Efficiency, Durability and Ordinality in 2-Person Bargaining with Incomplete Information1

We consider two-person bargaining problems in which (only) the players’ disagreement payoffs are private information and it is common knowledge that disagreement is inefficient. We show that if the Pareto frontier is linear, or the utility functions are quasi-linear, the outcome of an ex-post efficient mechanism must be independent of the players’ disagreement values. Hence, in this case, a bargaining solution must be ordinal: the players’ interim expected utilities cannot depend on the intensity of their preferences. For a non-linear frontier, the result continues to hold if disagreement payoffs are independent or if one of the players only has a few types. We discuss implications of these results for axiomatic bargaining theory and for full surplus extraction in mechanism design