Natural disasters trigger economic damages directly as well as indirectly, as negative shocks propagate through production networks or supply chains (Sheffi and Rice 2005). Such indirect damages are far from negligible and often constitute a large share of the total damages (Pelling et al. 2002, Tierney 1997). For example, after the Great East Japan Earthquake in March 2011, a large number of firms that were not directly affected by the earthquake – including those in other countries such as the US, South Korea, Thailand, and France – had to cease operations due to shortages in supply and demand (Nikkei Shimbun 2011). Therefore, examining the total loss due to such indirect effects and the speed of their propagation is an important agenda for policymakers to deal with regarding risks of natural disasters.
Although many existing studies have tried to answer these questions, there are several drawbacks. Most of them rely on input-output (IO) tables that represent sectoral interactions of input and outputs (Okuyama et al. 2004, Rose and Liao 2005) so that the mechanism of supply chain disruption at the firm level may not be adequately captured. Moreover, these studies cannot show how quickly negative shocks propagate through supply chains. To overcome the shortcoming of the sector-level analysis, Henriet and Hallegatte (2008) and Henriet et al. (2012) employ firm-level analysis, simulating an agent-based model in which firms are connected through supply chains. However, due to a lack of actual data on supply chain relations among firms, they rely on hypothetical random networks. The use of hypothetical networks (rather than actual ones) in simulation analysis is clearly a drawback, because the literature in network science shows that a structural difference in the network can lead to a substantial difference in the behaviours of agents in the network (Barabási 2016, Newman 2010).
Therefore, in a recent paper we use the actual supply chain data for more than one million firms in Japan collected by Tokyo Shoko Research (TSR) and simulate an extended model of Henriet and Hallegatte (2008) and Henriet et al. (2012) (Inoue and Todo 2017). In particular, we examine how a negative shock from a hypothetical earthquake that directly destroys 0.5% of production sites in the economy propagates through supply chains.
The benchmark simulation result using the actual supply chains is represented by the red line in Figure 1, showing that value added declines by 12% in 30 days after the shock and by 48% in 100 days. Note that the simulation does not assume any recovery from destruction due to the earthquake. This is obviously not the case in practice, however. Even after the Great East Japan Earthquake, the fourth largest earthquake in the past century, most plants that were directly hit restarted their operations within three months. The index of manufacturing production in the impacted region dropped to 65% of the pre-earthquake level immediately after the earthquake, but recovered to 85% three months afterward (Small and Medium Enterprise Agency of Japan 2012). Thus, we focus on short-run effects within 100 days in our simulation analysis, but we still conclude that indirect effects propagate so rapidly that even short-run effects are enormous.
Figure 1. Simulation results (1)
Next, we compare results from the actual network and a randomly generated network as used in the previous studies. The blue line in Figure 1 is given using a hypothetical network of firms that is randomly generated while the total number of supply chain ties is the same as in the actual network. It is clear that propagation of negative shocks is substantially slower in the random network than in the actual network. This is because supply chains in Japan are a scale-free network (Fujiwara and Aoyama 2010) in the sense that some ‘hub’ firms have an extremely large number of supply chain partners. Propagation through the actual network is fast because firms are indirectly connected within a relatively small number of steps of supply chains through such hubs.
We run the same simulation using a hypothetical network in which firm links are randomly changed while the number of supply chain partners of each firm remains the same as in the actual network (i.e. this network is scale-free). The simulation result shown by the green line in Figure 1 is not very different from the red line in the actual network, suggesting that scale-freeness is the key to rapid propagation. However, the loss of value added in the red line is smaller in the short run (within a month or two) than that in the green line. The difference may come from another important network characteristic, namely, the level of clustering. Firms tend to form clusters or groups in a network, as supply chains in Japan are often characterised by dense long-term relations within firm groups, or keiretsu. In a clustered network, propagation of negative shocks is slow because shocks quickly propagate within the cluster but not outside it.
Finally, we examine the importance of production substitution in shock mitigation. In the simulations hitherto, we assume that supplies from firms affected by the shock can be substituted by those from other current suppliers that produce the same product. Information on each firm's products can be obtained in the data, using 190 product categories. Now, we assume two hypothetical networks where production substitution is more difficult than in the actual network. In one case, products of firms are randomly determined while preserving the distribution of products as in the actual network. Then, because firms tend to utilise a larger variety of inputs, production substitution is more difficult. In another case, the product of each firm is assumed to be completely differentiated from those of other firms, so that substitution among suppliers is impossible. Figure 2 shows the results using the benchmark case (the red line) and assuming random assignment of products (green) and completely differentiated products (blue). This figure indicates that in the latter two cases where substitution among suppliers is more difficult, propagation of negative shocks is substantially faster. Therefore, substitution of undamaged suppliers for damaged ones is shown to be an important channel of mitigation of shock propagation.
