Voting on a sharing norm in a dictator game

https://doi.org/10.1016/j.joep.2009.10.006Get rights and content

Abstract

A large body of experimental research has documented that fairness and cooperation in experimental games are often enhanced when subjects can (and do) make non-binding pre-play agreements. Many studies have demonstrated this phenomenon for verbal agreements reached via open discussion. In contrast, little evidence exists concerning the effects of agreements reached using formal procedures that do not allow for discussion. This paper begins to fill this gap by investigating the effectiveness of an agreement reached by majority voting. Specifically, I investigate how voting on a sharing norm affects subsequent behavior in a dictator game. The voting procedure takes place behind a ‘veil of ignorance’, and the result of the vote is referred to as a non-binding agreement. I find that this procedure does not induce higher offers than are observed under a no-vote baseline. In fact, dictators were significantly more likely to offer nothing under the voting treatment. This evidence suggests that formal decision procedures may be less effective than open deliberation in creating effective non-binding agreements.

Introduction

A large and growing literature in experimental economics shows that fairness and cooperation in experimental games are enhanced when subjects can engage in pre-play cheap talk communication (Bornstein et al., 1989, Brosig et al., 2003, Charness and Dufwenberg, 2006, Dawes et al., 1977, Ellingsen and Johannesson, 2004, Kerr and Kaufman-Gilliland, 1991, Orbell et al., 1988, Ostrom et al., 1992). Indeed, as emphasized by Brosig et al. (2003), “one of the few variables that is known to have a robust and strong positive effect on the level of cooperation is the opportunity to communicate” (see also Sally, 1995, Walker and Ostrom, 2007).

As a typical example, Orbell et al. (1988) report on a public good experiment in which subjects were given the opportunity to engage in group discussion prior to interacting. Relative to a no communication benchmark, contributions were significantly enhanced, especially when promises had been exchanged. Ellingsen and Johannesson (2004) find that the exchange of promises enhances trust and trustworthiness in a two-player setting, suggesting that people have a preference for keeping their word. Charness and Dufwenberg (2006) collect similar evidence and suggest that the theory of guilt aversion (Dufwenberg and Gneezy, 2000, Battigalli and Dufwenberg, 2007) may account for players’ tendency to uphold agreements.1

Notions such as commitment or guilt aversion may potentially explain why people feel motivated to abide by the terms of a contract once is has been established. However, a point worth emphasizing is that neither of these concepts helps to identify the procedures necessary for creating something that will be perceived as constituting a contract. Applying these theories to the experiments mentioned above requires us to accept the additional premise that a certain type of verbal behavior (“making a promise”) causes expectations and/or obligations to be established. However, the theories themselves are silent as to what actually constitutes an instance of promise-making. Similarly, both guilt aversion and commitment may explain norm-abiding behavior. However, neither concept helps to identify the process by which a social norm is established. Therefore, even if we were to accept one or the other of these theories, we would still be left with the question: how are effective agreements actually created?

Existing studies have focused mostly on agreements reached through verbal (spoken or written) communication.2 This procedure is justified if the researcher’s goal is to prove the basic point that non-binding agreements can be effective. Open discussion seems intuitively to maximize the chance that subjects can establish an agreement with which they identify. Once the effectiveness of such agreements has been established, a question worth turning to is whether ‘weaker’ decision procedures can have similar effects. Such procedures are interesting because there are many real-life situations in which agreements in a group cannot feasibly be reached via discussion, and instead must rely on formal procedures.

Arguably the most common mechanism for collective agreement in real-world situations is majority voting. The current paper seeks to investigate the effectiveness of this procedure for creating a non-binding agreement or norm. Specifically, I investigate whether behavior in a dictator game is affected when, prior to interaction, subjects vote on a non-binding sharing norm. Previous research has shown that verbal communication in the form of written messages can have a significant positive effect on dictator offers (Mohlin & Johannesson (2008)). My goal here is to investigate whether a more formal procedure not involving verbal communication can have similar effects.3

The advantage of the dictator game setting is that it is a non-strategic situation and removes all coordination aspects of an agreement. The only reason for a subject to abide by an agreement in this context is that she feels obligated to do so. The design is based on the idea that participants may prefer to establish a norm of sharing when placed behind a “veil of ignorance”. Therefore, voting takes place before subjects are told whether they will be the dictator or the recipient in the subsequent interaction.

My main hypothesis is that the opportunity to establish a norm using this mechanism causes dictator subjects to behave more ‘fairly’ in the subsequent game. Specifically, I test the hypothesis that the distribution of dictator offers is shifted to the right by the voting treatment, as compared to a no-vote benchmark. An alternative approach would have been to test whether the distribution of dictator offers is shifted in the direction of whatever agreement is reached. The problem with this is that the agreement is endogenous, and thus correlated with unobserved characteristics of the respective session’s subject population. Therefore, it becomes impossible to causally interpret a correlation of behavior with the agreement itself. For this reason, I instead test whether offers are positively affected by the voting treatment as such.4

Surprisingly, the results of the experiment point in the opposite direction. Offers under the voting treatment were somewhat lower than under the baseline. A significantly higher proportion of dictators offered zero, as predicted by the standard model of narrowly self interested behavior. This evidence suggests that agreements arrived at through a formal mechanism such as the one employed may be less effective than those produced by open and informal deliberation.

The next section presents the experimental design. Section 3 describes the results, comparing behavior as well as measures of beliefs and feelings of obligation under the two treatments. Section 4 concludes and briefly outlines a possible explanation for the absence of a positive effect on dictator offers. The Appendix contains a translation of instructions.

Section snippets

Baseline treatment

Subjects entered the laboratory and were randomly assigned to isolated terminals. Instructions (reproduced in the appendix) were handed out and read aloud by the experimenter (myself). The instructions were the same for all subjects and informed them that they would be randomly matched with another subject, that one of them would then be randomly chosen to be ‘subject A’, and that this subject would then have the opportunity to choose an integer-valued distribution of 10 EUR between the two

Voting on the norm

Table 1 summarizes the voting stages in each of the voting sessions. In 4 of the 6 sessions (V1, V2, V5, and V6) subjects agreed upon a norm of sharing 50% of the pie. In sessions V3 and V4, the collective agreement was to share 40%. Table 1 also shows the number of subjects voting to ‘stay’ in the final round of voting. According to this measure, the norm of 4 was associated with a lower level of agreement in the groups. (Recall that this round is essentially about confirming the decision that

Conclusion

The experiment reported on in this paper sought to test whether majority voting can enhance fairness concerns and induce higher offers in a dictator game. Previous studies have documented that (written) verbal communication can enhance fairness in this setting. The hypothesis to be tested was that similar effects would be observed when agreement is reached using a formal procedure without verbal communication, namely majority voting. Surprisingly, the data suggests the opposite conclusion. In

Acknowledgements

I thank Luis Miller for the discussions which lead to the idea for this project. The comments of an anonymous referee helped to greatly improve the paper. Thanks to Frederic Bertels for assistance with programming and organizing experimental sessions. This research was funded by the Max Planck Society.

References (23)

  • T. Ellingsen et al.

    Promises, threats and fairness

    Economic Journal

    (2004)
  • Cited by (0)

    View full text