Please use this identifier to cite or link to this item: http://hdl.handle.net/1946/10557
In this thesis a modeling framework to aid Icelandic pension funds in their asset allocation decisions is introduced. The framework is based on stochastic programming and asset liability management methodologies, where emphasis is placed on the uncertainty in the liabilities of the pension funds as well as the current Icelandic regulatory and supervisory environment.
The modeling framework is based on three interacting components: modeling the underlying stochastic variables; a scenario generator; and a stochastic programming model which is formulated as a multistage recourse problem in discrete state space and is solved with a scenario-based technique. Computational experiments presented in this thesis are carried out for a hypothetical pension fund, where market valued liabilities and premiums are used and the current capital controls are not accounted for.
The credibility of the model is analyzed in terms of in-sample stability and analysis on spurious changes in the optimized asset allocation. This vital attribute has received little attention in similar studies. The model performance is measured by comparison with partly dynamic fixed-mix investment strategies in terms of stronger actuarial position at the horizon. In addition, analysis of the most influential constraints is carried out with sensitivity analysis.
The weakest component of the framework proved to be the scenario generator which was unable to generate a sufficient number of scenarios for satisfactory in-sample stability. In terms of performance and credibility, the model outperformed the partly dynamic fixed-mix benchmarks, in terms of stronger actuarial position at the horizon with no drastic changes in the optimal asset allocation. The most influential constraint is maximum allowed purchases in each year. Finally, comparison based on actuarial position, where liabilities and premiums are valued with fixed 3.5 percent interest rate, suggests that the interest rate risk is underestimated in the actuarial valuation methods currently used by the Icelandic Financial Supervisory Authority.