Why would you calculate VaR using Monte Carlo simulations?

Answer 1

Monte Carlo simulation is useful for calculating Value at Risk (VaR) when the relationship between market movements and the value of your portfolio is too complicated for a simple analytical formula.

VaR asks a question such as:

“With 99% confidence, how much could this portfolio lose over the next day?”

With Monte Carlo, you specify a model for the relevant risk factors—stock prices, interest rates, exchange rates, volatility, etc.—and simulate thousands or millions of possible future market outcomes.

For each simulated outcome:

1. Generate possible changes in the market risk factors.
2. Revalue the portfolio under those conditions.
3. Calculate the resulting profit or loss.
4. Rank all simulated profits and losses.
5. Take the appropriate percentile of the loss distribution.

For example, suppose you run 100,000 one-day simulations. For 99% VaR, you look near the loss threshold exceeded by only the worst 1% of simulated outcomes. If that threshold is a $5 million loss, your one-day 99% VaR is approximately $5 million.

The major advantage is flexibility. If a portfolio contains only simple linear positions, such as stocks, an analytical variance-covariance VaR calculation may be sufficient. But portfolios containing options and other nonlinear instruments can respond very differently depending on how the underlying market moves. Monte Carlo allows you to reprice those instruments under each simulated scenario.

It can also model multiple correlated risk factors, changing volatility, non-normal distributions and other complexities, depending on how sophisticated the simulation model is.

The disadvantage is that Monte Carlo can be computationally expensive and, more importantly, its answer is only as good as its assumptions. Running one million simulations does not make an incorrect probability model correct. Poor assumptions about correlations, volatility or tail behavior can still produce misleading VaR estimates.

So Monte Carlo is not necessary simply because you want to calculate VaR. It is particularly valuable when the portfolio or risk-factor behavior is sufficiently complicated that simpler methods cannot adequately represent the distribution of potential losses.