Connectedness between <scp>G10</scp> currencies: Searching for the causal structure

This paper presents a new approach for modelling the connectedness between asset returns. We adapt measure of Diebold and Yilmaz, which is based on forecast error variance decomposition VAR model. However, their hinges critical assumptions with regard to variance–covariance matrix terms. propose use more agnostic empirical approach, machine learning algorithm, identify contemporaneous structure. In Monte Carlo study, we compare different measures discuss advantages disadvantages. an application analyse G10 currencies. Our results suggest that US dollar as well Norwegian krone are most independent currencies in our sample. By contrast, Swiss franc New Zealand have negligible impact other Moreover, cluster analysis suggests can be divided into three groups, classify as: commodity currencies, European safe haven/carry trade financing