Version: November 18, 2018 Dynamic Escalation in Multiplayer Rivalries Fredrik Ødegaard Ivey Business School, The University of Western Ontario, London, ON, Canada, N6A 3K7,
[email protected] Charles Z. Zheng Department of Economics, The University of Western Ontario, London, ON, Canada, N6A 5C2,
[email protected] Many war-of-attrition-like races, such as online crowdsourcing challenges, online penny auctions, lobbying, R&D races, and political contests among and within parties involve more than two contenders. An important strategic decision in these settings is the timing of escalation. Anticipating the possible event in which another rival is to compete with the current frontrunner, a trailing contender may delay its own escalation effort, and thus avoid the instantaneous sunk cost, without necessarily conceding defeat. Such a free-rider effect, in the n-player dynamic war of attrition considered here, is shown to outweigh the opposite, competition effect intensified by having more rivals. Generalizing the dollar auction framework we construct subgame perfect equilibria where more than two players participate in escalation and, at critical junctures of the process, free-ride one another's escalation efforts. These equilibria generate larger total surplus for all rivals than the equilibrium where only two players participate in escalation. Key words : dynamic escalation, multiplayer rivalry, subgame perfect equilibrium, markov games, auction theory 1. Introduction In recent years crowdsourcing R&D through open online challenges has become a popular busi- ness innovation strategy. The first and most celebrated of such initiatives was the 2006 Netflix Prize, where the online movie streaming company offered one million dollars to the best performing movie recommendation algorithm.1 Like an ascending bidding process, the online challenge openly updated the submissions and their performances (\bids") so that contenders could dynamically up their efforts to outperform each others and most importantly the frontrunner.