Social position and distributive justice: experimental evidence.

1. Introduction

In A Theory of Justice (1971) John Rawls states that men are born into various social positions that yield "different expectations of life" but "cannot possibly be justified by an appeal to the notions of merit or desert" (p. 7). Rawls further hypothesizes that if men could choose behind a "veil of ignorance" with regard to social position, they would unanimously agree to social institutions that allocate social benefits to maximize the income of the lowest income group, and they would reject institutions or principles (such as utilitarianism) that lead to gains to some members of society if such gains were to come at the expense of the least advantaged members.

It is not possible to directly test the Rawlsian hypothesis in the field. However, the hypothesis has been tested in laboratory economics and political science experiments (e.g., Frohlich, Oppenheimer, and Eavey 1987; Frohlich and Oppenheimer 1990, 1992; Bond and Park 1991; Lissowski, Tyszka, and Okrasa 1991; Jackson and Hill 1995; de la Cruz-Dona and Martina 2000; Oleson 2001; Herne and Suojanen 2004). The conclusion to be drawn from these studies is that subjects' choices do not conform strictly to the Rawlsian principle of distributive justice but rather to a mixture of Rawlsian and utilitarian principles.

While related to previous studies, in that we test if subjects' choices under a laboratory veil of ignorance are consistent with the Rawlsian hypothesis, our experiments extend the analysis by determining the extent to which subjects' expectations of distributive shares are influenced by unearned social position. Broadly speaking, social position refers to the resources, power, education, prestige, or occupation that significantly impacts one's current and future well-being. We use a simple two-player dictator experiment and measure social position by the initial distributive shares (resources) and the subjects' ability to determine the final distributive shares (power). Our hypothesis is that increasing a subject's power and resources will result in an income distribution choice to that subject's greater favor. Thus, while we contribute to the literature on Rawlsian justice, our primary focus is on the importance of differences in unearned social position with regard to expectations, as opposed to the absence of differences. In this way we gain insight into how the removal of such differences may impact choices in the field. Our experiment also contributes to the understanding of behavior in dictator experiments. In particular, ours complements related experiments on social distance and entitlement effects.

The remainder of the paper proceeds as follows. Section 2 briefly describes the experimental design, subjects, and setting and compares our experiment to other laboratory dictator and Rawlsian experiments. Section 3 provides a discussion of the results, followed by concluding remarks in section 4.

2. The Experiment

In a dictator experiment (Kahneman, Knetsch, and Thaler 1986), one subject determines how a fixed sum of money is to be divided between her and one or more other players. We conducted a simple, double-blind (Hoffman et al. 1994; Hoffman, McCabe, and Smith 1996), one-shot, two-person $20 dictator experiment with no show-up fee. (1) We implemented five protocols in which we systematically changed language and procedures to vary the decision maker's social position in the experiment. We obtained approximately 30 observations for each of the five protocols, using 211 subjects. We conducted sessions with approximately 15 subjects in each of two rooms for all treatments. Participants were students at the U.S. Naval Academy in various years of study and various majors. Students were invited to participate via E-mail solicitation. All decisions were recorded privately by an experimenter who had no contact with the subjects.

Protocol Giving Dictator (GD) is a standard dictator experiment. Player 1 is given $20 (an envelope with 20 one-dollar bills and 20 slips of paper) and the option of sending any portion of this endowment to an anonymous second person with whom she has been randomly matched. In Protocol Probabilistic Dictator (PD), both players allocate $20 between two envelopes. One player's choice is chosen randomly and implemented with that player retaining the Player 1 envelope. In Protocol Taking Dictator (TD), Player 2 receives the $20 envelope and Player 1 is instructed to indicate how much of the money she wants to take. Player 2 is instructed to place this amount in a second envelope to be transferred. In Protocol Veiled Dictator (VD), only one player (not both players, as in the PD protocol) divides the money between envelopes marked Player 1 and Player 2, knowing that there is a 50% chance she will be either Player 1 or Player 2. The subjects are then randomly assigned their role (Player 1 or Player 2). In addition to these four protocols, we asked the recipient (Player 2) in Protocol GD "What allocation would you choose if you were Player 1?" and "What allocation do you predict Player 1 will make?" Protocol Hypothetical Dictator (HD) refers to these hypothetical choices.

