Revealed altruism without independence of irrelevant alternatives

Abstract

Revealed altruism theory (Cox, Friedman, and Sadiraj, 2008) is a theory of reciprocity that is grounded in neoclassical preference theory. The unconditional preferences included in revealed altruism (RA) theory are neoclassical; they satisfy the generalized axiom of revealed preference (GARP) and are independent of irrelevant alternatives. The reciprocal preferences included in RA theory do not satisfy GARP and they can be dependent on choices not made by another. This paper introduces an inconsistency with independence of irrelevant alternatives (IIA) into the non-reciprocal preference component of RA theory. This dependence is inspired by the axiom of monotonicity (Kalai and Smorodinsky, 1975) introduced into bargaining theory as an alternative to Nash's IIA axiom. The extended RA theory can rationalize patterns of giving and taking in dictator games (List, 2007; Bardsley, 2008) previously thought to be anomalous. It also rationalizes effects of earned vs. unearned endowments on dictators' choices (Cherry, et al., 2002). And the theory continues to rationalize data from many games involving reciprocity.