3

I have trouble understanding the following exercise:

Assume a setting in which the assumption of strict exogeneity holds, but random effects estimation and fixed effects estimation give very different parameter estimates. Which estimates do you trust most? Briefly explain why.

My answer to this question was that we should trust the random effects model, because $\sigma_{RE}$ is smaller than $\sigma_{FE}$ in the case of strict exogeneity right?

The correct answer to the question however is:

Trust the fixed estimation because the covariates are correlated to the fixed effects

Q: Why would we trust fixed effects instead of random effects in this case? (I don't understand the explanation given).

Pere
  • 5,875
  • 1
  • 13
  • 29
titusAdam
  • 365
  • 1
  • 10

1 Answers1

0

You would trust the Fixed Effects because it is consistent under weaker assumptions. This is not about the efficiency so the fact that variance is smaller is secondary. If the estimates are different, chances are strict exogeneity is violated and RE becomes biased, while FE remains consistent even without strict exogeneity.

Lucas
  • 1