Stress Testing and Emerging Risk

Stress testing asks "what happens to our balance sheet if X occurs?" while emerging-risk analysis asks "what plausible X's have we not yet fully priced or reserved for?" Together they form the forward-looking complement to backward-looking reserving and pricing, and are a core ORSA deliverable.

Scenario design

A well-designed stress scenario is: (1) plausible — grounded in historical precedent or a coherent causal narrative, not an arbitrary shock; (2) internally consistent — a recession scenario should move unemployment, interest rates, asset values, and claim frequency/severity together in directions consistent with actual recessions, not just one variable in isolation; (3) calibrated to a return period or narrative severity (e.g., "a 1-in-100-year hurricane season" or "a repeat of 1979–1981 medical cost inflation"); and (4) actionable, i.e., tied to specific management or capital responses (reinsurance purchase, rate action, capital raise) rather than an academic exercise.

Scenarios are typically built from historical replay (rerun a past event's shocks against today's book), hypothetical/what-if construction (combine expert-elicited shocks to key variables), or statistical extreme scenarios (tail draws from a fitted joint distribution, see Dependence and Copulas).

Reverse stress testing

Rather than asking "what is the impact of scenario X," reverse stress testing starts from an unacceptable outcome — e.g., insolvency, breach of a rating-agency capital threshold, or breach of a regulatory action level — and works backward to identify the combination of events that would produce it. This surfaces vulnerabilities that forward stress tests (built from an analyst's list of plausible shocks) may never think to test, because the starting point is the failure mode itself rather than a specific cause. A reverse stress test on a long-tail casualty writer, for instance, might reveal that a moderate combination of social inflation and a modest increase in interest rates (impairing both reserve adequacy and asset values simultaneously) is sufficient to breach a rating threshold — a combination not obviously "extreme" on either dimension alone.

Emerging risk survey

Medical and social inflation. Medical cost trend (utilization, unit cost, new technology) compounds directly into WC and liability medical claim costs. Social inflation is the broader (and harder to quantify) trend of rising liability severity driven by litigation funding, expanding tort theories, larger jury awards ("nuclear verdicts"), and eroding claims-handling advantages — it behaves like a trend but is driven by legal/social rather than purely economic forces, and is notoriously difficult to trend reliably from historical data because it can shift discontinuously.

Legislative change. Statutory changes — WC benefit level reform, tort reform (or its reversal), assignment-of-benefits legislation, presumption laws (e.g., COVID-19 occupational presumptions for first responders) — can retroactively change the cost of existing claims (not just prospective business), a risk reserves must anticipate through legislative-change loads distinct from pure loss-development uncertainty.

Recession effects on WC frequency. Historically, WC claim frequency tends to fall during recessions (fewer new/inexperienced workers, reduced overtime and hours, "discretionary" claims suppressed by job insecurity) and rise coming out of a recession as hiring resumes with less-experienced workers — a well-documented pro-cyclical pattern that argues for adjusting frequency trend selections around economic turning points rather than assuming a constant secular frequency trend.

Cyber. Aggregation risk (a single vulnerability affecting many insureds simultaneously — e.g., a widely-used software supply-chain compromise) makes cyber one of the least diversifiable modern exposures; silent cyber (unintended cyber coverage embedded in traditional property/casualty wordings) is a related emerging-risk concern requiring explicit exclusion or affirmative pricing.

Pandemic. COVID-19 demonstrated correlated shocks across underwriting (business interruption, event cancellation, WC presumption claims), reserving (claims handling delays distorting development patterns), and the asset side (simultaneous equity and credit spread shocks) — the textbook case for why stress scenarios must be constructed to move all balance sheet drivers together rather than testing underwriting and investment risk in isolation.

Worked example: medical-inflation shock on a WC reserve

A WC book carries a held case reserve for open medical-only-plus-indemnity claims of 50,000,000 at current cost levels, with average remaining claim duration of 4 years and base medical trend of 6% per year already embedded in the reserve. Management wants to stress-test a scenario where medical trend accelerates to 9% per year for the remaining claim duration (a plausible "return to early-2000s medical inflation" scenario).

The incremental annual trend differential is $9\% - 6\% = 3\%$, compounding over the 4-year remaining duration. Applying the differential compounding to the base reserve (a standard simplification for a stress load, treating the existing 6% as already reflected):

$R_{\text{stressed}} = R_{\text{base}} \times (1.03)^{4} = 50{,}000{,}000 \times 1.1255 = 56{,}275{,}000.$

Incremental reserve strengthening required = 56,275,000 − 50,000,000 = 6,275,000, roughly a 12.6% increase in the affected reserve segment.

If this book represents 40% of total company reserves of 200,000,000, and surplus is 60,000,000, the stress reduces surplus by 6,275,000/60,000,000 = 10.5% — a material but survivable hit in isolation. Combining this with a correlated stress (e.g., a simultaneous 100bp rise in interest rates reducing bond portfolio value by, say, 4% of a 150,000,000 bond portfolio = 6,000,000) more than doubles the surplus impact to roughly 12,275,000, or 20.5% of surplus — illustrating why correlated, multi-driver scenario construction matters more than testing each driver alone.

Pitfalls

• Stressing severity/trend in isolation from correlated frequency, asset, and reinsurance-recoverable effects.
• Treating social inflation as a smooth, extrapolable trend rather than a regime-shift risk requiring scenario (not just point-estimate) treatment.
• Failing to reverse stress test — only running management's preferred "plausible" scenarios rather than searching for the scenario that breaks the company.
• Ignoring that reinsurance recoverables themselves are exposed to the same stress (counterparty credit risk under a correlated market-wide event).

Exam relevance

Stress testing, ORSA, and emerging risk are covered on CAS Exam 9 and appear throughout SOA's ERM curriculum.

Further reading

• NAIC, ORSA Guidance Manual
• CAS Committee on Reinsurance Research, various emerging-risk papers
• Swiss Re sigma series on social inflation and litigation funding