Stressed VaR is ordinary Value at Risk recalculated on a period of severe market stress instead of recent calm. For a company with currency exposure, it answers a question that normal VaR cannot: how large could the loss on today's positions be if the market behaved the way it did in its worst recent episode? Used together, the two figures give a finance team a daily baseline and a crisis ceiling for hedging decisions and limits. Our platform brings VaR analysis, live exposure tracking and hedging strategy simulation into one place, so a treasury team can look at both numbers on the same positions.
TL;DR
- VaR estimates a loss for a stated exposure, horizon and confidence level. Stressed VaR keeps that method but calibrates it to a period of significant financial stress.
- Regulators calibrate stressed VaR to a continuous 12-month stress window. A corporate can borrow the idea and pick the window that hurt its own currency pairs most.
- Stressed VaR does not replace stress tests. Run sensitivity analysis, historical scenarios and hypothetical shocks alongside it.
- Backtest the normal VaR every day and watch for clustered exceptions, not only the count.
- Set limits on both figures, and add simple operational limits such as nominal exposure caps.
Table of Contents
- What VaR and stressed VaR measure
- A worked example with a currency receivable
- How to build a stressed VaR for your FX exposure
- Stress tests: the companion to stressed VaR
- Backtesting: is your VaR telling the truth?
- Pitfalls and data quality
- Setting and monitoring VaR limits
- Our perspective
- Bring VaR, stress tests and hedging into one view
- Regulatory context in the UK
- FAQ
- Sources
What VaR and stressed VaR measure
Value at Risk is a loss estimate. The IMF's working paper on exchange rate risk defines FX VaR as the loss for a stated exposure over a specified horizon at a given confidence level, under the model's conditions. Three choices shape the number: the holding period, the confidence level and the currency you report in. The same paper gives one day as a typical horizon and 95% and 99% as the usual confidence levels.
In plain terms, a one-day 99% VaR of £100,000 means that on 99 days out of 100 you would expect to lose less than £100,000 on that position, if the model's assumptions hold. It says nothing about the size of the loss on the one remaining day. The IMF paper makes this point explicitly: VaR does not describe losses beyond its threshold.
That gap is where stressed VaR comes in. Ordinary VaR learns from a recent window, often one to two years of returns. If that window was quiet, the estimate is small, exactly when complacency is highest. Stressed VaR swaps the quiet window for a turbulent one.
The European Banking Authority's 2012 guidelines, written for banks' regulatory capital, describe the approach precisely. Stressed VaR is calibrated to a continuous 12-month period of significant financial stress relevant to the institution's portfolio, and that 12-month rule applies even if the stress event itself was shorter. The same guidelines say stressed VaR should build on the current VaR methodology, adjusted as needed, and reflect every risk factor in the VaR model. In other words, you do not invent a second model. You run the same one on a harsher history.
A worked example with a currency receivable
Take a hypothetical UK exporter with a $12.5m receivable, roughly £10m, falling due in about two weeks. The figures below are illustrative, not market data.
Suppose daily GBP/USD volatility in the last two years averaged 0.5%. In the stress window you chose, it averaged 1.2%. Using a simple parametric calculation, where the 99% one-tailed multiplier is about 2.33:
| Measure | Daily volatility | 1-day 99% VaR | 10-day 99% VaR |
|---|---|---|---|
| Normal VaR | 0.5% | about £116,000 | about £368,000 |
| Stressed VaR | 1.2% | about £279,000 | about £883,000 |
The 10-day figure scales the one-day number by the square root of 10, a common simplification that assumes returns are independent from day to day. The stressed figure is more than twice the normal one, and the gap is the information. If your board is comfortable with a £368,000 swing but not an £883,000 one, the hedge ratio on that receivable should reflect it.
Parametric maths is the easiest way to see the logic. Many treasury teams prefer historical simulation, which revalues the actual positions on each past day's currency moves and avoids assuming a bell-shaped distribution. We cover that approach in our guide to daily historical VaR for FX books.

How to build a stressed VaR for your FX exposure
1. Fix the exposure list first
Stressed VaR is only as good as the position data it runs on. List every currency exposure: receivables, payables, intercompany loans, forecast cash flows and existing forwards or options. Net them by currency pair and by date. If you need a refresher on which exposures belong in the list, our overview of FX risk types and mitigation walks through transaction, translation and economic exposure.
2. Keep the normal VaR settings
Choose the horizon and confidence level that match how quickly you can act. A company that can hedge within a day may use one day. A company that needs two weeks to adjust a pricing policy or renegotiate a contract should look at a longer horizon. Use 99% for the stress figure if your normal VaR uses 99%, so the two are comparable.
