Robustness of the international oil trade network under targeted attacks to economies
aa r X i v : . [ ec on . E M ] J a n Robustness of the international oil trade network under targeted attacks toeconomies
Na Wei a , Wen-Jie Xie a,b, ∗ , Wei-Xing Zhou a,b,c a School of Business, East China University of Science and Technology, Shanghai 200237, China b Research Center for Econophysics, East China University of Science and Technology, Shanghai 200237, China c Department of Mathematics, East China University of Science and Technology, Shanghai 200237, China
Abstract
In the international oil trade network (iOTN), trade shocks triggered by extreme events may spread over the entirenetwork along the trade links of the central economies and even lead to the collapse of the whole system. In thisstudy, we focus on the concept of “too central to fail” and use traditional centrality indicators as strategic indicatorsfor simulating attacks on economic nodes, and simulates various situations in which the structure and function ofthe global oil trade network are lost when the economies suffer extreme trade shocks. The simulation results showthat the global oil trade system has become more vulnerable in recent years. The regional aggregation of oil tradeis an essential source of iOTN’s vulnerability. Maintaining global oil trade stability and security requires a focuson economies with greater influence within the network module of the iOTN. International organizations such asOPEC and OECD established more trade links around the world, but their influence on the iOTN is declining. Weimprove the framework of oil security and trade risk assessment based on the topological index of iOTN, and providea reference for finding methods to maintain network robustness and trade stability.
Keywords:
Global oil market, oil security, international oil trade network, network robustness, targeted attackJEL: C1, P4, Z13
1. Introduction
Oil as the primary fossil fuel is a vital strategic resource and widely used in the chemical industry, transportation,and other industries. The uneven distribution of oil on a global scale leads to the separation of supply and consumptionmarkets. As a result, international oil trade becomes a critical complement to the balance of oil supply and demand[1–3].Oil trade is not a simple point-to-point transaction, displaying the characteristics of a chain network structureacross economies and regions [4]. Therefore, the oil trade system can be abstracted as an international oil trade net-work (iOTN),in which the economies are nodes and the trade relationships between them are edges [5]. Fluctuationsin oil consumption, supply, demand, storage, and price can all cause changes in the global trade pattern and worldeconomic situation [6–8]. It is of great significance for the government to know the development trend of oil trade,recognize the economy’s trade status, and prevent oil supply risks [9–11].Oil trade promotes energy cooperation and economic development. However, the iOTN as a complex systemalso faces shocks and challenges. In the closely connected iOTN, not only oil resources but the trade shocks andsupply risks can be transmitted through trade relationships between economies. In 2014, the sharp increase in shaleoil production in the United States caused the U.S.-Africa oil trade volume to plummet to its lowest point in 40 years,and Africa became the worst hit economy. The plunge in the U.S.-Africa oil trade has not only disrupted the mostimportant economic ties between the U.S. and Africa, but has also had a ripple effect globally. Many economies ∗ Corresponding author. Corresponding to: 130 Meilong Road, P.O. Box 114, School of Business, East China University of Science andTechnology, Shanghai 200237, China.
Email addresses: [email protected] (Wen-Jie Xie), [email protected] (Wei-Xing Zhou)
Preprint submitted to Elsevier February 1, 2021 uffered from potential trade shocks brought about by changes in oil imports of the U.S. This phenomenon of risktransmission and diffusion can be attributed to the cascade effect of network [12, 13]. The above facts show that whenunexpected events impact trade activities, the structure and functions of the iOTN will be affected and the globaloil security may be endangered. Therefore, it is of great practical significance to integrate the study of trade risktransmission mechanism based on complex network theory into the framework of energy security assessment.In recent years, focusing on economic fragility caused by factors such as oil supply interruption and politicalinstability, there has been a growing literature on global oil security [14–16]. A large body of literature providesmulti-angle, multi-dimensional refinements for the energy security assessment framework [17, 18]. However, thereis little literature on integrating the topological indicators of the iOTN into the traditional oil security assessmentframework [19]. How the roles and status of economies in the iOTN affect oil security and trade stability is still anopen problem worth exploring.Considering the oil security and trade structural stability implies the need to explore the stability of the iOTNagainst extreme events or the robustness when encountering cascade failures. In other words, the iOTN needs tohave the ability to maintain certain structural integrity and functions when the economies suffer from terrorist attacks,foreign policy changes, and the energy trade channels between economies are blocked or interrupted. Therefore,robustness is an essential dynamic characteristic of the network system. The higher the robustness of a network, themore stable it is [20].The international oil market is turbulent, and the political stance of the economies and international relations arecomplicated. It is difficult to capture and detect the real changes of the oil trading system and the cascade effect afterthe impact, and it is even more difficult to conduct quantitative analysis and research. Firstly, we construct iOTNsbased on international oil trade data sets. Secondly, we use the numerical simulation method to simulate attacks on theiOTNs and analyze the network’s robustness changes under various targeted attacks. Finally, through the comparativeanalysis of different attacks, we find the effective way to maintain network robustness and trade stability, and providesuggestions for oil trade policy formulation and market supervision.This article is organized as follows: Section 2 reviews the literature; Section 3 introduces research data andmethods; In section 4, we construct the international oil trade networks and uses the related method to conduct theempirical analysis; Section 5 is the discussion and application.
