Modeling the Dynamics of Distribution Networks: A Data-Driven Approach to Supply Chains

Abstract

This thesis examines the formation, growth, and resilience of large-scale distribution systems. We investigate the interactions among manufacturers, distributors, and consumers, and show how these interactions shape the growth and resilience of these systems. Our study begins with an empirical analysis, where we reconstruct the complete distribution networks of opioids in the United States using data from nearly half a billion shipping records. We then examine the main topological properties of these networks and analyze their stability over a nine-year period. Surprisingly, we find that despite the increasing demand for opioids, the main topological properties of the distribution networks remain stable. To investigate how distribution systems form and evolve, we develop an evolutionary network growth model that simulates strategic link formation between firms. Testing the model against the empirical data, we show that two mechanisms are essential for the emergence of the observed networks: centralization and multi-sourcing. While centralization enhances efficiency, multi-sourcing fosters local resilience to shocks. Next, we discuss firm growth dynamics and examine how previous economic theories can be applied to the supply chain domain. Finally, we analyze system resilience to possible disruptions. We model the propagation of supply shocks at the firm-level and discuss various system responses to mitigate them. Our focus is on the role of supply substitution as a quick strategy that we show can effectively reduce the shock impact. Our research offers a valuable tool for managers and policymakers, enabling them to devise effective mitigation strategies that can be implemented after disruptions occur. Through a rigorous approach that combines both empirical analysis and data-driven modeling, we are able to unveil the underlying mechanisms that govern these systems. Our results contribute to both network science and supply chain management. In our attempt to bridge the gap between the two fields, we provide new methodologies based on high-resolution data to study the dynamics of large-scale distribution networks.