Social Network Analysis in Encouraging Role-Players in the Beef Market to Take Breeding Decisions: A Methodological Study

Abstract

The aim of the research is to apply the social network analysis (SNA) to the Italian Romagnola beef meat market in order to better understand the interrelationships of the supply chain. Knowledge of the market exchange system is very important in order to be able to choose the most suitable breeding approach. The analysis was carried out using data concerning the heifers exchanged during the year 2008 within the framework of the activities of the Italian consortium of white bullock producers of the central Apennine region of Italy (Vitellone Bianco dell’Appennino Centrale). The social network analysis shows that very few of the possible links between the agents are used and that this leads to a lack of competitiveness. At the same time, the agents have shown to be able to place themselves at the most important points along the supply chain. Furthermore, the results of the analysis show the need for a stronger farming phase with a greater emphasis on the farmers, given that, from an economic point of view, they are the most important role-players.

Keywords:

• Social network analysis
• Supply chain
• Romagnola

Introduction

Economics in the local meat market

Genetic innovation plays an important role in the development and performance of meat supply systems (Wever et al., 2010). The profitability of any innovation drives the decision to adopt it (Harry, 2011). Therefore, market forces provide the main stimulus for innovation of products and processes. While the consumers’ expectations have an increasing influence on the shaping of companies’ activities and in directing the innovation process (de Barcellos Marcia et al., 1998), the animal product supply systems in turn organise and carry out the design, the production and the implementation of product and process innovation (Grunert and Valli, 2001; de Barcellos Marcia et al., 1998). Within the meat supply system, the relationships among the agents influence both the collection and the processing of information about the consumers’ expectations and the opportunities for exchange. Namely, the system of the relationships structures the exchange opportunities and the different options available to each company. Therefore, this system contributes to the development of market strategies and to choosing what should be the main target for breeding innovation. The animal breeding strategies are mainly devoted to improving the economic value of the production (e.g. milk, meat, etc.), In this framework, the price of the product plays a very important role and, therefore, knowledge of the market factors defining it is crucial.

Today, the Italian beef cattle breeds (Chianina, Marchigiana, Romagnola) provide a high quality meat product guaranteed by an EU-designated Protected Geographical Indication (PGI) of the white bullock producers of the central Apennine region of Italy (Vitellone Bianco dell’Appennino Centrale). The selection strategies in these breeds are based on both quantitative (weight) and qualitative (tenderness, colour, fat content) traits and much research has been carried out to define the quality of their meat and to study the factors affecting it (Sarti et al., 2005; Lasagna et al., 2007; Panella et al., 2009, 2011).

The specific requirements of the market are the main factors to be considered in the production chain. Therefore, the exchange system has to be well understood in order to make the most suitable choices for breeding.

Furthermore, it is not clear how this product is priced because of the many different market role-players, such as retail purchasing groups, small shopkeepers, and farmers involved at different levels of the production chain. Among the three Italian beef cattle breeds, the Romagnola is the most appropriate to use to study the Italian commercial beef chain. In fact, this breed has the lowest number of heads and this facilitates a more detailed monitoring of the product. Also, its geographical area is characterized by a very traditional commercial system while at the same time being close to highly efficient commercial networks.

The aim of the study was to assess the effectiveness of the social network analysis to understand the market demand for Romagnola beef so that more appropriate selection strategies can be designed. The results obtained can also give indications for the other Italian beef breeds. The innovation in agri-food chains is the creation of new profiles based upon repetition of a process in which production is adapted to the market through continuous strategies of differentiation (Allaire and Wolf, 2004). The engagement of all the role-players in the supply chains shows how important the concept of identity is to support many supply systems, particularly those targeting quality products, such as the EU-designated PGI white bullock producers of the central Apennine region of Italy (Vitellone Bianco dell’Appennino Centrale) (Ilbery and Kneafsey, 2000; Allaire and Wolf, 2004).

Social network analysis characteristics

This study examines the organisation of the exchange of high-quality bovine meat in agrifood supply chains. It aims to show that the objectives of research into bovine meat are directed towards taking advantage of the opportunities provided by a network organisation of the exchange process. The basic characteristics of the organisational frameworks will be evaluated, since these influence company strategy and, indirectly, the supply strategies for animal agri-food resources. The supply of genotypes, raw materials and final products is usually based on the concept of the centrality of the competitive spot market. Therefore, it is assumed that information about consumers’ expectations and product quality both flow through the chain on the basis of indications of price. Nevertheless, this linear perspective has been progressively challenged. Research has shown that in developed countries there are a variety of organisational forms in the agri-food supply chains (Frank and Henderson, 1992; Saccomandi, 1998; Mènard and Klein, 2004), and that these chains involve complex relationships between farmers, processors, distributors, traders and the other agents (Omta et al., 2001a, 2001b; Fritz and Schiefer, 2008). The more complex the systems become, the less the price reflects the information required for strategies to be developed at each stage of the chain. These are not the only elements to become critical, and the different agents in the chain need to work together to adapt to changes in the main forces of demand and supply, and deal with specific circumstances as they present themselves (Williamson, 1985; Omta et al., 2001a, 2001b).

