Stress Testing, Scenario Analysis and ORSA

Introduction

Insurance solvency is not established merely because an insurer satisfies a capital requirement on the reporting date. Insurers are exposed to events that may develop over several years: inflation, falling interest rates, catastrophe accumulation, cyber disruption, litigation, adverse mortality, policyholder lapse, and deterioration in the quality of assets. A sound solvency framework therefore asks two related questions:

1. What would happen to the insurer’s financial position under adverse but plausible conditions?
2. What decisions would management take if the insurer moved outside its risk appetite or capital plan?

Stress testing and scenario analysis answer the first question. The Own Risk and Solvency Assessment (ORSA) connects the answer to governance, strategy, risk appetite, business planning and management action. ORSA is not simply a regulatory report; it is a forward-looking management process that evaluates whether an insurer’s risk profile and capital resources remain appropriate over its planning horizon.

This chapter develops sensitivity analysis, deterministic and stochastic scenarios, reverse stress testing, multi-year solvency projections and ORSA integration. Numerical examples use simplified values for teaching. Actual implementation must reflect the applicable jurisdiction, valuation basis, tax rules, accounting standards and regulatory capital regime.

Learning objectives

By the end of this chapter, the reader should be able to:

• distinguish sensitivity analysis, stress testing, scenario analysis and reverse stress testing;
• design credible single-factor, multi-factor and narrative scenarios;
• project own funds, capital requirements and solvency ratios over several years;
• interpret management actions and their timing, credibility and implementation constraints;
• explain the purpose, governance and contents of an ORSA;
• evaluate model, data, behavioural and regulatory limitations; and
• use stress results to support risk appetite, business planning and board decisions.

1. The conceptual architecture

1.1 Sensitivity analysis and stress testing

Sensitivity analysis changes one input, or a small group of inputs, while holding other assumptions constant. For example, an insurer may measure the effect of a 100-basis-point increase in interest rates on the value of its bond portfolio. It is useful for understanding local drivers and marginal exposures, but it can be unrealistic when risks move together.

Stress testing applies a severe change or set of changes to assess resilience. A stress may be historical, hypothetical or prescribed by a supervisor. Scenario analysis generally describes a coherent economic or business story in which several variables change consistently. The terms are sometimes used interchangeably, but the distinction is analytically useful.

1.2 Solvency measures

A simplified solvency ratio is:

where OF_t is eligible own funds at time t, and RC_t is the applicable capital requirement at that time. In many regimes the denominator may be called a solvency capital requirement, risk-based capital requirement or prescribed capital amount. The ratio is meaningful only when numerator and denominator use compatible valuation, eligibility and aggregation rules.

A simplified surplus is:

A stress is material when it causes a significant reduction in surplus, breaches a risk-appetite limit, approaches a regulatory threshold, or creates an unacceptable liquidity or operational position. Solvency is therefore multidimensional: capital adequacy does not guarantee liquidity, operational continuity or policyholder service.

2. Designing useful stress tests

2.1 Define the decision and scope

A test should begin with a decision, not with a convenient dataset. The analyst should specify:

• the portfolio, legal entity and consolidation scope;
• the valuation date and projection horizon;
• the base case and valuation basis;
• the risk measures to be monitored;
• whether the test is gross or net of reinsurance and tax;
• available management actions; and
• the decisions that the result may trigger.

For example, a board may ask whether the insurer can maintain its dividend policy after a catastrophe and a period of high inflation. That question requires more than a one-day asset shock: it requires a multi-period projection of claims, expenses, reinsurance recoveries, capital, liquidity and management response.

2.2 Build a causal risk map

Risk identification should link risk drivers to financial and non-financial consequences. A catastrophe may increase claims, receivables from reinsurers, liquidity demands, operational workload and reputational risk. Inflation may increase claims severity, repair costs, wage expenses and reserve uncertainty. A lapse shock may reduce future profits while increasing the proportion of guarantees remaining in force.

