Control industry*year-fixed effect rather than separately industry-effect or year-fixed effect

[Hoa Phan]

Hi there,
Can I control industry*year-fixed effect rather than separately industry-effect or year-fixed effect?
I saw some papers they do that but I do not understand clearly,
Thank you!

[Carlo Lazzaro]

Hoa:
the issue here is how ton explain your methodological choice to your audience.
Personally, I would consider the two fixed effects (replacing -industry. with -firms-, though. If you go -fe-, time-invariant predictors like -industry- will be wiped out by the -fe- estimator) separately and code something like:

xtset firm year
xtreg <depvar> <indepvars> i.year, fe vce(cluster firm)

As an aside, to invoke non default stanadard errors positivey, you should have >=30 panels in your dataset.

[Hoa Phan]

Hi Carlo,
I mean how can we run regress for industry-time fixed effects conditions out time-invariant differences across time-varying industry factors.
You can see in this paper.
https://www-sciencedirect-com.ezprox...04405X21000957
Thanks!

[Hoa Phan]

Hi Carlo,
I mean how can we run regress for industry-time fixed effects conditions out time-invariant differences across time-varying industry factors.
You can see in this paper.

Ding et al. (2021), Corporate immunity to the COVID-19 pandemic
Thanks!

[Carlo Lazzaro]

Hoa:
it seems that Authors performed two different panel data fixed effect regression at country and a at firm level.
How this approach is in line with the tribal rules of your research field, I cannot tell.

[Hoa Phan]

Thank Carlo,
I may have look it back my study