Granger Causality Tests with Panel Data

[sanjeev]

Hi,
Could you also tell me how we can choose the lag length for the VECM with panel data?

Thanks.

[TomDoan]

Hi,
I have following two questions to ask:
1. While conducting joint granger causality tests, if we get a large negative value for the statistic, what could be the probable reasons?
2. Under what conditions do the joint statistic and the sum of the individual statistics not match?
3. What would be the corresponding p-value of the sum of the individual t-statistics?
Please reply soon!
Thank you.

Didn't you already ask that? https://estima.com/forum/viewtopic.php?p=14178#p14178

[sanjeev]

Thanks Tom for your reply!
What about selection of lag length?

Please reply!
Thanks.

[TomDoan]

Given that you have some individuals with missing observations, probably just pick a reasonable, but not very large, number and go with it.

[sanjeev]

Dear Tom,
Could you please clarify if the panel granger causality test discussed in the e-course on panel data assume cross-sectional dependence or does away with this assumption?
Please reply soon!

Thanks.

[TomDoan]

I have no idea what you mean. This isn't a unit root test. And there is no connection between individuals in the test.

[sanjeev]

Hi,
Could you let me know if and how we could conduct diagnostic tests like JB test for normality, LM test for serial correlation,etc after we estimate the VECM in panel the way we do it for time series.For instance,the following paper claims to have done that:
https://onlinelibrary.wiley.com/doi/abs ... 8268.12140.

Please reply!

Thanks.
Regards.

[TomDoan]

That paper is behind a paywall. However, a test for normality in a panel data situation (particularly a multivariate panel data situation) is going to depend a lot on what very specific assumptions are being made about the residuals. Unless they document rather carefully what they did, I would assume they just took an off-the-shelf test and applied it (probably incorrectly). At any rate, how is that even a "diagnostic" test? What about your model depends upon whether the residuals are Normal or not.

[sanjeev]

Dear Tom,
I was wondering as to how the degrees of freedom in the joint test for granger causality is determined.In my sample, with N=4 and T=35 with three lags,can degrees of freedom be 16?
Please reply!

Thanks and Regards.

[sanjeev]

Dear Tom,
In continuation with my previous query, I have some more queries:
While conducting granger causality tests also included ecm term in the regression for joint test.Should I include the ecm term while I conduct the individual granger causality tests also?

[TomDoan]

Dear Tom,
I was wondering as to how the degrees of freedom in the joint test for granger causality is determined.In my sample, with N=4 and T=35 with three lags,can degrees of freedom be 16?
Please reply!

Thanks and Regards.

T is irrelevant. Why would it be 16? It's N x the number of restrictions in each individual, so 12.

[TomDoan]

Dear Tom,
In continuation with my previous query, I have some more queries:
While conducting granger causality tests also included ecm term in the regression for joint test.Should I include the ecm term while I conduct the individual granger causality tests also?

Yes.

[sanjeev]

Dear Tom,
I was wondering if you could send me the code to test heterogeneity across cross-sections in a panel.

Please reply soon.

Thanks and Regards.

[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.

[sanjeev]

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?

Please reply.

Thanks and regards.