Abstract
Theory provides limited understanding of auctions in complex, multi-unit settings. This absence became apparent when performance concerns were raised over wholesale electric power auctions in England and California. This research used experiments involving 234 subjects to partially bridge the gap in knowledge with regard to bidding behaviour. Using a complex procurement setting motivated by power markets, bidding in the last accepted and first rejected uniform price auctions and the multi-unit Vickrey were examined across group sizes of 2, 4, and 6. To allow opportunity for tacit collusion, constant cost and demand conditions were maintained for 75 periods. Group size and experience were found to be most significant in explaining the differences between bids and costs. For groups of 2, this difference was significantly higher than with the other sizes, suggesting market power. Across sizes, after experience there was no significant difference in bidding behaviour between the uniform price auctions, with subjects in both bidding well above costs. Differences in the multi-unit Vickrey decreased after experience but, unexpectedly, bids were often below costs. Under no combinations did bids equal costs as would be desired by market designers. Results should give concern to policy analysts, and research suggestions to theorists.
Notes
A third large set of problems involves the complexities of power markets themselves, including transmission constraints, minimum up and down times, and others. These issues are beyond the scope of this research.
Dynamic auctions such as the English were not included since they were believed to be infeasible given the aforementioned complexities of power markets. For a consideration of uniform price auctions and discriminatory auctions see Bower and Bunn (Citation2001).
Ausubel's auction uses a ‘clinching’ rule that is arguably easier for subjects.
Instructions for all three auctions are available from the authors on request.
This lack of full capacity revelation is also evident in the figures. A separate analysis on this found no significant differences across auction type or group size. It was thus assumed these reductions were due to experience showing subjects the difficulty in selling those units, rather than collusion.