year fixed effects & industry fixed effects

[Finley Voigt]

Hello everyone,

I read this sentence: "We also included year fixed effects to control for time trends and industry fixed effects on the two-digit NAICS level to control for industry patterns in our analysis" in many diffrent papers now and researched the web for it.

My solution looks like, i used "i.app_year" & "i.ManufacturingFirm". I was just wondering if this is the correct way to do this, do you have any tipps? (I use a fixed effect model with xtpoisson and robust standard errors)

Thanks for your help.

[Carlo Lazzaro]

Finley:
how did you -xtset- your data?

[Finley Voigt]

Hi Carlo Lazzaro ,

I did it like that: xtset gvkey (Company Identifier) app_year (Year)

Does that help or is more information needed?

Thank you

[Carlo Lazzaro]

Finley:
thanks for clarifying.
It makes sense, but sounds a bit different from the papers that you mention did (as far as I can get it, at least).
It is not clear whether those papers were interested in retrieving the fixed effect of both firms and industries. If that were the case, and you're interested in mimicking the same approach, the Stata community-contributed module -reghdfe- can do it, but for linear regression.
Just a closing-out sidelight: the -fe- estimator available from -xtpoisson- is a conditional -fe- estimator, due to the incidental parameter bias (PII: S0304-4076(99)00044-5).

[Finley Voigt]

Thank you for your help :D