A causal map helps prevent double counting. For instance, an interest-rate shock may already affect both asset values and technical provisions. Applying separate unconnected shocks to each may overstate or understate the economic effect.

2.3 Select scenario types

Historical scenarios replay an event such as a market crash, major hurricane season or pandemic period. They are transparent but may not represent current exposures, inflation, portfolio composition or emerging risks.

Hypothetical scenarios combine plausible assumptions, such as a cyber incident, rating downgrade and lapse increase. They are flexible but require judgement.

Regulatory scenarios follow prescribed parameters or templates. They support comparability but may not capture the insurer’s distinctive risk profile.

Reverse stress scenarios begin with a failure condition and work backwards to identify the combination of events that produces it.

2.4 Correlation and dependence

Risk factors should not be combined using arbitrary correlations. Dependence may be nonlinear and may change in the tail. For example, policy lapses can rise when customers face financial pressure, while asset values fall and replacement costs increase. Reinsurance recoveries may also become less reliable when a major event affects many cedants simultaneously.

A scenario need not estimate a complete probability distribution, but it should explain why assumptions move together. Where probabilities are used, the analyst should distinguish empirical evidence, expert judgement and model-implied dependence.

3. Quantitative techniques

3.1 Balance-sheet projection

A simplified period-end balance-sheet identity is:

where A_t denotes assets, L_t liabilities and OF_t own funds. In a projection, changes in own funds can be represented as:

where P_t+1 is recognised profit after tax, D_t+1 is dividends or distributions, K_t+1 is new external capital, and V_t+1 represents valuation movements not already included in profit. The equation is a planning identity, not a substitute for a full regulatory balance-sheet calculation. All terms must be measured on a consistent basis.

A simplified profit bridge may be written as:

The symbols denote premiums, investment income and gains, claims and reserve movements, expenses, reinsurance costs, reinsurance recoveries and tax. In practice, premium revenue, incurred claims, discounting, acquisition costs and reserve releases require accounting and actuarial definitions; they cannot be substituted casually into this bridge.

3.2 Discounting and interest-rate sensitivity

For a set of deterministic cash flows, present value is:

where CF_t is the cash flow at time t, and i_t is the appropriate annual effective discount rate for that maturity. If a single rate i is used, the expression becomes PV = ∑ CF_t/(1+i)^t. Cash-flow timing, compounding convention and currency must be consistent.

A first-order duration approximation for a small parallel yield change is:

where D is modified duration in years and ∆y is the decimal change in yield. This approximation is local and excludes convexity. For large shocks, nonparallel curves, embedded options or negative-rate effects, full revaluation is preferable.

3.3 Capital requirement projection

The future capital requirement should be projected rather than held constant without justification. A stylised risk-based model may use:

where C_i,t is the capital charge for risk module i, and ρ_ij is the assumed dependency parameter. This formula is applicable only when the model’s aggregation assumptions justify a covariance-style approach. The resulting figure is not automatically a regulatory capital requirement; regulatory rules may impose floors, adjustments, stress calibrations and eligibility constraints.

3.4 Scenario severity and probability

A stress test is not necessarily a forecast. A severe scenario may have a low or unassigned probability but still be valuable for resilience analysis. Conversely, a high-probability adverse trend may be strategically more important than a remote catastrophe. Reports should therefore present severity, plausibility, time horizon and confidence separately rather than implying false precision.

4. Reverse stress testing

Reverse stress testing starts with a specified failure state. Examples include:

• eligible own funds falling below the regulatory minimum;
• the solvency ratio falling below the board-approved limit;
• inability to meet claims and collateral payments when due;
• loss of a critical licence, rating or outsourcing capability; or
• a breach of a debt covenant.

The analyst then searches for the smallest, most plausible combination of adverse developments that reaches the failure state. A useful formal statement is:

Here x is a vector of stress factors, g(x) measures distance from the failure boundary, and C(x) is a severity or plausibility cost function. There is no universal definition of C(x); it may be based on standardised deviations, expert scoring or scenario likelihood. The optimisation is therefore a decision-support device, not an objective truth.

