What Does It Cost to Be Prudent? : Tail-Risk Optimisation and Constraint Analysis for a Swedish Insurance Group

Sammanfattning

Insurance companies invest assets to support policyholder obligations,maintain solvency, and generate adequate long-term returns. This createsa constrained strategic allocation problem in which investment-policyguidelines, duration rules, return requirements, and internal risk tolerancejointly define the feasible set of portfolios. This thesis studies how suchconstraints shape the attainable trade-off between expected return and tail riskfor an anonymised Swedish insurance group with two regulated entities: alife-insurance entity and a non-life insurance entity.The thesis develops a scenario-based asset-allocation model at asset-classlevel. Thirteen asset classes are represented by public proxy data in order topreserve confidentiality. Annual return scenarios are generated from assetlevel log-return models with Gaussian, Student’s 𝑡, and skewed Student’s𝑡 marginal innovations, combined through a Student’s 𝑡 copula dependencemodel. The annual log-return scenarios are converted to simple returns andused in a constrained Conditional Value-at-Risk optimisation. Value-at-Riskis then evaluated ex post as the governance risk measure, consistent with thecase company’s internal risk framework.The analysis identifies feasible mean–Conditional Value-at-Risk frontiersfor both entities. The target-return minimum-CVaR portfolios remainwithin the applicable Value-at-Risk limits, while higher-return frontier pointsprogressively exhaust the Value-at-Risk budget. The one-at-a-time constraintvalue analysis ranks policy constraints by the reduction in minimum CVaR atthe target-return portfolios. Under this measure, the alternatives limit is mostlocally binding for the life-insurance entity, while the duration constraint ismost locally binding for the non-life entity, followed by the alternatives limit.A comparison with a normal-distribution benchmark further shows that tailrisk estimates are sensitive to the distributional framework.The framework is intended as a decision-support complement to thecompany’s internal asset-liability management process. Its contribution isto make the trade-offs between expected return, tail risk, policy flexibility,and governance consistency explicit under a transparent and reproduciblemodelling structure.