Models & quantification

Monte Carlo simulation

A Monte Carlo simulation is a calculation technique that makes the uncertainty in a model visible by repeating the calculation many times with randomly drawn values for the uncertain inputs, each following a chosen probability distribution. The result is not a single outcome but a distribution of possible outcomes. The standard IEC 31010:2019 classifies it among the techniques for understanding the consequences and likelihood of risks.

Source: IEC 31010:2019 (Annex B.5.10); NIST, Monte Carlo Tool version 1.2 (2025); NIST SP 800-30 Rev. 1 (2012), section 2.3.2; Dutch Decree on the Financial Assessment Framework for Pension Funds (Besluit financieel toetsingskader pensioenfondsen) Articles 22 and 23b

Instead of a single point estimate, a Monte Carlo simulation shows how likely it is that a target will be met and how large a setback could become in adverse cases. The approach: determine which variables are uncertain, choose a probability distribution for each, draw random values, run the model and repeat this a large number of times. For its own Monte Carlo tool, NIST names the triangular, normal and uniform distributions as common choices and leaves the number of iterations to the user. The set of outcomes supports statements such as: with a chosen level of confidence, the loss will stay below a given amount.

IEC 31010:2019, the international standard on risk assessment techniques, describes Monte Carlo simulation in Annex B.5.10, among the techniques for understanding consequences and likelihood, alongside Bayesian analysis and Markov analysis, among others. A related approach appears in the Dutch pension rules: a pension fund performs its feasibility test on the basis of a stochastic analysis (Article 22 of the Decree on the Financial Assessment Framework for Pension Funds, Besluit financieel toetsingskader pensioenfondsen) and uses a uniform set of 10,000 economic scenarios provided by the Dutch Central Bank (De Nederlandsche Bank, DNB) for scenario analysis (Article 23b).

The output is only as good as the input. Poorly chosen distributions, underestimated dependencies between variables or too few iterations create a false sense of precision. NIST SP 800-30 Rev. 1 (2012) notes more generally that the rigour of quantification is significantly reduced when subjective judgements are buried in the calculation or when the values are highly uncertain. Directors should therefore ask about the distributions used, the source of the parameters, the dependencies between variables and how sensitive the outcome is to those choices. A Monte Carlo simulation differs from scenario analysis: the latter works with a small number of considered, coherent scenarios, the simulation with a large number of random draws.

Last updated: · Editor: ONE Risk Advisory

Frequently asked questions

What is the difference between a Monte Carlo simulation and a stress test?

A Monte Carlo simulation draws a large number of random combinations of input values and produces a distribution of outcomes. A stress test deliberately selects one or a few severe but plausible circumstances and calculates their effect on capital, liquidity or results. The simulation says something about likelihoods; the stress test about the consequences of a chosen setback, even when its likelihood is hard to estimate.

Can organisations outside the financial sector use Monte Carlo simulation?

Yes. IEC 31010:2019 describes the technique for risk assessment in a wide range of situations, not for one sector. Examples include the likelihood that a project stays within budget, the spread of claims costs or the range of a cash flow forecast. The condition is that the uncertain variables and their spread can be described with data or well-founded estimates.

← All terms