pooled ols - standard errors

#1
Hi,

has anybody an idea why the standard errors of a pooled ols estimation may be inappropriate if the error term contains an individual-specific component, even if it is uncorrelated to the regressors?
I mean if that component would be positively correlated to a regressor then I would say the coefficient of this regressor would be underestimated, would that be right? but what if they aren't correlated?

cheers!