Allocative Efficiency in Online Auctions: Improving the Performance of Multiple Online Auctions via Seek-and-Protect Agents

Much of the prominent literature describing behavior in eBay-like marketplaces emphasizes the successful use of "sniping" agents that wait until the last moments of an auction to bid (truthfully) on behalf of a human user. These agents fare well against "naive" agents (typically assumed to be those who bid incrementally on the most profitable open auction) who do not get the chance to respond to the snipe-bid placed in the final seconds. This reasoning, however, tends to ignore the effect of the poor coordination that occurs as more and more players attempt the sniping agent strategy, thereby raising prices above their minimum possible competitive equilibrium levels. Using proprietary data purchased from eBay, encompassing all bids submitted on four specific product types over a 3-month period, we analyze the allocative efficiency, price, and bidder surplus using a software agent and compare this to the historical performance. After showing a significant amount of "money left on the table" in the historical record, we proceed to demonstrate how bidders can significantly improve their surplus (i.e., observed profit) by adopting a "seek-and-protect" agent. If bidders go further and implement sequential-auction shading strategies, they can incrementally improve their surplus, but sometimes at the expense of allocative efficiency. Acknowledging that each bidder's time window of interest is inherently unobservable, we vary the length of bidders' consumption windows and find similar results.