Title: Modeling realized volatility in big data panel
Authors: Michele Costola - SAFE, Goethe University Frankfurt (Germany) [presenting]
Massimiliano Caporin - University of Padova (Italy)
Mauro Bernardi - University of Padova (Italy)
Abstract: A Bayesian approach to the problem of variable selection and shrinkage in high dimensional sparse vector-HAR model is proposed. The regularisation method is an extension of a previous lasso. The model allows us to include the realized volatility of a large number of assets by taking into account in each equation of the system the HAR and co-jumps components for all the considered assets. The empirical analysis is performed on high-frequency data from the US market. The model estimates, performed on a rolling basis (and with a future time-varying parameter specification) will be used to retrieve the information needed for the identification of financial networks and to evaluate, ex-post, the estimated network structures.