Who can change what? Self-perceived, attributed and structural influence among actors in the Swedish grain legume system

Self-perceived and attributed influence refer to how actors see the power and agency that they and others possess. For an actor to knowingly wield their influence, they must perceive and draw on their own agency in taking action, which is a key component of the ability to respond to change in one’s environment (Brown and Westaway 2011, Barnes et al 2020), including dietary shifts. Being perceived as influential may indicate the framing power of the influential actor to shape agendas and disseminate ideas (Morrison et al 2019), whether or not the actors who are attributed influence recognize that influence themselves. Both self-perceived and attributed influence may exist separately or together.

Structural influence refers to the ability of actors to exercise power and agency through the structure of their relationships with other actors in the larger networked system (Bodin 2017). Different patterns of relationships between actors can impact the ways in which actors can exercise their influence. Social network analysis can aid in evaluating the relational structures among actors (Newman 2018), allow us to understand how these interactions may support actions to address transformation (Bodin and Crona 2009, Bodin et al 2019), and identify opportunities and constraints for system change (Hileman et al 2020).

We derive these types of influence based on previous research of actor roles in transformative change processes. Lam et al (2021) combined the leverage points perspective with social network analysis to classify how existing relationships might contribute to transformative change based on where they intervene in a system, finding that the method can help identify key actors and their attributes. Other research has explored how actors’ self-perceived and attributed influence in a network can impact governance systems with transformative implications, finding that influence and centrality are related but not synonymous (García and Bodin 2021) and that alignment of self-perceived and attributed influence may have important implications for collective action (Jericó-Daminello et al 2021).

Social network studies have found that open and closed network structures facilitate an array of actions related to transformative processes, including social learning, resource distribution, establishment of trust and shared norms, and cooperation (Berardo and Scholz 2010, Hileman et al 2018, Bodin et al 2020). Closed structures refer to clusters of actors with many ties between them, and can help strengthen cooperation among actors through building shared norms and trust (Berardo and Scholz 2010, McAllister et al 2017). Open structures refer to bridges between actors in otherwise loosely connected regions of the network, and can promote sharing of knowledge and material resources (McAllister et al 2015, Hileman and Lubell 2018). A balance of both structures can provide joint benefits on multiple social processes needed to realize transformative potential (Barnes et al 2017, Constantino et al 2022).

This study builds on previous research showing how the leverage points framework can help uncover the transformative potential of different actions to change food system behavior (Abson et al 2017, Dorninger et al 2020, Lam et al 2021), as well as the structural elements of the food system that support these actions (Iannetta et al 2021, Williams et al 2023). A recent analysis of peer-reviewed and grey literature (Scheuermann et al 2024) identified 96 system-level actions that could influence grain legume consumption in Sweden through changes that either: take a value chain or system perspective; change production to meet a future increase in demand; or directly interface with consumers or increase the health and sustainability of diets. The authors applied Abson et al’s (2017) framework, then used open and axial coding to place the actions into 15 categories, examples of which are provided in table 1. This context-specific information provides a rich foundation for exploring value chain actor influence over actions with the potential to increase grain legume consumption.

We focus on the grain legume value chain for products consumed within and produced wholly or partly in Sweden. To compile the list of relevant actors, we used the online retail sites of the major grocery chains (e.g. Coop, Hemköp, ICA, mathem) to identify grain legume products and their manufacturers. Expertise from the authors yielded additional processors selling through alternative channels. Next, we used the Swecris database, which aggregates research funding at the national level, to identify large projects funded by the Swedish government related to grain legumes using English and Swedish terms for grain legume varieties (Vetenskapsrådet 2022). These funding recipients, along with their project partners, were also added to the actor list. This phase of the data collection occurred over Fall 2022 and Spring 2023. Later stages of the data generation process employed snowball sampling techniques to identify additional actors.

