Granger Causality Tests with Panel Data

[TomDoan]

What type of heterogeneity? You've asked that question before, and I pointed out that Baltagi has an entire chapter on different types of heterogeneity tests.

Hi,
The panel causality tests offered by RATS relax the assumption of homogeneity across cross-sections and therefore allow for heterogeneity.I am referring to this heterogeneity. I have earlier used RATS' code to conduct this heterogeneity test but I am not able to locate that now. One more thing, are you referring to Baltagi's book on Panel data Econometrics?

I don't recall ever describing a test for heterogeneity in this (Granger causality) context. First of all, it's unlikely to be very interesting. Almost all work with time series panel data allows at least the short-term dynamics to differ across individuals as there is no good theoretical reason to believe that they're the same. Also, even if the lag coefficients were the same, there is really, really no good reason to believe that the variances are the same and similarly (unless the series are assumed to be mean zero), no good reason to believe the intercept is the same.

[sanjeev]

Probably, I didn't frame my question properly. Please refer to the following test for granger causality.Since the test allows for heterogeneity in the coefficients and variances,so there might be a test to confirm this heterogeneity....

[TomDoan]

I understand your question. I answered it. This is not a test for heterogeneity/homogeneity, it's a test for causality. Allowing for heterogeneity means the test works either way, but (as I describe) heterogeneity is to be expected in this type of situation anyway.

[sanjeev]

Hi,
Sorry to bother you so much! Please refer to the following output:

Heterogeneity Test
Chi-Squared(6)= 249.830914 with Significance Level 0.00000000
Could you let me know the code for the above test? I cant locate the corresponding code.

[TomDoan]

Use Help--Find in Files to look for Heterogeneity Test in your program files. There's a test for heterogeneity (actually null of homogeneity) in the Pedroni JAE 2007 replication, so maybe you applied that to your own data.

[sanjeev]

Dear Tom,
While conducting gragnger causality tests for certain variables, I am getting following error message:
Heterogeneous Panel Causality Test
Chi-Squared(21)= -77.575088 with Significance Level NA
The Error Occurred At Location 624, Line 5 of loop/block
## X13. Redundant Restrictions. Using 0 Degrees, not 3
The Error Occurred At Location 624, Line 5 of loop/block
## X13. Redundant Restrictions. Using 0 Degrees, not 3

1.000000 14.888 0.00191
1.000000 53.462 0.00000
1.000000 NA NA
1.000000 2.761 0.43000
1.000000 39.242 0.00000
1.000000 NA NA
1.000000 14.046 0.00284
OVERALL -77.575 NA

Could you help me find the probable cause for the above error?
Please reply!
Thanks.

[TomDoan]

You had the same question before:

https://estima.com/forum/viewtopic.php?p=14173#p14173

except now you clearly have an even more serious problem with missing data---it looks like some of your individuals don't have enough data to even do the regressions, much less the test.