1 Introduction and Background

Cooperative agreements have long attracted the interest of scholars for their “great efficiency in moving information, innovation, routine experience, and resources that enable firms to increase their organization learning, adaptation and competitive advantage” (Baum et al. 2003). Inter-firm cooperation has been addressed widely in the industrial literature, particularly in the automotive industry (Hüttenrauch et al. 2008), the high technology industry (Gomes-Casseres 1996; Patibandla and Petersen 2002; Roijakkers 2003), iron and steel production (Rosegger 1992), research and development (Staropoli 1998; Grønhaug 1989), and cooperative advertising (Li et al. 2002). However, as observed by Chen and Yahalom (2013), although cooperative agreements and alliances are main trends in the shipping industry, scant analyses have focussed on this topical industrial strategy.

The shipping industry, which is noted for its freight rate instability, imbalanced cargo flows and high degree of required financial commitment (i.e. significant capital investments in assets) is thus a volatile and risky sector with strong competition among carriers (Graham 1998; Stopford 2009; Notteboom et al. 2010; Satta et al. 2013); for instance, only five major ocean carriers (Maersk, MSC, CMA-CGM, Evergreen Line, and APL) operate nearly 50 % of the cellular fleet in the world. Carriers have therefore used various cooperative agreements to gain competitive advantage, share investment risk, and reduce costs. In the late 19th century carriers started to negotiate agreements in order to stabilize the supply chain and assure steady services on different maritime routes (Stopford 2009). Furthermore, consortia were introduced in the 1960s with the aim to lower costs and achieve efficiency in the production cycle due to the introduction of new typologies of vessels such as the deployment of mega-vessels (Drewry Shipping Consultants 2005; Imai et al. 2006). From the 1990s onward, cooperative agreements in shipping have had the twin goals of sharing risk and widening carriers’ markets in order to develop economies of scale and scope (Brooks et al. 1993; Alix et al. 1999), strengthen global, vertical (Soppé et al. 2009) and/or horizontal integration (Midoro and Pitto 2000; Panayides and Cullinane 2002; Ferrari et al. 2008; Lam and van de Voorde 2011) and sometimes test if there are sufficient economic conditions for mergers (Gardellin et al. 2011).

Within this framework, the container shipping industry is particularly interesting because in the last twenty years it has experienced tremendous growth rates triggered by the delocalization of manufacturing activities and the expansion of the world seaborne trade of components and finished products (UNCTAD 2012). To keep pace with market demand and decrease risks through diversification across the different business segments (liquid bulk, dry bulk, container, general cargo, etc.), as well as defend their market share, container companies have begun to follow aggressive corporate growth strategies. Whereas if traditionally, shipping companies have entered into the market largely alone, more recently this trend has been overturned by more and more cooperative agreements (Rimmer 1998; Alix et al. 1999; Lorange 2001). The container shipping industry has thus increasingly introduced cooperative strategies into its organization, resulting in a growing market concentration (Lorange 2001). Several types of cooperative strategies, from informal consortia to formal strategic alliances, are present in the container maritime market. Strategic alliances are agreements where carriers manage several joint shipping services worldwide, and by doing so, alliance members also share the investment risks. Two major strategic alliances currently operate in the market: G6 Alliance (derived from the merger between Grand Alliance and New World Alliance) and CKYH Alliance; it is noteworthy that these two alliances involve many of the world’s leading carriers. In June 2013 (Financial Times 2013) a new agreement known as P3, among Maersk, MSC and CMA-CGM concerning the Far East-Europe shipping route, was announced. Since the agreement involves the “big three” shipping companies in the world, it will be subject to the approval of the antitrust agencies of Asia, Europe and the US. Whatever the final decision turns out to be, this trend clearly shows how shipping cooperative agreements will play a crucial role in the growth and development of the maritime industry. And indeed, as observed by various scholars, the economic and financial crisis started in 2008 has determined a significant decrease in freight rates as well as chronic overcapacity, but has also resulted in an increase in cooperation among container shipping carriers (Hoffmann 2010; Panayides and Wiedmer 2011; Notteboom and Rodrigue 2012).

