The effect of collusion on efficiency in experimental auctions

This paper examines the effect of collusion on allocative efficiency in a second-price sealed-bid auction, in which bidders' valuations have both private and common value components. We present a theoretical model which shows that explicit collusion improves average efficiency. Furthermore, a reduction in common value signal variance increases the efficiency of allocations when a cartel is present. We test for the presence of these patterns in a laboratory experiment. Subjects can choose whether to compete or to form a cartel. Colluding bidders can communicate and make side payments using a knockout auction. Our results show that a large majority of bidders joins a cartel and collusion has a negative impact on efficiency.