Quantitative Optimization of Insurance Structures : A Data-Driven Case Study in Risk Management

Sammanfattning

Corporate insurance purchasing is commonly evaluated through historical loss experience, marketbenchmarking, and line-of-business specific assessments. However, for large industrial corpora-tions, insurance decisions can also be formulated as a quantitative portfolio optimization problem,where external premium costs are balanced against retained loss distributions and the firm’s in-ternal cost of capital. This thesis develops and applies a stochastic combinatorial optimizationframework for evaluating alternative corporate insurance structures across Property Damage andBusiness Interruption (PDBI), General and Products Liability (GLPL), and Marine transit. The model is based on Monte Carlo-simulated retained loss distributions corresponding to theevaluated insurance alternatives for each line of business. Since the available simulation filescontained different numbers of iterations, the distributions are harmonized to a common lengthof 10,000 stochastic scenarios to support portfolio-level aggregation. Each candidate portfolio isevaluated using the Economic Cost of Risk (ECoR), defined as the sum of external premium cost,expected retained loss, and a Conditional Value at Risk (CVaR)-based capital charge. Feasibilityis determined by operationalizing the company’s one-off impact bands as Value-at-Risk (VaR)constraints at the 50%, 75%, and 99% confidence levels. The optimization evaluates 64 discrete portfolio configurations over 10,000 harmonized stochasticscenarios. The ECoR-minimizing feasible portfolio is identified as GLPL: Current, PDBI: Option 2,and Marine: Expiring. Compared with the current framework, the optimized portfolio reducestotal ECoR from 25.57 MEUR to 24.46 MEUR, an improvement of approximately 1.11 MEURdriven primarily by lower external premium expenditure. The optimized portfolio has a VaR99% of34.95 MEUR against a 125 MEUR threshold and a CVaR99% of 78.86 MEUR, with both measuresremaining effectively unchanged relative to the current framework. The results show that the proposed framework can identify economically more efficient insuranceconfigurations within a finite set of available market alternatives. Diversification benefits are foundto be limited, reflecting concentration of tail risk in the dominant industrial exposure. The thesisdemonstrates how actuarial risk measures, stochastic simulation, and corporate finance principlescan be integrated into a transparent decision-support tool for corporate risk financing.