Title: Identifying uncertainty shock: A Bayesian mixed frequency VAR approach
Authors: Fabio Parla - Central Bank of Ireland (Ireland) [presenting]
Alessia Paccagnini - University College Dublin (Ireland)
Abstract: The aim is to investigate the transmission of COVID-19-induced financial uncertainty shocks to proxy global financial conditions and real economic activity by extending a previous empirical analysis to a mixed-frequency data sampling (MIDAS) approach. In detail, we proxy global financial uncertainty shocks by using the VIX. Global financial conditions and real economic activity are proxied, respectively, by the global financial cycle index and the world industrial production index. We estimate a Mixed-Frequency Vector Autoregressive model fitted to daily/weekly VIX and to monthly observations on the GFC index and the WIP. The model is estimated over January 1990-April 2019. To account for parameter proliferation, we estimate the model by adopting Bayesian techniques. Overall, the comparison of the impulse responses obtained from the estimation of an MF-VAR with those from a standard (common frequency) VAR suggests moderate evidence of a temporal aggregation bias corroborated by differences in the magnitude of the responses and in the uncertainty around the estimates. These differences are more pronounced when we increase the discrepancy between high- and low-frequency variables (e.g. daily vs monthly data).