FE with large T and small N

[Sundus Al Riyami]

I have a panel data of 17 countries over 37 years. I am planning to estimate it using a fixed effects model. My data appears to be non-normal. Should I be worried about this? How can I fix it?

[Joro Kolev]

The fixed effects model is consistent with fixed N and T going to infinity. Should you be worried about what?

[Sundus Al Riyami]

Time trends. I am getting heteroskedastic standard errors. Should I add in more countries or will using Driscoll and Kraay's standard errors help give me robust results?

[Joro Kolev]

Yes, in a dataset where T is long you should worry about the time series properties of your data, such as deterministic and stochastic trends.

Driscoll and Kraay's standard errors are a good idea, but they do not resolve problems with deterministic trends and nonstationary data.

Time trends. I am getting heteroskedastic standard errors. Should I add in more countries or will using Driscoll and Kraay's standard errors help give me robust results?

[Sundus Al Riyami]

Thank you for your response! What do you suggest I do to resolve problems with deterministic trends and non stationary data?

[Keigo KAMEDA]

We used "xtcointreg" command.

(Our paper)
https://www.sciencedirect.com/scienc...22142522000056

I would be happy if this could be helpful.