Scenario Analysis

The discipline of the technique lies in coherence: within a single scenario, every input must be consistent with the same underlying narrative. An adverse case built around a demand shock cannot pair collapsing output with rising policy rates and tightening credit spreads unless that combination is economically defensible. Analysts typically begin with a small set of drivers: growth, unemployment, policy rates, credit spreads, equity levels, exchange rates, commodity prices, property values: then propagate those drivers through the portfolio or financial model to revalue positions, re-estimate default probabilities and recompute earnings.

Scenarios are sourced in two ways. Historical scenarios replay an observed episode, applying the market moves of a past crisis to today's holdings. Hypothetical scenarios are constructed, allowing exposure to risks with no precedent in the data sample. A third variant, reverse stress testing, inverts the exercise: start with a defined failure outcome and work backwards to identify the combinations of conditions that would produce it.

Where subjective probabilities can be attached to each state, the scenario set becomes a coarse discrete distribution. The expected outcome is the sum, across scenarios, of each scenario's probability multiplied by its result; the variance is the probability-weighted sum of squared deviations from that expected outcome, and its square root gives a standard deviation. In capital budgeting this converts a base, best and worst NPV into an expected NPV and a dispersion measure that can be compared across competing projects.

Where probabilities are not assigned, the output is read directly: the loss under the severe case, the resulting capital ratio, the funded status of a pension plan, the maximum drawdown a mandate would sustain. This is often the more useful reading, because the point of the exercise is not a central estimate but whether the institution survives the tail and remains above its regulatory or policy constraints.

Bank supervision is the most visible application, but the method is embedded far more widely. Expected credit loss accounting under IFRS 9 and CECL requires lenders to weight loss estimates across multiple forward-looking macroeconomic scenarios, so provisions reported on the income statement are literally a probability-weighted scenario output. Insurers run own risk and solvency assessments on analogous lines, and pension trustees test funding ratios against combined shocks to discount rates, asset returns and longevity.

On the investment side, scenario analysis complements value at risk. VaR quantifies a loss threshold at a stated confidence level under prevailing statistical relationships; it says little about behaviour when correlations break, liquidity evaporates and hedges fail simultaneously. A liquidity-crisis scenario applied to a leveraged multi-asset book captures exactly those joint effects, which is why risk committees typically review both.

At Level I the expectations are practical. Given a table of scenarios with probabilities and outcomes, compute the expected value and the standard deviation across scenarios, and interpret the spread as a measure of project or portfolio risk. Explain why shifting several correlated variables together yields information that single-variable perturbation cannot, and identify which technique a described exercise represents.

Equally important is stating the limitations. Results depend entirely on the scenarios chosen, so risks not imagined are not measured. Probabilities attached to hypothetical states are judgmental and rarely verifiable. A small scenario set gives no view of the density of outcomes between the cases examined. These constraints explain why scenario analysis is presented as one component of a risk framework alongside simulation, sensitivity testing and position limits, not as a standalone risk measure.