Global minimum variance portfolio : an application to the Singapore stock market

This paper studies the performance of the Global Minimum Variance Portfolio (GMV Portfolio) constructed using three different models of estimating the covariance matrix; the Historical Model, Constant Correlation Model and Market Model. Unlike other studies that do not explore the effects of optimal look-back and rebalancing periods, we use an experimental method to vary these periods in estimating the covariance matrix. We then examine the differences in portfolio performances, so as to identify the GMV Portfolio with the best performance. We find that GMV Portfolios constructed using a 3-year look-back period generally perform better than those constructed with a 5-year look-back period. Within the 3-year look-back period, the Market Model, rebalanced annually performs the best with a Sharpe ratio of 0.53 compared to the market benchmark of 0.17. Lastly, the mean returns of the three models and are not significantly different from the market index (MSCI Singapore Free Index) when unadjusted for risk. Furthermore, the look-back periods have no significant impact on the mean returns of the portfolio within each model.BUSINES