Should Supply-Chain Stress Tests Use Monte Carlo Simulation?
September 14, 2026
Altsets
Research by Altsets Research
Usually yes when the goal is portfolio risk. Monte Carlo can represent uncertainty in disruption severity, duration, substitution, correlated shocks, and recovery instead of forcing one deterministic network-loss scenario.
Data used:Altsets Supply Chain Intelligence: 90k+ entities, 400k+ relationships, 20+ years of history.
Key findings
- The supplied ASML-Micron relationship provides observed economic structure and materiality, but its relationship metrics should not be multiplied into a deterministic loss formula because shock pass-through, substitution, inventory, duration, and market repricing remain separate uncertain variables.
- Supply-chain risk research uses Monte Carlo simulation to generate loss distributions under uncertain and correlated network disruptions, making probabilistic stress testing a natural extension of relationship-aware portfolio risk analysis.
Yes. A probabilistic supply-chain stress test is usually more informative than one deterministic shock because the size of the disruption, the amount that reaches each connected company, and the market response are all uncertain. Monte Carlo simulation can turn one network scenario into a distribution of possible portfolio outcomes, as long as the model does not pretend that relationship percentages are transmission coefficients.
Relationship data can define the structure without defining the shock size
The supplied Altsets view shows ASML connected to Micron with a 3B USD relationship size, 7.64% of ASML revenue, and 11.91% of Micron's cost base. Those metrics establish that the relationship is economically meaningful on both sides. They do not say that a 10% ASML disruption should mechanically create a 1.191% Micron earnings loss or any particular stock return.
A Monte Carlo model should therefore separate observed network features from assumed transmission parameters. The graph identifies which firms are connected and which relationships appear more economically material. The scenario model separately specifies distributions for disruption severity, duration, substitutability, inventory buffers, pass-through, recovery time, and market repricing. Those assumptions can be broad when evidence is weak instead of being hidden inside a false deterministic formula.
Simulation is useful because several uncertain mechanisms interact
A single point scenario might assume an ASML disruption lasts four weeks and Micron absorbs one fixed loss. In reality, the disruption could be shorter or longer, alternative tools or inventory could offset part of the effect, another supplier could be hit at the same time, or the market could have already priced part of the risk.
Supply-chain risk research has used Monte Carlo methods for exactly this reason. Published models simulate disruption intensity, duration, network propagation, and loss distributions rather than treating one shock path as certain. Other work shows that network diversification can create counterintuitive risk outcomes once correlated hazards and multi-stage propagation are included.
Correlation between shocks matters more than multiplying edge weights
A portfolio stress model should not assume node disruptions are independent. A geopolitical event can affect several semiconductor suppliers at once. A natural disaster can hit multiple facilities in one region. A customer-demand shock can reach several suppliers simultaneously. Independent random draws can make the simulated portfolio look safer than the real network.
The graph can help identify where correlated shocks are plausible, while external data or explicit scenario assumptions determine the correlation structure. The researcher can compare independent-shock simulations with cluster-shock simulations and observe how expected shortfall changes. The exercise becomes a risk-model sensitivity analysis rather than a claim that the true probability distribution is known.
The output should be a distribution, not one dramatic number
A useful simulation can report median portfolio impact, loss quantiles, expected shortfall, probability of exceeding a chosen drawdown threshold, and which nodes contribute most often to bad outcomes. It can also compare mitigation choices such as reducing exposure to one dependency cluster or adding holdings with genuinely different network paths.
The exact dollar loss should not be presented as a forecast unless the transmission model is independently validated. Monte Carlo is valuable because it exposes how assumptions change the distribution. If a portfolio only looks safe under one optimistic pass-through assumption, that sensitivity is the finding.
The conclusion is to simulate uncertainty around the network, not certainty through it
Yes, Monte Carlo simulation is a natural extension of supply-chain stress testing. Use the relationship graph to define where shocks can plausibly travel, keep transmission and recovery assumptions separate from observed relationship metrics, and examine a distribution of portfolio outcomes rather than multiplying percentages into one false-precision scenario. The model is strongest when it makes uncertainty visible instead of hiding it.
The same-shock-different-outcomes guide explains why identical external events can produce different company outcomes. The tail-risk guide explains how network structure can matter more in extreme states than in average covariance.
For relationship definitions and evidence limits, read the Altsets methodology.