What is fair? Experimental evidence.

• Author: Dickinson, David L.

1. Introduction

   While economists have traditionally focused on efficiency, there has been a consistent, if not growing, interest within the discipline in the role of equity in the distribution of resources (e.g., Yaari and Bar-Hillel 1984; Rabin 1993; Gaertner 1994; Konow 1996, 2000; Fehr and Schmidt 1999; Bolton and Ockenfels 2000). Concerns about fairness are present in many different economic environments, and the desire to achieve a "fair" outcome has been offered as an explanation for many outcomes that do not support the theoretical predictions of purely self-interested utility maximization. In fact, a large body of experimental economic evidence has generated a significant literature regarding the nature of fairness in nonhypothetical environments since subjects are consistently willing to sacrifice personal cash payoffs for other (fair?) allocations (e.g., Guth, Schmittberger, and Schwarze 1982; Kahneman, Knetsch, and Thaler 1986; Forsythe et al. 1994; Andreoni 1995; Roth 1995; Clark 1998). Though most of the experim ental literature on fairness examines self-interested decision makers, this paper studies disinterested third-party decision makers as a way of removing the confounding effects of self-interest from much of the experimental data that explore fairness perceptions. (1)

   While we are obviously not the first to concern ourselves with the issue of fairness perceptions, the research to date typically uses either survey methodology or controlled experiments in which individuals bargain over their own payoff allocations. We present results from experiments that focus exclusively on individual preferences over payoffs to other individuals, unlike the typical experimental bargaining game in which two individuals bargain over how to divide a monetary pie among themselves. (2) Additionally, our experiments actually consummate the subjects' resource allocation decision by having third-party decision maker allocations generate real cash payoffs for real subject beneficiaries, which is a key difference from this paper and the survey literature. (3) Specifically, the final cash payoff of two individuals (beneficiaries) in our experiments is determined by decision makers whose cash payoff is independent of the allocation of scarce tokens. These tokens are inputs that determine the benefici aries' cash payoffs via differing (and nonlinear) payoff functions. Therefore, a simple 50/50 split of the tokens is not necessarily evident since it leads to unequal cash payoffs for the recipients (see section 3 for more details). Our design ensures that the decision maker's notion of fairness is what motivates the division of tokens, and it also allows one to distinguish between multiple concepts of fairness in the data.

   We also explore how fairness perceptions may be affected by effort--whether one recipient "merits" an advantageous position. Sometimes referred to as natural law or desert theories, the idea that higher effort merits a larger entitlement of resources is supported by existing research (e.g., Schokkaert and Lagrou 1983; Hoffman and Spitzer 1985; Mellers and Hartka 1989; Schokkaert and Overlaet 1989). We also collect demographic information on each subject so that we can investigate the demographic determinants of fairness beliefs, such as gender effects. Gender effects on perceptions of fairness and the role of effort have been previously explored in bargaining environments (e.g., Bolton and Katok 1995; Eckel and Grossman 1997; Andreoni and Vesterlund 2001) and in the household distribution of resources (Lundberg and Pollak 1996). Our findings are largely consistent with results from survey and experimental research showing that gender significantly affects fairness perceptions.

   These results have important policy implications. The fact that earning the right to a more favorable position may merit an individual a larger allocation of resources may explain many courtroom dialogues. Alternatively, if one can be cast as a victim who is undeserving of a less favorable position, an outsider's (i.e., judge's) perception of fairness is more likely to promote equal "payoff" outcomes--a topic of debate in many divorce settlements. Another implication stemming from our (and others') results is that men are significantly more likely to choose the most efficient (i.e., highest joint payoff) outcome than are women (see Andreoni and Vesterlund 2001). Given this, men and women are likely to view the justice of court decisions, transfer policies, and even the household distribution of resources differently.

