Why Adding Fixed Effects May Increase Bias

Why adding (more) fixed effects is not a silver bullet for the problem of unobserved heterogeneity and 3 things you can do about it.
A common approach to controlling for unobserved heterogeneity is to run a linear regression with fixed effects. This post shows that this strategy may lead to strong bias in more realistic settings, even if you specify the fixed effects correctly. It is also about some things you can do about it.