Using the actor roles in figure 1, we coded the actors identified through retail sites and the Swecris database as belonging to one of five roles: producers (e.g. farmers, farmer organizations), processors (e.g. manufacturers, packers), distributors (e.g. retailers, wholesalers), consumers (e.g. municipalities, consumer organizations), or wider food system (e.g. government agencies, universities, funders). Several actors had more than one role, which we coded as primary, secondary, and tertiary. We then contacted organizations within each of these roles to conduct semi-structured interviews aimed at getting deeper insight into each value chain role, elicit ties to other actors (e.g. information, collaboration, ownership, funding), and how actors perceive their and others’ influence with regards to increasing grain legume consumption. Interviews of 1–1.5 h were conducted in English or Swedish during Fall 2023 and Spring 2024, recorded with participant permission, and machine-transcribed using Amberscript or Whisper. We used the information from the interviews to expand the list of actors (snowball sampling), while continuing to conduct interviews with previously identified actors. Ultimately, we were able to interview 21 individuals representing 18 unique organizations spanning the value chain roles.

For the perceived influence component of the interviews we used the 15 action categories from Scheuermann et al (2024) to describe actions with the potential to increase grain legume consumption. This allowed us to be specific about types of actions while also embedding the transformative potential from the leverage points framework. Participants rated their perceived influence for each action category on a Likert scale from 1 (less influence) to 5 (most influence); then they were asked what organization or actor role had the most influence over the action category. However, only 15 individuals representing 13 unique organizations (i.e. producer organizations, processors, distributors, consumer organization) agreed to participate in the influence exercise during the interviews.

Details from the interviews included herein, including individual actor’s perceptions of influence and knowledge about connections, are limited to those that were publicly available and known or willingly shared by the participants. For the purposes of analysis, we group actors by their value chain role, which precludes identification of individual actors. Lastly, we further refined the ‘wider food system’ categorization from initial actor identification based on the participant responses to more accurately reflect their perceptions of influence. The additional roles include interest organizations (e.g. trade or advocacy organizations), EU and national governments, society, funders, universities, and large projects (i.e. long-term, multi-stakeholder programs). This did result in multiple actor roles for which we lack interview data since only actors that were part of the value chain were targeted for interviews. The additional actor roles are included in the network analysis. We note that while ‘society’ is not a specific actor, several participants named ‘society’ as having the most influence over different actions so we retain this term for their responses. This term also includes actors who do not fit in other actor roles.

We used the initial list of actors, plus those actors generated through the interviews, as a starting point for building the network of relationships that constitute the Swedish grain legume AKIS (figure 1). In cases where actors were membership organizations, their organizational members and organizational board members were also added. Where actors were companies, their owner organizations were added. This information about actor ties was obtained from organizational websites, industry publications, and/or LinkedIn. The purpose of exploring these secondary actors in more detail was to better capture the group membership of the organizations named as being influential in the interviews, as the structure of their ties within the system might help support the creation of norms, facilitate the distribution of material resources, or contribute to other forms of transformative potential (Barnes et al 2017).

We identified a total of 543 unique actors through the interviews and website review. However, we removed all actors possessing only one connection unless they were named as having influence during the interviews (n = 253). This procedure is routinely performed on large networks (e.g. Hileman and Lubell 2018), and has been demonstrated to preserve the core structures at the heart of the network (Hanneman and Riddle 2005, Yi and Scholz 2016). We then removed an additional 14 actors that were secondary connections (i.e. members of organizations named by interviewees), and for which we could not find any information online. This resulted in 276 remaining actors included in the analysis. These actors, and their 908 ties to one another, comprise the Swedish grain legume value chain and AKIS and are the object of our structural analysis herein.

We used the ‘sna’ and ‘igraph’ packages in the R software environment to calculate descriptive network statistics. Measures of centrality and clustering were calculated at the network and group (actor role) levels in order to determine the presence of open and closed structures in the Swedish grain legume value chain and AKIS. We examined average path length at the network level and betweenness centrality at the group level (Wasserman and Faust 1994) as measures of open structures. Average path length is a measure of the mean number of connections separating actors in a network (Wasserman and Faust 1994), while betweenness centrality represents the number of times a particular actor is located on the shortest path between other actors in a network. These measures are indicative of actors serving as intermediaries or bridges across the network. We examined local clustering coefficient at both the network and group levels (Wasserman and Faust 1994) as measures of closed structures. This measure represents the fraction of an actor’s partners who are themselves connected to one another, and can indicate the existence of subgroups in the network. The balance of open and closed structures in a network is captured by the small-world quotient (Watts and Strogatz 1998), which is a ratio of path length and clustering.

