Panel data unit root testing for cross sectionally invariant variables

[Neha Verma]

Hello,

I am working on a quarterly panel data with N=6 (countries) and T = 84 (2000q1 to 2020q4). I am looking at the determinants of capital flows in emerging markets and some of the variables like US GDP, Global risk aversion etc are same for all panels. I have created these variables such that same values/data is pasted for each country under a single variable name. For example, US GDP has same values for each country pasted under a single variable titled 'US GDP'.

I wish to use FMOLS estimation for which unit root testing is required. My question is how to do unit root testing for cross sectionally invariant variables. Should I used -xtunitroot- or -dfuller- (in case of dfuller, I can pick data of just one country and run the command)?

Thanks
Neha

[Lise Pe]

Hello,
I have the same problem. Did you find the answer to this question?
Thanks

[Sebastian Kripfganz]

If the variables do not vary across countries, then you have only a single time series for each variable. Conventional time series unit root tests can then be applied (e.g. dfuller).