Article S12.3 Improving Portfolio Management: Research Focus on Computational Models Md Washim Raja, Dillip Kumar Rath Doctoral Candidates, Xavier University, Bhubaneswar, INDIA washim[at]stu.ximb.ac.in Published Online: September 21, 2015 Note. This report is based on the “Topic Registration Seminar” delivered by Biplab Mahapatra, Doctoral Candidate, Xavier Institute of Management, Xavier University, Bhubaneswar, India, on August 11, 2015. Doctoral Scholar Biplab Mahapatra began his seminar by discussing artificial intelligence. It is a field of study that seeks to mimic human (or animal) intelligence through computational means. One of the applications of artificial intelligence has been in the area of financial management, especially portfolio management (i.e., managing a set of investments for optimal risk-return outcomes). Mahapatra described how genetic algorithm, a class of techniques in artificial intelligence, can be used for portfolio optimisation process. Portfolio optimisation is a multiobjective optimisation problem due to the presence of conflicting objectives of maximising return and minimising risk. With the publication of the seminal paper of Harry Markowitz (Markowitz, 1952), the so-called father of modern portfolio theory, the portfolio selection process went through a paradigm shift. Prior to his work, the risk of the portfolio was considered to be the summation of the risks of the individual assets, measured by volatilities and the emphasis was on picking single high-yield and low-risk stocks without any regard to their effects on the portfolio. But Markowitz developed the concept of correlations among the individual assets and the diversification effect created by those correlations. Markowitz claimed that he could create a portfolio with the same expected return but lesser risk compared to the portfolio created by ignoring the correlation among these assets. Genetic Algorithm Genetic algorithm was proposed by John Henry Holland in the year 1975. It is a search heuristic that mimics the process of natural selection. This heuristic is used to solve optimisation problems. Mahapatra highlighted the effectiveness of genetic algorithm for single-objective optimisation problems (Fogel, 1999; Goldberg, 1989; Schwefel, 1981) and for multiobjective optimisation problems (Zitzler, Deb, & Theile, 2000). He also discussed the use of genetic algorithm for solving decision problems such as making financial trade decisions under currency volatility, scheduling of events in Barcelona Olympics, optimising telecommunication networks, and designing aircraft engines (Grupe & Jooste, 2004). Genetic algorithm has also been used by several researchers for portfolio optimisation. However, as indicated by Grupe and Jooste (2004), genetic algorithm is not free from limitations. It is not always possible to formulate problems as required by genetic algorithm. It does not guarantee an optimal solution. As genetic algorithm uses random numbers, different runs can give different solutions. Review of the Literature Some of the original assumptions of Markowitz such as financial returns are normally distributed were proved wrong by other researchers (Maringer & Parpas, 2009; Mills, 1997). The use of variance as a measure of risk was later modified by Markowitz (1991) himself; he proposed semivariance as a better measure of risk. The sensitivity of genetic algorithm based models to inputs such as risk and return was criticised by Oberuc (2004). Portfolio management models based on genetic algorithm had limited acceptance in industry due to their distance from reality, as the models assumed zero taxes and transaction costs. Another assumption was that the assets are infinitely divisible, which is not true because of practicality and trading restrictions. Similar to various other models in finance and economics, these portfolio management models also make unrealistic assumptions about how investors think and decide. Statman (2014) has highlighted the difference between standard finance and behavioural finance. He has discussed the impact of the behaviour of investors, who are considered to be “normal people,” not always rational in the same way, as assumed in standard finance theory. Subsequently, researchers have tried to add practicality constraints such as round lot constraint, floor-ceiling constraint, buy-in threshold constraint, and asset class constraints, to bring models closer to reality. Some researchers have attempted to add more realistic objectives such as 12-month performance, 3-year performance, dividend yield, and Standard & Poor’s star rating. The constraints can be loosely classified into regulatory constraints, guideline constraints, trading constraints, risk management constraints, and discretionary constraints. Regulatory constraints are the constraints which cannot be violated (such as insider trading guidelines). Guideline constraints are imposed by the organisation on investment advisors as well as the traders to protect the organisation as well as to maintain reputation. Trading constraints are those constraints which are imposed by trading rules such as round lot constraint, floor-ceiling constraint, and buy-in threshold constraint. Risk management constraints are those which help in reducing risk by limiting the exposure in a particular asset, sector, or geography. With the addition of these constraints, the portfolio optimisation process becomes a non-deterministic polynomial-time hard (NP-hard) problem. NP-hard problems are those problems which can be solved in theory but these are not useful in practice due to the time needed to solve the problem (Hopcroft, Motwani, & Ullman, 2007). Moreover, additions of constraints also make the problem space more complex where classical optimisation techniques fail. Research Focus Mahapatra aims to study the potential of genetic algorithm based models to create an optimum portfolio under realistic assumptions. This research attempts to create the construct to explain the components of the portfolio optimisation problem. The seminar discussions focused on the following possible future directions in this field. (a) The scope and limitations of heuristics in portfolio optimisation problems (b) The impact of adding more constraints on portfolio optimisation models (c) Integrating behavioural finance aspects such as “loss aversion” (Kahneman & Tversky, 1979) in the model (d) Evaluating the efficiency and effectiveness of genetic algorithm based models under various conditions References Fogel, L. J. (1999). Intelligence through simulated evolution: Forty years of evolutionary programming. New York, NY: John Wiley. Goldberg, D. E. (1989, June).Sizing populations for serial and parallel genetic algorithms. In Proceedings of the 3rd International Conference on Genetic Algorithms (pp. 70-79). Morgan Kaufmann. Grupe, F. H., & Jooste, S. (2004). Genetic algorithms: A business perspective. Information Management & Computer Security, 12(3), 288-297. Hopcroft, J. E., Motwani, R., & Ullman, J. D. (2007).Introduction to automata theory, languages, and computation. Boston, MA: Pearson. Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica: Journal of the Econometric Society, 47(2), 263-291. Maringer, D., & Parpas, P. (2009). Global optimization of higher order moments in portfolio selection. Journal of Global Optimization, 43(2), 219-230. Markowitz, H. (1952). Portfolio selection. The Journal of Finance, 7(1), 77-91. Mills, T. C. (1997). Stylized facts on the temporal and distributional properties of daily FT-SE returns. Applied Financial Economics, 7(6), 599-604. Oberuc, R. (2004). Dynamic portfolio theory and management. New York, NY: McGraw-Hill. Schwefel, H. P. (1981). Numerical optimization of computer models. New York, NY: John Wiley. Statman, M. (2014). Behavioural finance: Finance with normal people. Borsa Istanbul Review, 14(2), 65-73. Zitzler, E., Deb, K., & Thiele, L. (2000). Comparison of multiobjective evolutionary algorithms: Empirical results. Evolutionary Computation, 8(2), 173-195. Suggested Citation: Washim Raja, M., & Rath, D. K. (2015). Improving portfolio management: Research focus on computational models. Research World, 12, Article S12.3. Retrieved from http://www1.ximb.ac.in/RW.nsf/pages/S12.3
Copyleft Xavier Institute of Management, Xavier Square, Bhubaneswar 751013, India Research World (ISSN 0974-2379) http://www1.ximb.ac.in/RW.nsf/pages/Home | ||
 |