Author here. What started as a numerical experiment for a paper turned out to be something else.
The model: I take UN Comtrade data for crude trade flows, build an oil flow routing matrix (with countries as nodes, extending it to include maritime chokepoints) and simulate how the network is stressed by the closure of the Strait of Hormuz. Each country has oil reserves that deplete sequentially, in an event-driven simulation.
The math is basically Eisenberg-Noe interbank contagion, but for oil. Instead of bank defaults, nodes are limited by their “capped” production limits. The inventories evolve as a multi-dimensional Skorokhod problem, which sounds complicated but just means "inventory can't go negative and the shortfall propagates through the routing matrix."
What’s really surprising: In the Hormuz scenario, even though France’s sourcing doesn’t come from Hormuz at all, the endogenous pricing mechanism (prices spike, so every country inflates its safety-stock targets) pushes its reserve depletion 4.5 weeks EARLIER than in a fixed-price model. Nothing physical happened to it, deliveries dropped 0.0%. Its buffer gets eaten purely by the precautionary math of its own hedging policy responding to the global price. Meanwhile the countries that actually got hit (Japan loses 85% of flow) survive slightly LONGER, because high prices destroy demand and stretch their reserves.
Preemptive comments:
- “Exporter reserves depleting in 5 weeks is absurd." Yes if you consider them as storage. Here, they are export-capacity slack (calibrated as a percentage of geological reserves). Importer buffers are the real story (IEA 90-day stockholding mandates).
- There is no sanctioned/unreported trade: Fair, for example Iran-China trade is underrepresented. The tool is a stress-test and not a forecast. Adding the data is potentially an interesting follow-up exercise.
- There is not explicit rationing, buffers drain mechanically: The model is just a baseline.
- Costs are constant: In reality they are variable and a function of the oil price.
Engineering notes (since it’s HN):
Frontend is globe.gl, backend is ~600 lines of Flask with LLM assistance from the manuscript numerical example (outputs reproduce the paper’s tables)
Paper with proofs: [https://arxiv.org/abs/2607.17491]
Try it yourself! Close Malacca or Suez instead (the sliders also tell an interesting pricing story).
Well, the model is less of a prediction and more of a stress testing tool. But under the hypothetical closure scenarios it shows the timing the oil reserves of distant countries exhaust, as well as the systemic effects on pricing (the France paradox).
>What developments in pricing/other would indicate that your model is wrong or incomplete?
The model has a stylized way of incorporating pricing as a function of the total supply. In practice, when countries ration their oil that's beyond the scope of the model. That being said, the implied pricing trajectory is estimated and could be tested (the staircase graph showing prices constant while countries absorb the shock with their reserves and rebalanced whenever there is a reserve depletion).
>Nice website regardless, but I'm a bit skeptical that the dynamics of the global oil/energy market can be accurately predicted.
Thank you! Indeed, but I think having at least a stylized testing tool might be useful for policymakers :$ (assuming decisions are ever data-driven lol)
Author here. What started as a numerical experiment for a paper turned out to be something else.
The model: I take UN Comtrade data for crude trade flows, build an oil flow routing matrix (with countries as nodes, extending it to include maritime chokepoints) and simulate how the network is stressed by the closure of the Strait of Hormuz. Each country has oil reserves that deplete sequentially, in an event-driven simulation.
The math is basically Eisenberg-Noe interbank contagion, but for oil. Instead of bank defaults, nodes are limited by their “capped” production limits. The inventories evolve as a multi-dimensional Skorokhod problem, which sounds complicated but just means "inventory can't go negative and the shortfall propagates through the routing matrix."
What’s really surprising: In the Hormuz scenario, even though France’s sourcing doesn’t come from Hormuz at all, the endogenous pricing mechanism (prices spike, so every country inflates its safety-stock targets) pushes its reserve depletion 4.5 weeks EARLIER than in a fixed-price model. Nothing physical happened to it, deliveries dropped 0.0%. Its buffer gets eaten purely by the precautionary math of its own hedging policy responding to the global price. Meanwhile the countries that actually got hit (Japan loses 85% of flow) survive slightly LONGER, because high prices destroy demand and stretch their reserves.
Preemptive comments: - “Exporter reserves depleting in 5 weeks is absurd." Yes if you consider them as storage. Here, they are export-capacity slack (calibrated as a percentage of geological reserves). Importer buffers are the real story (IEA 90-day stockholding mandates). - There is no sanctioned/unreported trade: Fair, for example Iran-China trade is underrepresented. The tool is a stress-test and not a forecast. Adding the data is potentially an interesting follow-up exercise. - There is not explicit rationing, buffers drain mechanically: The model is just a baseline. - Costs are constant: In reality they are variable and a function of the oil price.
Engineering notes (since it’s HN): Frontend is globe.gl, backend is ~600 lines of Flask with LLM assistance from the manuscript numerical example (outputs reproduce the paper’s tables)
Paper with proofs: [https://arxiv.org/abs/2607.17491] Try it yourself! Close Malacca or Suez instead (the sliders also tell an interesting pricing story).
What concrete predictions does your model make?
What developments in pricing/other would indicate that your model is wrong or incomplete?
Nice website regardless, but I'm a bit skeptical that the dynamics of the global oil/energy market can be accurately predicted.
>What concrete predictions does your model make?
Well, the model is less of a prediction and more of a stress testing tool. But under the hypothetical closure scenarios it shows the timing the oil reserves of distant countries exhaust, as well as the systemic effects on pricing (the France paradox).
>What developments in pricing/other would indicate that your model is wrong or incomplete? The model has a stylized way of incorporating pricing as a function of the total supply. In practice, when countries ration their oil that's beyond the scope of the model. That being said, the implied pricing trajectory is estimated and could be tested (the staircase graph showing prices constant while countries absorb the shock with their reserves and rebalanced whenever there is a reserve depletion).
>Nice website regardless, but I'm a bit skeptical that the dynamics of the global oil/energy market can be accurately predicted. Thank you! Indeed, but I think having at least a stylized testing tool might be useful for policymakers :$ (assuming decisions are ever data-driven lol)