To aggregate market risk exposures across multi-currency asset allocations, organizations use Value at Risk (VaR) models. VaR provides an estimate of the maximum potential financial loss over a specific timeframe within a given confidence level (typically a 99% confidence interval over a 10-day horizon for central bank portfolios).
The Parametric (Variance-Covariance) VaR Model
The parametric model assumes that portfolio returns follow a standard normal distribution. It utilizes historical price variations to compute the mean and standard deviation for each asset class, applying correlation metrics to estimate portfolio risk.
- Strengths: Highly efficient to calculate; excellent for stable, normal market regimes.
- Weaknesses: Fails to predict fat-tailed non-normal events, leading to a significant understatement of tail risk.
The Historical Simulation VaR Model
The historical simulation approach drops normal distribution assumptions. Instead, it applies actual price movements from a historical window (such as the preceding 250 or 500 trading days) directly to the current portfolio asset allocation.
[Current Asset Allocations] ---> Apply Past 500 Days of Real Price Changes ---> Sort Outcomes Low to High ---> Extract Target 99th Percentile Value
This model naturally captures real-world market distributions, skewness, and fat-tailed shocks, providing a highly reliable risk assessment during volatile economic cycles.
Â