Distributive fairness and the social responsibility of the representative of a group

ABSTRACT

Using an ultimatum game experiment where a representative makes a decision on behalf of the group members and equally shares the outcomes with them, we investigate whether the representative’s social responsibility exists and has a systematic relationship with his or her individual distributive fairness. The experimental results show that the representative of a respondent group tends to change his or her individual willingness to accept due to social responsibility. More importantly, we find that the minimum fairness of other members in the group tends to be the representative’s group standard for aggregating fairness, and that representatives whose individual fairness is higher [lower] than the minimum fairness of other members has a strong [negligible] tendency to incorporate their group members’ fairness preferences. According to our conceptual framework, this tendency can be only consistent with a positive correlation between the representative’s social responsibility and his or her individual distributive fairness. The results suggest that an incentive mechanism for a representative making a public decision would need to consider such a positive correlation between his or her social responsibility and distributive fairness.

KEYWORDS:

• Social responsibility
• representative
• distributive fairness
• social role
• ultimatum game

JEL CLASSIFICATION:

• D71
• C91
• C92

Acknowledgments

The authors are very grateful to the editor and anonymous referee for their helpful comments and suggestions which greatly improved this study.

Disclosure statement

We declare that we have no conflict of interest.

Supplementary material

Supplemental data for this article can be accessed here.

Notes

¹ Delegating to a third party, which we do not deal with in this study, means that the third party is free from the outcomes of his or her decisions on behalf of the delegating party. For this delegation problem, refer to Füllbrunn and Luhan (2017b) and its references.

² We, here, define social preferences as preferences based on social interaction. Thus, fairness, altruism, cooperation, trust and punishment are all included in the social preferences.

³ We adopt UG and focus on the respondent’s behaviour for the purpose of this study rather than adopting a dictator game, because the dictator game environment is known to be a ‘weak situation’ wherein the offer made by the dictator is very sensitive to the decision-making environment. It is well known that the respondents’ behaviour in UG is quite robust to stake size, degree of anonymity, context and culture (Henrich and Smith 2004; Camerer and Fehr 2014). Thus, focusing on the respondent’s behaviour would make us test our hypotheses in a decision-making environment wherein the interpretation is more robust. The respondent’s WTA is also called the minimum acceptable offer (MAO) in the literature.

⁴ Füllbrunn and Luhan (2017b) elicit a non-incentivized belief before the main task and find that the representative’s decision tends to move towards the belief. However, this procedure may be susceptible to an anchoring bias.

⁵ Some neuroscientific studies support this idea (e.g. Sanfey et al. 2003). Various social preference models incorporating outcome comparisons, norm following, or intentions could be regarded as the specification of this general model (e.g. Dufwenberg and Kirchsteiger 2004; Falk and Fischbacher 2006; Dillenberger and Sadowski 2012).

⁶ By assuming x≤ E/2, we only consider WTA≤E/2. This assumption makes us simplify and clarify the analyses on the relationship between social responsibility and distributive fairness by always making a respondent’s higher distributive fairness more distant from the subgame perfect NE.

⁷ We give a brief discussion on alternative interpretations of ${\delta }_i^{REP}$ in Section 5.

⁸ In the case of ${\alpha }_i={\alpha }_{-i}^{GS}$, ${\delta }_i^{REP}$ will not be identifiable.

⁹ ${\delta }_i^{REP}<0$ would imply that the representative has an antisocial preference rather than social responsibility. The representative can have the positive utility of $-{\delta }_i^{REP}\{x-{\alpha }_{-i}^{GS}(x_j-x)\}$ by making a decision which is opposite to that of other group members.

¹⁰ Equation (2)(2) $WT{A}_i^{REP}=\frac{({\alpha }_i+{\delta }_i^{REP}{\alpha }_{-i}^{GS})E}{(1+2{\alpha }_{-i}^{GS}){\delta }_i^{REP}+1+2{\alpha }_i}$(2) would allow to directly estimate both the social responsibility parameter and the fairness parameter if data is collected from an experiment designed for the purpose. For example, collecting a $WT{A}_i$ and a few $WT{A}_i^{REP}$of each subject where the degree of other members’ distributional fairness are varied, estimating ${\delta }_i^{REP}$ and ${\alpha }_i$ for each subject using the equation, and investigating crucial factors affecting the heterogeneity of ${\delta }_i^{REP}$ and ${\alpha }_i$ could be a worthwhile future study.

¹¹ See the online appendix A for the instructions and the screenshots.

¹² The decision procedure for REP was done in terms of the group payoffs: the group endowment was set to 3E. The representative of the proposer [respondent] group decided an offer [minimum acceptable point] between 0 and the group endowment. However, it was clearly informed that the three members in the group equally share the group payoffs and hence that each member’s payoff is exactly one third of the group payoffs. That is, subjects were informed that the representative’s offer [minimum acceptable point] out of 3E for the group is exactly matched his or her offer [minimum acceptable point] out of E for each member of the group.

¹³ In four out of eight sessions for REP&NOINFO, the order of Tasks 1 and 2 was reversed to control a possible order effect. We pool the data for the analyses because there was no significant order effect. For INFO treatments, reversing the order was impossible because information on other members’ individual fairness had to be collected from Task 1 before Task 2 began. For IND&NOINFO, Tasks 1 and 2 were exactly identical.

¹⁴ Here, we exclude observations whose WTAs are above 50 in order to be consistent with the conceptual framework (see footnote 6). Although we include these observations, the results are qualitatively similar.

¹⁵ A group-level test gives a qualitatively similar result. This is summarized in the online appendix C.

¹⁶ A similar analysis for IND&INFO shows that there is no significant correlation between WTA₂ and a possible group standard (see the online appendix D). This confirms that there is no general conforming effect. Moreover, a similar analysis for REP&NOINFO also shows that there is no significant correlation between WTA₂ and a possible group standard. Given that the information on others’ distributive fairness is not provided in REP&NOINFO, this result would confirm the noisiness of the representative’s belief in a possible group standard.

Additional information

Funding

We declare that we have no funding from an organization.