FX Stress Testing: Historical, Hypothetical and Factor-Based Scenario Design

FX stress testing asks how a portfolio behaves under market conditions that are severe enough to challenge ordinary risk models, operating limits and management assumptions.

A useful stress programme does not rely on a single kind of scenario. It combines historical replay, hypothetical shocks and factor-based scenarios, then compares the results with ordinary VaR, Expected Shortfall, liquidity capacity and risk appetite.

The purpose is not to predict the next crisis. It is to make vulnerabilities visible before the market forces the institution to discover them.

Historical replay

Historical replay applies a past market event to the current portfolio.

The attraction is straightforward: the shock actually happened. The scenario therefore carries an empirical relationship between exchange rates, volatility, rates, spreads and other market variables.

Historical replay can answer questions such as:

• What would today's portfolio lose if a past dislocation happened again?
• Which positions now create more exposure than they did during the original event?
• Has the institution become more dependent on a correlation or funding assumption that failed previously?

Its main weakness is also obvious: history contains only the events that have already occurred.

A historical scenario should therefore be treated as evidence of what can happen, not as a complete catalogue of what will happen.

A scenario library should contain events that represent different transmission mechanisms rather than simply the largest daily FX moves.

Useful categories can include:

• abrupt currency devaluation;
• policy-rate divergence;
• funding-market stress;
• sovereign or political risk;
• global risk-off episodes;
• liquidity dislocation;
• volatility spikes;
• breakdown of a long-standing correlation;
• settlement or operational disruption combined with market stress.

For each event, record:

• event window;
• principal FX moves;
• volatility change;
• rates and basis movement;
• liquidity conditions;
• credit-spread behaviour;
• relevant market closures or intervention;
• rationale for retaining the event.

The scenario should then be applied to the current portfolio, not to the historical positions that existed at the time.

A hypothetical scenario begins with a risk narrative rather than a historical date.

For example:

Illustrative Example / Hypothetical Scenario

Illustrative Example / Hypothetical Scenario.

>

A sharp repricing of monetary-policy expectations causes a major currency to fall 12%, implied volatility to double, cross-currency basis to widen and liquidity in selected emerging-market pairs to deteriorate.

The value of a hypothetical scenario is that it can combine risks that have not previously occurred together.

The design should still be disciplined. A scenario should state:

1. the initiating event;
2. the affected risk factors;
3. the size and direction of each shock;
4. the assumed time path;
5. the dependence between factors;
6. the affected books and legal entities;
7. the management actions assumed — or deliberately not assumed;
8. the reason the scenario is severe but plausible for the intended purpose.

A practical design separates the scenario narrative from the factor matrix.

The figures are illustrative. The important control is that every shock has an explicit rationale and can be challenged independently.

Factor-based testing isolates specific weaknesses before combining them into a full scenario.

Carry unwind

A carry strategy can perform well while volatility is low and funding remains stable. A stress test can reverse the assumptions simultaneously:

• funding currency strengthens;
• target currency weakens;
• implied volatility rises;
• liquidity deteriorates.

The scenario tests whether apparent carry income is supported by a hidden short-volatility exposure.

Volatility shock

A volatility scenario is especially important for options portfolios.

It can shock:

• at-the-money volatility;
• smile/skew;
• term structure;
• correlations between currency pairs.

A simple parallel-volatility shock is easy to run but may miss the part of the volatility surface where the portfolio is most exposed.

Correlation breakdown

Correlation can create apparent diversification in ordinary VaR.

A stress test should ask what happens if:

• correlations move toward +1 or -1;
• historically offsetting positions stop offsetting;
• safe-haven relationships reverse;
• regional currency baskets become highly correlated.

The point is not to pick a single “correct” stress correlation. It is to expose how much risk reduction depends on the current correlation regime.

A one-day instantaneous shock and a ten-day disorderly repricing can produce different outcomes even if the final market levels are the same.

Path dependence matters when the portfolio contains:

• barrier options;
• margin calls;
• collateral thresholds;
• dynamic hedging;
• liquidity constraints;
• stop-loss rules;
• funding dependencies.

The scenario specification should therefore state whether the shock is instantaneous, staged or path-dependent.

The stress engine should use a valuation method appropriate to the portfolio.

For largely linear positions, factor sensitivities may be adequate for rapid screening.

For material options or structured exposures, full revaluation is normally more informative because gamma, vega and path dependence can dominate the stress result.

The methodology should explain where approximations are used.

Stress testing should sit beside VaR and Expected Shortfall, not disappear inside them.

A useful dashboard can compare:

• operating VaR;
• Expected Shortfall;
• historical replay loss;
• hypothetical scenario loss;
• reverse-stress break-even point;
• liquidity impact;
• relevant risk-appetite threshold.

Divergence is often the useful information.

If ordinary VaR remains stable while a hypothetical correlation-breakdown scenario rises sharply, the stress result is identifying a vulnerability that the daily model is not designed to represent.

A scenario library should evolve when:

• the portfolio changes materially;
• a new currency or product is introduced;
• market structure changes;
• an observed event exposes an unmodelled transmission channel;
• correlations or liquidity regimes change;
• a regulatory or supervisory exercise introduces a relevant scenario;
• repeated exceptions show that an old scenario no longer challenges the portfolio.

A scenario should not remain in the library merely because it is familiar.

For each approved scenario, retain:

• scenario ID and version;
• narrative;
• factor matrix;
• calibration sources;
• selected event window where historical;
• treatment of correlations;
• valuation methodology;
• assumptions about management action;
• owner and approver;
• approval date;
• result history;
• changes and retirement rationale.