A Data-Driven Exploration of Bidder Strategies in Continuous Combinatorial Auctions

Combinatorial auctions, in which bidders can bid on combinations of assets (commonly referred to as bundles or packages), represent an innovative yet complex market mechanism that can enhance the ability to allocate multiple assets efficiently when the values of the assets exhibit synergies (e.g., “I will pay $100 if I get both matching end tables but only $40 for either one separately”). Although combinatorial auctions have been gainfully employed for resource allocation in industrial settings (e.g., by Sears, Home Depot, and Compaq), they are relatively absent in business-to-consumer (B2C) and consumer-toconsumer (C2C) electronic commerce, primarily due to the absence of a continuous mechanism that allows asynchronous bidding. The computational complexity of determining winners (the problem is NPhard) and the cognitive complexity of formulating efficient combinatorial bids (in an n-item auction, the number of biddable item bundles is 2-1) have acted as hurdles for the development of a continuous mechanism (Adomavicious and Gupta 2005; Kwasnica et al. 2005).