Skip to content

Risk Management in Banking and Insurance · Market Risk Management

Stress Testing and Expected Shortfall in Market Risk

Updated 11 October 2026 · Fact-checked

Stress testing and scenario analysis ask what a bank would lose under severe but plausible shocks. Expected shortfall (ES) is the average loss in the worst tail beyond the VaR cutoff. Both capture tail risk that VaR ignores. To solve questions, define the scenario or confidence level, compute losses, then compare with capital and limits.

Understand Stress Testing and Expected Shortfall

Value at Risk (VaR) gives one number: the loss that will not be exceeded at a stated confidence level over a stated horizon. For example, a 1-day 99% VaR of ₹10 crore means that on 99 days out of 100 the loss is expected to be within ₹10 crore. It says nothing about how bad the remaining 1 day could be.

Expected shortfall (ES), also called conditional VaR, fills that gap. It is the average loss on the days when the loss exceeds VaR. So ES is always at least as large as VaR at the same confidence level. It tells you how heavy the tail is, not only where it starts. ES is also subadditive, so diversification never makes the combined ES larger than the sum of the parts. VaR does not always have this property. The Basel market risk framework for internal models moved from VaR to ES for this reason.

Stress testing measures the loss from an extreme move in risk factors such as interest rates, exchange rates, equity prices or credit spreads. It does not rely on a statistical distribution, so it can show losses outside historical experience.

Scenario analysis is a closely related tool. It builds a coherent story, such as a sharp rise in yields combined with rupee depreciation, and revalues the portfolio. Scenarios can be historical (repeat a past crisis), hypothetical (designed by management) or based on sensitivity shocks (for example, a 200 bp parallel shift). Reverse stress testing starts from an unacceptable outcome, such as breaching capital limits, and works backwards to find what would cause it.

The three tools work together. VaR monitors normal-day risk, ES describes the tail, and stress tests show vulnerability to events the data may not contain. Results go to senior management and the board, and they feed limits, capital planning and contingency actions.

Key rules to remember

Expected shortfall (discrete, equally likely outcomes)
ES = Average of all losses that are greater than or equal to the VaR cutoff, i.e. the worst (1 − c) share of outcomes
c is the confidence level. At 97.5%, ES uses the worst 2.5% of outcomes.
Relation of ES to VaR
ES ≥ VaR at the same confidence level
ES is the average of tail losses. Never present ES as smaller than VaR.
Number of tail observations
Tail observations = N × (1 − c)
With 200 observations at 95%, the tail has 10 observations.
Stress loss from a sensitivity shock
Loss ≈ Σ (Exposure × Shock)
For a bond portfolio: Loss ≈ Modified duration × Δy × Market value, for small yield changes.
Stress loss as share of capital
Stress loss ÷ Capital × 100
Used to judge whether capital is adequate after the shock.

How to solve Stress Testing and Expected Shortfall questions

Use this order for any question on stress testing, scenarios or expected shortfall.

  1. 1Identify what is asked: VaR, ES, a stress loss, or a discussion of why the tools are needed.
  2. 2Note the confidence level, horizon and number of observations or scenarios given.
  3. 3For ES, sort losses from worst to best and find the tail: N × (1 − c) observations.
  4. 4Average the tail losses to get ES. Pick the VaR cutoff as the boundary loss as the question defines it.
  5. 5For stress tests, apply each shock to each exposure, sum the losses and add any hedge offsets.
  6. 6Compare the result with capital, limits or risk appetite and state whether it is acceptable.
  7. 7Close with a recommendation: reduce exposure, hedge, raise capital or revise limits.

Quickest way: Tail average and shock table

When to use it: For numerical MCQs on ES or stress loss when time is short.

  1. Count the tail size using N × (1 − c) first.
  2. Write the worst losses only. Ignore the rest.
  3. Add them and divide by the tail count.
  4. For stress loss, multiply exposure by shock for each line and add. Use duration for bond shocks.
  5. Check that ES is not below VaR before marking the answer.

Common mistakes in Stress Testing and Expected Shortfall

  • Stating that ES is the loss at the VaR point.

    Both use the same confidence level, so they seem the same.