2. Literature review
According to the resource dependence theory, close trade relations are needed between economies to exchangeresources to meet their needs. Those complex relations often form a trade network pattern and promote the process oftrade globalization. Network science become an essential method for analyzing the pattern of trade network [21–23].As the most crucial energy resource, oil’s global trade always attracts the attention of researchers. The inter-national oil trade system is evolving into a stable, integrated, and more efficient system [4, 23].Economies are moreinterconnected, but global competition for oil resources has intensified. The addition of industrialized economies suchas China and India increase the uncertainty of the development of oil trade [24]. In addition to factors such as supplyand demand, technological advances and energy efficiency, geographic location, and competition and dependencebetween economies have increasingly prominent influence on oil trade [6, 25, 26]. The researches about the iOTN atthe system and individual levels reveal the general characteristics of oil trade and deepen the understanding of the oiltrade system [7, 8]. The researches also have important practical guiding significance for the position recognition andtrade strategy formulation of the economies in the iOTN.The complexity and vulnerability of the oil trade system still leaves many issues to be explored and resolved [27].Therefore, the stability of the iOTN and trade risks are closely related to energy security and the stable developmentof global economy, and are hot research issues. The risk and instability arise from the chain impact of the failure ofeconomies through the trade relationships in the iOTN, which can also be called cascade failure. When the cascadeshock reaches a certain level and scope, it will cause the outbreak of systemic risks in the iOTN. It may eventuallycause the collapse of the entire system and have a continuous impact on the global economy [28]. Then, the stabilityof the iOTN is crucial to the development of economies and the global oil trade market [29, 30].From the concepts of “too big to fail” and “too connected to fail” to the concept of “too central to fail”, numerousresearchers have paid attention to these concepts and have discussed them extensively. In this article, we focus on “too2 anada
China
France
Germany
Iran
Iraq
Italy
Japan
Rep. of Korea
Netherlands
Nigeria
Norway
Russian Federation
Saudi Arabia
India
Spain
United Arab Emirates
USA
Venezuela
Figure 1: The international oil trade network in 2017. Each node represents an economy, and a directed edge represents the oil trade relationshipfrom one economy to another. The size of a node implies the number of the economy’s partners. Different colors of the nodes indicate that theeconomies belong to different modules. central to fail” [31, 32]. Central economies are very important for maintaining the structural stability and play a bridgerole in the formation of trade networks [27]. Traditional measures of network node importance or influence, such asdegree, betweenness, and closeness, can be used to determine the economies with dominant positions [33, 34]. Basedon these indicators, many researchers propose new methods to measure the influence of network nodes [10, 32, 35].Different from the methods of directly selecting or designing node influence measures to identify the most influen-tial nodes in previous studies [10, 35], this article selects a series of classic node influence measures, and uses them asstrategic indicators for simulated attacks on the economic nodes in the iOTN. This article explores the changes in thestability of the iOTN structure when important economies are impacted, and compare those indicators. The findingscan better the understanding of the role of economies with different critical positions in the transmission of oil traderisks.
3. Data and Methodology
The oil trade data is downloaded from the UN Comtrade database, and the HS code of data is 270900. The datacomes from official data reported by both trade parties. We chose the import data with broader coverage and betterintegrity [36], which contains the oil import and export trade values of 256 economies from 1988 to 2017. We removeda small amount of economic trade data whose reporting source cannot be identified, such as “Other Asia, nes”. Thedeleted data is small in amount and will not affect the research results.
The global oil trade system can be abstracted as an international oil trade network (iOTN). We define the un-weighted directed iOTN that can be represented by the N × N adjacency matrix A = ( a i j ), where N is the number ofeconomies in the iOTN. The matrix element a i j = i to j . We constructthe yearly iOTNs from 1988 to 2017.Figure 1 shows the iOTN in 2017. The nodes are economies, and the edges with direction indicate the oil traderelationships. The sizes of nodes are set according to the number of the economies’ partners. The more the tradepartners of the economy, the larger the node size. The different colors of the nodes indicate that the economies belongto different modules. The method of module division will be explained in the method section.3 .3. Models of attack to economies in iOTN The scale-free characteristic of the iOTN enables the trade system to resist random attacks but is highly vulnerableto targeted attacks [19]. Previous studies prove that attacks on a single important node can trigger large-scale cascadefailures [37]. A targeted attack model of the economies needs to be constructed to explore the structural stability ofthe iOTN in different attack situations.A targeted attack involves a specific rule or indicator. We take the influence measures as the strategic indicatorsfor the attacks. Simulating the destruction of the network under targeted attacks can serve as a prediction for thecrisis of system collapse. Knowing under which specific circumstances the system is more prone to collapse can be apreparation for the reinforcement and protection of the subsequent system.Based on these strategies and indicators, we select the economies with the higher importance rankings, accountingfor q of the total number of economies in the iOTN, to attack. Then, the selected economic nodes can be removedfrom the network. By changing the selection basis and value of q , we can simulate network structure damage underdifferent attack strategies and degrees. The importance of economies usually depends on the analysis of node centrality indicators [33]. Different cen-trality indicators reveal the influence of nodes in the network from different perspectives. Considering the types andscopes of the network topological connection, we choose three types of influence measures that were commonly usedin previous studies [4, 19]. Then, we use them as strategic indicators for subsequent economic attacks. The impactsof critical economies identified by traditional methods on the stability of the iOTN can be measured, and differentmeasurement methods can be horizontally compared.
