Endogenous variables

[Oliver Gatland]

Hi all,

I am slightly confused around some of the assumptions of exogenous variables when using different regression techniques. To my understanding pooled OLS method assumes the independent variables are exogenously determined. Im not sure if this is the same for fixed effects regressions or not?

Apologies if this is a mundane question but I can't seem to find much information on this anywhere.

Thanks in advance for any help!

[Carlo Lazzaro]

Olivier:
predictors exogeneity is the general rule in regression model, as we are interested in the contribution of each predictor (when adjusted for the remaining ones) to change in the regressand. Whenever enedogeneity is detected, this goal is out of reach and what we obtain is biased and uninformative.
As far as panel data regression model are concerned, -fe- specification allows a sort of weak endogeneity, in that correlation between the idiosyncratic error and the vector of regressors is permitted.
Conversely, the -re- specification assumes 0 correlation (although this is sometimes an untenable assumption) between between the idiosyncratic error and the vector of regressors.