We study, using the non-parametric data envelopment approach, we investigated the long-run profit efficiency dynamics and the short-run market reaction of nine pre-classified merger deals of merging and non-merging U.S. banks over the time period from 1992 to 2003. Our main results are as follows: First, merger deals that match least efficient acquirers with the least efficient targets could improve their profit efficiency four years following the merger event, unlike all other merger deals. Second, we find that mergers match least efficient acquirers with the least efficient targets could also achieve significant positive cumulative access returns (CARs) while all other deals were followed by significant negative CARs. Third, we find that, in general, that large-size acquirers have and maintain higher and efficiency scores than targets and non-merging banks. Fianally, the value-maximizing mergers are mostly large in size and match banks with clear chances to increase their future efficiency rankings.