Charta is on Google Play: practise on the go. Get the app
Charta

Outline the challenges and points to bear in mind when putting the market-driven scenario approach into practice.

FRM® Part II · Risk Management and Investment Management · Market-Driven Scenarios: An Approach for Plausible Scenario Construction · FRM2.IM.15.d

The idea

Putting MDS into practice raises several issues. The covariance matrix is estimated from history, so the window, the weighting and whether it spans calm or stressed periods drive the results. Correlations typically rise in crises, so calm-period estimates understate co-movement. Real returns have fat tails and nonlinear dependence, so a normal-based distance and chi-square probabilities overstate how rare a scenario is. With many risk factors and few observations the sample covariance matrix is noisy or singular, so inversion is unstable and the number of factors must be limited or the matrix regularised. Conditional expectations are linear averages, so portfolios with options need full revaluation and perhaps volatility factors in the scenario. The plausibility threshold is a judgment, not a statistical bright line, and the horizon must match the data frequency. The exam tests whether you see MDS as a disciplined aid to judgment, not an objective forecast.

A practice question

From the question bank, with the reason each wrong option is wrong.

Which is a key practical challenge when implementing the MDS approach?

  1. MDS cannot be applied to portfolios that mix several asset classes.Why not: MDS is built for multi-asset portfolios; the approach handles several asset classes in one covariance matrix.
  2. MDS requires a full simulated distribution of portfolio returns before a scenario can be defined, as in Monte Carlo VaR.Why not: Confuses a scenario method with simulation-based VaR; no simulated distribution of portfolio P&L is needed.
  3. MDS leaves no room for analyst judgment, so scenario narratives cannot be incorporated.Why not: Reverses the facts: the analyst chooses the narrative, the shocked variables and their sizes.
  4. Results depend on covariance estimation: window and weighting matter, and calm data may understate stressed co-movement. - correct

The covariance estimate is the engine of MDS. Window length, weighting and the regime in the data affect the derived moves, and correlations usually rise in stress.

A harder one

A portfolio contains large short option positions. The MDS gives expected moves in the underlying risk factors. What is the practical implication for implementing MDS?

  1. A delta approximation is sufficient, because conditional expectations already capture the convexity of the option positions in the portfolio.Why not: Conditional expectations are linear and say nothing about convexity; delta alone misses gamma and vega effects.
  2. Options should be left out of the scenario, since they have no covariance with other factors.Why not: Ignores that implied volatilities and underlyings are risk factors with covariances.
  3. Fully revalue at the scenario's factor moves, including volatility factors, since expectations are linear and payoffs are not. - correct
  4. The scenario should be applied in several small steps to remove any need for revaluation.Why not: Does not fix the issue; the nonlinear portfolio needs full revaluation at the scenario point.

Because the conditional expectations are linear while option payoffs are not, full revaluation, including implied volatility moves, is needed to measure the scenario loss.

The hardest tier

In an MDS, equity is shocked by -20% (horizon volatility 16%). The credit spread has a horizon volatility of 80 basis points. The calm-period correlation is -0.4, while the correlation in stressed markets is -0.7. By how much does the calm-period estimate understate the derived spread widening, in basis points?

  1. 24Why not: Omits the shock-to-volatility ratio (20 / 16): 0.3 x 80.
  2. 30 - correct
  3. 37.5Why not: Squares the ratio of shock to volatility: 0.3 x 80 x 1.25².
  4. 70Why not: Reports the stressed-correlation spread move itself, not the difference from the calm estimate.

Calm: -0.4 x (80 / 16) x (-20) = +40. Stressed: -0.7 x 5 x (-20) = +70. The difference is 30 basis points, which is 0.3 x 80 x 1.25.

In the app: 6 practice questions on this outcome, spaced review that brings it back before you forget it.

Start free →

More in Market-Driven Scenarios: An Approach for Plausible Scenario Construction