Asset allocation in the Athens stock exchange: a variance sensitivity analysis

Abstract

ABSTRACTThis paper provides an analysis of asset allocation using univariate portfolio GARCH models applied on daily data for the period January 1999 to December 2009 on stocks traded in the Athens Stock Exchange, a recently monitored emerging market. Our analysis adopts the variance sensitivity analysis (VSA) methodology due to Manganelli (2004) and we are able to recover from the univariate approach the multivariate dimension of the portfolio allocation problem. The main results of the analysis are: First, we demonstrate that using a two asset portfolio consisting of blue chips traded in the Greek capital market the estimated variance is a parabolic and convex function of the estimated weights providing evidence that diversification produces significant gains in terms of risk reduction. Second, based on the shape of the first and second derivatives the model misspecification due to the fitting univariate GARCH models is insignificant. Third, we compare the performance of VSA against that of three popular multivariate GARCH models and it is shown that the adopted methodology provides more efficient results than the competing models. The gains in efficiency get larger as the size of the portfolio increases. Finally, with the application of the Kupiec's test for out‐of‐sample forecasting performance we demonstrate that the VSA outperforms all three alternative models at both the 95% and 99% confidence interval independently of the trading position. Copyright © 2011 John Wiley & Sons, Ltd.