Volatility forecasting using a decision-based attribution framework

This research develops a portfolio volatility forecasting method for absolute return equity strategies with consideration of managers' investment skills. Besides the portfolio holdings and prices that are commonly used by existing volatility forecast methodologies, the method proposed in this research takes account of investment skills and their volatility attribution. Investment skills are indicated by decisions of constructing portfolio overtime. Portfolio volatility is attributed to investment decisions through use of decision-based performance attribution model. It is shown that tracking the information contained in the time series of investment decision attribution leads to better volatility forecasts than commonly used forecasting methods which directly use returns and holdings. The forecasting method proposed has advantage of explaining risk forecast in terms of actual investment decisions, and changes to those decisions in real time.