2. Fairness in the Literature

   What is fair? The theoretical literature on fairness has shown that there are many potential fair allocations with each meeting a different set (but not all) of widely accepted axioms (such as envy freeness, anonymity, and Pareto optimality). The theoretical literature on fairness makes it clear that in any situation there are many allocations that could be considered fair. However, which of these perceptions of fairness are actually held by real people? The general conclusion from the existing empirical literature (mostly survey data) is that one dominant fairness concept does not exist and that fairness concepts differ with the context of the situation and with the individual. Konow (1996) argues that the dispersion in fairness concepts is not that chaotic. His argument is that people are influenced by three key items--accountability, altruism, and efficiency--when determining a fair allocation. The dispersion in what people report as fair stems from people weighing these criteria differently when making th eir decisions. Other theoretical models define fairness on the basis of reciprocity (Rabin 1993) or comparative payoff outcomes (Fehr and Schmidt 1999; Bolton and Ockenfels 2000). Concerns for reciprocity should not affect third-party decision making, and concerns for egalitarian outcomes would imply equal outcomes as a focal point for third-party allocations.

   Konow's (1996) "accountability principle" states that an unequal distribution is fair if the individual with more has earned his superior position. Konow's (1996, 2000) original survey results and more recent experimental results support the accountability principle, as do results from a variety of other studies (e.g., Schokkaert and Lagrou 1983; Hoffman and Spitzer 1985; Clark 1998). In his more recent work, Konow (2000) finds that benevolent dictators (decision makers paid a fixed fee for participation in the experiment) allocated a fixed payout between two subjects in proportion to "credits" when credits were obtained through effort but ignored the distribution of credits between the subjects when credits were arbitrarily assigned. Experimental evidence from Hoffman and Spitzer (1985), Guth (1988), Burrows and Loomes (1994), and Hoffman et al. (1994) also support the notion that differential payoffs are seen as fair if linked to effort. Hochschild (1981) finds that effort is a key criterion in whether gov ernment interhousehold distribution policies are seen as fair by the American public. Schokkaert and Lagrou (1983) and Schokkaert and Overlaet (1989) find additional survey data support for effort-based differences in payoffs. There is much less evidence on the importance of need and efficiency in forming fairness considerations. Need is offered as an important fairness criterion by the social psychology literature (Yaari and Bar-Hillel 1984; Gaertner 1994). The "needs rule" implies that those at the bottom be brought up at the expense of those at the top. Engle (1990) points out that need is likely to be a particularly important criterion when an individual is severely constrained. There is also some survey research evidence that efficiency concerns influence what people view as fair (e.g., Yaari and Bar-Hillel 1984; Konow 1996).

   Fairness concepts (or the weight given to Konow's three criteria) may also differ across individuals. Culture, background, and other personal characteristics may influence an individual's morals and values and, therefore, his notion of fairness. Evidence from the interdisciplinary literature on fairness finds that gender, education, and cultural background affect what an individual thinks is fair. Harbert and Scandizzo (1982) find that education increases the likelihood that a mother splits resources equally among her children. In reviews of the social psychology literature, Engle (1990) and Engle and Nieves (1993) find that women are more likely to equally split resources than to allocate on the basis of effort, and equal split is a more likely norm in noncapitalistic cultures.

   Experimental economics research has further substantiated gender effects. Croson and Buchan (1999) find that women reciprocate more than men in East Asian countries, and in the United States, Eckel and Grossman (in press) survey numerous studies that have shown somewhat mixed gender effects, but their results indicate that women are less selfish than men in double-anonymous (anonymous from counterpart and experimenter) dictator games. Andreoni and Vesterlund (2001) find that women are more likely to equalize payoffs and that men are more likely to maximize total payoffs in their modified dictator game. The weight of the evidence points toward women being more interested than men in equalizing payoffs.

   Farmer and Tiefenthaler (1995) theoretically examine a third-party allocation problem of parents' allocating food among their children, and their paper motivates many of the details of our experimental design. The authors point out the fairness concepts that likely exist in this simple problem of dividing a single, perfectly divisible good. Standard possibilities for fairness include equal split, proportional split, and equal loss of the resource. Equal and proportional split refer to how the resource is divided between two parties, either equally or in proportion to the individual's needs (for maximum...