Stress tests: model the chain of failure

Ask what can break, not just what is likely

A stress test asks how a portfolio or process behaves under a deliberately adverse set of conditions. It is not a forecast, and attaching a precise probability to every imaginative scenario can create false comfort. Its usefulness lies in exposing vulnerabilities that an average day or a volatility estimate conceals.

Start with an explicit decision: whether available cash can meet collateral needs, whether a concentration dominates losses, or whether a planned exit is feasible. A test without a decision often becomes a dramatic spreadsheet with no practical consequence. Institutional stress-testing principles emphasize objectives, governance, data quality and meaningful challenge; a small educational portfolio can adopt that discipline without pretending to replicate bank regulation.

Use coherent stories and isolated sensitivities

A sensitivity test changes one input, such as an interest rate or exchange rate, while holding others fixed. This reveals a clean mechanical relationship. A scenario combines changes that could occur together: falling asset prices, wider spreads, weaker collateral and reduced access to funding. Both views are useful, but they answer different questions.

Avoid assembling a superficially extreme story from mutually inconsistent assumptions without labeling it. Conversely, do not assume familiar hedges always work. A bond-equity relationship observed during one economic shock may reverse under another. Describe the economic mechanism, time horizon and instruments affected. Include a scenario in which the portfolio's apparently diversifying positions move adversely together.

Separate valuation loss from cash need

A mark-to-market loss reduces equity, but a margin demand is an immediate cash requirement. They are related without being identical. Some instruments settle losses daily, while others require additional collateral or permit discretion over haircuts. Cash in an inaccessible account may not be transferable before the deadline.

Map each obligation by currency, location and timing. A portfolio can appear solvent on a consolidated statement while lacking the right cash in the right place. Do not assume proceeds from selling an asset will arrive instantly or at the displayed price. Settlement timing, banking hours, withdrawal restrictions and platform outages can all disrupt a seemingly straightforward funding plan.

Collateral haircuts connect market value to usable funding. If an asset worth $10,000 is accepted at a 10% haircut, its collateral value is $9,000. If the asset then falls to $8,000 and the haircut rises to 25%, collateral value becomes $6,000. Funding capacity has declined by $3,000, although the asset’s mark-to-market loss is $2,000. The additional reduction reflects less borrowing capacity, not an additional trading loss. The actual cash demand depends on the obligation the collateral supports.

Let liquidity change with the shock

A constant exit-cost assumption is especially weak during stress. Spreads can widen, depth can disappear and multiple positions may compete for the same limited liquidity. A stop order should be modeled at plausible executable prices rather than its trigger alone. Options and leveraged products can introduce nonlinear changes that a simple percentage shock misses.

Use at least two liquidation conventions: immediate stressed execution and a slower disposal path with continuing price risk. A larger displayed volume does not necessarily mean a large position can exit cheaply. The relevant question is the price concession and time required for the specific size, not whether some trades occurred elsewhere in the market.

Reverse the problem and predefine responses

Reverse stress testing begins with an unacceptable outcome, such as a cash shortfall or a specified loss, and asks what combination of shocks could cause it. This can reveal that several modest problems together are more dangerous than one spectacular event. Challenge dependencies: a hedge may fail because the counterparty or venue is also under strain.

Record potential responses before the event, including which actions are operationally possible and who has authority. Then rerun the test after material portfolio changes. A spreadsheet that assumes flawless discretionary action is testing an imaginary portfolio. Preserve the original assumptions and results so revisions can be distinguished from attempts to explain away an uncomfortable finding.

Walk the scenario through time, not just to an endpoint

An endpoint portfolio value can miss the event that ends the strategy. Imagine a position that suffers a large intraday loss and later recovers. If the account has to post collateral during the decline, the final closing price does not establish that the account could have held on. Build a sequence of events: price shock, collateral calculation, payment deadline, possible asset sales, settlement and any recovery. The ordering determines whether the plan remains feasible.

Keep a cash-flow ledger for that sequence. In the existing example, $7,000 of available cash faces a $4,000 cash-settled loss and a separate $5,000 collateral increase. The ledger reaches −$2,000 before any rescue transaction. If an asset sale produces $3,000 only after the payment deadline, it does not cure the earlier shortfall. If the sale can settle in time but costs $500 under stress, usable proceeds are $2,500. These are different scenarios and should not be collapsed into a single optimistic assumption.

Contingent actions need their own dependencies. Selling a hedge to raise cash may reduce immediate funding pressure while increasing the remaining portfolio’s exposure. Moving collateral from another venue may require that venue, the bank and the transfer channel to function at the same time. A line of credit may be conditional rather than committed. Model a response as available only under the conditions that make it executable, then rerun the portfolio after the response instead of assuming the original risk profile remains unchanged.

A reverse stress exercise can be stated as a threshold question: what combination of settlement losses, extra collateral and unavailable cash makes timely resources fall below obligations? Several combinations may cross that boundary. Report the boundary and the assumptions, not a claimed probability of failure unless a defensible probability model exists. The practical lesson is often specific: a portfolio is vulnerable to the timing of one transfer, the value of one collateral pool, or a funding source that disappears during the same shock it is supposed to cover.