Nash's bargaining solution when the disagreement point is random (Q2485452)

The paper studies the Nash bargaining problem in cooperative game theory, allowing for a random disagreement point. In contrast to extant literature where the realized value of the random disagreement point becomes the actual outcome if the players are unable to agree on a division and, therefore, under the expected utility framework, the random disagreement point may be replaced by its expectation, in this paper the interpretation of a random disagreement point is different. If the players are unable to agree then they are notified of the realized disagreement point at which point they have a second chance to come to an agreement. Within this framework players may have an incentive to come to an agreement in the first stage and, \textit{a priori}, it may not be justifiable to replace the random disagreement point by its expectation. A two person bargaining problem with random disagreement point is formulated as a pair \(\left( \widetilde{a},S\right) \) where \(S\subset \mathbb R^2\) and \(\widetilde{a}\in \widehat{D}\), the set of all random variables with values in \(\mathbb R^2\). \ The interpretation is that for any \(a=\left( a_1,a_2\right) \) in the support of \(\widetilde{a}\), \(a_{i}\) is the utility level which is received by player \(i\) if the players do not cooperate and if \(a\) is the realization of \(\widetilde{a}\) . \(S\) represents the pairs of utility levels which are possible to achieve through agreement. Let \(U\) denote the set of all bargaining problems \((\widetilde{a },S)\) where \(S\) is non-empty, compact and convex and \(\widetilde{a}\) has a finite support contained in \(S.\) A bargaining solution is a function \( \varphi :U\rightarrow \mathbb R^2\) satisfying \(\varphi(\widetilde{a} ,S) \in S.\) In addition to standard axioms regarding the solution \( \varphi \), the author also adds a ``continuity'' axiom on \(\varphi \), an axiom which he calls ``betweenness'' which excludes a situation where one of the players would surely prefer to wait for the realization of the disagreement point before cooperating with the opponent and an axiom which states (see below) that the bargaining problem is, conceptually, resolved by the players in two stages. In the first stage the players replace the random disagreement point with a (virtual) fixed point and in the second stage they resolve the new bargaining problem which arises as a result. Axiom of ''\textit{Replaceability of random disagreement point}'': There exists a function \(\mathcal{R}\):\(\widehat{D}\rightarrow \mathbb R^2\) such that for all \(\left( \widetilde{a},S\right) \in U,\varphi \left( \widetilde{a},S\right) =\varphi \left( \mathcal{R}\left( \widetilde{a} \right) ,S\right) \) and \(\left( \mathcal{R}\left( \widetilde{a}\right) ,S\right) \in U.\) Let \(N(a,S) \) denote the Nash solution, i.e., \[ N(a,S) =\arg \max_{\{ x_{1}\geq a_{1},x_{2}\geq a_{2},( x_{1},x_{2}) \in S\} }( x_{1}-a_{1})( x_{2}-a_{2}) \] and let \(E(\cdot)\) denote the expectation operator. The following theorem is proved. Theorem: The unique solution \(\varphi \) satisfying all of the axioms is \(\varphi ( \widetilde{a},S) =N( E( \widetilde{a}) ,S) \) for all \(( \widetilde{ a},S) \in U.\) In general, however, the replacement of the random disagreement by its expectation is not warranted. The weaker Axiom of Replaceability is needed for the result. An example (Example 4 of the paper) is provided which demonstrates that without it several solutions are possible.