Sequential claim games (Q6049401)

The authors discuss a non-cooperative approach to the division of an estate. This kind of problem can arise in the context of bankruptcy or -- after suitable transformation -- in the context of advertising, rental of space (for exhibition), and voting, to say the least. The model for estate division is an extensive form game of perfect information. The players are ordered (by some extrinsic means, including random mechanism). On \(j\)th stage of the game, Player \(j\) observes the decisions of players from earlier stages of the game and claims a finite number of disjoint subintervals of \([0,1]\) (representing the estate) with total length corresponding to the \(j\)'s entitlement. The resulting subdivision of \([0,1]\) is further subdivided proportionally according to the entitlements on parts claimed by more than one player. The authors restrict their investigation to a myopic way of playing: claims of Player \(j\) are put on the least claimed parts of the estate. It turns out that the myopic play leads to a Nash equilibrium in a simultaneous version of the claim game. However, it may not lead to a Nash equilibrium in the sequential play; it is optimal only for the last claimer. The main question is the following: how to turn a myopic strategy profile into a subgame perfect equilibrium. The main result of the article roughly states that if myopic players select the least claimed parts of \([0,1]\), scanning it from the left to the right, and they `punish' a claimer who makes `holes' when choosing subintervals, then this leads to a subgame perfect equilibrium. Since the estate division problem in the case of equal entitlements resembles the fair division of cake, it would be interesting to see any approach to estate division in the spirit of `I cut, you choose' procedures for multiple cake-eaters.