental effects. This result is consistent with earlier findings in Swiss water policy, where industrial and agricultural actors associated market-based instruments such as tax incentives to address micropollution in water bodies with rising costs for themselves (Metz and Leifeld Reference Metz, Leifeld, Bréthaut and Schweizer2018). Third, actors’ policy core beliefs on the relative importance of agricultural workload played a role in their preferences for voluntary industry agreements. This result is in line with the finding of Kaiser and Burger (Reference Kaiser and Burger2022) that depending on the crop protection practice, Swiss farmers’ participation in voluntary pesticide-low/free production programs is more likely if the program does not require additional work. As a conclusion to these results, identifying the individual role of specific policy core beliefs in relation to potential consequences of policy instruments allows for developing consensus solutions. For example, for repressive market-based instruments targeting the use of hazardous pesticides, Finger et al. (Reference Finger, Möhring, Dalhaus and Böcker2017) suggested a redistribution of tax revenues to financially support pesticide-low/free agriculture. Ideally, the redistribution would reduce the negative consequences for agro-economic productivity and thus increase the support for tax incentives. Ultimately, we found no pattern between the coerciveness of policy instruments and the relative importance of specific policy core beliefs. For example, the relative importance of environmental protection played a role in preferences for both persuasive and regulatory instruments. Although coercive policy instruments may be perceived as more efficient, the individual question remains as to which policy core belief is effectively satisfied and with what consequences for other policy core beliefs. The consequences of a stimulative instrument may be less costly to a high-priority (relatively important) policy core belief, even though less effective at satisfying another policy core belief.
Our results also supported the second hypothesis that the actors’ preferences for particular policy instruments depend on their domain-specific risk attitudes. We found relationships between the actors’ preferences for regulatory and stimulative market-based instruments and specific risk attitudes in all four domains. For example, in the case of repressive regulatory instruments, actors who were more risk-tolerant of uncertain environmental consequences regarding the protection of non-target organisms and water bodies were more likely to prefer additional pesticide approval requirements compared with actors who were less willing to accept such uncertainties. This finding might be counterintuitive at first sight. However, risk-tolerance should not be equated with accepting negative environmental consequences, but rather with believing that there is a low risk that the consequences are likely to be worse than the status quo. Similar observations were made among a sample of farmers (C. McCallum, personal correspondence, 04.10.2024): Farmers who did not use pesticides were more willing to take risks with respect to the environment. Current environmental risk assessments of pesticides have been criticized for focusing on individual substances rather than on substance mixtures and for misrepresenting exposure dynamics (Topping et al. Reference Topping, Aldrich and Berny2020). Therefore, some actors believe that improving the approval process will likely be no worse than the status quo. Under the current authorization regime, an increasing number of agricultural pesticides are being withdrawn from the market due to new evidence of health or environmental risks. According to Topping et al. (Reference Topping, Aldrich and Berny2020), an advanced risk assessment based on an integrated systems approach would imply that pesticide authorization would be time- and context-specific, rather than a binary decision between “safe” and “unsafe”. The remaining uncertainties and risks would have to be transparently communicated. In this respect, it is perhaps less surprising that those actors in our study who were more willing to accept (transparently communicated) risks preferred an advancement of the pesticide approval process. As another example, the actors’ risk attitudes in agro-economic productivity played a key role in their preferences for stimulative market-based subsidies. Actors who were more risk-tolerant of uncertain consequences for food security and the economic viability of farming were less likely to prefer direct payments for low-emission application techniques or pesticide-low/free production. In Switzerland, direct payments are used to compensate for not using pesticides and thus taking the increased risk of yield and income losses. These payments are a key decision factor for farmers to participate in pesticide-low/free production programs, such as the Swiss Extenso program (Mack et al. Reference Mack, Finger, Ammann and Benni El2023; Möhring and Finger Reference Möhring and Finger2022). Not surprisingly, actors who were less willing to accept uncertain consequences for farm income and yield tended to prefer direct payments. Based on our results, it remains to be explored whether the domain-specific risk attitudes of actors play a more decisive role in forming preferences for more coercive but less repressive instruments. We found that domain-specific risk attitudes played a role in preferences for regulatory rather than persuasive instruments using stimulative rather than repressive policy strategies. However, this trend needs further empirical verification. Our explanation for this trend is that the uncertainty in the consequences of policy instruments decreases as the repressiveness of the instrument increases. This is also reflected in the association between instruments’ effectiveness and repressiveness (van der Doelen Reference van der Doelen, Bemelmans-Videc, Rist and Vedung1998). Because of their optional adaptability and uncertainty in the response of the target actors, stimulative instruments are perceived as less effective, with less certain consequences. However, the consequences of policy instruments are inherently uncertain, and the potential consequences of regulatory instruments are typically more severe than those of persuasive instruments. Therefore, potential benefits and risks and thus the relevance of domain-specific risk attitudes, even for non-targeted policy core beliefs, increase with the degree of coercion.
