Two-way fixed effects model in three-dimensional model

[Deric Kim]

Hello,

I'm confused by a critic's comment on my model specification, so I need some help.

My model has three dimension: i(region) , j(industry), t(year).
I initially had three-way fixed effects for the baseline model, including all three dummies (separate).
But the critics suggested a two-way fixed effects model using i#j interactions and t dummies.
I understand that i#j captures more dimensions so that it would be a more restrictive model, but why would the three-way fixed effects model be uncommon?

What would be the appropriate explanations?
Any comments or the related literature would be helpful.
Thanks a lot.

Best,
Deric

[Carlo Lazzaro]

Deric:
welcome to this forum.
I'm not clear with the estimator you used (-fe- or -re-).
If you go -fe-, -region. and -industry- are in all likelihood time-invariant and, as such, no coefficient will be estimated.