Monte Carlo VaR

Monte Carlo value-at-risk (Monte Carlo VaR) is a risk measurement method that simulates thousands or millions of possible future market scenarios using probabilistic models of price movements, correlations, and volatility. The value-at-risk is then calculated from the distribution of simulated portfolio losses.

How Monte Carlo VaR works

Step 1: Define models for price movements. For each asset in the portfolio, define how its price evolves. Common model: the geometric Brownian motion, where price drifts at some rate and is buffeted by random shocks (volatility).

Example: Stock price follows: dS = μS dt + σS dW

Where μ is the drift (expected return), σ is volatility, and dW is a random shock.

Step 2: Generate random scenarios. Use a random number generator to create thousands (or millions) of possible future paths for each asset. Each path represents one possible future over the time horizon (e.g., 10 days).

Step 3: Calculate portfolio value in each scenario. For each simulated scenario, calculate the portfolio’s value at the horizon (e.g., 10 days from now). Account for:

• Changes in underlying asset prices.
• Changes in correlations (if modeled).
• Non-linear effects (e.g., options become more or less in-the-money).

Step 4: Calculate loss distribution. Sort the simulated portfolio values from best to worst. Calculate the loss in each scenario.

Step 5: Extract VaR. For 99% VaR, find the worst 1% of simulated scenarios and use the loss in the 99th percentile worst scenario as the VaR.

Advantages of Monte Carlo VaR

Flexibility. Can handle any instrument, any distribution, any dynamic. Options, bonds with embedded options, exotic derivatives — all can be simulated.

Realism. Can model fat tails, jumps in prices, volatility clustering, and correlation breakdowns — all empirically observed in markets.

No distributional assumption errors. Unlike parametric-var, there is no assumption that returns are normal. The distribution emerges from the simulations.

Non-linear instruments. Options are non-linear; their value does not move linearly with the underlying. Monte Carlo naturally handles this.

Multiple risk factors. Can model many correlated risk factors (rates, FX, volatility, credit spreads, etc.) simultaneously.

Disadvantages of Monte Carlo VaR

Computational cost. Simulating a million scenarios for a portfolio of 100 assets requires billions of calculations. It is expensive and slow.

Model risk. The accuracy of Monte Carlo VaR is entirely dependent on the quality of the underlying models. If the price movement model is wrong, the VaR is wrong.

Parameter estimation. You still need to estimate volatility, correlation, and drift. Errors in these estimates feed into simulation errors.

Scenario quality. The simulated scenarios are only as realistic as the model. If the model does not capture fat tails, the simulated worst scenarios will not either.

Black swan blindness. Monte Carlo simulates based on historical patterns. A new type of tail event (a black swan) will not appear in simulations.

Example: Monte Carlo VaR for a stock portfolio

Suppose a $100M portfolio of 2 stocks, A and B, each $50M.

Models:

• Stock A: daily return ~ Normal(0.05%, 1.5%)
• Stock B: daily return ~ Normal(0.05%, 1%)
• Correlation: 0.5

Simulation:

1. Generate 100,000 random scenarios of 10-day future returns for A and B.
2. For each scenario, calculate portfolio return.
3. Calculate portfolio loss in each scenario.
4. Sort losses from worst to best.
5. 99% VaR = the 1,000th worst loss (worst 1% of 100,000).

Suppose the 1,000th worst scenario resulted in a loss of $3.5M.

Result: 10-day 99% Monte Carlo VaR = $3.5M.

When Monte Carlo VaR is essential

• Options portfolios. Options are non-linear; Monte Carlo handles this naturally.
• Complex instruments. Structured products, exotics, callable bonds — anything with embedded optionality.
• Multiple risk factors. When interest rates, FX, volatility, and credit spreads all matter and are correlated.
• Stress scenarios. You can run the simulation under “what if” conditions (e.g., 10% stock decline) to see portfolio impact.

Model risk in Monte Carlo VaR

The greatest risk in Monte Carlo VaR is model-risk: the model for price movements is wrong.

Example: A model assumes stock prices follow a geometric Brownian motion (continuous, no jumps). In reality, prices can jump sharply on earnings announcements or crises. The model’s simulated scenarios do not include jumps, so Monte Carlo VaR underestimates tail risk.

This is why sophisticated risk managers pair Monte Carlo with stress-testing and scenario-analysis. Monte Carlo gives a baseline; scenarios test extremes the model might miss.

Practical use

Large banks and hedge funds use Monte Carlo VaR extensively because:

1. Portfolios are complex; simpler methods do not work.
2. Risk frameworks allow Monte Carlo as an acceptable method.
3. Computing power is cheap relative to the cost of mispricing risk.

But they also:

• Validate the underlying models rigorously.
• Backtest to ensure actual losses do not exceed predictions.
• Supplement with stress-testing and expected-shortfall.
• Use multiple methods (parametric, historical, Monte Carlo) to cross-check.

See also