Asset and Liability Management: Optimization using Least-Squares Monte Carlo

Sammanfattning: This thesis aims to examine an efficient asset and liability management method under Solvency II regulations, and to find an optimization framework that takes complex interactions between assets and liabilities into account. The investigated approach consists of a least-squares Monte Carlo method, where least-squares regression is used to obtain a proxy function for future net asset values. A fairly close approximation is achieved, and the computational burden is significantly reduced compared to a traditional full nested Monte Carlo simulation method. By allocating capital into several asset classes with different risk attributes, the effects on the risk adjusted net asset value are studied when moving from low to high risk assets. Restrictions on risk and asset return are introduced, and an optimal allocation compatible with the Solvency capital requirement is obtained. For comparison, a similar study is conducted using a mean-variance optimization approach