Panel

[Ava Rooijakkers]

Hi guys,

I'm fairly new to Stata and have a question regarding fixed effects. I tried finding the answer by reading other questions but I'm a bit confused as I have seen various different answers.

For my thesis I want to run a baseline OLS regression for panel data. If I am not mistaken, this would be done by
reg y x1 x2

Now, I want to keep both industry and year fixed effects into account. Online I have seen various answers. Now I am confused when to use the ,fe option for xtreg and when to use the i.year or i.industry options.

I first set my data to panel by using
xtset gvkey year
And now want to run a regression where industry (identified by SIC code) fixed effects and year fixed effects are taken into account. What should the code be for this?
When I try the following codes
xtreg y x1 x2 i.year, fe
xtreg y x1 x2 i.year i.sic
I get very different answers.

Thanks in advance!

[Andrew Musau]

help xtreg

will tell you that xtreg by default estimates a random effects model. You need the option -fe- to estimate a fixed effects model. Your data is at firm level (gvkey), so if you

xtset gvkey year

you get firm fixed effects if specifying the -fe- option. For what you want, either

xtset industry
xtreg y x1 x2 i.year, fe robust

or

regress y x1 x2 i.industry i.year, cluster(industry)
regress y x1 x2 i.year, absorb(industry) cluster(industry)

[Ava Rooijakkers]

Thank you Andrew!

I have another question, a different subject.

In a paper by Bris, Koskinen, and Nilsso in 2009, they mention:
"Observations with a negative EBITDA larger than total assets are removed"
Could you please help me out with the code for this? I'm not sure how to specify the 'negative EBITDA' part.

I tried googling, but didn't know exactly how to Google it either.

Thanks in advance.

[Andrew Musau]

The value of assets is strictly positive (unless they refer to net assets, in which case if liabilities> assets, then net assets are negative). Therefore, assuming that you have the earnings variable (EBITDA), then it should be something like:

gen todrop = ebitda<0 & abs(ebitda)> assets
drop if todrop