Reverse stress testing can reveal concentration risk, dependence assumptions and delayed management responses. It should not be used to claim that the identified path is the only way failure can occur. Multiple paths may reach the same boundary.

5. Multi-year solvency projections

5.1 Projection structure

A robust projection normally contains:

1. opening balance sheet and own funds;
2. business volumes, premiums, expenses and claims;
3. asset allocation, investment returns and reinvestment;
4. technical provisions and discount curves;
5. reinsurance terms, recoveries and counterparty default;
6. tax, dividends and capital movements;
7. capital requirements by risk module;
8. liquidity and collateral requirements; and
9. management actions and their implementation dates.

The base case should be consistent with the approved business plan. Scenarios should alter drivers transparently rather than quietly changing outputs.

5.2 Worked example: three-year projection

Illustrative assumptions. An insurer begins with own funds of 120 and a capital requirement of 80, both in monetary units. The opening solvency ratio is therefore:

Assume the base case has the following projected profit, distributions, capital requirement and external capital:

Using OF_t+1=OF_t+Profit−Distribution, own funds are 126 at the end of year 1, 133 at the end of year 2 and 140 at the end of year 3. The projected ratios are:

Now impose an illustrative adverse scenario: profit is reduced by 25 in year 1, 18 in year 2 and 10 in year 3; the year-1 capital requirement rises by 20%, and subsequent requirements are 10% above base. No management action is initially assumed.

The arithmetic illustrates an important result: the year-1 ratio remains marginally above 100%, but the continuing deterioration creates a later breach and may violate a board buffer much earlier. If the board limit is 120%, the insurer breaches that limit in year 1 despite not breaching the regulatory threshold.

Suppose management can raise 25 of capital at the end of year 2 and suspend the year-3 distribution of 5. Year-2 own funds become 115, producing 115 ÷ 96.8 × 100% ≈ 118.8%. At year 3, own funds become 120 rather than 87, and the ratio is approximately 120 ÷ 101.2 × 100% ≈ 118.6%. The actions avoid the regulatory breach in this simplified calculation but do not restore the 120% board limit. The result demonstrates why an ORSA should distinguish regulatory minimums, internal appetite limits and management triggers.

6. Management actions and credibility

Management actions may include repricing, underwriting restrictions, claims settlement changes that remain fair and lawful, reinsurance purchases, asset reallocation, hedging, expense reduction, dividend suspension, capital raising and disposal of non-core business. A model should record:

• the trigger and decision authority;
• the earliest implementation date;
• financial benefit and implementation cost;
• operational capacity and legal constraints;
• policyholder, conduct and reputational effects;
• market capacity and counterparty availability; and
• whether the action is already reflected in the base plan.

Actions must not be counted twice. An action assumed in the base case cannot also be claimed as incremental mitigation. Actions that require board approval, regulatory consent, new funding or market capacity should be time-lagged and assigned a confidence level.

7. ORSA as an integrated management process

7.1 Purpose and principles

ORSA is the insurer’s own forward-looking assessment of its risks, capital needs and solvency position. Although detailed requirements differ by jurisdiction, effective ORSA commonly demonstrates that the insurer:

• understands its material risks and risk profile;
• assesses whether capital is adequate over its business horizon;
• considers stress and reverse stress outcomes;
• links risk appetite to strategy and planning;
• identifies credible management actions; and
• reports conclusions to the board and relevant supervisor.

ORSA should be proportionate. A small specialist insurer does not need the same modelling infrastructure as a multinational group, but it must still identify material risks and support its conclusions with reliable evidence.

7.2 Risk appetite linkage

Risk appetite states the amount and type of risk the insurer is willing to accept in pursuit of objectives. It may include solvency ratio limits, liquidity survival days, concentration limits, reinsurance credit exposure, earnings volatility and operational tolerance. A useful hierarchy is:

• risk capacity: the maximum risk the insurer can bear before legal, financial or operational failure;
• risk appetite: the risk level the board is willing to accept;
• risk limits: measurable constraints allocated to businesses and risk types; and
• triggers: early-warning levels that prompt investigation or action.

