|Institution:||University of New South Wales|
|Keywords:||Optimisation; Multiple criteria decision making; Portfolio selection problems; AHP; ELECTRE; MAUT|
|Full text PDF:||http://handle.unsw.edu.au/1959.4/53445|
Multiple criteria decision making (MCDM) is a growing field that helps tackle complex problems under multiple and often conflicting criteria. The portfolio selection decision is a fundamental problem in financial investment where the future performance of assets is usually uncertain, but can be related to different attributes and factors. Thus, MCDM has a large potential to contribute in portfolio selection decisions. This thesis studies the applications of MCDM in portfolio selection problems of insurance linked securities (ILS) and common stocks. An approach utilising MCDM methods is developed and applied to real data: portfolio selection of ILS, and of the ASX200 stocks. The proposed model includes two main steps, screening based on MCDM methods to rank available assets in the investment universe according to a predefined criteria set, then standard portfolio optimisation techniques are utilised to construct optimal portfolios from the top 20 highest ranked assets. Two popular MCDM techniques are: Analytic Hierarchy Process (AHP) and ELECTRE III. Both of these techniques are considered in the screening step. Linear optimisation with constraints is used for ILS portfolio selection, while standard mean-variance optimisation is used for stock portfolio selection in the optimisation step. Results obtained from applying the model with AHP and ELECTRE are analysed and compared against each other, and also against market portfolios. The analysis shows that optimal portfolios based on the proposed model dominate the market portfolios. This demonstrates the relevance and usefulness of MCDM in a portfolio selection context. The final part of the thesis examines the link between AHP and another well established MCDM method: Multiple Attribute Utility Theory (MAUT). Different techniques to link AHP rating priority scales with the utility functions in MAUT are suggested. They are based on piecewise linear functions and curve fitting of parametric utility functions. The study indicates that AHP (direct rating mode) could be viewed as a simplified version of MAUT in additive form. Results from applying the two methods to ILS portfolio selection are very similar. This highlights the similarity and efficiency of AHP compared to MAUT when using the proposed model to incorporate MCDM into portfolio selection.