A Comparative Study of Computational Finance and Behavioral Finance Approaches in Multi-Period Asset Allocation

Chen Xiaoyang

Abstract

The traditional multi-period asset allocation focus on (a) Markowitz’s single period mean-variance model, (b) multi-period linear programming models based on scenario generation, (c) dynamic programming models and (d) models based on Markov property. This thesis will focus on the investor’s utility and takes into account the risk preference of the investor. Expected utility theory and prospect theory are two famous utility theory in in contrast with each other. In this thesis, we study the multi-period assets allocation from both computational finance approach and behavioural finance approach. Computational model is focused on the optimal strategy, while the behavioural model is focused on real-life behaviour. First we propose an optimization model with rebalancing options to get the optimal strategy based on expected utility theory. Second, we propose a simulation model to simulate human’s behaviour in multi-period asset allocation based on prospect theory.

The method we adopt is statistics sampling, scenario generation, quadratic programming model, simulation modelling. We use bond, stock and mutual fund as three kinds of assets to assist us in justifying the model. We also investigate the influence of changing required returns and varying time horizons costs to the final investment policy. By compare and contrast between two models, we find the strategy from computational finance approach can help investors get a significantly better result than investing based on intuition.