A0925
Title: Time-varying global VARs with application to interconnectedness, structural analysis, and nowcasting
Authors: Francesco Meglioli - University of York (United Kingdom)
L Vanessa Smith - University of York (United Kingdom) [presenting]
Abstract: A time-varying parameter global vector autoregressive (TVP-GVAR) model is developed that leverages Kalman filtering techniques with forgetting factors to address the computational challenges typically associated with large-scale time-varying systems. The flexibility and broad applicability of the modelling approach are demonstrated through three distinct empirical applications, highlighting its practical value for regulators and policymakers. First, a baseline version of the TVP-GVAR model is employed, and is used to examine the evolution of interconnectedness between European banks and non-bank financial institutions over the past two decades. Second, the baseline model is extended to enable structural analysis, applying it to investigate the dynamic evolution of government spending multipliers across countries. Finally, the forgetting factors are incorporated within an unscented Kalman filter framework to estimate a MIDAS TVP-GVAR model, making the approach computationally feasible for nowcasting purposes. High-frequency macroeconomic indicators, including industrial production and prices, are used to produce timely nowcasts of quarterly economic activity.