The influence of economies based on the local structure is mainly measured according to the local topologicalcharacteristics of iOTN such as degree centrality, local clustering coefficient. Some other methods are also derivedaccording to the different local network structure and scope [38]. We select the most used degree and local clusteringcoefficient as strategic indicators.We define indegree k in and outdegree k in of economy i , which represent the number of import and export rela-tionships. K in i = N X j = a ji and K out i = N X j = a i j , (1)where N is the total number of economies in the iOTN. Local clustering coefficient C c of the economy i is ratio of the number of actual trade relationships between itsneighboring economies to all possible trade relationships [39]. It is defined as: C c ( i ) = |{ a jk : i , j ∈ N i , a i j ∈ A }| K i ( K i − , (2)where N i is the set of neighboring economies of economy i , K i = K in i + K out i . The node’s influence is measured according to the network’s global information. The most used methods includethe shortest path (betweenness and closeness) and random walk (such as PageRank).
Betweenness B t ( i ) refers to the ratio of the number of shortest paths passing through a specific economy i to thetotal number of shortest paths between any two economies [40]: B t ( i ) = X st n ist g st , (3)where g st is the total number of shortest paths from node s to node t and n ist is the number of those paths pass through.4 loseness C out considers the average length of the shortest path from each economy to other economies [41]. C out ( i ) = (cid:18) D i N − (cid:19) C i , (4)where D i is the number of economies that can be reached from economy i (excluding i ), N is the number of economiesin the iOTN. C i is the sum of the distances from economy i to all reachable economies, and for isolated nodes, C out ( i ) =
0. Considering that the iOTN is a directed network, Eq. (4) calculates outcloseness. Taking the tradedirections into consideration, incloseness can also be calculated based on the Eq. (4). Incloseness is the reciprocal ofthe sum of the distances from all economies to the economy i . PageRank is an indicator for measuring the influence of economies in the directed network [42]. When calculatingPageRank, the number of iterations is set to 100.
Authorities and Hubs are two indicators involved in the HITS algorithm, which were first proposed by Kleinberg,and are inseparable in measuring the centrality [43]. An economy with a high authority score is defined as the economypointed by many economies with high hub centrality. Correspondingly, an economy with a high hubs score is definedas it points to many economies with high authority scores. The HITS algorithm is exquisite and can provide moreinformation for node centrality in theory, so the index is relatively complicated. In this article, the hub or authorityscores are converted according to the proportion of the economy in all economies, and the sum of the scores (hub orauthority) is equal to 1.
Economies in the iOTN with similar functions have similar topological properties [44]. Therefore, there is often amodular structure. The connections between members within each module are relatively close, while the connectionsbetween the modules are relatively sparse [5, 45]. The economic influence index based on the module structureconsiders the individual characteristics of the economy and mines the group information based on the network module.We apply the module division method based on network structure to identify the critical position of each economyin the iOTN [46]. Firstly, the network needs to be divided into modules. Two indicators of within-module degree andparticipation coefficient can be proposed to measure the influence of economies on the internal and external moduleswhere they are located. We use the classic modular algorithm to divide the iOTN into modules [47–49].
Within-module degree Z i measures how well-connected economy i is to other nodes in its module [46]. Z i = k i , s − k i , s σ s , (5)where k i , s is the number of trade relationships between economy i and other economies in module s . k i , s is the averagevalue of k i , s for economies in the module s . σ s is the standard deviation of k i , s in s . If the economy i has trade relationswith many economies in its module s , the value of Z i will be larger, implying that the trade changes of the economy i can affect other economies in the module to a greater extent. Participation coefficient P measures the uniformity of connections between the economy i and the economies indifferent modules [46]. P i = − N M X s = k i , s K i ! , (6)where k i , s is the number of relationships of economy i to economies in module s , and K i is the total degree of economy i . N M is the total number of modules. The P of an economy is close to 1 if its trade partners are uniformly distributedamong all the modules. Furthermore, if its partners are within its module, P = P cannot measure the strength of the economy’s influence in oil trade outside its module. Therefore, we introducethe outside-module degree [50]. Outside-module degree of economy i measures how well-connected i is with ports outside its own module [50].It is defined as: B i = m i , s − m i , s σ s , (7)5 able 1: Top 10 economies identified based on different influence indicators in 2017 Rank Pagerank Outegree Indegree Outcloseness Incloseness Betweenness1 Netherlands USA Netherlands USA Netherlands Netherlands2 India Russian Federation USA Russian Federation USA USA3 Spain United Kingdom India United Kingdom India United Kingdom4 USA Nigeria China Nigeria Spain United Arab Emirates5 France Saudi Arabia Spain Saudi Arabia France India6 Italy Kazakhstan Rep. of Korea Algeria Italy China7 Singapore Iraq Singapore United Arab Emirates Canada Germany8 Rep. of Korea United Arab Emirates France Iraq Germany France9 Canada Algeria Italy Norway Sweden South Africa10 Germany Germany Germany Azerbaijan United Arab Emirates AzerbaijanRank Authorities Hubs Clustering Within-module Outside-module Participation1 Netherlands USA Saudi Arabia Netherlands Netherlands Netherlands2 China Russian Federation Norway USA USA Israel3 USA Saudi Arabia Venezuela China China USA4 Spain Nigeria Angola Russian Federation Russian Federation Finland5 India United Kingdom Papua New Guinea South Africa India El Salvador6 France Iraq Kuwait India South Africa Mozambique7 Italy Algeria Iraq United Arab Emirates Italy Zambia8 Rep. of Korea Kazakhstan Brazil France United Kingdom Italy9 Singapore Libya Chad Spain Rep. of Korea Sweden10 Germany Norway Canada Rep. of Korea Singapore Libya where m i , s is the number of connections of economy i to other economies out of its own module s . m i , s and σ s arerespectively the average and standard deviation of m i , s . To measure the function and structural integrity of the iOTN after an attack, we use the fraction of economies inthe giant connected component (GCC) after removing a ratio of q economies as the measure indicator [51], whichcan be defined as S ( q ). Considering the various situations where the network suffers enormous damage but does notcompletely collapse, we introduce robustness measure R to measure the function and structural integrity of the iOTNunder different attack strategies [51]: R = N N X n = S (cid:18) nN (cid:19) , (8)where N is the total number of economies in the iOTN, and n = qN . The normalization factor N ensures that therobustness of networks with different sizes is comparable.