Of all the protocols, in GD, the decision maker is in the greatest position of resources and power. She is given the endowment, full authority to implement the distribution of her choice, and there is no uncertainty that her choice will be implemented. She realizes that she is determining not only her own payoff, but also the payoff of a powerless person with whom she is matched.

PD levels the playing field between matched participants. There is no difference in resources because neither player is initially allocated the endowment. Furthermore, both participants have an equal probability of being assigned to the position of power. Therefore, the participants may perceive themselves to have identical social position relative to each other.

While TD is equivalent to GD in that it grants, with certainty, full power to the decision maker, she is in a weaker position relative to GD with regard to resources. In this protocol, the decision maker is not allocated the endowment and is required to "take" money from the other player to increase her own payoff. Inaction, in this situation, will leave the entire payoff to the other player.

VD most closely reproduces Rawls' veil of ignorance in the laboratory. Should the income distribution choice be unequal, the decision maker does not know on which end of the income distribution she will ultimately fall. Therefore, she cannot use this information to promote her self-interest.

HD is the hypothetical response of the powerless player, Player 2, when asked what she would choose to do if in a position of power. Therefore, HD is the opposite of GD in that the decision maker is in the weakest position of resources and power.

A standard game-theoretic prediction based on strict self-interest is that decision makers in all protocols will seek to maximize their own monetary payoff. This implies that decision makers in protocols GD, PD, TD, and HD (assuming an honest response) will all choose a payoff of $20 to themselves, leaving $0 to the other player. In VD, any distribution choice yields the same expected payoff. As implied by the Rawlsian hypothesis, risk-averse subjects should choose equal splits, while risk-seeking subjects should choose unequal splits. Models of pure altruism (as in Levine 1998) or inequity aversion (Fehr and Schmidt 1999; Bolton and Ockenfels 2000) would predict no difference in the distribution choice in protocols GD, PD, TD, and HD, though the choice may involve a positive payoff to the other player.

The simple variations in our experimental instructions and procedures are intended to generate perceived differences in unearned social position in the laboratory. The primary objectives are (i) to test Rawls' general assumption that "expectations" of distributive shares are influenced by "social position" and (ii) to quantify the extent of such effects. Our simple hypothesis is that protocols PD, TD, VD, and HD will result in lower Player 1 payoffs (that is, more egalitarian payoffs) than the benchmark protocol GD, because decision makers are in a lower social position in terms of power and/or resources. (2) Strictly interpreting the Rawlsian hypothesis, all subjects should implement the equal share outcome in VD, the "veiled" protocol.

While the simple dictator experiment does not allow the Rawlsian hypothesis to be evaluated in a richer way, (3) other laboratory experiments have been used to explicitly test Rawls' predictions. In the majority of these studies, participants complete a questionnaire in which they make choices over explicit income distributions with or without veil-of-ignorance conditions. These studies have consistently rejected the strict Rawlsian hypothesis. For example, Frohlich, Oppenheimer, and Eavey (1987); Frohlich and Oppenheimer (1990); and Herne and Suojanen (2004) all find that a majority of veiled individuals, given the opportunity for group discussion, do not prefer a Rawlsian income distribution, but rather choose a distribution that maximizes the average income subject to a floor constraint. Unlike these experiments, our experiment does not focus on the social contract chosen by groups of individuals under veil-like conditions. Rather, we are interested in the extent to which deviations from the veil condition impact individual choices of an income distribution in a dictator experiment. Using a trust game, Vyrastekova and Onderstal (2005) find that subjects do not behave differently when they play both roles behind a "veil" versus when they play the standard trust game.