In addition to assessing open and closed structures, we examined average degree as a measure of the overall activity of different actor roles. Average degree highlights which actor roles tend to have greater, or fewer, ties in the network. Lastly, we used the UCINET program to calculate the external-internal (EI) index at the group (actor role) level. The EI index shows the extent to which connections exist within and across groups in a network, and can indicate where additional relationship building is needed. For this aspect of the analysis, we focused on connections between different actor roles across the value chain and AKIS.

Results from the semi-structured interviews are presented in figure 3, and show how value chain actors attribute influence over actions to increase grain legume consumption to other actors. This diagram does not include self-perceived influence—some actors named others within their value chain role as having the most influence. Every value chain role is named at least four times as having the most influence over an action category, affirming the call for broad participation and the importance of engaging all stakeholders in transformative processes. However, AKIS roles are mentioned more often than value chain roles. The size of the text boxes on the right indicates the frequency of the roles being attributed influence.

Zoom In Zoom Out Reset image size

Actors outside of the value chain are most often named as having influence over actions with the potential to increase grain legume consumption. As shown in figure 3, there are some actors mentioned often by every value chain actor, such as the national government, but others that seem to be particularly important to certain value chain roles. These perceptions may help us understand particular aspects of how the food system drivers and institutions impact different value chain roles and ultimately the ability to shift dietary patterns towards more grain legumes, discussed in more detail in section 6.1.

The grain legume value chain and AKIS analyzed here consists of 276 actors and 908 ties between them. The average degree (6.6) indicates that some actors have substantially more ties than others—the maximum number of ties associated with any one actor is 52—indicating some actors are considerably more active. However, this says little about how these ties are structured between actors. In terms of closed structures, the local clustering coefficient (0.086) indicates a relatively low level of closure at the network level. In terms of open structures, the average path length (3.9) indicates a degree of openness in the network. The small-world quotient (5.3) is above 1.0, indicating the network does embody the balance of closed and open structures characteristic of ‘small-world’ networks, but the result is quite low compared to other empirical networks.

The group-level network measures (table 2) provide details about structural differences between value chain and AKIS roles. The average degree results can be considered how actively each actor role engages with other actors in the Swedish grain legume system. Here, large projects are by far the most active, followed by interest organizations, universities, and producers. The closed and open structures columns are grouped by high, mid, and low categories for ease of interpretation. These categories are relative; they distribute the observations into three roughly equal groups, and line up with cut-points in the data. In terms of closed structure, producers, the EU government, society, and large projects have high clustering. This suggests they are embedded in the system in ways that can help build trust and collaboration. In terms of open structures, producers, interest organizations, the national government, universities, and large projects have high betweenness centrality. This suggests they are likely to act as bridges between actor roles and help facilitate distribution of knowledge and resources across the wider system. Large projects and producers exhibit both high clustering and high betweenness, indicating they are well-positioned to engage in multiple social processes needed to realize transformative potential in the system.

In figure 4, we present a synthesis of the interview and network analysis results showing self-perceived, attributed, and structural influence by action category. The icons in the column headings correspond to the primary actor role shown in the legend, and the colors correspond to the type(s) of influence. Actors with structural influence are those with high clustering or high betweenness (shown in table 2), and these results apply to all action categories.

Zoom In Zoom Out Reset image size

Producer organizations are the only role with all three types of influence, displayed in dark grey. Processors tend to perceive their own influence more often than other value chain actors do, and are often attributed influence by others. Distributors and consumers are attributed influence more often than they identify their own influence. There are only two action categories for which value chain actors do not perceive their own influence, and these both relate to changing regulations and the regulatory process. For the AKIS actors, the national government and interest organizations are attributed influence for nearly all of the categories, and only funders lack structural influence.

Producer organizations tend to point to actors in the AKIS, and stand out for naming ‘society’ multiple times. This indicates a perception among producer groups that wider consumer opinion and choices are among the most important drivers for change. They point to the socio-cultural and perhaps demographic drivers of the food system as determining what changes are possible.