In the present paper we focus on the container industry, and in particular on the cooperative agreements based on the sharing of on-board slots: which may be either slot-charter (S-C) or vessel-sharing (V-S). Slot-charter agreements are those in which a shipping companyFootnote 1 books a certain number of slots on vessels operated by a partner without deploying any (owned) ship (Panayides and Wiedmer 2011); this type of agreement is implemented in order to streamline operational costs. Whereas in a vessel-sharing (V-S) agreement, each carrier deploys its own fleet and allows other partners to place a number of slots at their own disposal.

Against this background, in this study we apply network analysis and construct the Cooperative Container Network in order to study how shipping companies integrate and coordinate their activities. In constructing our network, we assume that, as discussed previously, carriers join inter-firm networks for several reasons, including the achievement of economies of scale (supply of services) and scope (joint utilization of equipment and know-how) and in order to enter into new markets (Caves and Porter 1977; Goerzen and Beamish 2005; Hoetker and Mellewigt 2009). The contribution of this study is therefore given by the fact that network applications in the maritime literature—particularly of analyses relating to container shipping—are relatively few. Some studies have applied network analysis to worldwide movements of cargo disaggregated at port level (Bergantino and Veenstra 2002; Koluza et al. 2010; Ducruet and Notteboom 2012; Parola et al. 2013); other studies investigate port choice behaviour (Tang et al. 2011), but most contributions focus on port efficiency, global cargo networks and maritime freight rates (Barros 2005; Cisic et al. 2007; Ducruet et al. 2010; Márquez-Ramos et al. 2011). In our network analysis application we aim to demonstrate that the cooperative container network has a small world structure, i.e. a network where firms cluster into dense cliques that foster communication and cooperation (Phelps and Schilling 2005). Through a stepwise analysis we reach different interesting results that support our idea that network analysis is an important tool for understanding and mapping specific patterns of cooperative agreements in the container shipping industry.

The paper is organized as follows. Section 2 addresses a literature review on inter-firm networks. Section 3 describes the dataset and network modelling applied to construct the cooperative container network (CCN). Sections 4 and 5 present the network analysis. In Section 6 we discuss the results of our study, set out the main implications for the container industry, and propose future research developments.

2 Literature Review on Inter-firm Networks

Network analysis provides a solid background to study the impact of agreements on firms, and scholars have often applied network analysis in order to disentangle the mechanisms at the centre of complex economic and financial cooperative interdependence among firms (Grandori and Soda 1995). Network theory is an effective paradigm for analyzing cooperative agreements between firms (nodes in the network) connected by links which may, for example, be franchising agreements, joint ventures, sub-contracts, and interlocking directorates. Network analysis has allowed scholars to scrutinize the strategic behaviours and partnering of firms that are involved in inter-firm agreements.

Network theory has its roots in the work of Euler in the first half of the 18th century, Solomonov (1951) and Erdos and Renyi (1960). Their work with the development of the network paradigm represents the first successful attempts to understand and study real world phenomena through the lens of the network (Newman et al. 2003). Another important advance is the introduction by Watts and Strogatz (1998) of a new network class known as small world. Small world networks are characterized by a high local connectivity (i.e. clustering coefficient C) and by a small topological distance (i.e. shortest path l) between each pair of nodes in the network, whereas regular networks have a high C and a high l, and random networks have a low C and a low l. Soon after the seminal contribution of Watts and Strogatz, Barabasi and Albert (1999) proposed a model for the growth of networks based on the notion that “the rich get richer.” In this formulation, networks grow according to power laws and are known as scale free. Networks such as the Internet, the WWW, airline networks, cross-collaboration in science, and many others all belong to the class of scale free networks (Albert and Barabási 2002; Newman 2003; Boccaletti et al. 2006).

