Time-Varying Vector Autoregressive Models with Structural Dynamic Factors
|Speaker:||Julia Schaumburg, Vrije Universiteit Amsterdam|
|Date: ||Wednesday 8 November 2017|
|Location: ||Streatham Court B|
We develop a transparent methodology for the estimation of time-varying parameters in vecto autoregressive models. In contrast to the widely used Bayesian approaches, we base our analysis on a combination of time-varying autoregressive coefficient matrices depending on a flexible set of stochastic dynamic factors, and of time-varying variance matrices depending on score-driven factors. The resulting method for estimating static parameters and extracting the different factors is insightful, robust and computationally fast, while being easy to implement. In a simulation study, we demonstrate the good performance of the method. We further show that our approach is promising in the empirical modelling of time-varying macro-financial linkages using a data set of U.S. macroeconomic and financial variables.