A0249
Title: Improved density forecasts using mixed frequency data: A focused Bayesian approach
Authors: Andrew J Patton - Duke University (United States)
Ole Maneesoonthorn - Monash University (Australia)
Ruben Laoiza Maya - Monash University (Australia) [presenting]
Abstract: A modeling framework is constructed that allows for the flexible contribution of intraday information to predict the conditional distribution of asset returns. A flexible spline specification is proposed to capture the contribution of intraday returns inside a GARCH-type specification for the daily return, and the "focused Bayesian'' prediction framework is adopted to allow the intraday contribution to be trained according to the region of the support that is of interest. An empirical analysis of three broad market indices and ten industrial indices demonstrates the gains from using (i) high-frequency data, (ii) incorporated flexibly into the low-frequency model, (iii) estimated using a focused Bayes approach.