Regarding our control variables, the instrument specificity and actor type played a role only in the case of preferences for market-based subsidies and for regulatory and cooperative instruments, respectively. In the case of market-based instruments, the actors tended to prefer subsidies for research on the effects of agricultural pesticides and alternative crop protection practices over direct payments to farmers. We did not find clear differences in the preferences for particular examples of other instruments. We therefore applied our model results to all policy instrument examples in the respective categories except for market-based subsidies. In the case of regulatory or cooperative instruments, the instrument preferences of some actor types clearly differed. For example, interest groups were more likely to prefer cooperative policies than administrative actors. We explain this result by stating that voluntary industry agreements allow interest groups to have more influence on policy design than other instruments (Metz and Leifeld Reference Metz, Leifeld, Bréthaut and Schweizer2018). Moreover, science actors were less likely to prefer additional pesticide approval requirements than administrative actors. One explanation for this finding could be that scientists criticize the pesticide approval process, particularly the decision-making process after risk assessment, for not being transparent (Hunka et al. Reference Hunka, Meli, Palmqvist, Thorbek and Forbes2015). Our findings are consistent with those of Bolognesi et al. (Reference Bolognesi, Lieberherr and Fischer2024): The role of the actor type in policy instrument preferences depends on the policy instrument. Finally, the fact that we did not find a relationship between the preferences of actors and those of their collaboration partners may be attributable to the institutionalized context. In the Swiss federal democracy, evaluations to prepare new policies are highly institutionalized, and the relevant information is made available to all actors (Varone and Ingold Reference Varone, Ingold, van Gerven, Rothmayr Allison and Schubert2023). Functional interdependencies and rules for consensus decision-making also support collaboration between opposing actors. If the information does not match one’s policy core beliefs or instrument preferences, this can be a barrier to the systematic uptake of the information (Hofmann et al. Reference Hofmann, Ingold, Stamm, Ammann, Eggen, Finger, Fuhrimann, Lienert, Mark and McCallum2023). Consequently, actors might strategically use the information from collaboration partners to advocate for their instrument preferences instead of adjusting them based on the received information.
Conclusion
Various policy instruments can be used to address pressing sustainability problems such as the use of pesticides in agriculture. For successful implementation of new instruments, a majority of relevant actors must support these instruments. Support is negotiated but driven by individual instrument preferences. A better understanding of these preferences is thus key to anticipating potential conflicts and finding broadly supported policy instruments for solving the problems we face. The aim of this study was to shed light on actors‘ instrument preferences, and to explain differences therein in the context of sustainability problems. In this context, when forming instrument preferences, actors must navigate trade-offs between satisfying different policy core beliefs, which is also a risky endeavor due to uncertainty about policy consequences.
We found that not only the relative importance of specific policy core beliefs but also domain-specific risk attitudes are relevant to explain why actors display different preferences for policy instruments. So far, domain-specific risk attitudes have been less explored in policy studies, but our study shows that they can play a decisive role in forming preferences for more stimulative and coercive policy instruments (e.g., subsidy, contract, or order). Our explanation is that risk attitudes become more relevant as the uncertainty and severity of consequences of instruments for satisfying policy core beliefs increase. Based on our study findings, we argue that modeling policy instrument preferences without considering the actors’ risk attitudes in addition to the relative importance of policy core beliefs is inadequate. The relative importance of policy core beliefs reflects the evaluation of trade-offs between conflicting consequences of policy instruments. The risk attitudes capture actors’ responses to uncertain consequences in a specific domain. In the case of Swiss pesticide risk reduction policy, for example, it seems important to consider the relative importance of policy core beliefs and risk attitudes regarding objectives related to agricultural workload, protection of non-target organisms and water bodies, food security, economic viability of farming, innovation potential in crop protection, and cost fairness, as these beliefs and attitudes were relevant for the actors’ instrument preferences.
