A NORMATIVE PORTFOLIO DECISION SYSTEM

Abstract

The aim of this research is to demonstrate the usefulness of Decision Analysis, which is a normative analysis, for a fund manager to make a decision. In this thesis, the basic investment analysis method and the basic concept of decision analysis are briefly described. The influence diagram which is an effective and intuitive way to represent decision problem is introduced Shachter's (1986) influence diagram evaluation algorithm is extended by introducing the "null information mode" and the characteristic that the deterministic node which does not necessarily have to precede the value node, can be removed. The algorithm for generating the risk profile, which is not provide by the Shacter's algorithm but is important for the decision maker, is produced. Chance and decision nodes are extended to include values. Value node can accept both separate data and expressions as a function of the variables in the influence diagram. Finally, a Portfolio Decision System (PDS) prototype which is a decision system for equity investment is implemented under Microsoft Windows environment. The system includes tools such as a decision model editor, a probability assessment tool, and evaluation engine and a knowledge database. The system can suggest optimal policy and expected profit return for the decision-maker as well as the analyst. As a case study, the Shanghai Stock Market is chosen due to tits relative young age, and its volatility. Investment model for a decision-maker on the Shanghai Stock Market is constructed in the form of influence diagrams. With the evaluation of the investment model, the recommendation that investors should invest in equity rather than money market instruments can be obtained with respect to the investor's risk tolerance. With sensitivity analysis, how much the uncertainty influences the financial outcome is easily found. Further research should concentrate on how to extract relevant information from financial data such as database and graphs to help decision makers assess probabilities. A more complete implementation of the PDS has to be designed with the inputs from a real-world portfolio manager.