It has been shown time and time again that volatilities and correlations are unstable. Depending on the sample used to estimate them, they may differ significantly—which can lead investors to underestimate their exposure to loss, and to form portfolios that are not sufficiently resilient to turbulent markets.
To better manage risk, we propose a methodology for partitioning historical returns into distinct samples that are characteristic of quiet and turbulent markets. By separating these samples, investors can:
- Stress test portfolios by evaluating their exposure to loss during turbulent conditions
- Structure portfolios that are more resilient to turbulent markets
- Shift their portfolios’ risk profiles dynamically to accord with their assessment of the relative likelihood that market conditions will be quiet or turbulent
To identify risk regimes, we use a statistical procedure to isolate periods in which returns are unusual. A period may qualify as unusual because two assets whose returns are positively correlated generate returns that move in opposite directions. Periods with unusual returns represent statistical outliers and are typically associated with turbulent markets. We create a subsample of these outliers and estimate volatilities and correlations from these turbulent subsamples. We also estimate volatilities and correlations from the remaining returns, which are associated with quiet markets.
How We Apply Risk Regimes
The most common application of the risk regime methodology is to stress test portfolios. Investors may solve for optimal portfolios based on the full sample of returns. However, once they have identified the optimal portfolio, they substitute the volatilities and correlations from the turbulent subsample and estimate its probability of loss and value at risk, assuming a turbulent regime will prevail.
For investors that are particularly sensitive to turbulent markets, they can identify portfolios that are likely to be more resilient to turbulence by blending the quiet and turbulent volatilities and correlations in a way that emphasizes the turbulent risk parameters. They can then use these blended parameters in the optimization process.
Another option for investors is to use regime-specific risk parameters to shift a portfolio’s weights dynamically by forecasting the relative likelihood of a quiet or turbulent regime. These relative probabilities can be used as the weights to blend volatilities and correlations in the optimization process.