4. Empirical analysis
The indicators to measure the influence of nodes in the network are applied to the real trade data set [33, 34]. Thecritical nodes in the network may vary depending on the indicators and network types. However, there is relativelylittle literature on comparison and analysis of these indicators and methods themselves in the study of the trade risktransmission mechanism under the framework of energy security,We calculate the oil trade influence of economies in different years based on the economic influence measuresproposed in Section 3. The trade influence of economies is continually changing over time. Taking the iOTN in 2017as an example, we select the top 10 economies with the most considerable trade influence under different measures asshown in Table 1.In Table 1, the ranking results under different measures are indeed different. The USA and the Netherlands bothhave a strong influence on the network’s local, global, and community structure. In 2017, the Netherlands surpassedthe USA in most centrality measures and was considered as the most critical economy. The surpass may be relatedto the expansion of both port oil trade in the Netherlands and the American energy transformation in recent years.Due to complex geopolitical relations and continuous regional conflicts, the dependence of the USA on traditional6 able 2: Correlation of the influence indicators of oil trade economies in 2017.
Pagerank Outdegree Indegree Outcloseness Incloseness AuthoritiesPagerank 1.000Outdegree 0.274 ∗∗∗ ∗∗∗ ∗∗∗ ∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ Betweenness 0.919 ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Clustering 0.006 0.617 ∗∗∗ ∗∗∗ ∗ Within-module 0.671 ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Outside-module 0.650 ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Participation 0.252 ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Hubs Betweenness Clustering Within-module Outside-module ParticipationHubs 1.000Betweenness 0.315 ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗ p < . , ∗∗ p < . , ∗∗∗ p < . . oil-producing regions in the Middle East and Africa has been significantly reduced. The USA has gradually movedfrom the largest oil importer to energy self-sufficiency [9]. Emerging economies such as China and India have alsodemonstrated their importance in the iOTN. Although different influence indicators measure the influence of economies in the iOTN from different aspects, thesettings of the centrality indicators are all based on the topological structure of the iOTN, and there may be correlationbetween those indicators. We calculate the correlation between the 12 indicators and the results are shown in Table 2.Table 2 shows the Spearman correlation coefficients between the influence measures. When measuring the tradeinfluence of economies, most of the indicators have a significant correlation. The outdegree and hubs have a significantand highly positive correlation, and the correlation coefficient is 0.959. The correlation coefficients of indegree andPageRank, and of indegree and authority are 0.930 and 0.929 respectively.Although there are close relationships between the economic influence measures, their calculation basis is differentand cannot replace each other. The optimal influence measure needs to be selected according to specific issues. Thisstudy aims to find the economies that have a more significant impact on the stability of the oil trade network structure.Therefore, the horizontal comparison of the influence measures is of great significance for selecting the key economiesthat deserves more attention and maintaining the structural stability of the iOTN.
