DisruptSupplyChain

Aims. DisruptSupplyChain quantifies a disaster's indirect impacts by assessing the economic loss related to the perturbations of supply chains. It can be used to (1) evaluate the impact of a disaster, (2) run stress tests on the transport network to identify critical links, and (3) assess resilience-enhancing measures.

Scope. The model applies to one country and has a weekly time step.

Modeling paradigms

• Agent-based: macroeconomic aggregates emerge from the bottom-up through the modeling of multiple heterogeneous agents with behavioral rules, here firms, countries, and agents
• Input–output: the model has a sectoral structure based on input-output tables and is consistent with national account aggregates
• Dynamic: the model is a dynamical map X_t+1 = f(X_t), in which each time step represents one week
• Spatial: agents and infrastructure are spatially mapped
• Network: transport infrastructure and supplier-buyer interactions form two interconnected networks

Key features

• Agents. Three types of agents are modeled in the domestic country: firms, households, and countries. Each firm belongs to one economic sector. We model one firm per sector and relevant administrative unit (e.g., district, county) and one household per administrative unit. They are spatially assigned to the administrative units’ capital city or its most populated city. Countries represent the trading partners. Typically, they are aggregated at the level of continents, but specific countries can be kept separated (e.g., neighboring countries or countries with particularly significant trade relationships).
• Infrastructure. A multimodal transport network is represented. It allows firms to exchange goods. It consists of roads and, if relevant, inland waterways, railways, open-sea maritime routes, and airways.
• Supply chain. Firms are linked through supplier–buyer interactions. When no network data are available, supplier–buyer interactions are reconstructed using a probabilistic gravity model adjusted with the input–output table. Firms have a Leontief input mix defined by the input–output technical coefficient. They choose suppliers based on geographical distance and size. Firm sizes are derived from micro data (e.g., business registry).
• Disaster. The initial shock primarily consists of the complete but temporary disruption of a transport link or a transport node. The duration of the disruption can be adjusted. No goods can be transported on a disrupted link. If a transport node is disrupted, then the modeler has two options. Either all firms associated with this node can no longer transport goods out of this node (e.g., mild flood affecting the transport infrastructure only), or those firms have a reduced production capacity (e.g., severe flood affecting the productive assets).
• Dynamic. At each time step, firms plan production, order supplies, produce, adjust prices, deliver output and collect orders. Firms hold inventories, which they try to maintain to a target level. If a delivery route is unavailable, the firm chooses another, more expensive route or keeps the shipment on its premises if it is too costly. When a more expensive route is used, the increased costs are passed on to the clients. Similarly, any increase in input costs is passed on to the clients. If inventory is depleted, production is rationed.
• Indirect losses. Both effects—higher costs and rationing—can propagate along the supply chain down to households and international trade partners. Indirect losses are calculated as the sum of these two impacts.

Data. The model uses input–output tables, business census (e.g., location of firms, their sector, and size), population census, trade data (e.g., ComTrade), inventory data from surveys, infrastructure data (e.g., from OpenStreetMaps), logistic data (e.g., main ports, handling costs, speed). Additional data can be flexibly added, e.g., agriculture production, and mine locations.

Publications

Colon, C. , Hallegatte, S., & Rozenberg, J. (2021). Criticality analysis of a country’s transport network via an agent-based supply chain model. Nature Sustainability 4 (3) 209-215. https://doi.org/10.1038/s41893-020-00649-4