Author:

Abstract:

We present a revealed preference analysis of the testable implications of the Nash bargaining solution. Our specific focus is on a two-player game involving consumption decisions. We consider a setting in which the empirical analyst has information on both the threat points bundles and the bargaining outcomes. We first establish a revealed preference characterization of the Nash bargaining solution. This characterization implies conditions that are both necessary and sufficient for consistency of observed consumption behavior with the Nash bargaining model. However, these conditions turn out to be nonlinear in unknowns and therefore difficult to verify. Given this, we subsequently present necessary conditions and sufficient conditions that are linear (and thus easily testable). We illustrate the practical usefulness of these conditions by means of an application to experimental data. Such an experimental setting implies a most powerful analysis of the empirical goodness of the Nash bargaining model for describing consumption decisions. To our knowledge, this provides a first empirical test of the Nash bargaining model on consumption data. Finally, we consider the possibility that threat point bundles are not observed. This obtains testable conditions for the Nash bargaining model that can be used in non-experimental (e.g. household consumption) settings, which often do not contain information on individual consumption bundles in threat points.