To get rid of unobserved heterogeneity, all individual-specific variables which are time-constant are removed by FE model. So to my knowledge, whether we need to include lots of dummies to include FE variables, it doesn't matter because we are gonna subtract all those variables while getting FE estimator.
To put it briefly, I don't get the degrees of freedom issue when it comes to pros and cons of RE and FE models respectively.
What are your thoughts?(I'm just a beginner)
