Hi, I have an unbalanced panel data (cross-country included 18 and periods 20 years). When I run panel unit root tests (LLC, Breitung, Im Pesaranm, ADF-Fisher, PP-Fisher) I get mixed results. I have difficulties to interpret them. Can anyone help?
(The only thing I have read that separates these tests is that the latter 3 allow for unbalanced panel, however, even between these results I get mixed results.)
Thanks for any comments in advance.
I recommend you book, entitle: " Econometric Analysis of Panel Data' (Baltagi, Third Edition), See chapter 12 from page 239 to 250. Base on strong theoretical, chose one unit root test and follow it.
for unbalanced panel the most appropriate test is Fisher test: ADF-Fisher and PP-Fisher
Dear David,
We can select the Panel Unit root tests based the type of data set such as balanced or unbalanced. For unbalanced panel data set we can use the Fisher type tests of ADF and Phillips Perron these are sufficient for the test the stationarity. However, if the data is balanced these methods gives misleading results. For that we can use the Levin-lin-ch, Harris-Tzavalis, and Breitung tests are necessary to test the unitroot in balanced panel data.
According to your data you can go with the ADF and PP tests.
I hope you will understand.
Dear David,
I share an image that can be useful for you.
The reference of this image is "User's Guide - Stata".
Best regards,
Reza
Dear Reza
Your answer is useful for solving David's question
Fisher test is recommended with de-mean unit root test for unbalanced panel data
Dear Hari Venkatesh , do you have a paper that I can cite for your provided argument (that PP and ADF tests are better for unbalanced panel)? Thanking you.
Dear Nayaz Rojid
Please go through the STATA manual 13.
https://www.stata.com/manuals13/xtxtunitroot.pdf