In the global energy governance system, organizations such as the Organization of Petroleum Exporting Coun-tries (OPEC) and the Organization for Economic Cooperation and Development (OECD) are important internationaleconomic institutions, performing an important role in global economic governance and global economic order co-ordination. In the 21st century, emerging economies have grown rapidly, and the global economic landscape hascontinuously changed. Exploring the changes of energy trade influence of different international economic organiza-tions can deepen the understanding of the changes in the energy market patterns and the international situation.We take several typical international organizations as the research objects, and take the average influence value ofthe economies included in these organizations as their oil trade influence. Fig. 2 shows the trade influence evolutionof the OPEC, OECD, European Union (EU), G7, and G20 based on different influence indicators.It can be seen from Fig. 2 that the oil trade influence of international organizations has a relatively apparent trendunder most measures. The influence measured by the network’s global structure, such as PageRank, authority, and7
988 1992 1996 2000 2004 2008 2012 201600.010.020.030.04 -3 -3 Figure 2: The evolution of the oil trade influence of international organizations from 1988 to 2017. hubs, has a clear downward trend. The influence based on the local structure, such as degree and local clusteringcoefficient, shows an upward trend. The influence based on the modular structure was an upward trend in the earlystage and then gradually stabilized and fluctuated slightly. The influence with the participation coefficient as anindicator fluctuates significantly with the year.It can be concluded that as more and more economies join the process of trade globalization, international organi-zations have developed closer trade ties with neighboring economies, and their influence on a local scale is increasing.However, as far as the global situation is concerned, the international status, trade policies and oil trade patterns ofeconomies are constantly changing. The impact of a single international organization on global oil trade is declining.The nature of international organizations and the characteristics of their membership determine the trend of oiltrade influence. After the first oil crisis hit Western economies, the G7 was born to respond to the global economic8
Figure 3: The structural integrity of the iOTN under attacks of economic nodes. crisis and to stabilize the international economic system. The outbreak of the Asian financial crisis in 1997 highlightedthe importance of developing countries. With the participation of major developing countries, the G20 came intobeing. Therefore, the emergence of the G7 and the G20 are closely related to the economic crisis. Similarly, theEU and the OECD aim to promote international cooperation and joint economic development, and they have manycommon members. The trade influence measured by the more relevant centrality indicators has the same trend. Thesefacts explain the similar trends in the oil trade influence of different organizations and indicators.OPEC has certain peculiarities among several organizations, being the earliest established and the most influentialproducer and exporter of raw materials. Its purpose is to coordinate and unify oil policies, maintain the stability ofinternational oil market prices, and safeguard oil-producing countries’ interests. Therefore, its impact on oil trade is9ot limited to its trade, but it also balances world power by affecting oil prices. With the strengthening of internationalcooperation, the trade cooperation of various economies is more extensive.However, based on the oil trade network’stopological structure, OPEC’s trade influence has not been significantly improved from a multidimensional perspec-tive. Its global trade influence based on trade export capacity as measured by the hubs indicator is declining. Thisresult is also consistent with OPEC’s declining impact on global oil production mentioned in previous studies [52].
The oil trade of the economies is continually diversifying and the structure of the iOTN is becoming increasinglycomplex. Trade shocks and risks caused by unexpected events such as oil supply interruptions, wars, economicsanctions may pass along intricate trade links. From the perspective of network structure, when the iOTN suffersfrom trade shocks (random attacks or targeted attacks), it needs to be robust enough to maintain trade stability andsustainable global economic development. To measure the robustness of the iOTN under economic attacks, which canalso be called the anti-risk ability of the network structure, we use the influence measures of economies as strategicindicators of node attacks. The changes in the integrity of the network structure after the network is attacked areshown in Fig. 3.In Fig. 3, we show the results of simulated attacks in 2003, 2008, 2014 and 2017. For comparison, the black dashedlines represent the results of randomly removing economies with a ratio of q . Under the random attack, the GCC sizedecreases linearly as the attack intensity increases. Under the targeted attacks, the iOTN falls apart at a faster rate.Different targeted attacks have roughly the same influence on the network structure, but the destruction speed of theiOTN is different. In general, when the proportion of attacks on the economies reaches 40%-60%, the network will beclose to collapse. The global oil trade network is a scale-free network, and a small number of economies have a largenumber of trade relationships [19]. When the economies with more trade relations and trade influence are attacked,the structure of the iOTN will be more severely damaged.It can be seen from the results of any simulated attacks that the iOTN has become more vulnerable to attacks overtime. In other words, the iOTN becomes more fragile. Although the results in different years under the same attackstrategy are not much different, the number of economies in the GCC decreases at a faster rate. The structure andfunction of the iOTN are losing faster.Figure 3 shows the vulnerability of the iOTN under specific attack strategies. It is necessary to calculate the entirenetwork’s robustness to compare different attack strategies and find the critical economies that significantly impactnetwork stability. The robustness considers the destruction of the network under under all attack situations. Accordingto Eq. (8), we calculate the robustness of the iOTN in different years under various strategic attacks. To make the curvesmoother, we use the cubic spline interpolation method to process the data. Its evolution over time is shown in Fig. 4.It can be seen from Fig. 4 that under the random and local clustering coefficient-based attack strategies, thestability of the iOTN is the strongest, basically maintained at 0.4-0.5. In most targeted attacks, the anti-risk abilityof the iOTN is low, and the network robustness is basically below 0.3. In most years, the attack strategy based onthe within-module degree indicator results in the lowest robustness of the network, and it is the most effective attackstrategy.The within-module degree measures an economy’s ability to establish connections with other economies withinits module. The economy with high within-module degree plays a vital role in the aggregation of oil trade and is acritical connector. The failure of the connector will cause the collapse of regional trade cooperation, which will resultin the loss of the structure and function of the entire trade network. These results also indicate that the concentrationof regional oil trade may result in higher fragility of network structure. Strengthening the control of the connectors ineach module is of great significance to maintaining the stability of the entire oil trade.To observe the robustness rankings of the iOTNs under different strategic attacks, we arrange the robustness valuesover the past 30 years by year and draw the box plot, as shown in Fig. 4, to observe its distribution.From the 30-yeartime scale, it can be clearly seen that when the oil trading system is attacked based on the within-module degree,the network robustness index is distributed evenly and basically stable at the level of 0.2. The second thing needs toobserve is the authority strategy attack based on the global structure of iOTN.The structure of the iOTN is complex, and its formation is affected by many factors and cannot be designedaccording to subjective will. Therefore, protecting economies that are critical to system stability is an importantway to maintain oil trade stability. The numerical simulation experiments are of great significance for maintaining10
988 1992 1996 2000 2004 2008 2012 20160.050.10.150.20.250.30.350.40.450.5
Figure 4: The robustness evolution (Upper) and box plot (Lower) of the iOTN from 1988 to 2017. oil security and avoiding oil trade crises and global economic crises caused by systemic risks. For global policyresearchers, finding the right indicators is a key factor, and the differences between indicators make it difficult to makeoptimal decisions in the decision-making process. Through horizontal comparison, we can find the indicator with themost significant impact on network connectivity. Of course, there are also many indicators of network stability, whichcan be measured not only by the proportion of the GCC, but also by other network functional indicators. We can alsouse the same method in this article to perform numerical simulations, determine the most influential attack indicators,and make corresponding decision support.