Our findings are relevant to most sustainability problems, which come with conflicting policy core beliefs and uncertainty about policy consequences, and where any policy choice inevitably involves making trade-offs and dealing with risks. However, further research exploring the relationship between policy core beliefs and risk attitudes in preference formation is necessary. Particularly, the observed pattern between the coerciveness and repressiveness of policy instruments and the role of actors’ risk attitudes needs further empirical verification. Our Bayesian framework allowed studying the individual role of specific policy core beliefs and risk attitudes in a complex decision context based on a quite small data sample. But our results need to be confirmed in other cases.
Our study contributes to a better understanding of policy instrument preference formation through an innovative combination of the ACF from policy studies and MAUT from MCDA. The role of policy core beliefs and risk attitudes for policy instrument preferences under conflicting beliefs and uncertainty are undertheorized in the ACF (Kammermann and Angst Reference Kammermann and Angst2021; Weible et al. Reference Weible, Ingold, Nohrstedt, Henry and Jenkins-Smith2020). MAUT complements the ACF theoretically in terms of how actors evaluate trade-offs and deal with conflicting policy core beliefs, how they respond to uncertainty with domain-specific risk attitudes, and how these evaluations and responses are translated into preferences for policy instruments. Moreover, our adapted MCDA tools to elicit the relative importance of policy core beliefs and actors’ risk attitudes could be an interesting methodological contribution to policy studies. Our theoretical and methodological approaches can be applied to other sustainability problems to assess instrument preferences and find instruments with the potential for sustainable policy transition.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0143814X25100664
Data Availability Statement
Replication materials are available in the Journal of Public Policy Dataverse at https://doi.org/10.7910/DVN/MVVXPF
Acknowledgments
We thank all the actors and representatives who participated in our survey and workshop. We are grateful to Ofelia Freidhof, Laura Jaeggi, Benjamin Hofmann, Adrian Mettler, Ueli Reber, and Michael Zysset for their assistance during the workshop and acknowledge the valuable feedback on policy core beliefs, lottery questions, and policy instruments by various experts at Agroscope, apisuisse, BAFU, BLW, Eawag, ETH Zurich, Kalaidos University of Applied Sciences, the Ecotox Centre, VSA water quality, science industries, WSL, FiBL, SBV, SCAHT, and SECO. We thank Mario Angst and Andreas Scheidegger for their statistical support and the scholars participating in the Political Science Research Seminar at the University of Bern for their feedback on an initial draft. We are also grateful for the constructive and valuable comments we received from the two anonymous reviewers.
Author contributions
Milena Wiget: conceptualization, formal analysis, investigation, writing (original draft), visualization; Judit Lienert: conceptualization, investigation, writing (reviewing and editing), supervision; Karin Ingold: conceptualization, investigation, writing (reviewing and editing), funding acquisition, project administration, supervision.
Funding statement
This work was supported by the Swiss National Science Foundation [grant number: 193762]. The foundation had no role in the design, execution, analysis, or interpretation of data.
Competing interests
The author(s) declare none.
Appendix
Figure A-1. Variance in the importance of policy core beliefs. The scatter plot illustrates the overall importance (y-axis) and the variance in importance (x-axis) of policy core beliefs across all actors (N = 24). The sum of the relative weights assigned to a policy core belief indicates its overall importance, and the variance of the assigned weights indicates the variance in importance. The policy core beliefs cover the domains of human health (violet), environmental protection (green), agro-economic productivity (yellow), and socio-political costs (dark blue).
Figure A-2. Posterior predictive checks. The bar plots illustrate how well the data simulated by the Bayesian ordinal logistic regression models for regulatory instruments (A), market-based tax incentives (B), market-based subsidies (C), cooperative instruments (D), and persuasive instruments (E) fitted the observed data.
Table A-1 Decision-, position-, and reputation-relevant actors in Swiss pesticide risk reduction policy who participated in the survey (S; N = 46) and the workshop (W; N = 24). Administrative actors (AD), science actors (SC), and interest groups (IG) participated in the workshop. The table is adapted from Wiget (Reference Wiget2024)
Table A-2. Result of the Bayesian ordinal logistic regression models for regulatory, market-based, cooperative, and persuasive instruments. The table includes the model estimates for the intercepts and parameters of the explanatory (CB: policy core beliefs; RA: risk attitudes) and control variables (CPSI: collaboration partner support index; actor type: science actors (SC), interest groups (IG); policy instrument specificity) with the upper and lower limits of the 95% credibility interval (CI).
Open access
