What is Stress Testing?

By EC Assets Research Team · Published 19 August 2026 · Updated 19 August 2026

Stress Testing: Evaluating a portfolio under specified severe conditions instead of statistical distributions. It exists because the events that matter most are too rare to estimate from data - so they are imposed rather than inferred.

What Stress Testing Does

A stress test asks a direct question: if this happens, what do we lose? It replaces the statistical machinery of volatility and value-at-risk with a named scenario and works out the consequence.

The reason it exists is the reason tail risk cannot be measured well. Rare events are, by definition, scarce in the sample. A ten-year history contains perhaps two or three genuine crises, which is far too few to fit a distribution to. Rather than infer the tail from data that barely contains it, stress testing imposes the tail and examines the result.

That changes the nature of the exercise. A value-at-risk figure is an estimate with a confidence level. A stress result is a conditional statement — no probability attached, and none implied.

How It Works

Three approaches are used, and they answer different questions.

Historical replay applies the actual moves of a past episode to today's book. It has the advantage of internal consistency: the correlations, the sequencing and the liquidity conditions all come from something that genuinely happened.

Hypothetical scenarios construct a plausible future that has not occurred — a specific rate shock, a sovereign event, a counterparty failure. They cover risks no history contains, at the cost of resting entirely on judgement about how variables move together.

Reverse stress testing inverts the question. Instead of asking what a scenario costs, it asks what scenario would break the firm, then examines how plausible that is. It is generally the more useful discipline, because it does not require guessing the shock in advance.

Worked Example

A book runs 200 percent gross and 60 percent net equity exposure with a 12 percent annualised volatility.

Under a March 2020 replay the net exposure alone implies roughly a 20 percent loss. But the same episode saw long and short books converge — crowded shorts rallied while longs fell — so the gross exposure, not the net, drives part of the damage. Add a widening of financing spreads and a doubling of margin requirements, and the question stops being about mark-to-market loss and becomes about whether the position can be held at all.

That progression is the point. A statistical model reports a number; a stress test reveals a sequence — loss, then margin, then forced liquidation, then the loss realised at the worst prices of the episode.

When It Applies (and Limitations)

Scenario selection is the whole exercise. A stress test only covers what someone thought to test. The 2008 crisis was survived by institutions that had modelled a housing decline and destroyed those that had not, and no statistical property of the test distinguished them — only the imagination behind it.

It fights the last war. Historical replays are calibrated on shocks that already happened and to which markets have adapted. The next episode rarely repeats the last one's mechanism.

Correlations must be assumed. The most consequential input is how assets move together under stress, and that is precisely the parameter that behaves differently in the tail than in the sample. Applying calm-period correlations to a crisis scenario understates the loss systematically.

Liquidity is usually omitted. Most stress tests revalue positions at stressed prices while implicitly assuming they could be sold there. In the episodes that matter, the exit price and the screen price are different numbers.

No probability is attached, and that is a feature. A stress result answers what if, not how likely. Presenting stress losses alongside probabilistic measures without marking the distinction invites the reader to average them mentally, which is meaningless.

Why It Matters for Institutional Investors

It is where leverage limits should be set. The binding question is not the expected loss but whether the worst plausible one is survivable at current gearing. Stress testing is the only tool that answers it directly.

Regulation requires it. Bank capital under the Basel framework rests on supervisory stress scenarios, and fund managers face liquidity stress testing obligations under AIFMD and the associated ESMA guidelines. For a regulated firm the exercise is a compliance artefact as well as a risk tool — which is a hazard, since compliance exercises tend to converge on the scenarios everyone runs.

It belongs in operational due diligence. Asking a manager which scenarios they run, when they last changed them, and what the reverse stress test says reveals more about their risk culture than any ratio on the factsheet.

It disciplines the redemption question. For an institution with spending obligations, the relevant stress is not only the drawdown but the drawdown coinciding with a withdrawal — the case where a paper loss becomes permanent.

References

1. Basel Committee on Banking Supervision. Stress Testing Principles. (https://www.bis.org/bcbs/publ/d450.htm)
2. European Securities and Markets Authority. Guidelines on Liquidity Stress Testing in UCITS and AIFs. (https://www.esma.europa.eu)
3. Berkowitz, J. (2000). A Coherent Framework for Stress Testing. Journal of Risk, 2(2), 5–15.
4. Embrechts, P., Klüppelberg, C., & Mikosch, T. (1997). Modelling Extremal Events for Insurance and Finance. Springer.