On the (non-)existence of efficient partnership dissolution mechanisms and the performance of combinatorial auctions

The difference between the willingness-to-pay (WTP) and the willingness-to-accept (WTA) has been widely studied and established in the literature. A WTP-WTA disparity is particularly relevant when the agents do not know whether they will be the buyer or the seller of an object, because then they are forced to assess both their WTP and their WTA. This phenomenon occurs in mechanisms to dissolve a partnership, i.e., an indivisible object that is jointly owned by more than one agent. Agents do not know ex-ante whether they will be buying the shares of other agents or selling their own share. Dissolution mechanisms decide which agent is to be awarded the partnership and how the others are to be compensated. The first section of this dissertation identifies a necessary and sufficient condition for the existence of an efficient dissolution mechanism in an independent private values model. In our setup - in contrast to the standard case, where the WTP and the WTA coincide - the existence of such a mechanism cannot be guaranteed. The second section analyzes a dissolution mechanism that is used in practice and is known to be efficient if there is no difference between the WTP and the WTA: the k+1-price auction. We show that even for an arbitrarily small difference between the WTP and the WTA, the k+1-price auction is inefficient. In addition, if the difference exceeds a certain threshold, agents are not willing to participate in the auction. Because of that, we analyze variations of the k+1-price auction which guarantee voluntary participation. The last section provides an experimental analysis of the performance of (combinatorial) core-selecting and Vickrey auctions. It contributes to the literature by explicitly testing equilibrium predictions under the realistic assumption of incomplete information. In the experiment, the proportion of efficient assignments is high for all auctions. The core-selecting auction performs better than the combinatorial Vickrey auction in generating revenue and in selecting stable (core) allocations. Interestingly, in an independent private values setting, the core-selecting auction is predicted to select fewer (eponymous) core allocations than the Vickrey auction and the data point in the opposite direction. The core-selecting auction provides incentives for overbidding the own valuation and - under certain conditions - also for bid-shading, which can hamper performance. The participants react to these incentives, though less pronouncedly than predicted