Title: A mixed-frequency combination approach to forecast crude oil and gold future returns volatility and correlation
Authors: Malvina Marchese - Cass Business School (United Kingdom)
Ioannis Kyriakou - Cass Business School (United Kingdom)
michael tamvakis - bayes business school (United Kingdom)
Francesca Di Iorio - University of Naples Federico II (Italy) [presenting]
Abstract: To forecast the covariance matrix of crude oil and gold futures returns, a novel forecast combination approach is proposed based on mixed information, i.e. high and low-frequency data. Specifically, the combination strategy identifies the optimal predictor using several loss functions via an iterative procedure based on the Model Confidence Set. The findings suggest that combined forecasts of multivariate GARCH and Realize Covariance models outperform each individual model and their equally weighted mean from a statistical as well as an economic perspective, indicating that low-frequency data improve volatility forecasting even when high-frequency data is available.