Definition
Monte Carlo simulation estimates a quantity by repeatedly drawing outcomes from a specified probabilistic model and aggregating the resulting values. In portfolio risk, the model can generate a distribution of losses rather than only one point estimate. The result describes the chosen model and its sampling error, not reality independent of those choices.
How it is calculated
Monte Carlo estimate = (1/M) times the sum of simulated outcomes Y_m for m from 1 to M.
| Symbol | Meaning | Unit |
|---|---|---|
| M | number of simulations | count |
| Y | simulated outcome | same as target quantity |
| mu | target model expectation | same as Y |
-
Average the simulated outcomes.
-
Under independent finite-variance draws, divide the outcome variance by the number of simulations.
-
Take the square root to obtain the Monte Carlo standard error.
Interpretation
Imagine asking the same carefully specified model to play out many plausible futures. The average of those futures settles down as the number of draws grows, but it can still be precise about the wrong model. A tail estimate also needs enough simulated tail observations to be informative.
Assumptions and conventions
-
The simulation draws follow the stated model.
-
For the usual standard-error formula, draws are independent with finite variance.
The estimate has the units of Y; M is dimensionless.
Limitations and common misreadings
-
More simulations do not repair model misspecification.
-
A simulated loss distribution is not a realized forecast.
References
- Paul Glasserman. (2004). Monte Carlo Methods in Financial Engineering. Stochastic Modelling and Applied Probability. Chapter 1, Introduction. doi:10.1007/978-0-387-21617-1. Source

