Global Minimum Variance Portfolio : A Horse Race of Volatilities

We conduct a horse race using three asset return volatility estimates: the sample variance, the exponential smoother used by RiskMetrics, and the generalized autoregressive conditional heteroskedasticity (GARCH). Our results are performed in both univariate and multivariate analysis. Our goal is to test whether using econometric tools achieves better economic performance. Economic performance is captured by portfolio performance as in DeMiguel et al (2009), who use three metrics to evaluate portfolios: risk-return, Sharpe-ratio, and turnover. Our portfolio of interest is the global minimum variance (GMV) portfolio since its inputs are determined solely by the variances and covariances of the asset returns. Computing the three portfolio performances for the three volatility candidate, we provide an evidence in support of the GARCH approach; however, this edge is undermined by greater portfolio turnover and larger estimation error, when the number assets in the portfolio is large