Hybrid optimisation and formation of index tracking portfolio in TSE

Assadi, M

Abstract

Asset allocation and portfolio optimisation are some of the most important steps in an investors decision making process. In order to manage uncertainty and maximise returns, it is assumed that active investment is a zero-sum game. It is possible however, that market inefficiencies could provide the necessary opportunities for investors to beat the market. In this study we examined a core-satellite approach to gain higher returns than that of the market. The core component of the portfolio consists of an index-tracking portfolio which has been formulated using a meta-heuristic genetic algorithm, allowing for the efficient search of the solution space for an optimal (or near-optimal) solution. The satellite component is made up of publicly traded active managed funds and the weights of each component are optimised using mathematical modelling (quadratics) to maximise the returns of the resultant portfolio.
In order to address uncertainty within the model variables, robustness is introduced into the objective function of the model in the form of risk tolerance (Degree of uncertainty). The introduction of robustness as a variable allows us to assess the resultant model in worst-case circumstances and determine suitable levels of risk tolerance. Further attempts at implementing additional robustness within the model using an artificial neural network in an LSTM configuration were inconclusive, suggesting that LSTM networks were unable to make informative predictions on the future returns of the index because market efficiencies render historical data irrelevant and market movement is akin to a random walk. A framework is offered for the formation and optimisation of a hybrid multi-stage core-satellite portfolio which manages risk through the implementation of robustness and passive investment, whilst attempting to beat the market in terms of returns. Using daily returns data from the Tehran Stock Exchange for a four-year period, it is shown that the resultant core-satellite portfolio is able to beat the market considerably after training.
Results indicate that the tracking ability of the portfolio is affected by the number of its constituents, that there is a specific time frame of 70 days after which the resultant portfolio needs to be re assessed and readjusted and that the implementation of robustness as a degree of uncertainty variable within the objective function increases the correlation coefficient and reduces tracking error.
Keywords: Index Funds, Index Tracking, Passive Portfolio Management, Robust Optimisation, Core Satellite Investment, Quadratic Optimisation, Genetic Algorithms, LSTM, Heuristic Neural Networks, Efficient Market Hypothesis, Modern Portfolio Theory, Portfolio optimisation

Citation

Assadi, M. Hybrid optimisation and formation of index tracking portfolio in TSE. (Thesis). University of Salford

• Thesis Type: Thesis
• Deposit Date: Nov 2, 2022
• Publicly Available Date: Nov 2, 2022
• Award Date: Feb 9, 2022