Sector Omitted because of collinearity with fixed effects

[Simone De Sanctis]

Dear Stata community, I have a question. I'm studying the effect of CDPscore (A to D, with F = to a company not scored) on firm evaluation (Firm Value). I'm interested to understand how this effect change for environmentally sensible industries, thence I've created a dummy varibale (ESI_dummy) that equals to 1 if a company is in a sector that is sensible to environment. This is what Stata gives me

However, ESI_dummy is omitted because the sector remain the same for each company in all the years considered (2016-2020). Can I interpret the regression withouth including the ESI_Dummy but only by including the dummy as an interaction term for each score? Or is there other wat to solve this problem?

[Clyde Schechter]

If you wish to know whether there is effect modification by sector, then, yes, you can include sector interaction with the variable(s) whose effect might be modified. But using a fixed effects model, it is mathematically impossible to estimate the effect on the level of the outcome variable of a time-invariant covariate like sector or any sector-level variable like ESI_dummy. To do that you must go to a random effects model or a hybrid (aka correlated random effects or Mundlak) model (-xthybrid-, available from SSC can estimate these).

As an aside, it is unclear what the A-F and ESI_A-ESI_F variables are. The fact that one of the ESI_* variables was omitted due to colinearity makes me wonder if these are in fact 0/1 indicators that are indicator ("dummy") variables for having received grades A through F, respectively. If that is what they are, you can simplify your code by instead creating a single variable that gives the grade A, B, C, D, or F, then -encode- it, and use factor variable notation on the -encode-d variable in your regression command. See -help fvvarlist- for details.