5. Discussion and application
We incorporated the topology indicators of the iOTN based on complex network theory into the traditional oilsecurity and trade risk assessment framework, exploring the impact on the structural stability of the iOTN wheneconomies with different important positions are attacked. Through the horizontal comparison of traditional centralitymeasures, the basis for maintaining network robustness and trade stability can be found.By analyzing the different centrality positions of economies and international organizations in the iOTNs, wefind that most influence measures have significant correlations, but the results of identifying the trade influence ofeconomies by different node centrality measures are actually quite different. Optimal node impact measures also needto be selected based on the specific problem to be solved. At the individual level, the position of the economies inthe global oil trade network is constantly changing with the network structure. In 2017, the Netherlands surpassed11he USA to occupy a more important position and have higher structural importance. At the organizational perspec-tive, international organizations such as the OPEC and the OECD have increased their trade relationships with othereconomies in the iOTN. However, their overall influence has shown a downward trend in recent years.The simulated attacks results show that the international oil trade system has become more vulnerable in recentyears. The system has lower network stability when it is attacked by an economy based on the within-module de-gree. The agglomeration of regional oil trade is a major reason for the vulnerability of the iOTN. The oil tradingsystem is also fraught with uncertainty, with trade protection, national strategies, economic sanctions and even war allpotentially putting the trading network at risk. Maintaining the trade system’s stability requires a focus on the key con-nectors in the modular structure of the iOTN. The impact of the economies with the most significant influence withinthe module will cause the most incredible damage to the trade system’s structure. Therefore, more consideration needbe given to the trade gathering areas in the iOTN in the future, focusing on the economies with high within-moduledegree such as the Netherlands, the USA, and China.The stability and systematic risk in the iOTN can be explored and identified based on the change of networkstructure [11, 29]. However, to reduce the possibility of the outbreak of systemic risks in the global oil trade network,it is necessary to prevent the spread of trade shocks and improve the stability of the iOTN. In recent years, theinternational oil trade center has moved to the Asia-Pacific region, and European economies seek to reduce theirdependence on Russian oil imports. Brexit (Britain exiting from the EU) and the American shale oil revolutionhave significantly impacted the oil trade. These facts fully reflect that oil trade status and policy changes in criticaleconomies can cause instability and even endanger energy security. Therefore, identifying and protecting the economywith a critical position is of great significance for stabilizing the oil trade and prevent trade risks.We provide a framework for selecting critical trade economies and maintaining the stability of the global oil tradesystem. However, the paper still has certain limitations. Firstly, the risks faced by the oil trade system come from manyaspects, but our research only focuses on the situations in which the network structure can resist unexpected eventssuch as trade interruptions. Secondly, there are many indicators for measuring network stability and robustness, but theindicators we choose only consider the integrity of the network structure. There has not been a more comprehensivemeasure of the impact on the actual function of the network. In the future, more realistic factors can be used to expandthe assessment framework for the stability of the oil trading system and introduce more mechanisms for economictrade interactions so as to better inform global economic development and oil security in conjunction with networktopology.
Acknowledgment
This work was supported by the Shanghai Outstanding Academic Leaders Plan and the Fundamental ResearchFunds for the Central Universities.
Data Availability
Oil data sets related to this article can be found at https://comtrade.un.org/, an open-source online data repository.
References [1] X. Xi, J. Zhou, X. Gao, D. Liu, H. Zheng, Q. Sun, Impact of changes in crude oil trade network patterns on national economy, Energy Econ.84 (2019) 104490. doi:10.1016/j.eneco.2019.104490 .[2] P. Caraiani, Oil shocks and production network structure: Evidence from the OECD, Energy Econ. 84 (2019) 104559. doi:10.1016/j.eneco.2019.104559 .[3] Q. R. Sun, H. Z. An, X. Y. Gao, S. Guo, Z. Wang, S. Y. Liu, S. B. Wen, Effects of crude oil shocks on the PPI system based on variancedecomposition network analysis, Energy 189 (2019) 116378. doi:10.1016/j.energy.2019.116378 .[4] W.-J. Xie, N. Wei, W.-X. Zhou, Evolving efficiency and robustness of global oil trade networks, Energy Policy (2020) submitted.[5] Z. J. Zhang, H. Q. Lan, W. L. Xing, Global trade pattern of crude oil and petroleum products: Analysis based on complex network, IOPConference 153 (2018) 022033. doi:10.1088/1755-1315/153/2/022033 .[6] H.-Y. Zhang, Q. Ji, Y. Fan, What drives the formation of global oil trade patterns?, Energy Econ. 49 (2015) 639–648. doi:10.1016/j.eneco.2015.02.017 .[7] T.-H. Le, Y. h. Chang, Oil price shocks and trade imbalances, Energy Econ. 36 (2013) 78–96. doi:10.1016/j.eneco.2012.12.002 .
