Network analysis of organic seed systems: a systems-level analysis for resilience

Survey responses:

• Out of those we surveyed (initial and snowball sample), we have responses from 95 seed producers, 49 companies, 60 academic researchers and 43 organizations for a combined response rate of 33% (248/762) from regions across the US. Details of survey response rates are in Table 1.

Table 1. Number of seed stakeholders in the population databases and survey responses

Preliminary results and discussion:

• Q1: How do organic seed stakeholders define resilience for the system, and what are the needs and priorities of each sector?

Data summaries and figures for this question are available at the interactive data page: https://liza-wood.shinyapps.io/actors-organic-seed-innovation/

The “resilient seed system” definitions provided by seed stakeholders included 35 coded themes and six broader thematic groups. These thematic groups include: the social, environmental, and economic challenges faced by the seed system, various seed characteristics that contribute to high quality seed, the role of the community, attributes of the supply chain that make the system viable, the system's ideal structure and tendencies in the face of stress, and different values that a resilient seed system should uphold. Regarding the theme of challenges, environmental and climate-related challenges were the most-named challenges. With respect the supply chain, three priorities stood out: growing enough seed to meet producers' demands, maintain a wide diversity of crops, and seed banking and continuously breeding in order to keep stock of viable and adaptable seed. This latter point matches with the seed qualities that are of top priority: stakeholders believe seeds need to adapted to their local environment and be consistently productive. Regarding the seed community, stakeholders emphasized having stakeholders who are informed and knowledgeable about seed production; including a diversity of people — professionally, geographically, and demographically; and that these communities must be cooperative and work together. This collaborative spirit was further identified under the theme of system structure, where a resilient system was often described as operating at the regional level, and as such required a decentralized network populated by multiple stakeholders at multiple scales. Last, these systems must uphold a number of values all weighted evenly across the stakeholders: seed production as an interconnected system that supports environmental health, food security, and empowers growers. As one producer defined it, "A resilient seed system is one that can succeed in the face of challenges of climate, political, and market force impacts. It utilizes the decentralized network of growers, provides widespread education and communication among all participants, and fosters cooperation and sharing for the benefit of all." Together, seed producers identify that a resilient seed system should strike the balance of having diverse, decentralized regional seed networks while maintaining enough connectedness to support efficient flows of information and resources across the country. 

Beyond resilience definitions, stakeholders who work directly with producing seeds express that challenges related to yields and managing biotic and abiotic stressors are among the top issues for organic seed production. Similarly, top priorities for breeding are disease tolerance, yield, germination and vigor, and abiotic stress resistance. This matches the resilience definitions that climate challenges and yield functionality are the most important priorities for the organic seed system at this stage. 

• Q2: What are the structures of two co-occurring networks (knowledge and supply chain) in the organic seed system?

Data summaries and figures for this question and Q3 are available at the interactive data page: https://liza-wood.shinyapps.io/network-organic-seed-innovation/

How is the seed system network structured? When asked about their organic seed connections – those who they get information from, research with, acquire and exchange seed with, and work with along the supply chain – survey respondents named a whole range of stakeholders, including other seed producers, processors and retailers, organizations, university researchers, and government agencies. These connections help us understand key steps in the organic seed innovation system: how organic seed-related knowledge is developed and diffusion, how resources are mobilized to support research and supply chain activities, and how products (seed, equipment, etc.) are moved through the supply chain in the development of markets. Throughout the process of this research we decided to map these networks onto four key 'functions' of an 'innovation system' (Bergek et al. 2008): knowledge development, diffusion, resource mobilization, and market formation. All four of these connections make up the organic seed innovation system. 

At the national level, the seed producer network is diverse, decentralized, and cooperative. The 610 stakeholders identified in the network make over 1,000 connections with only 11 stakeholders completely disconnected, signaling the interdependence of those in the seed system. On average, stakeholders have between 2-4 connections (producers generally have fewer than other actors). According to exponential random graph model results (Figure 1.), the organic seed innovation system is more decentralized (anti-coordination and more cooperative than would be expected from a random network with shared structural features. In each region, however, the structure and composition vary. The Western region is the largest network (322 stakeholders) and is moderately decentralized and cooperative.  Similarly, the North Central region has 236 stakeholders and also moderately decentralized and cooperative. The Southern network, however, is the smallest (152 stakeholders) and much more decentralized than the other regions, meaning there are not as clear of leaders within their network. And in the Northeast, also on the smaller side with 162 stakeholders is the least decentralized (more decentralized) and the least cooperative. This means that the Northeast network operates more like a hub-and-spoke than other regions. 

Figure 1. Exponential Random Graph Model coefficient plot for different regional innovation system networks. Estimates are reported as log-odds and represented by shapes, and the lines represent confidence intervals of the estimate.

