Running a panel estimation (using reghdfe) with only dummy time invariant independent variables

[Corine EL Habr]

Hello everyone,

I have panel data on companies throughout years such as: Their profits, their own funds, their organizational forms (big complex corporates, medium sized corporates, ...), and other entity specific data. (The true data I have is confidential so this is similar to what I have, but not exactly it)
As a first test, I want to see if there is a significant difference in the profits of the firms depending on their organizational forms.

Here is my question: Can I have a regression with a continuous dependent variable (the profits), and only dummy time invariant variables (the organizational forms) on the right hand side? And even more, can I add entity fixed effects to the regression?
I have run a reghdfe using these variables, and I have gotten results, but I don't understand why it would work. Aren't I trying to predict changes based on time invariant variables?

To conclude, my point is to see if there is a structural difference between the profits of each organizational form (one form has a cnsistently higher level of profits than the other), while trying to exclude all other confounding variables.

Thank you very much,

Corine

[Carlo Lazzaro]

Corine:
1) -xtreg,fe- wipes out all time- invariant variables, like organizational model, if it does not change within panels across years;
2) if you want to compare profit variations among a set of panels (firms), -fe- specification is probably not what you're searching for, and you should consider -re- or, less frequently in applied research, -be- estimators;
3) in order to increase your chances of getting more positive replies, as per FAQ please post what you typed and what Stata gave you back and an example/excerpt of your dataset via -dataex-. If your dataset is confidential, just change the name of your variables. Thanks.