8] S. Rafiq, P. Sgro, N. Apergis, Asymmetric oil shocks and external balances of major oil exporting and importing countries, Energy Econ. 56(2016) 42–50. doi:10.1016/j.eneco.2016.02.019 .[9] Q. An, L. Wang, D. Qu, H. Zhang, Dependency network of international oil trade before and after oil price drop, Energy 165 (2018) 1021–1033. doi:10.1016/j.energy.2018.09.098 .[10] R. J. Du, Y. Wang, G. G. Dong, L. X. Tian, Y. X. Liu, M. G. Wang, G. C. Fang, A complex network perspective on interrelations and evolutionfeatures of international oil trade, 2002-2013, Appl. Energy 196 (2017) 142–151. doi:10.1016/j.apenergy.2016.12.042 .[11] R. J. Du, G. G. Dong, Y. G. Wang, L. F. Zhao, X. Zhang, A. L. Vilela, H. E. Stanley, Identifying the peak point of systemic risk in internationalcrude oil importing trade, Energy 176 (2019) 281–291. doi:10.1016/j.energy.2019.03.127 .[12] D. H. Kim, D. A. Eisenberg, Y. H. Chun, J. Park, Network topology and resilience analysis of south korean power grid, Physica A 465 (2017)13–24. doi:10.1016/j.physa.2016.08.002 .[13] W.-X. Wang, G. Chen, Universal robustness characteristic of weighted networks against cascading failure, Phys. Rev. E 77 (2) (2008) 026101. doi:10.1103/PhysRevE.77.026101 .[14] S. P. A. Brown, H. G. Huntington, Assessing the US oil security premium, Energy Econ. 38 (2013) 118–127. doi:10.1016/j.eneco.2013.03.010 .[15] A. Cherp, J. Jewell, The concept of energy security: Beyond the four As, Energy Policy 75 (2014) 415–421. doi:10.1016/j.enpol.2014.09.005 .[16] B. W. Ang, W. L. Choong, T. S. Ng, Energy security: Definitions, dimensions and indexes, Renew. Sust. Energ. Rev. 42 (2015) 1077–1093. doi:10.1016/j.rser.2014.10.064 .[17] B. K. Sovacool, Evaluating energy security in the Asia pacific: towards a more comprehensive approach, Energy Policy 39 (11) (2011)7472–7479. doi:10.1016/j.enpol.2010.10.008 .[18] B. K. Sovacool, I. Mukherjee, Conceptualizing and measuring energy security: A synthesized approach, Energy 36 (8) (2011) 5343–5355. doi:10.1016/j.energy.2011.06.043 .[19] L. Liu, Z. Cao, X. Liu, L. Shi, S. Cheng, G. Liu, Oil security revisited: An assessment based on complex network analysis, Energy 194 (2020)116793. doi:10.1016/j.energy.2019.116793 .[20] Y. M. Zhou, J. W. Wang, G. Q. Hang, Efficiency and robustness of weighted air transport networks, Transp. Res. Pt. e-Logist. Transp. Rev.122 (2019) 14–26. doi:10.1016/j.tre.2018.11.008 .[21] M. A. Serrano, M. Bogu˜n´a, Topology of the world trade web, Phys. Rev. E 68 (2003) 015101(R). doi:10.1103/PhysRevE.68.015101 .[22] G. Fagiolo, J. Reyes, S. Schiavo, World-trade web: Topological properties, dynamics, and evolution, Phys. Rev. E 79 (2009) 036115. doi:10.1103/PhysRevE.79.036115 .[23] H. Z. An, W. Q. Zhong, Y. R. Chen, H. J. Li, X. Y. Gao, Features and evolution of international crude oil trade relationships: A trading-basednetwork analysis, Energy 74 (2014) 254–259. doi:10.1016/j.energy.2014.06.095 .[24] Y. Yu, P. H. Jessie, B. S. Sharmistha, Small and flat worlds: A complex network analysis of international trade in crude oil, Energy 93 (2015)534–543. doi:10.1016/j.energy.2015.09.079 .[25] A. Kharrazi, B. D. Fath, Measuring global oil trade dependencies: An application of the point-wise mutual information method, EnergyPolicy 88 (2016) 271–277. doi:10.1016/j.enpol.2015.10.017 .[26] T. Kitamura, S. Managi, Driving force and resistance: Network feature in oil trade, Appl. Energy 2017 (208) (2017) 361–375. doi:10.1016/j.apenergy.2017.10.028 .[27] Q. Ji, H. Y. Zhang, Y. Fan, Identification of global oil trade patterns An empirical research based on complex network theory, EnergyConversion and Management 85 (2014) 856–865. doi:10.1016/j.enconman.2013.12.072 .[28] S. Battiston, J. D. Farmer, A. Flache, D. Garlaschelli, A. G. Haldane, H. Heesterbeek, C. Hommes, C. Jaeger, R. May, M. Scheffer, Complexitytheory and financial regulation, Science 351 (6275) (2016) 818–819. doi:10.1126/science.aad0299 .