We analyze a Bayesian merger game under two-sided asymmetric information about firm types. We show that the standard prediction of the lemons market model–if any, only low-type firms are traded–is likely to be misleading: Merger returns, i.e. the difference between pre- and post-merger profits, are not necessarily higher for low-type firms. This has two implications. First, under very general conditions, equilibria exist where mergers take place, and there is no presumption that there is ineffciently low trade. Second, in these equilibria it is typically not the case that only low-type firms enter an agreement.