Processors mostly attribute influence to distributors, political and institutional actors, and themselves. This likely reflects their perceived dependence on distributors in the highly consolidated Swedish retail industry (Ghosh and Eriksson 2019). They are also the only actor role to name funders multiple times as being the most influential over several action categories, likely reflecting their dependence on both privately and publicly funded research and innovation. Processors also stand out for naming themselves as having influence over nearly every action category. This may reflect a techno-optimism that views specific grain legume innovations (e.g. processing techniques that closely mimic meat textures) as key solutions to dietary changes to support health and sustainability.

Distributors name nearly every actor role included in this study as having the most influence, and more often than other actors attribute influence to large projects and universities. One explanation could be that the distribution portion of the value chain is so consolidated that the few actors are engaged in many of the larger research projects, whereas other value chain sectors with less consolidation have a greater number of actors to participate in different projects. Several other actors referred to distributors as having ‘extreme amount [s] of influence’ due to the high market concentration. However, they were not identified as having any particular structural influence as we did not evaluate market share or trading relationships in this study. Distributors are moderately connected to one another, and grocery stores and wholesalers each have their own interest organization. This type of forum may be perceived by other stakeholders as reinforcing control of the value chain, and it could be used to facilitate or hinder changes that impact healthy and sustainable diets.

The consumer organization mostly attributes influence over actions to increase grain legume consumption to the national government. Here, control over public procurement policy, dietary guidelines, and subsidies/incentives could shift dietary patterns. They do not mention producers at all, which may reflect a traditional disconnect between production and consumption in policy, research, and practice.

Value chain actors often named the national government as being influential. The analysis of structural influence also showed it has cross-system connections and a more open structure to serve as a bridge across actor roles. In general, the participants acknowledge the role of regulations in creating the conditions for the market to function and for consumers to trust companies. However, many actors also highlighted the constant tension between the level of influence provided by the government and the amount of freedom private actors have, such as concern for a reduction to producer and consumer agency if the state were to influence the market too much, for example by increasing taxes on unhealthy foods or prescribing production amounts for grain legumes.

On the other hand, participants reflected on the disconnect between the market incentives that push them to consider only their business interests and profits, and the common good that the government should promote, such as certain information shared between value chain actors to promote healthy products and purchases:

• [information flows] should not be company owned. So that should be on an [larger] organization level… [otherwise] it is dangerous, because a single supplier cannot own any part of that flow. We will only contribute to the flow. Otherwise, we would try to fix it to go our way, right? It must be on an aggregated level… it is extremely important… I am not sure that you would get the same answer if you talk to other industry people.

Several participants pointed to actions that they wish the Swedish government would take to support dietary shifts towards grain legumes. In particular, several participants suggested increased investments in general marketing of plant-based eating as a way to influence social norms, which has been done by the Danish government (Ministeriet for Fødevarer, Landbrug og Fiskeri 2023).

At the consumer level, the role of municipalities was also highlighted for the potential to spur development in the value chain to meet public demand through coordinated public procurement:

• Municipalities are super consumers, basically. We consume 800 billion SEK [Swedish krona, approximately € 70 billion] each year, so it is a lot of money and we can have a lot of impact in how we choose what to consume… the market needs to adapt to meet our requirements.

However, the distributed decision-making at the municipal (290) and county (21) levels in Sweden makes this coordination difficult. Conversely, this decentralization could result in the ability of municipal and county governments to experiment with other local actors, generating new practices to change consumption and production patterns, creating niche ‘pockets of future in the present’ (Sharpe et al 2016, Metelerkamp et al 2020). Thus, the distributed authority in Sweden may be a barrier to substantial change without specific political mandates at the national level, but may hold other potential to change institutions if local actors work together and commit to specific, shared goals.

Our findings echo previous research showing self-perceived and attributed influence have some overlap, but are not synonymous, with structural influence (García and Bodin 2021). Previous research suggests that alignment between self-perceived and attributed influence may increase the likelihood of collective action (Jericó-Daminello et al 2021). The rows in figure 4 illuminate where targeting misalignment between producer organizations and the rest of the value chain may be most fruitful: green cells in the first column indicate producer attributed and structural influence, and orange cells in the subsequent columns indicate self-perceived and attributed influence. Changing producer organization perceptions so that they acknowledge their influence over and can take action on ‘Change the food environment,’ ‘Increase investments,’ and ‘Develop new tools, facilities, and/or products’ could create synergy with other value chain actors for actions with the potential to increase grain legume production and consumption. For example, a combination of changes to the food environment by actors across the value chain could support change together: if producers leverage their policy influence to advocate for changes to regulations and fiscal policies, it may enable processors and distributors to more easily make changes to packaging, price, and the physical food environment—actions with the potential to increase grain legume consumption (Scheuermann et al 2024).