In its focus on inter-firm cooperation studies, network analysis has been carried out mainly in accordance with two approaches. The first approach examines the formation, evolution and adaptation of firm networks in order to calculate the levels of performance and allocative efficiency due to economies of scale and scope (Sydow and Windeler 1998; Koza and Lewin 1999). For instance, Gimeno (2004) has argued that competitive strategies evident in network analysis can explain how a firm decides to cooperate and build an alliance with a third party. But the literature is not consistent in its results. Network formation in inter-firm cooperation does not always add competitive advantage to the industrial relationship; sometimes very diverse firms involved in an agreement, as well as their national or corporate culture, can impact negatively on the economic performance of the alliance (Brown et al. 1989; Biggart and Guillen 1999; Goerzen and Beamish 2005). The second approach applies network analysis to the characterization of the structure of firm relationships (Stuart et al. 1999; Dittrich et al. 2007); in this approach one examines the inter-firm network topology so as to understand firm behaviour in the choice of partners and impacts of the strategies of firms on alliance networks (Ahuja 2000; Goerzen and Beamish 2005; Zaheer and Bell 2005; Dittrich et al. 2007).

From our perspective of analysis it is interesting to observe how in the literature the level of interconnectedness of a firm’s partners can influence a firm partnering strategy. According to Rowley et al. (2000), highly interconnected firms can generate local clusters that may force a firm to cooperate in order to avoid being excluded from commercial agreements. Firms may therefore use existing cooperative agreements to connect to a third party (i.e. indirect links) in order to maximize inter-firm network benefits such as an increase in innovation, output and maintenance costs (Ahuja 2000). Furthermore, firms may also gain competitive advantage from their position within a network; a firm acts as a conduit to connect disjointed clusters and exploit strategic information from both sides (Zaheer and Bell 2005). Various authors have also extended this analysis by examining the global interconnectedness of inter-firm networks and the influence on firms’ commercial strategies. Garcia-Pont and Nohria (2002) have argued that there is no evidence to confirm whether the global interconnectedness of a network can impact on individual firm behaviour; this means that each firm acts according to its own commercial advantage without considering the effect of its commercial cooperative strategy on other firms.

However, an interesting element of network inter-firm cooperation is that it can be identified as a small world network (Watts and Strogatz 1998) because small world networks often encapsulate the many characteristics of industrial cooperation. Small world networks are sparsely connected networks with high local connectivity, and therefore firms identified as small world networks are more likely to disseminate information and implement best practice among them (Watts and Strogatz 1998). Following these findings, our approach will examine container shipping companies as firms that cooperate and which are also identified as small world networks. In so doing, we are in line with several scholars (Goerzen and Beamish 2005; Hoetker and Mellewigt 2009; Sullivan and Tang 2012) who have verified that small world networks are best able to capture inter-firm cooperation. Baum et al. (2003) in particular have evaluated the origin and evolution of small world networks in inter-firm alliances in order to bring to light the factors underlying the formation of these network structures. They argue that in the case of firm alliances, the evolution of a network is neither random nor due to coordinated behaviours, but is instead driven by chance partnering and strategic manoeuvring (insurgent and control partnering), and this is key to our analysis as well. Thus, the small world network structure is none other than the result of local interactions among firms that generate complex non-local cooperative behaviours. In the next section we discuss the data set we have exploited to investigate cooperation in the container industry and present the cooperative container network (CNN).

