Reverse causality test in panel data with small period

[Ari Anggrayni Ramadhan]

Hello. I have strongly-balanced panel data with 34 provinces and 6 years. I want to test the impact of government spending on pollution. I tried to check reverse causality with granger (xtgcause) but a warning appears that there must be more than 8 periods to run the test. Is there another way to prove the existence or absence of reverse causality in panel data with a few periods? Thanks in advance

[Maxence Morlet]

As far as I know you cannot prove it. There are circumstances under which it is unlikely, e.g. the regressor is defined on a coarser level than the dep. var., but otherwise you will want to argue it.