Asset Liability Management i pensionssektoren: En analyse af udvalgte ALM modeller

Martin Borberg Jensen

Student thesis: Master thesis

Abstract

The overall objective of a pension fund is to earn an adequate return on investments while maintaining a comfortable surplus of assets beyond liabilities. This is known as Asset Liability Management (ALM). To overcome the ALM requirements the pension fund can choose between various ALM models. The purpose of an ALM model is to support the pension funds long-term strategic decisions (e.g. investment and bonus strategies). However, such decisions are not straightforward due to the constantly changing financial markets. The model has to incorporate these uncertainties while incorporating the interdependencies between assets and liabilities in a way that the fund can base its strategic decisions on the model. This leads me to the main research objective that reads: In what extent can selected ALM models contribute to support the pension funds strategic decision making? I have analyzed four theoretical model approaches that can help the pension funds model and solve ALM. The approaches are: Deterministic optimization, deterministic simulation, stochastic optimization and stochastic simulation. The first approach is deterministic optimization which uses immunization and cash flow matching. These models have their strength in the setup, as they are relative simple and the asset portfolio is pure bonds. The models focus on minimizing the interest rate risk between the assets and liabilities. The second approach is deterministic simulation which is often used for stress testing where the main focus is on down-side risk. This approach is also used in the standard model for the upcoming Solvency II, where Value-at-Risk plays a big factor. The third approach is stochastic optimization where I focus on the mean-variance optimization and the dynamic-stochastic optimization. The last approach is stochastic simulation which consists of two models to simulate the outcome of the assets; a vector autoregressive model and stochastic differential equations. Based on my theoretical analysis I found that the stochastic simulation approach is the best way to model ALM. The approach is very flexible and is capable of incorporating almost any interdependency between assets and liabilities that the pension fund may wish to analyze.

EducationsMSc in Mathematics , (Graduate Programme) Final Thesis
LanguageDanish
Publication date2011
Number of pages80