3. Choose the stress window
The regulatory template is a continuous 12 months. For a corporate, pick the 12 months in which your own pairs moved most violently against your positions. A sterling exporter and a euro-area importer may need different windows. Document why you chose it, so that the number is explainable to an audit committee a year later.
4. Run the same model on that window
Use the same risk factors, the same revaluation method and the same confidence level, replacing only the data window. This follows the EBA principle of building on the current VaR methodology. If a currency pair did not trade or has no history in the stress period, choose a proxy and record the choice.
5. Compare, report and decide
Report normal VaR, stressed VaR and the ratio between them. A rising ratio means your book is becoming more sensitive to a crisis regime, even while daily VaR looks calm. Tie hedging decisions to the stressed number: for example, raise the hedge ratio if stressed VaR exceeds a share of budgeted margin. Our guide to VaR-based hedging and the page on hedging based on value at risk show how this link is made in practice.
Stress tests: the companion to stressed VaR
Stressed VaR still relies on history. A shock that has never happened is not in the data. That is why stress testing is done alongside it, with three tools of rising imagination.
Sensitivity analysis moves one factor by a set amount and reads the result. The Bank of England's institutional stress-testing guidelines advise sensitivity analysis at exposure, portfolio, business-unit and institution-wide levels, proportionate to complexity, and recommend several degrees of severity to reveal nonlinearities and thresholds. For a treasury, that means testing a 5%, 10% and 20% move in sterling, not only one round number. Options and barrier products often behave differently at 20% than at 5%.
Historical scenarios replay a past episode, such as a sharp sterling sell-off, on today's positions. This is close to stressed VaR but shows the path and the cash impact, not only a statistical quantile.
Hypothetical scenarios describe events that may not have happened. The same Bank of England guidelines say scenarios should include hypothetical as well as historical events, be forward-looking, cover the main risk factors and the institution's own vulnerabilities, and document coherent assumptions about how factors move together. The PRA's January 2026 statement on the ICAAP and SREP expects firms to run sensitivity and scenario analysis at portfolio and firm-wide levels, with scenarios relevant to the business model, and to calibrate severe hypothetical downturns against the most adverse movements in individual drivers over a long historical period.

These are rules for regulated firms, not for a manufacturer in Leeds. They are still a good checklist. The key corporate question is not capital but liquidity: would a stressed FX move trigger margin calls, covenant breaches or a funding gap? Our article on FX stress scenarios that reveal funding gaps shows 1x to 2x shock sizes applied to treasury cash flows.
Backtesting: is your VaR telling the truth?
Backtesting compares each day's VaR forecast with what the portfolio actually did. A loss bigger than VaR is an exception. The Bank of England's working paper on VaR models explains the standard approach: compare each forecast with the corresponding profit and loss, check exception frequency against the confidence level, and check whether exceptions cluster in time.
At 99% over about 250 trading days, you should expect two to three exceptions. Far more suggests the model understates risk. Far fewer suggests it is too conservative, which means hedging capital or credit lines tied up needlessly.
Two classic tests formalize this. The paper describes the Kupiec test as comparing observed exceptions with the number expected, while Christoffersen's test also checks whether exceptions occur on adjacent days. Clustered exceptions usually point to a volatility model that reacts too slowly or to correlations that broke down. A single isolated breach needs investigation in context, not an automatic model change.
The FCA Handbook's standard on internal-model assessment is a useful benchmark for discipline. It looks for independent validation, assumptions that neither underestimate nor overestimate risk, backtesting specific to each risk and portfolio, and hypothetical portfolios that test features such as basis and concentration risk. A corporate can reasonably have someone outside the daily process review the model once a year.
Backtest the normal VaR, because it has a daily forecast to compare against. Stressed VaR is not meant to be breached by ordinary days. For its check, compare the figure with the loss your stress scenarios produce and investigate when the two are far apart. For a deeper treatment of the 95% and 99% rules, see our guide to FX VaR modeling and backtesting.
Pitfalls and data quality
Most VaR failures are data or assumption failures, not maths failures. The recurring ones:
- Stale or patched data. Missing days filled with the previous close flatten volatility. Use one consistent rate source and timestamp, and record how gaps are handled.
- Wrong positions. A VaR run on last month's exposure list is precise and wrong. Forecast cash flows change with sales, and hedges roll off.
- Correlations that move in a crisis. Diversification between pairs often shrinks when markets are stressed. Stressed VaR captures this only if the stress window contains such behavior, which is another reason to choose the window carefully.
