We propose the sharp identifiable bounds of the distribution functions of potential outcomes using a panel with fixed T. We allow for the possibility that the statistical randomization of treatment assignments is not achieved until unobserved heterogeneity is properly controlled for. We use...