Experimental evidence from dictator games and simple choice situations indicates concerns for fairness and social welfare in human decision making. At the same time, models of inequality averse agents fail to explain the experimental data of individuals who reduce their payoff below a fair split in order to maximize social welfare. This paper presents a linear model of altruism and inequality aversion which reconciles inequality aversion with departures from distributional fairness if welfare is thereby gained. It also establishes a unique link between altruism and interest in social welfare in the proposed model.
For more than a decade, models of altruism, inequality aversion and reciprocity have provided distinct explanations of non-selfish behavior (e.g., Fehr and Schmidt, 2006) and competed for legitimate application (e.g., Cooper and Kagel, forthcoming and Camerer, 2003). Altruism, inequality aversion or reciprocity rationalize the sharing of a surplus as, for instance, observed in ultimatum game (UG) experiments while only inequality aversion or reciprocity are consistent with money rejection. Positive money transfers, observed in unilateral dictator games (DGs), are exclusively explained by altruism and inequality aversion.1 Altruists enjoy their own as well as others’ well-being, and substantially inequality averse agents object to unequal payoff allocations. As altruism and inequality aversion mitigate or reinforce each other, the unconditional unselfishness and distributional preferences appear confounded to the outside observer. To deconstruct the interplay of both other-regarding motives, I extend a linear model of inequality aversion (Fehr and Schmidt, 1999; FS) by altruism.2 The trade-offs between altruism and inequality aversion explain how a more parsimonious model of inequality aversion, in some contexts, can abstract from the other behavioral motive without failing to replicate observed behavior. Explicitly accounting for altruism organizes extant experimental results (e.g., Charness and Rabin, 2002; CR) that, so far, have contradicted the simpler inequality aversion theory. In a complimentary approach, Tan and Bolle (2006) combine altruism with quadratic inequality aversion (Bolton and Ockenfels, 2000) and estimate the relative strengths of altruism and fairness using the experimental data from distinct DGs. For four sets of games, they report significant estimates for altruism as well as quadratic inequality aversion. Unlike their estimation of the preference parameters for a representative agent, I investigate the interaction of altruistic, fair and self-interested individuals in the interactive UG and assess the model's predictive potential against reciprocity and welfare preferences.
The first part of the paper calibrates the model in the UG, in which social welfare is fixed or destroyed, and demonstrates how inequality aversion interferes with altruism. The second part revisits the study of CR designed to compare the predictive power of inequality aversion to reciprocity and welfare preferences. In the CR experiment, subjects were confronted with decisions in which social welfare can be increased solely in combination with an increase in inequality, or in which inequality reduction costs welfare. Therefore, pure inequality aversion models, by construction disregarding the conflict between distinct social objectives, fell short in explaining their data. In contrast, altruism and inequality aversion preferences, which unlike the social-welfare preferences of CR exclude reciprocity, achieve the highest consistency value, an upper predictive success bound, of 98 percent. CR's decision tasks were also constructed to trigger reciprocal action which prevents actual achievement of the almost perfect fit of the suggested altruism and inequality aversion model. Despite this natural difficulty of distributive preferences with this data, due to the reciprocity found, this paper shows that a substantial degree of the inappropriateness of pure inequality aversion on the CR data is also attributable to altruism rather than reciprocity.
Section 2 presents experimental evidence of welfare concern. It reviews a formal depiction of altruism and then links it to inequality aversion. In Section 3, the new preferences are calibrated in the UG. Section 4 confronts the model with CR's experimental data from simple decision games in which final welfare, in contrast to the UG, varies in a broad surplus range with the players’ choices. Section 5 concludes.