- Reading VaR as the worst case. It is a threshold, not a maximum. The IMF paper notes firms often use operational limits such as nominal exposure limits or stop-loss orders alongside VaR for this reason.
- Scaling shortcuts. The square-root-of-time rule is convenient but assumes independence. Check it against actual multi-day moves.
- A window that is not documented. If nobody can say why the stress period was chosen, the number will be challenged and rightly so.
A simple data-quality routine helps: reconcile the position list to the ledger weekly, check rate feeds for gaps and outliers daily, and log every manual override with a reason.
Setting and monitoring VaR limits
A limit turns a measurement into a control. A workable structure for a corporate treasury has three layers:
- A VaR limit on the normal figure, for day-to-day monitoring, expressed in your reporting currency and as a share of budgeted profit or cash.
- A stressed VaR limit, set higher but tied to what the company could absorb without breaching covenants or cutting investment.
- Operational limits, such as maximum unhedged nominal per currency, maturity ladders and counterparty caps, which hold even if the model is wrong.
Define in advance what happens at each trigger: at 80% of a limit, notify the treasurer. At 100%, hedge or seek approval to exceed. Record the escalation path and who can authorize an exception. Limits that are only reviewed quarterly are not limits.
Our perspective
Most finance teams we talk about this topic with have a VaR number somewhere, often in a spreadsheet that is updated when someone has time. The harder problem is not the formula. It is having current positions, live rates, a stress view and a record of what was decided, all in one place.
We built CorpHedge around that problem. The platform offers currency position tracking, hedging strategy simulation, Value at Risk analysis, portfolio reporting and live market data, with automated notifications and operational expert advice. Our aim is that a CFO or risk manager can see the exposure, test a hedge against it and report the result without rebuilding the analysis each time.
Our view is simple: look at normal VaR and a stress figure together, and let the harsher one set how much risk you keep. A calm month is the best time to agree your limits, not the week the market moves.
Bartas
Bring VaR, stress tests and hedging into one view
If your team wants VaR analysis, position tracking and hedging simulation on the same live data, the CorpHedge platform is built for that, and our operational expert advice and Risk Safari Tours are available for teams that want guidance on their own exposures. The platform also connects with Corpay through the Corpay integration.

For teams building the skills in-house, our FX hedging course is a one-off purchase at €220, and the Academy lists the learning material.
Regulatory context in the UK
Stressed VaR is a regulatory concept first. For UK banks and investment firms, the Bank of England's updated market-risk statement SS13/13 takes effect on 1 January 2028, following the PRA's January 2026 policy statement. It includes expectations on expected shortfall, non-modellable risk factors, the default risk charge and stressed VaR. If your company borrows from or trades with regulated banks, you may see their pricing and credit terms reflect these rules over time. Non-financial companies are not bound by SS13/13, but its direction, towards tail-aware measures and stricter backtesting, is a good guide to where internal practice is heading.
FAQ
What is the difference between stressed VaR and VaR?
Both use the same model, horizon and confidence level. Normal VaR is calibrated on recent market data, while stressed VaR is calibrated on a period of significant financial stress, which regulators set as a continuous 12 months. The stressed figure is usually higher and shows how your positions would behave in a crisis regime.
What does 95% VaR mean?
A 95% one-day VaR is the loss you expect not to exceed on 95 days out of 100, under the model's assumptions. On the other five days the loss could be larger, and VaR does not say by how much. Many treasuries use 99% for limits and 95% for daily monitoring.
Is stressed VaR the same as a stress test?
No. Stressed VaR is a statistical estimate from a historical stress window. A stress test applies a specific shock or scenario, which can be hypothetical, and reads the effect on cash flow and profit. Companies should run both, because each catches what the other misses.
How often should we backtest FX VaR?
Compare forecast and outcome daily, and review exceptions at least quarterly, looking at both their count and whether they cluster. Investigate an isolated breach in context before changing the model. Repeated clusters are the signal to revisit volatility and correlation assumptions.
Sources
- Exchange Rate Risk Measurement and Management, IMF Working Paper 2006
- Guidelines on Stressed Value at Risk, EBA 2012
- The ICAAP and the SREP, PRA supervisory statement, January 2026
- Guidelines on institutions' stress testing, Bank of England
- Filtered historical simulation VaR models and their competitors, Bank of England Working Paper No. 525
- FCA Handbook, Article 9 General assessment standards
- SS13/13 Market risk, effective 1 January 2028, Bank of England
