Dumitrescu & Hurlin (2012) Granger non-causality test

[Gabriel Temesgen]

Hello Dears,
I am trying to see the granger causality between government revenue and government spending. I am using panel data of 40 countries and 20 years dataset. Since the variables have to be stationary, I use the first difference of revenue and level value of government spending. I also balanced the panel. However, the p-value of Z-bar and the p-value of Z-bar tilde have different levels of significance. Which one should I have to choose [Z-bar / Z-bar tilde]? Does the result indicate granger causality? Thanks

xtbalance, range(2000 2019) miss( dlrev lspending )
xtgcause lspending dlrev , lags(1)

Dumitrescu & Hurlin (2012) Granger non-causality test results:
Lag order: 1
W-bar = 1.4082
Z-bar = 1.8257 (p-value = 0.0679)
Z-bar tilde = 0.8740 (p-value = 0.3821)
H0: dlrev does not Granger-cause lspending.
H1: dlrev does Granger-cause lspending for at least one panelvar	(id).

xtgcause lspending dlrev , lags(2)

Dumitrescu & Hurlin (2012) Granger non-causality test results:
Lag order: 2
W-bar = 2.8715
Z-bar = 2.7560 (p-value = 0.0059)
Z-bar tilde = 1.0146 (p-value = 0.3103)
H0: dlrev does not Granger-cause lspending.
H1: dlrev does Granger-cause lspending for at least one panelvar	(id).

[Fei Wang]

This may help.

[Devashish Singh]

heyy did you get the answer to your query? i too am facing a similar issue

[Manh Hoang Ba]

"Z-bar tilde" should be use in sample with large N and small T. When large T, Z-bar can be reasonably considered (Luciano & Sylvain, 2017).