We also suppose that the error terms are homoskedastic and serially un-correlated (across the time). On the other hand, by random effect estimation, we suppose that the un-observed effect is un-correlated with the independent variables in all time periods.