Fixed effect model, insignificant values help

[Karabo Mnisi]

Good day, I hope you're having a wonderful day.

I'm currently trying to model the effects of the determinants (gdpgrowth, corruption etc) of inward fdi on inward fdi.

I ran the fe and re models using

xtreg fdi debt gdpgrowth inf trade corr, fe
xtreg fdi debt gdpgrowth inf trade corr, re

and got back the following.

Fixed effect model

Random effects

as you can see that for fe the values are insignificant but for re it is significant but when i run the hausman test, I reject the null hypothesis and use fe which has insignificant values

Hausman test

So i tried re modeling the fe and adding logs to fdi and i now have less insignificant variables

Log Fdi (independent variable) fixed effects model

My question is would I need to log all my variables to get significant variables although i read somewhere that over logging is bad? Or is there another way to fix my fixed model and get significant variables like in my random effects model?

Thank you for your time

[Carlo Lazzaro]

Karabo:
the issue rests on a too limited sample size, that implies high standard errors and lack of statistical significance.
In addition:
1) you're dealing with a T>N panel dataset. Therefore, it would be wise to switch from -xtreg,fe- to -xtregar,fe-;
2) as per FAQ screenshots are discourared on this forum; please use CODE delimiters to share what you typed and what Stata gave you back. Thanks.

[Karabo Mnisi]

Thank you for your response.

My apologize, I did not know and will use code next time.

I ran -xtregar,fe- and still get insignificant values, could I just use the random effects model instead although hausman suggests I use fixed effects?

[Carlo Lazzaro]

Karabo:
again, the issue rests on your sample size.
Is there any chance that you can increase teh number of your observations?

[Karabo Mnisi]

If I increased the number of observations from 1990 to 2020 from 2000 to 2020 for example

inward FDI (dependent variable) would have 120 compared to 84
Debt would have 102 compared to 84
GDP growth would have 117 compared to 84
inf would 120 compared to 84
trade would have 119 compared to 84
corruption would have 84 compared to 84

making it strongly unbalanced and I would not be able to run my unit root tests but ive found a paper that deals with fixed model and insignificant variables and how they went about writing the paper.

[Carlo Lazzaro]

Karabo:
significant and insignificant coefficients are equally informative.
In your case, my guess is that the limited sample size causes the lack of statistical signbiifcance.

[Karabo Mnisi]

Carlo
Yes that is something, I will be pointing out in my caveats.