3 Modelling and Data Description

We select 24 major ocean carriers (that cooperate with 41 other carriers) on the basis of total fleet capacity, and construct the network of all the cooperative agreements for the given set of carriers (65 in total). Information has been collected through the Containerisation International Journal and refers to January 2010. In this way we provide a picture of a specific moment in time and avoid any bias, since there is commonly variability in the supply of shipping services. In order to design the Cooperative Container Network, we calculate the Weekly Container Transport Capacity (WCTC) deployed by each carrier of the selected set per container service, as defined by Frémont and Soppé (2004). In addition, the data set includes for each container service: frequency, fleet capacity (Twenty-Foot Equivalent Units—TEUs) deployed by one or more carriers on the same service, and existence of vessel-sharing (V-S) or slot-charter (S-C) agreements. As mentioned above, container carriers cooperate in accordance with either vessel-sharing (V-S) or slot-charter (S-C) agreements. Figure 1 shows a simplified representation of container cooperation schemes, where carriers α, ß, and γ operate container service S among ports A, B, C, D, E, F, and G with a Weekly Container Transport Capacity (WCTC). Let us notice in the figure that one slot-charter, carrier δ, does not use its fleet to provide the service but instead rents container slots from carriers α, ß, and γ (vessel-sharing). The complete data set is thus comprised by a subset of 603 container services split into two categories: services managed by single carriers (205) and services managed by multiple carriers (398).

Fig. 1
figure 1

Schema of the construction process of the Cooperative Container Network (CCN). Blue nodes represent carriers that share V-S relationships and red nodes indicate S-C relationships in the S service

Full size image

The Cooperative Container Network (CCN) is represented by an indirect graph G (N, L) composed of two sets: N is the set of nodes, N ≡ {n1, n2, … nN} and L is the set of links, L ≡ {l1, l2, … ln}. We use a mathematical representation for the CCN composed of a NxN adjacency matrix A, whose off-diagonal entries aij are equal to 1 when we have one or more shared service between companies i and j (i ≠ j or j ≠ i), and 0 otherwise. Diagonal elements aii are set to 0 because there is no intra-firm relationship in our modelling. Since we have constructed an indirect graph, the adjacency matrix A is symmetrical. We extend the topological representation provided by the adjacency matrix by constructing the weighted adjacency matrix W, whose off-diagonal elements wij are the sum of service capacity Q S WCTC shared by carriers i and j. The weighted matrix W provides us with a richer description of the CCN because it considers the topology in addition to quantitative information on the market share between pairs of carriers. The two fundamental types of agreements (V-S and S-C) yield the mutual interactions between pairs of carriers in our network modelling. Each node in the network corresponds to a carrier and the links represent cooperation between carriers. Total vessel capacity available for each service is thus represented by tvc.

In our depiction of five cooperative and fleet characteristics of the major carriers of our data set shown in Table 1, two main patterns emerge: all of the carriers in our set provide liner services through cooperative agreements, but in varying degrees, and only the MSC operates more services independently than in cooperation with other carriers. A second pattern is apparent when we observe the number of independently operated services; after the sixth major carrier (Cosco Container Lines), the number of digits of services operated independently drops off by one unit. Data shows that cooperative agreements are the preferred organizational and logistics solutions of minor carriers, whereas mega-carriers such as the MSC still operate a number of services independently due to the global services they have built up over the years.

Table 1 Characteristics of the main carriers ranked according to total fleet capacity (TEUs) in January 2010. On the far left row, letters in parentheses identify membership in one of the global alliances: CKYH Alliance (A), G6 (B)
Full size table

In Table 2 we rank the ten most frequent cooperative agreements in our data set. The Asian players clearly dominate the ranking and are the most likely to cooperate with other carriers through strategic alliances.

Table 2 Ranking of the leading cooperative agreements in the CCN. Wij represents the sum of service capacity shared between carriers
Full size table

Figure 2 displays the complete CCN composed of N = 65 carriers (nodes) and L = 287 cooperative agreements (links). The colour of each node is drawn as a function of the number of connections from red (high number of connections) to light yellow (low number of connections); the size of each node (i.e. carrier) is proportional to the number of TEU operated in January 2010; and the width of each link is proportional to the level of cooperation between nodes i and j.

