Theoretical IO models of horizontal mergers and acquisitions make the critical assumption of efficiency gains. Without efficiency gains, these models predict either that mergers are not profitable or that mergers are welfare reducing. A problem here is the empirical observation that on average mergers do not create efficiency gains. We analyze mergers in a model where firms cannot equalize marginal costs and marginal revenues over all dimensions in their action space due to constraints. In this type of model mergers can still be profitable and welfare enhancing while they create a loss in efficiency. The merger allows a firm to relax constraints. Further, this set up is consistent with the following stylized facts on mergers and acquisitions: M\&A's happen when new opportunities have opened up or industries have become more competitive (due to liberalization), they happen in waves, shareholders of the acquired firms gain while shareholders of the acquiring firms lose from the acquisition. Standard IO merger models do not explain these empirical observations.
