How Individual Preferences Get Aggregated in Groups - An Experimental Study

Abstract

This paper experimentally investigates how individual preferences, through unrestricted deliberation, get aggregated into a group decision in two contexts: reciprocating gifts, and choosing between lotteries. In both contexts we find that median group members have a significant impact on the group decision, but particular other members also have some influence. Non-median members closer to the median tend to have more influence than other members. By investigating the same individual's influence in different groups, we find evidence for relative position in the group having a direct effect on influence. We do not find evidence that group choice exhibits a shift in a particular direction that is independent of member preferences and caused by the group decision context itself. We also find that group deliberation not only involves bargaining and compromise, but it also involves persuasion: preferences tend to shift towards the choice of the individual's previous group, especially for those with extreme individual preferences.

Are Teams Less Inequality Averse than Individuals?

“…Finally, the move toward more selfish team decisions can be partly explained by the fact that median team members make on average 27 less inequality averse initial proposals than the mean, at least for the α parameter (mean difference=0.110, p=0.064) if not for β (mean difference=-0.007, p=0.401). These results are consistent with those of Ambrus et al (2013).…”

“…Fourth, the higher heterogeneity of team members in terms of guilt and envy increases the number of iterations until convergence. Consistent with Ambrus et al (2013), we find that in the aggregation process, individuals with median preferences drive the team decision, while selfish and more inequality averse individuals converge more slowly to the proposals made by the median members. Finally, if the guilt parameter predicts the advantaged team's behavior in the Production Game when choices are not anonymous, the envy parameter shows no predictive power in any configuration of this game.…”

“…In their design, unanimity must be reached in five rounds maximum, while in our case, we apply a time constraint. 8 Kocher et al, 2006;Kugler et al, 2007;Sutter et al, 2009;Ambrus et al, 2013). Bosman et al (2006) find that the combination of the decision rule and the distribution of players' types determines the differences.…”

“…Franzen and Pointner (2013) find no difference in a dictator game with communication. 3 Ambrus et al (2013) argue in favor of the third explanation. In groups, median members have a stronger influence because extremes on both sides neutralize each other.…”

“…Using a within-subject design allows us to compare individual's decisions made in isolation and their initial proposals within a team. Thus, we contribute to the literature on how individual preferences are aggregated in groups (Zhang and Casari, 2012;Ambrus et al, 2013). Another contribution is studying how social image affects team preferences.…”