Two-way fixed effects for multiple units and time periods DiD

[Marijn Borg]

Dear Statalisters,

For my master thesis I am currently working on the question whether the availability of ESG ratings impact analyst forecast accuracy. My dataset exists of the dependent variable "ForecastError" and independent variable "Treatment" which is 1 for the year since the ESG rating is available, also i have control variables. Now I plan to do a TWFE first and thereafter a Callaway-Sant'Anna approach (whereas i expect my treatment effect to be heterogenous). For my TWFE, I am hesitant on the correct commands, is the following correct:

using xtreg:
xtset panel_id Year

xtreg ForecastError Treatment (control variables) i.Year, fe
OR
reg ForecastError Treatment (control variables) i.Year i.panel_id

Is one of these correct? Or do i miss something?

Thank you in advance!

[Maxence Morlet]

both the equations you wrote are identical by the Frisch Waugh Lovell theorem. You may want to cluster your standard errors. For that, see https://academic.oup.com/qje/article/138/1/1/6750017.

[Marijn Borg]

Hi Maxence,

Thank you! If i include the clustered standard errors are both correct application of the Two-way fixed effects methods which used to be the default method for DiD with multiple time periods and units?

Kind regards!