Constructing coherent cross-asset scenarios

Begin with the vulnerability and the horizon

Define what the exercise is meant to expose: simultaneous equity and bond losses, commodity dependence, foreign-currency liabilities, or a funding shortfall. Set the holding horizon and balance-sheet boundary. An instantaneous market shock and a two-year macro path are different experiments. A terminal valuation alone may hide an earlier margin demand or a temporary breach that matters more than the final outcome.

Federal Reserve stress-test scenarios explicitly distinguish hypothetical exercises from forecasts. IMF stress-testing guidance similarly frames scenarios as tools for investigating vulnerabilities and transmission. Their institutional settings differ from a research portfolio, so borrow the methodological distinction rather than imply that an official bank scenario certifies a private portfolio or supplies the correct shock for every purpose.

Write the economic mechanism before assigning market moves

A negative demand disturbance can reduce commodity consumption and earnings while affecting policy expectations. A supply-driven inflation disturbance can raise some commodity prices and pressure margins while producing a different rate response. State which mechanism the scenario assumes and why. The same initial oil-price increase can have different cross-asset implications depending on whether it reflects stronger demand or disrupted supply.

Identify policy response, expectations, and lags as assumptions rather than automatic consequences. A central bank need not react immediately or exactly as it did in a selected historical episode. Write down alternative responses that could materially change the result. This prevents a story about inflation from quietly embedding an unsupported certainty about every point on the yield curve.

Translate the narrative into a factor path

Specify relevant yield-curve nodes, credit spreads, equity cash-flow or price assumptions, exchange-rate quotes, commodity maturities, and volatility. Match units and horizons. If equity prices are shocked directly, do not also add a separate earnings-and-discount-rate loss unless the method deliberately reconciles those components. Otherwise, the same economic effect can be counted twice under different labels.

Choose whether exposures remain fixed or adapt over the scenario. Rebalancing, hedge changes, and funding decisions require their own stated rules and feasibility assumptions. A fixed-position exercise is often easier to interpret, while a dynamic path can address a different question. Present these as alternative designs rather than silently allow only favorable management actions after losses appear.

Worked hypothetical example: a joint inflation stress

Consider an invented USD 100 million portfolio containing USD 50 million of bonds with modified duration six, USD 30 million of equities, USD 10 million of commodity exposure, and USD 10 million cash. Assume a parallel yield rise of one percentage point, an equity price fall of 12%, and a commodity return of plus 20%. Hold currency, spreads, positions, and cash income constant for this simplified exercise.

The first-order bond effect is approximately minus 6%, or minus USD 3 million. Equities contribute minus USD 3.6 million and commodities plus USD 2 million. The aggregate estimated change is minus USD 4.6 million, or minus 4.6%. These invented shocks are not calibrated forecasts. The bond estimate omits convexity and income, while any derivative commodity implementation would require its own collateral and cash-flow treatment.

Add funding and nonlinear effects without confusing them with loss

A collateral transfer can reduce immediately available cash without being an additional valuation loss. Track margin, financing, settlement, and redemptions on a dated cash ledger. Avoid adding the full collateral posting to market loss when it remains an asset. Conversely, do not assume a modest terminal loss guarantees the portfolio can meet a large interim cash demand.

The ECB's analysis of leveraged funds distinguishes notional exposures from more informative measures of market and liquidity risk. For options, use full repricing or appropriate nonlinear approximations where available. For bonds, large yield changes can make a duration-only estimate inadequate. Where the necessary model is unavailable, publish a bounded approximation and explain which risks the number does not cover.

Challenge internal consistency and severity

Review whether market shocks are economically compatible with the narrative, without requiring every scenario to resemble an observed episode. A deliberately extreme synthetic combination can investigate a breaking point if labeled clearly. Compare a historical replay, a mechanism-based hypothetical case, and a reverse stress aimed at a defined loss or liquidity threshold. Each answers a different question.

Vary the assumptions that dominate the outcome rather than perturb every input indiscriminately. Test curve shape, correlation, funding availability, and commodity basis when relevant. If a result improves only because an assumed hedge is purchased after the shock at an unchanged price, expose that dependency. A scenario should challenge implementability under stress, not let an optimistic response repair every vulnerability.

Communicate conditional findings and unresolved channels

Summarize the dominant loss drivers, path-dependent cash needs, approximations, and assumptions that could reverse the conclusion. Keep model output distinct from expert judgment. Use sensitivity ranges only where their construction is explained; a range is not automatically a confidence interval. Preserve the input scenario so a reviewer can reproduce the result and distinguish a later exposure change from a changed stress design.

Common failures include calling a scenario a prediction, summing incompatible worst historical moves, and double-counting price and factor shocks. Another is reporting a portfolio loss without identifying the data that were missing. A useful exercise can leave important uncertainty unresolved while showing exactly where the vulnerability sits. It should support further research and preparedness, not present a current trade recommendation.

What to retain

• A coherent scenario connects an economic mechanism to a stated factor path.
• Keep valuation losses, collateral movements, and interim funding requirements separate.
• Conditional results are not forecasts, probabilities, or portfolio certifications.

Source notes

1. Federal Reserve Board: 2025 Stress Test Scenarios, hypothetical exercise definitions
2. IMF: Stress Testing Financial Systems, scenario construction
3. ECB: Leveraged investment funds, a framework for assessing risks

Related research