Tutorial: Stress testing a portfolio against a 2008-style crash

In this tutorial we’ll replay a 2008-style market crash against a small multi-asset portfolio and see:

• How to create a scenario that moves equities, currencies, interest rates and volatility at the same time.

• Why the combined crash costs more than its parts added together.

• How to work out how much worse a crash would have to be to lose a particular amount.

The sample portfolio

Consider the following USD portfolio with a mix of assets chosen so that each one reacts to the crash in a different way.

On 1 October 2026, the portfolio is worth 2,349,065 USD, GBP/USD is 1.30, USD interest rates are a flat 4%, and Apple’s volatility is a flat 25%.

Creating the crash scenario

1. Navigate to Portfolio Management > Scenarios and click Create Scenario.

2. On the Basic Details tab, enter a Scope and Code, choose a Scenario type of Historical, and enter a Display name such as 2008-style market crash.

3. On the Shifts tab, click Add shift six times and fill in each shift as follows:

Note the following:

• Shift 1 uses a filter, so it reaches every equity without having to list them. Shifts 2 and 3 name an instrument directly. When a shift names an instrument and a filter-based shift hits the same price, LUSID applies only the one that names the instrument. So JPMorgan and HSBC fall 30%, not 20% and then a further 30%.

• The FX, rate and volatility shifts don’t name any holdings at all. They move market data, and every holding valued from that market data moves with it.

• Volatilities are decimals, so 0.05 means 5 volatility points. An amount of 5 would add 500 points.

• The web app can’t set a short code, so LUSID generates one from the display name. More information.

Valuing the portfolio under the crash

On the Portfolio Management > Valuations dashboard, choose 2008-style market crash from the Scenarios picker. More information.

LUSID revalues every holding under the shifted market and adds a column group for the scenario next to the base values. The portfolio falls from 2,349,065 USD to 1,853,879 USD, a loss of 495,187 USD, or 21.1%.

Each holding moves for a different reason:

Note the following:

• The option is the biggest surprise. A 20% fall in Apple’s shares cuts the option’s value by more than 60%, because an option magnifies moves in its underlying. Notice that no shift names the option: it moves because LUSID revalues it using the shifted Apple price, volatility and rates.

• The scenario shows PV in portfolio currency, so the HSBC and BP rows include the fall in sterling as well as the fall in their share prices.

• LUSID never changes your stored market data. Remove the scenario from the picker and the base values are exactly as before.

Comparing the crash with its parts

We could create separate scenarios for each of the four different shift types and then value the portfolio against all five scenarios (the aggregate 2008-style market crash and the four individual constituents:

The equity fall does most of the damage, and higher volatility on its own actually makes money, because it raises the value of the option. More interesting is the last two rows: the parts add up to a loss of 488,635 USD, but the crash loses 495,187 USD, which is 6,552 USD more. That difference is the result of the shifts interacting:

• The option loses 16,302 USD more than its parts suggest. On its own, a rise in volatility or rates adds value to the option. After Apple’s shares have fallen to 160, the option is further below its strike, so the same rises add much less.

• HSBC, BP and the GBP cash lose 9,750 USD less. On its own, the sterling fall applies to today’s share prices. In the crash, it applies to share prices that have already fallen, so it takes less off them.

This is why a scenario revalues every holding under all the shifts together, rather than adding up the effect of each shift. Adding up separate results, or separate risk metrics, misses interactions like these.

Seeing analytics under the crash

The Scenarios picker in the LUSID web app only shows PV in portfolio currency. You can also see many analytic metrics under a scenario, one column at a time. For example, to see how the crash changes the Treasury bond’s modified duration:

1. If your view doesn’t already include it, add the Modified Duration column.

2. Right-click its header, select Options · Scenario and choose 2008-style market crash. More information.

LUSID adds a scenario copy of the column next to the original. The bond’s modified duration falls from 7.09 to 7.00: at higher rates, the bond is slightly less sensitive to further rate moves. The other holdings have no modified duration, so their cells are blank in both columns:

Working out how bad a crash would have to be

So far we’ve asked what the crash costs. A risk committee often asks the opposite question: how bad would things have to get before we lost 1,000,000 USD? A reverse stress test answers this by scaling every shift in a scenario up or down until the portfolio loses your target amount. More information.

The only way to do this currently is calling the SolveReverseStress API with the portfolio, recipe, date, scenario and target loss:

curl -X POST 'https://<your-domain>.lusid.com/api/api/scenarios/$reversestress'
  -H 'Authorization: Bearer <your-API-access-token>'
  -H 'Content-Type: application/json-patch+json'
  -d '{
  "portfolioEntityIds": [{"scope": "ScenarioTutorial", "code": "CRASH-USD", "portfolioEntityType": "SinglePortfolio"}],
  "recipeId": {"scope": "ScenarioTutorial", "code": "Recipe"},
  "effectiveAt": "2026-10-01T00:00:00Z",
  "scenario": {"scope": "ScenarioTutorial", "code": "MarketCrash"},
  "targetPnl": -1000000,
  "metric": "Valuation/PvInPortfolioCcy",
  "maxScale": 3
}'

LUSID returns a scale of about 2.33. The crash would have to be 2.33 times as severe for the portfolio to lose 1,000,000 USD:

Note the following:

• The request uses Valuation/PvInPortfolioCcy because the portfolio holds GBP as well as USD assets. The default metric is in each holding’s own currency, which LUSID can’t add up.

• maxScale limits how far LUSID scales the crash. At the default of 5, bank shares would fall 150%, which is impossible, and the request would fail. At 3, they fall at most 90%.

• The response also includes a ladder showing the loss at each scale LUSID tried, from 311,580 USD at 0.6 times the crash to 1,207,999 USD at 3 times.