Disruption cascades in supply chain networks

Mohan, Shyam (2018) Disruption cascades in supply chain networks. Doctoral thesis, University of London: London Business School.

Abstract

Supply chain disruptions result in substantial losses that can arise not only due to direct shocks at firms (eg., due to natural disasters or factory fires), but also from the interruption of normal operations at the suppliers' or buyers' ends; that is, due to disruptions cascading from supply chain partners in the adjacent tiers and beyond. The extent of damages to a firm due to such disruption cascades is intricately related to the overall network structure comprising all the firms. Besides, firms in the supply chain network can make ex-ante investments in strategies that curtail such losses. In this thesis, I develop theoretical models to understand the relationship between disruption cascades, firm strategies to mitigate risk, and the topology of the network. In the first essay, given an exogenous specification of the network structure and firm strategies, I undertake an epidemiological characterization of disruption cascades, wherein I formalize the notion of supply chain resilience, as a measure of how quickly firms in the supply chain recover from a disruption. I also come up with a comparative metric termed relative vulnerability, to identify firms which are expected to suffer greater downtimes and consequently more losses due to disruptions. In the second essay, I build on the above framework by endogenizing firm strategies. To this end, I adopt a decentralized game-theoretic approach to characterize equilibrium investments of firms in an assembly network facing "large-scale" disruptions. I find that when direct shocks to individual firms are disjoint events, the optimal investment depends only on the attributes of the firm?s extended local neighborhood; that is, up to its tier 2 suppliers, thus making knowledge of the remaining network structure redundant. The findings help explain why "local" network information may often suffice for making equilibrium investments. In the third essay, I extend the above framework by characterizing the optimal investments in the above setting from a central planner's perspective. I use this characterization to relate inefficiency (gap between centralized and decentralized payoffs) to the network structure, and identify class of networks that stand to benefit the most by the intervention of a central planner. I find that networks whose nodes have a 'balance' between the in- and out-degrees are less inefficient than others. I also undertake a numerical investigation to study how the concurrence of direct shocks to firms can change these results, and observe that such concurrence increases the benefit from the potential intervention of a central planner.