Performance of global minimum variance portfolio.

Abstract

This research uses four different methods of variance-covariance estimation namely Traditional, Traditional (Constraints), Zero Correlation and Constant Correlation to construct the Global Minimum Variance Portfolio (GMVP). The portfolios generated are subjected to Monthly, Quarterly, Semi-Annual and Annual rebalancing. We find that the less we rebalance the portfolio, the greater the returns generated and lower the standard deviation of returns obtained. The paired t-test analysis indicates that no single method of portfolio creation significantly outperforms the other methods. Lastly, we discover that the Zero Correlation method performs the best as indicated by the Sharpe Ratio and it also shows the least standard deviation of weights of the constituent stocks in the GMVP.