Thèse numérisée par la Direction des bibliothèques de l'Université de Montréal. / This thesis defines collusion broadly as play in a repeated game which differs from play in a one shot game. The analysis of collusion is an important part of many branches of economics. In industrial organization, for example, if collusion were not present then we could restrict investigation to the study of oligopolistic and competitive markets. Another important part of modern economic theory is the analysis of situations where there exists some ldnd of privileged information. Game theorists would say that in such a scenario there exists incomplete information. The predominant theme of the three essays composing this thesis is the study of repeated games under incomplete information. Typically, repeated game analyses have assumed the presence of complete information. However, many examples of repeated interaction must be treated in an incomplete information scenario. Auctions, the subject of the first essay, are the most natural example of this. The goal of this study is to understand how bidders collude in auctions. The main innovation is an explicit treatment of the repeated nature of the game to endogenize the threats necessary to support non competitive behavior. This analysis yields several testable implications about the behavior of colluding agents in auctions which are not apparent from the few models which have studied collusion in auctions from a static point of view. Nearly every model constructed to study collusion makes predictions which are at odds with accepted stylized facts. The most prominent of these paradoxes is the stability which theoretical models predict yet empirical and allegorical evidence rejects. This is the subject of the second essay. Incomplete information takes the form of how players evaluate future payoffs. This is an important detail because future payoffs are the only tool a cartel can use to enforce the play of collusive equilibria. The developed model imposes that patience is private information and heterogeneous and develops predictions which contrast strongly with accepted models of collusion. The predictions of the model are supported with discussion of some empirical evidence on cartels. The third essay develops a model which can have sociological as well as economic applications, in that it studies how rational agents form surplus creating partnerships in a repeated, incomplete information environment. Previous work has assumed an exogenous production technology which partners use to create this surplus. Furthermore an agents type, which affects surplus creation, has always been assumed to be observable. This essay studies matching with an endogenous production technology, in the sense that the surplus is a function of the level of collusion which can be supported. Collusion can be supported to varying degree based upon the type (patience, or discount factor) of each agent in the partnership. A special attention is turned to contrasting the implications of the model in the presence of complete as opposed to incomplete information.