The time variation between independent and dependent variables are only in each cross-sectional observation. By this method, the constant unobserved effect over time is not now a problem, and we do not need to assume that the unobserved effect is uncorrelated with the independent variables. We also suppose that the error terms are homoskedastic and serially uncorrelated (across the time). On the other hand, by random effect estimation, we suppose that the unobserved effect is uncorrelated with the independent variables in all time periods.