Automated Investment Assessment : an investment decision making model for the ordinary person

Abstract

The ordinary person invests to achieve their long-term nancial goals. Choosing where to invest and what to invest in can be challenging for the ordinary person and paying a nancial broker can be expensive and reduces the returns received. Thus, it was decided that the ordinary person requires a decision support tool to aid them in deciding where to invest. It was decided to investigate investing in the shares of 40 Johannesburg Stock Exchange (JSE) companies. Factors that in uence which shares to invest in were investigated and it was discovered that there are four main metrics that should be considered simultaneously. These metrics are the market liquidity of a company, the market risk of the company, the nancial risk tolerance of the ordinary person and the expected nancial growth of the company. As these metrics must be considered together, it was discovered that this is a multi-objective optimisation (MOO) problem and a customised MOO model should be developed and built. Three simple MOO methods were selected, the lexicographic, weighted sum and weighted product methods, to ensure that the ordinary person using the model will be able to understand the model logic. The metric data used by the model was collected for three years, and as such it was decided that a weighed-moving average metric value should be calculated for each metric for each company to account for stochasticity. A mathematical model for selecting investment portfolios consisting of at most ten companies from the JSE 40 was developed. Furthermore, pseudocode was written that was used to develop a model that selects investment portfolios for the ordinary person. This model selects investment portfolios using each of the three suggested MOO methods as well as based on only individual metric values over a range of risk tolerance score (RTS) values, using six di erent weighting combinations of factors. As di erent methods are used to select investment portfolios, with six di erent sets of weightings over a range of 44 di erent RTS values, the model produced a total of 13 464 model portfolios. The return on investment was calculated for each portfolio and it was found that the majority of model portfolio had similar returns. This is as a result of the fact that the model only selected 16 of the JSE 40 companies and these were selected multiple times as part of multiple portfolios. These model returns were then compared to the returns on investment received by various unit trusts over the same time period. It was found that the model portfolios signi cantly outperformed the unit trusts and thus produces worthwhile results. The model was rerun three time with di erent assumptions and a sensitivity analysis was performed. It was found that the model is not very sensitive to its model assumptions, although this may again be result of the limited sample of companies. The payout received by the unit trusts and model portfolios for various investment periods were calculated and it was discovered that the model portfolios were signi cantly outperformed. For this reason, it recommended that the ordinary person should invest in unit trust with reasonable brokerage fees.