Fig. 2
figure 2

Visualization of the CCN

Full size image

From a network perspective, our CCN is sparse because its density is equal to 0.086,Footnote 2 but it has a relatively small average shortest path (2.2),Footnote 3 given that the value of maximum shortest path to connect the most remote nodes in the graph is five links. The analysis of shortest path shows us that, on average, nodes are close to each other; in other words, having obtained an average shortest path equal to 2.2 means that the cooperative container network is a tight network based on cooperative agreements among shipping companies. In the next sections we examine the CCN in greater detail in order to characterize the topological structure of the container shipping industry.

4 Analysis of the Cooperative Container Network (CCN)

In this section we examine the characteristics of the Cooperative Container Network (CCN); in doing so, we identify the carriers (nodes) that play a relevant role in terms of cooperation. Our aim in this section is to verify if highly connected carriers play leading roles in the organization of strategic container agreements and the management of services. Our first step is to calculate for each carrier i the level of connectivity, that is, the degree k (or number of cooperative partners in our specific case study) in the following expression:

$$ {k}_i={\displaystyle {\sum}_{j\in \vartheta (i)}}{a}_{ij} $$
(1)

where ϑ(i)  is the set of topological neighbours of carrier i, i.e. the carriers (nodes) directly connected to i, and aij is the element of the adjacency matrix A. The value of degree k in our set ranges between 1 and 31, and the average degree k is equal to 8.8.

Let us now estimate the cumulative probability distribution of degree k, which can give us an idea about the interactions among shipping carriers. The degree distribution in our analysis (Fig. 3) points to two different regimes: a power law in the first part and an exponential regime in the final part of the tail of the distribution. The black dots in Fig. 3 depict the log-log scatter plot. The first regime is well-approximated by a power law with an exponent of −0.58 (adjusted R2 = 0.845 for the complete distribution). Moreover, the cumulative probability distribution has a cut-off point at approximately degree k = 20 and p < 0.2. We can conclude that in our sample we have a higher probability of carriers with less than 20 cooperative partners, and that the CCN does not belong to the family of random networks. Cooperation is therefore not uniformly distributed among carriers. Nonetheless, we also need to emphasize that this conclusion may be influenced by the limited number of observations in our data set (65 nodes). Some authors have suggested that x-axis of a power law distribution should range over more than three orders of magnitude (Farmer and Geanakoplos 2008; Clauset et al. 2009). In the CCN the total number of nodes influences the variance of degree k (maximum possible degree k is 64).

Fig. 3
figure 3

Log-log scale plot of probability distributions of degree k for the CCN and power law fitting curve. The probability distribution is shown as black dots and the fitting curve is red

Full size image

When we analyze the relationship between the network connectivity and the commercial capacity of each carrier, the linear regression in Fig. 4 shows a significant correlation between degree k (i.e. number of cooperative partners) and total fleet capacity. In our set the carriers with a small operated fleet capacity (TEUs) have limited cooperation (k < 10); the majority of these carriers cluster together in the low-k and low-TEU region of Fig. 4. Carriers with higher carrying capacity than in the previous case are more scattered and clearly do not cluster. In this case we can infer that carriers use cooperation as an industrial strategy to establish their leadership roles in the market. Two cases are worthy of mention: the mega-carriers MSC with k = 9 and Maersk Line with k = 17. MSC, for instance, follows an industrial strategy characterized by a large total fleet capacity (i.e. high level of market share), which allows this company to behave as an independent operator. Carriers such as MSC and Maersk can operate independently due to the large portfolio of clients they have accumulated over time.

