an econometric question on staggered DIV models

[Saleh sharmah]

Dear members

I am assessing the staggered policy mandate effect on firm outcomes across countries and I’m using the following regression model:
reghdfe outcome treated controls, a(firm year##industry) vce(cluster firm) My question is that two of my controls are related to country level characteristics and are non-variant ( these variables were calculated in 2005 and since then never been updated). The question is there any concern of biased estimates give that these two controls might be correlated with the fixed effects specifications in the model?
thanks for your input

[Luca Calianno]

Hi,
usually Stata's commands automatically drop the coefficients of time constant regressors.
If you really want to add those regressors you can use a correlated random effect specification.

[Saleh sharmah]

Thanks Luca for your input. I’m using the staggered DID stata command reghdfe and it does not eliminate constant control variables at the firm or country level, given that I control for firm fixed effects and year x industry fixed effects. The results remain unchanged if I drop these non variant variables. My question is whether the regdfe command allows for constant variables to be observed within the regression estimation and hence eliminating any milticollinority problem in the model ?