Figure 2. Simulation Results (2)
Our results suggest several important implications for policymakers and firm managers. First, because propagation of negative shocks through supply chains is quick, policy support such as financial support for repair and replacement of capital goods should be implemented immediately after disasters to minimise their indirect effects. Second, production substitution is the key to slowing the propagation. Therefore, firms should diversify suppliers, rather than relying on a single supplier for a particular input. Also, it is useful to prepare business continuity plans to determine what firms should do in the wake of disasters, including how they substitute for disrupted supplies (Cole et al. 2015).
Before the Great East Japan Earthquake, Japanese firms were not prepared for shocks from natural disasters. For example, only 9% of small and medium enterprises in the region affected by the earthquake were equipped with business continuity plans (Hamaguchi 2013). However, Japanese firms learned lessons from the earthquake and started to change in line with the implications of our study. Toyota Motor Corporation, for example, created the Reinforce Supply Chain Under Emergency (RESCUE) database for full information of direct and indirect suppliers and their products (Fujimoto et al. 2016). Using this database, Toyota can easily search for alternative suppliers when supplies of certain materials and parts are disrupted due to a disaster. In addition, Toyota and other automobile manufacturers have diversified their suppliers according to the TSR dataset, departing from their traditional keiretsu relations (Matous and Todo 2017). In addition, large manufacturers have encouraged their suppliers to prepare business continuity plans to minimise their risk of supply disruption. These efforts seem to have paid off as, in the wake of the Kumamoto earthquake in 2016, firms recovered relatively quickly thanks to the use of business continuity plans and production substitution, according to interviews with firm managers in the affected region by our research team.
As the global production network has been developed across countries,, including developed and less developed ones (Baldwin 2016), firms are more likely to be exposed to the risk of disruption of supply even when they are located in natural disaster-free areas. As noted earlier, the Great East Japan Earthquake affected suppliers and assemblers in the US and Europe. Therefore, firms in any country should be prepared for the risk of supply disruption due to natural disasters, as long as they are connected with the global supply chains.
Baldwin R (2016), The Great Convergence, Belknap Press: Boston.
Barabási, A-L (2016), Network Science, Cambridge University Press: Cambridge.
Cole, M A, R J R Elliott, T Okubo and E Strobl (2015), "The effectiveness of pre-disaster planning and post-disaster aid: Examining the impact on plants of the Great East Japan Earthquake", RIETI, Discussion Paper 15-E-097.
Fujimoto, T, Y Kato and S Iwao (2016), "Strengthening purchasing and supply chain management at Toyota: Minutes of the lecture of Hiroaki Koda from the purchasing group, purchasing planning & Tnga Promotion Div, of Toyota Motor Corporation", Manufacturing Management Research Center tUoT (Ed), MMRC Discussion Paper.
Fujiwara, Y, H Aoyama (2010), "Large-scale structure of a nation-wide production network", The European Physical Journal, B 77: 565-580.
Hamaguchi, N (2013), "Impact of the Great East Japan Earthquake on local industries: Synthesis and analysis from a survey", RIETI, Policy Discussion Paper 13-P-001.
Henriet, F and S Hallegatte (2008), "Assessing the consequences of natural disasters on production networks: A disaggregated approach", FEEM Working Paper.
Henriet, F, S Hallegatte and L Tabourier (2012) "Firm-network characteristics and economic robustness to natural disasters", Journal of Economic Dynamics and Control, 36: 150-167.
Inoue,, H and Y Todo (2017), "Propagation of negative shocks through firm networks: Evidence from simulation on comprehensive supply chain data", RIETI, Discusson Paper 17-E-044.
Matous, P and Y Todo (2017), "Analyzing the coevolution of interorganizational networks and organizational performance: Automakers’ production networks in Japan", Applied Network Science 2, 5.
Newman, M (2010), Networks: An Introduction, Oxford University Press: Oxford.
Niekei Shamun May 3,
Okuyama, Y, G J Hewings and M Sonis (2004), "Measuring economic impacts of disasters: Interregional input-output analysis using sequential interindustry model", Modeling Spatial and Economic Impacts of Disasters, Springer: 77-101.
Pelling, M, A Özerdem and S Barakat (2002), "The macro-economic impact of disasters", Progress in Development Studies 2: 283-305.
Rose, A and S-Y Liao (2005),, "Modeling regional economic resilience to disasters: A computable general equilibrium analysis of water service disruptions", Journal of Regional Science 45: 75-112.
Sheffi, Y and J B J Rice (2005), "A supply chain view of the resilient enterprise", MIT Sloan Management Review 47: 41.
Small and Medium Enterprise Agency of Japan (2012), "White paper on small and medium enterprises in Japan".
Tierney, K J (1997) "Business impacts of the northridge earthquake", Journal of Contingencies and Crisis Management 5: 87-97.