On Variable Selection for Volatility Forecasting: The Role of Focused Selection Criteria

This paper is concerned with the issues of modeling and projecting the dynamics of volatility when a group of potentially useful predetermined variables is available. We predict realized volatility and value at risk (VaR) with a nested set of multiplicative error models for realized volatility. We make use of recently proposed focused model selection/combination strategies as well as the classic AIC/BIC. Focused strategies consist of choosing the model that minimizes the estimated MSE of a given function of the parameters of interest to the forecaster. Results show that VaR forecasts can significantly be improved upon using focused prediction strategies. Copyright The Author 2008. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org., Oxford University Press.