panel: Pooling vs Random vs Fixed (with large i & few t)

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

bokchy
Posts: 2
Joined: Sun Apr 09, 2017 11:13 pm

panel: Pooling vs Random vs Fixed (with large i & few t)

Postby bokchy » Mon Apr 10, 2017 9:51 pm

Hi,
I'm trying to conduct panel data regression.

Among 3 options(pooling, fixed, random),
I'd like to know which one would be best for my analysis.

My input data is unusual. It has a lot of individuals(i) but few periods(t)
Specific input data is as follows:

Periods included : 3
Cross-sections included : 8,194
Total panel (unbalanced) observations: 14,939


After conducting regression, my conclusions are as follows:

1) Pooling regression
- Every coefficient is significant with p-values under 1%.
- But, I think this method is relatively simple and dull...

2) Random effect
- Every coefficient is also significant.
- But fixed effect could be a more suitable model because Hausman test rejects null hypothesis,


3) Fixed effect
- According to the Hausman test, this would be suitable.
- But lots of coefficients are insignificant.
- Moreover, it might cause <incidental parameters problem>, owing to asymmetric input data(large cross-section, short periods).

So, I'd like to choose 2) random effect.

Is there any problem I might be missing if I choose 2) random effect ?
I'm concerned about the input data having a very short period.
I've never seen such panel data regression conducted with extremely short periods.

Return to “Econometric Discussions”

Who is online

Users browsing this forum: No registered users and 21 guests