Constrained Mean-Variance Portfolio Optimization with Alternative Return Estimation

This paper studies the problem of asset allocation in a mean-variance framework. The theoretical model of portfolio optimization is specified and then applied to a long panel data set from historic to most recent times, March 1990 – March 2013. The paper contributes in three ways. First, an alternative asset return model is proposed that combines the historical returns, capital asset pricing model (CAPM) and returns estimated based on firm fundamentals. These return estimates enter the optimization problem. The second contribution is the application of an improved covariance matrix estimator that has superior properties compared to the typical sample covariance estimator. Third, the paper proposes two investments strategies. The first proposition suggests always choosing the maximized Sharpe ratio portfolio and the second one, the portfolio with the highest information ratio. The nature of both strategies is designed for investors with different appetites for risk. The performance of these choices is analyzed in light of four types of constraints: upper/lower investment limits, group constraints and transaction costs. The one-period optimal investment portfolio is rebalanced at quarterly intervals. Both strategies are benchmarked against an alternative investment choice such as holding the S&P 500 index, or investing in a risk-free asset such as a bond. Portfolio analysis and backtesting reveal that the strategies are superior to simply holding an equally weighted portfolio, a risk-free asset or the S&P 500 index.