Seminar of the Thematic Project (2023/01728-0) Econometric Modlling and Forecasting in High Dimensional Models (HDEM&For)

Speaker: Dr. Igor Ferreira Batista Martins – Orebro Univserity – Sweden

Title: Stochastic Dynamic Correlations with Exogenous Shifts: Connecting Macroeconomic Events and Financial Risk

Abstract:
We propose a novel dynamic correlation model that allows for systematic shifts in correlations driven by exogenous variables while ensuring positive definiteness. Using matrix logarithm transformation, we obtain unbounded eigencorrelations that uniquely recover the correlation matrix. Our MCMC algorithm not only handles the non-linearity induced by the transformation but also easily incorporates information from realized correlations. Our approach enables separate treatment of correlations and volatilities, and we leverage this feature in our empirical application. Using data on the Australian Dollar and Japanese Yen exchange rate against the US Dollar, we find that correlation and volatilities generally increase on days when events occur, with announcements in the United States---and in particular FOMC meetings---being most impactful. Our findings highlight the dominant role of the US within the global financial markets.
 

Type: Online

(São Paulo local time)