    Fix: VaR is the cutoff. ES is the average of losses beyond the cutoff, so it is equal to or larger.

  • Averaging all losses instead of only the tail.

    Students forget to compute the tail count.

    Fix: Compute N × (1 − c) first and average only that many worst losses.

  • Saying stress testing predicts the probability of a loss.

    It is confused with VaR, which is probability based.

    Fix: Stress tests give the size of loss in a stated scenario. They do not assign a probability.

  • Treating scenario analysis and sensitivity analysis as identical.

    Both apply shocks to risk factors.

    Fix: Sensitivity shocks one factor at a time. Scenario analysis moves several factors together in a coherent story.

  • Ending the answer with a number and no action.

    Students stop once the calculation is done.

    Fix: Compare the loss with capital or limits and add a short recommendation.

Worked examples

Example 1

A bank's trading desk has 20 daily loss figures (₹ lakh) at the worst end of its 200-day history: 62, 55, 50, 47, 45, 43, 41, 40, 39, 38 and smaller losses below that. Using 200 observations and a 95% confidence level, compute the VaR cutoff and expected shortfall.

Show the solution
  1. Tail observations = 200 × (1 − 0.95) = 10.
  2. The worst 10 losses are 62, 55, 50, 47, 45, 43, 41, 40, 39 and 38.
  3. Taking the boundary of the tail as the 10th worst loss, the VaR cutoff is ₹38 lakh.
  4. Sum of the tail losses = 62 + 55 + 50 + 47 + 45 + 43 + 41 + 40 + 39 + 38 = 460.
  5. ES = 460 ÷ 10 = ₹46 lakh.
  6. ES of ₹46 lakh is greater than VaR of ₹38 lakh, as expected.

Answer: VaR cutoff = ₹38 lakh and expected shortfall = ₹46 lakh.

Example 2

A bank holds a government bond portfolio of market value ₹800 crore with modified duration 5. Management runs a stress scenario of a 150 bp parallel upward shift in yields. The bank's Tier 1 capital is ₹4,000 crore. Estimate the loss and state its share of capital.

Show the solution
  1. Yield shock Δy = 150 bp = 1.5% = 0.015.
  2. Loss ≈ Modified duration × Δy × Market value.
  3. Loss ≈ 5 × 0.015 × ₹800 crore = ₹60 crore.
  4. Share of capital = 60 ÷ 4,000 × 100 = 1.5%.
  5. This is a linear estimate. For a large shock, convexity would slightly reduce the actual loss.

Answer: Estimated loss is ₹60 crore, which is 1.5% of Tier 1 capital. This is manageable, but the bank should also test other shocks and combined scenarios.

Exam tips

  • Write the one-line contrast first: VaR gives a cutoff, ES gives the average beyond it, stress tests give losses in stated scenarios.
  • In descriptive answers, name the three scenario types: historical, hypothetical and reverse.
  • For numerical questions, show the tail count before averaging. Method marks depend on it.
  • End every answer with a link to capital, limits or board reporting.
  • Case-based MCQs often test why ES is preferred to VaR. Think of tail severity and subadditivity.

Practice questions from Market Risk Management

Stress Testing and Expected Shortfall in other exams

The same ground in other exams, if you are preparing for more than one or want another angle on it.

Stress Testing and Expected Shortfall: frequently asked questions

What is the difference between expected shortfall and VaR?

VaR is the loss level that is not exceeded at a given confidence level. Expected shortfall is the average loss in the cases that go beyond that level. So ES shows how severe the tail is, and it is at least as large as VaR.

Why do banks need stress testing if they already use VaR?

VaR depends on past data and a confidence level, so it can miss rare, extreme events. Stress testing applies severe but plausible shocks directly. It shows whether the bank could survive events outside its historical experience.

What is reverse stress testing?

It starts from an outcome the bank wants to avoid, such as failing a capital requirement. It then works backwards to find the combination of events that could cause it. This helps reveal hidden vulnerabilities.

Is expected shortfall used in Basel norms?

Yes. The Basel market risk framework for internal models uses expected shortfall in place of VaR for measuring market risk capital. The reason is that ES captures tail risk better.