Causality between oil shocks and exchange rate: A Bayesian, graph-based VAR approach

Our paper studies the casual relationship between oil and major bilateral exchange rates against US dollar via a novel Bayesian, graph-based approach. This approach is shown to be quite effective in dealing with identification in Vector Autoregression (VAR) model, in which the temporal causal structure is represented by a graph sampled by Markov Chain Monte Carlo (MCMC) method. Empirical evidence demonstrates that oil price leads the exchange market in the after-crisis period whereas vice versa before crisis, implying a potential impact from financial crisis on the causality between these two markets. We further show that in general, oil-market specific shock affects the dependence structure most, while aggregate demand shock plays a weaker role and supply shock contributes least. Specifically, these three oil shocks take effect during different periods, thus capturing some invisible information about market evolutions.