Stress testing translates abstract appetite into consequences. If a scenario breaches a trigger, the ORSA should explain the escalation process, not merely display a red indicator.

7.3 ORSA cycle

The ORSA cycle normally follows the sequence:

1. strategy and business plan;
2. risk identification and materiality assessment;
3. base-case capital projection;
4. stress, scenario and reverse stress analysis;
5. evaluation of risk appetite and management actions;
6. board challenge and approval;
7. supervisory communication where required; and
8. monitoring, event-driven refresh and lessons learned.

An ORSA document is the output of this cycle. It should not be prepared as an isolated annual compliance exercise.

7.4 Governance and the three lines

The first line owns risks and implements controls. The second line, including risk management and actuarial functions where applicable, challenges assumptions, coordinates the ORSA and monitors limits. The third line, internal audit, independently evaluates governance, controls and the reliability of the process. The board retains responsibility for oversight and must challenge optimistic assumptions, weak data and unsubstantiated management actions.

The actuarial, finance, investment, underwriting, reinsurance, liquidity, compliance and operations functions should contribute. A technically sophisticated model can still produce a poor ORSA if material operational or conduct risks are excluded.

8. Case study: illustrative Meridian General Insurance

This is a fictional illustrative case. Meridian General writes commercial property and motor insurance. It has strong growth in catastrophe-exposed regions, a high allocation to corporate bonds and a board solvency appetite of 130%, compared with a regulatory minimum of 100%.

Background and problem

The business plan assumes premium growth of 8% annually, stable reinsurance capacity and moderate claims inflation. Risk management identifies a plausible compound scenario: a severe windstorm, two years of claims inflation above plan, widening credit spreads and a 12% increase in policy lapses among financially stressed customers. The board asks whether growth should continue and whether the dividend policy is sustainable.

Analysis

The team constructs a three-year projection. It revalues assets and liabilities, increases claims severity, models reinsurance recoveries with a counterparty delay, reduces investment income and raises the capital requirement for catastrophe and spread risk. It also tests a reverse scenario in which the solvency ratio falls below 100%. The reverse analysis identifies catastrophe concentration and delayed repricing as the shortest path to failure, rather than the bond portfolio alone.

Approach and solution

The ORSA recommends four actions: restrict new business in the most concentrated territories; purchase additional event-limit reinsurance subject to market availability; suspend dividends if the ratio falls below 130%; and establish a quarterly claims-inflation trigger. The investment committee is required to reduce issuer concentration gradually, while liquidity buffers are increased to meet collateral and claims payments.

The board rejects an assumption that management can immediately sell illiquid assets at book value. It also requires a separate operational scenario involving claims-system outage, because the catastrophe scenario increases operational workload at the same time as liquidity pressure.

Outcome

Under the approved plan, the base case remains above the appetite limit. Under the compound scenario, the ratio falls below 130% in year 1 but remains above the regulatory minimum after the actions. The ORSA therefore concludes that the plan is viable only with controlled growth, contingent reinsurance and dividend flexibility. It does not describe the business as risk-free; instead, it identifies the conditions under which the strategy must change.

Learning points

The case shows that ORSA adds value when it changes decisions. It also demonstrates that capital, liquidity, underwriting, investment and operational resilience should be analysed together. A single solvency ratio cannot reveal whether a proposed mitigation is timely, executable or fair to policyholders.

9. Limitations, data and model risk

Stress results are conditional estimates. Important limitations include:

• incomplete or inconsistent historical data;
• parameter uncertainty and regime change;
• nonlinear asset and liability responses;
• uncertain policyholder and management behaviour;
• insufficient modelling of reinsurance disputes or counterparty default;
• accounting and regulatory basis differences;
• aggregation and diversification assumptions;
• model error, coding error and inappropriate granularity; and
• false confidence created by precise-looking outputs.

