39 Pages Posted: 23 Nov 2016 Last revised: 21 Apr 2017
Date Written: April 17, 2017
Despite their better revenue and welfare guarantees for repeated auctions, dynamic mechanisms have not been widely adopted in practice. This is partly due to their computational and implementation complexity, and also due to their unrealistic use of forecasting for future periods. We address the above shortcomings and present a new family of dynamic mechanisms that are computationally efficient and do not use any distribution knowledge of future periods. Our contributions are three-fold:
1. We introduce the concept of non-clairvoyance in dynamic mechanism design. A dynamic mechanism is non-clairvoyant if the allocation and pricing rule at each period does not depend on the type distributions in future periods. Our mechanism is non-clairvoyant and guarantees a 5-approximation compared to the optimal mechanism that knows all the distributions in advance.
2. We develop a framework for characterizing, designing, and proving lower bounds for dynamic mechanisms (clairvoyant or non-clairvoyant). In addition to the aforementioned positive results, we use this characterization to show that no non-clairvoyant mechanism can produce a better-than-2-approximation to the mechanism that knows all the distributions.
3. We present the first polynomial-time dynamic incentive-compatible and ex-post individually rational mechanism for multiple periods and for any number of buyers that is a constant approximation to the optimal revenue. Unlike previous mechanisms, we require no expensive pre-processing step and in each period we run a simple auction that is a combination of virtual value maximizing auctions.
Keywords: Dynamic Mechanism Design, Bank Account Mechanisms, Non-Clairvoyance, Dynamic Auctions, Approximations, Internet Advertising
JEL Classification: D44, C73, D82
Suggested Citation: Suggested Citation
Mirrokni, Vahab and Paes Leme, Renato and Tang, Pingzhong and Zuo, Song, Non-Clairvoyant Dynamic Mechanism Design (April 17, 2017). Available at SSRN: https://ssrn.com/abstract=2873701 or http://dx.doi.org/10.2139/ssrn.2873701