In the present study, we suggest that the animal products supply systems (APSS) involve a complex system of relationships, and that the companies involved face specific processes of competition both as consumers and as suppliers. More precisely, we suggest that the systems of relationships supporting the flow of products and information along the animal product chains can be represented in terms of network relationships. This hypothesis is tested and compared, in particular to the competitive hypothesis that the exchange is linearly organised by competitive spot markets. The importance of our conjecture relates to the implications of the alternative ways of organising the exchange once company strategies have been established. The case of high-quality bovine meat provides a useful field of observation. Our study adds two methodologies to the existing literature. Firstly, knowledge of the systems of exchange of animal products is widened by testing the hypothesis of alternative patterns of organisation. Secondly, the study illustrates the possibility of identifying the system of relationships among the agents in the chain by analysing product flow. Even though this approach can be identified within the framework of the social network analysis (Borgatti and Foster, 2003; Borgatti et al., 2009), to our knowledge, this has received little attention from researchers.

The social network analysis aims to identify the relationships among the agents, their evolution through time, and their impact on the activities of the agents themselves. Related data include contacts, links and connections and, therefore, do not summarise all the characteristics of the agents involved but rather represent aspects of the system as a whole (Scott, 1991).

The network approach is of particular interest in this field because: i) it allows the basic system of the relationships between the agents to be represented, given that this system has been constructed to meet specific commercial and social needs; ii) it allows us to identify the role of the agents within the system as a first step towards understanding the technological and economic opportunities it offers; iii) it clarifies the internal characteristics of the system (i.e. the subsets of relationships). This information should allow scholars, entrepreneurs and managers to put together a comprehensive view of the systems that will enable them to plan technological, managerial, selective and economic activities in a more reliable fashion. In this context, the network analysis could help explain the usually weak position of the breeders in the market and their dependence on the other agents. This could heighten their awareness of market forces and help them in their breeding selection.

The basic hypothesis of the study is that the links between the agents can be identified through an analysis of product flow. According to the social network analysis, the ties that connect relationships within the system can best be described as discrete events and categorised as interactions, i.e. an accumulation of transactions over a period of time (Borgatti and Li, 2009).

Materials and methods

The methodological approach

Two hypotheses about the organisational pattern of the market exchange can be formulated from the analytical framework introduced and these can be compared and tested (Figure 1): i) the network pattern summarises the idea that the network relationships and cohesion (Moody and White, 2003) influence the organization and product flow among the role-players; ii) the linear pattern refers to the idea that the market exchange may be equally accessed by each roleplayer, regardless of his or her interrelationships within the system. The two hypotheses are then empirically tested according to social network analysis (Borgatti et al., 2009) using data collected in the Vitellone Bianco dell’Appennino Centrale supply chain identified by the EU-designated PGI.

Nevertheless, agents also manage information flows and, therefore, ties have to be thought of as tools of channelling information and creating knowledge (Ancori et al., 2000; Borgatti and Li, 2009). The social network analysis approach involves the following initial steps: ai) identification of the constitutive elements (livestock farms, processing companies and traders); ii) characterisation of the relational system as product flows from one agent to another (Borgatti and Li, 2009); iii) analysis of the individual positions (cohesion, centrality, clustering). The empirical measures (Scott,1991) considered here are reported in Table 1.

A critical aspect in the analysis of the meat exchange process is the number of customers and suppliers a company deals with. These numbers reflect the competition process and they may highlight a company’s competitiveness. To investigate this, specific u and v coordinates of each company were calculated (Borgatti and Li, 2009): where:

u_i, is the score a supply company is given depending on the number of suppliers it competes with, then the company is given a high score on u if it supplies companies that in turn have a large number of suppliers themselves (e.g. a livestock farmer who sells to big companies in the same market);

λ is the single value (scaling factor);

v_i, is the score a company that buys on the market is given depending on the number of buyers it competes with, then the firm is given a high score on v if it buys from companies that have many customers (e.g. livestock farmer buying from big animal food companies);

x_ij in our data set has value 1 if there is an exchange between the nodes i and j and 0.

