This paper presents a model of ingroup favouritism in collective decisions. The focus is on the role of the institutional set-up for whether individuals' discriminatory preferences are mitigated or exacerbated. When are collective decisions less biased than individual decisions? Do diverse committees discriminate less than homogeneous ones? The analysis suggests that homogeneous committees can be expected to discriminate more than individual decision makers both under unanimity rule and under majority rule, but the reasons behind this are different under the two rules. Diversity in committees may help mitigate or avoid own group favoritism and can be expected to lead to less discrimination than decisions by homogeneous committees or by an individual decision maker.

This experiment focuses on two questions: First, are individuals willing to incur a cost to signal their true (or false) group identity to a person they are matched to interact with? Second, we test whether a true (or false) identity signal can help players to coordinate on a more socially efficient outcome in subsequent play of a coordination game involving a risk-return trade-off. We find that a significant proportion of subjects care about their group identity or believe that it matters enough to send a costly signal to others. Effort levels in the outgroup treatment increase significantly when participants are allowed to lie about their group identity compared to the case when they are not allowed to lie.

This paper presents an experiment investigating whether decision makers discriminate between members of their own group and members of another group. I focus on two aspects of this question: First, I compare behavior in individual and in joint decisions; Second, I test whether the identity of the co-decision maker matters in joint decisions. Substantial own group favoritism occurs in joint decisions in spite of there being no such favoritism in individual decisions. Decision makers strongly favor own group candidates when deciding with someone from their own group, but not when deciding with someone from the other group. The study suggests that higher-order beliefs about co-decision maker behavior may be a factor behind discrimination in collective settings and that diversity in committees might be helpful in counteracting own group favoritism.

This paper considers the use of categories to make predictions. It presents a framework to examine when decision makers may be better off using fewer rather than more categories, even without exogenous costs of using more. We study three cases: individual prediction, coordination of predictions, and the convex combination of the two. The analysis focuses on how the attempt to coordinate predictions with others affects incentives for coarse categorization in different environments. We show that while a coordination motive does not provide incentives for coarse categorization in deterministic environments, it could provide such incentives in stochastic environments.

Players may categorize the strategies available to them. In many games there are different ways to categorize one's strategies (different frames) and which ones players use has implications for the outcomes realized. This paper proposes a model of agents who learn which frames to use through reinforcement. As a case study we fit the model to existing experimental data from coordination games. The analysis shows that the model fits the data well as it matches the key stylized facts. It suggests a trade-off of using coarser versus finer representations of the strategy set when it comes to learning.

We study the impact of social identity and punishment in a Minimum Effort Game (MEG) experiment. In a first stage, subjects participated in some problem solving tasks, either as a group or individually. In a second stage, subjects played a one-shot two-player MEG, either with an Ingroup or an Outgroup counterpart, or as individuals without group identity. The MEG in this second stage was either with or without a punishment mechanism. In the treatments with punishment, subjects were given the opportunity to assign costly punishment points to their counterpart conditional on the effort levels chosen by their counterpart. We develop a measure of the strength of group identity based on consensus in the problem solving tasks. We find that the presence of group identity leads to higher effort levels. The threat of punishment leads to higher effort levels in the Outgroup treatment. There is no difference in punishment schedules towards In- and Outgroup members.