Volatility Forcasting and Portfolio Optimization

Abstract

This thesis investigates the relationship between volatility forecasting and portfolio performance. The aim is to use stylized facts about financial asset returns to improve the accuracy of volatility forecasts and see if better forecasts can improve portfolio selection – and performance. GARCH type models are used in order to forecast volatility over a rolling period of 1,008 trading days (4 years). The volatility forecasts are used to construct Markowitz mean-variance optimal portfolios by maximizing the Sharpe Ratio of the portfolios. We find that the ability to forecast volatility is linked with portfolio performance. The strategies that are able to forecast the volatilities with highest accuracy outperforms the other strategies in terms of cumulative returns and standard deviation of the returns.