Greetings! I am analyzing the impact of climate change of wheat crop across 4 districts for the last 30 years. It is imperative to check the stationarity of the panels. I performed LLC and IPS tests. The results showed that the panels are stationary.

Now please let me know:

  1. Whether I need to perform cointegration tests like Kao, Westerlund etc. I doubt that if the panels are stationary, there is no cointegration, and there is no need to go for cointegration tests.

  2. The LLC and IPS unit root tests assume the cross sectional independence. If their results show no unit roots, then, do I still need to test the cross sectional dependence through Pesaran CD or other LM tests. Or I can skip the cross sectional dependence tests

  3. If the panels are stationary, should I proceed directly to the model specification tests to find the best model among Pooled OLS, Fixed Effect and Random Effect Models.

I will be highly thankful to you all for your help and guidance in this regard.


1 Answer 1


Ad 1, indeed, if the series are stationary, there cannot be cointegration, as cointegration describes the situation of stationary linear combinations among nonstationary variables.

Ad 2, yes, as it is quite possible that the rejection you obtained from LLC/IPS is spurious, i.e., that it did not result because the series really are stationary but because an underlying assumption of the test is not met so that it does not give trustworthy results.

Ad 3, I am not sure what you mean by "directly", but that does not sound like a bad idea.

  • $\begingroup$ Thank you sir for guiding me in my analysis. I used "directly" as I intended to say that I should skip cross sectional dependence test and proceed to model specification tests. $\endgroup$ Feb 8, 2023 at 11:17
  • $\begingroup$ OK, thanks for clarifying. In 2, by yes I meant that you should test for CSD and if you find it, use robust panel unit root tests $\endgroup$ Feb 8, 2023 at 12:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.