Most of the existing research on auctions has been conducted within a non-cooperative framework with homogeneous risk neutral bidders where the risk neutral auctioneer can announce a reserve price. The contribution of this project comes from working in the relatively unexplored areas of heterogeneous bidders and bidder collusion. This research is motivated by observed behavior at auctions such as nested buyer ring formation, non-inclusive rings and lift-lining (phantom bidding) by auctioneers. The previous work explained the generation of phantom bids by practicing English auctions and the durability of bidder collusion. Further research is needed to generalize these results to more complex and realistic environments and to study other aspects of auctions. The high quality of the investigators' past research shows that they have the ability to do the work. The project investigates the behavior of heterogeneous bidders and the auctioneer at first price, Dutch, and second price auctions. Distributional information regarding bidders is usually unknown by the seller, so the processing of bid data by a Bayesian auctioneer and consequent strategic behavior are explored. The project explains why only all inclusive bidder coalitions are viable at first price and Dutch auctions and why coalitions of all sizes are observed at English auctions. The project investigates if certain bidder types are either undesirable to, or, unwilling to join a coalition in an English auction. Finally, the project explains how bidder coalitions try to avoid the incentive problems created by sharing the gains of coalition formation. Both simultaneous and sequential game models are used to study nesting of bidder coalitions and other ways of efficiently sharing the gains.