[29] W. Q. Zhong, H. Z. An, X. Y. Gao, X. Q. Sun, The evolution of communities in the international oil trade network, Physica A 413 (2014)42–52. doi:10.1016/j.physa.2014.06.055 .[30] Q. Sun, X. Gao, W. Zhong, N. Liu, The stability of the international oil trade network from short-term and long-term perspectives, Physica A482 (2017) 345–356. doi:10.1016/j.physa.2017.04.047 .[31] T.-S. Yun, D. Jeong, S. Park, “Too central to fail” systemic risk measure using pagerank algorithm, J. Econ. Behav. Organ. 162 (2019)251–272. doi:10.1016/j.jebo.2018.12.021 .[32] W. Zhong, H. An, L. Shen, W. Fang, X. Gao, D. Dong, The roles of countries in the international fossil fuel trade: An emergy and networkanalysis, Energy Policy 100 (2017) 365–376. doi:10.1016/j.enpol.2016.07.025 .[33] F. Dablander, M. Hinne, Node centrality measures are a poor substitute for causal inference, Sci. Rep. 9 (6846) (2019) 6846. doi:10.1038/s41598-019-43033-9 .[34] R. J. Richmond, Trade network centrality and currency risk premia, J. Financ. 74 (3) (2019) 1315–1361. doi:10.1111/jofi.12755 .[35] S. Battiston, M. Puliga, R. Kaushik, P. Tasca, G. Caldarelli, Debtrank: Too central to fail? Financial networks, the fed and systemic risk, Sci.Rep. 2 (2012) 541. doi:10.1038/srep00541 .[36] Y. Fan, S. Ren, H. Cai, X. Cui, The state’s role and position in international trade: a complex network perspective, Econ. Model. 39 (2014)71–81. doi:10.1016/j.econmod.2014.02.027 .[37] P. Crucitti, V. Latora, M. Marchiori, Model for cascading failures in complex networks, Phys. Rev. E 69 (4) (2004) 045104. doi:10.1103/PhysRevE.69.045104 .[38] D. Chen, L. Lu, M.-S. Shang, Y.-C. Zhang, T. Zhou, Identifying influential nodes in complex networks, Physica A 391 (4) (2012) 1777–1787. doi:10.1016/j.physa.2011.09.017 .[39] D. J. Watts, S. H. Strogatz, Collective dynamics of ‘small-world’ networks, Nature 393 (1998) 440–442. doi:10.1038/30918 .[40] L. C. Freeman, A set of measures of centrality based on betweenness, Sociometry 40 (1) (1977) 35–41. doi:10.2307/3033543 .[41] P. Hage, F. Harary, Eccentricity and Centrality in Networks, Soc. Networks 17 (1995) 57–63. doi:10.1016/0378-8733(94)00248-9 .
42] L. Page, The pagerank citation ranking: Bringing order to the web, Stanford InfoLab (1999) 1–14 doi:10.1016/j.tre.2018.11.008 .[43] J. Kleinberg, Authoritative sources in a hyperlinked environment, J. ACM 46 (5) (1999) 604–632. doi:10.1145/324133.324140 .[44] R. Guimer`a, L. A. N. Amaral, Cartography of complex networks: modules and universal roles, J. Stat. Mech.-TheoryExp. doi:10.1088/1742-5468/2005/02/P02001 .[45] P. Giudici, B. H. Huang, A. Spelta, Trade networks and economic fluctuations in Asian countries, Econ. Syst. 43 (2) (2019) 100695. doi:10.1016/j.ecosys.2019.100695 .[46] R. Guimer`a, L. A. N. Amaral, Functional cartography of complex metabolic networks, Nature 433 (2005) 895–900.[47] M. E. J. Newman, M. Girvan, Finding and evaluating community structure in networks, Phys. Rev. E 69 (2004) 026113.[48] M. E. J. Newman, Modularity and community structure in networks, Proc. Natl. Acad. Sci. U.S.A. 103 (23) (2006) 8577–8582. doi:10.1073/pnas.0601602103 .[49] V. D. Blondel, J.-L. Guillaume, R. Lambiotte, E. Lefebvre, Fast unfolding of communities in large networks, J. Stat. Mech.-Theory Exp.2008 (10) (2008) P10008. doi:10.1088/1742-5468/2008/10/P10008 .[50] M. Xu, Q. Pan, A. Muscoloni, H. Xia, C. V. Cannistraci, Modular gateway-ness connectivity and structural core organization in maritimenetwork science, Nat. Commun. 11 (1). doi:10.1038/s41467-020-16619-5 .[51] C. M. Schneider, A. A. Moreira, J. S. Andrade, S. Havlin, H. J. Herrmann, Mitigation of malicious attacks on networks, Proc. Natl. Acad.Sci. U. S. A. 108 (10) (2011) 3838–3841. doi:10.1073/pnas.1009440108 .[52] S. Al Rousan, R. Sbia, B. K. O. Tas, A dynamic network analysis of the world oil market: Analysis of OPEC and non-OPEC members,Energy Econ. 75 (2018) 28–41. doi:10.1016/j.eneco.2018.07.032 ..