Auction Design and Performance: An Agent-Based Simulation with Endogenous Participation

This paper presents results from computational experiments evaluating the impact on performance of different auction design features. The focus of the study is a conservation auction for water quality where auctions are used to allocate contracts for improved land management practices among landholders bidding to provide conservation services. An agent-based model of bidder agents that learn using a combination of direction and reinforcement learning algorithms is used to simulate performance. The auction design features studied include: mix of conservation activities in tendered projects (auction scope effects); auction budget levels relative to bidder population size (auction scale effects); auction pricing rules (uniform versus discriminatory pricing); and endogeneity of bidder participation. Both weak and strong bidder responses to tender failure are explored for the case of endogeneity in participation. The results highlight the importance of a careful consideration of scale and scope issues and that policymakers need to consider alternatives to currently used pay-as-bid or discriminatory pricing fromats. Averaging over scope variations, the uniform auction can deliver substantially higher budgetary efficiency compared to the discriminatory auction. This advantage is especially higher when bidder participation decisions are more sensitive to auction outcomes.