These coordinates allow the competitive position of each firm within the network to be investigated. The idea is to take into account the fact that within the supply chain a company is involved in two basic competition processes through its links. Namely, a company is connected to its own customers and the larger the number of the suppliers of the company’s customer, the more intense the competition process is. More precisely, a company with a high u value supplies to other companies with many suppliers. Inversely, a company with high v is supplied by companies with many customers (Borgatti and Li, 2009). Company A in Figure 2 faces intense competition through its connections with the customers as each of them has alternative suppliers. Company A is given a high u score and is labelled as a hub (Borgatti and Li, 2009). On the other hand, on the procurement side, a company that is supplied by companies with many customers faces intense competition in terms of procurement (see Company B, Figure 2). Company B is given high v scores and is labelled as an authority (Borgatti and Li, 2009).

According to Borgatti and Li (2009), the u and v scores were used to classify the companies of the network investigated.

In order to characterize the network and to explore our hypotheses (Figure 1) in more detail, attention is also given to the concept of structural cohesion. Moody and White (2003) elaborated the concept of cohesion showing its relationship with the concept of structural embeddedness (Granovetter, 1985). Structural cohesion is defined as the minimum number of role-players that, if removed from the group, would result in the group losing its collective function (Moody and White, 2003). A cut set of a graph is a collection of specific nodes that, if removed, would break the component into two or more pieces: a graph is said to be k-connected and is called k-component if it has a cut set with no fewer than a certain number of k-nodes. In the meat market, breeders could play an important role if it were demonstrated that their removal would disconnect the group; this effect would become progressively stronger according to the productive performance and genetic value of their animals. The concept of k-connectivity provides a clear, strict definition of variables into measurable factors (operationalisation) within the structural cohesion and embeddedness (Moody and White, 2003). The influence of the degree of structural cohesion on the product flows within the network was investigated by the following linear multiple regression: where:

y_i = product flow through the network;

μ = constant;

b₁ – b₄ = partial regression coefficients;

All_i, Lab_i, Dest_i = dummy variables accordin to characteristics of role-players;

Nest = nestedness,

e_i = random error.

Data analysis was carried out using UCINET software and, as for structural cohesion, according to the Moody and White (2003) algorithm, following the procedural steps illustrated at http://intersci.ss.uci.edu/wiki/index.php/Cohesive_blocking. Data were analysed by R routines (R Development Core Team, 2007) and PAJEK software (2011).

Figure 1. The structure of the inquiry.

Figure 2. Supply side and procurement side competition in the network.

Figure 3. Romagnola meat (slaughtered heifers) supply chain: agents’ networks.

Figure 4. Agents’ position according to hub and authority.

Figure 5. Cohesive blocking for the network of role-players involved in Romagnola exchanges.

Table 1. Empirical measurements of the agent’s interaction.

Variable	Measurement	Meaning
Cohesion	Density	Effective ties respect to the potential ones, given the number of total agents.
	Distance	The shorter length connecting 2 agents.
Centrality	Degree	Number of adjacent nodes.
	Closeness	Sum of the distances between a node and the remaining nodes.
	Betweenness	The frequency according to which a given node appears within the shorter path between 2 nodes.
Competitive position	Coordinate u	The number of suppliers in competition with the company.
	Coordinate v	The number of buyers in competition with the company.
SCA		Minimum number of role-players who, if removed from the group, would disconnect the group.

Table 2. Centrality: largest values of the networks. Only the largest values of each index are reported with the descriptive statistical indexes and the overall network measurements.

Agents	Degree of centralisation	Agents	Closeness	Agents	Betweenness Centrality
ALL3	43	ALL3	62.1	ALL3	2123.4
LAB5	38	INT5	56.6	LAB5	1288.7
INT5	38	INT1	56.2	INT5	927.8
LAB8	21	LAB8	49.7	LAB4	680.8
INT6	21	INT6	48.1	INT1	650.4
Mean	9.2		41.7		98.1
SD	7.9		6.01		310.8
Network centralisation	45.1		41.6		35.1

ALL, breeders; LAB, industrial meat processing plants; INT, marketing companies; SD, standard deviation.

Table 3. Effect of nestedness on the product flow.

