Hello,

I have unbalanced panel data (74 companies with a total of 800 observations over 18 years).

I am trying to do the following regression (OLS with standard errors adjusted for heteroskedasticity and firm-level clustering):

MBE = MBEt-1 + Year dummy + Industry Dummy

I have read many posts about this type of regression but I am in doubt if I am doing it correct / which regression I should use and why:

xtset firm year, yearly

regressions:
1) xtreg MBE MBEt-1 i.industry, fe
or:
2) xtreg MBE MBEt-1 i.industry i.year, fe vce(cluster firm)
or:
3) xtreg MBE MBEt-1 i.industry i.year, re vce(cluster firm)

Summed up i kindly ask for answers to the following questions:
a) Is my approach correct?
b) Could you explain to me in the simplest of terms what clustering means and when you should use it?
c) I have performed a Hausmann test which tells me to use 'fe'. However, I am unsure if 're' may be more correct for my model due to my small dataset?

Best,
Niels