COMPSTAT 2022: Start Registration
View Submission - COMPSTAT2022
Title: Generalized autoregressive conditional betas Authors:  Francesco Violante - ENSAE ParisTech (France) [presenting]
Stefano Grassi - University of Rome 'Tor Vergata' (Italy)
Abstract: A new class of multivariate models is introduced allowing for observation-driven time-varying slope coefficients in a system of conditionally heteroskedastic simultaneous multiple regressions. These processes, dubbed Generalized Conditional Autoregressive Beta (GCAB), introduce a structural layer tailored to the linear asset pricing framework, solving the problem of incorporating time variation in the exposure of assets to risk factors in the asset pricing equation. This class of models accommodate for large cross-sectional dimensions (both in terms of regressors and regressands), and allow parametric cross-sectional restrictions, which are key for validation of asset pricing models. The proposed dynamics naturally accommodate the coexistence of constant and time-varying betas (that can be validated via testable hypotheses), and introduce new economically meaningful mechanisms of propagation of shocks, tagged beta spillovers. We derive stationarity and uniform invertibility conditions and, to mitigate the problem of parameter proliferation in large dimensions, we also provide beta and covariance tracking constraints. We propose a variety of computationally convenient (parallel and sequential) quasi maximum likelihood estimators, and we investigate their finite sample properties by means of Monte Carlo experiments. Finally, the GCAB is used to illustrate the role of beta spillovers in the Fama-French three factors asset pricing framework.