Effect	Coef.±SE	P>t
ALL	51.71±26.19	0.05
LAB	-16.98±10.38	0.41
DEST	-38.04±16.98	0.03
Nestedness	4.85±1.15	0.00
Constant	3.95±19.15	0.84
R-squared model	0.30

Experimental data

Data were collected from in the period October - December 2008 at the Consortium of Italian Beef Cattle Producers (Consorzio Carni Bovine Italiane, CCBI) on slaughtered Romagnola heifers and referred to the exchanges between the following agent categories: 5 breeders (ALL); 13 industrial meat processing plants (LAB); 9 marketing companies (INT); 51 distribution companies (DEST). Details of these categories help characterise the supply chain system. It was decided to use heifers for the study rather than other slaughtered animals (veals, bullocks, steers) because their smaller numbers make it easier to test the suitability of social network analysis as a means to examine the beef market.

Data collection included: i) structured interviews with technicians and managers of the supply system; ii) collection of quantitative flows of products among the agents.

Connections between the categories were identified by the animal ear tags; in this way, every management phase from breeder to distribution of each animal could be followed. To do this, before analysing the data, an exchange array was compiled where each box identified the different phases of management of each animal through its ear tag; a square matrix was obtained where the number of rows (columns) is equal to the number of the agents (n. ALL+ n. LAB+ n. INT+ n. DEST). Array data were analysed with UCINET software (2010).

The differences within the categories are very important from the point of view of economic analysis. Therefore, the database was provided with the following additional data: farm (size, number of heads, locality); processing agents (slaughtering house, cutting plant); distribution companies (wholesalers, butchers, supermarkets). This information was used to analyse the results obtained by using the UCINET software (2010).

Results and discussion

The system of the relationships among the agents observed was identified through the product flow (Figure 3).

From the perspective of exchange process analysis, the higher the density of the network, the more widespread the organisation of the exchange process. This in turn involves a more competitive exchange. In the sample, only 9.87% of the total potential connections had been established; indicating that the exchange is characterised by a weak competition process.

Three distance measurements were calculated.

–	Average distance (among reachable pairs): the average of the reciprocal distance among the nodes. Sample value is 2.526; this suggests a clear degree of cohesion. In fact, values indicate that the pairs of roleplayers are on average close to each other. Therefore, the system can be considered to be quite close.
–	Distance-based cohesion (compactness): this varies between 0 and 1, and accounts for the different connective routes between the nodes. Sample value is 0.312; this suggests that there are few alternative connective routes for each node, even though the nodes are close together. In other words, the larger the index the larger the cohesion. It can be concluded that in this study there is no great cohesion.
–	Distance-weighted fragmentation (breadth):this varies between 0 and 1. The larger the index the lower the cohesion. More precisely, the index is based on the measure of fragmentation; this indicates the fraction of the pairs of nodes that cannot be reached from any other single node. Sample value is 0.698; this confirms the low degree of cohesion.

These three measurements indicate that the agents activate short distance relationships but that these relationships remain rather sparse. This structure reflects the typical short circuit organisation (Sonnino and Marsden, 2006), but also indicates the possibility that some agents may assume a more central role, thus exploiting the weak cohesion between agents. Figure 3 shows that a complex local system of exchange in most of the flow process is concentrated in one main subsystem of relationship, while the remaining relationships are organised in two other smaller subsystems. This clarifies the meaning of the cohesion, indicating that the short distances typical of the local exchange system are associated with an internal articulation of the system. This would suggest different economic means to channel the product flows and to promote the collection and processing of information both by and for consumers and producers. This is of critical importance as it illustrates the internal differentiation of the exchange system and thus suggests possibilities of differentiating strategies of technological innovation and assistance.

Measurements of centrality provide information about the importance of the node (Borgatti, 2005) and of the possibilities it offers for communication (Freeman, 1979). In our context, both these factors are indexes of the potential of an agent (e.g. somebody involved in meat processing or a farmer) to develop successful trade strategies (Table 2).

It was found that the average degree of centrality is approximately 9 (ties/node), but 20 agents manage at least 15 ties and 5 of these manage over 20 relationships (Table 2). Network centralisation is 45.1% indicating that almost half of the nodes have adjacent nodes. Therefore, the remaining nodes do not have many opportunities or alternatives within the exchange process. The degree of centrality indicates a certain asymmetry in the possibilities of exchange.

Furthermore, centrality is also calculated as betweenness, indicating the ability of the individual agent of being at the centre of relationships, i.e. the frequency with which each node found in the tie connects to two other nodes. In terms of product flow, the betweenness indicates how much a network agent may influence the behaviour of the other agents by affecting the competitive process and the development of strategy choice.

