A Stochastic Optimization Approach to Financial Decision Making

Thumbnail Image
Journal Title
Journal ISSN
Volume Title
School of Business | Doctoral thesis (article-based) | Defence date: 2004-05-25
Checking the digitized thesis and permission for publishing
Instructions for the author
Liikkeenjohdon systeemit
Management Science
Degree programme
[118] s.
Acta Universitatis oeconomicae Helsingiensis. A, 234
Stochastic optimization is an effective tool for analyzing decision problems under uncertainty. In stochastic optimization a decision problem is formulated as an optimization problem, where the objective is to find an optimal decision, while considering all the possible scenarios for the uncertain factors and dependencies between the decision variables through time. In stochastic optimization the decision problem is solved numerically and there are only minor limitations for decision criteria, constraints and distributions of random factors that can be used in the formulations. This thesis consists of an introductory section and four articles. The introduction summarizes the contents and findings of the four articles and provides an introduction to the main issues in stochastic optimization: formulation of the decision problem as a stochastic program, econometric modeling of the stochastic factors and discretization of the problem for numerical solution. The first two articles are related to Asset-Liability Management (ALM) problem of a Finnish pension company. Article 1 develops a stochastic model for assets and liabilities of a pension company. The model is used for producing long term forecasts for asset returns as well as company’s liabilities and cash-flows. The model is utilized in Article 2, where a stochastic optimization model for ALM of a Finnish pension company is developed. The model is used as a decision support tool for finding long-term dynamic investment decisions in an uncertain environment, where the aim is to cover the uncertain future liabilities with dynamic investment strategies. The last two Articles address the problem of discretization of stochastic programs for numerical solution. New scenario generation techniques based on deterministic and randomized integration quadratures, more precisely Quasi Monte Carlo methods, are developed and applied to financial portfolio optimization problems. Conditions that guarantee the convergence of the objectives and solutions of the discretized problems to the original one are derived for both, Quasi-Monte Carlo and Randomized Quasi-Monte Carlo methods
Supervising professor
Kallio, Markku, professor
Other note
Permanent link to this item