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

Combinatorial auctions – in which bidders can bid on combinations of goods – have been shown to increase the economic efficiency of a trade when goods have complementarities. However, the computational complexity of determining winners coupled with the cognitive complexity of formulating combinatorial bids has prevented this mechanism from reaching the online marketplace. Drawing from recent research in continuous combinatorial auctions, this study uses a data-driven approach to explore bidder behavior in such auctions using three experimental treatments that differ in the type of information feedback provided to participants. This study is the first to examine bidder strategies in complex combinatorial bidding environments. To do so, we collect not only the bids placed by bidders but also the clicks generated by the bidders as they explore different bundling options. Using cluster analysis, we find three stable bidder strategies across the three treatments. The different bidding strategies along with the nature of feedback have significant economic implications in terms of the surplus retained by the bidders themselves as well as the total revenue generated for the sellers. We believe that the enumeration of the strategies along with the analysis of their financial implications will help practitioners design better combinatorial auction environments.