Volatility forecasts are important inputs into financial decisions such as portfolio allocation. While the forecasts are often used in such economic applications, the parameters of these models are traditionally estimated within a statistical framework….
Volatility forecasts are important inputs into financial decisions such as portfolio allocation. While the forecasts are often used in such economic applications, the parameters of these models are traditionally estimated within a statistical framework. This leads to an inconsistency between the loss functions under which the model is estimated and under which it is applied. This paper examines the impact of the choice of loss function on model performance in a portfolio allocation setting. It is found that employing a utility based estimation criteria is preferred over likelihood estimation, however a simple mean squared error criteria performs in a similar manner. These finding have obvious implications for the manner in which volatility models are estimated when one wishes to inform the portfolio allocation decision.
Read more at http://d.repec.org/n?u=RePEc:qut:auncer:2010_01&r=for