A0393
Title: Dynamical quantile graphical modeling
Authors: Piergiacomo Andrea Carlesi - Univ. degli Studi di Padova, Dept. of Statistical Science (Italy) [presenting]
Mauro Bernardi - University of Padova (Italy)
Cristian Castiglione - Bocconi University (Italy)
Nicolas Bianco - University of Padova (Italy)
Abstract: An innovative approach is introduced for jointly estimating multiple quantiles within a dynamic framework. Essentially, the method acknowledges that the fixed-level quantile of a vector of response variables is influenced by both a fixed set of covariates and a random effect with an autoregressive dynamic. As a result, the proposed framework expands upon traditional univariate quantile models by integrating a vector autoregressive structure. To streamline the estimation of model parameters and the extraction of signals, Bayesian methodologies are developed that leverage approximate techniques and data augmentation strategies. These methodological advancements are instrumental in efficiently addressing the complexities inherent in estimating the model parameters and extracting meaningful signals from the data. A comprehensive panel comprising US equity market returns and macroeconomic indicators is utilized to empirically validate our approach. Through this analysis, the objective is to shed light on the dynamic evolution of spillover effects within individual Value-at-Risk estimates. By scrutinizing the interaction between macroeconomic variables and the autoregressive component, the aim is to uncover the intricate mechanisms that drive risk dynamics in financial markets.
