A usual assumption in the theory of collusion is that cartels are all-inclusive. In contrast, most real-world collusive agreements do not include all firms that are active in the relevant industry. This paper studies both theoretically and experimentally the formation and behavior of partial cartels. The theoretical model is a variation of Bos and Harrington’s (2010) model where firms are heterogeneous in terms of production capacities and individual cartel decisions are endogenized. The experimental study has two main objectives. The first goal is examine whether partial cartels emerge in the lab at all, and if so, which firms are part of it. The second aim of the experiment is to study the coordinated effects of a merger when partial cartels are likely to operate. The experimental results can be summarized as follows. We find that cartels are typically not all-inclusive and that various types of partial cartels emerge. We observe that market prices decrease by 20% on average after a merger. Our findings suggest that merger analysis that is based on the assumption that only full cartels forms produces misleading results. Our analysis also illustrates how merger simulations in the lab can be seen as a useful tool for competition authorities to back up merger decisions.