On the Informational Inefficiency of Discriminatory Price Auctions

We analyze bidding behavior in large discriminatory price auctions where the number of objects is a non-trivial proportion of the number of bidders. Bidders observe private signals that are affiliated with the common value. We show that the average price in the auction is biased downward from the expected value of the objects, even in the competitive limit. In particular, we show that conditional on relatively low signals, bidders bid the expected value of the objects conditional on their information and winning; while bids at higher signals flatten out and are below the expected value conditional on winning.