Many business and policy problems, such as allocation of spectrum rights and financial assets, involve allocation of heterogeneous objects among players with superadditive values. This paper uses laboratory experiments to study core-selecting auctions (clock-proxy auctions of Ausubel, Cramton, and Milgrom (2004)) recently proposed as a solution to this problem. Our experimental design involves three factors. The first factor is the auction design and we consider generalized Vickrey auctions, simultaneous ascending auctions, and clock-proxy auctions. The second factor is the value structure of agents. In addition to a benchmark case of additive values, we considered superadditive value structures which feature the exposure problem and the coordination problem. The third factor is subject characteristics. We ran experiments with professional traders and university students. We found that clock-proxy auctions outperformed generalized Vickrey auctions. Clock-proxy auctions outperformed simultaneous ascending auctions with the exposure problem value structure, and did statistically equally well with the additive and the coordination problem value structure. The result suggests a trade-off between efficiency improvements and complexity in package bidding. An ANOVA of outcomes demonstrated that auction designs were significant, and the interaction terms were often significant. We estimated the effect of auction design on efficiency and revenue and found that its magnitude depended on the valuation structure and subject characteristics. The result suggests that market design is not one-size-fits-all and a successful design builds on an understanding of fine details of the problem environments.