Fig. 4
figure 4

Relationship between degree k and total fleet capacity in TEUs (January 2010). The linear regression has a correlation coefficient of 0.86 but does not include the two players situated in the uppermost area of the figure (MSC and Maersk)

Full size image

At this point in the analysis we may ask, do carriers select their commercial partners on the basis of similar industrial strategy (i.e. cooperation or independence)? To answer this question we estimate what is known as network assortativity. This measure provides information on the tendency of nodes/carriers to connect to other nodes/carriers with a similar level of connectivity (Newman 2003). The assortativity coefficient ranges between 1 (the highest value) to −1 (the lowest value). The assortativity coefficient r is defined as follows:

$$ r=\frac{\mathrm{Tr}\ \left(\mathbf{e}\right)-\left|\right|{e}^2\left|\right|}{1-\left|\right|{e}^2\left|\right|} $$
(2)

where e is the vector of entries from the network mixing matrix aij, Tr(e) is the trace of the matrix A and ||e 2|| is the sum of the squared entries.

In the CCN we obtain a small assortativity coefficient (r = −0.16), and when we consider the top ten most connected carriers, we find a similar assortativity value (r = −0.14). These results demonstrate that container shipping companies do not select their partners on the basis of similar industrial strategies; in fact, with a value of assortativity at −0.14, the carriers which are the most connected do not necessarily consider connectivity as a determinant factor for entering into partnerships with other carriers.

In the next section we will compare these results with the capability of carriers to topologically cluster together, in other words, we will verify if our network is a small world network.

5 Cooperative Clusters in the CCN

We have shown that degree k represents local connectivity of the CCN from the viewpoint of a single node, but we do not yet know how the partners of this node link together. Therefore, at this stage of the analysis we need to calculate inter-local connectivity, i.e. the clustering coefficient. The clustering coefficient for carrier/node i is defined as the number of closed paths of length two, normalized by the maximum possible number of paths of length two.Footnote 4 Although the literature identifies at least four different mathematical expressions for the clustering coefficient (Saramäki et al. 2007), in this study we adopt the topological formulation of the clustering coefficient as proposed by Watts and Strogatz (1998) in their seminal work on small world networks, which yields the following definition:

$$ C(i)=\frac{2E(i)}{k_i\left({k}_i-1\right)} $$
(3)

where E(i) is the number of the nearest paths of node/carrier i and k i (k i - 1)/2 is the maximum number of possible links among the neighbours of node/carrier i. Clustering coefficient C in this formulation ranges between 0 and 1.

For our Cooperative Container Network we obtain values of clustering coefficient close to 1 for carrier i when its first neighbours (i.e. carriers with direct cooperation in the CCN) show a mutually higher number of links; otherwise we have values close to 0. Figure 5 shows values of degree k, versus clustering coefficient C, for each carrier. The trend is approximated by a linear decay law (correlation coefficient of 0.81). The regression excludes carriers with a clustering coefficient of zero.

Fig. 5
figure 5

Scatter plot of degree k, versus clustering coefficient C. For the purpose of clarity, we plot the names for a selected subset of carriers

Full size image

The results obtained relating to degree k with clustering coefficient C are in line with our findings in the previous sections. In this case we observe a main cluster in the data. The cluster includes carriers with low values of degree k and high clustering coefficient C (regional carriers), implying that carriers with only a few cooperative links and low levels of operated fleet capacity generally tend to link together. In other words, cooperative agreements are more easily established among regional players that tend to be specialized in local markets.

In Fig. 6 we depict the sub-networks of a sample of four cooperative carriers. We can observe that, as the clustering coefficient increases (from bottom to top), the links are more equally distributed among the carriers.

Fig. 6
figure 6

Representation of sub-networks of some carriers in the CCN

Full size image

After having carried out our stepwise analysis, we can now verify that the CCN is a small world network with an average shortest path (2.2) that scales as the logarithm of the number of nodes in the network (65); the CCN shortest path has the same order of a random graph with the same number of nodes/carriers, while the average clustering coefficient is equal to < C(i) > = 0.55, a very high value, higher than the case of a random graph with the same number of nodes/carriers (<Crand (i) > = 0.1). The result shows that, as a small world network, medium and small companies in the CCN are the most cooperative players. Small world networks are locally clustered networks with short path lengths connecting different clusters, meaning that apparently disconnected carriers (as in the case of small and medium carriers) may easily be indirectly connected through a few intermediary companies. In the case of the container industry we can confirm that the most cooperative companies are small-to-medium-size carriers that engage in commercial agreements in order to reduce costs (hedge risk and increase scale and scope economies), and also when in partnership with larger carriers, these cooperative companies are able to compete, especially against the three largest carriers which absorb around 40 % of the total fleet capacity (Table 1).