Regarding network diversity, actors can interact with those within or between their region (indicated by the Homophily Region-scale variable in Figure 1), and professionally (connecting with others in similar or diverse roles, indicated by the Homophily: actor-type variable). According to exponential random graph model results, all regions are more likely to form ties within their region rather that to other regions, though this is more true for the South and North Central regions compared to the West and North Central. This may have to do with the different kinds of connections that are prominent in those regions, which we will describe more in the next figure. All regions are also more likely to form ties with actors of the same profession (e.g. producer-producer rather than producer-government researcher). However, all actor-types are more likely to form a tie compared to seed producers, which are the baseline group for the Actor-type attribute in Figure 1. Also in these networks, we note the importance of the National actor attribute, which signals the probably to form a tie based on actors that span regions -- national actors that work to connect regions. Actors that have a national presence are significantly more likely to have ties for the Western and North Central regions, meaning there is a stronger presence of national actors. In the smaller regional networks, however, national-actor presence is lower.

How is information developed and shared? Of the different types of connections considered in this research, the seed system's knowledge development and diffusion are the least centralized, and in the case of knowledge diffusion, most cooperative (Figure 2). Knowledge development connections – that is, the people or groups that stakeholders go to for information and collaborate with on projects – are significantly more common for organizations, government agencies, and universities than seed producers (the baseline actor group). Knowledge diffusion -- that is, where stakeholders get information from -- comes from a wide variety of resources, though markedly not government agencies. We also see that both knowledge-related connections have no effect for actor-type homophily, meaning that there is diversity in who stakeholders develop and exchange knowledge with. For example, because knowledge often requires diverse expertise and new perspectives, one producer shared, "I think it's important we step out of everyday sources" for information. In this way, knowledge comes from experts but also from peers: "You just can't beat bouncing ideas/problems off other growers who are in a similar situation." 

Figure 2. Exponential Random Graph Model coefficient plot for different functional connections in the innovation system. Estimates are reported as log-odds and represented by shapes, and the lines represent confidence intervals of the estimate.

How are supply chain relationships organized? Supply chain connections involve fewer stakeholders than information-based networks. Supply chain relationships typically involve more costly transactions of mobilization resources and setting up markets, and therefore are pursued more purposively than relationships like knowledge sharing. For these connections, companies play an active role than other kinds of stakeholders, though governmental and university actors also play an important role in resource mobilization. This role of the government and university actors comes largely from research funding, as well as providing a centralized source of germplasm from their national repository. Compared to the knowledge-based connections, supply chain connections are slightly more centralized and less cooperative. However, there is still a strong regionally-based supply chain network, where stakeholders are more likely to be making supply chain connections within, rather than between regions. Based on stakeholders definitions, supply chains are fundamental to a resilient seed system, one in which "the public and farmers have the ability, infrastructure, and systems in place to supply the needs of gardeners and farmers in a given area." While these strong regional ties keep business operations within a shared geography, this also indicates that smaller regional networks like the South might be limited. This matches what we heard from seed producers regarding their challenges sourcing seed cleaning and harvesting equipment. In the South, the costs and equipment for seed cleaning and harvest rise to the top of the challenges list, which may be a consequence of the limited supply chain network in this region.

• Q3: What can the network structure tell us about resilience of the seed system and management interventions?

Based on responses to question one, a resilient seed system is one that should strike the balance of having diverse, decentralized regional seed networks while maintaining enough connectedness to support efficient flows of information and resources across the country. By and large, we see that the different regional networks in the organic seed system match this description. Centralization is low and cooperation is high across all four regions, though different functional relationships change the structural form to some degree. Having a centralized networks means that a small group of actors are the go-to people (think about a hub with many spokes). This kind of network can be really great for diffusing information and efficiently allocating resources, but may reduce diversity and create potential vulnerability to the system if a central actor were to leave or be removed from the network. When three actors are connected in a triangle, this indicates a "cooperative" relationship -- it is strong and there is trust and/or ability to monitor one another. The degree to which actors cluster together into triangles can help indicate the strength of cooperation in a network. These features are well-balanced throughout the different network functions.

We also see a strong presence of bridging, national actors in the North Central and Western regions, but let so in the South and Northeast. Increasing the presence of national-level organizations and government agencies in these regions could help make sure that the decentralization of these regional seed systems is not isolating. We also note that while a variety of actors take on more or less prominent roles across various functions, it is surprising that government actors play such a minor role in knowledge diffusion. While these stakeholders are an important part of knowledge development, they seem to not be a source of communicating that information. This finding points to the need for government agencies to either improve their connections to communicate their research, or to strategically partner with organizations that can communicate their research.