# An Operational Goal Programming Model of Mutual Fund Portfolio ... An Operational Goal Programming...

date post

11-May-2020Category

## Documents

view

0download

0

Embed Size (px)

### Transcript of An Operational Goal Programming Model of Mutual Fund Portfolio ... An Operational Goal Programming...

An Operational Goal Programming Model of Mutual Fund Portfolio Determination

John R. MacLeod

Indiana University – Purdue University Fort Wayne

James S. Moore Indiana University – Purdue University Fort Wayne

An operational investment model is designed and built in order to find the optimal combination of a set of pre-screened mutual fund investments to make over a specific time horizon, subject to the goals and constraints defined by the user. The model can recommend an optimal investment mix based upon the user’s goals with respect to desired return, desired risk exposure, Sharpe Ratio, or a combination thereof. INTRODUCTION

The purpose of this model is to find the optimal combination of initial and recurring investments into a given set of candidate mutual funds over five years, subject to certain goals and constraints. Sixteen ‘no- load’ funds were pre-selected by the user from two different fund families, Royce and USAA. The candidate funds were chosen for consideration based upon historical performance data. A model with many more funds and/or a distinct set of candidate funds may be appropriate to another investor; customization to the situation of the individual investor is straightforward, while retaining the basic design tenants used here.

This model uses goal programming to determine the allocation of assets among the 16 different mutual funds over five years based upon three goals: maximizing return, minimizing risk, and maximizing the Sharpe Ratio for the portfolio. These three goals are weighted in the objective function, using user-defined coefficient values to establish their relative importance. The model’s constraints formally account for return, risk, Sharpe Ratio, budget, minimum investments and diversification. The budget for the entire five year period is set at $200,000, with the initial investment pool of $100,000 on the first day of year one, followed by annual investment pools of $25,000 on the first days of years two through five.

The portfolio also must have a diversity of funds. These investor-specific requirements state that no more than 70% of the budget can be invested in Royce funds, no more than 60% of the budget can be invested in USAA funds, at least 20% of the budget must be invested in small cap funds, and at least 12% of the budget must be invested in international funds. The model also requires that at least eight different funds are invested in during each period.

The model accounts for minimum investments by ensuring that the investments meet the criteria for each individual fund’s minimum entry requirement; it then ensures that subsequent investments into that

20 Journal of Accounting and Finance vol. 12(2) 2012

fund meet the fund’s minimum additional allocation requirement. The model is submitted to the Simplex method, as implemented in the optimization software LINDO. LITERATURE REVIEW

Optimization of investment portfolios based upon multiple criteria, such as risk and return, has been a topic of research for over fifty years. Since his early work framing portfolio selection as a mean-variance optimization problem with an efficient frontier, Harry Markowitz is generally regarded as the founding father of Modern Portfolio Theory (MPT). Prior to his seminal work, there was no formal attempt to quantify portfolio risk. Markowitz (1952) proposed a method by which an investor could assemble a portfolio which would generate returns greater than the mere weighted summation of the returns of the individual components. Markowitz contended that diversification of a portfolio has a direct affect on the risk level of the portfolio (Aouni, 2009). His mean-variance framework argued that, when building a portfolio, the co-movement of assets with each other is often more important than the individual security attributes. He introduced measurement of the dependence structure of returns, via the variance-covariance matrix, and quantified the impact on portfolio selection of each asset vis-à-vis the others; thus, formalizing the arguments for diversification and the ‘efficient portfolio’.

Understanding and implementing the underlying statistical implications of the Markowitz optimization model presents challenges. Several approaches have been put forth to address the problems of estimating the many elements in the covariance matrix [Ledoit and Wolf, (2004); Jagannathan and Ma (2003); Chan et al (1999); Chopra (1993); Green and Hollifield (1992)]. DeMiguel et al (2009) develop a framework for establishing an optimal portfolio in the presence of estimation error for the means and covariances of asset returns. They solve the traditional minimum variance problem, based on sample covariances, but subject to an additional constraint that the norm of the portfolio-weight vector be smaller than a stated threshold. Sharpe (1972) proposed a single-index model that circumvented the issues with implementing the quadratic formulation of the Markowitz efficient frontier. He suggested a linear framework that provides the foundation for a linear programming approach to portfolio optimization.

