Traditionally empirical auction analysis has assumed that bidders enter a single auction for a single item. While this approach has generated useful insights, it does not seem to be a close approximation of economically important auctions including eBay, online advertising auctions, FCC spectrum auctions, wholesale car auctions, fish auctions etc. In many cases there are multiple items being auctioned in multiple auctions that are operating sequentially or simultaneously. More recently, a number of researchers have begun analyzing the impact of auction competition on bidding behavior and auction prices.
Estimating Demand with Competing Auctions considers a situation where differentiated products are sold simultaneously in competing auctions.
The papers shows that auctions prices from competing and non-competing auctions can be used to identify the bidder's valuations across multiple products. Non-parametric identification uses the copula function to combine information from single auctions with partial identification of the distribution from competing auctions.
A simple estimator is used to implement model and estimate bidder valuations for Ford F-150 and Chevy Silverado auctions ending in a particular one-hour time period. The results are used to show how the optimal reserve prices must be lowered to account for competing auctions when bidders consider items to be close substitutes - as they do for the F-150 and Silverado.