Validation should include independent review, reconciliation to financial statements, back-testing where possible, sensitivity to key assumptions, benchmark comparisons, code controls and documented expert judgement. Data lineage should show the source, transformation, owner and approval of each material assumption.

Climate change, cyber risk, pandemic effects and geopolitical disruption often lack stable historical samples. They should be addressed through scenario design, expert elicitation and qualitative analysis as well as quantitative models. Absence of data is not evidence of absence of risk.

10. Technology, AI and contemporary practice

Modern insurers use automated data pipelines, actuarial projection platforms, cloud computing and dashboards to update scenarios rapidly. Machine learning may assist claims severity analysis, anomaly detection or scenario generation, but it does not remove the need for causal reasoning and governance. An algorithm trained on benign historical conditions may understate a novel tail event.

AI-supported ORSA tools should have documented data provenance, version control, access controls, explainability appropriate to the decision, human review and monitoring for drift and bias. Generative AI may help draft scenario narratives or identify missing dependencies, but final assumptions and board conclusions require accountable professional judgement. Sensitive policyholder and commercially confidential information must be protected.

Contemporary best practice increasingly combines deterministic stresses with stochastic distributions, climate and nature-risk scenarios, liquidity stress testing, operational resilience testing and dynamic management-action modelling. The objective is not maximal complexity; it is decision-relevant insight that is proportionate, reproducible and challengeable.

Best-practice checklist

A high-quality stress-testing and ORSA process should:

• start with strategic decisions and material risks;
• use a clearly documented base case;
• combine sensitivities, coherent scenarios and reverse stresses;
• distinguish regulatory capital from internal capital needs;
• project both capital and liquidity over time;
• model dependencies and avoid double counting;
• test the credibility and timing of management actions;
• include operational, conduct, climate and emerging risks where material;
• obtain independent challenge and board approval; and
• record limitations, uncertainty and actions arising from the results.

Chapter Summary

Stress testing examines resilience under adverse conditions, while sensitivity analysis isolates the effect of individual drivers. Scenario analysis combines drivers into a coherent narrative, and reverse stress testing begins with a failure condition to identify vulnerable pathways. Multi-year projections connect underwriting, claims, expenses, investments, technical provisions, reinsurance, tax, distributions and capital requirements.

A solvency ratio is useful only when own funds and capital requirements are measured on a consistent basis. Management actions must be realistic, time-lagged and free from double counting. ORSA integrates these analyses with strategy, risk appetite, business planning, governance and board decisions. Effective ORSA is forward-looking, proportionate and decision-oriented. Its limitations—model risk, data uncertainty, behavioural change and emerging risks—must be made explicit. Technology can improve speed and consistency, but professional judgement, validation and accountability remain essential.

Key Terms

Sensitivity analysis: Assessment of the effect of changing one or a limited number of assumptions.

Stress testing: Evaluation of financial resilience under severe adverse conditions.

Scenario analysis: Assessment based on a coherent set of interacting assumptions.

Reverse stress testing: Analysis that begins with a failure outcome and works backwards to identify causes.

Own funds: Capital resources available to absorb losses under the relevant valuation and eligibility rules.

Capital requirement: Amount of capital required under a specified regulatory or internal framework.

Solvency ratio: Own funds divided by the applicable capital requirement.

Risk appetite: The amount and type of risk the board is willing to accept.

Risk capacity: Maximum risk that can be borne before unacceptable failure or constraint.

Risk limit: A measurable restriction allocated to a risk, business or portfolio.

Trigger: An early-warning level prompting investigation or action.

Management action: A decision intended to alter risk, capital, liquidity or performance.

ORSA: Own Risk and Solvency Assessment, an insurer’s forward-looking assessment of risk and solvency.

Technical provisions: Liabilities representing the insurer’s obligations to policyholders and claimants.

Duration: First-order sensitivity of a present value to a change in yield, expressed approximately in years.

Dependence: The way risks move together, including nonlinear tail relationships.

