Simulation Evidence on Granger Causality in Presence of a Confounding Variable

This paper provides simulation evidence on Granger causality between two variables when they are jointly caused by a third variable. Four Data Generating Processes (DGPs) are considered for testing causality by Granger method and two DGPs for testing causality by Toda and Yamamoto (1995) procedure. Our simulation involve three variables but causality has been tested only between two variable and the third variable (the real cause) has been ignored to show that its association which matters in these causality tests. Nevertheless, if we know that there are only two variables in economic dynamics and the true model is known then these causality tests work fine and for this we have carried out bootstrap simulation.