An Empirical Analysis of Informationally Restricted Dynamic Auctions of Used Cars

We analyze a dynamic informationally restricted auction mechanism a rental car company created to sell its cars. All bids are submitted via the internet in continuous time in ascending price auctions that last for two minutes. Unlike English or Japanese auctions, bidders do not observe the number of other bidders in the auction or their bids. At any time in the auction each bidder only knows a) the history of their own bids, and b) whether or not their bid is the current high bid. The bidder with the highest bid at the end of the auction is the winner and pays their winning bid. We develop theoretical and empirical models of bidding strategies in this auction and estimate bidders’ beliefs and distributions of valuations for used cars. We show that there is a substantial amount of early bidding in these auctions, even though a game-theoretic analysis suggests that the informational restrictions should create strong incentives for bid sniping — i.e. waiting to submit a bid only in the final second of the auction. If all bidders used non-informative bid sniping strategies the outcome of the auction would be equivalent to that of a static sealed bid auction. Bidders have an incentive to bid early in the auction to learn what the current high bid is (since they learn this only if their bid is the current high bid), but early bidding can operate to their collective disadvantage, leading them to bid more than they would in a static first price sealed bid auction. We use our model and estimated distribution of valuations for cars to compare auction revenues under the dynamic auction with counterfactual revenues from a static sealed bid auction with and without a reservation price under the same informational restrictions (i.e. that bidders do not know the number of competing bidders participating in any individual auction). Preliminary draft, not for quotation or public distribution. Direct comments to Sungjin Cho at Department of Economics, Seoul National University, Seoul, Korea, email: [email protected].