Model risk: Risk of error or inappropriate use arising from a model, data or implementation.

Liquidity stress: Analysis of the ability to meet cash and collateral obligations under adverse conditions.

Reverse failure boundary: The defined condition, such as a capital or liquidity breach, used in reverse stress testing.

Review Questions

Short Answer Questions

1. Distinguish sensitivity analysis from scenario analysis.
2. Why should a solvency projection include management-action timing?
3. Define risk appetite and explain how it differs from risk capacity.
4. What is the purpose of reverse stress testing?
5. State two important limitations of stress-testing models.

Long Answer Questions

1. Explain how an insurer should design a coherent multi-factor stress scenario.
2. Discuss the relationship between ORSA, business planning and board governance.
3. Explain why capital adequacy and liquidity adequacy should be assessed together.
4. Describe how dependence and aggregation assumptions can affect stress-test results.
5. Evaluate the advantages and limitations of deterministic and stochastic projections.

Analytical / Application Questions

1. An insurer has own funds of 150 and a capital requirement of 100. A scenario reduces own funds by 35 and increases the requirement by 20%. Calculate the stressed solvency ratio and interpret it against internal limits of 130% and 100%.
2. Design a reverse stress test for an insurer whose failure condition is a solvency ratio below 100%. Identify at least four drivers, their possible interactions and two management actions.
3. A board wants to approve rapid premium growth because the base-case solvency ratio remains above 140%. Explain what additional ORSA analyses should be required before approval.

Multiple Choice Questions with Answers

Question 1

Which method begins with a defined failure condition and works backwards to identify adverse pathways?

A. Sensitivity analysis

B. Reverse stress testing

C. Base-case forecasting

D. Experience rating

Correct Answer: B. Reverse stress testing

Question 2

An insurer has own funds of 120 and a capital requirement of 80. Its solvency ratio is:

A. 매 66.7%

B. 100.0%

C. 150.0%

D. 200.0%

Correct Answer: C. 150.0%

Question 3

In the simplified identity OF_t+1=OF_t+P−D+K+V, D represents:

A. Discount rate

B. Diversification benefit

C. Dividends or distributions

D. Deferred claims

Correct Answer: C. Dividends or distributions

Question 4

The approximation ∆PV/PV≈−D×∆y is primarily a measure of:

A. Interest-rate sensitivity

B. Lapse frequency

C. Operational loss frequency

D. Reinsurance recovery

Correct Answer: A. Interest-rate sensitivity

Question 5

Which statement best describes an ORSA?

A. A historical claims register only

B. A forward-looking assessment linking risks, capital and strategy

C. A replacement for the audited financial statements

D. A guarantee that no losses will occur

Correct Answer: B. A forward-looking assessment linking risks, capital and strategy

Question 6

If own funds are 115 and the capital requirement is 96.8, the solvency ratio is approximately:

A. 84.2%

B. 96.8%

C. 118.8%

D. 211.8%

Correct Answer: C. 118.8%

Question 7

Which is an example of a management action?

A. A historical exchange rate

B. A regulatory reporting date

C. Suspending dividends after a defined trigger

D. The opening balance-sheet identity

Correct Answer: C. Suspending dividends after a defined trigger

Question 8

Why can a historical scenario be insufficient by itself?

A. Historical data can never be used in insurance

B. Past events may not reflect current exposures or emerging risks

C. Historical scenarios always have zero severity

D. Historical scenarios cannot include market variables

Correct Answer: B. Past events may not reflect current exposures or emerging risks

Question 9

What is a key governance responsibility of the board in ORSA?

A. Approve assumptions without challenge

B. Challenge the assessment and oversee alignment with strategy

C. Perform every model calculation personally

D. Eliminate all management discretion

Correct Answer: B. Challenge the assessment and oversee alignment with strategy

Question 10

A stress raises a capital requirement from 80 by 20%. The stressed requirement is:

A. 64

B. 84

C. 96

D. 100

Correct Answer: C. 96