In formal optimization models, portfolio diversification is typically imposed via specific constraints limiting how much can be allocated (in actual dollars and/or percentages) into certain types of investments (Abdelaziz, Aouni, & Fayedh, 2007). In other models, binary variables are employed to impose quantity constraints (Mamanis & Anagnostopoulos, 2010). Diversification requirements such as these will be incorporated into the current model. Binary variables will play a role in setting the minimum number of chosen funds, as well as helping to establish the minimum investment allowable for each fund.

Markowitz’s bi-criteria portfolio selection model has evolved into multi-criteria models that include more than two objectives (Abdelaziz, Aouni, & Fayedh, 2007, Aouni, 2009; Mamanis & Anagnostopoulos, 2010). This research has brought about the creation of stochastic dominance models, multi-attribute utility models, multi-objective programming models, discriminant analysis, heuristic models, neural networks, optimization models, and multi-criteria decision aid methods (Aouni, 2009). Recently, the realization that adding more objectives to the model enhances its usefulness and dependability has motivated researchers to incorporate new criteria into selection models. These multi- dimensional models have helped to curb the assumptions that must be made by bi-criteria models, such as Markowitz’s (Mamanis & Anagnostopoulos, 2010).

Many portfolio models have failed to incorporate the preferences of the investor (Aouni, 2009). Investor preferences are to be explicitly included in the model, in order to account for the difference in acceptable risk and return (Davies, Kat, & Lu, 2005; Abdelaziz, Aouni, & Fayedh, 2007). This has shown to have significant effects on the reliability of the models. One model, built around hedge funds, seeks to optimize multiple conflicting objectives: maximizing expected return, minimizing return variance, maximizing skewness and minimizing kurtosis. Davies, Kat, & Lu (2005) show that the preferences of the investor have a profound impact upon the allocation of assets across the selection of hedge funds. Because of this, investor preference will be heavily incorporated into this optimization model for mutual funds via the coefficients in the objective function along with the constraints specific to the individual.

Journal of Accounting and Finance vol. 12(2) 2012 21

Further, the multi-criteria model of this paper is designed with multiple periods of investment. Some models are created to give recommendations over a specific time span, with investments made periodically throughout (Yu, Takahashi, Inoue, & Wang, 2010). This will be incorporated into this model via a time span of 5 years, with investments being made at the beginning of each year, following an initial investment in selected funds.

Sharpe (1966 ) developed a metric (then called ‘reward-to-variability’) for evaluating and predicting the performance of mutual fund managers. His capital asset pricing model (CAPM) extended the Markowitz theory to an equilibrium theory. The Sharpe Ratio measures performance relative to volatility, which is essentially the goal of the entire bi-criteria model. This value expresses the performance of a fund in relation to its risk (volatility) and can be used in a multi-criteria model to further establish a solid, dependable portfolio. The Sharpe Ratio will be the third parameter in the objective function for this model, establishing an even greater balance between risk and return. After nearly 50 years, this ‘Sharpe Ratio’ remains one of the most popular indexes of performance. The portfolio with the maximum Sharpe Ratio is the portfolio with the highest reward-to-risk ratio achievable from the available assets.

Nevertheless, alternatives to the Sharpe Ratio have periodically been advocated for portfolio selection [Choueifaty and Coignard, 2006; Szego, 2004; Pedersen and Satchell, 2002; Ortobelli, et al, 2003; Farinelli and Tibiletti, 2003; Uryasev, 2000; Sortino, 2000]. Unfortunately, these theoretically valid metrics seldom recommend the same portfolio solution, leaving the practicing investor at a loss as how to proceed. The Sharpe ratio, and it variations, have been criticized when the assumption of normality in the distribution of returns is relaxed; It can result in incorrect investment decisions when the returns exhibit kurtosis or skewness [Leland 1999]. Farinelli et al 2008 propose various asymmetri

Recommended

*View more*