Research Article: Propagation of negative shocks across nation-wide firm networks

Date Published: March 14, 2019

Publisher: Public Library of Science

Author(s): Hiroyasu Inoue, Yasuyuki Todo, Kazutoshi Sasahara.


This study examines how negative shocks due to, for example, natural disasters propagate through supply chains. We apply a simulation technique to actual supply chain data covering most Japanese firms. To investigate the property of the propagation in the network, we test different types of artificial negative shocks. We find that, first, network structures severely affect the speed of propagation in the short run, and the total loss in the long run. The scale-free nature of the actual supply-chain network—that is, the power-law degree distribution—leads to faster propagation. Second, more intensive damages—that is, more damages suffered by fewer firms—result in faster propagation than extensive damages of the same total size. Third, the actual supply-chain network has innate robustness that comes from substitutability of supplies. If the supply-chain network has severe substitutability, the propagation of negative shocks becomes substantially large. Fourth, direct damages in urban regions promote faster propagation than those in rural regions. Fifth, different sectoral damages show significant differences in the speed of propagation. Finally, we check the indirect damage triggered by a single firm’s loss: 9.7% of all firms contribute to significant loss, and this loss accounts for more than 10% of the damage to the entire production. The simulations conspicuously show that different direct damages, even if they have the same total magnitude of damages, can generate considerably different damages because of the structure of the supply-chain network.

Partial Text

Natural disasters trigger economic damages directly and indirectly because negative shocks are largely propagated through supply chains [1]. These indirect damages are far from negligible; they often constitute a large share of total damages [2, 3]. For example, let us consider the aftermath of the Great East Japan Earthquake in 2011. Although many firms, including those in foreign countries, were directly unaffected by the earthquake, they were compelled to cease operations due to supply shortages.

We use two databases collected in 2011 by Tokyo Shoko Research (TSR) (one of the two major corporate research companies in Japan)—the TSR Company Information Database and the TSR Company Linkage Database. The databases are commercially available. In our case, we have access to the databases that are licensed to the Research Institute of Economy, Trade and Industry (RIETI). The TSR data contain a wide range of firm information, including identification numbers of suppliers and clients of each firm. Although the maximum number of suppliers and clients reported by each firm is 24, we can capture more than 24 suppliers and clients by looking at the supplier-client relationships in the reverse direction. In other words, although a large firm such as Toyota reports only 24 suppliers, its suppliers are most likely to report Toyota as their client. Accordingly, we identify the supply chain network of firms in Japan to a great extent. The number of firms or nodes is 1,109,549, whereas the number of supplier-client ties or links is 5,106,081. This network is directed as it represents the flow of intermediate and final products.

We utilize the theoretical model proposed by [10] with some modifications. As we will indicate later, the major difference is in the rationing mechanism. Moreover, the target inventory size has a Poisson distribution, instead of a common constant. This is an agent-based model wherein agents, that is, firms and final consumers, follow specific rules. In the model, each firm in a sector produces a sector-specific product using a variety of intermediates, and delivers the product to its clients and final consumers. Further, we assume that firms have inventories of intermediates for dealing with a possible supply shortage.

We used Japanese nation-wide supply-chain network data and employed a modified version of Hallegatte’s model [10] to examine how negative shocks by artificial disasters propagate through supply chains. We obtained the following results. First, network structures severely affect the speed of propagation in the short run and the total loss in the long run. The scale-free nature of the actual supply-chain network, that is, the power-law degree distribution, leads to faster propagation than the random network. Second, a small number of firms with intense damages cause faster and larger propagation. Third, substitution among suppliers largely contributes to economic resistance. The pace of the propagation of negative shocks increases with an increase in substitution difficulties. Fourth, direct damages in industrial regions promote faster propagation than those in less industrial regions, although the total loss in value added in the long run is the same. Fifth, different sectoral damages cause large differences in the speed of propagation. Particularly, the effects of direct damages on the construction sector are quite small. Finally, an estimation of the indirect damage triggered by a single-firm loss shows that 86.6% of firms cause less than 10−5 of the damage to the entire economy. On the other hand, 9.7% of firms cause more than 10% of the damage to the entire supply chain. Thus, the actual supply chain has strong robustness against random failures, but it is vulnerable to selective attacks.




Leave a Reply

Your email address will not be published.