stress and scenario testing

Engineers do not just calculate that a bridge should hold; they load test it — pile on trucks, shake it, imagine the worst storm — to see what actually happens before real life finds out for them. Stress and scenario testing is the same instinct applied to an insurer's balance sheet. Rather than relying only on a model's neat average answer, the firm deliberately imagines bad worlds and recomputes its finances inside them, asking: 'if this happened, would we still be standing — and by how much?'

It comes in a few flavours. A stress test moves a single risk driver sharply — say, equities fall 40%, or interest rates drop 1%, or mortality worsens 20% — and measures the hit. A scenario test combines several stresses into a coherent story — a recession in which markets fall, defaults rise, and lapses spike all at once — which is more realistic because real crises are bundles, not single shocks. A reverse stress test runs the question backwards: instead of asking 'how bad is scenario X?', it asks 'what scenario would actually break us?' and works out the combination of events that would exhaust the firm's capital. Crucially, these tests do not need a probability attached; their value is exploring severity and identifying vulnerabilities, complementing the probabilistic capital models rather than replacing them.

These tests matter because models are built on past data and assumed relationships, and the events that kill insurers are usually the ones outside that history — correlations that 'were never supposed to' all go to one. Stress and scenario testing is a required part of the ORSA and of supervisory review precisely because it stress-tests the imagination, not just the spreadsheet. The honest caveat: a test is only as good as the scenarios chosen, and there is always a temptation to pick stresses the firm can comfortably survive. The most useful scenario is usually the uncomfortable one nobody wants to run.