Problem with xtcointreg and xtpmg for panel data

[Mjnh Pham]

Dear all,

In some literature reviews, I see that panel FMOLS/DOLS will be employed to estimate cointegration relationship (long-run relationship) between variables (in stata, xtcointreg is used to run these models).

However, with xtpmg, I can also gain the long-run cointegrating relationship between the variables.

I would like to ask whether or not there are difference between 2 approaches to estimate cointegration relationship between variables? If yes, which approach should I use? Because when I try both approaches, I am given very significantly different results

[Quynh-Anh Thi Nguyen]

Hi Minh,

As far as I know, xt cointreg is used to run cointegration tests among variables. You can use three main tests, check it out here: https://www.stata.com/new-in-stata/p...gration-tests/
If these tests say that variables have a cointegration, you need to estimate their relationship in long-term by FMOLS/DOLS.
Then, also estimate their relationship in short-term by VECM.
Both estimations can use xtpmg.

I am in the process of the step using xtpmg. So I also would like to ask you how to use xtpmg. Do you have materials to read about this command, in addition to the paper published by STATA? Could you please share or discuss the way to use xtpmg with me? Thanks!

[Mjnh Pham]

Dear Quynh Anh,

Thank you for your response.

I am afraid that you make a confused mistake. "xtcointreg" is used for panel cointegration regression between variables which are proved to be all integrated of the same order {for example, they are all I(1) or all I(2) } and cointegrated with each other. To test for panel cointegration relationship between variables, "xtcointtest " can be used with 03 types of test.

I also know that "xtpmg" can provide long-run relationship estimations apart from short-run relationship estimation. However, when I use "xtpmg" and "xtcointreg" to gain estimated long-run relationship between variables, I get very different results from 2 commands. Thus, my question is which command I should use to get more reliable estimated long-run cointegration relationship between variables.

To know how to use "xtpmg", you can refer to this material. It is very informative and handy.

Best,


Minh Pham