One action category is preventing further scaling up of grain legume production: ‘Sharing risk and profit differently’ across the value chain. Currently, producers bear much of the risk and receive less of the profit (Morel et al 2020, Schwarz et al 2021, Sweden Food Arena 2021), and while farmer cooperatives may absorb risk for some variations in production, for grain legumes it is largely the individual farmers that bear all of the risk. Participants expressed various views on this topic, all of which are rooted in retaining current economic structures as opposed to actually changing risk and profit:

• We will be one of those who get the profit. But we can also put demand for how to develop the crops. The farmer will always be the party who takes the most risk.In this worldview the producer responds to the demands of those in other value chain roles, limiting producer agency in making decisions about production yet retaining the risk involved.
• Absolutely it should be shared, and we should dare to say to consumers, ‘This here is more expensive’—that we need to explain better.Here the profit is not shared differently between companies, but the increased margin for farmers is passed to consumers, meaning other value chain actors do not actually change their risk or profit margins.
• It is a really difficult question, because there is no one that has the whole value chain. I do not see an actor that has it, that owns from plant breeding through to consumers.Here vertical integration surfaces as the only path for changing risk and profit and collective action is not even considered as an option.

While some value chain roles may have structural and attributed influence over changing how risk and profit are shared, the participant perspectives on this topic indicate that only increased prices or increased consolidation can address this issue rather than changes to business practices such as longer-term contracts with producers. These views reflect underlying values within the economy about minimizing exposure to risk and ensuring continued profitability, which are fundamentally incompatible with this action category, and which appears key to increasing grain legume production and consumption in Sweden.

The connections between value chain roles would be important in addressing this disconnect and creating a space for open dialogue about changing the current financial and risk model in grain legume production. Such a complex topic requires building trust, a precondition for actors tackling more difficult collective problems (Bodin et al 2020) which can be facilitated by closed structures, but the network-level clustering measure is overall low. While producers, processors, distributors, and consumers are each connected to one another, they are most strongly tied to interest organizations and large projects, suggesting these could be the actors with the connections to convene this type of dialogue. As a major funder of large projects, the national government can allot specific funding for these dialogues and ensuing pilot programs to test different strategies.

Collaborative processes are often required in efforts towards transformation (Scoones et al 2020). However, collaborative ways of working may conflict with business norms or incentives to consider information as trade secrets or as proprietary out of fear of losing market share, as suggested by one participant:

• To be honest we have tried… but it did not really happen because at the end of the day, people are scared, you know, everybody wants to be the one who could succeed, right? … so, it kind of fell apart… because people were really, really unwilling to share.

If private companies are reluctant to build trust and collaborate at a deeper level, any changes are likely to be small, rather than fundamental shifts in food system behavior. The national government has demonstrated its capacity to convene networks for specific functions such as Sweden Food Arena to support the National Food Strategy. Large projects can also function as networks depending on the scope, participation, and representation of stakeholders, but depend on a convening entity and funder to begin; and, as multi-stakeholder partnerships involving actors across the food system, projects may also be influenced by political, corporate, or other agendas (Dallas et al 2019). For lasting influence in a value chain dominated by private companies, these actors must have agency and ownership over the network developed in the project in order for it to last beyond the project funding timeline.

The relatively high connectedness of producer and other interest organizations with each other combined with their high degree of openness, which can serve as bridges to other roles, could mean that, if coordinated, their impact could be swift and large on redesigning the system rules and structures as part of changing food system outcomes. An important caveat to this potential is that benefits may only be realized for actors who are members of these organizations, which may be linked to particular self-selecting criteria (e.g. farm or company size, production system type, geographical scope). System changes that benefit particular actors and exclude others may create additional structural inequalities within the food system, which may make certain practices, such as collaboration, more difficult.

