Title: GAMM style volatility modeling
Authors: Giacomo Bormetti - University of Bologna (Italy) [presenting]
Giulia Livieri - Scuola Normale Superiore (Italy)
Fulvio Corsi - University of Pisa and City University London (Italy)
Abstract: In modeling daily volatility with high-frequency data, the econometrician faces two serious limitations: Realized measures are contaminated by noise and the overnight volatility is not observable. Both effects undermine forecasting ability and pricing performances. Inspired by the Generalized Autoregressive Method of Moments, we introduce a new reduced-form model for price returns which solves both issues. The observed close-to-open and open-to-close returns and realized volatilities step into the model via orthogonal conditions. As quantities with zero conditional expectation, the latter drive the dynamics of the latent volatility components and jump intensity. Our observation-driven specification is tailored for an effective and computationally undemanding filtering, that reveals the wandering behaviour of the overnight component and a sizeable increase of persistence of the latent volatility. Remarkably, the model belongs to the class of affine processes and thus inherits all advantages deriving from analytical tractability. By means of an extensive analysis of S\&P500 Futures time series and S&P500 Index options spanning more than two decades, we report the superior performances of our approach in comparison with competitor models.