Fixed effects model

[Jan Geurst]

Hello,

Can someone explain me the difference and the methodological reasoning behind including country, industry and time fixed effects or including country*time and industry*time fixed effects?

Thus:

egen country_time=group(country time)
egen industry_time=group(industry time)

reghdfe var1 var2..., absorb(country industry time)
reghdfe var1 var2..., absorb(country_time industry_time)

Thank you!

[Carlo Lazzaro]

Jan:
the -egen- code creates a new -panelid-; the second one absorbs multiple leves of fixed effect .

[Jan Geurst]

Thank you Carlo!

Can you elaborate on the difference between the two regressions with fixed effects? When should someone use the first regression and when the second?

[Carlo Lazzaro]

Jan:
usually, the -egen- code is invoked when you have multiple observations for the same id during in the same wave (say, year).
However, provided that you do not plan to use -tsvarlist- commands, such as lags and leads, things can be made simpler by -xtset-ting your data with -panelid- only:

xtset panelid

.

The second code can be useful if you want to include both time and country fixed effects.

[Jan Geurst]

Okay, so there's no methodological reason to include country, industry and time fixed effects separately or to interact them with time? Since they give different results?

[Carlo Lazzaro]

Jan:
no, it depends on your research goal.
Anyway, the aim should be to give a fair and true view of the data generating process.

[Jan Geurst]

Thank you, Carlo!