Propagation of economic shocks through supply chains
When an economic shock reduces production in a particular region or industry, the suppliers of the firms directly affected by the shock must reduce their production because of a lack of demand, and their customers must also shrink production because of a shortage of materials, parts, or components. As a result of this propagation through supply chains, a region- or industry-specific shock can have a substantial indirect effect on the entire economy, often more substantial than the direct effect of the shock itself.
The propagation of shocks resulting from natural disasters through supply chains has empirically been shown in several recent studies, including Barrot and Sauvabnat (2016) for the US, Calvarlho et al. (2016) for Japan, and Kashiwagi et al. (2018) across multiple countries. These studies also reveal that the propagation of shocks is more prominent when inputs are more difficult to substitute because of their specificity.
Another strand of the literature simulates agent-based models to examine how the structure of supply chains affects the propagation of shocks, comparing results from different hypothetical supply chains and shocks (e.g. Henriet et al. 2012). In an earlier paper, we use the actual supply chains of nearly one million firms in Japan and show that propagation is quicker and results in a larger production loss when there are ‘hub’ firms with an extremely large number of supplier and customer firms (Inoue and Todo 2019a).
Our recent paper revisits this issue, and simulates the production dynamics of Japan after the Great East Japan Earthquake in 2011, the fourth largest earthquake in the world since 1900 (Inoue and Todo 2019b). We estimate the parameter values in the model such that the average of simulated production from 30 cases, in which the initial conditions are randomly determined, can closely replicate the actual production dynamics after the earthquake (Figure 1). Since the simulations have a heavy computational load, we use the K computer, one of the fastest supercomputers.
Figure 1 Simulation results
Source: Figure 1 in Inoue and Todo (2019b).
Using these parameter values, our simulation shows that the economic shock quickly propagated through supply chains to most regions of Japan, and persisted there. Figure 2 illustrates the firms whose estimated production declined by 80% or more as red dots, while firms with a smaller production loss are depicted in a lighter colour. The figure demonstrates that 20 days after the Great East Japan Earthquake, the firms in major industrial regions, such as Tokyo, Nagoya, and Osaka, substantially reduced their production. This is because many firms in these regions were directly linked with firms in the disaster areas and were thus immediately affected by propagation through supply chains. After 40 days, wider regions were affected. We argue that this is because firms in these regions might have been indirectly linked with firms in the major industrial regions, even if they had no direct links with firms in the disaster areas. Even 60 days after the earthquake, a large number of firms across Japan still suffered from production shortages. A video of the simulation is available on YouTube.
Figure 2 Geographic propagation of the shock from the Great East Japan earthquake
Source: Figure 2 in Inoue and Todo (2019b)
We find that the production loss because of the indirect effect through supply chains reached 2.3% of GDP, while the loss because of the direct effect on firm production facilities was only 0.02%. In other words, the indirect effect of the earthquake on the entire economy was 100 times larger than its direct effect.
We also run the same model to forecast the effect of a predicted mega earthquake of magnitude nine, the Nankai Trough earthquake, which is expected to hit the major industrial regions of Japan with a probability of more than 70% within the next 30 years. Its direct and indirect effects are estimated to be 0.47% and 10.6% of GDP, respectively, confirming a far larger size of the indirect effect of disasters than the direct effect. A video of the simulation is available on YouTube.
We further use hypothetical supply chains to examine network characteristics that affect the propagation of shocks. In addition to the role of hub firms and input substitutability already found in the literature, our simulation analysis highlights the effect of ‘cycles’ or ‘loops’ in supply chains that are formed when the suppliers of primary and intermediate goods (for example, materials for printed wiring boards) utilise final goods (computers) for their production. In the actual supply chains, we find a large number of complex cycles. To examine the role of cycles for propagation, Figure 3 illustrates the production dynamics after the Great East Japan Earthquake, using the actual supply chain in panel d and five hypothetical supply chains with various levels of cycles and input substitutability. The different lines in each panel show the simulation results from 30 different sets of initially damaged firms (the total number of directly damaged firms is known and set as constant).
Figure 3 Dynamics of production using different supply chains
Source: Supplementary Figure 6 in Inoue and Todo (2019b).
These simulations show that complex cycles, particularly when combined with low input substitutability (panels d, f and h of Figure 3), make the propagation effect more substantial and persistent. Moreover, in many simulations using different initial conditions, even after the production starts to recover approximately two months after the earthquake, it drops again and often remains low. The persistency of low production and this second wave of production decline emerge possibly because economic shocks circulate and are amplified in cycles in supply chains.
In our previous Vox column, we suggested two policy implications from our earlier results to lessen the propagation effect: promoting input substitutability and supporting firms’ recovery immediately after a shock. Our new results indicate that shocks of the same size to different supply chains could lead to very different outcomes stochastically. Therefore, we should consider the possibility of devastating scenarios that are associated with second waves and a long duration of production loss. These may be far worse than expected based on past experience and academic studies.
In addition, we want to emphasise the importance of using data from actual supply chains in policy analysis. As indicated by our simulations, any hypothetical network structure leads to outcomes that are completely different from those using the actual networks in place. Therefore, the bottlenecks of supply chains can only be discussed when analysing the actual complex network, including the scale-free property and numerous loops.
Finally, although our analysis primarily uses natural disasters as a source of negative economic shocks, the results can be applied to other sources of supply-chain disruptions, including trade wars. Many scholars, including Blanchard (2019), argue that the effect of the current trade wars could be amplified by global value chains and hence that they could be quite large. Our results suggest that the total effect of trade wars on the global economy would be even larger than currently argued because of the complex nature of global value chains. However, we should also note that the effect can be examined using the actual data for global value chains.
Barrot, J and J Sauvagnat (2016), “Input specificity and the propagation of idiosyncratic shocks in production networks”, The Quarterly Journal of Economics 131(3): 1543-1592.
Blanchard, E (2019), “Trade wars in the global value chain era“, VoxEU.org, 20 June.
Boehm, C E, A Flaaen and N Pandalai-Nayar (2019), “Input linkages and the transmission of shocks: Firm-level evidence from the 2011 Tōhoku earthquake”, Review of Economics and Statistics 101(1): 60-75.
Carvalho, V M, M Nirei, Y U Saito, and A Tahbaz-Salehi (2016), “Supply chain Disruptions: evidence from the Great East Japan Earthquake”, Columbia Business School Research Paper, 17-5.
Henriet, F, S Hallegatte and L Tabourier (2012), “Firm-network characteristics and economic robustness to natural disasters”, Journal of Economic Dynamics and Control 36(1): 150-167.
Inoue, H and Y Todo (2019a), “Propagation of negative shocks across nation-wide firm networks”, PLOS ONE 14(3): e0213648.
Inoue, H and Y Todo (2019b), “Firm-level propagation of shocks through supply-chain networks”, forthcoming in Nature Sustainability.
Kashiwagi, Y, Y Todo and P Matous (2018), “International propagation of economic shocks through global supply chains”, WINPEC Working Paper, No. E1810, Waseda Institute of Political Economy, Waseda University.