Design and Testing of an Auction for Non-Convex Cost Environments

Van Boening and Wilcox ran experiments finding that the ordinarily robust DoubleAuction produced inefficient results in an environment characterized by a small numberof sellers with non-convex cost structures (i.e. large avoidable fixed costs, zeroincremental costs, and production quantity limits). Advances in computation provideopportunities for new ways to transact multilaterally, which may facilitate efficientproduction in such environments. In the context of electricity markets, novel competitiveinstitutions have evolved that execute multilateral trading in environments with nonconvexities.My experiments use an institution called a Quasi-Uniform Price Auction(“QUPA”), which is modeled after mechanisms that are currently used in electricityauctions. When tested in environments with non-convex cost structures, the QUPA is insome cases more efficient than the Double Auction and the Smart Market, which isanother computationally intensive multilateral trading institution. These results suggestthat further experimental research on QUPAs would provide useful lessons for the futuredesign of electricity auctions.