The mixed-methods approach in this study builds on a growing body of research examining self-perceived and attributed influence from actor interviews alongside structural influence using social network analysis. A key contribution of this study is evaluating the different forms of influence, which may make actions with the potential to shift deeper leverage points possible, such as changes to norms, values, and beliefs. By evaluating how different forms of influence are present for different actors it may be possible to see new possibilities for collaboration that can push the food system towards healthier and more sustainable diets. We assume that having more types of influence makes an actor more influential, but we do not evaluate the different combinations of influence in this study. A strength of this approach is including multiple ways of understanding actor positions with respect to actions that can change system outcomes. One difference in this study compared to previous work is that not all actors included in the social network analysis were interviewed, limiting the conclusions we can draw from the structural measures independent of the context of the interview data.

When conducting semi-structured interviews with value chain actors, it was challenging for several of them to name which part of the value chain they represented. The results presented in this paper represent the primary value chain role, the position from which the person(s) interviewed identified their work with grain legumes. However, it is notable that single organizations can navigate different spaces within the food system through multiple roles. For instance, retailers are distributors and may also have their own brand of grain legume products (processors). Municipalities are consumers and also closely aligned with national government regulations and processes. Producer organizations are cooperatives of individual farmers (producers) and also political actors (interest organizations).

Using the value chain perspective grounds this work in a market logic that aligns with current policy approaches at the EU and national levels. The strength of this approach is its pragmatic link to actions and actors in the interviews and the study findings. However, it may inadvertently exclude other logics that could result in different findings, such as community approaches where farmers may be conceptualized as stewards of nature or technological approaches that use the innovation system as a point of entry (Avelino and Wittmayer 2016).

This study focuses on the grain legume value chain to include as granular detail as possible in one part of the food system. An advantage to this choice is the ability for practitioners and policymakers to use the information, but some findings may not be applicable to other settings or sectors. Nevertheless, in the next section we suggest some key takeaways based on our analysis and expertise.

First, it appears the value chain may not be sufficient in framing policy priorities and discussions, but rather a focus on the relationships between the value chain and the AKIS (or wider system) is needed to support actions with transformative potential. Research using the multi-actor perspective could further develop this analysis to evaluate how different types of logic, particularly in third sector organizations, operate in the grain legume value chain and AKIS.

Second, processors perceive themselves as having influence over a wide range of actions across the leverage points framework. While they are active partners within many large research projects, they currently receive limited attention in the food system transformation literature compared to producers and consumers. As such, the results of our paper suggest that processors deserve more attention within the food systems research community as actors with the capacity to influence system parameters, feedback, design, and intent.

Third, while large, multi-stakeholder projects may have the connections, funding, and knowledge required to engage actors within a value chain and wider system in deep collaboration, their time-limited nature conflicts with the resources needed to build trust, negotiate trade-offs, and implement actions to change consumption and production patterns. Previous research suggests actors need to have some choice in these partnerships (Bodin and Crona 2009); that convening third parties need sufficient time, funding, and a shared commitment between actors to begin the process of changing risk and profit (Österblom and Bodin 2012, Bodin et al 2020); and that these type of projects may require adoption by formal authorities in order to succeed in the long term (Bodin 2017). Therefore, longer project funding opportunities and requirements to include dialogues to address specific elements, such as changes to risk and profit, could be implemented by private and public funders.

Lastly, experimentation and learning, including at a system level, involves trying new approaches even when there is uncertainty about the impact at different leverage points. How organizations perceive their and others’ roles in change processes depends on which of the core value chain elements—materials, labor, technology (Kogut 1985)—they emphasize, how they relate to other actors within the food system, and from where they derive benefits, including profits. Some of the action categories include actions that, if taken, would reduce or exclude existing actors from the value chain by, for example, more directly connecting producers to consumers and reducing the middle portion of the existing value chain. Reducing the role of processors would also impact the wider food system, such as reducing the profitability of funder investments in processing technology, infrastructure, and innovation. Other pathways to system change could emphasize either technology-heavy solutions that favor segmented manufacturing of legume-based products or labor-focused solutions that favor local grain legume farming (Gordon et al 2022). Changes of this scope will require resolving conflicts, managing trade-offs (Hebinck et al 2022), and negotiating sacrifices (Bodin et al 2020) as actors reconfigure in new patterns and develop different practices and processes reflecting the shift in norms, values, and beliefs.