A Complexity Economics paper investigating supply chain vulnerability has been named a '2024 Editors' Pick' by journal Nature Reviews Physics.
The research' Timeliness criticality in complex systems' responds to the 2021 global supply chain crisis; which saw empty supermarket shelves, limited supplies of essential medicines and unexpected issues like shortages of carbon dioxide used in soft drinks. The agent-based model developed by the research team identifies critical points at which networks break down. It is hoped that the research will guide global industries on buffers required to keep supply chains moving in times of stress to address the vulnerabilities of just-in-time approaches that caused the 2021 crisis.
First author José Moran of the Institute for New Economic Thinking at the Oxford Martin School at the University of Oxford, said that agent-based models were ideally suited to solving a problem such as global supply chains.
And in a review of the work for Nature Reviews Physics ,'Too much efficiency leads to delays', Senior Editor Zoe Budrikis praised the clarity and applicability of the model developed.
"Moran and colleagues’ model is simple enough to be applicable to a range of contexts, including supply chains and transit systems.” Zoe said.
“The system is a set of nodes that represents, say, a manufacturing company or a train service. At each time step, a node has a delay that depends on the sum of delays of a set of nodes it is contingent on (the company’s suppliers at that time, or the incoming trains that are needed to provide crew for the outgoing train), minus a buffer (inventory or staff on call), plus a random noise term. The dependencies between nodes change at each time step; in the simplest mean field model the dependencies are chosen at random at each step."
José Moran said he was continuing to work with the team, including Deb Panja and JP Bouchaud, on extending the model to make it more realistic.
"I think this is a very exciting topic to tackle via agent-based models, which are the only way to capture truly dynamical effects. “This approach is radically different to what people do in comparative statics, where they assume the economy is in a market clearing equilibrium, and then implement a "shock" that drives the economy to a new equilibrium whose properties can be studied, but the dynamics in between are often ignored”.
"Of course, the model we studied is extremely stylised, and likely unrealistic, but I think the key point is simple and quite universal. In many cases, some socio-economic systems have dynamics that make them intrinsically unstable (unless you do something about it). I do think that this is *the* way to start understanding fluctuations in firm networks, and how they become macroeconomic fluctuations themselves. My colleagues at INET Oxford did a fantastic job studying the impacts of the Covid lockdown, showing you can use this to make useful predictions, and now at Macrocosm we are working to improve agent based models and bring them even closer to data. Exciting times!" he added.
Director of the Complexity Economics programme at INET Oxford, Professor J. Doyne Farmer said: 'This is an outstanding and fascinating piece of work that fully deserves this recognition'.
Full links to the paper and the review can be found below.
Background
In complex systems, external parameters often determine the phase in which the system operates, that is, its macroscopic behaviour. For nearly a century, statistical physics has been used to extensively study systems’ transitions across phases, (universal) critical exponents and related dynamical properties. Here we consider the functionality of systems, particularly operations in socio-technical ones, production in economic ones and, more generally, any schedule-based system, where timing is of crucial importance. We introduce a stylized model of delay propagation on temporal networks, where the magnitude of the delay-mitigating buffer acts as a control parameter. The model exhibits timeliness criticality, a novel form of critical behaviour. We characterize fluctuations near criticality, commonly referred to as avalanches, and identify the corresponding critical exponents. The model exhibits timeliness criticality also when run on real-world temporal systems such as production networks. Additionally, we explore potential connections with the mode-coupling theory of glasses, depinning transition and directed polymer problem.