The reader is invited to peruse the Appendix, where we have developed a community detection analysis over the CCN so as to provide a detailed picture of the membership of families of carriers (clusters) created by cooperative agreements in the container industry. Although it is interesting as a descriptive tool, due to the low values of modularity Q provided by six different methods (Table 3), the community detection analysis does not allow us to reach conclusive results. For example, we had expected to detect clusters of carriers involved in the two major alliances active at the time of observation; this has happened in a few cases (Table 5). However, it is interesting to note that the three major ocean carriers have recently created the P3 alliance (mid-2013) and are coincidentally members of the same cluster. This situation could be the result of long-term cooperative agreements (sharing of on-board slots) between Maersk, MSC and CMA-CGM, which set the stage for the launch of the P3 alliance.

In conclusion, the various findings of our analysis have provided empirical evidence to indicate that small world cooperative networks have a positive impact on the container shipping industry. This organizational strategy has strong positive effects, particularly on small and medium size shipping companies that are now able to reduce financial risks and increase economies of scale and scope. But above all, through partnerships with global shipping companies, small and medium companies gain access to larger portfolios of clients that the international companies have built up over time in different parts of the globe.

6 Conclusion

The dramatic economic downturn occurring in 2008 has prompted carriers to increasingly seek cooperative schemes for their operations. In the mature and highly-competitive container shipping market, cooperation has strategic value because it allows for the reduction of investment in assets and the increase of load factor aggregating demand flows. In this work we have examined a sample of 603 container services distributed worldwide involving 65 carriers; we have then constructed a network of cooperative agreements of container services of our sample which we have labelled Cooperative Container Network (CCN). By applying network analysis, we have demonstrated that CCN belongs to the class of small world networks with high levels of local interconnectivity. We have shown that no random patterns emerge from cooperation among container carriers, thereby implying that they abide by rules in their cooperative relationships.

As a small world network, in the CCN we observe that carriers with small and medium total fleet capacity have cooperative links with other carriers sharing similar characteristics. However, when fleet capacity increases, the clustering coefficient decreases linearly. Interestingly from this point of view, is that a carrier with high capacity engages in cooperation with other carriers by merely looking for local carriers in order to increase its local and specialized market penetration. This result suggests how cooperative agreements among shipping carriers can indeed lead to greater diffusion of knowledge and information across the market and facilitate operational services, for example, in specific regional markets. However, when container carriers reach significant levels of total fleet capacity, they tend to achieve a level of connection that allows them to exert dominance in the market. The mega-carriers (MSC and Maersk) still operate a number of services independently due to the global services they have developed over many years. MSC and Maersk operate around 32 % of the container fleet deployed by top shipping firms (Table 1). From this perspective, on the one hand small and medium carriers use cooperative agreements as a way to reduce risk and reach economies of scale and scope. But when they are in partnership with international companies, small and medium carriers can use cooperative agreements as a means of achieving greater market scope and geographical coverage. On the other hand, large carriers tend to establish agreements with small carriers in order to penetrate the more specialized local markets.

Beneficial future research would be to evaluate cooperative attitudes among different shipping companies by examining the impacts of the maritime network structure in terms of hub-and-spoke solutions and geographical diversity. We suggest that it would then be possible to measure the level of investment risk that may affect cooperation agreements in the container shipping industry. The present paper has analyzed cooperative agreements from the carrier’s point of view; an extension of this work could apply network analysis to a study of container services.