Food supply chains, a network analytic approach

Current models for supply chains of food products are not fully able to describe production and marketing dynamics for they usually do not take into full account the network of relationships existing between production, processing, distribution, and even the disposal of food. This makes these models not very useful as tools for a good governance of the players in the food sector. Besides that, making predictions has become increasingly difficult due to the dynamics of the food market, more and more similar to that of a complex financial market. Even agro-food production processes have become very complex systems, involving many actors performing activities of different types and linked by relationships of different nature. Moreover, these relationships are no longer limited to those between the elements most closely linked along the chain but can include stakeholders anywhere in the chain (Yu & Nagurney 2013). This leads to the need to develop new models for evaluating the network of relationships between the actors of the food supply chain, both to assess the robustness of the organizational structure and to have more accurate measures of the role and the importance of each actor in the system. This can also allow to identify which of the actors occupy strategic positions in the network and which of them have only a redundant function. The objective of this paper is to develop a tool for a detailed analysis of the structure (actors and relationships) as well as of the dynamics existing in the supply chains of products of animal origin by providing a mapping of the productions as complete as possible, and as representative as possible of the relationships among the players. The basic hypothesis is that we face a complex system, mainly characterized by the dynamic nonlinear relationships existing between its elements. In the panorama of the many possible methods to approach the problem, the techniques developed in the framework of network science seem to be the most suitable for the purpose. A supply chain, in fact, can be seen as a network of stakeholders involved in growing, processing, and distributing the products, in which actors or actions are the nodes that are connected to each other by some kind of business relationship. These are directed links that can carry a weight which can be valued in different ways. The new model proposed is a complex network with topological (structural) characteristics defined by the connections between the vertices, to which we assign a weight that represent the monetary value of the exchange occurring. Such an abstract construct can map well many real networks (social, biological, technological), and the vast existing literature has shown it able to provide insights of great interest both from a theoretical and a practical point of view (da Fontoura Costa et al. 2011, Easley & Kleinberg 2010). Complex networks are suitable to represent supply chains for several reasons: i) they allow a visual (qualitative) and a quantitative analysis both at a global (whole system) and local (individual actors) level; ii) they allow highlighting possible substructures such as hierarchies or communities and measure the effects they have on the overall functioning of the system; iii) they allow comparing different configurations and highlight associated advantages and disadvantages, and; iv) they allow performing simulations thus giving the possibility to examine how global or local modifications can affect the system, and what configurations are the most effective with respect to some dynamic process unfolding on the network (Barrat et al. 2008, Newman 2010). Such a model can be an interesting and new analytical tool for an observatory on products of animal origin thus becoming a strong support to decisions for policy-makers. Moreover, it can also provide individual actors with useful advice on how to optimize own supply chains and improve efficiency. Finally, through a full and effective enumeration and evaluation of the relationships between all the actors, a network model can be highly helpful in developing policies and systems for tracking and tracing products. The case study presented here builds on the preliminary qualitative analysis presented at a previous EFITA conference (Nasuelli & Clemente 2013). The model examines the Italian supply chains of milk and dairy products (cow, goat, sheep and buffalo milk), and beef and pork meat along with their derived products. The network consists of the companies involved in the supply chains and their relationships are weighted by taking into account the monetary exchange between the different companies. The model takes into account some business choices such as that of the direct selling of products, as well as some aspects concerning the recovery of waste through rendering activities for the production of energy or by-products that are used in other industries, and aspects currently under the spotlight for their social valence such as the recovery of the unsold for humanitarian purposes. The resulting network is a directed weighted network exhibiting a skewed (i.e. show long tails) distribution of the links following an exponential shape, that makes it similar to many other complex natural networks. The different individual (nodal) measurements allowed to identify the most central actors as well as those whose position is critical for the connectedness of the whole system, and those whose neighborhood is of particular value. In addition a modularity analysis was performed. It provides a view into the mesoscopic structure of the network by highlighting groups of nodes more densely connected between them than with other parts of the network thus underlining the self-organization characteristics of the supply chain system (Fortunato 2010). Moreover, these clusterings, which can be interpreted as collaborative groupings, can be of great importance for policy actions directed towards an optimization of the whole system and, for individual stakeholders, in order to look for possible new relationships with the aim of improving operational and strategic activities. Finally, the limitations of the methods as employed here in our case study, the implications of the findings for the purposes stated above are described and discussed. Possible future extensions of this approach are also reviewed, mainly for what concerns the efficiency and effectiveness of such system.