The closeness centrality indicates a different situation. In fact, the largest values do not differ so much from the remaining ones (mean 41.7 which is very close to the highest value, with a small standard deviation). This indicates that, despite the centralisation, the distance from the nodes to their closest neighbours is similar.

As for the betweenness centrality, the sample value indicates that there really is a concentration process and that only some agents sustain their own betweenness: the average value is approximately 98, but just 9 agents have over the average values of betweenness (from 135 to 2123.4), whereas the centralisation index is very high (35.1%). Therefore, the nodes that are found between other nodes sum up an overall probability to be tied to 2 other nodes. This indicates that the agents corresponding to those nodes have a large probability of influencing the relationships between the other nodes.

Table 2 also shows that there is a difference in terms of centrality among the first 5 agents who manage a large number of relationships. ALL3 shows the largest degree of betweenness; this is approximately twice the large centrality of LAB5, while the remaining 3 agents are less important, even though their centrality is much higher than the average values. This indicates that only a few agents (only 2 in this case) are able to act at the centre of the system of relationships of exchange.

The hub score (relating to the fact that a company may supply another company that already has a large number of other suppliers) and authority score (which relates to the fact that a company receives supplies from companies with many customers) were also identified. Figure 4 shows each agent, with four groups being identified on the basis of their average hub and authority scores (Borgatti and Li, 2009). Five agents are in the Agile group (high hub and authority scores): these agents are operating in a very competitive environment. Two agents are Sales Oriented (above average hub scores and below average authority scores), 2 agents are Procurement Oriented (lower than average hub scores and higher than average authority scores), while most of the agents are in the Comfortable group (hub and authority scores both below average). This classification confirms the internal differentiation of the network examined. The differences in terms of the competitive processes undertaken suggest that just a few agents (Agile group) are in a better position to gather information from customers and then to process their knowledge (Ancori et al., 2000; Martino and Polinori, 2011) of consumers’ requirements and opportunities for suppliers. These agents are expected to be able to channel their perspectives and patterns of knowledge, potentially contributing to define the direction and the rate of innovation.

The nestedness structure consists of a hierarchy of 16 groups that at the lowest level includes 75 role-players who are members of 19 connected components, and 4 role-players connected within a unique group. ALL3 is involved in groups 5 and 9 (Figure 5) that include the largest number of role-players. This would suggest that the flow of products and market information appear to be mainly channelled through these groups. This supports the hypothesis that a network rather than a linear market pattern characterises the exchange process investigated.

Questions may arise about the influence of the nestedness on the intensity of the exchange flows. Membership of more deeply nested subsets could allow the role-player to manage a larger amount of product and information flow. Therefore, to interpret the structural cohesion, we can reason backwards from the effect to the cause by considering the flow of products (total amount of items exchanged by agents) in terms of the agent characteristics (ALL, LAB, INTERM, DEST) and the nestedness (measured by the position of the cut set in which the agents operate). We could expect that the more intensive the nestedness (i.e. the denser the relationships), the larger the flow. The agent characteristics are operationalised by dummy codes (the term INTERM was dropped to avoid confusion). Table 3 shows that product flows increase with the nestedness (a large and statistically significant coefficient is estimated). This supports the hypothesis that a more intensive nestedness allows larger flows to be managed. Notably, the coefficient of ALL is positive and statistically significant while the coefficient of DEST is negative; note that the coefficient for LAB is not statistically significant. These findings indicate that it is the breeders who are the main managers of product flows while these flows would be reduced by the final buyers (DEST). This emphasises the role of the breeders in structuring the market relationships according to the other role-players.

Conclusions

The Romagnola meat market reflects the general situation of the typical product market: i) it operates within a restricted area; ii) flows are, therefore, regulated by only a few agents; and iii) technical innovation (advanced selection methods and strategies) must take this into account.

Results of an analysis of nestedness are very interesting; they show the strong role of the breeders in this product. If confirmed by more practical research, this finding could allow more specific and targeted selection procedures and a more efficient production system to be set up.

The evidence in this study take into account the need for a stronger role for the farmer in directing flows and to concentrate the system around those farmers who are of greater economic importance.

This study has shown the relevance of social network analysis to interpret and understand market indications and to promote a greater awareness in the beef selection process. Other practical studies will be carried out on more popular beef categories (bullocks) to obtain the information necessary for planning selection strategies to meet market requirements.

Acknowledgments

The authors would like to thank the Consortium of Italian Beef Cattle Producers (CCBI) that provided all the data and supported the interpretation of results.

The research was financially supported by Fondazione Cassa di Risparmio di Perugia, Project N. 2012.0261.021.