if I understand your question, the cointegration tells about the long run and short run relationship among variables while casuality indicates that which variable cause other, we have two forms of causality uni-directional and bi-directional. Suppose we have two variables X and Y. The cointegration will tell us the relationship of long run and short among these two while causality indicates either X is causing Y or either Y is causing X or either both varianles are casuing each other.
Thanks for your answers. They helped a lot. I understand the idea of causality but I wasn't sure if cointegration implied or not causality in at least one direction.
i know this is probably to late to add an answer, but hopefully it worth reading:
cointegration analysis has nothing to do with Granger formalism of causality.
cointegration is just a way to find associational linear relationship among non-stationary time-series. that is finding a stationary linear combination among I(1) time-series. even the approach of Engle-Granger 1987 does not confirm that cointegration has causal influence.
Granger 1969 causal definition is based on the measure of how much knowing variable Y will decrease the volatility of VAR of X.
In case of Bidirectional causality the use of ARDL is not valid. Agreed? Granger causality test is only between two variables . Should be use Granger test /Toad Yamamoto on variables to qualify for ARDL or is there are better methodology?