There is a long literature on stock return predicabtility.
When I read papers I see lots of authors still using granger causality tests. It is still prominently taught in undergrad and grad classess.
But we do we continue to persist with this when many have shown that they are often not robust to even a little bit of ambiguity. Whether this is persistent variables or in the presense of heteroskedasticity.
Why do people continue to use it?
Why is it so important?