How do you build a scenario that has never happened?
Historical replays have one unavoidable defect: they can only test crises that already occurred. The next serious event will have its own causes, its own sequence and its own set of assets that surprise everyone. Hypothetical scenario design is the attempt to prepare for that without pretending to predict it.
Start with factors, not headlines
A usable scenario is not a story about geopolitics. It is a set of numerical shocks to risk factors, because that is what a portfolio can actually be revalued against. A typical scenario specifies moves in:
- Equity indices — by region, sometimes by sector
- Government yields — by curve point, so parallel shifts and steepenings differ
- Credit spreads — investment grade and high yield separately
- Currencies — against the portfolio's base
- Commodities — energy in particular
- Implied volatility — which usually rises when everything else falls
A scenario that names only "equities −30%" is not a scenario. It's a single shock, and it will miss every position that isn't equity.
Internal consistency is the hard part
The discipline of hypothetical design is making the shocks hang together. Markets do not move in unrelated directions.
A stagflation scenario, for example, cannot have inflation surging and government yields falling and the currency strengthening and equities flat. Those pieces contradict each other. The craft is checking that every number in the set could plausibly coexist with every other number — usually by reasoning through the mechanism that would connect them.
A useful test: can you write the two-sentence causal story? If you cannot explain why these particular moves would occur together, the scenario is a list, not a scenario.
A worked hypothetical
Scenario: "Inflation re-acceleration." The story: inflation surprises upward, central banks tighten faster than markets expect, growth expectations fall, and there is no flight-to-quality bid for bonds because bonds are the source of the problem.
The shocks:
- Developed equity indices: −25%
- 10-year government yields: +150 basis points
- Investment-grade credit spreads: +120 bp; high yield: +400 bp
- Base currency: +8% versus a trade-weighted basket
- Oil: +40%
- Equity implied volatility: doubles
Now apply it to a $1,000,000 portfolio of 55% equities, 35% investment-grade bonds (duration 7), 10% high-yield:
- Equities: 55% × −25% = −$137,500
- IG bonds: price change ≈ −duration × yield move = −7 × 1.5% = −10.5%; plus spread widening ≈ −7 × 1.2% = −8.4%; combined ≈ −18.9% → 35% × −18.9% = −$66,150
- High yield: spread +400 bp on shorter duration, say ≈ −16% → 10% × −16% = −$16,000
- Total ≈ −$219,650, or −22.0%
The arithmetic is deliberately simple — duration times yield change, weight times move. Its value is that every assumption gets forced into the open where it can be argued with.
Design habits worth copying
Shock things that "can't" move together. The scenarios that hurt are the ones violating a diversification assumption you didn't know you'd made.
Include a liquidity leg. Add an assumption about how much wider bid-ask spreads become and how long an exit takes. Unit 4 explains why.
Run a severity ladder. The same scenario at mild, severe and extreme calibrations shows you where the portfolio's behaviour turns non-linear.
Write the story down next to the numbers. A scenario without its narrative gets misapplied within a year by someone who wasn't in the room.
Try it now
- Design a one-page scenario of your own with at least four factor shocks and a two-sentence causal story linking them.
- Sanity-check your shock sizes against history. The S&P 500 fund's full history, below, covers every equity drawdown of this century — Measure the worst move over a window the length of yours. For the rate leg, the US 10-year yield over the last five years is under it, a yield in per cent rather than a price: Measure its fastest rise over a window the length of yours. Has it moved as far as your shock, as fast? Adjust for plausibility, not for comfort.
- Apply it to the simple portfolio you defined earlier and compute the loss. Then note the single assumption that, if wrong, changes the answer most. Describing exposure, not prescribing action.